REMOTE SENSING –

APPLICATIONS
Edited by Boris Escalante-Ramírez
Remote Sensing – Applications
Edited by Boris Escalante-Ramírez

Published by InTech
Janeza Trdine 9, 51000 Rijeka, Croatia

Copyright © 2012 InTech

All chapters are Open Access distributed under the Creative Commons Attribution 3.0
license, which allows users to download, copy and build upon published articles even for
commercial purposes, as long as the author and publisher are properly credited, which
ensures maximum dissemination and a wider impact of our publications. After this work
has been published by InTech, authors have the right to republish it, in whole or part, in
any publication of which they are the author, and to make other personal use of the
work. Any republication, referencing or personal use of the work must explicitly identify
the original source.

As for readers, this license allows users to download, copy and build upon published
chapters even for commercial purposes, as long as the author and publisher are properly
credited, which ensures maximum dissemination and a wider impact of our publications.

Notice
Statements and opinions expressed in the chapters are these of the individual contributors
and not necessarily those of the editors or publisher. No responsibility is accepted for the
accuracy of information contained in the published chapters. The publisher assumes no
responsibility for any damage or injury to persons or property arising out of the use of any
materials, instructions, methods or ideas contained in the book.

Publishing Process Manager Dragana Manestar

Technical Editor Miroslav Tadic
Cover Designer InTech Design Team

First published June, 2012

Printed in Croatia

A free online edition of this book is available at www.intechopen.com

Additional hard copies can be obtained from orders@intechopen.com

Remote Sensing – Applications, Edited by Boris Escalante-Ramírez

p. cm.
ISBN 978-953-51-0651-7
Contents

Preface IX

Section 1 Land Cover 1

Chapter 1 Narrowband Vegetation Indices for

Estimating Boreal Forest Leaf Area Index 3
Ellen Eigemeier, Janne Heiskanen, Miina Rautiainen,
Matti Mõttus, Veli-Heikki Vesanto, Titta Majasalmi
and Pauline Stenberg

Chapter 2 Crop Disease and Pest Monitoring by Remote Sensing 31

Wenjiang Huang, Juhua Luo, Jingcheng Zhang, Jinling Zhao,
Chunjiang Zhao, Jihua Wang, Guijun Yang, Muyi Huang,
Linsheng Huang and Shizhou Du

Chapter 3 Seasonal Variability of Vegetation

and Its Relationship to Rainfall and
Fire in the Brazilian Tropical Savanna 77
Jorge Alberto Bustamante, Regina Alvalá
and Celso von Randow

Chapter 4 Land Cover Change Detection in

Southern Brazil Through Orbital
Imagery Classification Methods 99
José Maria Filippini Alba, Victor Faria Schroder
and Mauro Ricardo R. Nóbrega

Chapter 5 Mapping Soil Salinization of

Agricultural Coastal Areas in Southeast Spain 117
Ignacio Melendez-Pastor, Encarni I. Hernández,
Jose Navarro-Pedreño and Ignacio Gómez

Chapter 6 Remote Sensing Based

Modeling of Water and Heat
Regimes in a Vast Agricultural Region 141
A. Gelfan, E. Muzylev, A. Uspensky,
Z. Startseva and P. Romanov
VI Contents

Chapter 7 High Resolution Remote Sensing Images

Based Catastrophe Assessment Method 177
Qi Wen, Yida Fan, Siquan Yang, Shirong Chen,
Haixia He, Sanchao Liu, Wei Wu, Lei Wang, Juan Nie,
Wei Wang, Baojun Zhang, Feng Xu, Tong Tang,
Zhiqiang Lin, Ping Wang and Wei Zhang

Chapter 8 Automatic Mapping of the Lava Flows at

Piton de la Fournaise Volcano, by Combining
Thermal Data in Near and Visible Infrared 201
Z. Servadio, N. Villeneuve and P. Bachèlery

Section 2 Climate and Atmosphere 221

Chapter 9 Coupled Terrestrial Carbon

and Water Dynamics in Terrestrial
Ecosystems: Contributions of Remote Sensing 223
Baozhang Chen

Chapter 10 Oceanic Evaporation: Trends and Variability 261

Long S. Chiu, Si Gao and Chung-Lin Shie

Chapter 11 Multi-Wavelength and Multi-Direction Remote

Sensing of Atmospheric Aerosols and Clouds 279
Hiroaki Kuze

Section 3 Oceans and Cryosphere 295

Chapter 12 Remote Sensing of

Submerged Aquatic Vegetation 297
Hyun Jung Cho, Deepak Mishra and John Wood

Chapter 13 Remote Sensing and Environmental

Sensitivity for Oil Spill in the Amazon, Brazil 309
Milena Andrade and Claudio Szlafsztein

Chapter 14 Satellite Remote Sensing of Coral Reef

Habitats Mapping in Shallow Waters at Banco
Chinchorro Reefs, México: A Classification Approach 331
Ameris Ixchel Contreras-Silva, Alejandra A. López-Caloca,
F. Omar Tapia-Silva and Sergio Cerdeira-Estrada

Chapter 15 Predictability of Water Sources

Using Snow Maps Extracted from
the Modis Imagery in Central Alborz, Iran 355
Seyed Kazem Alavipanah, Somayeh Talebi and Farshad Amiraslani

Chapter 16 Remote Sensing of Cryosphere 369

Shrinidhi Ambinakudige and Kabindra Joshi
 Contents VII

Chapter 17 Remote Sensing Application in

the Maritime Search and Rescue 381
Jing Peng and Chaojian Shi

Section 4 Human Activity Assessment 405

Chapter 18 Object-Based Image Analysis of VHR

Satellite Imagery for Population Estimation
in Informal Settlement Kibera-Nairobi, Kenya 407
Tatjana Veljanovski, Urša Kanjir, Peter Pehani,
Krištof Oštir and Primož Kovačič

Chapter 19 Remote Sensing Applications

in Archaeological Research 435
Dimitrios D. Alexakis, Athos Agapiou,
Diofantos G. Hadjimitsis and Apostolos Sarris

Chapter 20 The Mapping of the Urban Growth of

Kinshasa (DRC) Through High Resolution
Remote Sensing Between 1995 and 2005 463
Kayembe wa Kayembe Matthieu,
Mathieu De Maeyer and Eléonore Wolff

Chapter 21 Remote Sensing for Medical

and Health Care Applications 479
Satoshi Suzuki and Takemi Matsui

Chapter 22 Demonstration of Hyperspectral Image

Exploitation for Military Applications 493
Jean-Pierre Ardouin, Josée Lévesque,
Vincent Roy, Yves Van Chestein and Anthony Faust
Preface

Nowadays it is hard to find areas of human activity and development that have not
profited from or contributed to remote sensing. Natural, physical and social activities find
in remote sensing a common ground for interaction and development. From the end-user
point of view, Earth science, geography, planning, resource management, public policy
design, environmental studies, and health, are some of the areas whose recent
development has been triggered and motivated by remote sensing. From the
technological point of view, remote sensing would not be possible without the
advancement of basic as well as applied research in areas like physics, space technology,
telecommunications, computer science and engineering. This dual conception of remote
sensing brought us to the idea of preparing two different books. The present one is meant
to display recent advances in remote sensing applications, while the accompanying book
is devoted to new techniques for data processing, sensors and platforms.

Strictly speaking, remote sensing consists of collecting data from an object or

phenomenon without making physical contact. In practice, most of the time we refer to
satellite or aircraft-mounted sensors that use some sort of electromagnetic radiation to
gather geospatial information from land, oceans and atmosphere with increasingly high
spatial, spectral and temporal resolutions. Space agencies in charge of collecting
remotely sensed data have shown a notorious interest in making these data available for
research and social development. The confluence of remote sensing technology with
other sciences has resulted in an exponential growth of knowledge, technology
development and assessment of all kind of physical and natural phenomena, as well as
human activities that share a common ground: geospatial information. However, the
success of remote sensing influencing other areas of knowledge and human activity has
not always been a paved way. The variables of great interest to scientists in different
areas are not readily available from the raw remotely-sensed data. Even when the data
has been processed and converted to physical-related values, or even linked to human
and natural artifacts like crop fields, roads, urban areas, geomorphologic structures,
vegetation indices, etc., the relationship between these and the more abstract variables
that explain them such as human settlement dynamics, geophysical phenomena, climate
change, etc. remain a major field of study and research.

This book intends to show the reader how remote sensing impacts other areas of
science, technology, and human activity, by displaying a selected number of high
X Preface

quality contributions dealing with different remote sensing applications. Twenty two
chapters have been carefully collected and distributed in four areas. The first part
deals with land cover applications, and contains applications in vegetation indices,
crop and pest monitoring, rainfall and fire relationship with vegetation, change
detection, soil salinization, modeling water and heat regimes, catastrophe assessment
and lava flow mapping. The second part contains contributions on climate and
atmosphere, including carbon and water dynamics, ocean evaporation, and
atmospheric aerosols and clouds. The third part presents oceans and cryosphere
applications that include aquatic vegetation, oil spill assessment, coral reef habitat
mapping, water source predictability from snow maps, cryosphere study, and
maritime search and rescue. Last but not least, the last part presents contributions
dealing with human activity, including population estimation, archaeology, urban
growth, medicine and healthcare and military applications.

I am indebted to all authors who have contributed to this book. Without their
strongest commitment this book would not have been possible. I am also thankful to
InTech editorial team who has provided the opportunity to publish this book.

Boris Escalante-Ramírez
National Autonomous University of México,
Faculty of Engineering, Mexico City,
Mexico
Section 1

Land Cover
 1

Narrowband Vegetation Indices for

Estimating Boreal Forest Leaf Area Index
Ellen Eigemeier, Janne Heiskanen, Miina Rautiainen, Matti Mõttus,
Veli-Heikki Vesanto, Titta Majasalmi and Pauline Stenberg
University of Helsinki
Finland

1. Introduction
1.1 Leaf area index
The green photosynthesizing leaf area of a canopy is an important characteristic of the status
of the vegetation in terms of its health and production potential. At stand level, the amount
of leaf area in a canopy is represented by a variable called the leaf area index (LAI), which is
one of the key biophysical parameters in the global monitoring and mapping of vegetation
by satellite remote sensing (Morisette et al., 2006). In this paper we adopt the, by now
widely accepted, definition of LAI as the hemi-surface or half of the total surface area of all
leaves or needles in the vegetation canopy divided by the horizontal ground area below the
canopy. The definition is in line with the original definition of LAI, formulated for flat and
(assumedly) infinitely thin leaves (Watson, 1947), as the one-sided leaf area per unit ground
area. For coniferous canopies, the question arose on how to define the “one-sided” area of
non-flat needles. While projected needle area formerly often has been used erroneously as a
synonym to one-sided flat leaf area, it is now commonly accepted that the hemi-surface
needle area represents the logical counterpart to the one-sided area of flat leaves (e.g. Chen
& Black, 1992; Stenberg, 2006).
LAI controls many biological and physical processes, driving the exchange of matter and
energy flow. Because LAI responds rapidly to different stress factors and changes in climatic
conditions, monitoring of LAI yields a dynamic indicator of forest status and health. The
link between forest productivity and LAI, in turn, lies in that LAI is the main determinant of
the fraction of incoming photosynthetically active radiation absorbed by the canopy
(fAPAR). The absorbed photosynthetically active radiation (APAR) quantifies the energy
available for net primary production (NPP) and is thus a critical variable in NPP and carbon
flux models. NPP is related to APAR by the light-use-efficiency originally introduced by
Monteith (1977) for agricultural crops.
Traditionally, ground-based measurements of LAI have typically involved destructive
sampling and determination of allometric relationships, e.g. between leaf area and the basal
area of stem and/or branches carrying the leaves (the pipe model theory) (Shinozaki et al.,
1964; Waring et al., 1982). However, such “direct methods” are quite laborious and indirect
measurements of LAI using optical instruments are today the preferred choice (Welles &
4 Remote Sensing – Applications

Cohen, 1996; Jonckheere et al., 2004). They provide inverse estimates of LAI based on the
fraction of gaps through the canopy in different directions, which can be measured using
devices such as the LAI-2000 Plant Canopy Analyzer (LI-COR, 1992) or hemispherical
photography. A vast body of classical literature exists on the dependency between LAI and
canopy gap fraction underlying these techniques (e.g. Wilson, 1965; Miller, 1967; Nilson,
1971; Lang, 1986). In short, the inversion methods rely upon the assumption that leaves are
randomly distributed in the canopy, in which case Beer’s law can be applied to plant
canopies (Monsi & Saeki, 1953). However, as the organization of leaves (needles) in forest
canopies is typically more aggregated (“clumped”) than predicted by a purely random
distribution, the technique causes underestimation of LAI, especially in coniferous stands
(e.g. Smith et al., 1993; Stenberg et al., 1994). Instead of the true LAI, the inversion of gap
fraction data without correction for clumping yields the quantity commonly referred to as
the “effective leaf area index” (Black et al., 1991).
Monitoring LAI in a spatially continuous mode and on a regular basis is possible only using
remote sensing. Estimation of LAI from optical satellite images is considered feasible
because LAI is closely linked to the spectral reflectance of plant canopies in the shortwave
solar radiation range (Myneni et al., 1997). The physical relationships between canopy
spectral reflectances and LAI form the basis of retrieval algorithms used in current Earth
observation programs (e.g. MODIS, CYCLOPES, GLOBCARBON products) for mapping
LAI at global scales. They produce bi-weekly and monthly vegetation maps that are widely
used by biologists, natural resources managers, and climate modelers, e.g. to track seasonal
fluctuations in vegetation or changes in land use. The arrival of narrowband reflectance data
(also known as hyperspectral or imaging spectroscopy data) opens up new possibilities for
satellite-derived estimation/monitoring of variables connected to the status and structure of
vegetation, including LAI.

1.2 Spectral properties of boreal forests

The boreal forest zone, which spreads through Fennoscandia, Russia, Canada and Alaska, is
the largest unbroken forest zone in the world and accounts for approximately one fourth of
the world’s forests. The boreal zone is a major store of carbon and thus plays an important
role in determining global albedo and climate.
The reflectance spectra of coniferous forests (even if they have the same leaf area) are very
distinct from similar broadleaved forests. The reasons for the special spectral behaviour of
coniferous forests are versatile, yet primarily related to their structural, not optical,
properties. Firstly, a high level of within-shoot scattering of conifers was originally noted
nearly four decades ago (Norman & Jarvis, 1975). More recently, Landsat ETM+ data and a
forest reflectance model were used to show that the low near infrared (NIR) reflectances
observed in coniferous areas can largely be explained simply by within-shoot scattering
(Rautiainen & Stenberg, 2005). Secondly, absorption by coniferous needles is higher than
that by broadleaved species (Roberts et al., 2004; Williams, 1991), a phenomenon which can
partly contribute to the lower reflectances of conifer-dominated areas. Other explanations
include, for example, that the tree crown surface of coniferous stands is more heterogeneous
than in broadleaved stands (Häme, 1991; Schull et al., 2011). In other words, when surface
roughness (i.e. crown-level clumping) increases, the shaded area within the canopy
increases, thus leading to lower reflectances. Overall, these results highlight the importance
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 5

of various geometric properties as the main reason for the reflectance differences between
broadleaved and coniferous stands.
Remote sensing of the biophysical properties, such as LAI, of a boreal coniferous forest
canopy layer is further complicated by the often dominating role of the understory in the
spectral signal (Rautiainen et al., 2011; Rautiainen et al., 2007; Eriksson et al., 2006; Eklundh
et al., 2001; Chen & Cihlar, 1996; Spanner et al,. 1990). Coniferous forests that are regularly
treated according to forest management practices tend to have relatively clumped and open
canopies. Thus, the role of the understory vegetation in forming boreal forest reflectance
cannot be neglected (Pisek et al., 2011).

1.3 Vegetation indices in LAI estimation

Canopy biophysical variables, such as LAI, can be estimated from remotely sensed data by
two types of algorithms: empirical models and methods that use physically-based radiative
transfer (RT) models. In empirical algorithms, the estimation is based on statistical
relationships modelled between concurrent ground reference measurements and surface
reflectance data. These relationships are typically expressed in the form of vegetation
indices (VI). VIs include various combinations of spectral bands designed to maximize the
sensitivity to vegetation characteristics while minimizing it to atmospheric conditions,
background, view and solar angles (Baret & Guyot, 1991; Myneni et al., 1995). Operational
LAI algorithms at global-scale typically make use of RT models, but the empirical models
usually outperform them in more localized applications.
The design of a VI that is optimally correlated with a particular vegetation property requires
good physical understanding of the factors affecting the spectral signal reflected from
vegetation. The sensitivity of a VI to a vegetation characteristic is typically maximized by
including bands with high sensitivity (e.g. high absorption) to the monitored entity and
bands mostly unaffected by the same entity. The simplest forms of VIs are simple
differences (RB1–RB2), ratios (RB1/RB2) and normalized differences [(RB1-RB2)/(RB1+RB2)] of
the reflectances of two spectral bands (RB1, RB2). (In Table 2 we give examples of common
VIs used in this study.) The most apparent characteristic of the green vegetation spectrum is
the pronounced difference between the red and NIR reflectances, the so called red-edge
around 700 nm. For example, the normalized difference vegetation index (NDVI) utilizes
this difference and has been shown to correlate with many interrelated vegetation attributes,
such as chlorophyll content, LAI, fractional cover, fAPAR and productivity.
The most commonly used VIs were designed for broadband sensors (one spectral band
spans about 50 nm or more) having red and NIR bands, such as NOAA AVHRR and
Landsat MSS (e.g. Tucker, 1979). However, the basic VIs in red and NIR spectral range
suffer from three well-known problems in LAI estimation: (1) they are not sensitive to LAI
over its natural range but tend to saturate already at moderate levels of LAI, (2) they are
sensitive to canopy background variability, and (3) the VI-LAI relationships are dependent
on the vegetation type. These VIs are also sensitive to atmospheric noise and correction.
The saturation of NDVI occurs typically at LAI levels of 2 to 6 depending on the vegetation
type and environmental conditions (e.g. Sellers, 1985; Myneni et al., 1997). In general, NDVI
saturates as the fractional cover of vegetation approaches one, although LAI still increases
(e.g. Carlson & Ripley, 1998). Over conifer-dominated boreal forests, NDVI varies typically
6 Remote Sensing – Applications

in a narrow range and shows poor relationships with canopy LAI (Chen & Cihlar, 1996;
Stenberg et al., 2004). The reason for this is the green understory, which results in a non-
contrasting background in the visible part of the spectrum (Nilson & Peterson, 1994; Myneni
et al., 1997).
Many modifications of basic VIs have been suggested to give better sensitivity to LAI.
Typical modifications use other visible bands than red (e.g. the green vegetation index,
GNDVI, Gitelson et al., 1996), try to reduce soil effects based on the soil line concept (e.g. the
soil adjusted vegetation index, SAVI, Huete, 1988), or include short wave infrared (SWIR)
bands. Many modifications also attempt to reduce atmospheric effects (e.g. the enhanced
vegetation index, EVI, Huete et al., 2002). The soil line is based on the observation that soil
reflectances fall in a line in the red-NIR spectral space (e.g. Huete, 1988). Many VIs utilize
the parameterized soil line in their calculation, but these VIs have not been successful in
boreal forests as bare soil is rarely visible (e.g. Chen, 1996).
The sensitivity of shortwave infrared (SWIR) reflectance to forest biophysical variables has
been recognized for a long time (e.g. Butera, 1986; Horler & Ahern, 1986) and several VIs
utilizing the SWIR band have been designed. Rock et al. (1986) showed that the moisture
stress index (MSI), i.e. the ratio of SWIR reflectance to NIR reflectance, was an indicator of
forest damage. Later, the ratio has commonly been referred to as the infrared simple ratio
(ISR, Chen et al., 2002; Fernandes et al., 2003). The SWIR reflectance has also been used for
adjusting NDVI (Nemani et al., 1993) and SR (Brown et al., 2000). The reduced simple ratio
(RSR) has been used specifically for estimating LAI (Brown et al., 2000; Stenberg et al., 2004)
and has been employed also in regional and global-scale operational algorithms (Chen et al.,
2002; Deng et al., 2006). RSR seems to reduce the sensitivity to the type and amount of
understory vegetation, because background reflectance varies less in SWIR than in visible
and NIR (Brown et al., 2000; Chen et al., 2002). RSR has also some capability to unify
coniferous and broadleaved forest types, which reduces the need for land cover type specific
LAI algorithms. However, in comparison to ISR, the use of red band makes RSR sensitive to
atmospheric effects (Fernandes et al., 2003). However, although inclusion of SWIR
reflectance increases the sensitivity of VIs to LAI, these indices also have a tendency to
saturate at high levels of LAI (e.g. Brown et al., 2000; Heiskanen et al., 2011).
Imaging spectroscopy provides much narrower spectral bands than typical multispectral
sensors. Due to the more detailed sampling of the vegetation spectra, such data can detect
specific absorption features of vegetation and therefore improve the estimation of vegetation
biochemical properties. For example, the SPOT 5 HRG sensors capture a spectral range from
500 nm to 1750 nm with four broad bands, in comparison to Hyperion’s 242 (10 nm wide)
bands between 400 nm and 2500 nm. At the canopy scale, the contents of biochemical
components and LAI are highly inter-related (e.g. Asner, 1998; Roberts et al., 2004).
Therefore, imaging spectroscopy could potentially improve LAI estimates. Furthermore,
there is potentially complementary information outside the typical spectral bands of
broadband sensors.
One way to utilize imaging spectroscopy data is to calculate narrow-band VIs in a similar
fashion as for broadband data but using narrower bands. The aim is to improve the
sensitivity of the VI to a specific vegetation biochemical property. For example, Ustin et al.
(2009) give a comprehensive review on VIs used as indicators of plant pigments
(chlorophyll, carotenoids and anthocyanin). The methods of estimating the non-pigment
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 7

biochemical composition of vegetation (water, nitrogen, cellulose and lignin), on the other
hand, are reviewed by Kokaly et al. (2009). Many of the developed indices have been
designed to work at leaf level and do not necessarily upscale to canopy level, because of the
high sensitivity to canopy structure, background, solar and view geometry. Another
approach is to find iteratively the simple combinations of bands that give the best
correlation with empirical data (e.g. Mutanga & Skidmore, 2004; Schlerf et al., 2005).
Most chlorophyll indices exploit the information in the red edge around 700 nm (Ustin et al.,
2009). Imaging spectroscopy data also enables the estimation of the red edge position (REP),
which is particularly sensitive to changes in chlorophyll content (e.g. Dawson & Curran,
1998). Water indices, on the other hand, utilize the water absorbing regions in the SWIR
region of the spectrum (e.g. Gao, 1996; Zarco-Tejada et al., 2003). Those indices seem
particularly interesting for LAI estimation considering the importance of the SWIR spectral
region in estimating LAI using broadband indices.
There is growing evidence that imaging spectroscopy data can improve LAI estimates in
comparison to broadband data by reducing the saturation effects. Depending on the
vegetation type and range of LAI, different types of VIs have been found useful. However,
the red edge indices have been most effective in estimating LAI of crops (Wu et al., 2010),
grasslands (Mutanga & Skidmore, 2004) and thicket shrubs (Brantley et al., 2011). On the
other hand, indices based on NIR and SWIR bands have been successful in broadleaved (le
Maire et al., 2008) and coniferous forests (Gong et al., 2003; Schlerf et al., 2005; Pu et al.,
2008). The importance of the SWIR spectral region in estimating boreal forest LAI has also
been emphasized by multivariate regression analysis (e.g. Lee et al., 2004). However,
broadband sensors can also have advantages over narrowband sensors in LAI estimation,
for example, by being less sensitive to noise due to the sensor, atmosphere and background
(e.g. Broge & Leblanc, 2000). Although there are case studies from different biomes, the
performance of narrowband VIs has been poorly assessed over European boreal forests.

2. Case study
2.1 Aims
The aim of the study is to establish the extent to which vegetation indices can be used to
measure variation in LAI based on a test site in southern boreal forest in Finland. We
explore different VIs in LAI estimation during full leaf development. We compare the
performance of narrowband VIs to traditional broadband VIs. The objective is to identify
VIs, which are least sensitive to species composition and, on the other hand, perform well in
coniferous stands.

2.2 Materials and methods

2.2.1 Study area
The study area, Hyytiälä, is located in the southern boreal zone in central Finland (61° 50'N,
24°17'E) and has an annual mean temperature of 3°C and precipitation of 700 mm.
Dominant tree species in the Hyytiälä forest area are Norway spruce (Picea abies (L.) Karst),
Scots pine (Pinus sylvestris L.) and Silver birch (Betula pendula Roth). Understory vegetation,
on the other hand, is composed of two layers: an upper understory layer (low dwarf shrubs
8 Remote Sensing – Applications

or seedlings, graminoids, herbaceous species) and a ground layer (mosses, lichens). The
growing season typically begins in early May and senescence in late August. We measured
twenty stands from the Hyytiälä forest area in July 2010 (see Section 2.2.2, Table 1). The
stands represented different species compositions that are typical to the southern boreal
forest zone in Finland.

Site Vegetation Site type Tree height, Basal area, LAI

m m2/ha
A4 Pine mesic 15.8 20.4 1.77
A5 pine, understory broadleaf mesic 18.6 24.3 2.67
B2 spruce, understory birch mesic 7.5 10 2.64
D3 pine, understory spruce & birch sub-xeric 17.8 20.5 2.37
D4 spruce, 25% birch mesic 16.5 27.5 3.72
E1 birch, spruce understory mesic 19.1 10.7 2.58
E5 50% spruce, 50% birch mesic 23.1 27.2 4.12
E6 50% spruce, 40% birch, 10% pine mesic 10.2 22.2 3.34
E7 Spruce mesic 13.3 31.7 3.91
F1 birch, spruce understory mesic 13.8 20.9 3.37
G4 spruce, 15% birch, 10% pine herb-rich 15.5 29.1 4.57
H3 Birch herb-rich 14.9 10.7 2.63
H5 Birch herb-rich 14.1 20.6 2.77
birch, understory pine, spruce mesic
I4 2.4 4 2.61
seedlings
T Spruce mesic 24.6 56 3.43
U16 Birch mesic 14 21 2.69
U17 birch, 10% spruce herb-rich 11.7 27 3.35
U18 65% pine, 25% spruce, 10% birch sub-xeric 16.5 26 3.45
U26 20% pine, 70% spruce, 10% birch mesic 16.8 24.9 2.43
U27 5% pine, 90% spruce, 5% birch mesic 15.2 20.9 2.63
[pine = Scots pine, spruce = Norway spruce, birch = Silver birch]
Table 1. Study stands.

2.2.2 Ground reference measurements

The LAI-2000 Plant Canopy Analyzer (PCA) is one of the most commonly used optical
devices to measure LAI. The PCA’s optical sensor includes five concentric rings of different
zenith angles (θ) (together covering almost a full hemisphere), which measure diffuse sky
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 9

radiation between 320-490 nm (LI-COR, 1992). Measurements by the PCA performed below
and above the canopy yield canopy transmittances, T(θ), for each ring. Finally, LAI is
calculated by numerical approximation of the integral (Miller, 1967):

 /2
LAI  2  ln[T ( )]cos sin  d (1)
0

There are four fundamental assumptions behind the LAI calculation method: 1. leaves
(needles) are optically black in the measured wavelengths (implying that canopy
transmittance closely corresponds to canopy gap fraction), 2. leaves (needles) are randomly
distributed inside the canopy volume, 3. leaves (needles) are small compared to the area of
view of the PCA’s rings, and 4. leaves (needles) are azimuthally randomly oriented. The LAI
estimate produced by Eq. 1 is commonly called effective LAI as the foliage elements are not
randomly organized but typically clumped (or grouped) together, which causes the estimate
produced by the PCA to be smaller than the “true” LAI (Chen et al., 1991; Deblonde et al.,
1994).
The LAI measurements can be done either with one or two PCA instruments. One PCA is
used for small plants such as crops, but for taller plants (e.g. trees), two units are necessary.
When only one instrument is used, the measurement is at first taken below and then above
the canopy. If two instruments are used, one instrument remains above the canopy and the
other one below the canopy. The use of two instruments is preferable since data are logged
nearly simultaneously with both sensors. The LAI estimate is calculated by combining
below and above canopy data. The measurements should be conducted under diffuse light
conditions; for example, when the sky has a full cloud cover or the sun angle is low (less
than 16 degrees). The radius of the sample plot should be at least three times the dominant
tree height as the PCA instrument has a relatively large opening angle.
In this study, the ground reference LAI (Table 1) was acquired by operating two LAI-2000
PCA instruments simultaneously. The instruments were intercalibrated before
measurements were performed. The reference sensor was located above the forest canopy
and set at a 15-second logging interval, while the other sensor was used inside the forest.
The sampling scheme was a ‘VALERI-cross’ (Validation of Land European Remote Sensing
Instruments, VALERI) which consists of two perpendicular 6-point transects. The distance
between two measurement points was four meters, so that the sampling scheme
corresponded roughly to a 20 m x 20 m plot. Measurement height was kept constant at 0.7
meters.

2.2.3 Satellite data

In this study, we used narrowband spectral data obtained from a Hyperion satellite image.
Hyperion is a narrowband imaging spectrometer aboard the National Aeronautics and
Space Administration (NASA) Earth Observer-1 (EO-1) satellite launched in 2000. Hyperion
captures data in the ‘pushbroom’ manner in 7.7 km wide strips using 242 spectral bands.
The spectral range of Hyperion is 356-2577 nm with each band covering a nominal spectral
range of 10 nm. Each pixel in a Hyperion image corresponds to an area of 30 m x 30 m on
the ground. During an acquisition, a scene with a length of either 42 km or 185 km is
recorded. Hyperion is in a repetitive, circular, sun-synchronous, near-polar orbit at an
10 Remote Sensing – Applications

altitude of 705.3 km measured at the equator. Thus, it can image almost any point on Earth
and it flies over all locations at approximately the same local time. The nominal revisit time
is 16 days, but due to the possibility of tilting the sensor, the potential revisit frequency is
higher. The scene used in this study was captured on 03 July 2010, and was provided
courtesy of the U.S. Geological Survey (USGS) Earth Explorer service.
Out of the potential 242 spectral bands, several lack illumination (due to the absorption in
the atmosphere or a decrease of incident solar spectral irradiance in the longer infrared
wavelengths) or have a very low spectral response. This leaves the user with 198 usable
spectral bands: bands 8-57 in the visible and NIR (wavelengths 436-926 nm) and bands 77-
224 in SWIR (wavelengths 933-2406 nm) (Pearlman et al., 2003). Hyperion images have
several known deficiencies which can be corrected using algorithms given in scientific
literature. Firstly, Hyperion suffers from systematic striping in along-track direction of the
image. The stripes are characteristic to all pushbroom sensors. Instruments belonging to
this broad class have a different receiving element for each image line. Hyperion has thus
256 radiation-sensitive elements for each spectral band, each seeing a separate 30 m strip
of the ground, thus producing the 7.7 km wide image. The striping can be broadly
divided into two classes, completely missing lines (due to non-functioning receiving
elements) and actual stripes (arising from slightly different sensitivities of the 256
receivers). We removed the actual striping using Spectral Moment Matching (SpecMM),
outlined by Sun et al. (2008), which uses the average and standard deviation statistics
between highly correlated bands to remove stripes. Next, the missing lines containing no
information were identified and corrected using the values from spatially adjacent pixels
using local destriping (Goodenough et al., 2003). The results of the destriping can be seen
in Figure 1.

Fig. 1. Hyperion band 8 (436nm) uncorrected image (left), and corrected using Spectral
Moment Matching and local destriping (right).

The second known defect in Hyperion imagery is a shift in the wavelength of each column
in the across track direction from the band central wavelength. This shift, known as spectral
smile, is also characteristic to pushbroom sensors and is a result of different optical paths
leading to the different receiving elements. The shift is a function of wavelength and the
position of the receiving element in the receiving array. As is the case for most instruments,
the “smile” manifests itself in Hyperion imagery as a “frown”, with the wavelengths of the
columns near the edges of each band shifting negatively from the bands average wavelength
(Figure 2). The smile was corrected using the pre-launch laboratory measured spectral shift
(Barry, 2001). We used interpolation to bring each individual pixel to a common central
wavelength based on the pre-launch calibration measurements.
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 11

Fig. 2. Laboratory measured spectral shift of Hyperion (Barry, 2001).

The signal received by the Hyperion instrument consists of the photons scattered by the
atmosphere as well as the ground surface. To study surface reflectance, the influence of the
atmosphere needs to be eliminated in a process commonly known as atmospheric
correction. We performed this correction using an algorithm known as Fast Line-of-sight
Atmospheric Analysis of Spectral Hypercubes (FLAASH, Matthew et al, 2000). FLAASH is
an absolute atmospheric correction that incorporates the MODTRAN4 radiation transfer
code to model the scattering and transmission properties of the atmosphere at the time of
image capture (San & Suzen, 2010). The FLAASH algorithm is incorporated into the ITT
Visual Information Solutions (ITT VIS) ENVI software. For processing, FLAASH requires an
input value for visibility to estimate atmospheric aerosol levels, in addition to basic
geographic and temporal details about the scene. The visibility can be recalculated by
FLAASH, using a ratio between dark pixels at 600 nm and 2100 nm. However, a more
accurate estimate of visibility was achieved using ground based optical measurements from
a weather station in the area.
The final processing stage is to resample the image pixels into a geographic coordinate
system, known as geocorrection. This was done using a polynomial transformation to a
vector base map from the National Land Survey of Finland. The Hyytiälä area contains
numerous roads, providing a large number of easily identifiable potential ground control
points (GCPs) at intersections. Around 20 GCPs were selected, with a root mean square
error of 0.4 pixels being achieved. Bilinear interpolation was chosen for resampling the
image pixels due to the better geometric accuracy over nearest neighbour.
The final product is a geocorrected image of the surface hemispherical-directional
reflectance factors (HDRF) of the Hyytiälä area. To validate the atmospheric correction, we
compared the HDRF to a field measured reflectance factor. A soccer field of about 130 m by
60 m in the area was sampled during the summer of 2010 every two to three weeks using an
ASD handheld portable spectroradiometer covering a spectral range from 325-1075 nm. The
sampling was done using a transect approach with 42 measurements at around 1 meter
intervals. The final hemispherical-conical reflectance factor (HCRF) used for the comparison
is an average of the transect representing the average for the whole field. While no ground
measurements fell on the exact date of the Hyperion image, the ground measured spectra
was interpolated to dates between two measurements. After interpolation the ground
measured HCRF was binned into corresponding Hyperion bands using the spectral
response of each band.
12 Remote Sensing – Applications

Fig. 3. Comparison of a soccer field’s spectral reflectance factors from in situ radiometric
measurements and corrected Hyperion data.

Overall, there is a very good correlation between the field measured reflectance and the
fully processed Hyperion reflectance (Fig. 3). An overall RMSE of 1.8% is achieved, which
gives us confidence in the validity of the pre-processing and atmospheric correction.
However, as the in situ spectrum is considerably smoother than the one measured from the
satellite, a considerable amount of noise is also present in the satellite-derived HDRF.

2.2.4 Vegetation indices and statistical analysis

First, we studied how HDRFs in single bands are correlated with LAI. Next, in order to
evaluate narrow-band VIs for estimating LAI, we did regression analyses between various
VIs and LAI. We used two approaches to select narrowband indices: 1) We made a literature
survey for narrow-band VIs that have been designed to estimate foliage biochemical
components. (A collection of VIs showing the highest R² with LAI are shown in Table 2.) 2)
We calculated all the possible Ratio Indices (RI) and Normalized Difference Indices (NDI) of
Hyperion bands and correlated them with LAI. In other words, the first approach also
contains VIs combining several bands and the second approach aims to identify the simple
two-band VIs that best correlate with LAI.
To facilitate the comparison of narrowband VIs with broadband indices, we calculated
synthetic HDRFs based on Landsat 7 ETM+ bands. The HDRFs were calculated according to
Jupp et al. (2002) using the ETM+ spectral sensitivity functions, and Hyperion’s central
wavelengths and bandwidths. Four broadband indices were calculated for comparison, SR,
NDVI, ISR and RSR (Table 2). All these indices have been used for LAI estimation in various
biomes. SR and NDVI were included for reference, and ISR and RSR because they have
shown best performance over conifer-dominated boreal forests (see 1.3).
We analyzed the data both by grouping all the sample plots together and separately for
coniferous plots (> 75% of the trees were Scots pines or Norway spruces). In the birch-
dominated stands, the variation in LAI was too small for reliable regression analysis.
We studied only linear relationships. The strength of the relationship was assessed by the
coefficient of determination (R2) and the root mean square error (RMSE).
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 13

Abbr. Index Formula Reference Bands

applied
Indices concentrating on the red-edge
Rouse et al. (1974),
ETM+3,
SR Simple Ratio SR = RETM+4/RETM+3 Birth & McVey
ETM+4
(1968)
Normalized Difference NDVI = (RETM+4- ETM+3,
NDVI Rouse et al. (1974)
Vegetation Index RETM+3)/(RETM+4+RETM+3) ETM+4
REP = 700+ (((R773 +1,5
Danson & Plummer 773, 662,
REP Red Edge Position *R662) - R692) / (R733-R692))
(1995) 692, 733
*(740-700)
Indices concentrating on pigment content
Pigment-Specific
PSSRa PSSRa = R803/R681 Blackburn (1998) 681, 803
Simple Ratio – chla
Water sensitive indices
Rock et al. (1986),
MSI = Moisture Stress Index ETM+4,
ISR = RETM+5/RETM+4 Fernandes et al.
ISR = Infrared Simple Ratio ETM+5
(2002)
RSR = (RETM+4/RETM+3) *
ETM+3,
((RETM+5_min – RETM+5)
RSR Reduced Simple Ratio Brown et al. (2000) ETM+4,
/(RETM+5_max –
ETM+5
RETM+5_min))
Table 2. Vegetation indices investigated in this study. The symbol R refers to the HDRF.
Subscripts refer to the applied ETM+ band or the central wavelength (in nm) of the
Hyperion band

2.3 Results
2.3.1 General characteristics of forest spectra
Two examples of forest reflectance factors (HDRFs) are presented in Figure 4. To allow
relating the vegetation spectra to satellite signals, the sensitivity functions of the
corresponding ETM+ bands are shown. Note the correspondence of ETM+2 with the green
peak, ETM+3 with the red local minimum and ETM+4 with the plateau in the NIR. The red-
edge slope (between ETM+ bands 3 and 4) is not covered by ETM+ bands. ETM+5 and
ETM+7 catch the signal in the shortwave infrared region (SWIR-1 (here: 1470-1800 nm) and
SWIR-2 (here: 2030-2360 nm) respectively), avoiding the two strong water absorption bands
in-between.
The average reflectance of coniferous stands is slightly lower in the green region and
decidedly lower in the NIR than the reflectance of birch stands. In SWIR-1 (covered by
ETM+5) the reflectances become more comparable, and in SWIR-2 (covered by ETM+7) the
signals almost meet.
14 Remote Sensing – Applications

Fig. 4. Average conifer and birch-dominated stand spectra. The grey lines show the spectral
sensitivity of the ETM+ bands.

2.3.2 Regression analysis for single bands

The different average HDRF for the two forest types (Fig. 4) results in different correlations
of the satellite bands to LAI (Fig. 5).

Fig. 5. Correlation coefficient of LAI with ETM+ and Hyperion spectral bands for all sample
stands, and separately for conifer sample stands.

The correlation coefficients for all stands varied between -0.6 and -0.038. All correlations
were negative, except for the two Hyperion bands centred at 2345 nm and 2355 nm. Two
important regions (green and NIR) had almost no correlation with LAI. Only the absorption
peak of chlorophyll produced a strong negative correlation at 681nm. The SWIR correlations
were also mostly negative.
For conifer stands, correlation coefficients varied between -0.7 and 0.6. The first peak was at
549 nm, in the middle of the green band, followed by a strong negative correlation in the red
with a peak at 681 nm. In the NIR a strong positive correlation was observed again. A slight
shoulder began at 712 nm, with a plateau at 752 nm. In the SWIR, correlation coefficients
were very close to those of all stands.
Fig. 5 also shows the correlation of the ETM+ bands to LAI. The lower spectral resolution
averages wider wavelength ranges and therefore shows less variation in correlation
coefficients.
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 15

2.3.3 Correlation of vegetation indices to LAI for all sample plots

The best broadband index analysed here was the Infrared Simple Ratio (ISR, R2 = 0.56),
followed by the Reduced Simple Ratio (RSR, R2 = 0.40) (Table 3). The best narrowband
combinations (either RI or NDI) showed more potential with R2s exceeding 0.65 (Table 3,
Fig. 6). If there were several indices based on neighbouring bands (within 10 nm) we chose
the best one to Table 3.

VI Bands applied R2 RMSE RMSE RMSE

Conifer Broadleaf
broadband indices using simulated ETM+
ISR ETM+4, ETM+5 0.56 0.44 0.42 0.25
RSR ETM+3, ETM+4, ETM+5 0.40 0.52 0.59 0.31
NDVI ETM+3, ETM+4 0.09 0.64 0.68 0.51
SR ETM+3, ETM+4 0.04 0.66 0.73 0.46
narrowband indices using Hyperion
RI 1134, 1790 0.71 0.36 0.34 0.38
NDI 1134, 1790 0.68 0.38 0.36 0.39
RI 732, 1790 0.67 0.38 0.42 0.31
RI 1074, 1790 0.67 0.38 0.40 0.34
RI 885, 1790 0.67 0.39 0.37 0.35
RI 854, 1790 0.66 0.39 0.37 0.34
RI 1003, 1639 0.66 0.39 0.39 0.26
RI 1044, 1790 0.66 0.39 0.39 0.37
NDI 732 1790 0.66 0.39 0.42 0.33
NDI 1084, 1286 0.66 0.39 0.43 0.22
Table 3. Indices most correlated with LAI for all sample plots. RMSE was also calculated
separately for each forest class. Bands for Hyperion refer to the central wavelength (in nm).

The best band combinations for RI and NDI indices were very similar (Fig. 6). A strong
correlation with LAI existed for bands combining the region between 730 to 900 nm and
1130 to 1350 nm. Another interesting region was within SWIR-1; especially strong was the
correlation around 1780 and 1790 nm. These bands also showed up in the best performing
indices for all forest classes combined (Table 3).
The two best narrowband indices for all forest plots were the RI (R2 = 0.71, RMSE = 0.36)
and NDI (R2 = 0.68, RMSE = 0.38) based on bands centred at 1134 and 1790 nm (Table 3).
This is consistent with the best broadband index (ISR) which also combines NIR and SWIR.
The same spectral regions are used by all the other best indices except two cases including a
band in the red-edge (732 nm). Examples of the strongest relationships are shown in Fig. 7.
However, when looking at the RMSE for conifer and broadleaf stands (Table 3) it became
apparent that for some indices (e.g. NDI based on 1084 nm and 1286 nm: RMSE = 0.43 for
conifers and RMSE = 0.22 for broadleaf) their LAI was correlated differently to the same VI.
16 Remote Sensing – Applications

Fig. 6. Matrixes showing the R2 between LAI and simple narrowband indices calculated for
all possible combinations of Hyperion bands. The indices are defined as follows:
RI=Band1/Band2, and NDI=(Band1-Band2)/(Band1+Band2).

Fig. 7. The relationship of LAI and two best ratio indices (RI).

2.3.4 Correlations for coniferous dominated forest plots

The performance of the broadband indices for conifer-dominated stands was much better
than over all sample stands. R2 now ranged from 0.60 to 0.79, and NDVI showed the best
correlation with LAI, followed by SR.
The best performing narrowband index over coniferous forest was neither RI nor NDI but
REP (R² = 0.89) calculated according to the method of Danson & Plummer (1995) (Table 2).
This index combined four bands in the visible and NIR; an area also represented in several
of the other indices which best correlated with LAI in coniferous stands.
The matrixes for all band combinations of Hyperion bands over conifer-dominated stands
(Fig. 8) showed wider spectral regions of high correlation than for all stands (Fig. 6).
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 17

Fig. 8. Matrixes showing the R2 between LAI and two narrowband indices calculated for all
possible combinations of Hyperion bands for conifer-dominated stands.

Fig. 9. The relationship of LAI and the two best performing narrowband indices for conifer-
dominated stands.

Most of the indices with the highest correlations to LAI in coniferous stands used bands
around the red-edge. Almost all of them (e.g. the Pigment-Specific Simple Ratio Index for
chlorophyll a, PSSRa) applied the Hyperion band centred at 681nm, the peak of chlorophyll
a absorption. Exceptions were the RI and NRI using the bands centred at 1185 and 1790 nm
(i.e. combining NIR and SWIR), and RI and NDI using bands centred at 518 and 773 nm (i.e.
combining carotene absorption and NIR).
Scatterplots for the two best indices for coniferous stands are shown in Fig. 9. In both cases,
coniferous plots differed considerably from the other plots. This was indicated also by the
high RMSE for all stands (up to 1.42, Table 4). However, for indices using NIR and SWIR
(e.g. RI and NDI based on 1185 and 1790 nm) the differences were less pronounced. The VI
showing the lowest RMSE for all stands (0.49) was the RI (1185 and 1790 nm) with an R2 for
conifer stands of 0.86 and RMSE 0.29.
18 Remote Sensing – Applications

VI Bands applied R2 RMSE RMSE All

stands
broadband indices using simulated ETM+
NDVI ETM+3, ETM+4 0.79 0.36 1.20
SR ETM+3, ETM+4 0.78 0.36 1.56
ISR ETM+4, ETM+5 0.71 0.42 0.44
RSR ETM+3, ETM+4, ETM+5 0.60 0.50 0.90
narrowband indices using Hyperion
REP 671, 702, 742, 783 0.89 0.26 1.29
NDI 681, 773 0.88 0.27 1.02
RI 681, 773 0.88 0.28 1.01
RI 1185, 1790 0.86 0.29 0.49
NDI 1185, 1790 0.86 0.30 0.50
NDI 681, 742 0.85 0.30 1.01
NDI 681, 824 0.85 0.30 0.98
RI 681, 742 0.85 0.31 0.99
NDI 518, 773 0.85 0.31 1.42
PSSRa 803, 681 0.85 0.31 1.30
RI 518, 773 0.85 0.31 1.39
Table 4. Indices most correlated with LAI in conifer-dominated plots. and RMSE for
R2
conifer-dominated stands, and RMSE separately for all stands. Bands for Hyperion refer to
the central wavelength (in nm).

2.4 Discussion
In our case study, the narrowband VIs provided more accurate LAI estimates than the
broadband VIs synthesized from the same data in a boreal forest study site. The best
narrowband combinations showed relatively strong linear relationships with LAI (R2 >
0.65), although the Hyperion image was acquired in the middle of the growing season when
LAI is the highest. The relationships were even stronger if the analysis was restricted to the
conifer stands (R2 > 0.85). The results are promising as common broadband VIs tend to
saturate at the highest LAI values. The improvement of estimation accuracy is in agreement
with the previous studies, which have emphasized the potential of narrowband VIs for
estimating forest canopy LAI (e.g. Lee et al., 2004; Schlerf et al., 2005; Brantley et al., 2011;
Wu et al., 2010).
Most of the narrowband VIs showing the strongest relationships with LAI were based on
reflectances in the far red and at the red edge (680—740 nm), NIR (e.g. 885 and 1134 nm) and
SWIR (e.g. 1639 nm and 1790 nm) wavelength regions (Figure 10). Many of the most
important spectral regions are not covered by the ETM+ spectral bands, and the spectral
regions are very narrow in comparison to the ETM+ bands.
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 19

Fig. 10. Spectral regions used by the indices showing the strongest relationships with LAI
over all sample stands and conifer stands.

The NIR and SWIR spectral bands were particularly important when all sample plots were
analyzed together. This is in agreement with the best broadband indices, ISR and RSR. The
importance of NIR and SWIR bands has been emphasized also in previous studies testing
narrowband VIs for estimating forest LAI (e.g. Lee et al., 2004; Schlerf et al., 2005). The leaf
(needle) reflectance at those wavelengths is mainly controlled by water absorption, although
leaf biochemical components such as proteins, cellulose and lignin also contribute to
absorption in the infrared (e.g. Curran, 1989). The amount of water at the canopy level is
directly related to LAI, which explains strong correlations. The bands centered at 1134 nm
and 1790 nm are among the Hyperion bands, which are closest to the water absorption
regions centered at approximately 1200 nm and 1940 nm. The spectral bands close to the
water absorption regions at 970 nm and 1400 nm are also employed in some of the best
indices. The spectral bands of the broadband sensors are usually placed in the middle of the
atmospheric windows to avoid atmospheric absorption. However, it seems that narrow
spectral bands close to the water absorption regions are particularly interesting for
estimating LAI. In these wavelength regions, the reflectance seems to be relatively
insensitive to tree species or composition of the understory vegetation, as suggested earlier
by the studies using broadband indices (e.g. Brown et al., 2000).
When pure coniferous stands were studied separately, the relationships became stronger
and the far red and red edge spectral bands were included in several of the best VIs.
However, the improvement in accuracy relative to the best VI based on NIR and SWIR
reflectance (RI based on bands centered at 1185 nm and 1790 nm) was rather modest. The
best broadband indices were NDVI and SR, which are based on ETM+ red and NIR bands.
Usually, NDVI has shown relatively weak relationships with LAI in conifer dominated
boreal forest (e.g. Stenberg et al., 2004).
The strongest relationship with LAI was provided by the red edge position (REP) calculated
by the method proposed by Danson and Plummer (1995). In general, the REP is considered
to be sensitive to leaf and canopy chlorophyll content, so that increasing the amount of
chlorophyll, or LAI, is related to the longer REP wavelength because of the widening of the
chlorophyll absorption region at approximately 680 nm (Danson & Plummer, 1995; Dawson
20 Remote Sensing – Applications

and Curran, 1998; Sims & Gamon, 2002; Pu et al., 2003). In comparison to SWIR spectral
bands, the far red and red edge spectral region is sensitive to species composition, shown as
poor relationships over mixed vegetation. However, sometimes poor relationships between
the REP and LAI have been reported even for pure coniferous stands (Blackburn, 2002).
However, although the REP calculated in this study showed strong correlation with
coniferous LAI, the estimated wavelengths do not correspond to the Red Edge Inflection
Point (REIP), i.e. the steepest slope of the red-edge. The wavelengths are considerably
longer. Therefore, the unusual inverse relationship between REP and LAI in this study is
explained by the calculation method (Danson and Plummer, 1995). Alternative calculation
methods for REP are summarized, for example, by Pu et al. (2003).
Although many studies testing narrowband VIs for LAI estimation have stressed the
potential of the red edge and SWIR spectral regions, the specific spectral bands providing
the strongest relationships with LAI vary between the studies. Also in our case study, the
optimal band combinations provided stronger relationships with LAI than VIs collected
from the literature. This is somewhat expected, as the number of spectral bands and their
possible combinations is so large that empirically determined optimal band combinations
are likely to depend heavily on the local environmental conditions and type of satellite
image data. For example, approximately 150 useful spectral bands of Hyperion make more
than 20,000 two-band combinations. Because of this, the optimal indices cannot necessarily
be generalized very well. Furthermore, a large number of spectral bands combined with a
small number of sample plots increase the risk that the regression models are overfitted.
However, this should be mostly a problem with multivariate approaches (e.g. Lee et al.,
2004). Moreover, when comparing broadband and narrowband indices, it should be noted
that we used only synthesized ETM+ data and the results could differ to some extent if true
ETM+ data would have been used instead (Lee et al., 2004). This is because the synthetic
broadband data is affected by the lower signal-to-noise ratio of the narrow spectral bands,
even if data are averaged.

3. Future perspectives
Wider use of imaging spectroscopy data is hampered by the availability of the data. Today,
mostly airborne instruments are used to produce remote sensing data with high spectral
resolution. Airborne measurements are associated with relatively small spatial coverage and
high operating costs falling directly to data users. The Hyperion sensor used in this case
study is a rare exception: it is the only true imaging spectrometer in orbit today, providing
wide spectral coverage with uniform spectral resolution and contiguous bands. The scene,
however, is about to change. At the end of the decade (i.e., around 2020), NASA is planning
to launch the HyspIRI mission, providing narrowband data with routine global coverage
(Samiappan et al., 2010). Before HyspIRI, several national space programs are striving to
launch satellites with capability to produce narrowband data (e.g. the EnMAP instrument,
Segl et al., 2010). Therefore, the need for developing algorithms that would make use of the
advanced properties of narrowband data, compared to the more traditional multispectral
data, is evident.
In this case study, we used narrowband VIs to relate forest LAI to remotely sensed
reflectance signals. Historically, vegetation indices have been among the very first tools in
interpreting multispectral remote sensing data from vegetated areas. Later, physically-based
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 21

reflectance modelling has taken over the role of the preferred method in large-scale
retrievals of vegetation biophysical variables. Similar developments may take place in the
interpretation of narrowband imaging spectroscopy data. However, let us first take a closer
look at narrowband indices as they are used in the current study.
As discussed above (section 1.3), VIs are usually treated as empirical (or, at least semi-
empirical) tools in remote sensing. However, it has been known for a long time that the
reflectance indices convey also some information on the physical processes related to the
interaction of light with plant elements. Indeed, Myneni et al. (1995) showed that the
common indices are actually derivatives of canopy reflectance and are physically related to
abundances of absorbing pigments. For this reason, indices commonly make use of two
spectral regions: one inside the spectral region where the absorption of a pigment is strong,
and one outside the absorption band. The use of red and near-infrared wavelengths thus
corresponds to measuring the abundance of one of the most vital plant pigments,
chlorophyll.
Can such an interpretation be extended to narrowband indices? From the point-of-view of
the physics of radiative transfer, there is no fundamental difference between broad- and
narrowband indices. However, for calculating a spectral derivative, there is little use of well-
tuned and potentially much noisier narrow spectral bands. For detecting pigments whose
absorption spectra span tens, if not hundreds of nanometers, broadband indices seem a
much more robust tool. Further, vegetation indices, especially early ones like the NDVI,
have been shown both empirically and on the basis of theoretical studies, to be sensitive to
factors others than those of interest, such as soil brightness changes and atmospheric effects.
Most narrowband indices can be viewed as finely tuned versions of their older broadband
counterparts. Site-specific selection of wavelengths leads to a better explanatory power of
narrowband VIs as we also demonstrated in this case study. Unfortunately, the fine tuning
for eliminating environmental effects makes narrowband indices potentially even more site-
specific than broadband ones.
The comparison of narrowband and broadband VIs presented above did not concern indices
capturing truly narrowband effects, e.g. the photochemical reflectance index PRI (Gamon et
al., 1992) or various red edge parameters. Intrinsically narrowband VIs are based on effects
that cannot be detected from broadband data. These indices are not more site-specific than
broadband indices and do indeed, due to a finer spectral resolution, provide additional
information on vegetation cover on all scales. Similarly, the red edge parameters calculated
above make use of the high spectral resolution of narrowband data in a manner which is not
site-specific. Therefore, it is not surprising that they provide a good fit for estimating forest
stand variables regardless of dominating species.
An alternative to using narrowband indices would be to invert a full canopy reflectance
model: the goals of both methods are to retrieve information on some biophysical variable
of interest (Rautiainen et al., 2010). As discussed in this chapter, the theoretical
foundations of the two approaches are somewhat similar. However, obvious limitations of
index-based inversions lie in that it is not possible to define a spectral index sensitive to
only one process, nor is it possible to design a universal spectral index which would be
optimal for all applications everywhere and all the time (Verstraete & Pinty, 1996).
Further, since vegetation indices carry only part of the information available in the
22 Remote Sensing – Applications

original channel reflectances, they assume that the information of interest is contained
exclusively in the observed spectral variations. VIs also often neglect the effects of surface
anisotropy associated with the specific geometry of illumination and observation at the
time of the measurements (Govaerts et al., 1999). Last, but not least, a fundamental
shortcoming of the index-based approach lies in its potentially wide application area. A
user not directly working in the field of remote sensing science may be distracted by a
statistically strong dependence between a variable of interest (e.g. an ecological parameter
describing diversity) and a vegetation index. However, canopy reflectance signals can
carry information only on what are known as state variables of radiative transfer
(abundances of optically active substances, canopy amount and structure, etc.). Other
variables may be correlated with one or more of the state variables, but before drawing
conclusions based on such correlations, the nature and application range of the correlation
should be clarified.
Naturally, physical canopy reflectance models are immune to the problems listed above.
When working in the forward mode, a modern reflectance model can reliably predict the
spectral reflectance signal of a vegetation canopy given the required inputs (e.g.
Widlowski et al., 2007). When run in inverse mode, the models should be able to produce
an estimate of the state variables of radiative transfer based on measured spectral
reflectance values. Unfortunately, due to the large number of the state variables and the
mathematical nature of the inverse problem, a robust result is difficult to achieve (Baret &
Buis, 2008). Despite the present-day problems with inverting canopy reflectance models, it
is clear that physical models hold a clear advantage over index-based biophysical
parameter estimation, especially when using imaging spectroscopy data. Physical models
account for changes in environmental conditions and estimate all state variables
simultaneously. They also have the advantage of failing if unphysical data is fed to them
(e.g. due to sensor failure or preprocessing error) instead of producing unrealistic results.
The problem with the large number of state variables can be solved by the larger
information content of imaging spectroscopy data (compared with that produced by
multispectral sensors) and development of novel physically based parameterizations
allowing a more efficient description of canopy structure. However, until the full potential
of imaging spectroscopy has been utilized by the developers of physical models,
narrowband vegetation indices remain valuable tools in exploring the richness of high
spectral resolution data.

4. Acknowledgment
We thank Anu Akujärvi for assisting in the field measurements. This study was funded by
Emil Aaltonen Foundation, University of Helsinki Research and Postdoctoral Funds, and the
Academy of Finland. Hyperion EO-1 data was available courtesy of the U.S. Geological
Survey.

5. References
Asner, G.P. (1998). Biophysical and biochemical sources of variability in canopy reflectance.
Remote Sensing of Environment, Vol. 64, No. 3, (June 1998), pp. 234-253, ISSN 0034-
4257
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 23

Baret, F., Buis & S. (2008), Estimating canopy characteristics from remote sensing
observations: review of methods and associated problems, In: Advances in Land
Remote Sensing: System, Modeling, Inversion and Application, S. Liang (ed.), pp. 173-
201, Springer-Verlag, ISBN 978-1-4020-6449-4, Heidelberg
Baret, F., Guyot, G. (1991). Potentials and limits of vegetation indices for LAI and APAR
assessment. Remote Sensing of Environment, Vol. 35, No. 2-3, (February-March 1991),
pp. 161-173, ISSN 0034-4257.
Barry, P. (2001). EO-1/ Hyperion Science Data User’s Guide, Level 1_B, TRW Space, Defense &
Information Systems, Redondo Beach, California, USA
Birth, G. S. & McVey, G. R. (1968) Measuring the color of growing turf with a reflectance
spectrometer. Agronomy Journal, Vol. 60 No. 6, (March 1968), pp. 640-643, ISSN: 0002-
1962
Black, T.A., Chen, J.M., Lee, X. & Sagar, R.M.. (1991). Characteristics of shortwave and
longwave irradiances under a Douglas fir forest stand. Canadian Journal of Forest
Research Vol. 21, No. 7, (July 1991), pp. 1020–1028, ISSN 0045-5067
Blackburn, G. A. (1998). Spectral indices for estimating photosynthetic pigment
concentrations: a test using senescent tree leaves. International Journal of Remote
Sensing, Vol. 19, No. 4, (March 1998), pp. 657-675, ISSN 0143-1161
Blackburn, G. A. (2002). Remote sensing of forest pigments using airborne imaging
spectrometer and LIDAR imagery. Remote Sensing of Environment, Vol. 82, No. 2-3,
(October 2002), pp. 311-321, ISSN 0034-4257.
Brantley, S.T., Zinnert, J.C. & Young, D.R. (2011). Application of hyperspectral vegetation
indices to detect variations in high leaf area index temperate shrub thicket
canopies. Remote Sensing of Environment Volume 115, Issue 2, (February 2011), pp.
514-523, ISSN 0034-4257
Broge, N.H. & Leblanc, E. (2000) Comparing prediction power and stability of broadband
and hyperspectral vegetation indices for estimation of green leaf area index and
canopy chlorophyll density. Remote Sensing of Environment, Vol. 76, No. 2, (May
2000), pp. 156-172, ISSN 0034-4257
Brown, L., Chen, J.M., Leblanc, S.G. & Cihlar, J. (2000). A shortwave infrared modification to
the simple ratio for LAI retrieval in boreal forests: An image and model analysis.
Remote Sensing Environment, Vol. 71, No. 1, (January 2000), pp. 16-25, ISSN 0034-
4257.
Butera, M.K. (1986). A Correlation and Regression Analysis of Percent Canopy Closure
Versus TMS Spectral Response for Selected Forest Sites in the San Juan National
Forest, Colorado. IEEE Transactions on Geoscience and Remote Sensing, Vol. GE-24,
No. 1, (January 1986), pp. 122-129, ISSN 0196-2892.
Carlson, T.N. & Ripley, D.A. (1998). On the Relation between NDVI, Fractional Vegetation
Cover, and Leaf Area Index. Remote Sensing of Environment, Vol. 62, No. 3,
(December 1997), pp. 241-252, ISSN 0034-4257.
Chen, J.M. (1996). Evaluation of vegetation indices and modified simple ratio for Boreal
applications. Canadian Journal of Remote Sensing, Vol. 22, No. 3, (June 1996), pp. 229-
242, ISSN 0703-8992.
24 Remote Sensing – Applications

Chen, J.M. & Black T.A. (1991). Measuring leaf-area index of plant canopies with branch
architecture. Agricultural and Forest Meteorology Vol. 57, No. 1-3, (December 1991),
pp. 1-12, ISSN 0168-1923
Chen, J.M. & Black, T.A. (1992). Defining leaf area index for non-flat leaves. Plant Cell &
Environment, Vol. 15, No. 4, (May 1992), pp. 421-429, Online ISSN 1365-3040
Chen, J. & Cihlar, J. (1996). Retrieving leaf area index of boreal conifer forests using Landsat
TM images. Remote Sensing of Environment Vol. 55, No. 2, (February 1996), pp. 153-
162, ISSN 0034-4257
Chen, J., Pavlic, G., Brown, L., Cihlar, J., Leblanc, S., White, P., Hall, R., Peddle, D., King, D.,
Trofymow, J., Swift, E., Van der Sanden, J., & Pellikka, P. (2002). Derivation and
validation of Canada-wide coarse-resolution leaf area index maps using high-
resolution satellite imagery and ground measurements. Remote Sensing of
Environment, Vol. 80, No. 1, (April 2002), pp. 165-184, ISSN 0034-4257.
Curran, P.J. (1989). Remote Sensing of Foliar Chemistry. Remote Sensing of Environment, Vol.
30, No. 3, (December 1989), pp. 271-278, ISSN 0034-4257.
Danson, F. M., & Plummer, S. E., 1995, Red edge response to forest leaf area index.
International Journal of Remote Sensing, 1995, Vol. 16, No. 1, (January 1995), pp. 183-
188, ISSN 0143-1161
Dawson, T.P. & Curran, P.J. (1998). A new technique for interpolating the reflectance red
edge position. International Journal of Remote Sensing, Vol. 19, No. 11, (July 1998), pp.
2133-2139, ISSN 0143-1161
Deblonde, G., Penner, M. & Royer, A. (1994). Measuring Leaf Area Index with LI-COR LAI-
2000 in Pine Stands. Ecology. Vol. 75, No. 5, (July 1994), pp. 1507-1511, ISSN
00129658
Deng, F., Chen, J., Plummer, S., Chen, M. & Pisek, J. (2006). Algorithm for global leaf area
index retrieval using satellite imagery. IEEE Transactions on Geoscience and Remote
Sensing, Vol. 44, No. 8, (August 2006), pp. 2219-2229, ISSN 0196-2892
Eklundh, L., Harrie, L. & Kuusk, A. (2001). Investigating relationships between Landsat
ETM+ sensor data and leaf area index in a boreal conifer forest. Remote Sensing of
Environment, Vol. 78, No. 3, (December 2001), pp. 239-251, ISSN 0034-4257
Eriksson, H., Eklundh, L., Kuusk, A., & Nilson, T. (2006). Impact of understory vegetation
on forest canopy reflectance and remotely sensed LAI estimates. Remote Sensing of
Environment, Vol. 103, No. 4, (August 2006), pp. 408−418, ISSN 0034-4257
Fernandes, R., Butson, C., Leblanc, S. & Latifovic, R. (2003). Landsat-5 TM and Landsat-7
ETM+ based accuracy assessment of leaf area index products for Canada derived
from SPOT-4 VEGETATION data. Canadian Journal of Remote Sensing, Vol. 29, No. 2,
(April 2003), pp. 241-258, ISSN 0703-8992.
Fernandes, R., Leblanc, S., Butson, C., Latifovic, R. & Pavlic, G. (2002). Derivation and
evaluation of coarse resolution LAI estimates over Canada. In: Proceedings of
Geosciences and Remote Sensing Symposium, 2002. IGARSS 02.2002, IEEE
International Publications, Vol. 4, pp. 2097-2099, ISBN 0-7803-7536-X
Gamon, J., Penuelas, J., & Field, C. (1992). A narrow-waveband spectral index that tracks
diurnal changes in photosynthetic efficiency. Remote Sensing of Environment, Vol. 41,
No. 1, (July 1992), pp. 35-44, ISSN 0034-4257
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 25

Gao, B-C. (1996). NDWI–A normalized difference water index for remote sensing of
vegetation liquid water from space. Remote Sensing of Environment, Vol. 58, No. 3,
(December 1996), pp. 257-266, ISSN 0034-4257.
Gitelson, A. A., Kaufmann, Y. J. & Merzlyak, M. N. (1996) Use of a green channel in remote
sensing of global vegetation from EOS-MODIS. Remote Sensing of Environment, Vol.
58, No. 3 (December 1996), pp. 289-298, ISSN 0034-4257
Gong, P., Pu, R., Biging, G.S. & Larrieu, M.R. (2003). Estimation of forest leaf area index
using vegetation indices derived from Hyperion hyperspectral data. IEEE
Transactions on Geoscience and Remote Sensing Vol. 41, No. 6, (June 2003), pp. 1355–
1362, ISSN 0196-2892
Goodenough, D.G., Dyk, A., Niemann, K.O., Pearlman, J.S., Chen, H., Han, T. Murdoch, M.
& West, C. (2003). Processing Hyperion and ALI for Forest Classification. IEEE
Transactions on Geoscience and Remote Sensing, Vol. 41, No. 6, (June 2003), pp. 1321-
1331, ISSN 0196-2892
Govaerts, Y. M., Verstraete, M. M., Pinty, B. & Gobron, N. (1999). Designing optimal spectral
indices: a feasibility and proof of concept study. International Journal of Remote
Sensing, Vol. 20, No. 9, (June 1999), pp. 1853-1873, ISSN 0143-1161
Häme, T. (1991). Spectral interpretation of changes in forest using satellite scanner images.
Acta Forestalia Fennica, Vol. 222, The Society of Forestry in Finland & The Finnish
Forest Research Institute, ISBN 951-651-092-2, Helsinki
Heiskanen, J., Rautiainen, M., Korhonen, L., Mõttus, M., Stenberg, P. (2011). Retrieval of
boreal forest LAI using a forest reflectance model and empirical regressions.
International Journal of Applied Earth Observation and Geoinformation, Vol. 13, No. 4,
(August 2011), pp. 595-606, ISSN 0303-2434.
Horler, D.N.H. & Ahern, F.J. (1986). Forestry information content of Thematic Mapper data.
International Journal of Remote Sensing, Vol. 7, No. 3, (January 1986), pp. 405-428,
ISSN 0143-1161.
Huete, A.R. (1988) A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment,
Volume 25, Issue 3, (August 1988) , Pages 295-309, ISSN 0034-4257
Huete, A., Didan, K., Miura, T., Rodrigues, E.P., Gao, X. & Ferreira, L.G. (2002). Overview of
the radiometric and biophysical performance of the MODIS vegetation indices.
Remote Sensing of Environment, Vol. 83, No. 1-2, (November 2002), pp. 195-213, ISSN
0034-4257.
Jupp, D.L.B., Datt, B., Lovell, J., Campbell, S., King, E., et al. (2002). Discussions around
Hyperion data: background notes for the Hyperion data users workshop. CSIRO Earth
Observation Centre, Canberra. (accessed: 10.10.2011), Available from
ftp://ftp.eoc.csiro.au/pub/djupp/Hyperion/Workshop/Minimal_Set/Document
s/Hyp_Wsn.pdf
Jonckheere, I., Fleck, S., Nackaerts, K., Muys, B., Coppin, P., Weiss, M. & Baret, F. (2004).
Reviews of methods for in situ leaf area index determination. Part I. Theories,
sensors, and hemispherical photography. Agricultural and Forest Meteorology, Vol.
121, No. 1-2, (January 2004), pp. 19-35, ISSN 0168-1923
Kokaly, R. F., Asner, G.P., Ollinger, S.V., Martin, M.E. & Wessman, C.A. (2009).
Characterizing canopy biochemistry from imaging spectroscopy and its application
26 Remote Sensing – Applications

to ecosystem studies. Remote Sensing of Environment, Vol. 113, Supplement 1,

(September 2009), pp. S78-S91, ISSN 0034-4257
Lang, A.R.G. (1986). Leaf area and average leaf angle from transmission of direct sunlight.
Australian Journal of Botany, Vol. 34, No. 3, (May 1986), pp. 349 - 355. ISSN 0067-1924
Lee, K.-S., Cohen, W. B., Kennedy, R. E., Maiersperger, T. K. & Gower, S. T. (2004)
Hyperspectral versus multispectral data for estimating leaf area index in four
different biomes. Remote Sensing of Environment, Vol. 91, No. 3-4, (June 2004), pp.
508-520, ISSN 0034-4257.
LI-COR (1992). LAI-2000 Plant Canopy Analyzer: Instruction Manual. Lincoln, Nebraska, LI-
COR, Inc.
le Maire, G., François, C., Soudani, K., Berveiller, D., Pontailler, D., Bréda, N., Genet,
H.,Davi, H. & Dufrêne, E. (2008). Calibration and validation of hyperspectral
indices for the estimation of broadleaved forest leaf chlorophyll content, leaf mass
per area, leaf area index and leaf canopy biomass. Remote Sensing of Environment,
Vol. 112, No. 10, (October 2008), pp. 3846–3864, ISSN 0034-4257
Matthew, M. W., Adler-Golden, S. M., Berk, A., Richtsmeier, S. C., Levine, R. Y., Bernstein,
L. S., Acharya, P. K., Anderson, G. P., Felde, G. W., Hoke M. P., Ratkowski, A.,
Burke, H.-H., Kaiser, R. D., & Miller, D. P., (April 2000). Status of Atmospheric
Correction Using a MODTRAN4-based Algorithm. In: SPIE Proceedings, Algorithms
for Multispectral, Hyperspectral, and Ultraspectral Imagery VI. Vol. 4049, Accessed
(10.10.2011), Available from: http://www.dtic.mil/cgi-bin/
GetTRDoc?AD=ADA447767&Location=U2&doc=GetTRDoc.pdf
Miller, J.B. (1967). A formula for average foliage density. Australian Journal of Botany, Vol. 15,
No. 1, (February 1967), pp. 141 – 144, ISSN 0067-1924
Monsi, M. & Saeki, T. (1953). Über den Lichtfactor in den Pflanzengesellschaften und seine
bedeutung für die Stoff-production. Japanese Journal of Botany, Vol. 14, (March 1953)
pp. 22-52, ISSN 0075-3424
Monteith, J.L. & Moss, C.J. (1977). Climate and the efficiency of crop production in Britain.
Philosophical Tranactions of the Royal Soiety B. Vol. 281, No. 980, (November 1977),
pp. 277-294. Online ISSN 1471-2970
Morisette, J.T., Baret, F., Privette, J. L., Myneni, R.B., Nickeson, J., Garrigues, S., Shabanov,
N., Weiss, M., Fernandes, R., Leblanc, S., Kalacska, M., Sánchez-Azofeifa, G. A.,
Chubey, M., Rivard, B., Stenberg, P., Rautiainen, M., Voipio, P., Manninen, T.,
Pilant, A., Lewis, T., Iiames, J., Colombo, R., Meroni, M., Busetto, L., Cohen, W.,
Turner, D., Warner, E.D., Petersen, G.W., Seufert, G. & Cook, R. (2006). Validation
of global moderate resolution LAI Products: a framework proposed within the
CEOS Land Product Validation subgroup. IEEE Transaction on Geosciences and
Remote Sensing, Vol. 44, No. 7, (July 2006), pp. 1804-1817, ISSN 0196-2892
Mutanga, O. & Skidmore, A.K. (2004). Narrow band vegetation indices overcome the
saturation problem in biomass estimation. International Journal of Remote Sensing,
Vol. 25, No. 19, (October 2004), pp. 3999–4014, ISSN 0143-1161
Myneni, R. B., Hall, F. G., Sellers, P. J. & Marshak, A. L. (1995) Interpretation of spectral
vegetation indexes. IEEE Transactions on Geoscience and Remote Sensing. Vol. 33, No.
2, (March 1995), pp. 481-486, ISSN 0196-2892
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 27

Myneni, R.B., Nemani, R.R. & Running, S.W. (1997). Estimation of global leaf area index and
absorbed PAR using radiative transfer models. IEEE Transactions on Geoscience and
Remote, Vol. 35, No. 6, (November 1997), pp. 1380-1393, ISSN 0196-2892.
Nemani, R., Pierce, L., Running, S. & Band, L. (1993). Forest ecosystem processes at the
watershed scale: sensitivity to remotely-sensed Leaf Area Index estimates.
International Journal of Remote Sensing, Vol. 14, No. 13, pp. 2519-2534, ISSN 0143-
1161
Nilson, T. (1971). A theoretical analysis of the frequency of gaps in plant stands. Agricultural
Meteorology, Vol. 8, pp. 25-38, ISSN 0002-1571
Nilson, T. & Peterson, U. (1994). Age dependence of forest reflectance: analysis of main
driving factors. Remote Sensing of Environment, Vol. 48, No. 3, (June 1994), pp. 319-
331, ISSN 0034-4257.
Norman, J. & Jarvis, P. (1975). Photosynthesis in Sitka spruce (Picea sitchensis (Bong.) Carr.).
V. Radiation penetration theory and a test case. Journal of Applied Ecology, Vol. 12,
No. 3. (December 1975), pp. 839-877, ISSN 00218901
Pearlman, J. S., Barry, P.S., Segal, C. C., Shepanski, J., Beiso, D. & Carman, S. L. (2003)
Hyperion, a space-based imaging spectrometer. IEEE Transactions on Geoscience and
Remote Sensing, Vol. 41, No. 6, (June 2003), pp. 1160-1173, ISSN 0196-2892
Pisek, J., Rautiainen, M., Heiskanen,J. & Mõttus, M. (2012). Retrieval of seasonal dynamics of
forest understory reflectance in a Northern European boreal forest from MODIS
BRDF data. Remote Sensing of Environment, Vol. 17, pp. 464-468, ISSN 0034-4257
Pu, R., Gong, P., Biging, G.S. & Larrieu, M.R. (2003). Extraction of Red Edge Optical
Parameters From Hyperion Data for Estimation of Forest Leaf Area Index. IEEE
Transaction on Geosciences and Remote Sensing Vol. 41, No. 4, (April 2003), pp. 916-
921, ISSN 0196-2892
Pu, R., Gong, P. & Yu, Q. (2008) Comparative analysis of EO-1 ALI and Hyperion, and
Landsat ETM+ data for mapping forest crown closure and leaf area index. Sensors
2008, Vol. 8, No. 6, (June 2008), pp. 3744-3766, ISSN 1424-8220
Rautiainen, M., Heiskanen, J., Eklundh, L., Mõttus, M., Lukeš, P. & Stenberg, P. (2010),
Ecological applications of physically based remote sensing methods. Scandinavian
Journal of Forest Research Vol. 25, No. 4, (July 2010), pp. 325—339, ISSN 0282-7581
Rautiainen, M. & Stenberg, P. (2005). Application of photon recollision probability in
simulating coniferous canopy reflectance. Remote Sensing of Environment, Vol. 96,
No. 1, (May 2005), pp. 98-107, ISSN 0034-4257
Rautiainen, M.; Mõttus, M.; Heiskanen, J.; Akujärvi, A.; Majasalmi, T. & Stenberg, P. (2011).
Seasonal reflectance dynamics of common understory types in a northern
European boreal forest. Remote Sensing of Environment, in press, (available online
since 27 July 2011), ISSN 0034-4257
Rautiainen, M., Suomalainen, J., Mõttus, M., Stenberg, P., Voipio, P., Peltoniemi, J. &
Manninen, T. (2007). Coupling forest canopy and understory reflectance in the
Arctic latitudes of Finland. Remote Sensing of Environment, Vol. 110, No. 3, (October
2007), pp. 332-343, ISSN 0034-4257
Roberts, D., Ustin, S., Ogunjemiyo, S., Greenberg, J., Dobrowski, S., Chen, J. & Hinckley, T.
(2004). Spectral and structural measures of northwest forest vegetation at leaf to
landscape scales. Ecosystems, Vol. 7, No. 5, (May 2004), pp. 545-562. ISSN 1432-9840
28 Remote Sensing – Applications

Rock, B. N., Vogelmann, J. E., Williams, D. L., Vogelmann, A. F. & Hoshizaki, T. (1986).
Remote Detection of Forest Damage. Bio Science, Vol. 36, No. 7, (July-August 1986),
pp. 439-445, ISSN 0006-3568
Rouse, J. W., Haas, R. H., Schell, J. A. & Deering, D. W. (1974). Monitoring Vegetation
Systems in the Great Plains with ERTS. In: Third ERTS Symposium, pp. 309-317,
NASA SP-351 I, United States
Samiappan, S., Prasad, S., Bruce, L.M. & Robles, W. (2010) NASA's upcoming HyspIRI
Mission - Precision vegetation mapping with limited ground truth. Geoscience and
Remote Sensing Symposium (IGARSS), 2010 IEEE International, (July 2010), pp. 3744-
3747. ISSN 2153-6996
San, B. T., & M. L. Suzen, (2010), Evaluation of Different Atmospheric Correction
Algorithms for EO-1 Hyperion Imagery, International Archives of the Photogrammetry,
Remote Sensing and Spatial Information Science, Volume XXXVIII, Part 8, (August
2010), pp. 392-397, ISSN 1682-1777
Schlerf, M. et al. (2005). Remote sensing of forest biophysical variables using HyMap
imaging spectrometer data. Remote Sensing of Environment, Vol. 95, No. 2, (March
2005), pp. 177¬–194, ISSN 0034-4257
Schull, M., Knyazikhin, Y., Xu, L., Samanta, A., Carmona, P., Lepine, L., Jenkins, J., Ganguly,
S. & Myneni, R. (2011). Canopy spectral invariants, Part 2: Application to
classification of forest types from hyperspectral data. Journal of Quantitative
Spectroscopy and Radiative Transfer, Vol.112, No.4, (March 2011), pp. 736-750, ISSN
0022-4073
Segl, K., Guanter, L., Kaufmann, H., Schubert, J., Kaiser, S., Sang, B. & Hofer, S. (2010)
Simulation of spatial sensor characteristics in the context of the EnMAP
hyperspectral mission. IEEE Transactions on Geoscience and Remote Sensing, Vol.48,
No. 7, (July 2010), pp. 3046-3054, ISSN 0196-2892
Sellers, P.J. (1985). Canopy reflectance, photosynthesis and transpiration. International
Journal of Remote Sensing, Vol. 6, No. 8, pp. 1335–1372, ISSN 0143-1161
Shinozaki, K., Yoda, K, Hozumi, K. & Kira, T. (1964). A quantitative analysis of plant form -
the pipe model theory. I. Basic analyses. Japanese Journal of Ecology, Vol. 14, No. 3,
(June 1964), pp. 97-105, ISSN 0021-5007
Sims, D.A. & Gamon, J.A. (2002). Relationships between leaf pigment content and spectral
reflectance across a wide range of species, leaf structures and developmental
stages. Remote Sensing of Environment, Vol. 81, No. 2–3, (August 2002), pp. 337–354,
ISSN 0034-4257
Smith, N.J., Chen, J.M. & Black, T.A. (1993). Effects of clumping on estimates of stand leaf
area index using the LI-COR LAI-2000. Canadian Journal of Forest Research, Vol. 23,
No. 9, pp. 1940-1943, ISSN 0045-5067
Spanner, M.A., Pierce, L.L., Peterson, D.L. & Running, S.W. (1990). Remote sensing of
temperate coniferous forest leaf area index: the influence of canopy closure,
understory vegetation and background reflectance. International Journal of Remote
Sensing. Vol. 11, No.1, pp. 95-111, ISSN 0143-1161
Stenberg, P. (2006). A note on the G-function for needle leaf canopies. Agricultural and Forest
Meteorology, Vol. 136, No. 1-2, (January 2006), pp 76-79, ISSN 0168-1923
Narrowband Vegetation Indices for Estimating Boreal Forest Leaf Area Index 29

Stenberg, P., Linder, S., Smolander, H. & Flower-Ellis, J. (1994). Performance of the LAI-2000
plant canopy analyzer in estimating leaf area index of some Scots pine stands. Tree
Physiology, Vol. 14, No. 7-8-9, (July 1994), pp. 981-995, Online ISSN 1758-4469
Stenberg, P., Rautiainen, M., Manninen, T., Voipio, P. & Smolander, H. (2004). Reduced
Simple Ratio better than NDVI for estimating LAI in Finnish pine and spruce
stands. Silva Fennica, Vol. 38, No. 1, (March 2004), pp. 3-14, ISSN 0037-5330.
Sun, L., Neville, R., Staenz, K., & White, H.P. (2008). Automatic destriping of Hyperion
imagery based on spectral moment matching. Canadian Journal for Remote Sensing.
Vol. 34, Supplement 1, (May 2008), pp. 68-81, ISSN 0703-8992
Tucker, C. J. (1979) ETM+3 and photographic infraETM+3 linear combinations for
monitoring vegetation. Remote Sensing of Environment, 8, 127-150, ISSN 0034-4257
Ustin, S. L., Gitelson, A.A., Jacquemoud, S., Schaepman, M., Asner, G.P., Gamon, J.A., &
Zarco-Tejada, P. (2009). Retrieval of foliar information about plant pigment systems
from high resolution spectroscopy. Remote Sensing of Environment, Vol. 113,
Supplement 1, (September 2009), pp. S67-S77, ISSN 0034-4257.
Validation of Land European Remote Sensing Instruments (VALERI) Network. Cited:
16.8.2011. Available from: http://www.avignon.inra.fr/valeri/
Verstraete, M. M. & Pinty, B. (1996) Designing optimal spectral indexes for remote sensing
applications. IEEE Transactions on Geoscience and Remote Sensing, Vol. 34, No. 5,
(September 1996), pp. 1254-1265, ISSN 0196-2892
Waring, R.H., Schroeder, P.E. & Oren, R. (1982). Application of the pipe model theory to
predict canopy leaf area. Canadian Journal of Forest Research. Vol. 12, No. 3.
(September 1982), pp. 556-560. ISSN 0045-5067
Wilson, J.W. (1965). Stand structure and light penetration. I. Analysis by point quadrats.
Journal of Applied Ecology, Vol. 2, No. 2, (November 1965), pp. 383-390, ISSN:
00218901
Watson, D. (1947). Comparative physiological studies in the growth of field crops. I.
Variation in net assimilation rate and leaf area between species and varieties, and
within and between years. Annales of Botany, Vol. 11, No. 1, (January 1947), pp. 41-
76, ISSN 0305-7364
Welles, J.M. & Cohen, S. (1996). Canopy structure measurement by gap fraction analysis
using commercial instrumentation. Journal of Experimental Botany, Vol. 47, No. 9,
(September 1996), pp. 1335-1342, ISSN 0022-0957
Widlowski, J. L., Taberner, M., Pinty, B., Bruniquel-Pinel, V., Disney, M., Fernandes, R.,
Gastellu-Etchegorry, J. P., Gobron, N., Kuusk, A., Lavergne, T., Leblanc, S., Lewis,
P. E., Martin, E., Mottus, M., North, P. R. J., Qin, W., Robustelli, M., Rochdi, N.,
Ruiloba, R., Soler, C., Thompson, R., Verhoef, W., Verstraete, M. M. & Xie, D.
(2007). Third Radiation Transfer Model Intercomparison (RAMI) exercise:
Documenting progress in canopy reflectance models. Journal of Geophysical Research-
-Atmospheres 112, D09111, (May 2007), 28 pages, ISSN 0148–0227
Williams, D. (1991). A comparison of spectral reflectance properties at the needle, branch,
and canopy level for selected conifer species. Remote Sensing of Environment Vol. 35,
No. 2-3, (February-March 1991), pp. 79-93, ISSN 0034-4257
30 Remote Sensing – Applications

Wu, C., Han, X., Niu, Z. & Dong, J. (2010). An evaluation of EO-1 hyperspectral Hyperion
data for chlorophyll content and leaf area index estimation. International Journal of
Remote Sensing, Vol. 31, No. 4, (February 2010), pp. 1079-1086, ISSN 0143-1161
Zarco-Tejada, P.J., Rueda, C.A. & Ustin, S.L. (2003). Water content estimation in vegetation
with MODIS reflectance data and model inversion methods. Remote Sensing of
Environment, Vol. 85, No. 1, (April 2003), pp. 109-124, ISSN 0034-4257.
 2

Crop Disease and

Pest Monitoring by Remote Sensing
Wenjiang Huang, Juhua Luo, Jingcheng Zhang,
Jinling Zhao, Chunjiang Zhao, Jihua Wang,
Guijun Yang, Muyi Huang, Linsheng Huang and Shizhou Du
Beijing Research Center for Information Technology in Agriculture, Beijing
China

1. Introduction
Plant diseases and pests can affect a wide range of commercial crops, and result in a
significant yield loss. It is reported that at least 10% of global food production is lost due to
plant diseases (Christou and Twyman, 2004; Strange and Scott, 2005). Excessive pesticides
are used for protecting crops from diseases and pests. This not only increases the cost of
production, but also raises the danger of toxic residue in agricultural products. Disease and
pest control could be more efficient if disease and pest patches within fields can be
identified timely and treated locally. This requires obtaining the information of disease
infected boundaries in the field as early and accurately as possible. The most common and
conventional method is manual field survey. The traditional ground-based survey method
requires high labor cost and produces low efficiency. Thus, it is unfeasible for large area.
Fortunately, remote sensing technology can provide spatial distribution information of
diseases and pests over a large area with relatively low cost. The presence of diseases or
insect feedings on plants or canopy surface causes changes in pigment, chemical
concentrations, cell structure, nutrient, water uptake, and gas exchange. These changes
result in differences in color and temperature of the canopy, and affect canopy reflectance
characteristics, which can be detectable by remote sensing (Raikes and Burpee 1998).
Therefore, remote sensing provides a harmless, rapid, and cost-effective means of
identifying and quantifying crop stress from differences in the spectral characteristics of
canopy surfaces affected by biotic and abiotic stress agents.
This chapter introduces some successful studies about detecting and discriminating yellow
rust and aphid (economically important disease and pest in winter wheat in China) using
field, airborne and satellite remote sensing.

2. Detecting yellow rust of winter wheat by remote sensing

Yellow rust (Biotroph Puccinia striiformis), also known as stripe rust, is a fungal disease of
winter wheat (Triticum aestivum L.). It produces leaf lesions (pustules), which are yellow in
color and tend to be grouped in patches. Yellow rust often occurs in narrow stripes, 2–3 mm
wide that run parallel to the leaf veins. Yellow rust is responsible for approximately 73–85%
32 Remote Sensing – Applications

of recorded yield losses, and grain quality is also significantly reduced (Li et al. 1989).
Consequently, effective monitoring of the incidence and severity of yellow rust in
susceptible regions is of great importance to guide the spray of pesticides and to provide
data for the local agricultural insurance services. Fortunately, remote sensing technology
provides a possible way to detect the incidence and severity of the disease rapidly.
The interaction of electromagnetic radiation with plants varies with the wavelength of the
radiation. The same plant leaves may exhibit significant different reflectance depending on
the level of health and or vigor (Wooley 1971, West et al. 2003, Luo et al., 2010). Healthy and
vigorously growing plant leaves will generally have
1. Low reflectance at visible wavelengths owing to strong absorption by photoactive
pigments (chlorophylls, anthocyanins, carotenoids).
2. High reflectance in the near infrared because of multiple scattering at the air-cell
interfaces in the leaf’s internal tissue.
3. Low reflectance in wide wavebands in the short-wave infrared because of absorption by
water, proteins, and other carbon constituents.
The incidence and severity of yellow rust can be monitored according to the differences of
spectral characteristics between healthy and disease plants. In this chapter, we will report
several successful studies on the detection and identification of yellow rust in winter wheat
by remote sensing.

2.1 Detecting and discriminating yellow rust at canopy level

Hyperspectral remote sensing is one of the advanced and effective techniques in disease
monitoring and mapping. However, the difficulty in discriminating a disease from common
nutrient stresses largely hampers the practical use of this technique. This is because some
common nutrient stresses such as the shortage or overuse of nitrogen or water could have
similar variations of biochemical properties and plant morphology, and therefore result in
similar spectral responses. However, for the remedial procedures for stressed crops, there is
a significant difference between disease and nutrient stresses. For example, applying
fungicide to water-stressed crops would lead to a disastrous outcome. Therefore, to
discriminate yellow rust from common nutrient stresses is of practical importance to crop
growers or landowners.
The specific objectives of this study are to: (1) systematically test the sensitivity and
consistency of several commonly used spectral features to yellow rust disease during major
growth stages; (2) for those spectral features that are consistently sensitive to yellow rust
disease, we will further examine their sensitivity to nutrient stresses to determine whether
there are specifically sensitive to yellow rust disease, but insensitive to water and nitrogen
stresses.

2.1.1 Materials and methods

2.1.1.1 Experimental design and field conditions
The experiments were conducted at Beijing Xiaotangshan Precision Agriculture
Experimental Base, in Changping district, Beijing (40º10.6’N, 116º26.3’E) for the growing
seasons of 2001-2002 and 2002-2003. Table 1 summarizes the soil properties including
Crop Disease and Pest Monitoring by Remote Sensing 33

organic matter, total nitrogen, alkali-hydrolysis nitrogen, available phosphorus and

available potassium for both growing seasons. Three cultivars of winter wheat used in 2001-
2002 experiment (2002 Exp) were Jingdong8, Jing9428 and Zhongyou9507, while the
cultivars used in 2002-2003 (2003 Exp) were Xuezao, 98-100 and Jing411. All the cultivars
applied in both growing seasons included erective, middle and loose with respect to the
canopy morphology.

Disease inoculation
Items Nutrient stress experiment
experiment
Growth period Sep 2002-Jun 2003 Sep 2001-Jun 2002
Organic matter 1.42%-1.48% 1.21%-1.32%
Total nitrogen 0.08%-0.10% 0.092%-0.124%
Alkali-
hydrolysis 58.6-68.0 mg kg-1 68.8-74.0 mg kg-1
Top soil
nitrogen
nutrient status
(0-0.3m depth) Available 20.1-55.4 mg kg-1 25.2-48.3 mg kg-1
phosphorus
Rapidly
available 117.6-129.1 mg kg-1 96.6-128.8 mg kg-1
potassium
Jingdong8, Jing9428,
Cultivars Xuezao, 98-100, Jing411
Zhongyou9507
Normal: 200 kg ha-1 nitrogen,
450 m3 ha-1 water;
W-SD: 200 kg ha-1 nitrogen,
Normal; YR1: 3mg 100-1 225 m3 ha-1 water;
ml spores solution; YR2: W-SED: 200 kg ha-1 nitrogen, 0
9mg 100-1 ml spores m3 ha-1 water;
solution; YR3: 12mg 100 N-E: 350 kg ha-1 nitrogen, 450
-1
Treatments
ml spores solution (all m3 ha-1 water;
treatments applied 200 kg N-D: 0 kg ha-1 nitrogen, 450 m3
ha-1 nitrogen and 450 m3 ha-1 water;
ha-1 water) W-SED+N-E: 350 kg ha-1
nitrogen, 0 m3 ha-1 water; W-
SED+N-D: 0 kg ha-1 nitrogen, 0
m3 ha-1 water;
Spectral reflectance
measurements (on day after 207, 216, 225, 230, 233 196, 214, 225, 232, 239
sowing)
Table 1. Basic information of disease inoculation experiment and nutrient stress experiment

For 2002 Exp, six stress treatments of water and nitrogen were applied, and the treatments
were based on local conditions, which usually suffered from yellow rust in the northern part
34 Remote Sensing – Applications

of China. Each treatment was applied on 0.3 ha area, and the treatments were 200 kg ha-1
nitrogen and 225 m3 ha-1 water (slightly deficient water, W-SD),200 kg ha-1 nitrogen and no
irrigation (seriously deficient water, W-SED), 350 kg ha-1 nitrogen and 450 m3 ha-1 water
(excessive nitrogen, N-E), no fertilization and 450 m3 ha-1 water (deficient nitrogen, N-D),
350 kg ha-1 nitrogen and no irrigation (seriously deficient water and excessive nitrogen, W-
SED+N-E), and no fertilization and no irrigation (seriously deficient water and deficient
nitrogen, W-SED+N-D). A 0.3 ha reference area (Normal) was applied with the
recommended rate which received 200 kg ha-1 nitrogen and 450 m3 ha-1 water. Three
cultivars were evenly distributed in each treatment plot.
For 2003 Exp, according to the National Plant Protection Standard (Li et al. 1989), three
levels of concentration of summer spores of yellow rust were applied, and they were 3 mg
100-1 ml-1 (Yellow rust 1, YR1), 9 mg 100-1 ml-1 (Yellow rust 2, YR2) and 12 mg 100-1 ml-1
(Yellow rust 3, YR3), with a dosage of 5 ml spores solution per square meter. The reference
area (Normal) that was not inoculated yet was applied with the recommended amount of
fungicide to prevent the occasional infection. Each treatment involved 1.2 ha area, with even
constitution of three cultivars. All plots in 2003 Exp received the recommended rates of
nitrogen (200 kg ha-1) and water (450 m3 ha-1).
2.1.1.2 Canopy spectral measurements
A high spectral resolution spectrometer, ASD FieldSpec Pro spectrometer (Analytical
Spectral Devices, Boulder, CO, USA) fitted with a 25 field of view fore-optic, was used for
in-situ measurement of canopy spectral reflectance for both 2002 Exp and 2003 Exp. All
canopy spectral measurements were taken from a height of 1.3m above ground (the height
of the wheat is 90±3 cm at maturity). Spectra were acquired in the 350-2,500 nm spectral
range at a spectral resolution of 3 nm between 350 nm and 1,050 nm, and 10 nm between
1,050 nm and 2,500 nm. A 40 cm × 40 cm BaSO4 calibration panel was used for calculation of
reflectance. All irradiance measurements were recorded as an average of 20 scans at an
optimized integration time. Prior to subsequent preprocessing, all spectral curves were
resampled with 1 nm interval. All measurements were made under clear blue sky conditions
between 10:00 and 14:00 (Beijing Local Time).
The spectral measurements were taken 5 times from 196 days after sowing (DAS) to 239
DAS for 2002 Exp, which covered the growth stages of stem elongation, booting, anthesis
and milk development. For 2003 Exp, the spectral measurements were taken 5 times from
207 DAS to 233 DAS, which covered the growth stages of booting, anthesis and milk
development. The detailed measurement dates for both experiments were given in Table 1.
The stem elongation and anthesis stages are essential for the control of yellow rust
development, whereas the milk development stage is important for yield loss assessment.
2.1.1.3 Selection of spectral features
The spectral features that we adopted were related to several commonly used vegetation
indices (VIs), which were proved to be sensitive to variations of pigments and stresses.
Furthermore, in order to conduct a thorough investigation of various types of spectral
features, we also included a number of spectral features that were based on derivative
transformation and continuum removal transformation (Gong et al. 2002; Pu et al.
2003;2004). Therefore, the total 38 spectral features are shown in Table 2.
Crop Disease and Pest Monitoring by Remote Sensing 35

Variable Definition Description Literatures

Derivative transformed spectral variables
Blue edge covers 490-530nm. Db is a
Maximum value
maximum value of 1st order
Db of 1st derivative Gong et al., 2002
derivatives within the blue edge of 35
within blue edge
bands
λb Wavelength at Db λb is wavelength position at Db Gong et al., 2002
Sum of 1st Defined by sum of 1st order derivative
SDb derivative values values of 35 bands within the blue Gong et al., 2002
within blue edge edge
Maximum value Yellow edge covers 550-582nm. Dy is a
of 1st derivative maximum value of 1st order
Dy Gong et al., 2002
within yellow derivatives within the yellow edge of
edge 28 bands
λy Wavelength at Dy λy is wavelength position at Dy Gong et al., 2002
Sum of 1st
Defined by sum of 1st order derivative
derivative values
SDy values of 28 bands within the yellow Gong et al., 2002
within yellow
edge
edge
Red edge covers 670-737nm. Dr is a
Maximum value
maximum value of 1st order
Dr of 1st derivative Gong et al., 2002
derivatives within the red edge of 61
within red edge
bands
λr Wavelength at Dr λr is wavelength position at Dr Gong et al., 2002
Sum of 1st
Defined by sum of 1st order derivative
SDr derivative values Gong et al., 2002
values of 61 bands within the red edge
within red edge
Continuous removal transformed spectral features
DEP550-750 The depth of the In the range of 550nm-750nm
DEP920-1120 feature minimum In the range of 920nm-1120nm
Pu et al., 2003;2004
DEP1070- relative to the
hull In the range of 1070nm-1320nm
1320
WID550-750 The full In the range of 550nm-750nm
WID920-1120 wavelength In the range of 920nm-1120nm
Pu et al., 2003;2004
WID1070- width at half
DEP (nm) In the range of 1070nm-1320nm
1320
AREA550-
The area of the In the range of 550nm-750nm
750
absorption
AREA920-
feature that is the In the range of 920nm-1120nm Pu et al., 2003;2004
1120
product of DEP
AREA1070- and WID In the range of 1070nm-1320nm
1320
36 Remote Sensing – Applications

Variable Definition Description Literatures

VI-based variables
Zarco-Tejada et al.,
GI Greenness Index R554/R677
2005
Chen, 1996;
Modified Simple
MSR (R800/R670-1)/(R800/R670+1)1/2 Haboudane et al.,
Ratio
2004
(RNIR-RR)/(RNIR+RR), where RNIR
Normalized
indicates 775-825nm, RR indicates
NDVI Difference Rouse et al., 1973
650nm-700nm, that include most key
Vegetation Index
pigments
Narrow-band
normalised Thenkabail et al.,
NBNDVI (R850-R680)/(R850+R680)
difference 2000
vegetation index
Nitrogen
NRI (R570-R670)/(R570+R670) Filella et al., 1995
reflectance index
Photochemical
PRI Physiological (R531-R570)/(R531+R570) Gamon et al., 1992
Reflectance Index
The transformed
chlorophyll 3*[( R700- R670)-0.2*( R700- R550)*( R700/ Haboudane et al.,
TCARI
Absorption and R670)] 2002
Reflectance Index
Structural
Peñuelas et al.,
SIPI Independent (R800-R445)/(R800-R680)
1995
Pigment Index
Plant Senescence Merzlyak et al.,
PSRI (R680-R500)/R750
Reflectance Index 1999
The Physiological
PhRI (R550-R531)/(R550+R531) Gamon et al., 1992
reflectance index
Normalized
Pigment Peñuelas et al.,
NPCI (R -R )/(R680+R430)
Chlorophyll ratio 680 430 1994
Index
Anthocyanin Gitelson et al.,
ARI ARI=(R550)-1-(R700)-1
Reflectance Index 2001
Broge and Leblanc,
Triangular
TVI 0.5[120(R750-R550)-200(R670-R550)] 2000; Haboudane
Vegetation Index
et al., 2004
Chlorophyll
(|(a670+R670+b)|/(a2+1)1/2)x(R700/R670)
CARI Absorption Ratio Kim et al., 1994
a = (R700-R550)/150, b = R550-(a x 550)
Index
Crop Disease and Pest Monitoring by Remote Sensing 37

Variable Definition Description Literatures

Disease Water
DSWI (R802+R547)/(R1657+R682) Galvão et al., 2005
Stress Index
Hunt and rock,
Moisture Stress
MSI R1600/R819 1989; Ceccato et
Index
al., 2001
Shortwave
Fensholt and
SIWSI Infrared Water (R860-R1640)/(R860+R1640)
Sandholt, 2003
Stress Index
Red-Edge
Merton and
RVSI Vegetation Stress [(R712+R752)/2]-R732
Huntington, 1999
Index
Modified
Chlorophyll Daughry et al.,
MCARI (R701-R671)-0.2(R701-R549)]/(R701/R671)
Absorption in 2000
Reflectance Index
Peñuelas et al.,
WI Water Index R900/R970
1997

Table 2. Definitions of spectral features used in this study

2.1.1.4 Preprocessing and normalization of spectral reflectance data

Aggregating spectral reflectance data
As the first step, all spectra were processed with the following transformation to suppress
possible difference in illumination. The spectral regions with wavelength of 1330-1450 nm,
1770-2000 nm and 2400-2500 nm were removed due to strong absorption by water vapor.
We then normalized the spectral curves by dividing the mean band reflectance of the curve
(Yu et al., 1999). The normalized reflectance for the bandi is given as:

Ref i
Ref i 
1 n
(  Ref i )
n i 1

where Refi’ is the normalized reflectance for bandi; Refi is the original reflectance of the band; n
is the total number of bands. Fig. 1(a) shows a plot of unnormalized Refi versus band
wavelength for six observations (three YR3 curves and three Normal curves) on 233 DAS. Fig.
1(b) shows the corresponding curves in Fig.1(a) after normalization. The normalization clearly
separated the diseased spectra from the normal spectra especially over the near infrared region
(approximately from 770 nm to 1300 nm). The benefit of eliminating spectral difference caused
by the change of illumination conditions was also mentioned by Yu et al. (1999).
Normalization of the difference in measuring dates
As shown in Table 1, although both experiments conducted in five growth stages in 2002
and 2003, most measurement dates were not consistent, except for 255 DAS. Hence, to
improve the comparability of two datasets, we adapted the 2002 Exp data to match the dates
38 Remote Sensing – Applications

(a) Original spectra on 233 days after sowing

(b) Normalized spectra on 233 days after sowing

Fig. 1. Comparison between original spectra and normalized ones
Crop Disease and Pest Monitoring by Remote Sensing 39

of 2003 Exp, by using a linear interpolation method. The reflectance curve of a certain date
could be obtained based on the spectra from the adjacent data before and after the
measurement date (using days after sowing as a time scale). Each band of the spectra should
be processed as:

DAScurrent  DASbefore
Ref current  Ref before  (Ref before  Ref after )
DASafter  DASbefore

where Refcurrent represents the reflectance transformed from the date corresponding to an
ideal date in 2003 Exp; Refbefore and Refafter represent reflectances, respectively, from DASbefore
and DASafter; DAScurrent indicates an ideal date in 2003 Exp while DASbefore and DASafter are the
adjacent dates in 2002 Exp before and after the ideal date in 2003 Exp.
Fig. 2 provides an example of the progress of the normalization of measurement dates. The
averaged reflectance at central wavelengths of green band (560 nm) and near-infrared band
(860 nm) of Landsat-5 TM for normal samples were plotted against the measured dates in
both 2002 Exp and 2003 Exp. The date normalized reflectance values were marked as
triangle symbol in the graph. Through this step, the datasets collected in these two years
could be considered as acquired in the same dates, which thereby facilitated the subsequent
comparisons and analysis.

Adaptation of average reflectance of normal samples at 560 nm (central wavelengths of green band of
Landsat-5 TM) and 860 nm (central wavelengths of near-infrared band of Landsat-5 TM) to match the
dates of 2003 Exp, by using a linear interpolation method

Fig. 2. An example for normalization of measuring dates

40 Remote Sensing – Applications

Normalization of the difference from cultivars and soil backgrounds

The canopy spectra of winter wheat were not only supposed to respond to stresses, but
are also determined and influenced by several other aspects such as cultivars and soil
properties. Although the both 2002 Exp and 2003 Exp were conducted in the same fields
that had approximately identical climate and environmental conditions, the difference in
cultivars and soil properties between 2002 Exp and 2003 Exp should not be ignored (Table
1). To minimize this discrepancy, we calculated a ratio spectral curve for each of
measured dates (after the normalization of the measuring dates) by the averaged spectral
curve from normal samples in 2002 Exp divided by the averaged spectral curve from
normal samples in 2003 Exp, resulting in a total of five ratio curves corresponding to each
growth stage (Fig. 3). After that, all the spectral data measured at different growth stages
were multiplied by the corresponding ratio curves to yield a set of normalized spectra. It
should be pointed out that the present normalization processing to raw spectral
measurements will only enhance the comparability between the 2002 Exp and 2003 Exp
with little change in internal relations among different treatments because all the spectral
data at one growth stage were processed with the same ratio curve. The ultimate goal of
all these preprocessing and normalization steps above is to mitigate effects of the
variation of illumination conditions, measurement dates, cultivars and soil properties
between the 2002 Exp and 2003 Exp on target spectra.
2.1.1.5 Spectral features calculation and statistical analysis
With the spectra normalized using the methods above, we calculated 38 spectral features.
An analysis of variance (ANOVA) was employed to investigate the spectral differences
between the normal samples and all forms of stressed samples. Firstly, on different
measured dates, both the yellow rust disease data and nutrient stressed data were compared
with the normal data by ANOVA. For those spectral features that were consistently
sensitive to yellow rust disease, we not only tested their differences between the normal
treatment and different forms of stresses, but also tested the differences between various
kinds of nutrient stresses and varying levels of disease stresses with ANOVA. Statistical
analyses were conducted using SPSS 13.0 procedure.

2.1.2 Results
2.1.2.1 Spectra after normalizations
The spectral ratio curves in Fig 3 reflect the deviations between 2002 Exp and 2003 Exp’s
reflectance datasets at different wavelength positions. The ratio value close to 1.0 indicates
no difference in reflectance exists between the two years. Generally, the ratio values ranged
from 0.7 to 1.3, with an uneven distribution along the wavelength axis (Fig 3). The ratio
tended to deviate from 1.0 in the regions of 350 - 730 nm, 1450 - 1570 and 2000 - 2400 nm, but
stayed around 1.0 in the regions of 730 - 1330 nm and 1570 - 1770 nm. To assess the
improvement in comparability, we examined the difference of normalized datasets of
normal samples between 2002 Exp and 2003 Exp through an ANOVA with all 38 spectral
features. The result showed that the differences of all spectral features were insignificant at
all growth stages (p-value>0.05), with an average p-value (for all measuring dates) of 0.94,
indicating a relatively high level of similarity between two datasets. Therefore, we
confirmed that such normalization processes minimized the spectral difference originated
Crop Disease and Pest Monitoring by Remote Sensing 41

Fig. 3. Ratios of spectra for normalization with different years and varieties

from variation of illumination and different measurement dates, etc., and enabled more
rational comparisons among different treatments.
2.1.2.2 Spectral responses to different forms of stresses
The result of ANOVA between normal samples and different forms of stress samples
indicated that all spectral features had a response (defined as p-value<0.05) to at least one
type of stresses at one growth stage, except for the WID1070-1320, which had no response to
any form of stresses at all growth stages. Total 37 spectral features responded to water
associated stresses (W-SD, W-SED, W-SED+N-E, W-SED+N-D) at least at one growth stage,
followed by 35 spectral features to yellow rust disease, whereas only15 spectral features had
a response to solely nitrogen stress (N-E, N-D). As summarized in Table 3, most spectral
features were sensitive to yellow rust infection at least at one growth stage, except for λb, λr
and WID1070-1320. In addition, most spectral features tended to be more sensitive at later
growth stages than at the early stages. For example, several features such as DEP920-1120,
AREA920-1120, Dy, GI, NDVI and Triangular Vegetation Index (TVI) only had a response to
yellow rust at the last growth stage in our study (233 DAS). However, for the sake of
diagnosis, the spectral features with a consistent response to yellow rust during the
important growing period would be much more valuable. Therefore, those spectral features
that were sensitive to the yellow rust at 4 out of 5 growth stages were selected as candidates
for disease diagnosis. This yielded four vegetation indices (VIs): PRI, PhRI, NPCI and ARI.
2.1.2.3 One way ANOVA of four disease sensitive spectral features
Particularly for the four identified VIs that closely associated with yellow rust disease, a
throughout one way ANOVA was conducted to compare their differences between the
42 Remote Sensing – Applications

Days after sowing

Spectral features
207 216 225 230 233
DEP550-770 √ √ √
AREA550-770 √ √ √
WID550-770 √ √ √
DEP920-1120 √
AREA920-1120 √
WID920-1120 √
DEP1070-1320 √
AREA1070-1320 √
Db √ √
SDb √ √ √
Dy √
λy √
SDy √
Dr √
SDr √ √
GI √
MSR √ √
NDVI √
NBNDVI √ √
NRI √
PRI √ √ √ √
TCARI √ √
SIPI √
PSRI √ √ √
PhRI √ √ √ √
NPCI √ √ √ √
ARI √ √ √ √
TVI √
CARI √ √ √
DSWI √
MSI √
SIWSI √
RVSI √ √
MCARI √ √ √
WI √

Table 3. Responses of spectral features to yellow rust

Crop Disease and Pest Monitoring by Remote Sensing 43

normal sample and various kinds of stressed samples. Moreover, their differences among
each pairs of stress forms were also compared. We conducted this ANOVA based on the
data on 207 DAS, 225 DAS and 233 DAS respectively, which were essential growth stages
for carrying out fungicide spraying and yield loss assessing procedures. In addition to the p-
value of ANOVA, we also provided the change direction of spectral features. Positive sign
indicates the average spectral feature value of diseased or nutrient stressed samples is
greater than that of normal samples, and negative sign indicates the opposite cases to the
positive sign. As shown in Table 4, it was observed that for the treatments of N-E and N-D,
all four VIs failed to show any response at all growth stages. For the results of other
treatments, the responses of four VIs behaved in a varied pattern at three growth stages.
For the results on 207 DAS (Table 4a), compared to the normal samples, the NPCI and ARI
had responses to all three levels of yellow rust treatments (YR 1, YR 2, YR 3), and appeared
to be more sensitive than PRI and PhRI. For nutrient stresses, the PRI, NPCI and ARI were
sensitive to W-SED and W-SED+N-E treatments. Among them, NPCI and ARI showed
stronger responses (p-value<0.01) to W-SD, W-SED, W-SED+N-E and W-SED+N-D
treatments than the other two VIs. For the comparisons between diseased samples and
nutrient stressed samples, significant differences between W-SED and W-SED+N-E
treatments and YR2 and YR3 treatments were identified for PRI, NPCI and ARI. Moreover,
the change directions of the three VIs for diseased and nutrient stressed samples were
identical. At this 207 DAS growth stage, PhRI did not show a significant response to any of
three levels of disease treatments, but responded to W-SD, W-SED and W-SED+N-E
treatments. It is interesting that the change direction of diseased samples of PhRI was
contrary to that of the nutrient stressed samples, suggesting a discriminating potential of the
index.
For the results on 225 DAS (Table 4b), compared to the normal samples, all four VIs revealed
a clear response to level 2 and level 3 of yellow rust treatments (YR2, YR3). For nutrient
stresses, PRI, NPCI and ARI also appeared to be sensitive to W-SD, W-SED, W-SED+N-E
and W-SED+N-D treatments. However, PhRI was insensitive to all nutrient stresses. In
addition, when we looked at the difference of those VIs between diseased samples and
nutrient stressed samples, only PhRI showed clear differences between YR2 and YR3
treatments and W-SD, W-SED, W-SED+N-E, and W-SED+N-D treatments. Although a
significant difference between YR3 treatment and W-SED treatment also existed for ARI and
NPCI, the change directions of both treatments were identical. However, for PhRI, the
change directions of all levels of disease treatments were different from those of the nutrient
stress treatments.
For the results on 233 DAS (Table 4c), with further development of disease symptoms,
compared to the normal samples, all four indices showed responses to all three levels of
disease treatments. Comparing to YR1 treatment, the four VIs had shown a stronger
significant level (p-value<0.01) for YR2, YR3 treatments. For nutrient stresses, PRI, NPCI and
ARI exhibited clear responses to W-SED, W-SED+N-E and W-SED+N-D treatments as well.
For comparisons between diseased and nutrient stressed samples, PRI and NPCI appeared
to be significantly different between YR2 and YR3 treatments and W-SD treatment.
However, the change directions of both treatments were identical. Unlike the other three VIs,
PhRI remained insensitive to the nutrient stresses, but was significantly different among all
levels of disease treatments (YR1, YR2, and YR3) and all forms of nutrient stresses. More
44 Remote Sensing – Applications

(a) 207 DAS

(b) 225 DAS

(c) 233 DAS

*mean difference is significant at 0.950 confidence level; **mean difference is significant at 0.990
confidence level;*** mean difference is significant at 0.999 confidence level. (+) means the average
spectral feature value of diseased or nutrient stressed samples greater than that of normal samples; or
means the average spectral feature value of nutrient stressed samples greater than that of diseased
samples; (-) means the opposite cases to the case of (+). The definitions of treatments are as follows:
“Normal” represents normal samples; “W-SD” represents samples treated with slightly deficient water;
“W-SED” represents samples treated with seriously deficient water; “N-E” represents samples treated
with excessive nitrogen; “N-D” represents samples treated with deficient nitrogen; “W-SED+N-E”
represents samples treated with seriously deficient water and excessive nitrogen; “W-SED+N-D”
represents samples treated with seriously deficient water and deficient nitrogen
Table 4. ANOVA for four VIs separately on 207 DAS, 225 DAS and 233 DAS

importantly for the PhRI, the change directions of diseased samples were opposite to those
of nutrient stressed samples throughout the entire analysis.
In summary, all four VIs showed a significant sensitivity to yellow rust disease on 207 DAS,
225 DAS and 233 DAS. However, most of them also appeared to be sensitive to water
associated stresses to a varing extent, except for PhRI, which was only sensitive to disease
yet insensitive to any forms of nutrient stresses on 225 DAS and 233 DAS. More importantly,
the change directions of PhRI to disease treatments were always opposite to those to the
nutrient stress treatments at all relevant growth stages. This further confirmed the
discriminating characteristic of PhRI.
Crop Disease and Pest Monitoring by Remote Sensing 45

2.1.3 Conclusion
Combining with a dataset of yellow rust disease inoculation and a dataset of various forms
of nutrient stress treatments, we examined the responses of 38 commonly used spectral
features at five important growth stages from booting stage to milk development stage
using a one-way analysis of variance (ANOVA). There were 37 spectral features sensitive to
water associated stresses, 35 spectral features sensitive to yellow rust disease and only 15
spectral features sensitive to sole nitrogen stresses in at least one growth stage. It was
observed that more spectral features appeared to have a response to yellow rust disease at
later growth stages. A throughout ANOVA was conducted particularly on PRI, PhRI, NPCI
and ARI, which showed a consistent response to yellow rust disease at 4 out of 5 growth
stages. However, PRI, NPCI and ARI were also responsible for water associated stresses,
suggesting a risk of confusion in detecting yellow rust disease. Only PhRI was sensitive to
yellow rust disease, but insensitive to different forms of nutrient stresses. The discriminative
response of PhRI could provide a means of identifying and detecting yellow rust disease
under complicated farmland circumstances. This finding can serve the basis of remote
sensing system for detecting yellow rust disease.

2.2 Detecting yellow rust using field and airborne hyperspectral data
The aim of this study was to evaluate the accuracy of the spectro-optical, photochemical
reflectance index (PRI) for quantifying the disease index (DI) of yellow rust in wheat using
in-situ spectral reflectance measurements, and its applicability in the detection of the disease
using hyperspectral imagery.

2.2.1 Materials and methods

2.2.1.1 Experimental design and field conditions
Experimental design and field conditions was same as 1.1.1. Experimental data from 2002
Exp were used to establish the statistical models, and the data for 2003 Exp were used to
validate the models developed.
2.2.1.2 Inspection of disease severity
To quantify the severity of the disease of yellow rust, the leaves of plants were grouped into
one of 9 classifications of disease incidence (x): 0,1, 10, 20, 30, 45, 60, 80 and 100% covered by
rust. 0% represented no incidence of yellow rust, and 100% was the greatest incidence. The
disease index (DI) was then calculated using (Li et al. 1989):

DI (%) 
  x  f   100
n f

where f is the total number of leaves of each degree of disease severity and n is the degree of
disease severity observed (in this work, n ranged from 0 to 8). In each plot, 20 individuals
were randomly selected for check.
2.2.1.3 Canopy spectral measurements
The method of canopy spectral measurements and data was same as the part 1.1.1.2 above.
46 Remote Sensing – Applications

2.2.1.4 Airborne hyperspectral imaging

Airborne hyperspectral images of the trial field were acquired in 2003 using the Pushbroom
Hyperspectral Imager (PHI) designed by the Chinese Academy of Science (CAS) and flown
onboard a Yun-5 aircraft (Shijiazhuang Aircraft Manufacturing Company, China). The PHI
comprises a solid state, area array, and silicon CCD device of 780 × 244 elements. It has a field
of view of 21o, and is capable of acquiring images of 1 m × 1 m spatial resolution at an altitude
of 1000 m above ground. The wavelength range is 400–850 nm with a spectral resolution of 5
nm. Images of the target field were acquired in 2003 at the phenological growth stages of stem
elongation (April 18, 2003, Zadoks stage 3), anthesis (May 17, 2003, Zadoks stage 5) and milky
maturity (May 31, 2003, Zadoks stage 8). The inoculated wheat was adequately infected by
rust on April 18, obviously infected by May 17, and seriously infected by May 31.
Measurements of DI were made and in situ canopy reflectance spectra were also acquired on
the same dates. All images were geometrically and radiometrically corrected using an array of
georeferenced light and dark targets (5 m x 5 m) located at the extremes of the field site. The
aforementioned field spectrometer was used to calibrate these targets relative to BaSO4. The
location of each target, as well as field measurements of DI were recorded using a differential
global positioning system (Trimble Sunnyvale California, USA).
2.2.1.5 Photochemical reflectance index (PRI)
Because yellow rust epiphyte reduced foliar physiological activity by destroying foliar
pigments, the photochemical reflectance index (PRI) was selected as the spectrophotometric
method of estimating the disease index. PRI was calculated by the formula in Table 2.

2.2.2 Results
2.2.2.1 PRI versus DI
Fig. 4 shows a plot of the measured DI as a function of PRI for all varieties. The data points
associated with the variety Xuezao dominate in the top-left region of the scatter plot
(relatively high range of DI), while those associated with the variety 98-100 are located in the
mid region (mid-range DI) and those associated with Jing 411 dominate the lower right
region. This distribution trend is consistent with the relative susceptibility of these varieties
to rust; Xuezao is the least resistant and Jing 411 has the greatest resistance. The regression
equation of DI using PRI in 2002 Exp was obtained as following (n = 64):

DI %  721.22  PRI   2.40  0.14  PRI  0.02 ; r 2

 0.91 
An important feature in, the associated regression equation (Fig. 4) was that the spectrally-
derived PRI explained 91% of the variance observed in the disease index. This explanation
also encompassed the three varieties of wheat as well as the four stages of crop development
for each variety. In the subsequent validation of the PRI-DI regression equation with the
2003 Exp data (Fig. 5), the coefficient of determination (R2) between the estimated and
measured values was 0.97 (n = 80).
In Fig. 5, the locations of data points associated with individual varieties wew consistent
with the levels of resistance to rust. Xuezao dominated the top right-hand region of the
scatter plot (relatively high range of DI), the variety 98-100 had points scattered all along the
Crop Disease and Pest Monitoring by Remote Sensing 47

100
90
2
80 R = 0.91
70
60
DI (%)

50
40
30
20
10
0
-0.14 -0.12 -0.1 -0.08 -0.06 -0.04 -0.02 0 0.02
PRI
Fig. 4. Plot of measured disease index (DI) as a function of measured photochemical
reflectance index(PRI) for all varieties combined in 2002 Exp. Δ: Jing 411; +: Xuezao; □: 98–100

100
2
90 R = 0.97

80
70
PRI-derived DI (%)

60
50
40
30
20
10
0
0 20 40 60 80 100
measured DI (%)
Fig. 5. Comparison of measured DI and PRI-estimated DI for 2003 Exp; ‘Δ’ = Jing 411;‘+’ =
Xuezao; ‘□’ = 98–100

regression line (predominantly mid-range DI), and Jing 411 was concentrated in the central
lower-left region (lower range DI).
2.2.2.2 Application of multi-temporal PHI images for DI estimation
The DI was estimated on a pixel-by-pixel basis in each of the acquired PHI images using the
regression equation. To map the degree of yellow rust infection in the trial field, the DI was
48 Remote Sensing – Applications

binned into the following classes; very Serious (DI > 80%), serious (45% < DI ≤ 80%),
moderate (10% < DI ≤ 45%), slight (1% < DI ≤ 10%) and none (0 < DI ≤ 1%) (Fig. 6).

Fig. 6. Classified DI images derived from PHI airborne images of the trial site in 2003 Exp

Fig. 7 shows the relationship between the DI calculated from the multi-temporal PHI images
and the actual measured DI from the 120 sample sites located within the field (R2=0.91).

100
2
90 R = 0.91
80
PHI-derived DI (%)

70
60
50
40
30
20
10
0
0 20 40 60 80 100
measured DI (%)
Fig. 7. Comparison of PHI-derived estimates of DI and actual DI values for 2002 Exp. Data
were extracted from all three imaging times, although the DI values were< 20% for the April
18 image
Crop Disease and Pest Monitoring by Remote Sensing 49

2.2.3 Conclusion
The results of this work confirm PRI is a potential candidate for monitoring of yellow rust,
and could form the basis of an on-the-go sensor and variable-rate spray applicator or remote
detection and mapping process.

2.3 Detecting yellow rust in winter wheat by spectral knowledge base

In most cases, statistical models for monitoring the disease severity of yellow rust are based
on hyperspectral information. The high cost and limited cover of airborne hyperspectral
data make it impossible to apply such data for large scale monitoring. Furthermore, the
established models of disease detection cannot be used for most satellite images because of
the wide range of wavelengths in multispectral images (Zhang et al., 2011).
To resolve this dilemma, the study presents a novel approach by constructing a spectral
knowledge base (SKB) of winter wheat diseases, which takes the airborne images as a
medium and links the disease severity with band reflectance from moderate resolution
remotely sensed data, such as environment and disaster reduction small satellite images
(HJ-CCD) accordingly. To achieve this goal, several algorithms and techniques for data
conversion and matching are adopted in the proposed system, including minimum noise
fraction (MNF) transformation and pixel purity index (PPI) function. The performance of
SKB is evaluated with both simulated data and field measured data.

2.3.1 materials and methods

Experimental design and field conditions was same as the part of 1.1.1.1
2.3.1.1 Inspection of disease severity
Please refer to the part of 1.2.1.2 above.
2.3.1.2 Airborne hyperspectral imaging
Please refer to the part of 1.2.1.4 above about airborne hyperspectral imaging and image
processing.
2.3.1.3 Acquisition of moderate resolution satellite images
In this study, the SKB is designed to fit the charge coupled device (CCD) sensor, which is on
the environment and disaster reduction small satellites (HJ-1A/B). The basic parameters of
the CCD sensor (using ‘HJ-CCD’ in the following) are given in Table.5. The four bands of

Properties of HJ-CCD
Wavelength range
Band Spatial resolution (m) Swath (km) Revisit time (day)
(nm)
Blue 0.430–0.520
Green 0.520–0.600
30 360 2
Red 0.630–0.690
Near-infrared 0.760–0.900
Table 5. Properties of the environment and disaster reduction small satellites (HJ-CCD)
50 Remote Sensing – Applications

HJ-CCD covered the visible and near infrared spectral regions. The HJ-CCD sensor has
spectral and spatial characteristics that are similar to those of Landsat-5 TM, but the HJ-
1A/B satellites have more frequent revisit capability (2 days) than the Landsat-5 satellite (16
days), which is of great importance for agricultural monitoring.
2.3.1.4 Construction of the spectral knowledge base
The SKB in this study can be interpreted as a pool of relationships between spectral
characteristics and prior knowledge. Here, prior knowledge stands for the degree of severity
of yellow rust, and the spectral characteristics are the reflectance of the initial four bands of
the HJ-CCD image. Hence, there are two major steps involved in constructing the SKB. First,
the relationship between hyperspectral information and severity is obtained with a stable
empirical reversion model. Then, through the RSR function of the HJ-CCD sensor, the
hyperspectal data can be transferred to the wide-band reflectance. In this way, a one-to-one
correspondence between the disease severity of yellow rust and reflectances from the HJ-
CCD sensor is established at the pixel level. The SKB can represent disease severity in two
ways: the DI (%) value and the class of disease severity. The following sections describe each
step for establishing the SKB, including data selection, the reversion model, simulation of
the wide-band reflectance and estimating the degree of severity. A technical flow diagram of
SKB construction is summarized in Fig. 8.

Fig. 8. The flow chart for monitoring of DI(%) of winter wheat stripe rust, b1-b4 represented
the reflectance of the four bands of HJ-CCD images

As noted above, the SKB in this study comprised PHI pixels. The predicted accuracy
obtained by the SKB was determined primarily by the amount of prior knowledge, which
indicated the heterogeneity of disease severity. The design of the yellow rust fungus
inoculation ensured a considerable variation in disease severity within the experimental
field, from healthy plants to very diseased plants. In addition, to avoid using pixels on or
near the ridge in the field that are considered as mixed signals, we chose three rectangular
Crop Disease and Pest Monitoring by Remote Sensing 51

shaped areas that were within the field and comprised 7918 ‘crop-only’ pixels for
constructing the SKB.
2.3.1.5 Reversion model
The reversion model construction was the first step of establishing the SKB. Based on the
conclusion of the part above, PRI was a suitable vegetation index for monitoring the severity
of yellow rust disease in winter wheat. Therefore, in this study, PRI was used to establish
the linkage between the disease severity and the hyperspectral data. Specifically, the yellow
rust infection would be apparent at anthesis stage, and this should be closely related with
the subsequent yield loss. Therefore, we chose the PHI image at this stage to form the SKB.
To obtain a better fitting model, we reanalyzed the PHI-PRI and corresponding DI (%) data
at the anthesis stage specifically, and obtained a linear regression model. It should be noted
that the data range of DI must be between 0 and 100%. Any predicted DI results that
were＞100% or ＜0% were redefined as DI = 100% and DI = 0% to represent very infected
plants and healthy plants, respectively.
2.3.1.6 Simulation of the wide band reflectance
The second step of constructing the SKB is to transform the hyperspectral reflectance of PHI-
pixels to wide band reflectance of HJ-pixels. To achieve this goal, the best approach is the
inherent relative spectral response (RSR) function of the HJ-CCD sensor. By integrating the
hyperspectral reflectance of PHI-pixels on the RSR function, the band reflectance of HJ-CCD
sensor was thus obtained. Besides, although the wavelength range of the fourth band of HJ-
CCD sensor (760 nm-900 nm) was slightly exceeded the maximum wavelength of PHI
sensor (850 nm), for most ground measured spectra, the reflectance basically kept on steady
from 760 nm to 900 nm. Hence, the simulating results generated using the incomplete range
of wavelength (760nm-850nm) should approach to the true value. The integration can be
shown as follows:

bend
RTM   f ( x )dx
bstart

where RTM is the simulated reflectance of a certain band; bstart and bend indicate the
beginning and the end wavelength of this band respectively; f(x) indicates the RSR function,
which is obtained from CRESDA.
2.3.1.7 Spectral characteristics of different degrees of disease severity
Another way to define the disease severity of an undefined pixel, apart from the DI (%)
value, is to quantify disease severity by severity classes. The criterion of severity class
provided by Huang et al. (2007) was adopted, which corresponded to the major
physiological alteration of diseased plants. The DI (%) thresholds for each severity class
were: DI＜1% indicated not infected (NI), 1%＜DI＜10% indicated a low degree of infection
(LI), 10%＜DI＜45% indicated mid-range infection (MI), 45%＜DI＜80% indicated seriously
infected (SI) and DI (%)＞80% indicated very seriously infected (VI). The MNF
transformation and PPI function, which are used for noise reduction and end-member
identification, were applied here to select the most representative pixels from the PHI
image, and to form the typical spectrum for each severity class.
52 Remote Sensing – Applications

2.3.1.8 Spectral matching algorithms

The basic idea of spectral matching is to identify a set of pixels in the SKB that are the closest
to the undefined pixel in terms of spectral characteristics. Before matching, each pixel
should be standardized to eliminate systematic variation caused by aerosol conditions or
other factors as follows:
R  Rmin
R nor 
Rmax  Rmin

where Rnor is the standardized reflectance of a certain band, R is the original reflectance, and
Rmin and Rmax are the minimum and maximum band reflectance values, respectively, of the
corresponding pixel.
Mahalanobis distances (Mah) and Spectral angle (SA) were selected as the distance
measurement criterion. Both types of distance measurements had been proved to be with
high efficiency in reflecting the spectral discrepancy (South et al., 2004; Goovaerts et al.,
2005; Becker et al., 2007). The Mah distance can be written as:

DM ( x )  ( x  x R ) 1 ( x  x R )T x=(x1,x2,x3,x4), xR=(xR1, xR2, xR3, xR4)

where x1-4 are the reflectance of the pixel under test in band1 to band4, respectively; xR1-4 are
the simulated reflectance of a specific pixel in SKB. ∑ is the covariance matrix between x and
xR.SA can be calculated by the following formula:

4
 xi xRi  
i 1
  arccos   0, 
4 4  2
 xi2  xRi2
i 1 i 1

To determine the DI (%) or class of disease severity of an undefined pixel, we have to

calculate the Mah and spectral angle from this pixel to each pixel or class in the SKB. A
longer distance or larger angle indicates that the pixel deviated from the undefined pixel,
whereas a shorter distance or smaller angle indicates that it is similar to the undefined pixel.
By selecting the most similar pixel, the severity class of an undefined pixel can be
determined. To determine the DI (%) of a certain pixel, the weighted average method was
used. According to the distance criteria above, the five most similar pixels were selected
from the SKB. For each band of these pixels (here we used the hyperspectral bands extracted
from the PHI image), the reflectance was processed according to the following equation:

k
1
 Ri  d
i 1 i
RE  k
1
d
i 1 i

where RE is the estimated reflectance of a certain pixel through k-NN estimation; Ri is the
reflectance of the ith nearest pixel according to the ranking order of the distance; di is the
distance between the pixel under test to the ith nearest pixel.
Crop Disease and Pest Monitoring by Remote Sensing 53

2.3.1.9 Verification
To verify the performance of SKB in identifying and monitoring the severity of yellow rust
diseases, two datasets were used: the simulated data and the field-measured data with
corresponding satellite images.
1. Verification of SKB using simulated data
The simulated data comprised 50 randomly selected pixels in the same experimental field,
but outside the three regions selected for constructing the SKB. The hyperspectral
information of each pixel was used to create the reference DI (%) and severity class with
the empirical model and the corresponding threshold for each severity class. To test the
performance of SKB in terms of DI (%) value, we estimated the DI value with both
distance criteria described above. The samples were split into two: the pixels with a
reference DI between 1 and 100%, i.e. the ‘diseased’ pixels, and those with a reference
DI＜1%, i.e. ‘healthy’ pixels. For the diseased pixels, the estimated DIs were compared
with the reference DI by Pearson correlation analysis and the normalized root mean
square error (NRMSE). For the healthy pixels, we used ‘yes or no’ to determine whether
the estimated value indicated infection or not, which also provided an accuracy ratio. The
estimation of severity class was verified by overall accuracy and the kappa coefficient.
2. Verification of SKB using field surveyed data
The field surveyed data sets included the ground investigation of disease severity and the
corresponding HJ-CCD images. Between June 1–3, 2009, when the winter wheat was at the
anthesis stage, we conducted a survey in the southeast of GanSu Province. The climate of
the area surveyed is characterized by high humidity and rainfall, and yellow rust disease
occurs almost every year. This area has similar environmental conditions and cultivation
customs to those where we constructing the SKB in Beijing, and this makes it an
appropriate place for model verification. With the aid of the local Department of Plant
Protection, 26 plots were randomly selected and surveyed in the area (Fig. 9). To relate the
surveyed value to the pixel value of the HJ-CCD image, we defined the plot as a uniformly
planted winter wheat region with an area no less than 30 m in radius. The geographical
coordinates of each plot were measured by GPS at the centre of the plot. Disease severity
was measured as described above. We repeated the measurement in five evenly-
distributed sections in each plot, and 20 individual plants were included in each
measurement. The HJ-CCD images (ID: 122516, 122518) acquired on June 2, 2009
completely covered the surveyed area. The raw data from the HJ-CCD imagery was
calibrated based on the corresponding coefficients provided by CRESDA. The calibrated
data were atmospherically corrected with the algorithm provided by Liang et al. (2001),
which estimated the spatial distribution of atmospheric aerosols and retrieved surface
reflectance under general atmospheric and surface conditions. The images were also
geometrically corrected against historical reference images with the same geographical
coordinates. The images were rectified with a root mean square error of less than 0.5 pixels.
The spectrum of the each plot was extracted from the image according to the GPS records.
The estimated accuracy in this step followed the same process as the simulated data.

2.3.2 Results
There were 7918 pixels included in the process of constructing the SKB. The linear
regression model between DI (%) and PRI at anthesis stage could be illustrated as follows:
54 Remote Sensing – Applications

Fig. 9. The field surveyed area in Gansu Province. The base image is the HJ-CCD image
acquired on June 2, 2009

DI (%)  538.98  PRI  2.0983 (R2=0.88)

The pairs of DI (%) and PRI were plotted in Fig.4, which showed a significant correlation (R2
= 0.88). Based on the model, there were 85 pixels with a DI of 100% and 3991 pixels with a
DI between 1%and 100%, indicating 51.5%pixels infected to a varied degree of severity,
whereas the other 48.5% pixels (DI = 0%) were healthy plants. In the experimental field, the
variation in the degree of severity of yellow rust from totally healthy plants to very infected
plants provided the essential diversity or heterogeneity of infection, which then enabled
establishment of the SKB. The MNF transformation resulted in 9 leading eigenvectors with
eigenvalues greater than 4.0 (Fig. 10), and these were used for further analysis.
2.3.2.1 Performance of SKB for simulated data
In the simulated dataset, there were six healthy pixels and 44 diseases affected ones. When
estimating DI (%), one pixel with no infection was estimated as infected by the Mah distance
criterion, whereas with the SA criterion two were mislabeled. Fig.11 shows the scatter of the
disease affected pixels plotted in relation to reference DI and estimated DI; the average
reference DI is 36%. The reference DIs and estimated DIs were strongly and linearly
correlated for both the Mah distance (R2 = 0.90) and SA (R2 = 0.84) criteria. Further, the
NRMSE of Mah distance and SA were 0.20 and 0.24, respectively, indicating that the SKB
can estimate DIs accurately from the simulated multi-band reflectance.
Crop Disease and Pest Monitoring by Remote Sensing 55

Fig. 10. MNF eigenvalues variation trend

Fig. 11. Estimated DI(%) using simulated data

Table 6 gives the reference class of disease severity and the estimated class in the form of an
error matrix. The overall accuracy with Mah distance and the SA criterion were 0.80 and
0.76, respectively, whereas the kappa coefficients were 0.71 and 0.65, respectively. However,
we noticed that all the misclassified pixels were assigned to no more than one class adjacent
to the reference class. Therefore, for simulated data, the classification accuracy was
satisfactory in determining the severity class of yellow rust by SKB.
2.3.2.2 Performance of SKB for field surveyed data
Apart from the verification against simulated data, more importantly, the field surveyed
data can be also used to assess the performance of the SKB. The field investigation showed
that eight out of 26 plots were infected with DI ranged from 4 to 90%, whereas the other 18
plots were not affected by yellow rust. The estimation by DI (%) successfully identified the
eight infected plots when the Mah distance criterion was used, whereas the SA criterion
56 Remote Sensing – Applications

Reference
Mid Very
None Low range Serious Total
range serious
None 6 0 0 0 0 6
Low range 0 5 2 0 0 7
Mid range 0 1 20 2 0 23
Estimation
(Mah) Serious 0 0 1 10 1 12
Very
0 0 0 1 1 2
serious
Total 6 6 23 13 2 50
None 5 1 0 0 0 6
Low range 1 4 1 0 0 6
Mid range 0 1 20 2 0 23
Estimation
(SA) Serious 0 0 2 9 1 12
Very
0 0 0 2 1 3
serious
Total 6 6 23 13 2 50
Table 6. Error matrix for simulated data

resulted in one misestimated plot. Figure 7 shows the scatter of the eight data plotted in
relation to reference DI and estimated DI for both distance criteria. There was a significant
linear trend in graphs based on both the Mah distance and SA criteria. The R2 of Mah distance
and SA were 0.80 and 0.67, respectively, whereas the NRMSE were as high as 0.46 and 0.55. In
real circumstances, approximately 50% error in the estimated disease index is unsatisfactory.
On the other hand, however, most of the uninfected plots were correctly identified according
to DI (%) estimates (i.e. a DI<1%). For both the Mah distance and SA criteria, 15 out of 18 non-
infected plots had been identified correctly, resulting in an accuracy of 77.8%. The results for
estimating disease severity by severity class were even more encouraging. The overall
accuracy for the Mah distance and SA criteria were 0.77 and 0.73, respectively, whereas the
kappa coefficients are 0.58 and 0.49, respectively. Table 3 gives the error matrix for both
criteria. The misclassified pixels were also assigned exclusively to the adjacent class.
In general, the above results demonstrate that the proposed SKB scheme has great potential for
detecting the incidence and severity of yellow rust through multispectral images. As shown
from several previous studies, the image processing method of MNF transformation was
efficient in extracting the principle information from the images related to wheat disease
infection (Zhang et al. 2003; Franke and Menz 2007). For the present study, we found that
coupling MNF transformation with the PPI function was an appropriate way of extracting the
principle information on yellow rust disease. To estimate disease severity by DI (%), the
proposed SKB has achieved a satisfactory accuracy for simulated data. However, the estimated
accuracy for field surveyed data was unsatisfactory, implying that the method tends to
underestimate or overestimate the disease severity in practice. Nevertheless, to estimate disease
severity through disease severity class has achieved a satisfactory accuracy for both simulated
data and field surveyed data. Therefore, the disease severity class seems to be more robust in
Crop Disease and Pest Monitoring by Remote Sensing 57

determining the disease severity. This might be because it is more rough estimation than DI (%).
It is understandable that for the same sample, the less precise the criterion, the greater accuracy
it would achieve. Moreover, the 5-class disease severity quantification is enough to guide field
applications. We suggest that DI (%) should be used for detecting the disease severity of yellow
rust by SKB. For the distance criteria used in the process of matching with SKB, the Mah
distance criterion might be more appropriate because it performed better than SA in all the
analyses conducted in this study (Figs. 11, 12, Tables 6, 7). Some previous studies have already
emphasized the potential of hyperspectral imagery (Bravo et al. 2003; Moshou et al. 2004;
Huang et al. 2007) and the high-resolution of multispectral imagery (Franke and Menz 2007) for
detecting yellow rust disease. The development of SKB in the present study can be viewed as a
scaling up method, which has extended the capability of detecting yellow rust disease from
hyper- spectral imagery to the moderate resolution of multispectral imagery. However, it
should be noted that the task of monitoring the occurrence and degrees of infection of crop
diseases is far more complex than the cases described in this study. The spectral characteristics
of yellow rust infection might appear similar to other sources of stress. In addition, the impact
of phenology, cultivation methods, fragmentation of farmlands and other environmental
conditions would also increase the difficulty and uncertainty of the estimation process.
Therefore, the SKB developed in this study should correspond to the situation at the anthesis
stage exclusively, and is only suitable for those regions with similar environmental
characteristics and cultivation methods. For other regions with significantly different
environmental characteristics, this purposed SKB may not work well. The possible solution to
these problems may include incorporating suitable priors, which would require integration
strategies and understanding of the mechanisms underlying some fundamental processes.
Further research is required to address the problems mentioned above.

Fig. 12. Estimated DI(%) using field measurements

2.3.3 Conclusion
The low spatial resolution and few spectral bands have limited the application of moderate
resolution satellite images for monitoring yellow rust disease. The spectral knowledge base
developed enabled disease incidence and severity to be detected by moderate resolution
satellite images. The SKB supported two ways of estimating disease severity: the disease
58 Remote Sensing – Applications

Reference
None Low range Mid range Serious Very serious Total
None 16 0 0 0 0 16
Low range 2 2 1 0 0 5
Estimation Mid range 0 1 3 0 0 4
(Mah) Serious 0 0 0 0 1 1
Very serious 0 0 0 0 0 0
Total 18 3 4 0 1 26
None 15 0 0 0 0 15
Low range 3 2 1 0 0 6
Estimation Mid range 0 1 3 0 0 4
(SA) Serious 0 0 0 0 1 1
Very serious 0 0 0 0 0 0
Total 18 3 4 0 1 26
Table 7. Error matrix for ground measured data

index and disease severity class. Both methods of estimation achieved a satisfactory level of
accuracy for simulated data. For field surveyed data, estimation by DI (%) resulted in an
unsatisfactory level of accuracy, whereas it was satisfactory for severity class. The Mah
criterion performed better than spectral angle in all analyses. Therefore, the former should
be considered as the more appropriate distance criterion.
Generally, the purposed SKB has a great potential in extending the capability of detecting
yellow rust to multispectral remote sensing data, especially when the region of interest has
similar environmental conditions to where the SKB was developed. The uncertainties
caused by environmental differences should be further investigated in future studies.

2.4 Detecting yellow rust of winter wheat using land surface temperature (LST)
The air temperature and humidity are the most direct and important indicators of
occurrence of yellow rust fungal. Generally, weather stations can provide the dynamic
pattern of meteorological data for site sampled, yet not able to include the information of
spatial heterogeneity. Fortunately, remote sensing technology has great potential for
providing spatially continuous observations of some variables over large areas (Luo et al.,
2010). The aim of the study was to study preliminarily on the relationship between the
occurrence of wheat yellow rust and land surface temperature (LST) derived from
moderate-resolution imaging spectroradiometer (MODIS) in order to predict and monitor
incidence of the yellow rust on large scale.

2.4.1 Materials and methods

2.4.1.1 Survey area and field investigations acquisition
Field experiments of winter wheat were conducted during the growing seasons (form April
to June) of winter wheat in 2008 and 2009. The investigation locations included Longnan
Crop Disease and Pest Monitoring by Remote Sensing 59

district, Tianshui district, Dingxi district and Pingliang district in GanSu province and
Qingyang district in ShanXi province as well as Linxia district in Ningxia Hui Autonomous
Region (Fig.1), where the climates are semiarid and subhumid. Survey areas are located
between latitude 32º40’N to 35º39’N and longitude 103º10’E to 107º40’E, and the mean
altitude is over 2000 meter. The climate condition of surveyed area is characterized by high
humidity and rainfall, and yellow rust disease almost occurs every year. It is reported that
Longnan district is an important overwintering and oversummering area of yellow rust
fungal (Zeng, 2003).
With the aid of the local Department of Plant Protection, 151 plots, including 68 plots from
April to June in 2008, and 83 plots from April to June in 2009, were randomly selected and
surveyed in the areas. The geographical coordinates of each plot were measured by GPS
navigator at the middlemost of the plot. In addition, the disease severity was inspected.
2.4.1.2 MODIS land surface temperature (LST) products (MOD11)
Product description
MODIS Land Surface Temperature and Emissivity (LST/E) products (named starting with
MOD11) provide per-pixel temperature and emissivity values. Temperatures are extracted
in Kelvin with a view-angle dependent algorithm applied to direct observations. This
method yields the error less than 1 K for materials with known emissivity. The view angle
information is included in each LST/E product.
MOD11 acquisition and processing
24 MOD11A2 images（MODIS/Terra land surface temperature/emissivity 8-day L3 global
1km SIN grid v005）were acquired for free from Web (http://edc.usgs.gov/#/Find_Data)
from April to July in 2008 and 2009, which covered completely the survey area, and 4 scenes
images were acquired in every month. The raw data of MOD11A2 imagery were processed
and transformed by MRT tool, and LST products were extracted from MODII A2 images.
Then the survey area was cut by ENVI from LST images. Followed by that step, 4 scenes 8-
day LST images of every month were all averaged, and 6 average LST images, including
April, May, June in 2008 and 2009, were obtained. Finally, LST of 151 investigation points
were respectively extracted from 6 average LST images.

2.4.2 Result
2.4.2.1 Determining LST threshold of infected points
The spatial resolution of MODIS temperature products is 1 km, while the DI of every
investigation point only stands for the incidence of 30 m in semi diameter plots. Therefore,
the scale of MODIS temperature products seemed not satisfied the investigation points for
proper relationship between them. However, spatial variability of LST is slim, and the law
still exists. A series of results could be found by establishing a two-dimensional spatial
coordinate based on DI and LST, in which all investigation points were displayed (Fig 13).
Firstly, the DI ranged from 0% to 100%, and most of infected points ranged from 0% to 60%.
The LST values were between 292K and 310K with most of infected points distributed in the
range from 298K to 306K. In addition, the points in the region of less than 298K were not
infected by yellow rust basically; DI were less than 1% expect for one point (296.29K, 16%),
60 Remote Sensing – Applications

which was thought as abnormal point. In addition, the LST values of all investigation points
were less than 306K expect for one point (310.09K, 24%), which was abnormal because its
LST was far away from LST values of others.

Fig. 13. The distribution of the investigation points

Therefore, without considering other factors, It is concluded that yellow rust can occur
when LST is in the region from 298K to 306K.
2.4.2.2 Yellow rust incidence analysis based on LST
According to the results illustrated above, the advanced analysis was performed for
incidence and possible area of yellow rust. The points in different LST range were done
statistical analysis with all points’ numbers and the infected points’ number, and finally, the
incidences were obtained by the number of the infected points dividing the number of all
points in the different LST range (Table.8). The result showed that all investigation points in
the region of less than 298K were not infected by yellow rust, except for the abnormal point
(296.29K, 16%). On the other hand, in the LST region of more than 306K, there was only one
point, which was viewed as abnormal point (310.085K, 24%). Thereby, it is quite possible
that yellow rust fungus can not survive in the region of more than 306K. The conclusion was
consistent with the above result (Fig. 13).

LST≥2 LST≥2 LST≥ LST≥ LST≥ LST≥ LST≥ LST≥ LST≥ LST≥ LST≥
LST (K)
96 97 298 299 300 301 302 303 304 305 306
Total
126 112 99 79 61 34 25 16 12 8 1
number
Number of
infected 49 48 47 42 39 27 25 16 12 8 1
points
Incidence (%) 38.89 42.86 47.47 53.16 63.93 79.41 100 100 100 100 100

Table 8. Statistic analysis in different LST range

Crop Disease and Pest Monitoring by Remote Sensing 61

Furthermore, there was an increasing trend of incidences with the rising of LST in the region
from 296K to 302K. The incidence of yellow rust reached up to 100% when the LST was
graeter than 302K (Fig. 14).

Fig. 14. The incidence of yellow rust in different LST range

2.4.2.3 Dividing yellow rust suitable occurrence region based on LST

According to Table 8 and Fig. 14, the survey areas could be divided into yellow rust
unsuitable area (NSA), of which LST ranged from 298K to 306K, and yellow rust suitable
area (SA), of which the LST was less than 298K and more than 306K. Moreover, the SA was
divided into 3 levels according to the infected of yellow rust incidence and LST, and the LST
thresholds for each level were: 298K ≤ LST ≤ 299K the low suitable area (LSA), on which the
yellow rust occurs with very low possibility (incidence < 60%), 299K ≤ LST ≤ 301K the
medium suitable area (MSA), which had moderate possibility for the occurrence of yellow
rust (60% <incidence < 100%), and 302K ≤ LST ≤ 306K high suitable area (HSA), of which the
environment was highly favorable to yellow rust (incidence=100%).
2.4.2.4 Verification
Total 26 points (from May 2008) were applied for the verification the method of estimating the
incidence of yellow rust. It should be noted that those points were not used for the defining of
the LST thresholds. (Fig. 15). These 26 points were constituted by 18 infected points and 8 non-
infected points. Results showed the infected points were all in different suitable areas of wheat
yellow rust, while the non-infected points were all in the unsuitable area. Thus the infected
situation of yellow rust of these 26 points was consistent with forecast results. Geographically,
it seemed that the yellow rust was prone to be prevalent in the northeast of Pingliang,
southwest of Qingyang, northeast of Dingxi, the center part of Tianshui, and the west of
Longnan, because they all were located in MSA and HAS. This result was consistent with the
previous study (Xiao, et al, 2007). To prevent yellow rust from prevalence, more efforts should
be placed on the farmlands located in the MSA, HAS and LSA.

2.4.3 Conclusions
Plant disease is governed by a number of factors, and the habitat factors play a major role in
the development and propagation of fungal pathogens (Sutton et al., 1984; Hélène et al.,
62 Remote Sensing – Applications

Fig. 15. Forecast map of yellow rust and distribution of measured points in May, 2008 based
on LST

2002; Cooke et al., 2006). The yellow rust is no exception. The weather station can only offer
points data, and remote sensing, however, can be a promising means for acquiring spatially
continuous observations over large area. It has not been reported, if any, that the LST
derived from remote sensing data is used to forecast the development of yellow rust.
The study tried to present a method that could forecast the suitable areas of wheat yellow
rust by MODIS temperature products in a large scale. And it was proved that LST derived
from remote sensing data had potential for predicting the occurrence and development of
wheat yellow rust in a large area. From our results, it is clear that preventive measures of
yellow rust can been made over large scale area accordingly with different real-time
prediction methods based on LST derived from remote sensing data.

3. Detecting and discriminating winter wheat aphid by remote sensing

Wheat aphid, Sitobion avenae (Fabricius), is one of the most destructive pests in agricultural
systems, especially in temperate climates of the northern and southern hemispheres. Wheat
aphid appears annually in the wheat planting area of China, causing great economic
damage to plant crops as a result of their direct feeding activities. In high enough densities,
wheat aphids can remove plant nutrients, and potentially reduce the number of heads, the
number of grains per head, and overall seed weight. The damage is especially high when
wheat aphid occurs in the flowering and filling stage of wheat. It is reported that average
densities over 20 aphids per plant can cause substantial losses of yield and quality of wheat
(Basky & Fónagy, 2003). There are also indirect damages including excretion of honeydew
from aphids and as a vector of viruses, most notably two strains of the Luteovirus Barley
Yellow Dwarf Virus (BYDV-MAV and BYDV-PAV) (Susan et al, 1992). To prevent the
Crop Disease and Pest Monitoring by Remote Sensing 63

occurrence and prevalence of aphid, large amounts of insecticides are used, causing
environment pollution. Therefore, large-scale, real-time prediction and monitoring of wheat
aphid incidence and damage degree using remote sensing technology are extremely
important.

3.1 Detecting winter wheat aphid using hyperspectral data

The study aimed to identify spectral characteristics of wheat leaf and canopy infected by
aphid and find the sensitive bands to aphid at canopy level in filling stage of wheat, and to
establish an aphid damage hyperspectral index (ADHI) based on those sensitive bands for
detecting aphid damage levels in wheat canopy level in filling stage of wheat.

3.1.1 Materials and methods

3.1.1.1 Field experiments and field inventory
The field experiment plot was located at Xiaotangshan Precision Agriculture Experiment
Base, Changping distract, Beijing (40º10.6’N, 116º26.3’E). The experimental field was about
250 m in length and 80 m in width. The winter wheat was planted in the study area from
Oct 3, 2009, and harvested from June 25, 2010. Field inventory was conducted on June 7,
2010 when wheat was in the filling stage. Twenty five ground investigations including
different aphid damage levels were selected. Aphid damage level was surveyed according
to the investigation rule.
3.1.1.2 Canopy spectral measurements
Please refer to 1.1.1.2 part above.

3.1.2 Results
3.1.2.1 Leaf spectral characteristics of wheat infested by aphid
Representative reflectance measured from wheat aphid-infested and uninfested wheat
leaves are shown in Fig. 16. It was evident that the spectral response of the wheat leaf was
significantly affected by wheat aphid feeding (Fig. 16). The reflectance of wheat leaf infested
by aphid was higher in the visible spectrum and short-wave infrared region and lower in
near-infrared region than that of uninfested leaf. A significant increase in the reflectance
from the wheat aphid-infested leaf in the visible region (400-700 nm) was observed,
evidently due to reduction of photosynthetic pigment concentrations in particular
chlorophylls caused by wheat aphid feeding (Richardson et al., 2004).
3.1.2.2 Canopy spectral characteristics of wheat infested by aphid
Compared with the canopy spectra of the healthy wheat, the canopy reflectance of aphid-
infested wheat was gradually decreased in the range from 350 nm to 1750 nm, especially in the
near infrared region (Fig. 17). Previous researches indicated that wheat had higher reflectance
at visible wavelengths than the healthy vigorously growing wheat because the photoactive
pigments (chlorophylls, anthocyanins, carotenoids) were destroyed. In this study, aphid
occurred in the filling stage of wheat and the honeydew excreted by aphid absorbed dust or
others from surrounding environment and contaminated (darkened) the leaf surface. As a
result, the absorption at light slight wavelengths became stronger instead of weaker.
64 Remote Sensing – Applications

Fig. 16. The spectral reflectance of winter wheat leaf uninfested and infested by aphid

40
35
Healthy
30 Slight
Reflectance/%

25 Moderate

20 Severe

15

10

5
0
350 650 950 1250 1550 1850 2150 2450
Wavelength/nm

Fig. 17. The spectral reflectance of healthy wheat and wheat infested by various aphid
damage levels. (Healthy: the average spectra of healthy wheat samples; Slight: the average
spectra of aphid damage level 1and 2; Moderate: he average spectra of aphid damage level
3and 4; Severe: the average spectra of aphid damage level 5 and 6).

3.1.2.3 Aphid damage hyperspectral index for detecting aphid damage degree
Sensitive band selection of aphid infestation based on canopy reflectance
The sensitive bands were selected out by relevance analysis between reflectance and aphid
damage levels. The reflectance ranges were from 400 nm to 690 nm, from 700 to 1300 nm
and from 1500 to 1800 nm. The most sensitive bands to aphid were 551 nm (R2=0.741) in
Crop Disease and Pest Monitoring by Remote Sensing 65

visible light, 823 nm (R2=0.865) in near infrared (NIR) and 1654 nm in short-wave infrared
(SWIR) (R2=0.668), respectively (Fig. 18).

Fig. 18. Correlation coefficient between reflectance and aphid damage levels

Aphid damage hyperspectral index (ADHI) was established based on the most sensitive
bands from hyperspectral data in the visible light region, NIR and SWIR and weight
coefficient calculated according to rate of change of reflectance between healthy wheat and
aphid-infected wheat, respectively.

R551normal -R551infested R823normal - R823infested

ADHI  0.32   0.51 
R551normal R823normal
R1654normal - R1654infested
0.17 
R1654normal

where R551normal，R823normal and R1654normal are reflectance in 551 nm, 823 nm and 1654 nm
of healthy wheat; R551infested , R823 infested , R1654 infested are reflectance in 551 nm, 823 nm and
1654 nm of aphid-infected wheat; 0.32, 0.51 and 0.17 are weight coefficients calculated by the
contribution to change rates.
Further more, the correlation analysis was conducted between ADHI and aphid damage
level from 25 investigation points (Fig. 19). It was concluded that ADHI exhibited high
relationship with aphid damage levels (R2=0.839). Therefore, ADHI was an important index
to estimate aphid damage level in winter wheat.

3.1.3 Conclusions
Hyperspectral remote sensing has gone through rapid development over the past two
decades and there is a trend toward the use of hyperspectral image in the application of
remote sensing for precision farming. The study analyzed the spectral characteristics of
wheat infested by aphid and selected the sensitive bands to aphid damage level. Then, an
ADHI was developed using the most sensitive bands in visible light region, NIR and SWIR.
66 Remote Sensing – Applications

6
R2 = 0.839

Aphid damage level

5

0
-0.01 0.05 0.11 0.17 0.23 0.29 0.35 0.41
ADHI
Fig. 19. The correlation between ADHI and aphid damage level

It was concluded that ADHI was a sensitive index to aphid damage levels, and could be
used to retrieve aphid damage levels in the filling stage of wheat.
Crop growth is very dynamic processes and monitoring the condition of agricultural corps
is a complex issue. It is possible that wheat damage symptoms caused by aphids and its
response of canopy reflectance are different in different wheat growth stages. This study
revealed that the reflectance of wheat infested by aphid was lower than healthy wheat in
filling stage probably because of honeydew excreted by aphid. This was not consistent with
previously published results in early detection of aphid infestation. Therefore, whether the
ADHI can effectively retrieve aphid damage levels in other wheat growth stages remains as
a task of future studies.

3.2 Detecting winter wheat aphid incidence using Landsat 5 TM

Wheat aphid occurrence and damage degrees are related to many factors including
temperature, humidity, precipitation, field management, enemies, etc.. Most of the present
studies on aphid prediction have been conducted based on meteorological data acquired
from weather stations, and aphid density was monitored using the spectral characteristics of
wheat infested by aphid. However, it is rare to investigate the relationship between
environmental parameters, vegetable information derived from satellite images and aphid
damage degrees. The aim of the present study is to investigate the relationships of aphid
occurrence and damage degree to LST, NDWI, and MNDWI, which are related to vegetation
water content derived from multi-temporal Landsat 5 TM. Another goal of the current
research is to distinguish the degrees of aphid damage using 2-dimension feature spaces
established by LST-NDWI and LST-MNDWI.

3.2.1 Materials and methods

3.2.1.1 Study areas
The study areas are selected in Shunyi district (116°28'—116°58' E，40°00'—40°18' N) and
Tongzhou district (116° 32'—116°56' E, 39°36' —40°02' N,) of Beijing, China (Fig.20-a). The
Crop Disease and Pest Monitoring by Remote Sensing 67

a) b)
Fig. 20. The study area and the spatial distribution of sample plots

study areas have flat topography, with elevation ranging from 20 m to 40 m. The study areas
have semi-humid warm temperate climate with yearly precipitation of 625 mm and mean
temperature of 11.5°C in the Shunyi district and yearly precipitation 620 mm and mean
temperature of 11.3°C in the Tongzhou district. Both districts are considered main winter
wheat planting areas in Beijing, and aphid infestations occur in both areas almost every
year.
68 Remote Sensing – Applications

3.2.1.2 Field inventory and data pre-processing

Field inventory was conducted during the growing seasons of winter wheat in 2010. The
winter wheat in the study areas were planted between September 25 and October 7, 2009,
and harvested between June 19 and June 25, 2010. Based on the combination of
representative sampling and random sampling scheme, 70 sample plots with size of 0.09 ha
(30 m × 30 m) each were collected as in Fig 1-a. These sample plots had different site
conditions, plant densities, and management conditions. Aphid density surveys were
carried out respectively on May 4 and May 6 for jointing stage, May 20 and May 21 for the
heading stage, and June 3 and June 4 for the filling stage. The geographical coordinates of
each plot were measured by global positioning system (GPS) ( GeoExplorer 3000 GPS, with
the error within 1m) at the middlemost of the plot.
Each sample covered with an area of 1 m2. Then, 10 tillers in each sample plot were
randomly selected, and the number of aphids was counted. The aphid densities were then
estimated as follows: total aphids /10 tillers.
The survey results were divided into three aphid damage degrees according to the aphid
density investigated for facilitating the study. They were S0: non-infested by aphid and no
damage to wheat, S1: aphid abundance/per tiller was about 3-10 and damage degree to
wheat was slight, and S3: aphid abundance/per tiller was more than 20 and damage degree
to wheat was severe.
3.2.1.3 Satellite image acquisition and pre-processing
Three Landsat-5 Thematic Mapper (TM) images (path 123/row 32) and three MOD 02 1
KM-Level 1B Calibrated Radiances Production (MOD 02) were acquired on May 4, May 20
and June 5, 2010, respectively. And all images were more than 90% cloud-free.
The Landsat-5 TM images were spectrally corrected to reflectance using the Landsat TM
calibration tool and FLAASH (Fast line-of-sight Atmospherics Analysis of Spectral
Hypercubes) was used to correct the image for atmospheric effects in ENVI 4.5. The
Landsat-5 TM images were geometrically corrected versus a reference IKONOS image
(equivalent scale map 1:10000) of the same area, available from a previous study. The
resulting root mean square error (RMSE) did not exceed 0.3 pixels, which was adequate for
the purposes of the present study.
3.2.1.4 Derivation of LST, NDWI and MNDWI from Landsat 5 TM
NDWI and MNDWI are both sensitive to changes in liquid water content of vegetation
canopies (Hunt and Rock, 1989). In the current research, both NDWI and MNDWI were
used to determine the threshold of aphid occurrence and the aphid damage degree. The
indices are of the general form, as shown in the following:

RNIR  RSWIR R -R
NDWI  MNDWI  GREEN SWIR
RNIR +RSWIR RGREEN +RSWIR

where RGREEN ,RNIR and RSWIR are the reflectance in the green band, near-infrared band and
short wave infrared band, respectively. For Landsat TM/ETM+, RGREEN ,RNIR and RSWIR
correspond to band2, band4 and band5, respectively.
Crop Disease and Pest Monitoring by Remote Sensing 69

LST is the radioactive skin temperature of the land surface, which plays an important role in
farm and ecological environment. The present paper aims to discuss the relationship
between LST and aphid occurrence and spread. LST was derived from the thermal infrared
band (10.4-12.5μm) data of Landsat-5 TM using generalized single-channel algorithm
developed by Jiménez-Muñoz and Sobrino (Jiménez-Muñoz and Sobrino, 2004). Surface
emissivity (ε) and atmospheric water vapor content (w) were important parameters in the
generalized single-channel algorithm. In the study, w was derived from the reflectance of
band2 and band19 of MOD02, (Kaufman and Gao, 1992), and ε was calculated by vegetation
coverage (Carlson and Ripley, 1997).
The NDWI, MNDWI and LST of all sample points were calculated and extracted from the
Landsat images.
3.2.1.5 Subset image selection and wheat extraction
We resized the subset areas with size of 7.2 km2 (3 km × 2.4 km) from the study area image
located in Tongzhou district and covered with 20 evenly distributed sample points, and the
aphid densities of the sample points were surveyed on May 6, May 20 and June 4, 2010,
respectively. The survey results showed that the aphid damage degree of all sample points
were S0 on May 6, 18 points for S1 and 2 points for S0 on May 20, and 16 points for S2 and 4
points for S0 on June 4, respectively. The subset areas were small enough and 20 sample
points evenly distributed, According to the survey result, the aphid damage degree of the
sample plots was basically same. Thus, the change of the aphid damage degree of wheat
pixels in the wheat plots was slim or even basically the same as the sample plots. The wheat
area of subset image selection area was extracted using classification of decision tree in
ENVI 4.5 (Fig 20-b). The LST, NDWI and MNDWI of 2000 wheat pixels were extracted.
3.2.1.6 Methods of accuracy assessment
One basic accuracy assessment currently being used is overall accuracy, which is calculated by
dividing the correctly classified pixels by the total number of the pixels checked. The Kappa
coefficient is a measure of the overall agreement of a matrix introduced to the remote sensing
community in early 1983. It has since become a widely used measure for classification
accuracy. In contrast to overall accuracy, the Kappa coefficient takes non-diagonal elements
into account (Rosenfield and Fitzpatrick-Lins, 1986), and it is calculated by the formula:

r r
N  Xii   Xi  X  i
i 1 i 1
K r
N 2   Xi  X  i
i 1

where r is the number of rows and columns in the error matrix; N is the total number of
observations; Xii is the observation in row i and column i; Xi+ is the marginal total of row I;
X+i is the marginal total of column i.

3.2.2 Results
3.2.2.1 2-dimensional feature space based on LST-VI
The minimum value, maximum value, mean values and standard deviations of LST, NDWI
and MNDWI with aphid damage degrees of wheat pixels in subset image selection were
70 Remote Sensing – Applications

listed in Table 9 and Table 10. And 2-dimensional feature space coordinates were
established with LST as the abscissa and NDWI and MNDWI as the vertical axis,
respectively (Figs. 2, 3). LST ranged from 287.5879 to 313.3448, NDWI ranged from 0.0226 to
0.5591 and MNDWI ranged from -0.3402 to -0.1077, respectively.
It is clear that LST was increasing from S0 to S1 to S2. LST was an important driving factor
for aphid occurrence and could distinguish wheat non-infected from infested by aphids (Fig.
21 and Table 9). The general trend of NDWI increased firstly and reduced afterward,
whereas MNDWI reduced firstly and increased afterward from S0 to S1 to S2.

Aphid LST NDWI MNDWI

Damage Minimum Maximum Minimum Maximum Minimum Maximum
Degree value value value value value value
S0 287.5879 296.2498 0.0226 0.4405 -0.3402 -0.1077
S1 297.8084 306.0133 0.2083 0.5591 -0.6506 -0.3326
S2 300.5391 313.3448 0.0473 0.4542 -0.4117 -0.1159
Table 9. Minimum and maximum values of LST, NDWI and MNDWI in S0, S1 and S2

Aphid LST NDWI MNDWI

Damage Standard Mean Standard Standard
Degree Mean value deviation value deviation
Mean value
deviation
S0 290.8578 1.4740 0.3029 0.0574 -0.2293 0.0296
S1 299.9236 1.0834 0.3998 0.0587 -0.4940 0.0362
S2 303.9424 1.7121 0.2979 0.0458 -0.2672 0.0402
Table 10. Mean value and standard derivation of LST, NDWI and MNDWI in S0, S1 and S2

Fig. 21. The distribution of S0, S1 and S2 in the LST-NDWI (left) and LST-MNDWI (right)
feature space
Crop Disease and Pest Monitoring by Remote Sensing 71

3.2.2.2 Discriminating aphid damage degrees using LST and MNDWI

In the 2-dimensional feature space coordinate system that was composed by LST and
MNDWI, the S0 samples mainly scattered on the left part of the coordinate system, whereas
S1 and S2 samples were distributed on the right part. As shown in Fig. 22, when LST was
lower than the certain value, aphid did not occur, suggesting that LST served as a key factor
of aphid occurrence and the MNDWI was sensitive to aphid damage degree.
Furthermore, LST0 and MNDWI0, which were the cutoff value of threshold values of LST
and MNDWI of S0, S1 and S2, were determined by mean values and standard deviations.
LST0 and MNDWI0 were calculated by formula as follows:

LST0 =LST_M1－2×LST_SD1

MNDWI0= (M_M1+3×M_SD1)+ [(M_M1+ 3×M_SD1)-(M_M2-3×M_SD2)]/2

where LST_M1 and LST_SD1 are the mean value and standard deviation of LST for S1;
M_M1and M_SD1 are the mean value and standard deviation of MNDWI for S1; and M_M2
and M_SD2 are the mean value and standard deviation of MNDWI for S2.
According to Table 3, LST0 = 297.7568 and MNDWI0 = -0.3866. Wheat was not infested by
aphid when LST< 297.7568, and aphid damage degree was S1 when LST≥297.7568K and -
0.6506≤MDNWI ≤-0.3866 and S2 when LST≥297.7568K and -0.3866 ≤MDNWI ≤-0.1077 (Fig.
22).

Fig. 22. Discriminating aphid damage degrees using LST and MNDWI

3.2.2.3 Verification
All survey samples, except 20 samples in the subset selection image were used to test the
aphid prediction accuracy of 2-dimensional feature space based on LST and MNDWI (Fig.
23).
The discrimination accuracy was assessed using overall accuracy and kappa coefficient
(Table 11). The results showed that the overall accuracy was 84%, and the Kappa accuracy
was 75.67%.
72 Remote Sensing – Applications

Fig. 23. Distribution of test sample plots in LST-MNDWI feature space

S0 S1 S2 Total
S0 17 0 0 17
S1 2 14 0 16
S2 4 2 11 19
Total 23 16 11 50
Kappa coefficient = 0.7567
Table 11. Error matrices of the verification samples

3.2.3 Conclusions
This study successfully investigated the relationship between aphid damage degrees and
several spectral features, such as NDWI, MNDWI and LST, through 2-dimensional feature
space method. The results indicated that LST was the key factor in predicting the occurrence
of aphid, and MNDWI was more sensitive to aphid damage degree than NDWI. In the 2-
dimension feather space composed by LST and MNDWI, the result showed that S0, S1 and
S2 were divided into three regions; S0 was distributed on the left of the space, and S1 and S2
on the right. Further, LST0 and MNDWI0 were calculated according the mean and
derivation of S1, S2 as the cutoff value of threshold value to discriminate S0, S1 and S0.
Through the verification of discrimination threshold value, it confirmed that the overall
accuracy of discrimination was 84% and Kappa coefficient was 0.7567, suggesting that LST
and MNDWI were of great potential in discriminating and monitoring the aphid damage
degree over a large area, only using thermal infrared band and multi-spectral satellite
images.

4. References
Basky Z. & Fónagy A. (2003).Glutenin and gliadin contents of flour derived from wheat
infested with different aphid species. Pest Management Science, 59, 426-430.
Crop Disease and Pest Monitoring by Remote Sensing 73

Becker, B. L., David, P. L., & Qi, J. G. (2007). A classification-based assessment of the optimal
spectral and spatial resolutions for Great Lakes coastal wetland imagery. Remote
Sensing of Environment, 108, 111–120
Bravo, C., Moshou, D., West, J., McCartney, A., & Ramon, H. (2003). Early disease detection
in wheat fields using spectral reflectance. Biosystems Engineering, 84, 137–145.
Broge, N. H., & E. Leblanc. 2000. Comparing prediction power and stability of broadband
and hyperspectral vegetation indices for estimation of green leaf area index and
canopy chlorophyll density. Remote Sensing of Environment, 76, 156–72.
Carlson T. N. & Ripley D. A.. On the relation between NDVI, fractional vegetation cover,
and leaf area index. Remote Sensing of Environment, 1997, 62(3): 241-252.
Ceccato, P., N. Gobron, S. Flasse, B. Pinty, & S. Tarantola. 2002. Designing a spectral index to
estimate vegetation water content from remote sensing data: Part 1. Theoretical
approach. Remote Sensing of Environment, 82, 188–97.
Christou, P., Twyman, R.M., (2004). The potential of genetically enhanced plants to address
food insecurity. Nutrition Research Reviews 17, 23–42.
Daughtry, C. S., C. L. Walthall, M. S. Kim, E. Brown de Colstoun, & J. E. McMurtrey.(2000).
Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance.
Remote Sensing of Environment, 74, 229-239.
Fensholt, R., & I. Sandholt.(2003). Derivation of a shortwave infrared water stress index
from MODIS near-and shortwave infrared data in a semiarid environment. Remote
Sensing of Environment, 87, 111–21.
Filella, I., Serrano, L., Serra, J., & Penuelas, J. (1995). Evaluating wheat nitrogen status with
canopyreflectance indices and discriminant analysis. Crop Science, 35, 1400–1405.
Franke, J., & Menz, G. (2007). Multi-temporal wheat disease detection by multi-spectral
remote sensing. Precision Agriculture, 8, 161–172.
Galvão, L. S., A. Formaggio, R. & Tisot, D. A. (2005). Discrimination of sugarcane varieties in
Southeastern Brazil with EO-1 Hyperion data. Remote Sensing of Environment, 94,
523–34.
Gamon, J. A., Penuelas, J., & Field, C. B. (1992). A narrow-waveband spectral index that
tracks diurnal changes in photosynthetic efficiency. Remote Sensing of Environment,
41(1), 35–44.
Gitelson, A. A., Merzlyak, M. N., & Chivkunova, O. B. (2001). Optical properties and
nondestructive estimation of anthocyanin content in plant leaves. Photochemistry
and Photobiology, 74, 38–45.
Goovaerts, P., Jacquez, G. M., & Marcus, A. (2005). Geostatistical and local cluster analysis of
high resolution hyperspectral imagery for detection of anomalies. Remote Sensing of
Environment, 95, 351–367.
Gong, P., Pu R., Heald R.C.(2002). Analysis of in situ hyperspectral data for nutrient
estimation of giant sequoia. International Journal of Remote Sensing, 23, 1827-1850.
Haboudane, D., Miller, J. R., Tremblay, N., Zarco-Tejada, P. J., & Dextraze, L. (2002).
Integrated narrowband vegetation indices for prediction of crop chlorophyll
content for application to precision agriculture. Remote Sensing of Environment, 81,
416–426.
Haboudane, D., J. R. Miller, E. Pattery, P. J. Zarco-Tejad, & I. B. Strachan. (2004).
Hyperspectral vegetation indices and novel algorithms for predicting green LAI of
74 Remote Sensing – Applications

crop canopies: Modeling and validation in the context of precision agriculture.

Remote Sensing of Environment, 90, 337–52.
Hélène L., Frédéric F., Pierre D., Gaétan B., Isztar Z. (2002),Estimation of the spatial pattern
of surface relative humidity using ground based radar measurements and its
application to disease risk assessment. Agricultural and Forest Meteorology, 111, 223-
231.
Huang, W. J., David, W. L., Niu, Z., Zhang, Y. J., Liu, L. Y., & Wang, J. H. (2007).
Identification of yellow rust in wheat using in situ spectral reflectance
measurements and airborne hyperspectral imaging. Precision Agriculture, 8, 187–
197.
Hunt E.R., Rock B.N. (1989). Detection of changes in leaf water content using Near- and
Middle-Infrared reflectances. Remote Sensing of Environment, 30, 43-54.
Jiménez-Muñoz J. C., Sobrino J. A. & Leonardo P. (2004). Land Surface Temperature
Retrieval from Landsat TM 5. Remote Sensing of Environment, 90, 434-440.
Kaufman Y. J. & Gao B. C. (1992). Remote Sensing of Water Vapor in the Near IR from
EOS/MODIS. IEEE Transactions on Geoscience and Remote Sensing, 30, 871-884.
Kim, M. S., C. S. T. Daughtry, E. W. Chappelle, & J. E. McMurtrey. 1994. The use of high
spectral resolution bands for estimating absorbed photosynthetically active
radiation (APAR). In Proceedings of the 6th International Symposium on Physical
Measurements and Signatures in Remote Sensing, 299–306.
Li, G. B., Zeng, S. M., & Li, Z. Q. (1989). Integrated Management of Wheat Pests (pp. 185–
186). Beijing: Press of Agriculture Science and Technology of China (in Chinese).
Luo, J. H., Zhang, J. C., Huang, W. J., Xu, X. G., Jin, N. (2010). Preliminary study on the
relationship between land surface temperature and occurrence of yellow rust in
winter wheat. Disaster Advances, 3, 288-292.
Luo, J. H., Huang, W. J., Zhang, J. C., Xu X. G. & Wang D. C. (2011). The preliminary study
on spectral response of wheat under different stresses between field and satellite
remote sensing. Sensor Letters, 9, 1225-1228.
Merton, R., & J. Huntington.(1999). Early simulation of the ARIES-1 satellite sensor for
multi-temporal vegetation research derived from AVIRIS. In Summaries of the Eight
JPL Airborne Earth Science Workshop, 9–11 February, 299–307. Pasadena, CA: JPL
Publication 99-17.
Moshou, D., Bravo, C., West, J., Wahlen, S., McCartney, A., & Ramon, H. (2004). Automatic
detection of‘yellow rust’ in wheat using reflectance measurements and neural
networks. Computers and Electronics in Agriculture, 44, 173–188.
Peñuelas, J., Baret, F., & Filella, I.(1995). Semi-empirical indices to assess
carotenoids/chlorophyll a ratio from leaf spectral reflectance. Photosynthetica, 31,
221–230.
Merzlyak, M. N., Gitelson, A. A., Chivkunova, O. B., & Rakitin, V. Y. (1999). Non-destructive
optical detection of pigment changes during leaf senescence and fruit ripening.
Physiologia Plantarum, 106, 135–141.
Peñuelas, J., Gamon, J. A., Fredeen, A. L., Merino, J., & Field, C. B. (1994). Reflectance indices
associated with physiological changes in nitrogen- and water-limited sunflower
leaves. Remote Sensing of Environment, 48, 135–146.
Crop Disease and Pest Monitoring by Remote Sensing 75

Peñuelas, J., J. Piñol, R. Ogaya, & I. Filella. (1997). Estimation of plant water concentration by
the reflectance water index WI (R900/R970). International Journal of Remote Sensing,
18, 2869–75.
Pu, R., Ge S., Kelly N.M., Gong P. (2003). Spectral absorption features as indicators of water
status in coast live oak (Quercus agrifolia) leaves. International Journal of Remote
Sensing, 24, 1799-1810.
Pu, R., Foschi L., Gong P. (2004). Spectral feature analysis for assessment of water status and
health level in coast live oak (Quercus agrifolia) leaves. International Journal of
Remote Sensing, 25, 4267-4286.
Rouse, J. W., R. H. Haas, J. A. Schell, & D. W. Deering.(1973). Monitoring vegetation systems
in the Great Plains with ERTS. Proc 3rd ERTS Symp 1, 48–62.
Rosenfield G. & Fitzpatrick-Lins K. (1986). A coefficient of agreement as a measure of
thematic classification accuracy. Photogrammetric Engineering and Remote Sensing, 52,
223-227.
Rules for Resistance Evaluation of Wheat to Diseases and Insect Pests Part 7: Rule for
Resistance Evaluation of Wheat to Aphids. NY/T 1443.7-2007
Raikes, C., & L. L. Burpee. (1998). Use of multispectral radiometry for assessment of
Rhizoctonia blight in creeping bentgrass. Phytopathology, 88, 446-449.
South, S., Qi, J. G., & Lusch, D. P. (2004). Optimal classification methods for mapping
agricultural tillage practices. Remote Sensing of Environment, 91, 90–97
Susan E. Halbert, June Connelly B., Bishop G. W., et al. (1992).Transmission of barley yellow
dwarf virus by field collected aphids (Homoptera: Aphididae) and their relative
importance in barley yellow dwarf epidemiology in southwestern Idaho. Annals of
Applied Biology, 121, 105-121.
Strange, R.N., Scott, P.R., (2005). Plant Disease: A Threat to Global Food Security. Annual
review of Phytopathology, 40, 83–116.
Sutton, J.C., Gillespie, T.J., Hildebrand, P.D. (1984). Monitoring weather factors in relation to
plant disease. Plant Disease, 68, 78–84.
Thenkabail, P. S., Smith, R. B., & De Pauw, E. (2000). Hyperspectral vegetation indices and
their relationships with agricultural crop characteristics. Remote Sensing of
Environment, 71, 158–182.
West, J. S., Bravo, C., Oberti, R., Lemaire, D., Moshou, D., & McCartney, H. A. (2003). The
potential of optical canopy measurement for targeted control of field crop disease.
Annual Reviews of Phytopathology, 41, 593–614.
Xiao, Z. Q., Li, Z. M., Fan M., Zhang, Y. Ma, S. j. (2007), Prediction model on stripe rust
influence extent of winter wheat in Longnan Mountain area. Chinese Journal of
Agrometeorology, 28, 350-353.
Yu, B., Ostland, I.M., Gong, P., Pu, R. L. (1999) Penalized discriminant analysis of in situ
hyperspectral data for conifer species recognition. IEEE Transactions on geosciences
and remote sensing, 37, 2569-2577.
Zarco-Tejada, P. J., A. Berjón, R. López-Lozano, J. R. Miller, P. Martín, V. Cachorro, M. R.
González, & A. Frutos. (2005). Assessing vineyard condition with hyperspectral
indices: Leaf and canopy reflectance simulation in a row-structured discontinuous
canopy. Remote Sensing Environment, 99:271–87.
Zeng, S. M. (2003). Simulation study on oversummering process of wheat stripe rust caused
Puccinia striiformis west. In China, Acta Phytopathologica Sinica, 33, 267-278.
76 Remote Sensing – Applications

Zhang, M. H., Qin, Z. H., Liu, X., & Ustin, S. L. (2003). Detection of stress in tomatoes
induced by late blight disease in California, USA, using hyperspectral remote
sensing. International Journal of Applied Earth observation and Geo information, 4, 295–
310.
Zhang, J. C., Huang, W. J., Li, J. Y., Yang, G. J., Luo, J. H., Gu, X. H., & Wang, J. H. (2011).
Development, evaluation and application of a spectral knowledge base to detect
yellow rust in winter wheat. Precision Agriculture, 12, 716-731.
 3

Seasonal Variability of Vegetation

and Its Relationship to Rainfall and
Fire in the Brazilian Tropical Savanna
Jorge Alberto Bustamante, Regina Alvalá and Celso von Randow
Brazilian National Institute for Space Researches
Brazil

1. Introduction
The Brazilian savanna, named locally Cerrado, is the second largest Brazilian biome, covering
approximately two million km2, especially in the Central Highlands (Ratter et al., 1997). This
biome is composed predominantly of tropical savanna vegetation and is considered as one of
the world's biodiversity hotspots, a priority area for biodiversity conservation in the world
(Myers et al., 2000). The Cerrado region is considered the last agricultural frontier in the world
(Borlaug, 2002), which has been converted in the last 50 years especially for agriculture and
pasture purposes, where natural and mainly anthropogenic annual burning is a common
practice. Currently, around 50% of natural vegetation in the Cerrado region has been converted
to pastures and crops (PROBIO-MMA, 2007).This conversion has impacted the biological
diversity, the hydrological cycle, the energy balance, the climate and the carbon dynamics at
local and regional scales due to habitat fragmentation, invasive alien species, soil erosion,
pollution of aquifers, degradation of ecosystems and changes in fire regimes (Klink & Machado,
2005; Aquino & Miranda, 2008). The knowledge of spatial distribution, temporal dynamics and
biophysical characteristics of the vegetation types, are important elements to improve the
understanding of what is the interaction like between vegetation, precipitation and fire.
The objective of this study is to determine the relationship of environmental variables, such
as precipitation and fire, with spatial and temporal distribution patterns of main vegetation
type of the Brazilian tropical savanna. Thus, we seek to answer the question: how
environmental variables, like rain and fire, influence the main vegetation types, like
herbaceous, shrubs, deciduous trees and evergreen trees, in the Cerrado biome taking in
account the seasonal patterns of the variables involved?
In this study, the potential of multi-temporal satellite data, like TRMM data for
precipitation, MODIS vegetation indices products for land cover mapping, and others
sensors like GOES and MODIS for fire detection is explored by the use of remote sensing
and geographic information systems (GIS) techniques.

1.1 Seasonality of Cerrado vegetation

Phenological parameters of vegetation, such as start and end of the growing season, are
strongly influenced by atmospheric conditions (like precipitation, temperature and humidity)
78 Remote Sensing – Applications

at different time scales (intrannual, inter-annual, interdecadal, and so on). Atmospheric

conditions at intrannual scale influence the main phenological events that the plant
experiences during the annual cycle of growth (Reed et al. 1994). At greater time scales, climate
influence on the spatial and temporal distribution of vegetation (Schwartz, 1994). On the other
hand, the vegetation influence atmosphere while maintaining or modifying the flows of matter
and energy, albedo, roughness, CO2, which in turn affect the regional and/or global climate.
Savanna ecosystems that cover approximately 20% of the global land surface have
mechanisms that control the flow of matter and energy in tropical savannas. These
ecosystems are not well understood, which has hindered the inclusion of this biome in
studies of regional and global modeling (Law et al., 2006).

1.2 Climate and precipitation regime

Climate patterns from intra-seasonal to decadal and century scales directly influence the
timing, magnitude (productivity), and spatial patterns of vegetation growth cycles, or
phenology (Reed et al., 1994; Schwartz, 1994).
The Savanna biome has a wet/dry climate. Its Köppen climate group is Aw. The A stands
for a tropical climate, and the w for a dry season in the winter and the rainy season in the
summer. During the dry season of a savanna, most of the plants shrivel up and die. Some
rivers and streams dry up (Parker, 2000; Ritter, 2006). In the wet season all of the plants are
lush and the rivers flow freely. The temperature of the savanna climate ranges from 20° to
30° C. In the winter, it is usually about 20° to 25° C. In summer the temperature ranges from
25° to 30° C. The savanna temperature does not change a lot, although when it does, it is
very gradual and not drastic.
Because of its latitudinal position, the Brazilian savanna region is characterized by the transition
between the warm climates of low latitudes and mesothermal climates of middle latitudes
(Nimer, 1989). This region is considered almost homogeneous on the length and location of the
dry and rainy periods (Rao & Hada, 1990). However, Castro et al. (1994) show that this region
has a certain degree of heterogeneity due to the variation of length in the dry and rainy periods.
This heterogeneity is determined by the interaction of atmospheric circulation systems in the
lower and upper troposphere over the region. Some of these systems are: The South Atlantic
anticyclone also known as South Atlantic Convergence Zone (SACZ), Polar anticyclone and
Chaco low. SACZ is one of the main phenomena that determine the rainfall across the region
(Satyamurty et al., 1998). In general, rainfall in the region ranges from 1000 to 1500 mm.
The climate of the Cerrado is tropical warm and semi-humid, with just two seasons, a dry
one from May to September and a rainy one from October to April. Monthly rainfall in dry
season (that include fall and winter) reduces considerably, reaching zero, resulting in a dry
period that varies from three to five months duration (Coutinho, 2000). The rainy season
(spring and summer) sometimes has short dry periods named locally "veranicos". The mean
annual temperatures vary between 22 and 27°C and the mean annual precipitations between
600 and 2.200 mm.

1.3 Fire regime and detection

Fire is one of the most important drivers that influence vegetation function and structure.
Fire incidence, in a given area or ecosystem, is part of a fire regime which has specific
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 79

patterns of fire occurrences, frequency, size, severity, and sometimes vegetation and fire
effects as well. For example, savanna fires are often of low intensity and high frequency
(often annual), while forest fires are often of low frequency (once every few centuries) and
very high intensity (Bowman & Murphy, 2010). Most of the wildland fires occur by the
combination of edaphic, climatic and human activities (Roy, 2004). Natural fires are generally
started by lightning, with a very small percentage started by spontaneous combustion of dry
fuel such as sawdust and leaves. This kind of fire is insignificants in comparison to number of
fires started by humans (Roy, 2004). Most tropical fires are set intentionally by humans
(Bartlett 1955, 1957, 1961) and are related to several main causative agents (Goldammer,
1988): deforestation activities (conversion of natural vegetation to other land uses, e.g.
agricultural lands pastures, exploitation of other natural resources); traditional, but
expanding slash-and-burn agriculture; grazing land management (fires set by graziers,
mainly in savannas and open forests with distinct grass strata); use of non-wood forest
products (use of fire to facilitate harvest or improve yield of plants, fruits, and other forest
products, predominantly in deciduous and semi-deciduous forests); wildland/residential
interface fires (fires from settlements, e.g. from cooking, torches, camp fires etc.); other
traditional fire uses (in the wake of religious, ethnic and folk traditions; tribal warfare) and
socio-economic and political conflicts over questions of land property and land use rights.
Satellite-borne sensors can detect fires in the visible, thermal and mid-infrared bands. These
sensors have been used most extensively for detecting and monitoring fire activity from
landscape to global scales (Justice et al., 2003; Diaz-Delgado et al., 2004; Allan et al., 2003;
Brandis & Jacobson, 2003; Miller et al. 2003; Rollins et al., 2004; Bowman et al., 2003). Justice
et al. (2003) analyzed global remote sensing data and showed that occurrence of landscape
fire is not random across the world, which is strongly influenced by climatic variables, like
moisture deficit, wind speed, relative humidity and air temperature.

2. Methodology
2.1 Study area
The study area represents almost all (more than 90%) of the Brazilian savanna (Cerrado)
biome, excluding only the southern region, which is characterized by few small isolated
patches of savannas with intense anthropic activities like agriculture and ranching. The
Cerrado vegetation exhibits a wide range of physiognomies. Following the "forest-ecotone-
grassland" concept (Coutinho, 1978), the Cerrado ranges from campo limpo, a grassland,
to cerradão, a tall woodland. The intermediate physiognomies (campo sujo - a shrub
savanna, campo Cerrado - a savanna woodland, and Cerrado sensu stricto - a woodland) are
considered ecotones of the two extremes.
The soil surface dries out during the dry season, leading the herbaceous and sub shrub plants
suffering water stress. Thus, leaves dry out and die, while the underground plant structures
are kept alive. The presence of dead leaves by water stress and also by frost greatly increases
the litterfall and, consequently, the risk of fire (Nimer, 1977; Coutinho, 2000).

2.2 Methodology
The methodology involves the use of two spatial approaches, regional and local, to analyze
the spatio-temporal relationships between environmental variables (precipitation and fire)
and vegetation (NDVI).
80 Remote Sensing – Applications

The analysis unit at the local approach is the point, a specific pixel, which is obtained from
the grid of points that were selected using a stratified random sampling. This grid contains
separately the following types of vegetation: herbaceous, shrubs, deciduous trees, and
evergreen trees of the Brazilian savanna in our study area.
At the regional approach, the entire region is considered another analysis unit, which means
the Cerrado vegetation was not classified into four vegetation types. In this case, we
calculated a NDVI mean value, keeping together all vegetation types (from grassland to
forest) to each 16-days composite of the NDVI time series data.
The procedure applied to the vegetation data is also applied to the precipitation and fire
data. The results are seasonal profiles to each variable along the annual cycle which were
related using correlation and regression techniques. These seasonal profiles allow
calculating a gradient of vegetation seasonality, which is defined by the difference of
highest and lowest values of NDVI, precipitation, or fire. In the case of vegetation, the
degree of seasonality is directly related to the degree of deciduousness, that is, the degree
of leaf biomass loss during the dry season, when most plants suffer some degree of water
stress.
The spatial and temporal resolutions of the data are: 250m and 16-day, 1km and 1-day,
~20km and 3 hours, for MODIS NDVI, fire hotspot and precipitation, respectively. These
data are arranged to standardize them in the same 16-day temporal scale. Data from 2002,
2005 and 2008 were collected since they are considered as years under normal climatic
condition, without the influence of El Niño-Southern Oscillation events.

2.2.1 Vegetation seasonality

The Normalised Difference Vegetation Index (NDVI), normalised ratio between near
infrared reflectance (NIR) and red reflectance (red), has been widely used in satellite-based
vegetation monitoring and modelling. NDVI is computed as:

NDVI = (NIR - red)/(NIR + red) (1)

Index values can range from -1.0 to 1.0, but vegetation values typically range between 0.1
and 0.7. Higher index values are associated with higher levels of healthy vegetation cover,
whereas clouds and snow will cause index values near zero, making it appear that the
vegetation is less green.
Six Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference
Vegetation Index (NDVI) tiles (h12v09, h12v10, h12v11, h13v09, h13v10 e h13v11) were
joined to create a mosaic for the entire study area. Three annual time-series were prepared
for the following years: 2002, 2005 and 2008. Each annual dataset consists of 23 MODIS
NDVI data, at 16-day composite intervals, and 250 m spatial resolution. These data were
used to classify and analyze seasonal and phenology profiles of the Brazilian Cerrado
vegetation.
At this point, the methodology consists first in the classification of vegetation, from the
annual NDVI time-series data and using the tree decision technique, into four types: grasses
and herbs, shrubs, deciduous trees and evergreen trees. This classification uses the
phenological parameter named end of vegetation growing season, which corresponds to the
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 81

following ranges of NDVI for each vegetation type: grasses and herbs (E1) from 100 to174,
shrubs (E2) from 175 to 199, deciduous trees (E3) from 200 to 219 and evergreen trees (E4)
from 220 to 255. Ground truth data was used to validate this classification.
The second part consists of selecting representative spatial points of vegetation types (Figure
1), which are obtained from the vegetation classification image. Each point corresponds to a
pixel on the image and is defined as our unit of analysis. A stratified random sampling
technique was used for the selection of points in the classification image. The number of
points to each vegetation types was proportional to its spatial coverage in the study area. So,
herbaceous (E1) represents 52% of the points, shrubs (E2) 24%, deciduous trees (E3) 15% and
evergreen trees (4) 9%.
The total number of points identified in the study area was N = 639, which are distributed as
follows: 251 points of herbaceous, 318 of shrubs, 59 of deciduous trees and 11 of evergreen
trees (Figure 1).

Fig. 1. Location of Brazilian biomes highlighting the savanna (Cerrado) biome. The shaded
area is a mosaic of 4 MODIS-13Q1 tiles. Dots of different colors correspond to stratified
random sampling of the following vegetation types: herbaceous (E1), shrubs (E2), deciduous
trees (E3), and evergreen trees (E4).
82 Remote Sensing – Applications

2.2.2 Fire seasonality

First, daily data of fire hot spot obtained as latitude/longitude coordinates, in ASCII format,
are converted in XYZ vector format data using a geographical information system (GIS) tool.
These daily vectors were used to create a new vector data set of 16-days composites
accumulating these daily data. Each composite were used to create a raster data set of fire
hotspot density by the use of the Kernel density estimator, according to the following
equation (Silverman, 1986):

 n
f ( x ; H )  n 1  K H ( x  X i ) (2)
i 1

Where:
 X1, X2,…,Xn is sample of n data points (fire hot spot)
 H is bandwith matrix
 KHx - Xi is normal probability density function (pdf) with mean Xi and variance H
Kernel Density calculates the density of point features around each output raster cell. The
kernel function is based on the quadratic kernel function as described in Silverman (1986).
Conceptually, a smoothly curved surface is fitted over each point. The surface value is
highest at the location of the point and diminishes with increasing distance from the point,
reaching zero at the Search radius distance from the point. Only a circular neighborhood is
possible. The volume under the surface equals the Population field value for the point, or 1
if NONE is specified. The density at each output raster cell is calculated by adding the
values of all the kernel surfaces where they overlay the raster cell center.

2.2.3 Precipitation
We used two kinds of data for precipitation in the study area for the years 2002, 2005 and
2008. First, Tropical Rainfall Measuring Mission (TRMM) multisatellite rainfall data (3B42
product), which has 0.25 degree spatial resolutions and 3-hours temporal resolution.
Second, meteorological station rainfall data scattered throughout the study area, which has
1-hour temporal resolution.
These two datasets (TRMM and observed data) are combined following the approach of Vila
et al. (2009), which use the Barnes objective analysis (Barnes, 1973; Koch et al., 1983) for data
interpolation. This analysis allows the incorporation of observed data in a grid of estimated
data and also improves its spatial resolution. As result, the new precipitation data has 0.2-
degree spatial resolution and 1-day temporal resolution.

3. Results and discussion

3.1 Regional analysis
Figure 2 shows seasonal profiles of vegetation and precipitation in the Cerrado region for
the three years analyzed (2002, 2005 and 2008). These results show that the Cerrado
vegetation seasonality is well defined, which in turn has a direct relationship to the
seasonality of precipitation. However, there is a time lag ranging from 1 (16 days) to 3 (48
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 83

Fig. 2. Annual seasonality of vegetation and precipitation in the years 2002, 2005 and 2008
for the Cerrado biome. Each year consists of 23 16-days composite periods. Precipitation is
the daily mean rainfall values for a 16-days composite period in mm (first y axis) and
vegetation the mean NDVI values for the same period (second y axis).

days) periods between the beginning of the rainy season and the beginning of the vegetation
growing season.
Figure 3 shows seasonal profiles of vegetation and fire in the Cerrado region for the three
years analyzed (2002, 2005 and 2008). These results show, as in Figure 2, that the fire
occurrence in the Cerrado has well-defined seasonality, which in turn has a direct negative
relationship to the seasonality of vegetation. That means, the highest fire occurrence during
the growing cycle of fire is related to the greatest loss of plant cover during the dry season,
with a time lag ranging from 0 to 3 periods (0 to 48 days).

Fig. 3. Annual seasonality of vegetation and fire in the years 2002, 2005 and 2008 for the
Cerrado biome. Each year consists of 23 16-days composite periods. Fire is the daily mean
value of the density of hotspot within a 10km radius for a 16-days composite period and
vegetation the mean NDVI values for the same period (second y axis).
84 Remote Sensing – Applications

3.2 Local analysis

The results presented show the seasonality of vegetation, rainfall and fire in places (points)
defined by the grid points representing the four vegetation types analyzed in the study.

3.2.1 Seasonality of vegetation

Figure 4 shows the seasonal profile of four vegetation types over the three years analyzed.
These results show a clear difference, regarding the degree of vegetation seasonality, among
the four types of vegetation analyzed, according to the following gradient: herbaceous (E1),
with strong seasonality, shrubs (E2), deciduous trees (E3), and evergreen trees (E4), with
weak seasonality.
The vegetation phenology metrics are shown in Figure 5. Figure 5a shows annual maximum and
minimum NDVI values, indicating the highest and lowest vegetation productivity respectively,
for each type of vegetation in the three years analyzed. Figure 5b shows the difference between
the maximum and minimum NDVI as a percentage, indicating the degree of seasonality. Also
Figure 5a shows a slight difference between the maximum NDVI values, high plant productivity
in the four vegetation types, while the difference between the minimum NDVI values, lower
productivity, in the four vegetation types is significant. In general, the degree of seasonality of the
vegetation (Figure 5b) was consistently detected in the four vegetation types. That is, small plants
with low canopy (shrubs and herbaceous) have higher degree of seasonality than tall one with
high canopy, which in turn have lower degree of seasonality.

3.2.2 Seasonality of precipitation

Figure 6 shows the seasonal profile of rainfall recorded in the same sampling points of the
four vegetation types over the three years analyzed. These results show, in the beginning of
the year during the rainy season, lower rainfall at sites where herbaceous and shrubs were
registered than at sites where deciduous and evergreen trees are predominant. This result is
a first indicator that shows a relationship between rainfall gradient and vegetation cover
gradient. These gradients range from sites with higher precipitation, associated with high
canopy plants (evergreen trees), to those with less precipitation, associated with a lower
canopy plants (herbaceous).

3.2.3 Seasonality of fire

Figure 7 shows the pattern of the fire season recorded in the same sampling points of the
four vegetation types over the three years analyzed. The results of fire occurrence
throughout the Cerrado region show that there is a pronounced seasonality in all vegetation
types analyzed with a peak in the months of greatest drought in the dry season.
The results show a well-defined gradient of fires in the four types of vegetation. This
gradient varies from lower fire density in evergreen trees (E4), with shorter periods of time
(12 to 22) throughout the annual cycle, to higher fire density in herbaceous plants (E1), with
more periods of time (1 to 23, except 2), as seen in Figure 7. Most of the fire occurrences in
the four types of vegetation were recorded in 2005 and 2008 indicating the occurrence of an
inter-annual variability of fire. The higher fires were recorded between the periods from 15
to 21 taking into account the four vegetation types and the three years analyzed.
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 85
86 Remote Sensing – Applications

E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees.
Fig. 4. Annual seasonality of vegetation derived from NDVI data for the years 2002, 2005
and 2008 in the four vegetation type analyzed. Each year consists of 23 16-days composite
periods.
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 87

E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees
Fig. 5. Metrics of vegetation phenology derived from NDVI data used in Fig. 4. Maximum
and minimum NDVI values indicate periods of higher and lower plant productivity (left)
respectively, and the difference of both, as a percentage, indicates the degree of seasonality
of each vegetational type in the three years analyzed.
88 Remote Sensing – Applications
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 89

E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees.
Fig. 6. Annual seasonality of precipitation for the years 2002, 2005 and 2008, in places where
we sampled the four vegetation types analyzed. Each year consists of 23 16-days composite
periods. Precipitation in mm is the daily mean rainfall values for a 16-days composite
period.
90 Remote Sensing – Applications
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 91

E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees.
Fig. 7. Annual seasonality of fire for the years 2002, 2005 and 2008, in places where we
sampled the four vegetation types analyzed. Each year consists of 23 16-days composite
periods. Fire is the daily mean value of the density of hotspot within a 10km radius for a 16-
days composite period.
92 Remote Sensing – Applications

3.2.4 Relationship between vegetation (NDVI) and environmental variables

(precipitation and fire)
The results showed in Table 1 indicate significant positive correlation between NDVI and
precipitation in herbaceous, shrubs and deciduous trees, and negative correlation between
NDVI and fire in the same three vegetation types. In the case of evergreen trees, the
correlation between NDVI and precipitation is positive but not significant, and between
NDVI and fire is negative, but also not significant. These results are corroborated in
subsequent analysis.

E1 E2 E3 E4 E1 E2 E3 E4
Prec Prec Prec Prec Fire Fire Fire Fire

E1-NDVI 0.60 0.58 0.67 0.70 -0.69 -0.68 -0.66 -0.41

E2-NDVI 0.52 0.51 0.60 0.65 -0.68 -0.72 -0.70 -0.43

E3-NDVI 0.20 0.19 0.31 0.36 -0.61 -0.73 -0.75 -0.51

E4-NDVI 0.00 -0.01 0.16 0.09 -0.43 -0.57 -0.66 -0.47

E1-Fire -0.40 -0.39 -0.43 -0.49 1.00 0.77 0.65 0.35

E2-Fire -0.31 -0.30 -0.34 -0.42 0.77 1.00 0.65 0.41

E3-Fire -0.25 -0.24 -0.31 -0.38 0.65 0.65 1.00 0.47

E4-Fire -0.09 -0.08 -0.19 -0.20 0.35 0.41 0.47 1.00

Table 1. Correlation matrix of vegetation, rainfall and fire variables, highlighting the
significant correlations between the following couple of variables: NDVI and rainfall, NDVI
and fire, and rainfall and fire; which taking into account the four types of vegetation
analyzed (E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees).

Figure 8 shows the result of the linear regression analysis between vegetation and
precipitation for each vegetation type. Each line in this figure with a specific color shows the
degree of fit between the points distributed for both variables by type of vegetation.
Although this degree of fit between both variables is low, the results indicate that there is a
gradient of fit between precipitation and vegetation, here named as precipitation gradient,
which ranges from high to low coefficient of correlation (R2) following the sequence:
herbaceous-E1 (high R2), shrubs-E2, deciduous trees-E3 and evergreen trees-E4 (low R2).
Thus, as the R2 value increases the influence of precipitation on vegetation increases, so
herbaceous is more dependent on rainfall, in the annual cycle, than the other types of
vegetation analyzed. That means, herbaceous are strongly dependent on rainfall in order to
increase its vegetation cover. In the dry season, these kinds of species lose their leaves or
even die.
At the opposite end of the precipitation gradient, where the evergreen trees-E4 are
positioned, precipitation has weak influence on the vegetation cover, which means that in
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 93

E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees.
Fig. 8. Regression of precipitation (independent variable) and NDVI (dependent) for each
vegetation type analyzed. NDVI values range from 0 to 255 (x-axis). Precipitation in mm is
the daily mean rainfall values for a 16-days composite period (y-axis). N = 69.

the dry season, evergreen trees are able to capture water from the vicinity of river courses,
as occurs in gallery forests, or from deep soil, where the length of tree roots reach deep and
moist soil layers, allowing these trees to replace their leaves throughout the year, which
gives them their evergreen nature.
An analyses of variance (ANOVA) performed to evaluate these regressions is shown in table
2. Results indicate that, except for the regression between NDVI and Precipitation for the
evergreen trees (E4) class, all regressions are significant at the 0.99 confidence level.
Moreover, the relationships between NDVI and Fire were significant for all classes.

NDVI x Prec NDVI x Fire

R² F p R² F p

E1 0.3558 37.01 <0.01 0.4795 61.72 <0.01

E2 0.2608 23.64 <0.01 0.5172 71.77 <0.01

E3 0.0959 7.11 <0.01 0.5575 84.40 <0.01

E4 0.0072 0.49 0.49 0.2218 19.10 <0.01

Table 2. Analysis of Variance (ANOVA) of linear regression NDVI x Precipitation (Prec) and
NDVI x Fire. Bold values indicate the case where regression was not significant.
94 Remote Sensing – Applications

Figure 9 shows the result of the linear regression analysis between vegetation and fire for
each vegetation type. Each line in this figure with a specific color shows the degree of fit
between the points distributed for both variables by type of vegetation. The results indicate
that there is a gradient of fit between fire and vegetation, here named as fire gradient, which
ranges from high to low coefficient of correlation (R2) following the sequence: deciduous
trees-E3 (high R2), shrubs-E2, herbaceous-E1, and evergreen trees-E4 (low R2).

E1: herbaceous; E2: shrubs; E3: deciduous trees; and E4: evergreen trees.
Fig. 9. Regression of fire (independent variable) and NDVI (dependent) for each vegetation
type analyzed. NDVI values range from 0 to 255 (x-axis). Fire is the daily mean value of the
density of hotspot within a 10km radius for a 16-days composite period. N = 69.

The fire gradient identified above indicates that there is direct relationship between NDVI of
the main vegetation types (herbaceous, shrubs and deciduous trees), which make up the
Cerrado vegetation, and fire, indicating the role of fire in the maintenance of these
vegetation types.
Fire occurs with greater intensity at the end of dry season. First of all, fire consumes part of
the burk and organic matter of the plant, after the first rains, in the beginning of the rainy
season, these partially burned plant sprouts new shoots with greater vigor.
At the opposite end of the fire gradient, where the evergreen trees-E4 are positioned, the fire
occurs in lower proportion in these trees, however, unlike what happens with other types of
vegetation, the effect of fire is pernicious, it can damage or even eliminate some species in
this vegetation type according to the intensity level.
The multiple regression analysis indicates that there is a direct relationship between
precipitation and fire, and vegetation index (NDVI) in the four vegetation types of the
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 95

savanna vegetation. The multiple coefficients of determinations (R2) show that the
environmental variables as a whole (precipitation an fire) follow a gradient of high influence
in vegetation types with low vegetation cover (herbaceous R2=0.67 and shrubs R2=0.65) to
low influence in that with high vegetation cover (deciduous trees R2= 0.55 and evergreen
trees R2=0.27). Results from the ANOVA of the multiple regression presented in Table 3
indicate that, when the analysis is performed considering both independent variables, the
multiple regression gives statistically significant parameters, for all classes of vegetation.
However, an univariate test of significance performed for each independent variable show
that precipitation alone is not significant correlated to the vegetation index for both tree
classes (E3 and E4).

Whole model R Univariate test of significance

R² F p F_prec F_fire p_prec p_fire

E1 0.6001 49.52 <0.01 19.90 40.31 <0.01 <0.01

E2 0.6141 52.52 <0.01 16.59 60.43 <0.01 <0.01

E3 0.5648 42.83 <0.01 1.12 71.11 0.29 <0.01

E4 0.2220 9.42 <0.01 0.01 18.22 0.92 <0.01

Table 3. Analysis of Variance (ANOVA) of the multiple regression between NDVI

(dependent variable) and precipitation and fire (independent variables). Bold values
indicate the cases where regression was not significant.

4. Conclusions
The response of vegetation NDVI is more related to the variation of fire than to variations in
precipitation in Cerrado region. Vegetation NDVI responds to variation of precipitation
with a time lag ranging from 16 to 48 days, while vegetation NDVI responds to variation of
fire with a time lag ranging from 0 to 48 days.
The relationship between vegetation types, derived from NDVI, and precipitation, derived
from TRMM, shows a gradient of positive correlations in vegetation types with low
vegetation cover, herbaceous (r= 0.60) and shrubs (r= 0.51), to very little or none with high
vegetation cover, deciduous trees (r= 0.31) and evergreen trees (r= 0.09). On the other hand,
the relationship between vegetation and fire hotspot shows a gradient of negative
correlation, which is stronger in herbaceous (r= 0.72), shrubs (r= 0.74) and deciduous trees
(r= -0.73) than in evergreen trees (r= -0.52).
Our analyses show that vegetation cover increases are related to increases in precipitation
and decreased in density of fire hotspots. We also found high density of fire hotspot in the
dry season in deciduous trees, shrubs and herbaceous which suggesting the high removal of
CO2 (greenhouse gas) of the land cover to the atmosphere somehow influencing the
dynamic equilibrium of this (atmosphere) in the region of the Brazilian tropical savanna.
96 Remote Sensing – Applications

5. References
Allan, G.; Johnson, A.; Cridland, S. & Fitzgerald, N. (2003). Application of NDVI for
predicting fuel curing at landscape scales in northern Australia: can remotely
sensed data help schedule fire management operations? International Journal of
Wildland Fire 12, 299–308.
Aquino, F.G.; Miranda, G.H.B. (2008). Consequencias ambientais da fragmentação de
habitats no cerrado. In Sano, S.M.; Almeida, S.P.; Ribeiro, J.F. (Eds.). Cerrado:
ecologia e flora. Brasília: Editora Embrapa, p.385-398.
Barnes, S. L. (1973). Mesoscale objective analysis using weighted time-series observations,
NOAA Tech. Memo. ERL NSSL-62 National Severe Storms Laboratory, Norman,
OK 73069, 60 pp. [NTIS COM-73-10781.], 1973. 2359
Bartlett, H.H. (1955, 1957, 1961). Fire in relation to primitive agriculture and grazing in the
tropics: annotated bibliography, Vol. 1-3. Mimeo. Publ. Univ. Michigan Bot.
Gardens, Ann Arbor, USA.
Borlaug, N.E. (2002). Feeding a world of 10 billion people: the miracle ahead. In: R. Bailey
(ed.). Global warming and other eco-myths. pp. 29-60. Competitive Enterprise Institute,
Roseville, EUA.
Bowman, D.M.J.S. & Murphy, B.P. (2010). Fire and Biodiversity. In Sodhi, N.S.; Ehrlich, PR.,
eds 2010. Conservation Biology for all , pp. 163-180.
Bowman, D.M.J.S., Zhang, Y., Walsh, A. & Williams, R.J. (2003). Experimental comparison of
four remote sensing techniques to map tropical savanna fire-scars using Landsat-
TM imagery. International Journal of Wildland Fire 12, 341–48.
Brandis, K. & Jacobson, C. (2003). Estimation of vegetative fuel loads using LandsatTM
imagery in New South Wales, Australia. International Journal of Wildland Fire, Vol.
12, 185–94.
Castro, L.H.R.; Moreira, A.M. & Assad, E.D. (1994). Definição e regionalização dos
padrões pluviométricos dos Cerrados brasileiros. In: ASSAD, E.D. Chuvas nos
Cerrados: análise e espacialização. Brasília, Embrapa-CPAC/Embrapa-SPI, 1994.
423p.
Coutinho, L.M. (1978). O conceito de cerrado. Revista Brasileira de Botânica 1(1):17-23.
Coutinho, L. M. (2000). O bioma do cerrado, in: Klein, A. L. Eugen Warming e o cerrado
brasileiro: um século depois. UNESP, 156 p.
Diaz-Delgado, R.; Lloret, F. & Pons, X. (2004). Statistical analysis of fire frequency models for
Catalonia (NE Spain, 1975–1998) based on fire scar maps from Landsat MSS data.
International Journal of Wildland Fire 13, 89–99.
Goldammer, J.G. (1988). Rural land use and fires in the tropics. Agroforestry Systems, Vol.6:
235-253
Justice, C.O.; Smith, R.; Gill, A.M. & Csiszar, I. (2003). A review of current space-based fire
monitoring in Australia and the GOFC/GOLD program for international
coordination. International Journal of Wildland Fire, 12, 247–258.
Klink, C.A. & Machado, R.B. (2005). A conservação do cerrado brasileiro. Megadiversidade,
Vol.1: 147-155.
Seasonal Variability of Vegetation and Its
Relationship to Rainfall and Fire in the Brazilian Tropical Savanna 97

Koch S. E.; desJardins, M. & Kocin, P. J. (1983). An interactive Barnes objective map analysis
scheme for use with satellite and conventional data. J. Climate Appl. Meteor., Vol.22,
1487–1503.
Law, R. M.; Kowlaczyk, E. A. & Wang, Y.P. (2006). Using atmospheric CO2 data to assess
a simplified carbon-climate simulation for the 20th century. Tellus, Vol.58B, 427-
437.
Miller, J.D.; Danzer, S.R.; Watts, J.M.; Stone, S. & Yool, S.R. (2003). Cluster analysis of
structural stage classes to map wildland fuels in a Madrean ecosystem. Journal of
Environmental Management, Vol. 68, 239–52.
Myers, N. et al. 2000. Biodiversity hotspots for conservation priorities. Nature 403: 853-858.
Nimer, E. (1989). Climatologia do Brasil. Secretaria de Planejamento e Coordenacao da
Presidencia da Republica e IBGE, Rio de Janeiro, 421p.
Nimer, E. 1977. Clima. Pp 47-48 in: IBGE. Geografia do Brasil: região Nordeste. Rio de Janeiro.
Parker, A. (2000). Wet-Dry Tropical Climate (Aw), In: Climate Savanna, 01.06.2000,
Available from
http://www.blueplanetbiomes.org/savanna_climate_page.htm
PROBIO-MMA (Projeto de Conservação e Utilização Sustentável da Diversidade Biológica
Brasileira) do Ministério do Meio Ambiente. (2007). Mapeamento de Cobertura Vegetal
do Bioma Cerrado. Technical Report. Brasilia, Brasil.
Rao, V.B. & Hada, K. (1990). Characteristics of rainfall over Brazil: annual variations and
connections with the southern oscillations. Theoretical and Applied Climatology, Vol.2,
p.81-91.
Ratter, J.A.; Ribeiro, J.F. & Bridgewater, S. (1997). The brazilian Cerrado vegetation and
threats to its biodiversity. Annals of Botany, Vol.80: 223-230.
Reed, B.C.; Brown, J.F.; Vanderzee, D.; Loveland, T.R.; Merchant, J.W. & Ohlen, D.O. (1994).
Measuring phenological variability from satellite imagery. J. Veg. Sci. Vol.5, 703-
714.
Ritter, M.E. (2006). The Physical Environment: an Introduction to Physical Geography. Available
from
http://www.uwsp.edu/geo/faculty/ritter/geog101/textbook/ title_page. html
Rollins, M.G.; Keane, R.E. & Parsons, R.A. (2004). Mapping fuels and fire regimes using
remote sensing, ecosystem simulation, and gradient modeling. Ecological
Applications, Vol.14, 75–95.
Roy, P.S. (2004). Forest fire and degradation assessment using satellite remote sensing and
geographic information system. In: Sivakumar, M.V.K.; Roy, P.S.; Harmesen, K.;
Saha, S.K., eds, 2010. Satellite Remote Sensing and GIS Applications in Agricultural
Meteorology pp. 361-400.
Satyamurty, P.; Nobre, C. A. & Silva Dias, P. L. (1998). South America. Meteorological
Monographs, Vol.27, n. 49 (Southern Hemisphere Meteorology, cap. 3C), p. 119-
139.
Silverman, B.W. (1986). Density estimation for statistics and data analysis. London; New
York: Chapman & Hall, 1986. 175 p.
Schwartz, M.D. (1994). Monitoring global change with phenology: the case of the spring
green wave. Int. J. Biometeorol. Vol.38, 18-22.
98 Remote Sensing – Applications

Vila, D.A.; de Goncalves, L.G.G.; Toll, D.L. & Rozante, J.R. (2009). Statistical Evaluation of
Combined Daily Gauge Observations and Rainfall Satellite Estimates over
Continental South America. J. Hydrometeor. Vol.10, 533–543.
 4

Land Cover Change Detection in

Southern Brazil Through Orbital
Imagery Classification Methods
José Maria Filippini Alba, Victor Faria Schroder
and Mauro Ricardo R. Nóbrega
Brazilian Agricultural Reasearch Corporation (Embrapa)
Brazil

1. Introduction
Remote sensing has been considered a promising technology as support for agriculture since
its beginnings, due to its contribution for a climatic perspective or for understanding of
processes related to land. However, significant applications occurred only in the late twentieth
century, as result of the creation of best orbital systems, with higher spatial resolution, more
bands and stereoscopic capture. Several orbital platforms, as AQUA/TERRA, Quickbird and
Ikonos are examples in that sense (Moreira, 2005; Embrapa, 2009).
Engineering innovations, new sensors and methods of digital image processing must be
performed simultaneously so that the advances in remote sensing will be achieved.
Anyway, the incorporation of orbital images on geographic information systems (GIS) and
their post-processing appear as significant application since a daily life perspective,
specially when classification methods are involved, because of their relation to land use,
land cover and easy interpretation.
This chapter considers classification methods applied on orbital imagery in Southern Brazil,
in the coastal plain of Rio Grande do Sul state (Fig. 1), where a sequence of lagoons and
lakes of different sizes occurs in the context of subtropical to temperate climate with cold
winters and hot summers, being organized according to the following four sections:
 About classification methods.
 Evaluation of rice planting area in the vicinity of Caiuba lagoon (1981 – 2009).
 Analysis of land cover evolution in the municipality of Montenegro (1993 – 2008).
 Comparison and evaluation of errors.
All the exposed data are related to research projects of the Embrapa Temperate Climate
Research Center, Pelotas, Rio Grande do Sul state, one of the 45 research centers of the
Brazilian Agricultural Research Corporation (Embrapa) spread on the national territory.

2. About classification methods

Classification methods were created in the statistical context, when a collection of objects or
samples could be characterized and separated in different classes (Davis, 1986). The method
100 Remote Sensing – Applications

Fig. 1. Location of the areas of study in the context of South America, Brazil and the State of
Rio Grande do Sul. Montenegro municipality as gray levels composite of the bands 1, 4 and
5 TM/Landsat 5 (Miranda, 2002, 2005) and Caiuba lagoon in gray levels composite of the
bands 3, 2 and 1 ASTER (NASA, 2004) on september 26th, 2000.

was extended for processing of digital images considering the pixels as objects to be
classified (Crosta, 1993, Lillesand & Kiefer, 1994, Jensen, 1996). The application of the
classification methods on satelite imagery is affected by two main factors:
i. Intervention of user.
ii. Criteria of definition for the groups.
The “supervised classification” is included in topic (i), when the user defines the groups
through digitalization of uniform spectral answer. Statistics are calculated for each group, so
the classification is performed for all the image (pixel by pixel). The unsupervised
classification eliminates the intervention of the user; then, the software defines the groups
by means of scattergrams “band versus band”, where isolines related to distribution density
of the pixels are analysed (Crosta, 1993).
Different options for case (ii) are posible, by instance, criteria can use the standard deviation
for the parallelepiped method or a minimum distance when an eliptical form is defined for
each group or a combination of the later with statistical probability, that is, the maximum
likelihood method. Jensen (1996) presented other criteria of classification, as Isodata method
and the Fuzzy method.
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 101

Spectrometric methods measure the response of target materials in the laboratory or field.
Then the spectral patterns are simulated for a specific sensor through a specialized software,
so that a sequence of orbital images is classified according to the pattern generated by that
software (Lillesand & Kiefer, 1994; Pontara, 1998; Moreira, 2005).
Classification of remote sensing images appear as useful tool in terms of land use, whether
in local scale or in regional scale. Filippini-Alba and Siqueira (1999) classified land use in the
municipality of Pelotas, Rio Grande do Sul state, Brazil, according to nine classes:
agriculture, clay soils, forestry, natural forest, pastures, soil without vegetal cover, urban,
water and wetlands. Natural forest and pastures occupied 23% and 30% of the territory
respectively, with intense interference between the classes “urban” and “soil without
vegetal cover”. Similar classes were considered by Bolfe et al. (2009) for land use in Rio
Grande do Sul state, but with different results. Agriculture and pastures occurred 32% and
50% of the territory respectively with only 3% for natural forest. The differences between
both studies are easy explained in term of scale, because in the two occasions Landsat
images were used and a municipality was considered at the former and a state at the latter,
with territory difference of 1 to 155 times in size.
Lu et al. (2004) discriminated seven categories of change detection techniques: (i) algebra; (ii)
transformation; (iii) classification; (iv) advanced models; (v) Strategies with geographic
information systems (GIS); (vi) visual analysis and (vii) miscellanea. The classification
methods are detached, with six different modalities. One of them, the “Post-classification
comparison”, is predominantly used in this chapter. That is, multi-temporal images are
classified separately into thematic maps, then the classified images are compared pixel by
pixel. The “Post-classification comparison” minimizes the atmospheric impacts, the
environmental differences among multi-temporal imagery, as well as differences related to
the sensor kind, providing a complete matrix of change information. However, some
disadvantages can be appointed, because a great amount of time and expertise is required
and, by other side, the final accuracy depends on the quality of the classified image due to
the weather condition on that date.
Guild et al. (2004) quantified the areas of deforestation in the Amazonian forest, state of
Rondonia, Brazil. The tasselled cup transformation (Crist & Kauth, 1986) was applied with
the Landsat imagery from the years 1984, 1986 and 1992. The variables brightness, greenness
and wetness were evaluated for each year, then, a file integrated the nine levels of
information (three variables by three years). These data were processed through principal
components and classification methods with overall accuracy of 79.3, 68.4% and 71.4%, for
tasselled cap land cover change classification, tasselled cap with principal components land
cover change classification and tasselled cap image differencing, respectivelly. Final classes
were a combination between land cover and time, so change detection was quantified.
The two applications present in this chapter consider the Supervised classification method
with maximum likelihood as criteria for definition of the classes. The proximity of the study
areas and knowledge of the territory justify this option to take advantage of available
information. Unsupervised classification is a fast proccess, good for unknown or outlying
areas, when truth of field is unavailable and most time-consuming after processing, due to
the need of class identification. Maximum likelihood criteira is restricted by software and
time-consuming but it represents a improvement in relation simple criteria as the
parallelepiped or the minimum distance.
102 Remote Sensing – Applications

Acording to Lu et al. (2004), methods (iii) and (v) were considered in this chapter.
Classification (iii) was applied in both conditions, Caiuba lagoon and Montenegro
municipality. The extraction of the poligon corresponding to the “potential area for
agriculture” in the vicinity of Caiuba lagoon represents a tipical strategy of GIS (v).
Softwares of digital images processing and GIS are very similar. Both can execute multilayer
processing, including raster/vector files and logic/mathematical algorithms, but digital
images processing is more specific for raster format and GIS for vector format.

3. Evaluation of rice planting area in the vicinity of Caiuba lagoon (1981-2009)

The Caiuba lagoon is part of the litoral lacunar complex of Rio Grande do Sul state,
southern Brazil and it extends by 3300 hectares, in the municipality of Rio Grande, 15
kilometers to north extreme from the Taim Ecological Reserve and 45 kilometers to south
from Patos lagoon (Fig. 1). This significant source of water is used mainly for irrigation of
rice, specially when the Merin lagoon is further. Accordingly, the Foundation for Research
Support in State of Rio Grande do Sul (FAPERGS) funded a research project leadered by
the Federal University of Rio Grande (FURG) attempting to study the sustainability of the
productive system, as well as the effects on local biodiversity. The Embrapa Temperate
Climate Research Center collaborated to the evaluation of the agricultural area in the
period 1973 to 2009 by satellite images. Imagery of Landsat satellite of different years was
considered for similar times (Table 1), for the scenes corresponding to orbit 237 points 82
and 83 of the worldwide reference system 1 (MSS sensor) and for the scenes
corresponding to orbit 221 points 82 and 83 of the worldwide reference system 2 (TM
sensor). Thus, the atmospheric conditions were more or less equivalent, deriving in
comparable image quality. Each image was evaluated for the various land uses and the
areas occupied for the different classes were calculated in order to study the historical
evolution of the process during the above period.
The first satellite of Landsat series was launched in 1972 with the multispectral scanner
(MSS), with four bands in visible - near infrared and one in thermal infrared and

Sensor Date Range of wavelengh and IFOV

MSS Sep. 6th, 1973 500 – 600nm, 600 – 700nm and 700 - 800nm, IFOV = 79m
MSS Mar. 13th, 1981
TM Jan. 22nd, 1991 630 – 690nm, 760 – 900nm and 1550 – 1750nm, IFOV = 30m
TM Dec. 21st, 1996
TM Dec. 19th, 2001
TM Jan. 20th, 2002
TM Jan. 28th, 2005
TM Jan. 2nd, 2007
TM Jan. 7th, 2009
Table 1. Description of basical parameters of the images of Landsat series used for
evaluation of the planting area of rice in the vicinity of tha Caiuba lagoon.
IFOV = instantaneous field of view. Source: INPE, 2010b.
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 103

instantaneous field of vision (IFOV) of 79 meters and 240 meters respectively. Improvements
of the system included more bands (short-medium infrared) and reduction of IFOV to 30
meters and 120 meters respectively, for the thematic mapper (TM) in 1982 (Jensen, 1996). A
panchromatic band was developed for the Landsat 7 satellite, with the TM plus sensor, but,
the series reached to the end. The Landsat 5 satellite was an engineering success, the
platform was launched in 1984 and is still on orbit. Anyway, the Landsat series represents
the greatest collection of terrestrial images for environmental applications, specially, since a
historical point of view.
Composites of three bands were used, with green band (500 – 600 nm), red band (600 – 700
nm) and near infarred band (700 - 800 nm) for the MSS sensor and the red band (630 – 690
nm), the nearinfarred band (760 – 900 nm) and the shortmedium infrarred band (1550 – 1750
nm) for TM sensor. These games of bands are not equivalent, then similar patterns of colour
were adjusted by visual observation.
Digital imagery was registered for the Universal Transverse of Mercator projection(UTM),
zone 22 South with the datum WGS84, after that, a mosaic of pairs of scenes was composed,
by instance, scene 237/82 and 237/83 for MSS sensor. So the mosaic was cutted evolving the
study area and a file with the mentioned three bands was created for each date. Initially,
data were processed by the supervised classification according to the maximum likelyhood
criteria. Eigth poligons of homogeneous features were digitalized with the software ER-
Mapper (1995), deriving in the test areas, then each pixel of the corresponding image was
classified according to its similarity with the parameters of each test area (beach/dunes,
forestry, rice crops, pastures, sandy fields, soil without vegetal cover, water and wetlands).
A second strategy was developed to improve results, so the "potential area for agriculture",
that is rice crops, pastures and soil without vegetal cover, was isolated and classified by
similar way.
Results of the preliminar process of classification considered a rectangle of 30 km wide and
65 km long for the images of 2001, 2002 and 2005 (Fig. 2). The "potential zone for
agriculture" is represented by a “central zone” in direction south - north to the East of Merin
Lagoon, where agricultural areas are discriminated. A confusion between rice crop class and
wetlands class is observed in the west - north sector of the study area. Sandy fields are long
structures related to old movements of the sea (Atlantic Ocean), where a low charge of
livestock is a common use and forestry is developed eventually, as observed in the images.
The area occupied by water bodies was almost constant, that is 19 - 21% (Table 2), but, the
wetlands were reduced in area in 2005, a year of drought probably, then, there was an
increment in the area occupied by the class "Soil without vegetal cover" and a reduction of
the area occupied by the class “Pastures”.
When the "potential zone for agriculture" was isolated, the precison of evaluation of the area
occupied by pastures, rice crops and soil without vegetal cover (SWVC) was improved. The
kind of sensor, the date of the image and the meteorological conditions induced diferences
among the imagery of different dates (Fig. 3). The images of 1973 and 1981 present a
different characteristics due to captation with the MSS sensor. The first image corresponds
to september, when the culture had not been implanted yet. Some agricultural areas showed
different pattern in 2001 and 2002 (same harvest) related to waterlogged soils, probably, due
to intense rain in that time. A differencial answer of the vegetation in the agricultural areas
was observed since 2005, what suggests a evolution of the vegetal development of the rice
104 Remote Sensing – Applications

Fig. 2. Preliminary classification in the Caiuba region.

Image date Dec. 19th, 2001 Jan. 20th, 2002 Jan. 28th, 2005
Water, % 19.2 21.3 20.0
Wetland, % 17.8 18.5 11.3
Pastures, % 20.2 12.4 8.9
Rice crops, % 10.2 9.1 10.4
Sandy fields, % 17.7 13.6 15.1
Beach/dunes, % 6.7 6.3 10.5
Florestry, % 4.3 3.6 4.6
Clouds, % 1.5
SWVC,% 3.9 13.8 19.3
Total area, ha 202,777 204,088 208,907
Table 2. Preliminary areas of land cover calculated by classification methods in the vicinity
of Caiubá region through Landsat-TM images in the period 2001 – 2005. SWVC = Soil
without vegetal cover.

varieties or, perhaps, the introduction of a new crop. All the images show a intense rotation
among pastures, rice and fallow lands, what lets a reduction of inputs, rest of the soil and
improvement of productivity.
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 105

Fig. 3. Images Landsat corresponding to the “Potential zone for agriculture” for different
dates.
106 Remote Sensing – Applications

The poligonal area was classified according to five classes: (1) Undefined; (2) Pastures; (3)
Rice crops; (4) Soil without vegetal cover; (5) Water. The class “Undefined” represents rice
crops or pastures depending on the year, thus it was incorporated to class “Pastures” in
1973, 1981 e 2001 and to class “Rice crops” in 1991, 1996, 2002, 2005, 2007 e 2009, accordingly
the interpretation of the images. So, the classes “Pastures”, “Rice crops”, “SWVC” and
“Water” were evaluated for occupied area (Fig. 4; Table 3).
The occurrence of water is almost insignificant inside the “potencial zone for agriculture”,
because the irrigation is performed through the water of the lagoons Caiuba and Mirim. The
area occupied by the class “Rice crops” seems to depend on the vegetal developping, with
restricted values when months previous to January are evolved. This fact was checked with
the images of 2001 and 2002, corresponding to the same harvest, Dezember and January
respectively.
By this reason, only the data corresponding to the months of january and march, when the
vegetal developping of rice is reached, were consider in the graphic of “occupied area” as a
function of time (Fig. 5).

Year Water Pastures Rice crops SWVC Total

06/09/1973 173 19057 12856 13565 45652
13/03/1981 85 28593 13915 6592 49185
22/01/1991 52 17110 18751 15886 51798
21/12/1996 825 22534 12042 22534 57935
19/12/2001 56 30523 14404 8516 53498
20/01/2002 120 17579 20090 16299 54087
28/01/2005 49 13144 21963 20246 55402
02/01/2007 57 25062 21054 9467 55640
07/01/2009 80 5302 21029 31124 57535
Table 3. Area evaluation of land cover classes for the "potential zone for agriculture". The
class "Undefined" was incorporated to the class "Pastures" or the class "Rice crops" according
to the year. SWVC = Soils without vegetal cover. Data in hectares.

The area of the "Potential zone for agriculture" was delimited by digitalization, but a soft
and constant increment is evident during the period 1981 to 2009. By other side, the area
occupied by the class "Rice crops" was evaluated by classification methods; after a period of
increment, the class reached a maximum in 2005 with 22 thousand hectares, then there was a
stabilization in 2007 - 2009 with about 21 thousand hectares. The classes "Pastures" and
"SWVC" showed oscillation in complementary way, because the sum of both classes was
almost constant. As classes "Potential zone for agriculture" and "Rice crops" presented linear
behavior in the graphic Area against time, thus, linear regression models were adjusted
(Table 4).
The parameter R2 is the correlation coefficient between the real variable and the adjusted
variable by the model. So, a value near zero indicates bad adjust of the model and a value
near one indicates a good adjust of the model. The parameter A indicates the annual
growing rate for of the occupation area of the respective class. The area occupied by the
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 107

Fig. 4. Images Landsat post-classified corresponding to the “Potential zone for agriculture”
for different dates.
108 Remote Sensing – Applications

Fig. 5. Areas of land cover as function of the year in the “Potential area for agriculture”
(Total), Caiuba region. SWVC = Soils without vegetal cover.

Class of land cover A B R2 Period

Potential zone for agriculture 271 486996 0.97 1981 - 2009
Rice crops 302 584736 0.93 1981 - 2005
Table 4. Parameters of the linear regression models for the area of classes "Potential zone for
agriculture" and "Rice crops" as a function of time (Area = A*year - B, in hectares).

"Potential area for agriculture" grew with a rate of 271 hectares by year, little inferior than
the growing rate for the area occupied by the class "Rice crops", that is 302 hectares by year.
Parameter B is the value of area in the year zero without real significance in this case.
Data of the municipality of Pelotas (Filippini Alba & Siqueira, 1999) and data for the state of
Rio Grande do Sul (Bolfe et al., 2009) were compared to data presented here, after legend
conversion. The correlation coefficient of the data discussed here was 0.54 with data of the
first paper and 0.77 with data of the second one. Some classes showed significant
differences, by instances Bolfe et al. (2009) evaluated 50% of area occupied by "pastures" in
the state, but the value was about 30% for the other works. The area occupied by water was
19-20% in the Caiuba region, due to the occurrence of the lagoons. This value was 1% in the
municipality context and 3% in the regional one.

4. Analisys of land cover evolution in the municipality of Montenegro, Estado

do Rio Grande do Sul (1993-2008)
The municipality of Montenegro is located 55 kilometers South from Porto Alegre (state
capital), with a territorial area of 420 square kilometers and population about 59,557
inhabitants. Thirty-three municipalities, including Montenegro, integrate the vegetal
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 109

carbon productive pole. The production of black acacia for the manufacture of tannin is an
important activity for the economy of the municipality since 1948, when the first factory
of tannin derived from the bark of acacia was installed (TANAC, 2010). Recently, the fruit
production is becoming increasingly important in the context of local economy. The
intense forest exploitation, the occurrence of new uses of land and a moderate urban
occupation oriented the choice of the municipality of Montenegro for this research,
focusing on the detection of temporal changes in the territorial organization, during the
period 1993 to 2008, in the context of the project “Development and evaluation of
products and co-products of the vegetal carbon productive chain in the State of Rio
Grande do Sul, aiming for sustentability”, with coordination of Embrapa Temperate
Climate Reasearch Center.
The topography of the municipality is complex when compared to the previous case, while
in the southeastern region occurs a flat terrain changing for slightly wavy; in the north
sector occur a basalt plateau with a rugged relief.
Imagery of the Landsat 5 satellite were used, corresponding to the scene of orbit 221, points
80 and 81 for WRS-2 (INPE, 2010b), for three different dates: September 8th, 1993; August 8th,
1999 and October 3th, 2008. The initial data processing was performed with the software
Marlin (INPE, 2010a), after that, the software ER-Mapper (1995) was used for classification
according to isoclass likelyhood criteria. The images were registered with known ground
control points, considering terrestrial features of easy identification, so that, the coordinates
systems were uniformized and small errors eliminated. The projection used was the
Universal Transector of Mercaptor (UTM), region 22 South, datum WGS 84.
Eigth classes were defined by the supervised classification process according to maximum
likelihood criteria. The classes "Annual crops", "Perennial crops" and "Pastures/SWVC"
were mapped together in gray tones (Fig. 6). The annual crops reached a maximum area of
production in 1999 (Table 5) with poor production in previous and posterior times. By
another hand, the perennial crops reached a maximum in 2008, after a significant increment
in the previous years, as consequence of an important citrus production. Pastures and
SWVC were mapped together due to the dinamic process of changes evolving both classes.
A little reduction of the area occupied for both classes was observed.

Class of land cover 1993 1999 2008

Annual crops 13.9 5835 21.2 8884 12.1 5084
Forestry 19.2 8076 5.5 2299 18.3 7686
Native forest 18.9 7948 23.0 9675 21.3 8957
Pastures/SWVC 27.0 11340 26.8 11246 23.2 9733
Perennial crops 8.9 3737 18.2 7655 19.6 8226
Unevaluated 3.3 1370 1.9 810 0.4 173
Urban 6.5 2729 2.1 878 3.3 1377
Water 2.3 968 1.3 554 1.8 767
Units % hectares % hectares % hectares
Table 5. Areas calculated with TM/Landsat 5 imagens for the period 1993 – 2008 through
classification methods for Montenegro municipality (Schroder & Filippini-Alba, 2010a).
110 Remote Sensing – Applications

Fig. 6. Evolution of land cover related to agriculture and pastures/SWVC in the

municipality of Montenegro based on Landsat 5 imagery (1993 – 2008).
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 111

The class “Forestry” showed a minimum of planting area in 1999, what is evident in the
map of spatial distribution (Fig. 7.), but the class “Native forest” showed a maximum that
year (Fig. 8). The class “Urban” includes other features besides the urban regions, by
instance outcrops, which explains its high value in 1993. The density of the central spot in
the image of 2008 suggest a real increment of urban population that year.

Fig. 7. Evolution of land cover related to the classes “Native forest”, “Forestry” and “Urban”
in the municipality of Montenegro based on Landsat 5 imagery (1993 – 2008).
112 Remote Sensing – Applications

Fig. 8. Evolution of the land cover classes considered in this chapter in the municipality of
Montenegro, Rio Grande do Sul state, Brazil.

5. Comparison and evaluation of errors

Two strategies were used to analize the errors of the classification methods: (a) Duplication
of process with new test areas for the same classes in Caiuba region. (b) Confusion matrixs
by truth of field for Montenegro municipality. Each strategy is related to a different error
condition, that is, error of processing and error of the method respectively.
The maximum error for case (a) was for the class "Soils without vegetal cover", more or less
2% when the overall area was evaluated (Table 6). The interference of the clouds was of the
same order (value insered with the class “Beach/dunes/clouds”). A confusion between
dunes and water (sediments) occurred in the Caiuba region (central part of Fig. 9). Other
sectors appear very similar for both images.

Class Image A Image B

Water, % 20.3 21.3
Wetland, % 18.9 18.5
Pastures, % 12.2 12.4
Rice crops, % 7.7 9.1
Sandy fields, % 13.4 13.6
Beach/dunes/clouds, % 8.4 7.8
Forestry, % 3.6 3.6
SWVC,% 15.6 13.8

Table 6. Errors derived from classification with new test areas in the Caiuba region for the
image of Jan. 20th, 2002 in a total area of 204088 hectares. SWVC = Soil without vegetal cover.
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 113

Fig. 9. Comparison of the classification in Caiuba region with Landsat image of 2002 for
different test areas.

Data from Montenegro municipality considered the confusion matrix constructed with the
truth of field for the Landsat 5 image of april 13, 2009. Thus, 48 control points were selected
in the image, trying a “randomly - homogeneous” distribution on the territory of the
municipality. Each point was verified at field in september-november 2009 and historical
informations were collected with the local farmer when posible.
The accuracy of the method was moderate, that is, 42% for the full process (Table 7).
Forestry, Pastures/SWVC, Perennial crops and Urban/outcrops showed the better results,
114 Remote Sensing – Applications

with values greater or equal than 50%. The rest of the classes presented low accuracy with
values in the interval 0 – 25%. The correlation coefficient of the quantity of control points
and the accuracy was 0.43, suggesting few dependence between both variables. Anyway, a
critical case occurred with the class "Native forest" with 11 control points and only two hits.
An explanation for the low accuracy of the classification process and some specific classes is
the shadow derived from the steep topography, causing confusion among classes and
inconsistent results. A improvement of the results is obtained when principal components
are considerer before classification, with a potential increment of accuracy of 10% (Schroder
& Filippini-Alba, 2010b).

Control points SCC Accuracy

Forestry 9 5 56%
Pastures/SWVC 11 6 55%
Water 2 0 0%
Perennial crops 7 4 57%
Native forest 11 2 18%
Urban/outcrops 4 2 50%
Anual crops 4 1 25%
Total 48 20 42%
Table 7. Results of the confusion matrix for the process of classification in the municipality
of Montenegro, Rio Grande do Sul state, Brazil (Schroder & Filippini-Alba, 2010b). SCC =
samples correctly classified.

6. Conclusion
Two categories of change detection techniques (Lu et al., 2004) were considered in this
chapter, all of them including classification methods: Post-classification comparison and
strategy with GIS.
The strategy with GIS isolated the poligon corresponding to the “Potential area for
agriculture”, then, the interference between some pair of classes was eliminated, by
instance, wetlands and rice crops. The post-classification comparison allowed a rapid
approach about the region with minor accuracy (preliminary results). Definition of the
method used depends on the ratio between cost and efficiency according to the designed
objectives.
Errors associated to classification methods are mainly due to the spectral answer, by
undefinition of classes or occurrence of pixels of transition, because the errors derived from
digitalization were insignificant. Atmospheric conditions and the regional topography also
influence the process of classification.
Land cover changes in a dynamic way, sometimes with significant transformation rates of
one class to another, as the discussed cases confirm. Truth of field appears as an optimal
method to improve results, but the cost of process, in time, financial and human resources is
incremented.
Land Cover Change Detection in Southern
Brazil Through Orbital Imagery Classification Methods 115

7. Acknowledgments
The authors are grateful to Dr. Angela Campos Diniz (Embrapa) and Dr. Cleber Palma
(Federal University of Rio Grande), leaders of the projects that funded research activities.
Partial support was obtained through the Foundation for Research Support in the State of
Rio Grande do Sul (FAPERGS).

8. References
Bolfe, L.E., Siqueira, J.O.W. de, Pereira, R.S., Filippini -Alba, J.M. & Miura, A.K. (2009). Uso,
ocupação das terras e banco de dados geográficos da metade sul do Rio Grande do
Sul. Ciência Rural, Vol. 39, No. 6, (September 2009), pp. 1729 – 1737, ISSN 0103-8478.
Crist, E.P. & Kauth, R.J. (1986). The Tasseled Cap De-Mystified. Photogrammetric Engineering
and Remote Sensing, Vol. 52, No. 1, (January 1986), pp. 81-86, ISSN 5201-0081.
Crosta, A. P. (1993). Processamento digital de imagens de sensoriamento remoto. State University
of Campinas, ISBN 85-85369-02-7, Campinas, Brazil.
Davis, J. (2002). Statistics and data analysis in geology. Wiley, ISBN 04-71172-75-8, New York.
Embrapa (2009). Sistemas orbitais de monitoramento e gestão territorial. In: Embrapa
Monitoramento por Satélite, 27/06/2011, Available from
http://www.sat.cnpm.embrapa.br.
ER-Mapper (1995). ER Mapper Reference. Earth Resource Mapping, West Perth, Australia.
Filippini-Alba, J.M. & Siqueira, O.W. de. (1999). Caracterização da ocupação das terras em
escala municipal, através do processamento de imagens Landsat. Área piloto,
Pelotas, RS. Agropecuária Clima Temperado,Vol. 2, No. 2, (December 1999), pp. 223 –
231, ISSN 1415-6822.
Guild, L.S.; Cohen, W.B.; Kauffman, J.B. (2004). Detection of deforestation and land
conversion in Rondônia, Brazil, using change detection techniques. International
Journal of Remote Sensing, Vol. 25, No. 4, (February 2004), pp. 731-750, ISSN 0143-
1161.
Inpe (n.d.). Catálogo de imagens. In: National Institute of Spatial Research (Inpe),
10/jun/2010b. Available from <http://www.dgi.inpe.br/CDSR/>.
Inpe (n.d.). Marlin. In: National Institute of Spatial Research (Inpe), 01/03/2010a. Available
from <http://www.dgi.inpe.br/CDSR/marlin_PT.php>
Jensen, J.R. (1996). Introductory digital image processing: a remote sensing perspective. Prentice
Hall , ISBN 0-13-205840-5, Upper Saddle River, USA.
Lillesand, T. M.; Kiefer, R.W. Remote sensing and image interpretation. Wiley, ISBN 0-471-
57783-9, New York: 1994. 750p.
Lu, D.; Mausel, P.; Brondízio, E.; Moran, E. (2004). Change detection techniques. International
Journal of Remote Sensing, Vol. 25, No. 12, pp. 2365-2407, ISSN 0143-1161.
Miranda, E. E. (2002). Brasil visto do espaço: Rio Grande do Sul. In: Embrapa Monitoramento
por Satélite, 1 CD-rom.
Miranda, E. E. (2005). Brasil em Relevo. In: Embrapa Monitoramento por Satélite, 02/08/2010.
Available from <http://www.relevobr.cnpm.embrapa.br>.
Moreira, M. A. (2005). Fundamentos de sensoriamento remoto e metodologias de aplicação.
Universidade Federal de Viçosa, ISBN, Viçosa, Brazil.
Nasa (2004). ASTER: Advanced spaceborne thermal emission and reflection radiometer. In:
Nasa/JPL, 27/06/2011. Available from < http://asterweb.jpl.nasa.com>.
116 Remote Sensing – Applications

Pontara, R.C.P. (1998). Análise do comportamento espectral dos filitos carbonosos para interpretação
de imagens. Brasilia University, Academic dissertation, Brasília, Brazil.
Schroder, V.F.; Filippini-Alba, J.M. (2010). Classificaçção de imagens orbitais com auxílio da
análise por componentes principais no município de Montenegro – RS, 2009. In:
Infoteca/Embrapa, 11/10/2011. Available from
http://www.infoteca.cnptia.embrapa.br/handle/doc/884278
Schroder, V.F.; Filippini-Alba, J.M. (2010). Potencialidade do uso de imagens orbitais para
detecção de mudanças temporais: estudo de caso no município de Montenegro –
RS, 1993 – 2008. In: Infoteca/Embrapa, 11/10/2011. Available from
http://www.infoteca.cnptia.embrapa.br/handle/doc/884267
Tanac (n.d.). Histórico. In: Tanac Cia. 08/07/2010, Available from
<http://www.tanac.com.br/PT/institucional.php?codCategoriaMenu=148&nomA
rea=Hist%F3rico&codDado=2&menu=138>.
 5

Mapping Soil Salinization of

Agricultural Coastal Areas in Southeast Spain
Ignacio Melendez-Pastor, Encarni I. Hernández,
Jose Navarro-Pedreño and Ignacio Gómez
Department of Agrochemistry and Environment,
University Miguel Hernández of Elche
Spain

1. Introduction
Soil salt content is a key factor that determines soil chemical quality together with soil
reaction, charge properties and nutrient reserves (Lal et al., 1999). An adequate salt supply is
essential for an optimum development of photosynthetic mechanism and other biochemical
processes in plants (Sitte et al., 1994). Soil salt content constitutes an environmental problem
when salt accumulation generates drastic changes in soil physical and chemical properties,
adversely affecting soil productivity and plant growth (Richards, 1954; Qadir et al., 2000).
Salinization affects about 30% of the irrigated land of the world, decreasing this area
approximately 1-2% per year due to salt-affected land surfaces (FAO, 2002). In Europe,
about 1-3 million hectares of the land are affected by salinization (European Commission,
2003), and most of these areas are situated in the Mediterranean basin. In Spain, about 18%
of the 3.5 million hectares of irrigated land are severely affected or at serious risk of soil
salinization (European Commission, 2002). Soil salinization is a frequent problem in arid
and semiarid regions like Southeast Spain (Hernández Bastida et al., 2004). In these areas,
agriculture with a great water requirement combined with high water tables and an adverse
climate (increased occurrence of extreme drought events) have forced irrigation with poor
quality water, causing processes of soil degradation and salinization, limiting crop growth
and the production capacity (Pérez-Sirvent et al., 2003; Acosta et al., 2011).
Evaluating the spatial variability of basic soil properties in saline soils, and mapping
spatial distribution patterns of these soil properties helps to make effective site-specific
management decisions (Ardahanlioglu et al., 2003). Accordingly, remote sensing
techniques and geographic information systems (GIS) have introduced a new era for soil
resources assessment and monitoring in terms of information quality (Mermut and
Eswaran, 2001). A priori knowledge of spectral characteristics of remotely sensed materials
is fundamental to any valuable quantitative analysis (Ben-Dor et al., 1997). The variety of
absorption processes occurring in the soil and their wavelength dependence allow us to
derive information about the chemistry of the minerals composing it from the reflected or
emitted light (Clark, 1999). Reflectance spectra of soils are attributed to numerous soil
properties. There are no narrow absorption bands linked to soil salinity status, since it is
118 Remote Sensing – Applications

determined by soil properties such as pH, electrical conductivity, salt content and
exchangeable sodium percentage (Csillag et al., 1993; Farifteh et al., 2008). In this sense, soil
reflectance is derived from the particular spectral behaviour of the heterogeneous
combination of minerals, organic matter and soil water (Ben-Dor and Banin, 1994). Salt-
affected soils cations (Na+, Mg2+, K+, and Ca2+) and anions (Cl-, SO42-, CO32- and HCO3-) can
be detected by optical spectrometers since salt minerals have diagnostic spectral features
occurring in the visible and near infrared (VNIR) and short-wave infrared (SWIR) spectral
regions (Farifteh et al., 2008). Saline soils usually have evaporate minerals, which spectral
features that can be explained by vibrational absorption due to water molecules chemically
bound as part of the crystal structure (Howari et al., 2000). In this sense, the spectral
differences of evaporates of single salt compounds are determinant of the type and
mineralogy of the soils (Howari et al., 2000).
Remote sensing has been extensively employed in soil salinity studies. Data from aerial
photography, videography, and optical, thermal, microwave or geophysical sensors has
been used in soil salinity mapping (Metternich and Zinck, 2003). Perhaps, the most widely
used remote sensing data in recent decades have been provided by multispectral
(Landsat, SPOT, IRS, ASTER) or hyperspectral (DAIS, HyMap, AVIRIS, Hyperion) sensors
in the spectral range approximately between 400 and 2500 nm. Researchers have
frequently employed remote sensing data to map soil salinity with multispectral
(Metternich and Zinck, 1997; Dwivedi et al., 2001; Melendez-Pastor et al., 2010a) and
hyperspectral images (Dehaan and Taylor, 2002, 2003; Schmid et al., 2009, Ghrefat and
Goodell, 2011). Pioneering studies in the 1970s employed air-borne and satellite-borne
multispectral scanners to detect soil salinity, indicating the better capability of infrared
bands over visible bands to locate saline soils and the low contribution of thermal bands
to improve the delineation of saline areas (Richardson et al., 1976; Dalsted et al., 1979).
Nowadays, imaging spectroscopy techniques are employed for the automatic detection of
soil salinization with airborne or satellite sensor (Dehaan and Taylor, 2002, 2003;
Dutkiewicz et al., 2009; Schmid et al., 2009; Weng et al., 2009; Melendez-Pastor et al.,
2010a; Ghrefat and Goodell, 2011). Imaging spectroscopy deals with the mapping of
ground materials by detecting and analysing reflectance/absorbance features in
hyperspectral (or multispectral) images (Clark, 1999). Imaging spectroscopy adds a new
dimension of remote sensing by expanding point spectrometry into a spatial domain and
under field conditions, which is a very good approach for the study of soil properties
(Ben-Dor et al., 2009).
The aim of this chapter is the application of remote sensing for the study of soil salinity of an
agricultural area in southeast coast of the Iberian Peninsula. Different digital image
processing techniques were applied to satellite multispectral images (Landsat TM).
‘Conventional’ hard classification techniques were combined with spectral mixture analysis
and soil properties to achieve a better understanding of the soil salinization process in the
study area.
Multispectral satellite images such as those obtained by the Landsat program provide low
or free cost worldwide coverage for four decades. Moreover, salinization problems are
concentrated in arid and semi-arid regions, often in developing countries with few economic
resources. Although there are more advanced sensors that can provide a more precise
quantification of the extent of soil salinity (e.g. hyperspectral), their high cost difficult its
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 119

extensive use. Therefore, it is necessary to continue investigating the application of

multispectral image repositories as a tool to assist in the monitoring and management of
saline soils.

2. Material and methods

This study will evaluate the applicability of various remote sensing techniques for studying
salinization processes in an agricultural coastal area. One of the greatest difficulties in the
application of remote sensing techniques to the study area is the fragmentation of the
territory by the existence of small plots and buildings that create a dispersed mixture of
spectral signals to the scale of a moderate spatial resolution multispectral remote sensing
image as those acquired by the Landsat Thematic Mapper sensor. This difficulty motivates
the need to evaluate various techniques and methodological approaches to carry out this
study as necessary to help monitoring the processes of salinization.
Representative soils of the area were sampled and their properties were characterized at the
laboratory by standard methods. Predominant land cover classes at the soil sampling plots
and at additional land cover validation points were identified. Land cover is a fundamental
variable that impacts on and links many parts of the human and physical environments
(Foody, 2002) with a great influence on soil properties (Caravaca et al., 2002; Majaliwa et al.,
2010; Biro et al., 2011). Both kinds of information in a GIS database were included. In this
sense, the effect of land cover on soil properties was statistically evaluated. Then,
multispectral images were employed for a hard land cover mapping with a supervised
approach using the k-nearest-neighbour classifier. Accuracy assessment methods
highlighted the need to employ a mixed pixel focus to deal with the particularities of the
study area. Spectral unmixing techniques allowed the identification of representative
spectral endmembers and the obtainment of their corresponding fraction images. Finally,
fraction endmembers were employed to characterize land cover classes and to predict soil
properties with various statistical methods.

2.1 Description of the study area

The study area is located in a coastal zone of Southeast Spain, in the province of Alicante. It
is located around 38.14°N and 0.73ºW, at the south of the cities of Elche and Alicante. The
study area (Figure 1) comprises alluvial plains resulting from the accumulation of sediments
from the Segura and Vinalopó rivers. During most of the Holocene (~10,000 years ago to
present) the study area was a large lagoon (Blázquez, 2003). In the last centuries, the ancient
lagoon was transformed into an irrigated agricultural land draining the wetland.
Nowadays, this area is a mixture of small-size cities, coastal urban areas, scattered
residential houses, irrigated crops and isolated and scattered wetlands. The perimeter of the
study area was delimited according to natural or man-made features in order to enclose a
large coastal plain area primarily occupied by irrigated agricultural activities. The study
area lies in the north with the natural parks of El Hondo and the Salinas de Santa Pola. Both
natural areas are wetlands included in the RAMSAR list of wetlands of international
importance. The east and south boundaries are the Sierra del Molar and the Segura River
respectively. Urban areas and sclerophyllous vegetation mainly occupy the Sierra del Molar,
while the Segura River is the most important watercourse in southeast of Iberian Peninsula
providing water for irrigation agriculture and to fill the reservoirs that currently comprise
120 Remote Sensing – Applications

Fig. 1. Study area with the Landsat scene (false-colour composite RGB:742) and
superimposed cartographic information (soil samples, urban areas, natural parks and
roads).

the wetland of El Hondo. The western boundary of the study area is a motorway that cuts
north to south the floodplain.
This coastal region has a semiarid Mediterranean climate, with a mean annual rainfall of less
than 300 mm and a mean annual temperature of 17 ºC and defined by the Köppen climate
classification system as Bsk class (dry climate with a dry season in summer and a mean
annual temperature about 18 ºC). The climate is arid or semiarid according to the aridity
index of Martonne (De Martonne, 1926) and the aridity index of UNEP (1997) respectively.
Figure 2 shows the daily climatic diagram of mean temperature, precipitation and
evapotranspiration (by the Penman-Monteith method) for the hydrological year 2010-2011
(from October to September) at Catral meteorological station. Mean daily temperature (blue
line) varies from approximately 9ºC in winter to more than 25ºC in summer. Rain events
(red bars) mainly occurred from December 2010 to May 2011 with total accumulated
precipitation of 182 mm. This very scarce precipitation joint with an accumulated
evapotranspiration of 1115 mm implied that the hydrological year was very dry.
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 121

Fig. 2. Daily values of precipitation, mean temperature and evapotranspiration for the
hydrological year 2010-11 at Catral station. Source data from the Spanish Ministry of
Environment and Rural and Marine Affairs (MARM).

Predominant soil classes are Entisols according to the Soil Taxonomy (Soil Survey Staff,
2006) but affected by agriculture practices along years. They are characterized by a massive
presence of carbonates and soluble salt content. In the studied area, irrigation is essential to
support agriculture. The water deficit during several months requires irrigation while low
quality water is used in the poorly drained soils of these coastal plains, being soil
salinization an environmental problem. Thus, the study area soils are subjected to severe
risk of physical, chemical and biological degradation (De Paz et al., 2006) that endanger
agriculture sustainability.

2.2 Field survey

Field survey was done in the late spring and summer months of the hydrological year 2010-
2011 to collect soil samples and identify land cover classes. An extensive soil sampling was
done, and 116 samples were collected and geographically referenced. Samples were
obtained from the upper 5 cm as solar radiation in VNIR spectral range has limited
penetration capabilities. Soil samples were dried at room temperature and a 2 mm sieve was
used to separate the fine fraction to be analysed. Analysed soil characteristics included in
the study were electrical conductivity (EC) (1:5 w/v water extraction), pH and organic
carbon (OC) by wet chemical oxidation (Walkley and Black, 1934) with potassium
dichromate oxidation (Nelson and Sommers, 1982).
A land cover validation campaign was also conducted along with the soil survey in order to
allow accuracy assessment of generated land cover maps. Land cover validation points were
randomly generated in a GIS and a database with the land cover category generated. A total
of 205 land cover validation points were identified, combining field observation and recent
aerial orto-photography (0.5 m of spatial resolution). Land cover classes identified in the
122 Remote Sensing – Applications

study area were: water bodies, seasonal or permanent crops, saltmarshes and misused
agricultural field that tends to be saltmarshes, palm groves, marshes with almost permanent
inundation, and anthropic areas (Table 1).

Land cover ID Features

Water 1 Wetlands water tables and irrigation ponds
Arable land 2 Herbaceous (e.g. alfalfa, barley) and horticultural (e.g. melon,
broccoli) crops
Permanent crops 3 Fruit trees (e.g. orange, lemon, pomegranate)
Fallow/abandoned 4 Fallow or recently abandoned agricultural land.
Saltmarsh 5 Halophyte vegetation (e.g. Salicornia sp., Suaeda sp., Limonium sp.,
Halocnemum sp.)
Palm groves 6 Palm trees plantations and nurseries, mainly from Phoenix
dactilifera
Marsh 7 Phragmites australis dominated wetland vegetation
Man-made/urban 8 Urban areas, roads, farms or industrial areas
Table 1. Descriptions of land cover classes identified in the study area.

Land cover categories at soil sampling points were also identified and included along with
soil properties in a GIS database for the land cover classification training stage and for
further spatial analyses. Note that land cover (i.e. biophysical materials found on the land)
and land use (i.e. how the land is being used by human beings) (Jensen, 2007) are different
terms but often used together or interchangeably. In this chapter, we adopt the term land
cover because we are interested in knowing about the biophysical characteristics of the
study area, but the knowledge of both land use and land cover are important for land
planning and land management activities (Lillesand et al., 2003).

2.3 Satellite imagery preprocessing

Remote sensing data were acquired by the Thematic Mapper (TM) sensor on-board the
Landsat 5 satellite. Meteorological conditions and the satellite pass over the study area
conditioned the date of image acquisition. A scene acquired on 28th June 2011 (path 199 row
33) was employed for analyses. A vertical black line on Figure 2 indicates the time of
acquisition of the scene. No rain events happened 16 days prior to the scene acquisition
date. Typically summer meteorological conditions without cloud coverage and high
temperature were registered on the date of image acquisition, and thus the image quality
was optimal.
Satellite image preprocessing included geometric and atmospheric corrections with the aim
to ensure the spatial comparability with other data sources and to obtain at-ground
reflectance pixel spectra, respectively. Various georeferenced data types were used for the
geometric correction: aerial orthophotos (0.5 m of pixel resolution) and digital cartography
(scale = 1:10000). The Landsat 5 TM scene was geometrically corrected using Ground
Control Points (GCP) identified on the orthophotos and cartographic maps. A quadratic
mapping function of polynomial fit and the nearest neighbour resampling method were
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 123

used for the correction. The nearest neighbour resampling method was selected because it
ensures that the original (raw) pixel values are retained in the resulting output image, which
is an important requirement in any change detection analysis (Mather 2004). The maximum
allowable root mean square error (RMSE) of the geometric correction was less than half a
pixel, a reference value frequently cited (Townsend and Walsh 2001; Jensen, 2005).
Atmospheric correction involves the estimation of the atmospheric optical characteristics at
the time of image acquisition before applying the correction to the data (Kaufman, 1989).
This type of correction is a pre-requisite in many remote sensing applications such as in
classification and change detection procedures (Song et al., 2001). Radiometric calibration
was applied prior to the atmospheric correction. The conversion of raw digital numbers
(DNraw) of Landsat level 1 (L1) image products to at-satellite radiance values (Lsat) required
the application of current re-scaling values (Chander et al., 2010) by applying the following
expression (Chander and Markham, 2003; Chander et al., 2010):

 LMAX   LMIN  
Lsat    (DN)  LMIN  (1)
 255 

Where Lsat is at-satellite radiance [W/(m2 sr m)]; LMIN  is the spectral radiance that is scaled
to Qcalmin [W/(m2 sr m)] (Qcalmin is the minimum quantized calibrated pixel value, i.e.
DN=0, corresponding to LMIN ); LMAX  is the spectral radiance that is scaled to Qcalmax
[W/(m2 sr m)] (Qcalmax is the maximum quantized calibrated pixel value, i.e. DN=255,
corresponding to LMAX ); and DN are digital numbers of the L1 image product. Surface
reflectance values () were computed by using the image based COST method (Chavez Jr,
1996). Path radiance (Lp) values were computed by using the equation reported in Song et al.
(2001) that assumes 1% surface reflectance for dark objects (Chavez Jr, 1989, 1996; Moran et
al., 1992). The optical thickness for Rayleigh scattering (r) was estimated according to the
equation given in Kaufman (1989).

2.4 Land cover classification

Image classification procedures aim to automatically categorize all pixels in an image into land
cover classes or themes (Lillesand et al., 2003). Thematic mapping from remotely sensed data
can be defined as grouping together cases (pixels) by their relative spectral similarity
(unsupervised component) with the aim of allocating cases based on their similarity to a set of
predefined classes that have been characterized spectrally (supervised component) (Foody
2002). Multispectral images (like Landsat TM scenes) are frequently used to perform the
classification based on spectral pattern recognition methods that exploits the pixel-by-pixel
spectral information as the basis for automated classification (Lillesand et al., 2003). In this
study, a supervised land cover classification of the Landsat TM image was performed with a k-
nearest-neighbour clustering algorithm to obtain a discrete or ‘hard’ categorical land cover
map for the study area. K-nearest-neighbour (KNN) classifier searches away from the pixel to
be classified in all directions of the spectral space until it encounters k user-specified training
pixels and then assigns the pixel to the class with the majority of pixels encountered (Jensen,
2005). KNN algorithm has been successfully applied for land cover classification with remote
sensing data (Franco-Lopez et al., 2001; Haapanen et al., 2004; Blanzieri and Melgani, 2008).
Land cover classes assigned to the soil plots were employed in the training stage of the
124 Remote Sensing – Applications

algorithm. Major urban areas were digitized with a GIS and masked-out of the supervised
classification procedure since urban areas induce a great spectral confusion. Water areas
training points were also included in the training dataset.
The land cover validation database was employed to evaluate the performance of the
classification. Land cover map accuracy assessment was quantified with statistical
methods such as the error matrix and the kappa statistic. The error matrix is a square
array of numbers organized in rows and columns that express the number of sample units
(i.e. pixels) assigned to a particular category relative to the actual category as indicated by
the reference data (Congalton, 2004). Reference data are in the columns while the rows
indicate the map categories to be assessed. This form of expressing accuracy as an error
matrix is an effective way to evaluate both errors of inclusion (commission errors) and
errors of exclusion (omission errors) present in the classification as well as the overall
accuracy (Congalton et al., 1983). In addition to the error matrix, the Kappa coefficient
developed by Cohen (1960) was employed to quantify the accuracy of the land cover map.
Cohen’s Kappa (or KHAT) is a measure of agreement for nominal scales based on the
difference between the actual agreement of the classification (i.e., agreement between
computer classification and reference data as indicated by the diagonal elements) and the
chance agreement, which is indicated by the product of the row and column marginal
(Congalton et al., 1993).

2.5 Spectral unmixing

A mixed pixel results when a sensor’s Instantaneous Field of View (IFOV) includes more
than one land cover type on the ground (Lillesand et al., 2003). The spectrum of a single
pixel is a complex measurement that integrates the radiant flux from all the spatially
unresolved materials in the IFOV, regardless of whether or not we know their identities
(Adams and Gillespie, 2006). Spectral mixture analysis (SMA) has been developed as a
method to transform the reflectance in the bands of multispectral images to fractions of
reference endmembers, which are reflectance spectra of well-characterized materials that
mix to produce spectra equivalent to those of pixels of interest in the image (Adams et al.,
1995). As part of SMA techniques, linear spectral unmixing (LSU) models tread the radiation
recorded by a sensor as the result of a linear mixture of spectrally pure endmember
radiances (Small and Lu 2006). This method is based on the assumptions that: 1) the
recorded radiation by the sensor for each pixel is limited to the sensor’s IFOV, and assumes
no influences by reflected radiation from neighbouring pixels (Settle and Drake 1993), 2) the
overall global radiance is proportional to the surface occupied by each land cover type, and
3) the spectrally pure endmembers are valid for the whole study area (Quarmby et al. 1992).
LSU models describe radiation reflected by an individual pixel (i,j) of a band k as the result
of the product of reflectance for each land cover type by their respective mixture fraction
plus an additional associated error for each pixel. The general expression of the model is
presented in the following equation:

i , j , k   m  1, p Fi , j , m m , k  ei , j (2)

Where i,j,k is the observed reflectance of a pixel for row i, column j, and band k ; Fi,j,m is the
proportion of component m of a pixel for row i, column j, for each one of the pure
components; m,k is the characteristic reflectance for component m in band k ; and ei,j is the
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 125

error associated to the estimation of proportions for each pixel i, j. The Least Square Mixing
Model proposed by Shimabukuro and Smith (1991) is commonly used to resolve linear
spectral mixture models. The method proposed by Shimabukuro and Smith (1991) assumes
two initial restrictions for the computation of the proportions of spectrally pure
endmembers. The first one implies that pure endmember proportions must range between 0
and 1. This means that the proportions of the components are normalized to a common
range of potential values. The following expression summarizes this first restriction:

0  Fi , j ,m  1 (3)

The second restriction is that the sum of the fractions for every component is equal to the
total pixel surface. In this way, it is quite simple to express the individual contribution or
fraction of an endmember in relation to the total reflectance of the pixel.

 m1, p Fi , j ,m  1 (4)

The choice of a LSU model must consider both the landscape of the test site and the ability
of the model to depict the structure, shape and distribution of the basic landscape
components (Ferreira et al. 2007). Well-chosen endmembers not only represent materials
found in the scene, but provide an intuitive basis for understanding and describing the
information in the image (Adams and Gillespie 2006). Endmembers were obtained after
applying a spatial and spectral remote sensing data dimensionality reduction with the
minimum noise fraction (MNF) and pixel purity index (PPI) techniques, respectively. The
MNF is used to detect the inherent dimensionality of image data, segregating noise from the
signal in the data and reducing computational requirements for subsequent processing tasks
(Boardman and Kruse, 1994). The MNF as modified from Green et al. (1988) consists in two
steps: 1) applying a transformation, based on an estimated noise covariance matrix to
decorrelate and rescale the noise in the data (noise has unit variance and no band-to-band
correlations); and 2) performing a standard principal component transformation of the
noise-whitened data. A final dataset of coherent and almost noise-free bands are selected
from the MNF output and can be used for subsequent processing steps. Pixel Purity Index
(PPI) is a procedure for finding the most spectrally pure (extreme) pixels that typically
correspond to mixing endmembers in multispectral and hyperspectral images (ITT VIS,
2008). PPI is computed by repeatedly projecting n-dimensional scatterplots onto a random
unit vector; the extreme pixels in each projection (those pixels that fall onto the ends of the
unit vector) are recorded and the total number of times each pixel is marked as extreme is
noted. The selection of extreme pixels corresponding to analogous surface features is
complex due to the great number of pixels typically found in remote sensing image data.
The n-dimensional visualizer implemented in ENVI software (ITT Visual Information
Solutions) is a tool to locate, identify, and cluster the purest pixels and most extreme spectral
responses in a data set. The distribution of these points in n-space can be used to estimate
the number of spectral endmembers and their pure spectral signatures (Boardman, 1993).
Three endmembers were used in the LSU model of the study area, namely green vegetation
(GV), non-photosynthetic vegetation (NPV) and shade (S). The GV endmember represents
the signature of green dense vegetation, the NPV endmember is the signature of bare soil or
sparse non-photosynthetic vegetation, and the shade endmember represents the signature of
dark pixels and water bodies.
126 Remote Sensing – Applications

2.6 Statistical methods

The possible existence of differences in soil properties based on the land cover classes was
determined by the use of the analysis of variance (ANOVA). ANOVA is used to evaluate
significant differences between means of independent variables. The observed variance of
independent variables is partitioned into components by several explanatory variables
(factors). Land cover class was the factor employed in the analysis. Post-hoc analysis was
performed using Tukey method.
Relationships between fraction endmembers and soil properties were studied by the
principal component analysis (PCA). PCA is a technique of data dimensionality reduction
that performs an orthogonal transformation to convert potentially correlated input variables
into uncorrelated variables or principal components. The first components accumulate most
of the variance and therefore, the most useful information about the variables.
Regression analyses between soil properties and fraction endmembers were performed for
quantitative estimation of soil properties. A linear regression analysis applying a stepwise
method for variable entry and removal was the selected statistical technique. Model
selection was based on the lower typical error of the estimation and minimum collinearity.

3. Results and discussion

The relationship between various soil properties and land cover classes was analysed. Two
approaches to the study of land cover are presented. A first categorization based on discrete
land cover classes and another based on mixture fractions. Finally, statistical models for
predicting soil properties of interest in the study of soil salinity through the use of mixture
fractions are presented.

3.1 Soil properties

The study was focused on soil electrical conductivity (EC), pH and organic carbon (OC)
(Table 2). These properties are important in chemical and biological quality of soils (Lal et al.,
1999). Previous studies in semiarid areas combining remote sensing and soil analyses have
indicated significant differences in these properties in different land cover classes (Biro et al.,

Land covers EC (mS/cm) pH OC (%)

Arable land 1.38 ± 1.01 ab 8.23 ± 0.26 ab 1.92 ± 0.85 a
Permanent crops 0.70 ± 1.00 b 8.40 ± 0.26 b 1.29 ± 0.24 a
Fallow/abandoned 3.52 ± 2.43 ac 8.14 ± 0.25 ab 1.46 ± 0.36 a
Saltmarsh 3.98 ± 2.86 c 8.05 ± 0.07 ab 3.61 ± 0.81 b
Palm groves 3.82 ± 2.42 ac 8.20 ± 0.20 ab 2.11 ± 1.00 a
Marsh 4.71 ± 2.04 c 7.82 ± 0.09 a 5.24 ± 1.39 c
P-value <0.001*** <0.001*** <0.001***
Table 2. Descriptive statistics (mean ± standard deviation) of soil properties based on land
cover classes. The p-value and homogeneous subgroups (lower case letters; Tukey test, P <
0.05) resulting from the ANOVA test are included.
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 127

2011). Cultivated areas (i.e. arable land and permanent crops, homogenous subgroup b)
have lower electrical conductivity values than natural or semi-natural vegetation. The
construction of drainage systems at agricultural areas to encourage the leaching for salinity
control has been a traditional amelioration strategy (Qadir et al., 2000). This fact explains
that marshes and saltmarshes soils have higher EC values than the other land cover classes,
since they are areas with poor drainage and temporally flooded. Salinity increases when
farming finishes (i.e. fallow/abandoned) because irrigation water is not available to
promote salts leaching.
The pH values were slightly alkaline but significantly different for marshes. Wetland soils
are characterized by the permanent or seasonal inundation of the land, promoting
anaerobic conditions and thus reduced redox conditions (high concentration of H+ which
implies low pH)(Reddy et al., 2000). The organic carbon content was also different
depending on the type of land cover. Arable land and permanent crops soils have organic
carbon content ranging from 1.46 to 2.11% that is not very high (Pérez-Sirvent et al., 2003).
Opposite, wetland soils (i.e. stable saltmarshes and saltmarshes) exhibited the highest
organic carbon contents. Compared to upland areas, most wetland soils show an
accumulation of organic matter by the higher rates of photosynthesis in wetlands than
other ecosystems and the lower rates of decomposition due to anaerobic conditions
(Reddy et al., 2000).
All soil properties are significantly correlated (P<0.01) according to the Pearson bivariate
correlation test applied to the full dataset. Figure 3 shows two scatterplots of the land cover
classes average values (error bars represent the standard deviation) of pH and organics
carbon versus electrical conductivity (EC). EC is negatively correlated with pH (R=-0.61)
and positively correlated with organic carbon (R=0.34), while pH and organic carbon are
negatively correlated (R=-0.32). Two sets of distinct land uses mainly dependent on the EC
values are distinguished: 1) active cultures: with low EC and OC values, and 2) natural
vegetation and crops of low requirements (palm groves): with high EC and OC values,
increasing as the land cover is more similar to the wetland. Palm groves are the most
halotolerant crop and require little tillage.

3.2 Land cover classification

A land cover map was obtained with the k-nearest-neighbours algorithm (Figure 4).
Optimum results were obtained with k=4. The area occupied by the land cover classes
(hectares and percentage of the total area) was quantified (Table 3).
The study area is mainly agricultural but largely occupied by urban areas. Urban/man-
made areas represent 14.4% of the study area. There is a clear distinction between the
northeast portion (area between the two natural parks and the Sierra del Molar) with large
fields and less presence of buildings, and the rest of the study area, with numerous
buildings scattered, villages, small-size towns and smaller parcels. Dominant land cover
classes are also different at these two sectors. Close to the natural parks, there are many
saltmarshes (7.09% of the study area), marshes (1.95% of the study area), palm groves (2.21%
of the study area) and arable land (45.76% of the area and mainly forage, barley and
melons). The other sector has a massive presence of permanent crops (16.3% of the study
area and mainly citrus trees such as orange and lemon trees).
128 Remote Sensing – Applications

Fig. 3. Scatterplots with the average values pH and organic carbon versus electrical
conductivity for land cover classes (numbers in italics are the ID number of the class). X and
Y bars represent one standard deviation.

The distribution of land cover classes can be explained by the characteristics of the soils.
Generally, the closest soils to the wetland areas of the natural parks are more saline. These
soils have a poor drainage due to its lower altitude and very high-water tables, largely due
to the horizontal flow of water and salts from the nearby water bodies. Permanent crops
class dominates in areas that are close to the towns, being better drained and less saline.
Fallow/abandoned areas (12.11% of the study area) are spread throughout the study area as
a result of the abandonment of farming on individual fields. However, abandoned land is
more present in the proximal portion of the natural parks since the conditions of salinization
of soils led to their abandonment.
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 129

Fig. 4. Land cover map of the study area.

Land cover classes Area (ha) Area (%)

Water 17.37 0.20
Arable land 4061.52 45.76
Permanent crops 1446.93 16.30
Fallow/abandoned 1074.51 12.11
Saltmarsh 629.01 7.09
Palm groves 196.02 2.21
Marsh 172.71 1.95
Man-made/urban 1277.73 14.40
TOTAL 8875.80 100
Table 3. Area occupied by land cover classes according to the map obtained by k-nearest
neighbour.

The land cover map accuracy was evaluated with the data set of validation points. Overall
accuracy was a 68%, and KHAT value was 0.56. According to Landis and Koch (1977),
130 Remote Sensing – Applications

KHAT values ranging from 0.4 to 0.8 exhibit a moderate agreement. Inter-class confusion
was detected analysing the error matrix. A portion of arable land (78% of producer’s
accuracy, 65% of user’s accuracy) was wrongly classified as permanent crops,
fallow/abandoned land or saltmarshes. A great portion of palm groves (21% of producer’s
accuracy and 75% of user’s accuracy) was classified as arable land. The performance of the
automatic classification for marshes (90% of producer’s accuracy and 75% of user’s
accuracy) and water areas (100% of producer’s accuracy and 80% of user’s accuracy) was
highly satisfactory. The performance of the KNN algorithm for our land cover classification
approach was enough good and comparable with the accuracy obtained by Franco-Lopez et
al. (2001) classifying a forest stand (52% of overall accuracy with k=10), and the results of the
experiment carried on by Samaniego and Schulz (2009) classifying crop types (47% of
overall accuracy with k=5).

3.3 Spectral unmixing and land covers

Spectral mixture analysis was applied to obtain fraction images of green vegetation (GV),
non-photosynthetic vegetation (NPV) and shade endmembers. Spectral signatures of
selected endmembers are highly distinctive (Figure 5). These endmembers had optimal
spectral separability as measured with the transformed divergence method. GV endmember
is associated with vigorous vegetation, NPV endmember is associated with bare soil and dry
halophytic vegetation, and shade endmember is associated with water bodies and low
illuminated areas.

Fig. 5. Plot showing the spectral signatures of selected green vegetation (GV), non-
photosynthetic vegetation (NPV) and shade endmembers.

Fraction images of the three endmembers and the residual fraction of the spectral mixture
analysis were obtained (Figure 6). Values range from 0 for low high membership to the
image fraction (black colour) to 1 for high membership (white colour). Fractions images are
continuous variables that are graphically represented with a greyscale colour ramp. High
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 131

values of the shade fraction image are present in the wetland areas of the natural parks and
in a triangular area in the middle-right boundary of the study area that corresponds with a
small wetland. A white area in the right of the image corresponds with the Mediterranean
Sea. High green vegetation fraction values area scattered through the study area. They
correspond with active crops at the time of the image acquisition. Indeed, the white areas in
the NPV image fraction correspond with bare soil and saltmarshes which vegetation is quite
dry in summer and have a great spectral confusion with background soil. Urban areas were
also associated with this fraction image. Finally, high values of the residual fraction are
located in industrial areas, whose spectral signature was notably different respect to the
three endmembers of the unmixing model.

Fig. 6. Fraction images of shade/water, green vegetation (GV), non-photosynthetic

vegetation (NPV) and the residual component of the linear spectral mixture analysis.
White/black polygon represents the boundary of the study area.
132 Remote Sensing – Applications

Average values of the three fraction images for the land covers were computed and
represented in a ternary diagram (Figure 7). Water land cover has high shade fraction values
(>90%) and very low values for the GV and NPV fractions. Marshes have an important
fraction of shade (>55%) and around 30% of the GV fraction. This mixture composition is
highly indicative of the marshes structure with green Phragmites australis stands, growing on
flooded or water-saturated soils. Shade fraction has a low contribution in the other land
covers (<30%). Saltmarshes, permanent crops, arable land, fallow/abandoned and palm
groves land cover classes have GV fraction values between 20-40% and NPV fraction values
between 50-70%. This relative homogeneity in the mixture fractions values for different land
cover classes could be attributed to the lower water availability in summer, that promotes a
drying and browning of the vegetation and promotes spectral confusion. Melendez-Pastor et
al. (2010b) previously observed this phenomenon in the study area. They also employed
ternary diagrams, combining mixture fraction and land cover classes for a drought year and
an average year. Soil or NPV fractions increase their contribution in a dry weather scenario
(i.e. drought or summer) and the water and GV fractions have a lower contribution.

Fig. 7. Ternary diagram of the average mixture fraction values for the land cover classes.

3.4 Fraction endmembers to predict soil properties

Mixture fraction values were statistically related to soil salinity. Principal component
analysis provided valuable information about the relationship among soil properties and
spectral mixture analysis fractions. The first three principal components accumulated 75.8%
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 133

of total variance. PC1 was positively correlated with NPV fraction (factor loading = 0.988)
and negatively correlated with GV fraction (factor loading = -0.902). PC1 might be used to
separate vegetated from non-vegetated pixels. PC2 was positively correlated with electrical
conductivity (factor loading = 0.778) and the shade fraction (factor loading = 0.604) and
negatively correlated with the pH (factor loading = -0.785). PC2 might be used to
differentiate soil salinity status.
Salinization status seems to be related to the abundance of the shade fraction. This result
could be explained by the presence of water at the soil profile, which is an evidence of poor
drainage that could lead to salt accumulation. Thus, monitoring shade fraction values along
a year could be an indirect method to detect the evolution of soil electrical conductivity with
remote sensing. PC3 was positively correlated with the residual fraction of the spectral
unmixing (factor loading = 0.853) and organic carbon content (factor loading = 0.711).
Evident negative correlations with the PC3 were not found.

Fig. 8. Factor loadings plot for the measured soil properties and mixture fractions, and
average values factor scores plot for the land cover classes.
134 Remote Sensing – Applications

Previous studies assessed the relationships between PC factor loadings of soil properties
and PC factor scores of land cover classes (Biro et al., 2011). We also included the mixture
fraction values for soil plots in the principal component analysis. Soil properties and
mixture fractions factor loadings and average factor scores for land cover classes of the first
two components were plotted to explore their relationship (Figure 8). Factor loading values
range from -1 to 1. Factor scores of land cover classes were also included in the plot. Land
cover classes were differentiated from each other along PC1, mainly because of the high
positive factor loading of the NPV fraction and high negative factor loading of the GV
fraction. Also, land cover classes were differentiated from each other along PC2, mainly
because of the high positive loadings of the shade fraction, electrical conductivity and
organic carbon content, and high negative loading value of pH.
Finally, we tested the usefulness of mixture fractions to predict EC. Stepwise linear
regression was employed to model relationship among EC and mixture fractions. EC
variable was normalised with the natural logarithm, while the other variables were normal.
Table 4 summarizes the main parameters of two linear regression models. Moderate
adjustment was obtained from the regression with R=0.338 for the first model and R=0.408
for the second model. The ANOVA test (data not included in the table) indicated the
usefulness of the models with a p-value <0.001 for both cases. Model 1 included GV fraction
as the unique mixed fraction predictor variable, while model 2 included GV and also the
shade fraction. In both models, the coefficient B has a negative value for the variable GV,
suggesting an inverse relationship between the green covers and EC. By contrast, the
regression coefficient B was positive for the shade fraction, indicating a direct relationship
between the presence of shadows/water and EC. This latter observation corroborates the
interpretation given in the principal component analysis on the usefulness of the shade
fraction to predict soil salinity. Collinearity statistics revealed the absence of collinearity
problems as we obtained tolerance values much greater than 0 and VIF values much lower
than 15 (SPSS, 2009).

Unstandardized Collinearity
Coefficients Statistics
Model R R2 Adjusted R2 Predictors B Std. Error Sig. Tolerance VIF
1 0.338 0.115 0.107 (Constant) 7.483 0.106 0.000
GV -1.081 0.279 0.000 1.000 1.000
2 0.408 0.167 0.152 (Constant) 7.302 0.123 0.000
GV -1.005 0.273 0.000 0.989 1.011
Shade 1.320 0.492 0.008 0.989 1.011
VIF: variable inflation factor
Table 4. Summary of results of linear regression models.

Regression models with mixture fractions did not show collinearity that could lead us to
false predictions. A major constraint using proximal and remote sensing data for mapping
salinity is related to the fact that there is a strong vertical, spatial and temporal variability of
salinity in the soil profile (Mulder et al., 2011). Direct and precise estimation of the salt
quantities is difficult by using satellite data with a low spectral resolution because these fail
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 135

to detect specific absorption bands of some salt types, and the spectra interfere with other
soil chromophores (Mougenot et al., 1993). More research will help to improve the
prediction of soil properties with remote sensing data for a fast assessment of soil status
over large areas and at low cost.

4. Conclusions
This chapter provides an interesting case of study on the application of remote sensing to
soil salinity. The land use and management greatly affect soil salinity and land cover
mapping helps to delineate areas with different severity of salinization. The use of spectral
mixture analysis in combination with land cover maps and soil properties data is a more
advanced technique. Mixture fractions help to know the spectral behaviour of land cover
and their constituents using a simple three endmembers model. In addition, mixture
fractions can be used as predictors in regression models to predict the electrical conductivity
of soils. The results of regression models were encouraging but require further research to
improve them. Since mixture fractions are sensitive to spectral changes due to changes in
ground surface, they may be particularly useful for mapping the severity of soil salinization
processes over time with low coast satellite images. The combined use of soil properties
analytics, land cover maps and spectral mixture analysis is feasible for monitoring saline
soils and land management over large areas with a reduced cost.

5. Acknowledgment
The authors acknowledge the Spanish Ministry of Science and Innovation for the financial
support of the project ‘Efectos de la aplicación de un compost de biosólido sobre la calidad de suelos
con distinto grado de salinidad’ with the reference CGL2009-11194 that allowed this research.

6. References
Acosta, J.A.; Faz, A.; Jansen, B.; Kalbitz, K. & Martínez-Martínez, S. (2011). Assessment of
salinity status in intensively cultivated soils under semiarid climate, Murcia, SE
Spain. Journal of Arid Environments, Vol. 75, No. 11 (November 2011), pp. 1056-1066,
ISSN 0140-1963
Adams, J.B. & Gillespie, A.R. (2006). Remote Sensing of Landscapes with Spectral Images. A
Physical Modeling Approach. Cambridge University Press, ISBN 0521662214,
Cambridge, UK
Adams, J.B.; Sabol, D.E.; Kapos, V.; Filho, R.A.; Roberts, D.A.; Smith, M.O. & Gillespie, A.R.
(1995). Classification of Multispectral Images Based on Fractions of Endmembers:
Application to Land-Cover Change in the Brazilian Amazon. Remote Sensing of
Environment, Vol. 52, No. 2 (May 1995), pp. 137-154, ISSN 0034-4257
Ardahanlioglu, O.; Oztas, T.; Evren, S.; Yilmaz, H. & Yildirim, Z.N. (2003). Spatial variability
of exchangeable sodium, electrical conductivity, soil pH and boron content in salt-
and sodium-affected areas of the Igdir plain (Turkey). Journal of Arid Environments,
Vol. 54, No. 3 (July 2003), pp. 495-503, ISSN 0140-1963
Ben-Dor, E. & Banin, A. (1994). Visible and near-infrared (0.4-1.1 m) analysis of arid and
semiarid soils. Remote Sensing of Environment, Vol. 48, No. 3 (June 1994), pp. 261-
274, ISSN 0034-4257
136 Remote Sensing – Applications

Ben-Dor, E.; Inbar, Y. & Chen, Y. (1997). The reflectance spectra of organic matter in the
visible near-infrared and short wave infrared region (400–2500 nm) during a
controlled decomposition process. Remote Sensing of Environment, Vol. 61, No. 1
(July 1997), pp. 1-15, ISSN 0034-4257
Ben-Dor, E.; Chabrillat, S.; Demattê, J.A.M.; Taylor, G.R.; Hill, J.; Whiting, M.L. & Sommer, S.
(2009). Using Imaging Spectroscopy to study soil properties. Remote Sensing of
Environment, Vol. 113, No. SUPPL. 1 (September 2009), pp. S38-S55, ISSN 0034-4257
Biro, K.; Pradhan, B.; Buchroithner, M. & Makeschin, F. (2011). Land use/land cover change
analysis and its impact on soil properties in the Northern Part of Gadarif Region,
Sudan. Land Degradation & Development, Vol. in press DOI: 10.1002/ldr.1116, ISSN
1085-3278
Blanzieri, E. & Melgani, F. (2008). Nearest Neighbor Classification of Remote Sensing Images
With the Maximal Margin Principle. IEEE Transactions on Geoscience and Remote
Sensing, Vol. 46, No. 6 (June 2008), pp. 1804-1811, ISSN 0196-2892
Blázquez, A.M. (2003). L’Albufera d'Elx: evolución cuaternaria y reconstrucción paleoambiental a
partir del estudio de los foraminíferos fósiles. PhD dissertation. Universitat de Valencia,
Valencia, Spain.
Boardman, J.W. (1993). Automating Spectral Unmixing of AVIRIS Data Using Convex
Geometry Concepts, Summaries of the Fourth Annual JPL Airborne Geosciences
Workshop, Volume 1: AVIRIS Workshop, JPL Publication 93-26, pp. 11-14. Washington
D.C., USA.
Boardman, J.W. & Kruse, F.A. (1994). Automated spectral analysis: A geological example
using AVIRIS data, northern Grapevine Mountains, Nevada. Proceedings,
Environmental Research Institute of Michigan (ERIM) Tenth Thematic Conference on
Geologic Remote Sensing, pp. I-407 - I-418. Ann Arbor (MI), USA.
Caravaca, F.; Masciandaro, G. & Ceccanti, B. (2002). Land use in relation to soil chemical and
biochemical properties in a semiarid Mediterranean environment. Soil and Tillage
Research, Vol. 68, No. 1 (October 2002), pp. 23-30, ISSN 01671987
Chander, G. & Markham, B. (2003). Revised Landsat-5 TM Radiometric Calibration
Procedures and Postcalibration Dynamic Ranges. IEEE Transactions on Geoscience
and Remote Sensing, Vol. 41, No. 11 PART II, (November 2003), pp. 2674-2677, ISSN
0196-2892
Chander, G.; Haque, M.O.; Micijevic, E. & Barsi, J.A. (2010). A Procedure for Radiometric
Recalibration of Landsat 5 TM Reflective-Band Data. IEEE Transactions on Geoscience
and Remote Sensing, Vol. 48, No. 1 (January 2010), pp. 556-574, ISSN 0196-2892
Chavez Jr, P.S. (1989). Radiometric calibration of Landsat Thematic Mapper multispectral
images. Photogrammetric Engineering & Remote Sensing, Vol. 55, No. 9 (September
1989), pp. 1285 -1294, ISSN 0099-1112
Chavez Jr, P.S. (1996). Image-based atmospheric corrections - Revisited and improved.
Photogrammetric Engineering & Remote Sensing, Vol. 62, No. 9 (September 1996), pp.
1025-1036, ISSN 0099-1112
Clark, R.N. (1999). Chapter 1: Spectroscopy of Rocks and Minerals, and Principles of
Spectroscopy. In: Manual of Remote Sensing, Volume 3, Remote Sensing for the Earth
Sciences, A.N. Rencz, (Ed.), pp. 3-58. John Wiley and Sons, ISBN 0471294055, New
York, USA
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 137

Cohen, J. (1960). A Coefficient of Agreement for Nominal Scales. Educational and

Psychological Measurement, Vol. 20, No. 1 (April 1960), pp. 37-46. ISSN 0013-1644
Congalton, R.G. (2004). Putting the map back in map accuracy assessment. In: Remote
Sensing and GIS accuracy assessment, R.S. Lunetta & J.G. Lyon, (Eds.), pp. 1-11. CRC
Press, ISBN 156670443X, Boca Raton (FL), USA
Congalton, R.G.; Oderwald, R.G. & Mean, R.A. (1983). Assessing Landsat Classification
Accuracy Using Discrete Multivariate Analysis Statistical Techniques.
Photogrammetric Engineering & Remote Sensing, Vol. 49, No. 12 (December 1983), pp.
1671-1678, ISSN 0099-1112
Csillag, F.; Pasztor, L. & Biehl, L.L. (1993). Spectral band selection for the characterization of
salinity status of soils. Remote Sensing of Environment, Vol. 43, No. 3 (March 1993),
pp. 231-242, ISSN 0034-4257
Dalsted, K.J.; Worcester, B.K. & Brun, L.J. (1979). Detection of Saline Seeps by Remote
Sensing Techniques. Photogrammetric Engineering & Remote Sensing, Vol. 45, No. 3
(March 1979), pp. 285-291, ISSN 0099-1112
Dehaan, R.L. & Taylor, G. R. (2002). Field-derived spectra of salinized soils and vegetation
as indicators of irrigation-induced soil salinization. Remote Sensing of Environment,
Vol. 80, No. 3 (June 2002), pp. 406-417, ISSN 0034-4257
Dehaan, R.L. & Taylor, G.R. (2003). Image-derived spectral endmembers as indicators of
salinisation. International Journal of Remote Sensing, Vol. 24, No. 4 (February 2003),
pp. 775-794, ISSN 0143-1161
De Martonne, E. (1926). L’indice d'aridité. Bulletin de l’Association des geógraphes français, Vol.
9, pp. 3-5, ISSN 0004-5322
De Paz, J.M.; Sánchez, J. & Visconti, F. (2006). Combined use of GIS and environmental
indicators for assessment of chemical, physical and biological soil degradation in a
Spanish Mediterranean region. Journal of Environmental Management, Vol. 79, No. 2
(April 2006), pp. 150-62, ISSN 03014797
Dutkiewicz, A.; Lewis, M. & Ostendorf, B. (2009). Evaluation and comparison of
hyperspectral imagery for mapping surface symptoms of dryland salinity.
International Journal of Remote Sensing, Vol. 30, No. 3 (February 2009), pp. 693-719,
ISSN 0143-1161
Dwivedi, R.S.; Ramana, K.V.; Thammappa, S.S. & Singh, A. N. (2001). The Utility of IRS-1C
LISS-Ill and PAN-Merged Data for Mapping Salt-Affected Soils. Photogrammetric
Engineering & Remote Sensing, Vol. 67, No. 10 (October 2001), pp. 1167-1175, ISSN
0099-1112
European Commission. (2002). Towards a Thematic Strategy for Soil Protection. COM(2002) 179
final. European Commission, Brussels, Belgium.
European Commission. (2003). Extent, causes, pressures, strategies and actions that should be
adopted to prevent and to combat salinization and sodification in Europe. Directorate
General Environment, Directorate B, Erosion Working Group (Task 5; Topic:
Salinization and Sodification), Brussels, Belgium.
FAO. (2002). Crops and Drops, making the best use of water for agriculture, Food and Agriculture
Organization of the United Nations, Rome, Italy.
Farifteh, J.; Van Der Meer, F.; Van Der Meijde, M. & Atzberger, C. (2008). Spectral
characteristics of salt-affected soils: A laboratory experiment. Geoderma, Vol. 145,
No. 3-4 (June 2008), pp. 196-206, ISSN 0016-7061
138 Remote Sensing – Applications

Ferreira, M.E.; Ferreira, L.G.; Sano, E.E. & Shimabukuro, Y.E. (2007). Spectral linear mixture
modelling approaches for land cover mapping of tropical savanna areas in Brazil.
International Journal of Remote Sensing, Vol. 28, No. 2 (January 2007), pp. 413-429,
ISSN 0143-1161
Foody, G.M. (2002). Status of land cover classification accuracy assessment. Remote Sensing of
Environment, Vol. 80, No. 1 (April 2002), pp. 185-201, ISSN 0034-4257
Franco-Lopez, H.; Ek, A.R., & Bauer, M.E. (2001). Estimation and mapping of forest stand
density, volume, and cover type using the k-nearest neighbors method. Remote
Sensing of Environment, Vol. 77, No. 3 (September 2001), pp. 251-274, ISSN 0034-4257
Ghrefat, H.A. & Goodell, P.C. (2011). Land cover mapping at Alkali Flat and Lake Lucero,
White Sands, New Mexico, USA using multi-temporal and multi-spectral remote
sensing data. International Journal of Applied Earth Observations and Geoinformation,
Vol. 13, No. 4 (August 2011), pp. 616-625, ISSN 0303-2434
Green, A.A.; Berman, M.; Switzer, P. & Craig, M.D. (1988). A transformation for ordering
multispectral data in terms of image quality with implications for noise removal.
IEEE Transactions on Geoscience and Remote Sensing, Vol. 26, No. 1 (January 1988),
pp. 65-74, ISSN 0196-2892
Haapanen, R.; Ek, A.R.; Bauer, M.E. & Finley, A.O. (2004). Delineation of forest/nonforest
land use classes using nearest neighbor methods. Remote Sensing of Environment,
Vol. 89, No. 3 (February 2004), pp. 265-271, ISSN 0034-4257
Hernández Bastida, J.A.; Vela de Oro, N. & Ortiz Silla, R. (2004). Electrolytic Conductivity of
Semiarid Soils (Southeastern Spain) in Relation to Ion Composition. Arid Land
Research and Management, Vol. 18, No. 3 (July 2004), pp. 265-281, ISSN 15324982
Howari, F.M., Goodell, P.C. & Miyamoto, S. (2000). Spectral properties of salt crusts formed
on saline soils. Journal of Environmental Quality, Vol. 31, No. 5 (September 2002), pp.
1453-61, ISSN 00472425
ITT VIS. (2008). ENVI 4.5 User’s Guide. ITT Visual Information Solutions Boulder (CO), USA.
Jensen, J.R. (2005). Introductory Digital Image Processing (3rd Edition). Prentice Hall, ISBN
0131453610, Upper Saddle River (NJ), USA
Jensen, J. R. (2007). Remote Sensing of the Environment: An Earth Resource Perspective (2nd
edition), Pentice Hall, ISBN 0131889508, Upper Saddle River (NJ), USA
Kaufman, Y.J. (1989). The atmospheric effect on remote sensing and its corrections. In:
Theory and Applications of Optical Remote Sensing, G. Asrar, (Ed.), pp. 336-428, Wiley-
Interscience, ISBN 0471628956, New York, USA
Lal, R., Mokma, D. & Lowery, B. (1999). Relation between soil quality and erosion. In: Soil
quality and soil erosion, R. Lal, (Ed.), pp. 237-258, CRC Press, ISBN 1574441000, Boca
Raton (FL), USA
Landis, J.R. & Koch, G. G. (1977). The measurement of observer agreement for categorical
data. Biometrics, Vol. 33, No. 1 (March 1977), pp. 159-174, ISSN 0006341X
Lillesand, T.M.; Kiefer, R.W. & Chipman, J.W. (2003). Remote Sensing and Image Interpretation
(5th edition). John Wiley and Sons, ISBN 9812530797, Hoboken (NJ), USA
Majaliwa, J.G.M.; Twongyirwe, R.; Nyenje, R.; Oluka, M.; Ongom, B.; Sirike, J.; Mfitumukiza,
D.; Azanga, E.; Natumanya, R.; Mwerera, R. & Barasa, B. (2010). The Effect of Land
Cover Change on Soil Properties around Kibale National Park in South Western
Uganda. Applied and Environmental Soil Science, Article ID 185689, ISSN 1687-7667
Mapping Soil Salinization of Agricultural Coastal Areas in Southeast Spain 139

Mather, P.M. (2004). Computer Processing of Remotely-Sensed Images: An Introduction (3rd

edition). Wiley, ISBN 0470849185, Chichester, UK
Melendez-Pastor, I.; Navarro-Pedreño, J.; Koch, M., & Gómez, I. (2010). Applying imaging
spectroscopy techniques to map saline soils with ASTER images. Geoderma, Vol.
158, No. 1-2 (August 2010), pp. 55-65, ISSN 0016-7061
Melendez-Pastor, I.; Navarro-Pedreño, J.; Koch, M., & Gómez, I. (2010). Multi-resolution and
temporal characterization of land use classes in a Mediterranean wetland with land
cover fractions. International Journal of Remote Sensing, Vol. 31, No. 20 (October
2010), pp. 5365-5389, ISSN 0143-1161
Mermut, A.R., & Eswaran, H. (2001). Some major developments in soil science since the
mid-1960s. Geoderma, Vol. 100, No. 3-4 (May 2001), pp. 403 -426, ISSN 0016-7061
Metternicht, G.I. & Zinck, J.A. (2003). Remote sensing of soil salinity: Potentials and
constraints. Remote Sensing of Environment, Vol. 85, No. 1 (April 2003), pp. 1-20,
ISSN 0034-4257
Metternicht, G. & Zinck, J.A. (1997). Spatial discrimination of salt- and sodium-affected soil
surfaces. International Journal of Remote Sensing, Vol. 18, No. 12 (June 1997), pp.
2571-2586, ISSN 0143-1161
Moran, M.S.; Jackson, R.D.; Slater, P.N. & Teillet, P.M. (1992). Evaluation of simplified
procedures for retrieval of land surface reflectance factors from satellite sensor
output. Remote Sensing of Environment, Vol. 41, No. 1-2 (August-September 1992),
pp. 169-184. ISSN 0034-4257
Mougenot, B.; Pouget, M. & Epema, G.F. (1993). Remote sensing of salt affected soils. Remote
Sensing Reviews, Vol. 7, No. 3-4, pp. 241-259, ISSN 02757257
Mulder, V.L.; De Bruin, S.; Schaepman, M.E. & Mayr, T. R. (2011). The use of remote sensing
in soil and terrain mapping—A review. Geoderma, Vol. 162, No. 1-2 (April 2011), pp.
1-19, ISSN 0016-7061
Nelson, D.W. & Sommers, L. E. (1982). Total carbon, organic carbon, and organic matter. In:
Methods of Soil Analysis. A.L. Page; R.H. Miller & D.R. Keeney (Eds.), pp. 539-579,
American Society of Agronomy (ASA),Madison (WI), USA
Pérez-Sirvent, C.E.; Martínez-Sánchez, M.J.; Vidal, J. & Sánchez, A. (2003). The role of low-
quality irrigation water in the desertification of semi-arid zones in Murcia, SE
Spain. Geoderma, Vol. 113, No. 1-2 (April 2003), pp. 109-125, ISSN 0016-7061
Qadir, M.; Ghafoor, A., & Murtaza, G. (2000). Amelioration strategies for saline soils: A
review. Land Degradation & Development, Vol. 11, No. 6 (November-December 2000),
pp. 501-521, ISSN 1085-3278
Quarmby, N.A.; Townshend, J.R.G.; Settle, J.J.; White, K.H.; Milnes, M.; Hindle, T.L. &
Silleos, N. (1992). Linear mixture modelling applied to AVHRR data for crop area
estimation. International Journal of Remote Sensing, Vol. 13, No. 3 (February 1992),
pp. 415-425, ISSN 0143-1161
Reddy, K.R.; D’Angelo, E.M. & Harris, W.G. (2000). Biogeochemistry of Wetlands. In:
Handbook of Soil Science, M.E. Sumner, (Ed.), p. G-89-G-119, CRC Press, ISBN
0849331366, Boca Raton (FL), USA
Richards, L.A. (1954). Diagnosis and improvement of saline and alkali soils. Agriculture Handbook
No. 60. United States Department of Agriculture (USDA), Washington D.C., USA.
140 Remote Sensing – Applications

Richardson, A.J.; Gerbermann, A.H.; Gausman, H.W. & Cuellar, J.A. (1976). Detection of
Saline Soils with Skylab Multispectral Scanner Data. Photogrammetric Engineering &
Remote Sensing, Vol. 42, No. 5 (May 1976), pp. 679-684, ISSN 0099-1112
SPSS (2009). PSAW 18.0 for Windows Help System. IBM Corpotation, Armonk (NY), USA
Samaniego, L. & Schulz, K. (2009). Supervised Classification of Agricultural Land Cover
Using a Modified k-NN Technique (MNN) and Landsat Remote Sensing Imagery.
Remote Sensing, Vol. 1, No. 4 (December 2009), pp. 875-895, ISSN 2072-4292
Schmid, T.; Koch, M. & Gumuzzio, J. (2009). Applications of hyperspectral imagery to soil
salinity mapping. In: Remote Sensing of Soil Salinization: Impact and Land
Management, G. Metternicht & J.A. Zinck, (Eds.), pp. 113-140. CRC Press, ISBN
1420065025, Boca Raton (FL), USA
Settle, J.J. & Drake, N.A. (1993). Linear mixing and the estimation of ground cover
proportions. International Journal of Remote Sensing, Vol. 14, No. 6 (March 1993), pp.
1159-1177, ISSN 0143-1161
Shimabukuro, Y.E. & Smith, J.A. (1991). The least-squares mixing models to generate
fraction images derived from remote sensing multispectral data. IEEE Transactions
on Geoscience and Remote Sensing, Vol. 29, No. 1 (January 1991), pp. 16-20, ISSN 0196-
2892
Sitte, P.; Ziegler, H.; Ehrendorfer, F. & Bresinsky, A. (1994). Strasburger. Tratado de Botánica
(8th Spanish edition), Omega, ISBN 8428209790, Barcelona, Spain
Small, C. & Lu, J.W.T. (2006). Estimation and vicarious validation of urban vegetation
abundance by spectral mixture analysis. Remote Sensing of Environment, Vol. 100,
No. 4 (February 2006), pp. 441-456, ISSN 0034-4257
Soil Survey Staff. (2006). Keys to Soil Taxonomy (10th Edition).United States Department of
Agriculture (USDA), Washington D.C., USA
Song, C.; Woodcock, C.E.; Seto, K.C.; Lenney, M.P. & Macomber, S.A. (2001). Classification
and Change Detection Using Landsat TM Data: When and How to Correct
Atmospheric Effects? Remote Sensing of Environment, Vol. 75, No. 2 (February 2001),
pp. 230-244, ISSN 0034-4257
Townsend, P.A. & Walsh, S.J. (2001). Remote sensing of forested wetlands: application of
multitemporal and multispectral satellite imagery to determine plant community
composition and structure in southeastern USA. Plant Ecology, Vol. 157, No. 2
(December 2001), pp. 129-151, ISSN 1385-0237
UNEP. (1997). World Atlas of Desertification (2nd edition). United Nations Environment
Programme (UNEP), ISBN 0340691662, Nairobi, Kenya
Walkley, A. & Black, I.A. (1934). An Examination of the Degtjareff Method for Determining
Soil Organic Matter, and A Proposed Modification of the Chromic Acid Titration
Method. Soil Science, Vol. 37, No. 1 (January 1934), pp. 29-38, ISSN 0038-075X
Weng, Y.; Gong, P. & Zhu, Z. (2008). Reflectance spectroscopy for the assessment of soil salt
content in soils of the Yellow River Delta of China. International Journal of Remote
Sensing, Vol. 29, No. 19 (October 2008), pp. 5511-5531, ISSN 0143-1161
 6

Remote Sensing Based

Modeling of Water and Heat
Regimes in a Vast Agricultural Region
A. Gelfan1, E. Muzylev1, A. Uspensky2, Z. Startseva1 and P. Romanov3
1Water Problem Institute of Russian Academy of Sciences, Moscow
2State Research Center of Space Hydrometeorology Planeta, Moscow
3City College of City University of New York, New York, NY
1,2Russia
3USA

1. Introduction
In many parts of the world, solution of the problem of water security is closely associated
with possibility and predictability of access to soil water which accounts, globally, for
approximately 60% of total rainfall. Soil water, so-called “green water”, is fundamentally
significant for maintaining sustainability of ecosystems and providing water for agriculture
(in average, agriculture demands about 70% of global water use, and over 90% in some
regions (Global Water Security, 2010). In addition, changes in the store of soil water due to
evapotranspiration, infiltration of surface water and other water cycle processes strongly
affect land-atmosphere interactions on intraseasonal to interannual timescales (Yang, 2004)
and understanding physical mechanisms of these processes is central for effective modeling
mesoscale atmospheric circulation.
Taking into account a vital role of “green water” both for ecosystems and agriculture, there
is a need for physically based models allowing to describe land-surface processes in their
interaction with atmosphere and to predict soil water availability under changing
environment and climate. Processes of mass and energy exchange at the land surface,
linking the atmosphere and soil water are of interest to several geophysical disciplines;
particularly, the possibility of developing adequate models of these processes was studied
intensively by hydrological and meteorological communities for a few decades. Recently, a
huge contribution to our understanding these processes has been made by remote sensing
community. Combination of these efforts results in intensive developing so called land-
surface models (LSMs), considering as a key tool to predict successfully the likely future
states of the terrestrial systems under anthropogenic pressure and climate change. There are
a lot of reviews providing a detailed and comprehensive discussion of LSMs (e.g. Sellers et
al., 1997; Pitman, 2003; Overgaard et al., 2006). Most of them are concentrated on the
development of LSMs designed for use in weather and climate models. Below a review of
LSMs will be done with emphasis on their availability to capture processes controlling soil
water dynamics, and primarly, evapotranspiration. Additionally, focus will be on utilization
of remotely sensed data in LSMs.
142 Remote Sensing – Applications

According to Sellers et al. (1997), the LSMs could be broadly divided onto three generations
distinguishing largely on the level of complexity of the evapotranspiration process
description. The first generation, developed in the late 1960s and 1970s, used simple bulk
aerodynamic transfer formulations. Land surface considers as homogeneous one in the first-
generation LSMs and spatially uniform surface parameters (water-holding capacity, albedo,
roughness length etc.) are used, i.e. these models do not discriminate soil evaporation and
transpiration. Examples are the “bucket” model by Manabe (1969) (probably, the earliest
first-generation LSM) and more recent TOPUP model by Schultz et al. (1998), PROMET
model by Mauser & Schädlich (1998). Use of the single value of the aerodynamic resistance
regardless of the surface type is one of the common simplifications and major conceptual
limitations of the first-generation LSMs (Pitman, 2003). Another one is that these models
include a single soil layer for soil moisture and temperature simulations. According to the
findings of the Project for the Intercomparison of Landsurface Parameterisation Schemes
(PILPS; Henderson-Sellers et al., 1995), soil moisture dynamics simulated by the first-
generation LSMs can not be often reproduced adequately by these models because of
aforementioned conceptual limitations.
The second generation of the LSMs, developed in the early 1980s, explicitly represented, in
contrast to the first-generation models, the influence of vegetation on the interaction
between land surface and atmosphere. Taking into account difference between soil and
vegetation provided an opportunity to begin integrating satellite data into LSMs (Pitman,
2003). Also, these models improve soil moisture representation replacing the simple
conceptual scheme of Manabe (1969) by more sophisticated parametrizations (particularly,
the vertically distributed Richards equation is often used). Such a new direction in land
surface modeling took roots from a pioneering work of Deardorff (1978) developed a model
for simulating soil moisture and heat transfer in two layers and vegetation as a single layer
controlling heat balance on a land surface, as well as from the works of Dickinson (1986) and
Sellers et al. (1986) developed the Biosphere Atmosphere Transfer Scheme (BATS) and
Simple Biosphere Model (SiB) based on the ideas of Deardorff (1978). During the subsequent
decades, a huge number of the second-generation LSMs have been developed; examples
include SVAT (Kuchment & Startseva, 1991), VIC (Wood et al., 1992), BASE (Desborough &
Pitman, 1998), CLM (Dai et al., 2002), SWAP (Gusev & Nasonova, 2002), LaD (Milly &
Shmakin, 2002).
One of the principal advantages of the second-generation LSMs is their ability to consider
snow processes affecting a major part of terrestrial water balance in cold regions. Snow sub-
models of different vertical discretization have been implemented in LSMs and an intensive
work on their evaluation and intercomparison (e.g. the recent SnowMIP2 Project (Rutter et
al., 2009)) has been made. Additionally, some LSMs include parametrization of
hydrothermal processes in a frozen soil (Wood et al., 1992; Gusev & Nasonova, 2002 among
others) but an adequate description of these processes strongly affected soil moisture
content before vegetation season still remains problematic (Pitman, 2003).
A step forward of the third-generation LSMs is in explicit description of a canopy
physiology including biophysical mechanisms of stomatal conductance, photosynthesis,
plant growth, etc. This ability opens up new opportunities not only to improve reproduction
of evapotranspiration but, importantly, to address carbon exchange by plant. Description of
the other land surface processes (soil moisture and temperature dynamics, snow processes,
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 143

etc.) is similar to one utilized in the second-generation LSMs. One of the first LSMs of third
generation were developed by Collatz et al. (1991) and Sellers et al. (1992), some recent
examples representing improvements of the aforementioned second-generation LSMs
include modifications of the SVAT-model (Kuchment et al., 2006), the CLM-model (Oleson
et al., 2008), the SiBcrop model (Lokupitiya et al., 2009).
One of the major concerns restricting availability of LSMs, particularly on a regional scale,
is the issue of spatial heterogeneity of land surface characteristics (soil, vegetation,
topography) required for assigning the model parameters. Regional (mesoscale)
heterogeneity is not captured by the existing ground-based observational network that
leads to an excessive aggregation of the parameters and, as a rule, to decrease in accuracy
of reproducing spatial distribution of the desired processes. Significant improvement of
performance of LSMs can be reached by assimilation of information that is additional to
the ground observations, first of all, information on land surface provided by satellite
remote sensing. Remote sensing allows substituting the missed ground observations by
measurements of the incoming and outgoing land surface radiation fluxes conversed into
physical distributed parameters. No exaggeration to say that spatial data on land surface
derived from remote sensing is the only source of the distributed parameters for LSMs at
regional scale. Additionally, these data can be used for model evaluation purposes.
Applicability of satellite remote sensing for improvement of the land surface modeling is
reviewed, e.g. by Overgaard et al. (2006), including results of a number of field
experiments (FIFE, HAPEX, KUREX and others). However, in spite of the fact that
quantity and quality of satellite products have largely increased for the last decade and
they have recognized as a potentially valuable source of distributed information, the
majority of satellite products still needs considerable improvement and applicability of
these products, even being improved, should be verified both for the specific region and
for the used LSM.
This paper has three major objectives. First, the existing satellite-derived data of land surface
and snow cover characteristics will be overviewed in brief and the specific developed
technologies of the satellite data thematic processing will be presented for the study region
located within the agricultural Black Earth area of the European Russia. Secondly, structure
of the physically based distributed Remote Sensing Based Land Surface Model (RSBLSM)
developed for simulation of vertical water-and-heat transfer in vegetation, unfrozen and
frozen soil, snow cover will be described and the results of its testing against the available
ground-based observations will be shown. Finally, the results of utilizing satellite-derived
land surface and snow characteristics as the parameters and input variables of RSBLSM, as
well as for the model evaluation, will be demonstrated and discussed.

2. Case study
The case study has been carried out for agricultural Black Earth region of the European
Russia of 227,300 km2 located in the steppe-forest physiographic zone (Fig. 1)
Relief of the region is low middle-hilly plain dissected by broad river valleys, ravines and
gullies. Dominant slopes are ~1-4о. Absolute elevation marks of surface in the region are in
the range of 150-260 m. Annual net radiation is 27-32 kcal/cm2, and the sum of incoming
solar radiation during the summer months is 41-44 kcal/cm2. Annual precipitation is 519
144 Remote Sensing – Applications

mm, over 40% of which falls as snow. Snow water equivalent significantly varies over the
area. At the northern part of the region snow cover deceases in the middle of April, southern
parts are snow-free in March. Maximum rainfall (60-70 mm) falls as a rule in July. Ground
water level lies at the depth of 15-30 m between rivers and of 3-5 m in river valleys.
Soils are mainly chernozem (podzolized, leached and typical), small part of the territory is
occupied by floodplain meadow and gray soil. In texture, soils mostly relate to the loam.
Sandy loam and sand are found rarer. In the north-western part of the region gray forest-
steppe soils with patches of degraded chernozem are located. When moving to the south
gray soils give place to leached chernozem often occurring in combination with powerful
chernozem. The southeastern part of the region is occupied by ordinary chernozem. Most
of the region territory is under cultivation, the natural vegetation is preserved mainly in
river valleys and on slopes of gullies and ravines. Plough-lands compose the most part of
the region (78 %), forests occupy about 5%; pastures take up about 16%; urbanized lands
occupy less than 1%. The main crops are cereals (spring wheat and barley, winter wheat,
less corn, buckwheat and rye) as well as sugar beets, potatoes and forage grasses. Ratio of
crops in different households differs substantially. Region-averaged grain wedge is about
60% of all plough-lands. In the region there are 48 agricultural meteorological stations at
which observations on meteorological characteristics and soil moisture under different
crops, as well as snow cover characteristics are conducted. At several of them
measurements of the evaporation pans are also made.

Fig. 1. Study region (points at the right upper map represent location of meteorological
stations)

To assign the model parameters, most of which are the measured soil, vegetation and snow
cover characteristics as well as to calibrate and validate the model the data have been
attracted of above observations at agricultural meteorological stations for 21 years from 1971
to 2010.
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 145

3. Remote sensing of agricultural regions: products and algorithms of data

processing
3.1 Remote sensing of land surface and vegetation characteristics
This section contains a brief overview of satellite instruments and methods for remote
sensing of various land surface parameters, valuable in particular for hydrological
applications. During the past decades, a series of sensors have been developed and
launched, such as the Advanced Very High Resolution Radiometer (AVHRR) and the
MODerate resolution Imaging Spectroradiometer (MODIS), which are respectively onboard
the polar-orbiting satellites NOAA and NASA EOS Terra and Aqua. The AVHRR
instrument has 6 VIS and IR channels with spatial resolution of 1 km (NOAA KLM Users
Guide, 2005) that are informative with respect to different land surface and vegetation
parameters. The MODIS instrument has 36 channels in VIS and IR (Justice et al., 1998) with 1
km spatial resolution, which provides information on even more geophysical parameters of
land surface, vegetation, atmosphere, etc. In recent years new geostationary satellites have
been developed, such as European Meteosat Second Generation (MSG), i.e. Meteosat-8 and -
9. Meteosat main payload is the optical imaging radiometer, the so called Spinning
Enhanced Visible and Infrared Imager (SEVIRI). SEVIRI has 12 spectral channels from VIS to
IR (Schmetz et al., 2002). It has channels similar to AVHRR, and the benefit is that SEVIRI
provides measurements of the Earth-disc every 15 minutes comparing to the coverage of the
polar-orbiting satellites (2 times a day for the same territory).
Among the above mentioned geophysical parameters we will focus on: land surface
temperature ( LST or Tls ), air foliage temperature ( Ta ), emissivity ( E ), as well as
normalized difference vegetation index ( NDVI ), vegetation cover fraction ( B ) and leaf area
index ( LAI ). LST is one of the important geophysical parameters. Together with the land
surface spectral emissivity ( LSE ) the LST affect the heat and water transport between the
surface and the atmosphere. There is a strong need in the remote sensing LST , since the
conventional surface temperature observations are rather sparse (in space and time).
The possibilities of extracting LST and LSE information from thermal IR multichannel
measurements in the “atmospheric window” spectral range (3.7-4.0, 10.5-12.5 µm) has been
the subject of numerous investigations during last 20 years, see (Becker & Li, 1995).
Because of the land surface heterogeneity, the satellite measurements usually come from
mixed pixels. At satellite pixel scale, LSE refers to the area-weighted and channel-averaged
emissivity (E) and LST refers to the radiometric surface temperature corresponding to the
Field Of View (FOV) of a radiometer (Becker & Li, 1995).

3.1.1 AVHRR- and MODIS-based remote sensing products

The methods have been developed and tested for AVHRR/NOAA data processing (Muzylev
et al, 2002, 2005) that provide the retrieval of two types of LST (efficient radiation temperature
Ts.eff and land skin temperature Tg ) and emissivity E as well as the derivation of air-
vegetation temperature Ta and three vegetation characteristics, namely, vegetation index
NDVI , fraction B and leaf area index LAI . The algorithms for AVHRR-based estimation of
Ts.eff , Tg , Ta utilize cloud-free measurements in the split window channels 4 and 5 and linear
statistical regression similar to well-known split window technique (Becker et al., 1995; PUM
146 Remote Sensing – Applications

LST, 2008; Wan et al., 1996). The values of emissivity E4 and E5 are specified a priori using
one of alternate approaches. The required ancillary information is extracted from the
classification-based emissivity model (Snyder et al., 1998) as well as from the empirical
relationships between the emissivity and/or NDVI / B , see (Muzylev et al., 2002, 2005). To
estimate LAI , the empirical relationships between LAI and NDVI (established for different
land covers) have been applied (see Section 5.1, formulas (27) and (28)).
The threshold technique of cloud detection in the AVHRR FOV has been used that has
allowed increasing the reliability of cloud-free fragment detection (Volkova & Uspenskii,
2007). The developed software package has been applied for AVHRR/NOAA cloud-free
data thematic processing to generate named remote sensing products and
cloud/precipitation parameters for various dates of the 1999-2010 vegetation periods. The
error statistics of Ta , Tsg and Ts.eff derivation has been investigated for various samples
using comparison with synchronous collocated in-situ measurements that has given root-
mean square (RMS) errors in the range of 1.5-2.0, 3.5-4.5, and 2.5-3.5C respectively
(Muzylev et al., 2005, 2006, 2010). The archive of synchronous AVHRR/NOAA
measurements, remote sounding data, and in-situ hydro-meteorological observations has
been compiled for the study area and its separate parts for 1999-2010 vegetation seasons.
The dataset of MODIS-based remote sensing products has also been compiled on the base of
special technology using LP DAAC web-site https://lpdaac.usgs.gov, that includes
estimates of land surface temperature (LST) Tls , E , NDVI , LAI for the region of interest
and 2003-2010 vegetation seasons. Two types of MODIS-based Tls and E estimates have
been extracted (for separate dates of the named time period): LST/E Daily L3 product
(MOD11В1) with spatial resolution ~ 4.8 km and LST/E 5-Min L2 product (MOD11_L2)
with spatial resolution ~ 1 km. The verification of Tls estimates has been performed by the
comparison against analogous and collocated AVHRR-based ones (Muzylev et al, 2010).

3.1.2 SEVIRI-based land surface and land air surface temperature estimates
In the recent years there were a lot of studies on LST derivations from SEVIRI/MSG data,
see (PUM LST, 2008; Solovjev et al., 2009, 2010). In the State Research Center of Space
Hydrometeorology “Planeta” (Moscow, Russia) the new methodology has been developed
for the derivation of LST and LSE from cloud-free brightness temperatures measured in
the SEVIRI channels 9 (10.8 µm) and 10 (12.0 µm) at three different times based on
combination of two well-known techniques, i.e. split-window method (Wan et al., 1996) and
two temperature method, see (Faysash et al., 2000) with additional hypothesis that the
emissivity values E9 and E10 remain constant during the time interval between the first and
the last image cycles used. The detailed description of the proposed technique can be found
in (Uspensky et al., 2009; Solovjev et al., 2010). It is important to mention that unlike the
technique from (PUM LST, 2008), this method does not require the accurate knowledge of
emissivity (E) in the split-window channels.
The method described has been used for the period of 2009-2011 to produce LST estimates
over Europe and, in particular, for Black Earth zone of Russia. An example image of this
LST mapping is presented in Fig. 2.
The comparison with synchronous and collocated LST products from Satellite Application
Facility on Land Surface Analysis, SAF LSA (Lisbon, Portugal) (PUM LST, 2008) has been
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 147

Fig. 2. LST map, 27.09.2011, 07:00 UTC

performed to validate the results over the central Europe. Good level of correlation has been
reached, which could be treated as indirect proof of the method’s efficiency. RMS deviation
between the above mentioned LST estimates is in the range of 0.9-3.0C. The upper limit
could be shifted down to 2.4C by subtracting systematic biases.
An additional validation has been performed through the inter-comparison with MODIS-
based LST for the study region, see Fig. 3.
LST/MODIS 2010 .07.01, 12:00 UTC
LST/ SEVIRI
2010.07 .0 1, 1 2:00 -13 :0 0 UTC

Kurs k Kur sk

44.2°
T s.MOD IS – Ts.S EVIR I
35°E 36°E 37°E 38°E 39°E 40°E 41 °E 42°E 43 °E 23.7°
54°N

Sample size: 4341 pai rs

53°N

52°N

Kursk
51°N
bi as
0 -2 56%
2 -4 30%
50°N 4 -9 14%

Fig. 3 Inter-comparison results for MODIS- and SEVIRI-based LST estimates

It is obvious that both LST maps are similar to each other. The discrepancy is rather small
(0.0-2.0C) for the most of the territory.
148 Remote Sensing – Applications

Along with this, a very important parameter such as land surface-air temperature Ta can
also be derived from SEVIRI data. A new method of Ta derivation has been developed in
(Uspensky et al., 2011). A multiple linear regression model has been constructed that
estimates air temperature from satellite-observed LST , solar zenith angle (or related day of
the year number), and land elevation. Land surface temperatures estimates from above were
used in this scheme to calculate regression coefficients, as well as time-synchronous
collocated in-situ measurements from the ground observation network. The development
and validation experiments were carried out for the study area for the vegetation seasons of
2006-2009. Data from the above 48 agricultural meteorological stations were used for 8
standard synoptic times a day. All the data were subdivided on separate training and
validation data sets. For vegetation season of 2009, an overall bias and standard deviation of
calculations are approximately 0 and 1.9-2.1C, respectively. The performance of the method
is similar to the one presented in (Good, 2009).

3.2 Remote sensing of snow cover

Seasonal snow cover is among the most important factors for mid-latitude agriculture.
Water accumulated in the snow pack in winter and released through the snowmelt is critical
for crop development in early spring. For winter grain crops that are widely grown in
Canada, Russia and Ukraine, snow pack presents an insulation material protecting the
plants from freezing temperatures. Accurate information on the snow cover extent and
variations during the winter season helps to identify areas of potential winterkill and to
predict adverse conditions for crops development (e.g. Romanov, 2011).
Although most information on the snow cover distribution and properties for agriculture
is traditionally obtained from in-situ observations at ground-based meteorological
stations, the use of satellite-based snow products is becoming increasingly popular. First,
this section presents a brief overview of current satellite-based snow mapping techniques
and of snow products that can be used in agricultural applications. Then, a new technique
used in this study and based on synergy of optical and microwave snow cover products
available from MODIS and Advanced Microwave Sounding Radiometer (AMSR-E)
instruments onboard EOS satellites Terra and Aqua is described.

3.2.1 Existing snow mapping techniques and products

To derive information on snow cover from satellite data a number of different techniques,
both interactive and automated has been developed and is actively used. The most popular
interactive snow product based on visual inspection of satellite optical imagery is Northern
Hemipshere snow charts that have been generated by NOAA since 1972. Interactive maps of
snow cover are currently produced within a computer-based Interactive Multisensor Snow
and Ice Mapping System (IMS) that provides software tools and access to various datasets to
facilitate the image analysis and map drawing by human analysts (Helfrich et al, 2007).
Daily IMS snow maps are generated at 4 km resolution and are available from the National
Snow and Ice Data Center (NSIDC) at http://nsidc.org/data/g02156.html. Despite some
weaknesses associated primarily with subjectivity in the image analysis and interpretation
NOAA IMS snow maps present a robust and consistent product. The overall accuracy of
daily interactive snow maps is quite high with the yearly mean rate of agreement with
surface observations data exceeding 90% (e.g., Brubaker et al., 2005). Most errors occur
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 149

during fast and large-scale snow advance or retreat or when persistent cloudiness obscures
the land surface.
In contrast to interactive techniques the interest to automated algorithms is attracted due to
their low exploitation costs and ability to better utilize potentials of satellite data,
particularly their high spatial resolution and multi-spectral capability. The two major
techniques for mapping and monitoring snow cover from satellites are based,
correspondingly on passive observations in the microwave and in the optical (visible to
infrared) spectral range.
The advantage of using microwave sensors consists in their ability to “see” through most
clouds and to provide information (although quite limited) on snow depth and snow water
equivalent (SWE). The primary limitations of microwave measurements are associated with
their coarse spatial resolution of 25-50 km, poor sensitivity to shallow and melting snow
(Walker and Goodison, 1993) and difficulty to distinguish between snow and frozen rocks
and soil (Grody and Basist, 1996). Snow depth and SWE estimates from observations in the
microwave are limited only to dry snow packs while corresponding retrieval errors range
typically between 50 and 100% (e.g. Kelly et al., 2003). Global monitoring of snow with
microwave sensors data started in 1978 with the launch of Nimbus-7 with Scanning
Multichannel Microwave Radiometer (SMMR) onboard and continued with a number of
other sensors, including in particular, Special Sensor Microwave Imager (SSM/I) on Defense
Meteorological Satellite Program (DMSP) satellites since 1987 (Armstrong and Brodzik 2005)
and AMSR-E onboard EOS Aqua satellite since 2002 (Kelly et al, 2003).
As compared to satellite passive microwave measurements, observations in the optical spectral
range allow for more accurate mapping of snow cover at higher spatial resolution. The mean
accuracy of snow identification in optical bands usually exceeds 90%, but drops to 80-90% over
dense coniferous forests (Simic, et al, 2004, Hall & Riggs, 2007). Daily global snow cover maps
routinely generated with data from MODIS onboard NASA EOS Terra and Aqua satellites at
500 m spatial resolution (Hall et al, 2002) and from AVHRR onboard NOAA satellites at 4 km
resolution (http://www.star.nesdis.noaa.gov/smcd/emb/snow/HTML/snow.htm). It is
important that snow retrievals in the optical spectral bands are possible only under clear sky
conditions. Partial improvement in the map area coverage can be achieved with geostationary
satellites which provide multiple observations per day and hence increase the chance to see the
land surface cloud clear (e.g, De Wildt et al., 2007). With geostationary satellites, however, the
map coverage is only regional and is limited to the area within ~ 65° N and S.
Because of physical limitations of both principal snow remote sensing techniques, snow
products generated with single sensor data lack either continuity or sufficient accuracy and
spatial resolution and thus are hard to use in numerical model applications. In an attempt to
improve satellite-based snow cover characterization several techniques have been proposed
that combine snow cover observations in the optical and microwave spectral bands (e.g.,
Romanov et al. (2000), Brodzik et al. (2007), Foster et al. (2011)). The objective of these
techniques is to maximize advantages offered by optical and microwave observations, to
compensate for their weaknesses and to generate continuous snow maps at the highest
possible spatial and temporal resolution. Most often in these algorithms optical snow
retrievals are used in clear sky conditions, whereas microwave retrievals complement the
optical data when cloudy.
150 Remote Sensing – Applications

3.2.2 Snow cover mapping through synergy of optical and microwave products from
EOS satellites
A new technique used in this study is based on synergy of optical and microwave snow
cover products available from MODIS and AMSR-E instruments onboard EOS satellites
Terra and Aqua. The objective was to generate an advanced product providing continuous
(gap free) characterization of the global snow cover distribution at 5 km spatial resolution at
daily time step.
The algorithm utilizes two NASA daily snow products, MODIS snow cover map on a
latitude-longitude grid at 5 km resolution (labeled by NASA as MOD10C1 and MYD10C1
correspondingly for MODIS Terra and Aqua) and Aqua AMSR-E–based snow water
equivalent product AE_DySno. In the developed blending technique we took a cautious
approach to the microwave data: microwave retrievals indicating no snow as well as
retrievals over mountains were disregarded due to frequent omission of melting snow and
shallow snow and frequent overestimates of snow cover in the mountains by microwave
algorithms. The remaining microwave retrievals were used to complement snow cover
distribution mapped with MODIS data in clear sky conditions. Within this approach some
pixels in the daily map may remain undetermined. To eliminate these gaps in the coverage
and to achieve continuity in the derived snow cover distribution pixels that remain
undetermined in the current day snow map were filled in with the data from the previous
day’s blended snow map.
All available MODIS and AMSR-E snow products since 2002 have been reprocessed to
derive almost 10-years-long time series of daily global snow cover maps. Snow maps in
binary format are available at ftp://www.orbit.nesdis.noaa.gov/ pub/ smcd/ emb/ snow/
eos/. An example of a snow cover map generated through synergy of MODIS and AMSR-E
data is presented in Fig. 4. To estimate the accuracy of the new snow product we compared
it with available surface observations of snow cover. The comparison made for 9 consecutive
winter seasons from 2002-2003 to 2010-2011 has shown that the yearly mean agreement of
the blended MODIS and AMSR-E snow product to surface observations was 87%. This is
only about 3% less than the accuracy of NOAA interactive snow maps estimated using the
same method. The accuracy of the EOS blended product dropped to 80-85% in the middle of
the winter season and increased to close to 100% in late spring, summer and early fall.
In order to assess consistency of satellite snow retrievals derived from the different satellites,
we compared estimates of the derived snow covered area (SCA) from daily snow cover
maps of MODIS Terra with ones of MODIS Aqua. The analysis has shown that the SCA
derived from MODIS Terra and from MODIS Aqua changes synchronously.
Additionally, we compared dates of snow melt off as determined from satellite data with
the dates of snow melt off as determined from the available ground-based observations. The
comparison was performed for 48 meteorological stations. The results show that in most
cases these dates differ by 1 to 10 days, however in some cases the difference exceeded two
weeks (results of 2003 are shown in Fig. 5 as an example). Most probably the primary reason
for the difference between satellite and surface estimates of the snow melt-off date consists
in the difficulty of detecting shallow wet snow from satellites. However, the observed dates
of snow cover decease averaged over the whole area are appeared to be very close to
satellite-derived estimates of these dates.
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 151

Fig. 4. Example of a blended global daily snow cover map derived from combined data of
MODIS and AMSR-E instruments onboard Aqua and Terra satellites (upper) and zoomed in
portion of the map covering Eastern Europe. Light blue color and white color represents
snow cover identified correspondingly with microwave and optical satellite data (lower).

30

20

10
Days

0
Vo h
Sa ec

na
Ta sk
vo

vo
Bo n

v
yj od

Li c

Vo lac
O k
Pa ol

icu e l
ho

z
e

bo
ts

ja
sk

r
ri n

ne
l

lo
so

An
El
r

ve

Ka
pe

bo
St lgo

lk

-10
m
O

ro
Be

M
ar

-20

-30

Fig. 5. Difference of snow melt off dates determined from satellite data and from the
available ground-based observations
152 Remote Sensing – Applications

We consider the blended automated snow products as the most advanced ones providing
a reliable and robust characterization of snow cover distribution at the satisfactory
accuracy. Potential of using these products in conjunction with land surface model for
reproducing snow cover distribution over the agricultural area will be demonstrated in
section 5.2.
List of the satellite-derived data, both on land surface characteristics and snow, used in this
study is shown in Table 1.

Latitude-
Land Surface Sensor/ Time
Name of Product Longitude Time Period
Characteristic Satellite Resolution
Resolution
1 Apr to 31 Oct
NASA_MOD11_B1 0.05°x0.05° Twice per day
2003-2010
MODIS/ 1 Mar to 30 May
TERRA&AQUA 2002,
NASA_MOD11_L2 0.01°x0.01° Twice per day
1 Mar to 31 Oct
Land Surface 2003-2010
Temperature
1 Apr to 31 Oct
AVHRR/NOAA 1’x1.5’ Twice per day
1999-2010
Thematic processing
from Every 15min
SEVIRI/ 1 Apr to 31 Oct
0.06°x0.06° (at cloud-free
METEOSAT-9 2009-2011
condition)

Land Surface 16-day 1 Mar to 30 May

NASA_MOD043_C1 MODIS/TERRA 0.05°x0.05°
Albedo product 2003-2010

Snow Water 20 Jan to 30 May

NASA_AE-DySno MODIS/AQUA 0.20°x0.20° Once per day
Equivalent 2003-2010

Snow Covered 20 Jan to 30 May

NASA_MOD10_L2 MODIS/TERRA 0.01°x0.01° Once per day
Area 2002-2010

Vegetation Thematic processing 1 Apr to 31 Oct

AVHRR/NOAA 1’x1.5’ Twice per day
Cover Fraction from 1999-2010

Thematic processing 1 Apr to 31 Oct

AVHRR/NOAA 1’x1.5’ Twice per day
Leaf Area from 1999-2010
Index MODIS/ 1 Apr to 31 Oct
NASA_MOD15_A2 0.01°x0.01° 8-day product
TERRA&AQUA 2003-2010

Land Cover
Land Cover Type
Classification

Tree Cover
Tree Cover Fraction Static data generated from
Fraction AVHRR/NOAA 0.01°x0.01°
AVHRR data
Fraction of
Evergreen Tree
Evergreen Tree
Cover
Cover

Table 1. Satellite-derived products used in this study

Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 153

4. Remote sensing based land surface model: Structure, calibration and

validation by the ground-based data
The developed comprehensive Remote Sensing Based Land Surface Model (RSBLSM)
contains as major components the model of vertical water and heat transfer in the “Soil-
Vegetation-Atmosphere” system (SVAT) for vegetation season as well as the model of
vertical water and heat transfer in the “Soil-SNow-Atmosphere” system (SSNAT) for cold
season. The first versions of these models were developed in the 1990s (Kuchment et al.,
1989; Kuchment & Startseva, 1991; Kuchment & Gelfan, 1996). Later, conceptualization of
the model was improved through accounting for additional processes (Gelfan et al., 2004;
Kuchment et al., 2006), and, importantly, the methods of the model adaption to remote
sensing data on land surface were developed (Muzylev et al., 2002; 2005; 2010; Kuchment et
al., 2010). Below RSBLSM components used in this paper are described in brief.

4.1 SVAT component of the RSBLSM system

The SVAT model is intended for simulating evaporation from bare soil, transpiration by
vegetation, vertical latent and sensible heat fluxes, vertically distributed soil water and heat
content, soil surface and foliage temperatures, land surface radiation temperature as well as
other variables characterized water and heat regimes of soil-vegetation system during a
warm season. The model flowchart is shown in Fig. 6.
Land surface is considered as a two-component soil-vegetation system. Water/heat fluxes
incoming to and outgoing from bare soil and vegetation cover are accounted for separately.
So evapotranspiration ( Ev ) is described as sum of two fluxes: bare soil evaporation ( Eg )
and transpiration by vegetation ( E f ) as

Eg   a  (r  q * (Tg )  q a ) / rag (1a)

E f   a  (q * (T f )  q a )  LAI /(ra  rs ) (1b)

Sensible heat fluxes from surface of bare soil H g and from vegetation cover H f are
calculated as:

H g   a  c p  (Tg  T ) / rag , (2a)

H f   a  c p  (T f  T ) / ra (2b)

where Tg and T f are the soil surface and the foliage temperatures, respectively, T and q a
are the air temperature and specific air humidity at 2 m height, respectively; q * (Tg ) and
q * (T f ) are the specific air saturation humidity at the temperatures Tg and T f ,
correspondingly, r is the relative air humidity near the soil surface, rag and ra are the
aerodynamic resistance between soil surfaces and foliage and between foliage and
atmosphere, respectively,

rag  (C g  U )1 (3)

154 Remote Sensing – Applications

Fig. 6. The SVAT model flowchart

ra  (C e  U )1 , (4)

where C e and C g are the vapour transfer coefficients at the canopy level and at the ground
one, respectively, depending on the surface roughness; U is the wind velocity at 10 m
height, rs is the stomatal resistance calculated by formula

 ( wp )   f
rs  r0 (5)
 ( wp )   f

r 0 is the minimum stomatal resistance,  ( wp ) is the soil matrix potential corresponding to

wilting point  wp ,  is the coefficient, LAI is the leaf area index,  a is the air density, c p is
the specific heat at constant pressure.
The relative air humidity r near the soil surface in (1a) is defined by formula

Mg ( )
r  exp (6)
R(Tg  273)

where  is the volumetric soil moisture content,  ( ) is the soil matrix potential, M is the
molecular mass of water, g is the acceleration of gravity, R is the universal gas constant.
To describe water transfer in the soil, the equation of soil moisture diffusion is applied
taking into consideration water extraction by plant roots
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 155

    
 D    K     R  , z  (7)
t z  z 
where K ( ) is the hydraulic conductivity of soil, D  K   /   is the soil moisture
diffusion coefficient, R( , z) is the intensity of water extraction by plant roots

R( , z )  K ( )  [ r   ( )]  br  r ( z ) (8)

 r and  r ( z) are the rootage water potential and the density, respectively, br is the
coefficient.
The soil matrix potential    and the soil hydraulic conductivity K   can be assigned
using different parametrizations. We compared some of them (Brooks & Corey, 1964; Clapp
& Hornberger, 1978; van Genuhten, 1980) and found that they give similar simulation
results. Hereafter, van Genuhten’s formulas are used:

S 
1/n
1/m
1
     (9)


m 2



K  K 0S 0.5 1  1  S 1/m   (10)

1
where S     r  s  r  is the relative saturation;  s and  r are the residual and the
saturated water contents respectively; K 0 is the saturated hydraulic conductivity;   0 is
the parameter, which is related to the inverse of the air entry pressure; m  1  n 1 , n  1 is
the parameter, which is a measure of the pore-size distribution.
The foliage water potential  f is assumed to be expressed in terms of  r using the
relationship
Zmax
 f   r  rr  w  R  , z  dz (11)
0

where rr is the rootage resistance, Zmax is the maximum length of roots, and  w is the
water density.
Heat transfer within a soil layer is described by

T  T
C eff    ( ( ) ) (12)
t z z
where C eff   and  ( ) are the effective soil heat capacity and the soil heat conductivity.
The soil surface temperature Tg is calculated from Eq. (12). The foliage temperature T f is
obtained from the heat balance equation for the vegetation cover neglecting its heat content.
Both Tg and T f are used for assessing latent and sensible heat fluxes from bare soil and
vegetation surfaces (Eqs. (1a)  (2b)).
The land surface radiation temperature Ts is calculated from the long-wave radiation
balance equation
156 Remote Sensing – Applications

  Ts4  Ra  Rlf  Rlg (13)

where Ra is the atmosphere counterradiation, Rlf and Rlg are long-wave components of
radiation balance for vegetation and bare soil, respectively, calculated as functions of the
measured meteorological variables, σ is the Stefan-Boltzmann constant.
The input variables of the SVAT model are incoming radiation, air temperature, humidity
and pressure, cloudiness, precipitation, wind speed assigned from the standard
meteorological observations. Initial conditions for Eqs. (7) and (12), namely soil moisture
and temperature profiles at the beginning of the vegetation season, are calculated by Eqs.
(25), (26) (see the following section) through spin-up simulations of soil moisture and heat
regimes during a winter previous to vegetation season.
Eqs. (7), (12) are numerically integrated by an implicit, four-point finite difference scheme
with the time and spatial steps of 3 hours and 10 cm, respectively.
Spatial heterogeneity of land surface characteristics is taking into account in the model by
the mosaic approach, i.e. by selection of plots on the studied territory with different soils,
land-use and vegetation types which correspond to specific parameter values. Soils are
characterized by bulk density, maximum hygroscopicity, porosity, field capacity, and
saturated hydraulic conductivity. The heterogeneity of vegetation is represented by the
minimum stomatal resistance, the leaf area index LAI, vegetation cover fraction B, and the
aerodynamic resistance ra .
The principal model parameters have been adjusted by calibration against ground-
measured soil water content W , evapotranspiration Ev , vertical soil moisture profiles,
and soil surface temperature. Also, the values of several parameters have been retrieved
from the specific measurements at agricultural meteorological stations, some of them have
been estimated using satellite data, and the values of certain parameters have been
derived from literature sources. The spatial distributions of the most meteorological
parameter values (being input model variables) have been built using interpolation
procedures. The water and heat balance components for the entire considered territory
have been calculated as weight-averages accounting for size of the area occupied by
specific soil and vegetation.
In developing version of the model designed for utilizing satellite estimates of the land
surface characteristics (built in a quasi-regular grid nodes) the uniform grid with 3x3
AVHRR/NOAA pixel cells (~ 7x5 km2) has been superimposed on the entire investigated
territory divided into plots with different soils and vegetation. Grid size for other sensors
was assigned as close to one of AVHRR/NOAA. (Note, that pixel size for the IR channels of
the AVHRR radiometer is 1' in latitude and 1.5' in longitude, for similar MODIS channels
resolution is equal to 1 and 4.8 km, and for the same channels of SEVIRI it is 0.05º in latitude
and 0.06º in longitude) . For nodes of the grids there have been built AVHRR-derived
estimates of Ta , Tsg , T s.eff , NDVI , E , B , and LAI , MODIS-derived estimates of Tls , E ,
NDVI , B and LAI , and SEVIRI-derived estimates of Tls . Maps of some AVHRR-derived
land surface characteristics for part of the study area are shown in Fig. 7. There have been
also defined model parameters and input variables as well as there have been calculated the
values of Ev , W and other water and heat balance components together with T f , Tg and
Ts .
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 157

Fig. 7. Maps of some AVHRR-derived land surface characteristics

158 Remote Sensing – Applications

4.2 SSNAT component of the RSBLSM system

SSNAT model describes hydrothermal processes in soil and snow cover during a cold
season, including processes of snow accumulation and melt, changes of soil moisture and
temperature during soil freezing and thawing, as well as snowmelt water infiltration.
To simulate temporal changes of the snow depth and density, heat content of snow, water
phase transformation and other processes within a snowpack during snow accumulation
and melt, one-layer snow model proposed in (Kuchment, Gelfan, 1996) has been applied.
The main equations of the model are written as:

dH
  w  Xs 0 1  (S  Es )( i i )1   V (14)
dt

d 
(iH )  w  Xs  S  Es   Si (15)
dt i

d 
( wH )  Xl  S  El  R  i Si (16)
dt w

d
cs Ts H   Qa  Qg   wLS  i LSi (17)
dt
where H is the snow depth; i and w are the vertically averaged volumetric contents of ice
and liquid water, respectively; Ts is the vertically averaged temperature of snowpack; S is
the melt rate; Si is the rate of freezing of liquid water in snow, El is the rate of liquid water
evaporation from snow; Es is the rate of snow sublimation, Qa is the net heat flux at the
snow surface; Qg is the ground heat flux; Xs and X l are the snowfall and rainfall rates,
respectively (partitioning of the total precipitation, X , into solid and liquid phase is a
function of the air temperature); V is the snowpack compression rate; R is the snowmelt
outflow from snowpack; cs is the specific heat capacity of snow;  w , i , and  0 are the
densities of water, ice, and fresh-fallen snow, respectively; L is the latent heat of ice fusion.
The melt rate S is found from the energy balance of the snowpack at zero snow temperature
as:

S
  
 Q  Q   L 1  Q  Q  Q  Q  Q  Q  Q
a g w sw lw ls T E P g   wL 1 , Qa  Qg  0 (18)
0, Qa  Qg  0

where Qsw is the net short wave radiation; Qlw is the downward long wave radiation; Qls
is the upward long wave radiation from snow; QT is the sensible heat exchange; QE is the
latent heat exchange; QP is the heat content of liquid precipitation.
The heat flux components of Qa for an open agricultural site are calculated by the
empirical relationships using the observed meteorological variables (air temperature, air
humidity, wind speed, precipitation, and cloudiness) as the inputs.
The rate Si of freezing of liquid water in snowpack is calculated as:
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 159

 dw Qa  Q g dw
H , Ts  0 C  Qa  Qg  0  H
 dt i L dt

 Qa  Q g Qa  Q g dw
Si   , Ts  0 C  Qa  Qg  0  H (19)
 iL i L dt
 
0, Ts  0 C  Qa  Qg  0
 
Xl , Ts  0 C
The snowpack compression rate V (in cm s-1) is found from:

v1  s H2
V (20)
exp( v2Ts  v3 s ) 2

where  s is the density of snowpack (in g cm-3) equal to  s   i i   w w ; v1 , v2 , and v3 are

the coefficients equal to 2.810-6 cm2 s-1 g-1; -0.08 OC-1; 21 cm3 g-1, respectively.
The outflow of liquid water from snow is calculated as:

 dH
X  S  El  wmax , w  wmax
R l dt (21)
0, w  wmax

where wmax is the holding capacity of snowpack related to its density  s as

s
wmax  0.11  0.11 (22)
w

Numerical integration of the Eqs. (14)  (17) is carried out by an explicit finite-difference
scheme with the time-step of 3 hours.
To calibrate the snow pack model for the study area, the meteorological data from 48
stations for the period from 1 November 2001 to 31 May 2002 were used. Then the model
was validated against snow depth observations during the winter-spring seasons of 2002-
2003 and 2003-2004. The comparison of snow modeling results with the observed snow
depth at the ground-based stations has demonstrated a good correspondence between the
two datasets. The standard error of the simulated snow depth was 5.8 cm during the
calibration stage and 7.7 cm at the validation stage.
Water and heat transfer in a soil during the processes of soil freezing, thawing and
infiltration of water are described by the following equations (Gelfan, 2006):

W   I 
  D  DI K (23)
t z  z z 

T W   T    I  T
cT   wL      wс w  D  DI K (24)
t t z  z   z z  z
160 Remote Sensing – Applications

where W and I are the total water content and ice content of soil, respectively
( W     i I /  w ); cT  c eff   wL   / T  ; ceff is the effective heat capacity of soil equals
c eff   gc g (1  P )   wc w   ici I ;  and c are the soil density and the specific heat capacity,
respectively (indexes w, i and g refer to water, ice and soil matrix, respectively); P is the soil
porosity; D  K   /   I ; DI  K   / I  ;     , I  is the matrix potential of soil.
If soil is frozen ( I ( z , t )  0 ), than hydraulic and thermal characteristics of soil are functions of
ice content, i.e. K  K ( , I ) ,    ( , I ) . One can see that for an unfrozen condition
( I ( z , t ) =0), Eq. (23) reduces to Eq. (7) (neglecting water extraction by plant roots) and Eq.
(24) reduces to Eq. (12).
The matrix potential,     , I  , and the hydraulic conductivity, K=K(,I) are determined
from (Gelfan, 2006):

S 
1/n
1/m
f 1    r     r  
   1  8I 
2
 ( , I )    0  r 1  0 (25)
 
 0  I   r    0  I   r 

m 2

K  , I   K 0S 0.5

 1  1  S 1/m
 f   (26)
f   1  8I  
 
1
where S f    r  0  I   r  is the relative saturation of frozen soil.

Note, that under I ( z , t ) =0 formulas (25), (26) reduce to the van Genuchten’s formulas (9),
(10) for an unfrozen condition.
The values of 0 and r are assumed to be equal to the measured soil porosity and
maximum hydroscopicity, respectively. The values of α, n, cT, and  are calculated from the
measured soil characteristics, such as bulk density, field capacity, and wilting point by the
formulas presented in Gelfan (2006).
Equations (23) and (24) are numerically integrated by an implicit, four-point finite difference
scheme with the time and spatial steps of the finite difference scheme of 1 hour and 10 cm,
respectively.

5. Results
5.1 Satellite based modeling water and heat regimes of the study region during a
vegetation season
The model has been verified by comparing the calculated and measured vertical soil
moisture and temperature profiles, values of land surface temperature, radiation balance,
soil water content of one-meter soil layer, and evapotranspiration. Results of such a collation
for several characteristics above are shown in the Figs. 8-10 by the example of some
meteorological stations located in the study area.
Moreover, there has been executed comparison of modeled values of T f , Tg and Ts with
their satellite-derived analogues Ta , Tsg and Ts.eff for AVHRR and Tls for MODIS and
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 161

1
% of dry wieght 18.04 2 8.05 28.05
10 20 30 40 10 20 30 40
10 20 30 40
10 10
10
20 20
20
30 30 30
40 40
soil depth

40
50 50 50
60 60 60
70 70 70
80 80 80
90 90 90
100 100 100

18.06 10 20 8.07 30 40 28.07

10 20 30 40
10 10 20 30 40
10
20 10
20 20
30
30
40 30
40 40
50 50
50
60 60
60
70 70
70
80 80 80
90 90 90
100 100 100

18.08 8.09
10 20 30 40 10 20 30 40
10 10
20 20
30 30
40 40
50 50
60 60
70 70
80 80
90 90
100 100

Fig. 8. Modeled (1) and measured (2) vertical soil moisture profiles for perennial grasses at
water balance station Nizhnedevitsk for 2003 vegetation season
162 Remote Sensing – Applications

25
Soil temperature, С
0

20
15
10 1
3
5
0
15.04 23.04 2.05 11.05 20.05 28.05 6.06 15.06 24.06 2.07 11.07
Date

20
Soil temperature, С
0

15
2
10 4

5
0
15.04 23.04 2.05 11.05 20.05 28.05 6.06 15.06 24.06 2.07 11.07
Date

Fig. 9. Soil temperature under vegetation cover modeled for the depth 20 (1) and 40 (2) cm
and measured at the same depths for perennial grasses at Nizhnedevitsk water balance
station in 2003

SEVIRI. For most terms of each considered vegetation season the differences Ts.eff  Ts ,
Ta  T f , and Tsg  Tg have not exceed the standard errors of AVHRR-derived estimates of
Ts.eff , Ta and Tsg as one can see from Fig. 11 illustrating this result for the part of the study
area. Separate local spots with a difference of 20°C on these figures correspond to the clouds
above given plots.
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 163

1
Land surface temperature, оС

50 2
40
30
20
10
0
19.04 29.04 9.05 19.05 29.05 8.06 18.06 28.06 8.07 18.07 28.07

3
Foliage temperature, оС

4
40

30

20

10

0
19.04 29.04 9.05 19.05 29.05 8.06 18.06 28.06 8.07 18.07 28.07

Fig. 10. Modeled (1) and measured (2) soil surface temperatures, modeled foliage
temperature (3) and measured air temperature (4) for winter wheat at agricultural
meteorological station Petrinka for vegetation season 1999
52.5
52.5
Ï î í û ðè
Ï î í û ðè

20 20

52
52
Êóðñê 15
15 Êóðñê

Òèì Í èæí åä åâèöê Òèì Í èæí åä åâèöê 10

51.5 10 51.5

5
Î áî ÿí ü Î áî ÿí ü
5
0
51 51

0
-5

50.5 -5 50.5 -10

36 36.5 37 37.5 38 38.5 36 36.5 37 37.5 38 38.5

Fig. 11. LST difference map (T = Ts.eff (AVHRR) – Ts(MODEL)) for part of the study region
of 23,500 km2 with spacing of ~ 5 km, 11.00 GMT 27 May (left) and 12 August (right) 2009

Similar distributions have been also built for the difference of Tls ( MODIS )  Ts ( MODEL ) .
Results of comparing SEVIRI-derived temperature Tls defined by continuous measurements
during 6-day interval of the vegetation season 2009 with three-hour ground-based
observation data on air temperature T at agricultural meteorological station Livny and
with modeled temperature T f calculated using the same ground-based data are shown in
Fig. 12. As seen from this figure, the temporal behavior of these variables is quite close.
164 Remote Sensing – Applications

To discover relevance of the model parameters (stomatal resistance r0 , leaf area index LAI ,
soil and vegetation albedos ( Ag , A f ) and emissivities ( Es , Ep ) for describing the water
and heat exchange processes, sensitivity of the model (in particular, of the quantities Ev
and Ts ) to these parameter variations has been investigated. In (Kuchment & Startseva,
1991) it has been shown that Ev is strongly influenced by r0 . The present study has
confirmed this effect. Particularly, changing the values of r0 for winter wheat and perennial
grasses by 20 and 60 % results in changing the values of Ev by 4-5 and 8-10 mm per decade,
correspondingly. Strong sensitivity of quantities Ev and Ts to variations of LAI was also
found. It was particularly notable in periods of rapid plant growth, their yellowing,
mowing, as well as in the beginning of vegetation season.

40
1
35 2
3
30
Temperature, С
0

25

20

15

10

28.06 29.06 30.06 1.07 2.07

Date

Fig. 12. SEVIRI-derived LST (1), air temperature (2), and air vegetation temperature
modeled using three-hour ground observation data (3) on 28.06.-03.07.2009 at agricultural
meteorological station Livny

Changing the relative values of LAI by 0.1 and 0.2 resulted in changing the values of Ev by
2-2.5 and 3-3.5 mm per decade and Ts by 1.0-1.2 and 1.4-1.6°C, respectively, and changing
the maximum values of LAI from 3 to 5 leads to increase of Ev 7-8 mm per decade and
decrease of Ts ~ 2°C. Numerical experiments with a shift of time mowing led to changing
values of Ev for perennial grasses to 15-17 mm per decade and values of Ts to more than
3°C (fig. 13).
Such sensitivity makes it possible to select appropriately the values of LAI at the specific site,
that, in turn, leads to close to actual assessment of vegetation cover fraction B that is shown
below. The influence of the integral soil and vegetation albedos Ag and A f on the values of
Ev and Ts was different. Evapotranspiration was varied more when changing A f whereas
temperature Ts was changed significantly by variation of Ag . Particularly, change of Ag by
0.2 led to changes of Ev for several crops by 1.5-2.5 mm per decade and corresponding
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 165

60 1 60 5
2 b) 6
50 а) 50

Ev, mm/(10 days)

7
Ev, mm/(10 days)

3
40 8
40 4
9
30 30

20 20

10 10

0 0
1.05 21.05 11.06 1.07 21.07 1.05 21.05 11.06 1.07 21.07 11.08 1.09
Date

Fig. 13. Modeled values of evapotranspiration Ev for 1997 vegetation season at agricultural
meteorological station Petrinka for winter wheat (a); perennial grasses (b) and different
values of LAImax: 2.5 (1); 3.5 (2); 4.5 (3); 3.0 (5); 4.0 (6); 5.0 (7); (8) 4.5 with shift of mowing day
for 3 weeks. (4) and (9) corresponds to measured values of Ev for perennial grasses.

variations of Ts reached 3°С and more. Similar variations of A f led to changes of Ev by 4-5
mm and more per decade and also of Ts by 1.5-2°C. Direct effect of soil and vegetation
emissivities Es and Ep on Ev and Ts was negligible.
The main conclusion from all experiments described above is that the key parameters
affecting evapotranspiration and soil water content under wet soil conditions are r0 and
shading parameters LAI and B . At the same time the influence of the latter two is often
more substantial than of the first one. Under dry soil conditions (usually occurring when
increasing land surface temperature), this effect becomes less noticeable and the value of
evapotranspiration is mainly determined by soil water content of the upper one meter soil
layer.
High sensitivity of Ev and Ts to LAI variations as well as possibility to control current
values of LAI by comparing modeled T f , Tg and Ts with satellite-derived Ta , Tsg and
Ts.eff allowed specifying time behavior of LAI for several crops in the absence of
phytometry data. Satellite-based values of LAI were estimated using empirical
relationships between LAI and NDVI for grasslands (Biospheric Aspects, 1993) (27) and
for agricultural crops (Biftu & Gan, 2001) (28):

LAI  NDVI  1.71  0.48 (27)

LAI  2.5  ln(1.2  2  NDVI ) (28)

Specified in that way time behavior of LAI is presented at Fig. 14. Here sudden changes
correspond to time intervals of plant mowing.
In initial versions of the model the fraction B was calculated as follow

B  1  exp(  p  LAI ) (29)

Here p is empirical coefficient that was adjusted by comparing modeled magnitudes of T f ,

Tg , and Ts with satellite-derived values of Ta , Tsg , and Ts.eff using LAI estimates
determined from both satellite and ground data. Numerical experiments were carried out
under three scenarios of estimating LAI and B : 1) values of B were calculated by (29)
using LAI determined by ground observation data on phenological stage changes and plant
heights for different land-use; 2) values of B were also calculated by (29) using satellite-
166 Remote Sensing – Applications

2.0 1
2

2
1.5

LAI, cm /cm
3

2
1.0

0.5

0.0
1.04 18.04 8.05 29.05 18.06 8.07 29.07 18.08 8.09 28.09 18.10

Date

Fig. 14. Temporal behavior of LAI for vegetation season 2004 built by data of AVHRR/3 (1),
MODIS (3) and by ground observation data at water balance station Nizhnedevitsk for
perennial grasses (1, 2) and without allocation of cultures (3)

derived LAI ; 3) estimates of B and LAI were generated from satellite data. Values of Ta
and T f under these scenarios are close to each other: for the most of observation times their
differences do not exceed 2.5 and 3.5°C respectively (i.e. they are close to RMSE for satellite-
derived Ta ). These results are confirmed by rather high (0.65-0.75 at different seasons) T f
and Ta correlation coefficient. Similar results were also obtained when comparing Ts with
Ts.eff and Tsg with Tg under all the scenarios. Examples of comparing AVHRR-derived and
ground-based estimates of B for perennial grasses are presented in Fig. 15.

0.7
1
0.6 2
Vegetation cover fraction

0.5

0.4

0.3

0.2

0.1

0.0
1.04 18.04 8.05 28.05 18.06 7.07 29.07 18.08 8.09 30.09
Date

Fig. 15. Comparing AVHRR-derived (1) and ground-based estimates (2) of vegetation cover
fraction B for perennial grasses at the station Kursk for vegetation season 1999

Described results gave impetus to research possibility of direct use of satellite-based

estimates of vegetation cover and LSTs in the model. High desirability of such use while
model simulating water and heat balance components for vast territories is due to the
necessity of distributed estimates of the land surface characteristics, especially under the
lack of ground-based observation data. To assimilate satellite-based estimates of vegetation
and meteorological characteristics the updating of the SVAT model has been performed
including:
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 167

1. Replacing the ground point-wise estimates of the model parameters LAI and B by
their AVHRR- or MODIS-based analogues. The efficiency of such approach has been
proved through comparisons: between satellite-derived and ground-based data on LAI
and B behavior during vegetation season; between satellite-derived, modeled, and in-
situ measured temperatures; between modeled and actual values of evapotranspiration
Ev (Fig. 16) and of soil water content W for one-meter soil layer (Fig. 17). The

30
1
Evapotranspiration, mm / 5 days

2
25 3

20

15

10

0
5.04 6.05 5.06 5.07 5.08 5.09 5.10
Date

Fig. 16. Evapotranspiration Ev for vegetation season 2008 measured on grassland at water
balance station Nizhnedevitsk (1), modeled using AVHRR-derived LAI for perennial
grasses (3) and MODIS-derived LAI (2)
Kursk Ponyri

1
Soil water content W, mm

500 500
2
400 400
3
300 300
1
200 2 200

100 3 100

0 0
8.04 8.05 8.06 7.07 7.08 8.09 8.10 8.04 8.05 8.06 7.07 7.08 8.09 8.10

Nizhnedevitsk Tim

1 500
500
2
400 3 400

300 300
1
200 200
2
100 100 3
0 0
8.04 8.05 8.06 7.07 7.08 8.09 8.10 8.04 8.05 8.06 7.07 7.08 8.09 8.10

Fig. 17. Soil water content: ground measured for corn at four agricultural meteorological
stations (1), modeled using AVHRR-derived LAI and B for corn (2) and MODIS-derived LAI
(3) for vegetation season 2008.
168 Remote Sensing – Applications

discrepancies between Ta , Ts , Ts.eff and T f , Tg , Ts as well as between Tls and Ts and

ground-measured ones do not exceed standard errors of satellite-derived estimates Ta ,
Tsg and Ts.eff respectively while the modeled and measured values of Ev and W are
found close to each other within a standard error of their estimation.
2. Entering the AVHRR- or MODIS-based LST estimates as the input SVAT model
variables instead of their standard ground-based estimates if the time-matching of
satellite and ground-based observations takes place. The SEVIRI-derived Tls estimates
can also play the same role. Permissibility of such replacement has been verified while
comparing remote sensed, modeled and ground-based temperatures as well as
calculated and measured values of W and Ev (Fig. 18) The SEVIRI-based Tls
estimates are found to be very informative and useful due to their high (up to 15 min)
temporal resolution.
3. Inputting AVHRR- and MODIS-derived LAI and B , AVHRR-, MODIS-, and SEVIRI-
based LSTs in each grid cell of the model in order to account for the space variability
of vegetation cover parameters and meteorological characteristics. Ground-based
data on precipitation, air temperature and humidity prepared by Inverse Distance
Squared (IDS) interpolation method are also inserted into the model in each grid
node. The calculations of vertical water and heat fluxes, soil water and heat contents
and other water and heat balance components for above region of interest have been
carried out using the described updated SVAT model and the fields of AVHRR/3- or
MODIS-derived LAI and B estimates together with AVHRR/3- or MODIS- or
SEVIRI-derived Tls retrievals for 1999-2010 vegetation seasons. Fig.19 shows the
results of calculation of latent heat flux and soil water content at the same date of
2010 as shown in Fig. 7. The acceptable accuracy levels for above values assessment
have been achieved under all scenarios of parameter and input model variable
specification using satellite- and ground-based data.

0.80
0.70 1
Evapotranspiration, mm/3 h

2
0.60
0.50
0.40
0.30
0.20
0.10
0.00
28.06 28.06 29.06 30.06 1.07 2.07
Date

Fig. 18. Three-hour values of evapotranspiration modeled using SEVIRI-derived data (1)
and three-hour observation data at the agricultural meteorological station Livny (2) on
28.06.-03.07.2009
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 169

Fig. 19. Calculated latent heat flux and soil water content distributed over the study region

As follows from the above, the presented approach can be used for vast territories under the
lack or absence of ground observations. The most promising in this case is the utilization of
SEVIRI data due to their frequent occurrence.

5.2 Satellite based modeling of snow cover

The calibrated snow model (14)  (22) was applied to calculate snow cover characteristics in
each 0.01° pixel of the spatial grid within the study areas. The meteorological data
interpolated from the available meteorological stations to each pixel by the IDS method
were used as the input to the snow cover models. In pixels where MODIS data were
available the simulated land surface temperature and albedo were replaced by
corresponding satellite-derived variables. Fields of snow cover characteristics based on
satellite observations and the snow pack model were generated for the study area for the
time period from January 1st to May 31th of the years 2002  2005. The initial snow cover
distribution in the model was assigned following AMSR-derived SWE data on 1 January for
all pixels. For the pixels where the initial SWE values were unavailable because of the lack of
170 Remote Sensing – Applications

coverage or for some other reasons, these values were interpolated from neighboring pixels
by the IDS interpolation method.
Maps of the simulated distributions of these snow characteristics were compared with the
corresponding satellite maps. The simulated changes of SCA were appeared to be in
satisfactory correspondence with the satellite-derived SCA. As an example of this
correspondence, simulated and satellite-derived dynamics SCA are shown in Fig. 20 for
several sub-regions adjoining the meteorological stations of the study area.

Fig. 20. Simulated (solid dots) and MODIS-derived (circles) dynamics of SCA for some
polygons within the study area (winter-spring season of 2004)

However, the simulated maps of SWE substantially differ from the corresponding AMSR-
derived SWE maps as it is illustrated by Fig. 21. We assume that this difference is resulted
from by the fact that the accuracy of the SWE estimated from the radiometric satellite
measurements noticeably decreases during melt period when snowpack is saturated by
melted water so the AMSR-derived SWE maps may not adequately represent the real SWE
distribution for this period of the year.
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 171

Fig. 21. Simulated (upper raw) and AMSR-derived (lower raw) distribution of SWE (mm) on
25-26 March, 2003.

6. Conclusion
The paper highlights the importance of satellite remote sensing data used in conjunction
with a land surface model in describing water and heat regimes of vast agricultural regions.
Accuracy and robustness of the corresponding quantified assessments by LSM is strongly
restricted by insufficiency, both in space and time, of ground observations of highly
heterogenic soil, vegetation and snow characteristics, soil temperature, etc., which can be
used as the model parameters or input variables, as well as for calibration and validation of
the model. Satellite remote sensing is not only an additional source of land surface data
allowing substitution of the missed ground observations into the model, but, for majority of
regions, it is a unique data source. From the other side, in spite of the diversity of satellite
data on land surface characteristics and technologies of the data thematic processing,
opportunity of using these data in LSMs is often problematic because of many reasons, e.g.
irregularity of data, uncertain data accuracy, etc.
Moreover, there is no general approach allowing reliable choice of appropriate satellite data
and/or processing technique; this choice depends on both specific features of the region of
interest and the used model. In this study, we tried, firstly, to demonstrate opportunities of
utilizing land-surface/snow remote sensing products obtained from the different satellites
(NOAA, EOS Terra and Aqua, METEOSAT-9) and sensors (SEVIRI, AMSR-E, AVHRR,
MODIS) in the developed LSM. Secondly, we analyzed sensitivity of the simulation results
to different satellite remote sensing data. It is important that most of the used products were
172 Remote Sensing – Applications

derived by the originally developed processing techniques, especially new technique for
processing SEVIRI measurement and the technology based on synergy of optical and
microwave snow products. Other main emphasis of the study is to develop a new physically
based distributed Remote Sensing Based Land Surface Model (RSBLSM). The model is
aimed to simulation of vertical water-and-heat transfer and, importantly, it takes into
account hydrothermal processes in the “frozen soil-snow-atmosphere” system. These
processes are critical for cold region agriculture, as they define crop development in early
spring before the vegetation season beginning.
The obtained results allow us to conclude that utilization of the differently derived satellite-
based estimates in the developed physically based model, intensively calibrated and
validated against the available ground observations, provides an opportunity for
reproducing spatial fields of evapotranspiration, soil moisture and temperature at different
soil depths, temperature of soil/vegetation surfaces, snow and other water and heat
characteristics for the vast agricultural region.

7. Acknowledgment
We thank E.Volkova, A.Kukharsky, S.Uspensky, V.Solovjev from SRC “Planeta” for their
participation in the development of satellite data processing methods, as well as
M.Alexadrovich from Water Problem Institute of RAS for technical assistance in modeling
and data processing. The presented study was carried out under support of the Russian
Foundation of Basic Researches (grant № 10-05-00807).

8. References
Armstrong, R. L., Brodzik, M. J. (2005) Northern Hemisphere EASE-Grid Weekly Snow
Cover and Sea Ice Extent Version 3. Digital media, http://nsidc.org/data/nsidc-
0046.html, National Snow and Ice Data Center, Boulder, Colorado, USA.
Becker F, Li, Z.-L. (1995) Surface temperature and emissivity at various scales: definition,
measurement and related problems. Rem. Sens. Rev.,Vol. 12, pр. 225-253.
Biospheric Aspects of the Hydrological Cycle (BAHC). (1993) Report № 27, 103 p. Ed. by
BAHC Core Project Office, Institut fűr Meteorologie, Freie Universitat Berlin,
Germany.
Biftu, G.F., Gan, T.Y. (2001) Semi-distributed, physically based, hydrologic modeling of the
Paddle River basin, Alberta, using remotely sensed data. J. Hydrol., Vol. 244, pp.
137-156.
Brodzik, M. J., Armstrong, R., Savoie, M. (2007) Global EASE-Grid 8-day Blended SSM/I
and MODIS Snow Cover. Digital media.
http://nsidc.org/data/docs/daac/nsidc0321_8day_ssmi_modis_blend/index.html, National
Snow and Ice Data Center, Boulder, Colorado, USA.
Brooks, R.H., Corey, A.T. (1964) Hydraulic properties of porous media. Hydrol. Pap. Colorado
State Univ. № 3. 27 p.
Brubaker, K. L., Pinker, R. T., Deviatova, E. (2005) Evaluation and Comparison of MODIS
and IMS Snow-Cover Estimates for the Continental United States Using Station
Data. J. Hydrometeor., Vol. 6, pp. 1002–1017. doi: 10.1175/JHM447.1
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 173

Clupp, R.B., Hornberger, G.M. (1978) Empirical equations for some hydraulic properties.
Water Resources Research, Vol. 14, № 4, pp. 601-604.
Collatz, G.J., Ball. J.T., Grivet, C., Berry, J.A. (1991) Physiological and environmental
regulation of stomatal conductance, photosynthesis and transpiration: a model that
includes a laminar boundary layer. Agricultural and Forest Meteorology, Vol. 54, pp.
107–136.
Dai, Y., et al. (2003) The Common Land Model (CLM) version 1.0. Bull. Am. Meteorol. Soc.,
Vol. 84, pp. 1013– 1023.
de Wildt, M. D., Gabriela, S., Gruen, A. (2007) Operational snow mapping using
multitemporal Meteosat SEVIRI imagery. Remote Sensing of Environment, Vol. 109,
pp. 29-41.
Deardorff ,J.W. (1978) Efficient prediction of ground surface temperature and moisture with
inclusion of a layer of vegetation. Journal of Geophysical Research, Vol. 83, pp. 1889–
1903.
Desborough, C.E., Pitman, A.J. (1998) The BASE land surface model. Global and Planetary
Change, Vol. 19, pp. 3-18.
Dickinson, R.E., Henderson-Sellers, A., Kennedy, P.J., Wilson, M.F. (1986) Biosphere
Atmosphere Transfer Scheme (BATS) for the NCAR Community Climate Model, NCAR
Tech. Note, TN-275 + STR, 69 pp.
Faysash, A., Smith, E.A. (2000) Simultaneous Retrieval of Diurnal to Seasonal Surface
Temperatures and Emissivities over SGP ARM-CART Site Using GOES Split
Window. J. Appl. Meteor., Vol. 39, pp. 971-982.
Foster, J.L., Hall, D. K., Eylander, J.B., Riggs, G.A., Nghiem, S.V., Tedesco, M., Kim, E.,
Montesano, P.M., Kelly, R. E.J., Casey, K.A., Choudhury, B. (2011) A blended global
snow product using visible, passive microwave and scatterometer satellite data.
International Journal of Remote Sensing, Vol. 32, pp. 1371-1395.
Gelfan, A. N., Pomeroy, J. W., Kuchment, L.S. (2004) Modelling forest cover influences on
snow accumulation, sublimation, and melt. J. Hydrometeorology, Vol. 5, No. 5, pp.
785–803.
Gelfan, A. N. (2006) Physically based model of heat and water transfer in frozen soil and its
parametrization by basic soil data. In: Predictions in Ungauged Basins: Promises and
Progress. Proceedings of symposium S7 held during the Seventh IAHS Scientific Assembly
at Foz do Iguazu, Brazil, April 2005, Eds: M. Sivapalan,. IAHS Publ., Vol. 303, pp. 293-
304.
Global Water Security – an engineering perspective. Report of the The Royal Academy of
Engineering, Publ. by The Royal Academy of Engineering, London, UK, April 2010.
Good, E. (2009) Blending in situ and satellite data for monitoring land air temperatures.
Proc. of the 2009 Eumetsat Meteorol. Sat. Conf, Bath, UK, 21 - 25 September 2009, 5 p.
Grody, N.C., Basist, A.N. (1996) Global identification of snow cover using SSM/I
Measurements. IEEE Trans. Geosci. Remote Sens., Vol. 34, N 1, pp. 237-249.
Gusev, E. M., Nasonova, O.N. (2002) The simulation of heat and water exchange at the land–
atmosphere interface for the boreal grassland by the land-surface model SWAP.
Hydrol. Processes, Vol. 16, pp. 1893-1919.
Hall, D.K., Riggs, G., Salomonson, V., DiGirolamo, N.E., Bayr, K.J. (2002) MODIS snow
cover products. Remote Sensing of Environment, V. 83, pp. 181-194.
174 Remote Sensing – Applications

Hall, D.K, Riggs, G.A. (2007) Accuracy assessment of the MODIS snow-cover products.
Hydrological Processes, Vol. 21, pp. 1534–1547.
Helfrich, S.R., McNamara, D., Ramsay, B.H., Baldwin, T., Kasheta, T. (2007) Enhancements
to, and forthcoming developments in the Interactive Multisensor Snow and Ice
Mapping System (IMS). Hydrological Processes, Vol. 21, pp. 1576–1586.
Henderson-Sellers, A, Pitman, A.J., Love, P.K., Irannejad, P., Chen, T. (1995) The project for
intercomparison of land surface parameterization schemes (PILPS) phases 2 and 3.
Bulletin of the American Meteorological Society, Vol. 76, pp. 489–503.
Justice, C. O., Vermote, E., Townshend, J. R. G., et al. (1998) The Moderate Resolution
Imaging Spectroradiometer (MODIS): land remote sensing for global change
research. IEEE Trans. Geosci. Remote Sens., Vol. 36, pp. 228-1249.
Kelly, R. E., Chang, A. T., Tsang, L., Foster, J.L. (2003) A prototype AMSR-E global snow
area and snow depth algorithm. IEEE Transactions on Geoscience and Remote Sensing,
Vol. 1, pp. 230-242.
Kuchment, L.S., Motovilov, Yu., G., Startseva, Z.P. (1990) Modeling of moisture transport in
the soil-vegetation-surface boundary layer system for hydrological problems. Water
Resources, Vol. 16, N 2, pp. 121-128. (Translated from Russian, Vodnye Resursy
(1989), N 2, pp.32-39, Plenum Publishing Corporation, Consultants Bureau, New
York, USA).
Kuchment, L.S., Startseva, Z.P. (1991) Sensitivity of evapotranspiration and soil moisture in
wheat fields to changes in climate and direct effects of carbon dioxide. Hydrol. Sci.
J., Vol. 36, N 6, pp. 631-643.
Kuchment, L. S., Gelfan, A.N. (1996) The determination of the snowmelt rate and the
meltwater outflow from a snowpack for modeling river runoff generation. J.
Hydrology, Vol.179, pp. 23-36.
Kuchment, L.S., Demidov, V.N., Startseva, Z.P. (2006) Coupled modeling of the hydrological
and carbon cycles in the soil–vegetation–atmosphere system. J. Hydrology, Vol.323,
pp.4-21
Kuchment, L.S., Romanov, P., Gelfan, A.N., Demidov, V.N. (2010) Use of satellite-derived
data for characterization of snow cover and simulation of snowmelt runoff through
a distributed physicall basedmodel of runoff generation Hydrol. Earth Syst. Sci., Vol.
14, pp. 339–350.
Lokupitiya, E., et al. (2009) Incorporation of crop phenology in Simple Biosphere Model
(SiBcrop) to improve land-atmosphere carbon exchanges from croplands.
Biogeosciences, Vol.6, pp.969–986.
Manabe, S. (1969) Climate and the ocean circulation. The atmospheric circulation and
hydrology of the Earths surface. Mon.Wea. Rev., Vol. 97, pp. 739–774.
Mauser, W., Schädlich, S. (1998) Modelling the spatial distribution of evapotranspiration on
different scales using remote sensing. J. Hydrol., Vol. 212–213, pp. 250–267.
Milly, P.C.D., Shmakin, A.B. (2002) Global modeling of land water and energy balances. Part
I: the Land Dynamics (LaD) model. Journal of Hydrometeorology, Vol. 3, pp. 283–299.
Muzylev, E.L., Uspensky, A.B., Volkova, E.V., Startseva, Z.P. (2002) Simulation of
hydrological cycle of river basins using synchronous high resolution satellite data.
Russian Meteorology and Hydrology, N 5, pp. 52-63.
Remote Sensing Based Modeling of Water and Heat Regimes in a Vast Agricultural Region 175

Muzylev, E.L., Uspensky, A.B., Volkova, E.V., Startseva, Z.P. (2005) Using satellite
information for modeling heat and moisture transfer in river watersheds. Earth
Research from Space, N 4, pp. 35-44. (In Russ.)
Muzylev, E., Uspensky, A., Startseva, Z., Volkova, E. (2006) Modeling vertical heat and
water fluxes from river basin area with AVHRR/NOAA-based information on land
surface characteristics. Proc. of the 7th International Conference on Hydroinformatics,
Nice, France, 4-8 September 2006. Ed. P.Gourbesville, J.Cunge, V.Guinot, S.-Y.Liong,
Vol. 2, pp. 1163-1170, Research Publishing, Chennai, India.
Muzylev, E.L., Uspenskii, A.B., Startseva, Z.P., Volkova, E.V., Kukharskii, A.V. (2010)
Modeling water and heat balance components for the river basin using remote
sensing data on underlying surface characteristics. Russian Meteorology and
Hydrology, Vol. 35, N 3, pp. 225-235, DOI: 10.3103/S1068373910030106
NOAA KLM User's Guide. (2005) http://www.ncdc.noaa.gov/oa/pod-guide/ncdc/docs/klm/html/
c1/sec1-2.htm.
Oleson, K.W., Niu, G.-Y., Yang, Z.-L., Lawrence, D. M., Thornton, P. E., Lawrence, P. J.,
Stöckli, R., Dickinson, R. E., Bonan, G. B., Levis, S., Dai, A., Qian, T. (2008)
Improvements to the Community Land Model and their impact on the hydrological
cycle. J. Geophys. Res., Vol. 113, G 01021, doi:10.1029/2007JG000563.
Overgaard, J., Rosbjerg, D., Butts, M. B. (2006) Land-surface modelling in hydrological
perspective – a review. Biogeosciences, Vol. 3, pp. 229–241.
Pitman, A.J. (2003) The evolution of, and revolution in, land surface schemes designed for
climate models. Int. J. Climatol., Vol. 3, pp. 479–510.
PUM LST (Product User Manual. Land Surface Temperature) (2008) SAF/LAND/IM/
PUM_LST/ 2.1, 49p.
Romanov, P., Gutman. G., Csiszar, I. (2000) Automated monitoring of snow over North
America with multispectral satellite data. Journal of Applied Meteorology, Vol. 39, pp.
1866–1880.
Romanov, P. (2011) Satellite-Derived Information on Snow Cover for Agriculture
Applications in Ukraine, in Use of Satellite and In-Situ Data to Improve
Sustainability. NATO Science for Peace and Security Series C: Environmental Security.
Part 2, pp.81-91, DOI: 10.1007/978-90-481-9618-0_9
Rutter, N. et al. (2009), Evaluation of forest snow processes models (SnowMIP2), J. Geophys.
Res., 114, D06111, doi:10.1029/2008JD011063
Schmetz, J., Pili, P., Tjemkes, S., Just, D. et al., (2002) An introduction to Meteosat Second
Generation (MSG). Bull. Amer. Meteor. Soc., Vol. 83, pp. 977-992.
Schultz, K., Franks, S., Beven, K. (1998) TOPUP – A TOPMODEL based SVAT model to
calculate evaporative fluxes between the land surface and the atmosphere, Version 1.1,
Program documentation, Department of Environmental Science, Lancaster
University, UK.
Sellers, P.J, Mintz, Y., Sud, Y.C., Dalcher, A. (1986) A Simple Biosphere model (SiB) for use
within general circulation models. Journal of the Atmospheric Sciences, Vol. 43, pp.
505–531.
Sellers, P.J., Berry, J.A., Collatz, G.J., Field, C.B., Hall, F.G. (1992) Canopy reflectance,
photosynthesis and transpiration. III. A reanalysis using improved leaf models and
a new canopy integration scheme. Remote Sensing of the Environment, Vol. 42, pp.
187–216.
176 Remote Sensing – Applications

Sellers, P. J., Dickinson, R. E., Randall, D. A., Betts, A. K., Hall, F. G., Berry, J. A., Collatz, G.
J., Denning, A. S., Mooney, H. A., Nobre, C. A., Sato, N., Field, C. B., Henderson-
Sellers, A. (1997) Modeling the exchanges of energy, water, and carbon between
continents and the atmosphere. Science, Vol. 275, pp. 502-509.
Simic, A., Fernandes, R., Brown, R., Romanov, P., Park, W. (2004). Validation of
VEGETATION, MODIS, and GOES C SSM/I snow-cover products over Canada
based on surface snow depth observations. Hydrological Processes, Vol. 18, No. 6, pp.
1089–1104.
Snyder, W.C., Wan, Z., Zhang, Y., Feng, Y.-Z. (1998) Classification-based emissivity for land
surface temperature measurement from space . Int. J. Rem. Sens., Vol. 19, No. 14, pp.
2753-2774.
Solovjev, V.I., Uspensky, S.A. (2009) Monitoring of land surface temperatures based on
second generation geostationary meteorological satellites, Earth Studies from Space,
No. 3, pp.17-25, (in Russian).
Solovjev, V.I., Uspensky, A.B., Uspensky, S.A. (2010) Derivation of land surface temperature
using measurements of IR radiances from geostationary meteorological satellites,
Russian Meteorology and Hydrology, Vol. 35, No. 3, pp. 159-167, DOI 10.3103/
S1068373910030015.
Uspensky, S., Solovjiev, V., Uspensky, A.,(2009) Monitoring of land surface temperatures
based on SEVIRI/METEOSAT-9 measurements. Proceedings of 2009 EUMETSAT
Meteorological Satellite Conference, 2009, EUMETSAT, pp 55-60.
Uspensky, S., Uspensky, A., Rublev, A. (2011) Land surface air temperature retrieval
capabilities with geostationary meteorological satellite data. Proc. “Intern. Symp. on
Atmospheric Radiation and Dynamics”, St.Petersburg, June 21-24, 2011,pp. 37-38.
van Genuchten, M.Th. (1980) A closed form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Sci. Soc. Am. J., No. 44, pp.892-898.
Volkova, E.V., Uspenskii, A.B. (2007). Detection of clouds and identification of their
parameters from the satellite data in the daytime. Russian Meteorology and
Hydrology. Vol. 32, No. 12, pp. 723-732, DOI: 10.3103/S1068373907120011.
Walker, A.E., Goodison, B.E. (1993) Discrimination of a wet snow cover using passive
microwave satellite data. Annals of Glaciology, Vol. 17, pp. 301–311
Wan, Z., Dozier, J. (1996) A generalised split-window algorithm for retrieving land surface
temperature from space. IEEE Trans. Geosci. Rem. Sens., Vol. 34, No. 4, pp. 892-905.
Wood, E.F, Lettenmaier, D.P, Zartarian, V.G. (1992) A land-surface hydrology
parameterization with subgrid variability for general circulation models. Journal of
Geophysical Research Vol. 97, pp. 2717–2728.
Yang, Z.-L. (2004) Modeling land surface processes in short term weather and climate
studies, In: Observation, Theory and Modeling of Atmospheric Variability, Ed. X. Zhu, X.
Li, M. Cai, S. Zhou, Y. Zhu, F.-F. Jin, X. Zou, and M. Zhang, pp. 288-313, World
Scientific Series on Meteorology of East Asia, World Scientific, New Jersey.
 7

High Resolution Remote Sensing Images

Based Catastrophe Assessment Method
Qi Wen et al.*
National Disaster Reduction Center of China
China

1. Introduction
In recent years, there seems to be more and more occurrences of natural disasters happening
around the world due to abnormal climate change. To deal with natural disasters, disaster
assessment technology will provide technical support and help facilitate decision making for
disaster relief, disaster prevention and reduction, post-disaster recovery and reconstruction
(The six editing room of Press of China Standards, 2010). Airborne remote sensing and
satellite remote sensing, which feature no time limitation, no geographical restriction, wide
coverage and high accuracy, are widely used in disaster assessment scenario, because they
can provide prompt and accurate information (Xie & Zhang, 2000). After several earthquakes
happened around China (Xingtai, Haicheng, Tangshan, Longling, Datong) during 1970s and
1980s, China had widely implemented airborne remote sensing photography and seismic
damage interpretation (Zhang, 1993). Remote sensing had also played an important role in
disaster assessment of recent occurring disasters, including Wenchuan Earthquake, Yushu
Earthquake, Zhouqu Debris flow and Yingjiang Earthquake (Chen et al., 2008; Shi et al.,
2010). This chapter will mainly discuss the catastrophe assessment method and technical flow
used by National Disaster Reduction Center of China (NDRCC) during Wenchuan
Earthquake(2008), Yushu Earthquake(2010), Zhouqu Debris Flow(2010). Further discussion
and advises are also given.

2. The flow of catastrophe assessment method

Traditionally, catastrophe assessment process can be divided into three major steps: disaster
scope assessment, physical quantity assessment and direct economic losses assessment. In
addition, two assessment processes: rapid disaster assessment and ground investigation
process are needed to supplement the major steps. Rapid disaster assessment is usually
carried out to preliminarily judge the disaster condition after the disaster occurred. The
reported data from Rapid Disaster Assessment, combined with that from following Disaster
Scope Assessment, are integrated to have an overview of disaster scope and extent. Another
supplementary process, Ground Investigation Process, is usually carried out to cross-

*Yida Fan, Siquan Yang, Shirong Chen, Haixia He, Sanchao Liu, Wei Wu, Lei Wang, Juan Nie, Wei
Wang, Baojun Zhang, Feng Xu, Tong Tang, Zhiqiang Lin, Ping Wang and Wei Zhang
National Disaster Reduction Center of China, China
178 Remote Sensing – Applications

validate disaster condition after Rapid Disaster Assessment process. The rapid disaster
assessment result, ground investigation data, combined with high resolution remote sensing
data will provide a detailed comprehensive assessment report of disaster condition. Figure 1
illustrates the flow of catastrophe assessment method.

Disaster Scope Assessment Physical Quantity Direct Economic Loss

Assessment Assessment

Reported Disaster High Resolution Remote Social Economic Data

Information Sensing Data Related Standards
Remote Sensing Data Reported information

Administrative district as a Entity as a unit, statistic Damage area, length,

unit, judge disaster scope and the damage area, length, number multiply damage
extent number and so on. extent multiply price

Fig. 1. The flow of catastrophe assessment method

2.1 Preliminary judgment of disaster condition

When the disaster happens, the remote sensing system will response immediately. By
comparing pre-disaster and post-disaster high resolution remote sensing images, data on
damage of transportation lines, houses, variation of terrain will be generated, and
preliminarily overview of damage and disaster trend will be concluded based on these data.

2.2 Ground investigation

When the disaster situation is stable, according to preliminary judgment result, an expert
investigation group will be sent to investigate in-field the disaster situation of sample
region. Prior to ground investigation, the disaster region should be partitioned as multi-
level grids in high-resolution remote sensing images. The first level partitioning is to
partition the remote sensing images to functional zone according to administrative district,
based on which the second level partitioning is to partition each functional zone to smaller
unit, named “Entity”, according to street in order to facilitate the ground investigation. The
third level partitioning is to partition the “Entity” to even smaller grid unit according to
house and parcel of rice field. All the grids are indexed, so that the investigation team
members can fill out the corresponding forms. Investigation team uses Ground Disaster
Information Collection System (Figure 2) to retrieve disaster information of 3rd level grid
unit, including number of houses and their floors, estimated construction cost, and general
information of 1st level unit, including affected population, damage extent and field photos.
This information is collected and sent to the assessment team in NDRCC.
High Resolution Remote Sensing Images Based Catastrophe Assessment Method 179

Fig. 2. Ground Disaster Information Collection System

Ground Disaster Information Collection System is a GIS software with the client-server
architecture. The client side implemented in mobile phone is used to collect text and picture
disaster information, label the position and send information to the sever side. The server
side receives information from client side, interprets it and stores it to database.

2.3 Disaster scope assessment

When the disaster situation is stable, data reported by local government, assessment results
generated by investigation expert group, interpreted data from remote sensing images are
integrated and analyzed together to implement disaster scope assessment for sample
regions. Regression relations between the sample regions and the entire disaster area are
then established. Also, damage assessment data in adjacent regions are interpolated so as to
generalize assessment of the whole area based on that of sample region.
In the case of an earthquake, disaster loss assessment is graded using Comprehensive
Disaster Index. Disaster scope and extent of loss are assessed as a unit of administrative
district(such as “county” in China). The Comprehensive Disaster Index is usually
calculated as a weighted sum of the average seismic intensity, the toll of death and
missing, the rate of number of death and missing per ten thousands persons, the number
of collapsed houses, the rate of number of collapsed houses per ten thousands people, the
geological disaster risk, and the rate of number of relocated person per ten thousand
people.

DI   ( f k  DI k ) (1)

DI k is the kth normalized single indicator:

DI k  [DI k  min(DI k )] /[max(DI k )  min(DI k )] (2)

f k is the kth weight, and should be adjusted according to the specific condition of the
disaster.
180 Remote Sensing – Applications

Average seismic intensity can be calculated by weighted sum of rates of different seismic
intensity level for different area:

I   ( I i  Si / S ) (3)

I i is the intensity level. Si / S is the rate of different seismic intensity level for different area;
the toll of death and missing is the number of dead and missing persons in each county;
number of collapsed houses is counted separately for different counties; geological disaster
risk is calculated as a weighted sum of number of endangered residential areas, broken
roads, blocked rivers, collapsed bridges and reservoirs. The rate of number of relocated per
ten thousands of persons is calculated based on the base household population of each
county.

2.4 Physical quantity assessment

The physical quantity assessment is carried out based on the previous data gathered in the
first step of Disaster Scope Assessment. Based on extensive research and practical work
experiences, NDRCC has established NNDDCA (the National Natural Disaster Damage
Comprehensive Assessment Framework). Disaster loss is classified as five dimensions, and
further divided to two hundred and twenty nine indicators. The five dimensions include:
population loss, economic losses, residents’ housing and property damage, national
economy and industry loss, government offices and social undertakings loss. Several
indicators, which are suitable to be assessed in remote sensing images, are first selected from
the two hundred and twenty nine indicators and the damage of which is then assessed
using artificial visual interpretation method and automatically machine recognition
technology. The remote sensing index framework is then established.
One thing to notice is that, the affected population is not included in the remote sensing
assessment index framework, and the economic losses can be statistically calculated from
residents’ housing and property damage, national economy and industry loss, government
offices and social undertakings loss. The losses of residents’ property are also not included
in the remote sensing assessment index framework.
According to the characteristics of remote sensing technology, housing of three categories of
statistical objects is classified as housing type; select specific object from agriculture,
industry and services of state-owned enterprises; select water conservancy, environment,
public facilities and other departments from government offices and social undertakings.
Specific objects in the three categories can be divided into three new categories of remote
sensing index of physical quantity damage assessment: housing and buildings, important
infrastructure and natural resources. The first category, housing and buildings, includes
building, bungalow, chimney and water tower. In the second category, important
infrastructures include transportation, municipal, power, water, telecommunication and
radio communication, pipeline facilities, etc., natural resources include land resources and
water resources. Table 1 below shows a selection of assessment indicators.
In our case, data extracted from high-resolution remote sensing images in the step of
Physical Quantity Assessment is usually brought to the decision-making chain as major
information resource. By using pre-disaster and post-disaster UAV and airborne high-
resolution remote sensing images collected from members of NDRCC, we have
High Resolution Remote Sensing Images Based Catastrophe Assessment Method 181

Building
Housing and buildings Bungalow
Chimney and water tower
National road
Provincial road
Transportation County road
infrastructure Bridge
Railway
airport pavement
Municipal road
Municipal Public traffic facility
infrastructure Urban green land
Remote
sensing Important Street lamp
indexes of infrastructures Power generation, substation facility
physical Power facility
Overhead power line
quantity
damage reservoir
assessment Water conservancy channel
facility dam
embankment
telecommunication emission station
and radio
communication transmission line
facility
Pipeline resources Overhead and ground pipeline
Farmland
Land resources Woodland
Natural resources Building land
River
Water area resources
Lake

Table 1. Remote sensing indexes of physical quantity damage assessment

implemented disaster targets recognition and change detection, through which disaster data
is extracted and integrated. The data includes number and area index of houses, number
and length of roads, number and area index of power facilities, water conservation facilities,
and communication stations, municipal utilities, and land resources.

2.5 Direct economic losses assessment

Based on physical quantity assessment results and related standards, Direct Economic Loss
is assessed for houses, infrastructure of government, industry, and personal property within
disaster areas. For example, the economic loss of a building can be calculated by multiplying
182 Remote Sensing – Applications

build-up area, damage rate, and unit price. The damage rate is usually calculated using
damage extent grading method. Other indicators follow the same method in calculating that
of building.

3. Case studies of remote sensing catastrophe assessment

Catastrophe afflicted areas usually have complex terrain and frequently changing weather
in a short time after the disaster, which makes earth observation technology one of the best
methods for disaster emergency monitoring and disaster assessment due to the acquisition
capability of high resolution images. In this section, three cases in China will be discussed in
detail to illustrate the Disaster Assessment Technology NDRCC had used: Wenchuan
Earthqueake, Yushu Earthquake, and Zhouqu Debris Flow Disaster. During Wenchuan
Earthquake in May 2008, we used high resolution airborne remote sensing images, reported
data collected by local government and ground investigation data in combination for
Disaster Scope Assessment (Fan et al., 2008a); during Yushu Earthquake in April 2010, we
implemented housing damage assessment (Yang et al., 2011); during Zhouqu Debris Flow
Disaster in August 2010, we implemented Physical Quantity Assessment and Direct
Economic Loss Assessment (Qi et al., 2011). The Physical Quantity Assessment is
implemented based on disaster scope assessment result with indicators like houses and
building, important infrastructure, land resources, and the Direct Economic Loss
Assessment is implemented based on physical quantity assessment result.

3.1 Wenchuan earthquake

On May 12th, 2008, an earthquake of magnitude 8.0 Richter occurred in the countryside of
Wenchuan, Sichuan province, China. The maximum seismic intensity was 11 degree which
is higher than that of Tangshan Earthquake (7.8 Richter). Aftershocks frequently occurred
and the largest one was of magnitude 6.4 Richter. Four hundred and seventeen counties out
of ten provinces (Sichuan, Gansu, Shanxi, Chongqing, Guizhou, Yunnan, Henan, Hubei,
Shanxi, Hunan) suffered from this disaster, leading to 69197 deaths, 18289 missing and
374000 injuries by July 13th 2010. The highly affected areas are most in mountainous valley
where traffic was not convenient. The earthquake and following aftershock wrecked the
fragile transportation and communication facilities, making disaster relief a very difficult job
since rescue personnel, supplies, vehicles and large-scale disaster rescue equipments
couldn’t access the disaster scene in time.

3.1.1 Preliminary judgment of disaster scope

A seismic intensity distribution map was immediately made after the earthquake using
empirical data from USGS (Figure 3). Results showed that the maximum intensity was
beyond 9 degrees. Then, data of population and areas of affected towns and counties in the
affected regions is calculated according to the map.

3.1.2 Remote sensing assessment of housing damage

Short after the earthquake in Wenchuan, we rapidly collected 1277 remote sensing images
captured by 24 satellites from 12 countries using the International Charter “SPACE AND
MAJOR DISASTERS” and domestic satellites data sharing mechanism. Several airplanes of
High Resolution Remote Sensing Images Based Catastrophe Assessment Method 183

Fig. 3. Seismic intensity distribution map

National Administration of Surveying and Chinese Academy of Sciences flew to affected

area to take airborne remote sensing image, in a resolution of 0.5m and 2m.
Take the indicator of houses for example. With the help of remote sensing images, damage
condition of houses in satellite-covered regions was estimated. For uncovered regions,
damage condition of houses was estimated by proportional spatial extrapolation and
interpolation based on data from satellite-covered regions in a unit of village or town. Based
on destruction probability of general fortified buildings, combined with the seismic intensity
data, the relationship between collapse rate and damage rate was calculated. Damage
condition of housing in affected region was then estimated. Figure 4 illustrates the flow of
damage assessment process for the indicator of houses during Wenchuan Earthquake (Fan

Fig. 4. The flow of collapse and damage assessment process for houses
184 Remote Sensing – Applications

et al., 2008b). Figure 5, 6, and 7 show the damage assessment map of sample regions:
Maoxian County, Qingchuan County and the city of Dujiangyan. Figure 8 demonstrates the
comprehensive assessment result of damage of the whole disaster area for single indicator
of houses.

Fig. 5. The damage assessment result of Maoxian County for houses.

Fig. 6. The damage assessment result of Qingchuan County for houses.

Fig. 7. The damage assessment result of Dujiangyan City for houses.

High Resolution Remote Sensing Images Based Catastrophe Assessment Method 185

Fig. 8. Comprehensive assessment map of damage for houses

3.1.3 Disaster monitoring

In the following days after the quake, we continually and extensively monitored traffic,
main national roads and settlements in disaster area to fast and accurately identify and
locate traffic jam and landslide with the help of high-resolution remote sensing images.
EROS-B satellite image on May 15th illustrates several traffic jams occurred along 213 and
317 national highway in suburban area around Wenchuan due to secondary disaster
landslide (Figure 9).

Fig. 9. Remote sensing image of the area along 213 and 317 national highways in the
countryside of Wenchuan. Be noticed traffic jams are spotted with red lines
186 Remote Sensing – Applications

Quickbird satellite image on May 16th illustrates 5 landslides happened in a 2-kilometre

diameter area around Yongan County, which seriously interrupted traffic (Figure 10).

Fig. 10. Remote sensing image from Quickbird satellite about landslide around Yongan
County

Comparing EROS-B satellite image of 0.7m on May 27th and airborne remote sensing image
of 0.5m resolution on May 18th, about 10 large refugee settlements distributed along both
sides of urban main road in Qingchuan County (Figure 11), which increased significantly in
ten days.

Fig. 11. Remote sensing images about quantity increasing of refugee settlements
High Resolution Remote Sensing Images Based Catastrophe Assessment Method 187

3.1.4 Disaster scope assessment results

Combined with remote sensing interpretation results, ground investigation information and
reported data from local government, we built Comprehensive Disaster Index (CDI) for
Disaster Scope Assessment. The index will be demonstrated in detail as follows:
The weight of average seismic intensity was 0.3; the weight of rate of death and missing per
ten thousands persons was 0.15, and the total weight was 0.3; the weight of houses collapsed
rate per ten thousands was 0.1, and the total weight was 0.2; the weight for geological
disaster risk was 0.1; the weight of relocated rate per ten thousands was 0.1.
Disaster area was partitioned into three sub region (Figure 12): serious region, severe region,
and general region. The area and CDI of serious region was around 26000 km2, and 0.4
above; the area and CDI of severe region was around 90000 km2, and between 0.15 and 0.4;
the area and CDI of general region was around 384000 km2, and between 0.01 and 0.15.

Fig. 12. Disaster scope assessment map of Wenchuan Earthquake

3.2 Yushu earthquake

On April 14th, 2010, a magnitude-7.1 earthquake hit Yushu County, Yushu Tibetan
Autonomous Prefecture, in Qinghai Province of China. The epicentre (pointer on earth
surface directly above) was located in 14-kilometre depth at 33.2 degrees north latitude, 99.6
degrees east longitude, with a seismic intensity 9-degree at maximum. Aftershocks
frequently occurred with the largest one of magnitude-6.3. The earthquake wrecked Jiegu
town and its surrounding area, resulted in 2220 deaths, 70 missing and 12135 injuries by
p.m. 5, April 14th 2010. Many factors, such as high latitude, inconvenient traffic and cold
weather, make disaster relief very difficult.

3.2.1 Preliminary assessment

Firstly, a seismic intensity distribution map was made by earthquake intensity empirical
model according to earthquake magnitude, depth of epicentre and regional fracture zone
188 Remote Sensing – Applications

distribution. Jiegu Town, Yushu County was in the hard-hit area with IX degree. Figure 13
illustrates contour of seismic intensity and terrain distribution.

Fig. 13. Seismic intensity and terrain distribution map

By comparing pre-disaster airborne remote sensing images and post-disaster EROS-B high
resolution satellite images, we analyzed the damage condition of housings, and estimated
disaster scope and damage extent rapidly. Figure 14 illustrates the assessment result.

Fig. 14. Rapid assessment of disaster scope and extent of houses damage

3.2.2 Ground investigation and comprehensive assessment

Immediately after the quake, an airplane of Chinese Academy of Sciences (CAS) was sent to
Yushu to take airborne remote sensing images. Images with a total storage of 409G and 0.4
High Resolution Remote Sensing Images Based Catastrophe Assessment Method 189

meter spatial resolution were acquired, covering 4500 km2 hard-hit area. Then combined
with the result from preliminary assessment process, we partitioned residential region of
Jiegu Town into 3-level grids according to street, function, and structure. About 685 girds of
functional level were indexed.
First level (Figure 15) includes 4 categories: government office, community, industry area,
and restricted area; second level (Figure 16) includes 6 categories: government office, school,
general office, community, industry area, and restricted area; third level (Figure 17) includes
5 categories: building, bungalow, greenhouse, square, and stadium.

Fig. 15. Grids of first level

Fig. 16. Grids of second level

190 Remote Sensing – Applications

Fig. 17. Grids of third level

Hazard assessment expert group was then sent for field investigation. Damage extent,
building size, architecture style, and actual pictures of each third level grid were collected
and sent back to experts in NDRCC for interpretation and verification.

3.2.3 Disaster monitoring

At the same time, some key areas were monitored 24 hours a day using satellite and UAV
images, including landslide region in Road 214, traffic jams in Road 308, refugee settlements
area, terrain.
According to UAV monitoring images of 0.2m in Figure 18, 9861 refugee tents were placed
in Jiegu urban region by April 19th, 5 days after the quake.

Fig. 18. Tent settlement monitoring

High Resolution Remote Sensing Images Based Catastrophe Assessment Method 191

26 landslide regions were spotted, among which 5 were along Road 214 in the south of Jiegu
town, which resulted in traffic jam. The other 21 were mainly located in mountain region
and had little affect to traffic. Figure 19 illustrates traffic jam caused by landslide along Road
214.

Fig. 19. Traffic jam monitoring and assessment in landslide regions

3.2.4 Disaster scope assessment results

Figure 20 illustrates CDI about the disaster scope and extent of loss. CDI for this case is a
little different compared to that of previous Wenchuan Earthquake. For serious regions
covering around 992 km2, CDI was 1; for severe regions of around 7030 km2, CDI was [0.2,1];
for general regions of 27840 km2, CDI was [0.01,0.2].

3.2.5 Housing damage assessment results

The housing damage scope and extent of loss were defined again by integrating field
investigation data, data reported by local government (Figure 21). This is the first time
satellite-airplane-ground integration assessment mode used in national disaster assessment.
The damage scope was partitioned into three parts: collapsed building area, about 1,314,000m2;
serious damaged building area, about 2,332,000 m2; and minor damaged building area, about
680,000 m2. Collapsed building is defined as houses collapsed into ruins; serious damaged
building as structures of houses been destroyed, thus it must be reconstructed; minor
damaged building as structure been slightly affected, and it can still be used after
reinforcement. Figure 8 and table 1 show housing damage assessment results. Figure 22
illustrates the remote sensing hazard monitoring and assessment flow in Yushu Earthquake.
192 Remote Sensing – Applications

Fig. 20. Disaster scope assessment map of Yushu Earthquake

Fig. 21. Housing damage assessment results

High Resolution Remote Sensing Images Based Catastrophe Assessment Method 193

Collapse & Completely collapsed (ten Serious damage (ten thousand Minor damage (ten thousand
damage thousand square meters, %) square meters, %) square meters, %)
Bungalo Buildi Bungal Buildin Bungal Buildin
Total Ratio Total Ratio Total Ratio
type w ng ow g ow g
Residential 118.5 1.6 120.1 27.8 116.6 31.2 147.8 34.2 27.7 8.5 36.2 8.4
Office 0.8 4.9 5.7 1.3 3.8 62.7 66.5 15.3 4.2 17.3 21.5 4.9
School 0.7 0 0.7 0.2 0.2 1.4 1.6 0.4 0 0.9 0.9 0.2
Industrial
and mining 0.1 0.5 0.6 0.1 2.2 4.8 7 1.6 0.6 2.9 3.5 0.8
enterprises
Special 4 0.2 4.2 1.0 8.9 1.4 10.3 2.4 5.4 0.5 5.9 1.4
Total 124.1 7.3 131.4 30.4 131.7 101.5 233.2 53.9 37.9 30.1 68 15.7

Table 2. Housing damage construction area statistic table of Jiegu town urban area.

Fig. 22. Remote sensing hazard monitoring and assessment flow

3.3 Zhouqu debris flow

On August 8th 2010, a serious debris flow disaster occurred in Zhouqu County, Gannan
Tibetan Autonomous Prefecture, Gansu Province, and China. The debris flow breaks many
records since 1949, including number of death, extent of damage, and relief difficulty. The
casualty consists of 1447 deaths, 318 missing, and about 21000 residents were forced to
evacuate by p.m.4, August 24th. The disaster lasted for a long time, and repeatedly occurred
in some regions, resulting in several barrier lakes, and part of town under water. After 8th,
the disaster area had suffering from several heavy rains, which, triggered more debris flow,
and interrupted the traffic. The hard-hit region was in remote mountain area with narrow
road, making disaster relief a very difficult task.
194 Remote Sensing – Applications

3.3.1 Preliminary assessment

An UAV and airplane were sent to Zhouqu one day after the debris to acquire affected area
images, including UAV image of 0.2m resolution and aerial remote sensing images of 1m
resolution. Once the images were sent back, we compared pre-disaster and post-disaster
remote sensing images, continually monitored disaster scope, damage extent of housing,
residents , variation of river water level, newly happened landslide, and preliminary
analyzed and estimated the damage condition and disaster trend.
Figure 23 illustrates preliminary assessment result. Collapsed area was 0.14 km2, among
which more than 90% were bungalows; serious damaged area was 0.15 km2, among which
more than 79% were buildings; minor damaged area was 1.12 km2, among which more than
40% were buildings. Affected crop area was about 0.45 km2, and affected forest area was
about 0.27 km2.

Fig. 23. Preliminary assessment results of Zhouqu Debris Flow disaster

According to debris flow and landslide monitoring result (Figure 24), Sanyanyu debris flow
area was 0.66 km2; Luojiayu debris flow area was 0.2 km2. The 5 landslide regions in the
mountain area of Zhouqu had a total area of 17272 km2.
Figure 25 and Table 3 shows refugee settlement monitoring result. Tents number increased a
lot from August 15th to August 8th.

Tents number in 8th Tents number in

Settlement Area (m2) Increasing number
August 15th August
No.1 high school 2079 13 55 42
No.3 high school 3590 36 98 62
total 5669 49 153 104

Table 3. Comparison of tents number

High Resolution Remote Sensing Images Based Catastrophe Assessment Method 195

Fig. 24. Debris flow and landslide monitoring map

Fig. 25. Settlement monitoring map

The flooded area expanded obviously (Figure 26), with about 226,000 m2 in August 15th,
23,000 m2 more compared with that in August 10th, and 102,000 m2 more compared with that
in May 5th.
196 Remote Sensing – Applications

Fig. 26. Water area monitoring map

3.3.2 Ground investigation

Then according to the preliminary analysis results, we partitioned affection area into 2-level
grids similar with previous case, using house as unit in high-resolution UAV and airborne
remote sensing images. A total of 2457 girds were indexed (Figure 27). Hazard assessment
experts were sent for field investigation. The damage extent, building size, architectural
style, and field pictures of each grid were collected and sent back to experts in NDRCC for
interpretation and verification.

Fig. 27. Field investigation grids partition map

High Resolution Remote Sensing Images Based Catastrophe Assessment Method 197

First level includes 3 categories of 59 grids: seriously damaged area, seriously flooded area,
and minor affect area; Second level includes 2457 subgrids.

3.3.3 Disaster scope and housing damage assessment

The disaster scope and damage extent were defined again by integrating pre-disaster, post-
disaster remote sensing information, ground investigation data and data reported by local
government (Figure 28).
Disaster area was partitioned into three parts: serious affected region refers to the regions
which bore main force of landslide, resulting in serious damage to houses, infrastructure
and farmland; severe affected area refers the region which suffered from a long time soaked
in water and mud, causing damage to houses; minor affected region refers to the region
where houses and infrastructures are mildly damaged by floods and landslides.
The serious affected area was 1.2 km2, severe affected area was 0.2 km2 and minor affected
area was 1.0 km2.

Fig. 28. Disaster scope assessment map of Zhouqu debris flow disaster

3.3.4 Physical quantity assessment

In the following Physical Quantity Assessment process, with the help of data reported by
local government, we carefully assessed the physical quantity of houses, roads, power
facilities, communication facilities, water conservation facilities, municipal utilities and land
resources in UAV and airborne images to estimate approximate direct economic loss. It is
worth to mention that some damage quantity ignored in field investigation was detected in
remote sensing images, such as submerse street trees, telegraph poles, smart street pavilions.
According to remote sensing images and ground investigation verification result (Figure 29),
in rush destroyed region, collapsed residential building area was about 125,000 m2, seriously
damaged area was about 102,000 m2, minor damaged area was about 147,000 m2. In flooded
region, collapsed building area was around 419,827 m2.
198 Remote Sensing – Applications

Fig. 29. Housing damage assessment map

Roads were divided into 3 classes (Figure 30): provincial road, urban road and rural road.
According to remote sensing assessment result, 2 provincial roads were damaged, with a length
of 2 km; 25 urban roads were damaged, with a length of 7.5 km; 5 rural roads were damaged,
with a length of 4.8 km. Three bridges of a total length of 242 meters were out of function.

Fig. 30. Road and bridge damage assessment map

The physical quantity assessment results were used for direct economic losses estimation. It
is for the first time for China that remote sensing images are integrated into the national
catastrophe assessment flow as a major data source.
High Resolution Remote Sensing Images Based Catastrophe Assessment Method 199

4. Conclusion
Throughout the three cases discussed in this chapter, one interesting thing is that the
importance of remote sensing images is increasing in the whole process of disaster
assessment: preliminary assessment step, disaster scope and damage assessment step, and
physical quantity assessment. Also, the resolution of remote sensing images is improved,
from 0.5m to 0.2m, providing more accurate data for assessment and decision making.
Although the whole process of integration of remote sensing images and data from ground
investigation has been growing mature, the sample region choosing, physical quantity
assessment still need further improvement.
Another thing to notice is that, automatic operation system is needed to connect the whole
disaster assessment flow to provide efficient and accurate data for decision making. The
satellite-airplane-ground disaster monitoring and assessment system is the trend we are and
will realize in the near future.

5. Acknowledgment
This chapter is sponsored by High Resolution Earth Observation System Project of China.

6. References
Chen, S.; Ma, H. & Fan, Y. et al. (2008). Road Damage Assessment from High Resolution
Satellite Remote Sensing Imagery in Wenchuan Earthquake. Journal of Remote
Sensing, Vol.12, No.6, pp. 949-955, ISSN 1007-4619.
Fan, Y.; Yang, S. & Wang, L. et al. (2008a). Study on Urgent Monitoring and Assessment in
Wenchuan Earthquake. Journal of Remote Sensing, Vol.12, No.6, pp. 858-864, ISSN
1007-4619.
Fan, Y.; Yang, S. & Wang, W. et al. (2008b). Comprehensive Assessment of Disaster Scope,
In:Comprehensive Analysis and Assessment of Wenchuan Earthquake, Zhu, H., pp. (94-
103), Science Press, ISBN 978-7-03-023855-9, Beijing, China.
Shi, F.; He, H.; Zhang, Y. (2010). Remote Sensing Interpretation of the MS7.1 Yushu
Earthquake Surface Ruptures, Technology for Earthquake Disaster Prevention, Vol.5,
No.2, pp. 220-227, ISSN 1673-5722.
The sixth editing room of Press of China Standards (2010). Compilation of Earthquake and
Related Standards, Press of China Standards, ISBN 978-7-5066-5651-1, Beijing, China.
Wen, Q.; He H. & Wang X. et al. (2011). UAV remote sensing hazard assessment in Zhouqu
Debris Flow Disaster, Proceedings of SPIE European Remote Sensing, ISBN
9780819488022, Prague, Czech Republic, September, 2011.
Xie, L.; Zhang, J. (2000). Application of Satellite Remote Sensing Technology in Earthquake
Disaster Reduction, Journal of Natural Disasters, Vol.9, No.4, pp. 1-8, ISSN 1004-4574.
Yang, S.; Liu, S. & Wu, W. et al. (2011). Remote Sensing Applications in Qinghai Yushu
Earthquake Monitoring and Assessment, Spacecraft Engineering, Vol.20, No.6, pp.
90-96, ISSN 1673-8748.
200 Remote Sensing – Applications

Zhang, D. (1993). Preliminary Study on Visual Interpretation Marks of Building Damages

Caused by Earthquakes on Aero Photograph, Earthquake, Vol.13, No.1, pp. 26-30,
ISSN 1000-3274.
 8

Automatic Mapping of the Lava Flows at

Piton de la Fournaise Volcano, by Combining
Thermal Data in Near and Visible Infrared
Z. Servadio1,2, N. Villeneuve1 and P. Bachèlery3,4
1Laboratoire Géosciences Réunion, Université de la Réunion,
Institut de Physique du Globe de Paris, CNRS, UMR 7154,
Géologie des Systèmes Volcaniques, Saint Denis
2Institut de Recherche pour le Développement, US 140,

BP172, 97492 Sainte-Clotilde Cedex

3Clermont Université, Université Blaise Pascal,

Laboratoire Magmas et Volcans, CNRS, UMR 6524,

Observatoire de Physique du Globe de Clermont-Ferrand,
BP 10448, F-63000 Clermont-Ferrand
4IRD, R 163, LMV, F-63038 Clermont-Ferrand

France

1. Introduction
Knowing the eruptive history of a volcano is an essential key to the understanding of its
functioning, and therefore of the evolution of the character of dangerousness of its
eruptions. For an essentially effusive basaltic volcano such as the Piton de la Fournaise, the
spatial and temporal distribution of lava flows allows to deduct numerous parameters of its
activity, on a magmatic and a structural point of view. Satellite imaging brings more
advantages than the methods used in aerial pictures studies, especially by supplying bigger
temporal and spectral series. The revisiting of satellites over a region can allow the
generation of dynamic mappings of the implementation of the lava flow, and also bring
information on the phenomenology of the eruptions: Surface, volume, flow, spatial
distribution…
Furthermore, satellite images have the advantage of supplying data that grant a global
visualization of the study area, and information on not easily accessible areas. The
interpretation of these satellite data enables obtaining information on the surfaces and
volumes of the lava field flows, but also on its nature and behavior. In a tropical
environment such as La Reunion, where the climatological context presents a strong
cloudiness, a satellite revisit is statistically necessary.
The optical satellite images have already been successfully used to realize mappings of lava
flows. For example, in Nevado Subancaya in Peru (Legelay-Padovanie et al., 1997) or in Etna
in Italy (Honda et al.,2002), the combination of spectral and morphological properties
helped to elaborate surface lava flows mappings and also allowed to individualize the main
202 Remote Sensing – Applications

structures. Lu et al. (2004) propose an association of OPTICAL and RADAR imaging in

order to define more accurate outlines for the lava flows. In the later example, the
downstream part of the lava flow presenting vegetation, the discrimination was realized by
the infrared. The upstream part being a heavily snow covered zone, RADAR images
properties become a precious source of information. Thanks to the use of LIDAR (Airborne
Light Intensity Detection and Ranging) on volcanoes, Digital Ground Models of very high-
resolution can be generated and various retro-reflecting properties of the lava’s different
textures can be studied (Favalli et al., 2009) .The comparison of the different DGM produced
at different dates allow Favalli et al. (2009) to obtain mappings of the thickness and outlines
of the lava flow.
The use of thermal satellite imaging to characterize the relative chronology of the
implementation of lava flows was the aim of Kahle et al. (1998)’s work. Abrams et al. (1991),
used it in association with optical satellite imaging to realize a chronological mapping of the
lava flows in Hawaii. Other authors have realized mappings of lava flow’s temperature
(Hirn et al., 2005) and implementation maps by using a thermal camera (Harris et al., 2007;
James et al., 2007; Lombardo et al., 2009).
At the Piton de la Fournaise, the first research that used photogrammetry to realize
mappings of the outlines and thickness of lava flows took place on lava flows dated back to
1972 and 1976 (Lénat, 1987). Bonneville et al. (1989) mapped the main geological units by
using SPOT1 images. After which, Despinoy (2000) realized a mapping of the lava flows
above Les Grandes Pentes using a CASI hyperspectral sensor. Villeneuve (2000) realized
outlines, volume calculations and a chronological follow up of the implementation thanks to
stereophotogrammetry and the use of DGPS (Digital Global Positioning System). Lénat et al.
(2001) associated RADAR and SPOT images to map the field of lava of the Enclos Fouqué’s
caldera. Recently, De Michele et Briole (2007) used a technic of correlation of images to
extract lava flows which implemented between two series of aerial pictures. The study of the
incoherencies in the interferograms helped realize a dynamic follow up and a mapping of
the lava flows (Tinard, 2007 ; Froger et al., 2007 ; Froger et al., in press).
The aim of this article is to propose an original approach by combining thermal images with
ones acquired in the visible and near infrared in order to extract independent outlines for
each lava flow. The extraction is therefore independent of the operator’s subjectivity. Only
one automatic extraction is possible when associating thermal and optical images for the
implementation of a lava flow. An automatic extraction of the outlines of lava flow is
realized and then compared to a mapping of references realized by photo-interpretation. We
can therefore estimate the precision of the automatic extraction.

2. Context
The Piton de la Fournaise (Reunion Island, Indian Ocean) is a basaltic volcano whose
functioning is connected to the activity of a hot spot (Courtillot et al. 1986). The eruptions
take place inside the Enclos Fouqué’s caldera and have contributed to the creation of a
volcanic cone whose summit is occupied by two craters: the Bory on the east side and the
Dolomieu on the west side (figure 1).
During the past three decades and besides the 1986 and 1998 exceptions, the lava flows have
been taking place inside de Enclos Fouqué’s caldera (figure 1 and 4). They are mainly fed by
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 203

Fig. 1. Major morphological figures of the Piton de la Fournaise, the grey area is the rift zone
defined by Bachèlery (1981).
intrusions situated along the rift zones (Bachèlery, 1981). The intrusions are interconnected
at the same level as the central cone. Eruptions are of three kinds: summit zone eruptions,
proximal eruptions, and distal eruptions (Peltier et al. 2009). It is clear that in the recent
history of the Piton de La Fournaise, summit zone eruptions are the most frequent
(Villeneuve et Bachèlery, 2006; Peltier et al., 2009). Most of these eruptions take place
according to leveled cracks that progressively migrate by furthering from the central cone
(Bachèlery, 1981).
Degassing at the event during eruptions can generate lava fountains which cause pyroclastic
deposits and cones of several meters high. Two morphologies of lava flow are observed at
Piton de la Fournaise: ‘a’a type lava flows and pahoehoe type lava flows.
The occidental part of the Enclos Fouqué is largely recovered by a vast field of lava that
Lénat et al. 2001 name the CLEF (Champ de Lave de l’Enclos Fouqué). This field of lava,
essentially formed by a pahoehoe lava type flow, may have been constituted from slow
emissions from the volcano’s summit zone, between 1750 and 1794.
At the Piton de la Fournaise, the flow of lava emitted by eruptive cracks represent long
(several meters to several kilometers), thin, (about ten meters), shallow (one meter in the
slopes to 5 meters in flat zones (Letourneur et al., 2008)) lava flows, which shows the poor
viscosity of the emitted magma. The juxtaposition of several individual flows during a same
phase will contribute to the constitution of fields of lava, and particularly for long time
eruptions (more than one month). In this case, in regards to the initial lava flow, the new
income contributes essentially to the thickening and widening of the lava field flow. In this
case, the lava field will be considerated here as a same unit.
The spectral properties of the lava flows differ according to the type of the surface (mainly
‘a’a type at Piton de la Fournaise), but also according to the age, either because of a chemical
204 Remote Sensing – Applications

transformation of the rock on the surface, or because of the implantation (always very fast in
Reunion Island) of a vegetation cover (lichen, moss, shrub…). Lava flows which
implementations were separated by several years can therefore be distinguished by their
spectral properties. For the summit zone of the volcano, all the more in the Dolomieu crater,
where the rocks are superimposed with only a few years or a few months of interval, the
spectral properties of the diverse lava flows can then be very similar.
The oldest known eruption at the Piton de la Fournaise dates back to 1644. About 200 events
have been counted since that date thanks to archives, 95% of them took place in the caldera
(Lacroix, 1936; Stieltjes et al., 1989; Peltier et al. 2009). Never the less, this database is
incomplete, particularly in the case of short time and low scale eruptions, before 1980. The
mean average magma emission at Piton de la Fournaise estimated over a century, is 0,01
km3.an-1 (Lénat et Bachèlery, 1987), or 0.3 m3.s-1. The debit estimations show a temporal
estimation. For example, Stieltjes et al. (1989), calculate a mean debit of 0.3 m3.s-1 over 54
years (1931-1985), but obtain 0.78 m3.s-1 for a period of 25 years (1960-1985). These variations
are partly due to the existence of long periods of inactivity. For example, no eruption took
place during 1992 and 1998; witch is to say 6 years of inactivity. Also, another inactivity as
long was observed between 1966 and 1972 (Villeneuve, 2000). Peltier et al. (2009) illustrate
these debits variation and show a more important activity since 1998. Between January 1990
and January 2010, 61 eruptions have been registered with a total volume of emitted lava
estimated at 473 Mm3 (figure 2), and 33 eruptions between 1998 and 2010, with a total
volume of emitted lava of 313 Mm 3 (Peltier et al. 2009, OVPF 2009; 2010). From these
observations, we have calculated a mean debit estimated between 0.45 m³.s-1 and de 0.82
m3.s-1, from 1980 to 2010. These estimations are superior to those obtained by Stieltjes et al.
(1989), on former periods.

Fig. 2. Estimation of the cumulative volume of lava emitted from 1980 to 2010 by the Piton
de la Fournaise.
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 205

3. Method
The originality of this research states in the use of thermal data as an analyze mask. In spite
of its low spatial resolution (90m), thermal data brings essential information in our
automatic mapping method. It allows determining with certitude the zone where the newly
implemented lava flow is localized. The automatic extraction of the outline can be realized
in this analyze mask. Also, its utilization enables treating the lava flows separately from one
another because for one thermal image, only one lava flow is associated in this
methodology. This is particularly adapted in the case of the constitution of a lava field flow.

3.1 Optical and thermal data

The automatic extraction method has been realized by the combination of thermal and
optical data. SPOT and ASTER data have been used. SPOT data have a wavelength from the
visible to mid infrared, and a spatial resolution from 2,5 to 20 meters. ASTER data have a
spectrum from visible to thermal infrared and a spatial resolution that varies from 15m
(VNIR) to 90m (TIR).
The ASTER TIR thermal data have to be acquired at the end of the eruption or very little
time afterwards, in order for the thermal anomaly to be clearly visible on the entire zone.
The maximum post eruption delay of acquisition is variable and depends on the thickness of
the lava flow and therefore on its speed of cooling. In most of the cases, it is less than a
month. ASTER VNIR and SPOT data can be acquired long after the lava flow’s
implementation. The principal is not having a new lava flow implementing on the same
zone. The recent lava flows present low reflectances between the visible and mid infrared
wavelength. The basalt spectrum, in the visible and short wavelength of the infrared (0.4-2.4
µm), is dominated by the presence of iron, which, at different levels of crescent oxidation,
increases the reflectance (Despinoy, 2000). In the same way, the presence of lichen that
grows on the lava flow increases the reflectance. In near and medium infrared, the presence
of chlorophyll in the vegetation induces a strong signal (Kahle et al. 1995), permitting to
discriminate precise outlines in zones with vegetation cover, especially near the Grandes
Pentes area as for the Piton de la Fournaise.
The KALIDEOS project from the CNES (Centre National d’Etudes Spatiales, and GEO Grid
(AIST/METI), and the NASA, grant free satellite data in the case of a research program. The
data used in this article are from SPOT data from the KALIDEOS program, and from the
GEO Grid program for the ASTER (Advanced Spaceborne Thermal Emission and Reflection
Radiometer) data. The climatological conditions must be optimal because the presence of
clouds masks, deforms or reduces the thermal anomaly, which in a tropical zone and
especially in the volcano’s zone is frequent. Only seven ASTER TIR data (the ASTER satellite
was released in orbit in December 1999), acquired at the end of the eruption don’t present
these types of issues, and have therefore been used for the treatments (Table 1 and figure 3).
Optical data acquired after the eruption have been associated to the former images.

3.2 Error calculation and outlines precision of photo-interpretation

The precision of the outlines extracted by automatic method leans entirely on their
comparison with a base of outlines realized from very high-resolution satellite data’ photo
interpretation, and aerial pictures from IGN. This base is considerate as a reference.
206 Remote Sensing – Applications

Fig. 3. Chronology of the acquisitions of the thermal data from the eruptions at the Piton de
la Fournaise.

Table 1. SPOT and ASTER data used for automatic treatments.

The outlines realized after photo interpretation, also known as reference, are not without
errors. They contain errors linked to the operator’s subjectivity and from the resolution of
the data used to extract the outlines. One has to take into consideration these imprecisions
before interpreting the results. This will be approached in the photo interpretation chapter,
where outlines extracted by satellite data’ photo interpretation are tested and compared on
the base of those extracted from IGN’s aerial pictures.
The method chosen lies on the constitution of an error matrix, because it permits to know
the precision of all or part of the classification. It’s expressed in percent or area.
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 207

3.2.1 Error matrix

The error matrix enables comparing two thermal maps. These matrixes are constructed
according to a methodology developed by the Remote sensing Center of Canada
(http://ccrs.nrcan.gc.ca/glossary/index_e.php?id=3124).
In our study case, an outline is compared to another that will be considered as reference
(table 2).
In the matrix, numbers contained by cells correspond to areas. The total of the areas by
column represent the total area of the class obtained by automatic classification. The total of
the areas of each line corresponds to what has been correctly classified. The area values
present in the diagonal correspond to the ones correctly classified. The other cells represent
the areas which were classified wrong, either by omission, or by commission (table 2).

Table 2. An error matrix for two classes

(http://ccrs.nrcan.gc.ca/glossary/index_e.php?id=3124 (modifiée)).
It is therefore possible to calculate a global precision (in percent) of the classification, which
is to say the area of what has been classified correctly in regards to the total area of the zone
that has been classified
∑ Well class area
Pg ∗ 100
total area of the study
It is also interesting to look at the mean precision of each class of classification:
Pmi
Pm
i
The precision of each class is different. Some pixels can be attributed to classes that don’t
suit them, if the spectral properties are similar. The exactitude producer (table 2) bases itself
on the area of the pixels correctly classified in regards to the area of the lava flow considered
as reference. As for the exactitude user (table 2), it bases on the area of the pixel correctly
classified in regard to the area of the lava flow obtained by extraction.
The exactitude producer of the “lava flow” class allows knowing the accuracy of the outline
obtained by automatic extraction:
208 Remote Sensing – Applications

Well class area

Pm lava ∗ 100
Real lava flow area
The mean distance separating outlines vectors is calculated from the area which was not
correctly classified, divided by the smallest of the two compared lava flow outline’s
perimeter:
∑ Not correctly class area
Error β ∗ 100
Perimeter

3.2.2 Photo interpretation

The « referenced » outlines have been realized on a period from 1980 to 2010, by
vectorization on the base of IGN aerial data from 1997 and 2003. As for the eruptions post
August 2003, the research was done on a base of SPOT THR, or SPOT 4 AND 5 satellite data
(Figure 4). The vectorization of the lava flows by using aerial photos is one of the most
precise (Paparoditis et al. 2006), but it is also one of the most binding to realize. It could ask
for several worked days.

Fig. 4. Piton de la Fournaise lava flow cartography between 1980 and 2010 (Servadio et al.
2008 modified).
For the entire French territories, the temporal recurrence in the IGN's acquisitions is 5 years.
It is a problem when dealing with superimposed lava flows. Satellite imaging is then a
complementary tool because, even if the spatial resolution of SPOT data is lower (2,5m to
20m with optical SPOT data, or inferior or equal to 1 meter with aerial photos delivered by
IGN for the BDORTHOs), the temporal resolution grants data between eruptions and
defines outlines for the superimposed zones. Unfortunately, the resolution of the outlines is
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 209

then less precise. The SPOT satellite was put into function in 1986, the outline data base was
completed by the mappings of the OVPF and the BRGM (Stieltjes et al., 1985 et 1989 ; Billard,
1974 ; Bussière, 1967 ; Mc Dougall, 1971).
Our methodology leans on different types of satellite data, it is important to know the
influence of the spatial resolution on the extracted outlines. A comparison between the
outlines obtained by photo-interpretation of different data and the referenced outlines is
then done by using an error matrix and a mean distance between the outlines (table 3).
According to the used satellite data, the awaited error on the outline is about the same size
of the pixel (table 3). The classification’s precision can tell that a mapping by satellite data’
photo-interpretation is 85% more reliable for satellite data with a 10m to 20 m resolution,
and 95% reliable for THR SPOT data (table 3). The other error due to the referenced outlines
extraction can be the consequence of the operator’s subjectivity. It presents an error from 2%
to 5%. The same test was run from aerial photos, and the errors didn’t exceed 2%.

3.2.3 Automatic process: TIR- VNIR method

The lava flows which implement in zones where spectral properties are different, such as a
vegetation zone or a soil that presents high spectral reflectances, are the easiest to identify
(figure 5).

Fig. 5. Examples of automatic extraction of the lava flow outline when the reflectance of the
substrate is very different from that of the lava. A and B: lava flow from the eruption of
September 2009. C and D: lava flow from the eruption of 1986 off enclosures.
To automatically extract an outline, different methods can be used: classifications, threshold,
or automatic detection of change (Inglada et al., 2003; Habib et al., 2007). The distance
between the referenced outlines and the outlines automatically extracted is then
proportional to the pixel’s size, and the lava flow’s classification precision is between 95%
and 99%. On the other hand, lava flows which implement in low spectral reflectance zones,
such as the central cone of the Piton de la Fournaise or the upperstream part of the Grandes
Pentes, ask for more complex treatments. A data treatment methodology is then put
together by using ENVI software (Figure 6). The visible data Principal Composant Analysis
(PCA) is applied in order to maximize the data’s anti-correlation. The thermal bands, near
210 Remote Sensing – Applications

Fig. 6. TIR-VNIR Treatment sequences

infrared and the PCA are grouped in a multi-band data. A multi-level binary classification
named decisional tree classification is then applied by using the following steps:
 By using thermal data, distinguish « hot » from « cold » zones.
 Extract and keep only « hot » zones.
 Discretization of the bare soil and covered vegetation zones by using the Near Infra-red
band.
 Classify the pixels with low anti-correlation values.
The seven lava flows for which ASTER data were available (table 1) were tested. For half of
them, the tested eruptions are on a weak slope zone, with a substrate composed of lava flows
with similar spectral properties; the other half shows various substrates and slopes (figure 7).

Fig. 7. Localization of the studied lava flows

Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 211

Table 3. Continued
212 Remote Sensing – Applications

Table 3. Error matrices of outlines obtained by photo-interpretation, comparing those from the
photo-interpretation of aerial photographs with those of the different types of satellite data.
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 213

The classification achieved, it is then possible to export the « lava flow » class as a vector that
represents the outline of the lava flow (figure 8). The extraction is realized in less than an
hour, once the data collected.

Fig. 8. Automatic extraction of lava flows outlines.

4. Results and interpretations

In order to validate the outlines automatic extraction method, each extraction result has
been compared to the referenced outline. The error matrixes (table 4) represent the tests run
on the seven tested objects localized on figure 7.
The error matrixes enable calculating global errors of classifications between 77% and 96%
(table 4). We can observe a disparity between the summit zone eruptions and those
implemented on sloppy substrates with various spectral properties. The first ones show a
global precision between 77% and 88%, whereas the others variate from 91% to 96%. This is
partly due to low reflectances observed for the substrate at the summit of the volcano. For
example, the outlines of a lava flow newly implemented are hardly distinguished from the
intra Dolomieu lava effusion zone that presents similar ages. The mean precisions show the
same disparity, with values included between 77% and 84% at the summit zone, and
included between 86% and 90% for the lava flows situated on the cone’s flanks and on the
slopes. The area of the lava flows also play a role in the classification precision’s difference,
214 Remote Sensing – Applications

20001012 Lava Flow class No lava flow class

Lava Flow Aerial picture 3.422 0.659 4.081
No Lava Flow Aerial picture 0.667 11.602 12.269
Total Area of aerial picture 4.089 12.261 16.35
Global Accuracy 91.89%
Mean Accuracy 89.205%
Producer accuracy lava flow 83.85%
Producer accuracy no lava flow 94.56%
Mean distance between the outlines 28.41m
20010611 Lava Flow class No lava flow class
Lava Flow Aerial picture 2.620 0.425 3.045
No Lava Flow Aerial picture 0.696 9.039 9.735
Total Area of aerial picture 3.316 9.464 12.78
Global Accuracy 91.228%
Mean Accuracy 89.445%
Producer accuracy lava flow 86.04%
Producer accuracy no lava flow 92.85%
Mean distance between the outlines 28.54m
20030530 Lava Flow class No lava flow class
Lava Flow Aerial picture 0.2386 0.0445 0.2831
No Lava Flow Aerial picture 0.0784 0.3439 0.4223
Total Area of aerial picture 0.317 0.4223 0.7054
Global Accuracy 82.58%
Mean Accuracy 82.855%
Producer accuracy lava flow 84.28%
Producer accuracy no lava flow 81.43%
Mean distance between the outlines 30.45m
20040502 Lava Flow class No lava flow class
Lava Flow Aerial picture 1.8389 0.6246 2.4635
No Lava Flow Aerial picture 0.6610 26.2154 26.8319
Total Area of aerial picture 2.4999 26.84 29.2949
Global Accuracy 95.31%
Mean Accuracy 86.09%
Producer accuracy lava flow 74.64%
Producer accuracy no lava flow 97.70%
Mean distance between the outlines 31.55m
20050217 Lava Flow class No lava flow class
Lava Flow Aerial picture 3.291 0.901 4.192
No Lava Flow Aerial picture 0.816 17.002 17.818
Total Area of aerial picture 4.107 17.903 22.01
Global Accuracy 92.2%
Mean Accuracy 86.965%
Producer accuracy lava flow 78.51%
Producer accuracy no lava flow 95.42%
Mean distance between the outlines 29.65m
20051004 Lava Flow class No lava flow class
Lava Flow Aerial picture 0.2727 0.0761 0.3488
No Lava Flow Aerial picture 0.1462 1.2272 1.3734
Total Area of aerial picture 0.4189 1.3033 1.7222
Global Accuracy 87.1%
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 215

Mean Accuracy 83.8%

Producer accuracy lava flow 78.2%
Producer accuracy no lava flow 89.4%
Mean distance between the outlines 26m
20070218 Lava Flow class No lava flow class
Lava Flow Aerial picture 0.2960 0.0713 0.3673
No Lava Flow Aerial picture 0.0862 0.2617 0.3479
Total Area of aerial picture 0.3822 0.333 0.7152
Global Accuracy 77.98%
Mean Accuracy 77.9%
Producer accuracy lava flow 80.588%
Producer accuracy no lava flow 75.22%
Mean distance between the outlines 55.14m
Table 4. Error matrices for the seven outlines obtained by automatic extraction, by
comparing those from the photo-interpretation of aerial photographs with those from
automatic extraction.
the summit zone lava flows have less superficy than the ones observed along the slopes. An
error on the outline will echo as importantly as the aria of the mapped lava flow is low. As
to say, the more the mapped lava flow’s area is important, the less the error will echo on the
precision of the classification. The exactitude of the producer of the “lava flow” class is the
most important in the validation of an automatic lava flow outline extraction method. The
later varies from 74% and 87% on the entire tested lava flow. There is no observed difference
between the eruptions that occur in the Dolomieu crater and the ones that occur on the
flanks and the Grandes Pentes of the volcano. On another hand, the morphology of the lava
flows plays a role on this precision (figure 8). The more the lava flow’s superficy is
important and the more compact it is, like a cercle shape form or very large, the more
precise the “lava flow” class will be.
The distances between the automatic extracted lava flow outlines, in the case of
superimposed lava flows, and those considered as reference is about 30m (table 4 and figure
9). Now, the satellite data used in the automatic extraction have a visible, near and medium
infrared spatial resolution from 10m to 20m. If the extraction was realized without any
possible confusion, we should have mean distance of about the size of these pixels.
Nevertheless, in some low reflectances zones, only the thermal imaging is able to bound the
outline’s localization zone (figure 8 and 9). It therefore considerably reduces the precision
because this data has a 90m pixel. We notice that in zones where the lava regularly effuses,
the reflectance’s difference cannot allow a free from error extraction of the outline, and it is
therefore deducted from the thermal mask. This mask is essential because it allows the
definition of a zone of interest inside witch the outline can be determinated. There is no
possible confusion: to one thermal data, only one lava flow can be associated.
In vegetalized zones, the extraction can be considered as a simple threshold of the infrared,
the outline is then about the size of the pixel. In some zones, especially the Dolomieu crater
or certain parts of the cone, the differences of reflectances of the lava flows don’t allow to
differentiate them. The outline is then obtained thanks to the thermal data. It can represent
up to 40% of the summit zone lava flows outline, but only 5% to 0% of the proximal and
distal eruption’s outlines. The mean error observed on the outline’s extraction can then
reach twice the size of the optical data’s pixel. Some light diffusion and wedging effects can
interact on the precision.
216 Remote Sensing – Applications

Fig. 9. Automatic lava flow outline extraction of May 2004 (blue) compared with the
referenced outline (purple).
Lets considerate the example of the May 2004 lava flow (figure 9 and table 4). Three zones
are sharply distinguished in the extraction: a very low reflectance and high thermal zone (1),
a low reflectance with few or no thermal diffusion zone (2), and a various reflectance with
few or no thermal diffusion zone (3). The weakest precision is for the first zone, for the
second zone, the low reflectance is due to the substrate’s nature and to luminosity issues in
the zone, because the shadow projected by the rampart can interfere. A luminosity
parameter is to be taken into consideration when choosing and time acquiring SPOT data.
As for the extraction of the third zone, the thermal infrared band essentially obtains it,
because the lava flow implemented in a vegetalized zone with a high spectral signature
difference compared to the lava flow. It is the zone where the extraction is the most precise
because only based on the SPOT data.

5. Discussion and conclusion

The vectors of the outlines obtained by automatic extraction properly match with the
referenced outlines since we obtain a mean exactitude producer of 80% for the “lava flow”
class. If we compare our automatic extraction results to those obtained by DGPS, they are
less precise. The DGPS has a precision of about one centimeter at its antenna, now it’s the
operator who transports it. The error is due to the positioning of the antenna regards to the
outlines. The error then is about one meter, for pluri-meter in our extractions. Nevertheless,
the effusion zones are not all accessible, and cannot let realizing the outlines in their
integrality, especially in the Grandes Pentes. Computer assisted drawing methods are on the
other hand less reliable than our methodology for the data’ distortion are not taken into
account and can provoke hundred of meters error. The results of our classification are close
to those obtained in other contexts’ literature. (Azerzaq et al., 1997 ; Messar et Messar, 1997 ;
Yüksel et al., 2008).
The vectorization time by photo-interpretation can take several days, whereas the automatic
method enables obtaining this vector in less than one hour. There is therefore a considerable
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 217

gain of time. The gaps of observations due to clouds, or zone with high thermal diffusion
and low reflectances, could be filled by using RADAR imaging in treatment sequences
(Weisseil et al.2004). However, by adding Bi data will increase the treatment time, especially
by using coherence data.
The association of thermal and optical data has already been realized in other automatic
classifications with outlines extractions contexts: glaciology (Raciviteanu et al. 2008), the
canopy (Joshi et al., 2006), agriculture (Kasdan, 1979; Saito et al., 2001). For similar spectral
resolution data, the error matrix results are comparable to those obtained by our
methodology.
This methodology was developed in order to automate lava flow outline’s extraction and
therefore ensure a fast update of the Piton de la Fournaise’s database. The lava flow map
was updated thanks to photo-interpretation and automatic extraction (figure 4). It allowed
us to test the reliability of the outlines extracted according to each methodology used in this
article, and to know their precision.
The errors measured by the matrixes give us the extracted surface’s error, by comparing the
automatic lava flow’s area and the referenced one. We saw that: 1/ the difference of
interpretation between two operators for the same data can be of 2% for the aerial photos
and it varies between 2% and 5% with satellite data. 2/if the outlines extracted from satellite
data’s photo interpretation are compared to those extracted from aerial data, the exactitude
producer of the “lava flow” class can reduce to 80%, but is at an average of 90%. 3/ Those
obtained with the same type of data varies from 74% to 87%, which is to say an average of
less than 10%. It influences the obtained area in less than 10%, exception made for low scale
summit zone eruptions which represent an error of 20%, which remains modest and without
major consequences on the volume and production estimation rates for the volcano.

6. Acknowledgements
These reasearch were financed by the “Region Reunion” and the ministry of
“l’Enseignement Supérieur et de la Recherche” program. The authors would like to thank
the CNES (Centre National d’Etudes Spatial) for the free access to the SPOT data via the
KALIDEOS program (http://kalideos.cnes.fr).
This piece of work also used free ASTER data obtained thanks to AIST GEO Grid (Ministry
of Economy, Trade and Industry) programm.
Thanks to Rebecca Roger for the translation.

7. Note
All the lava flow outlines’ vectors will be available online on the Laboratoire de Geosciences
Reunion website.

8. References
Abrams M., Abbott E., KahleA., 1991. Combined Use of Visible, Reflected Infrared, and
Thermal Infrared Images for Mapping Hawaiian Lava Flows. Journal of
Geophysical Research, n°96 (B1), pp. 475-484.
218 Remote Sensing – Applications

Azerzaq M., Assafi M., Fahsi A., 1997. Utilisation des images HRV de SPOT pour la
détection du changement périurbain à Casablanca (Maroc). Télédétection des
milieux urbains et périurbains. Éd. AUPELF-UREF, pp. 275-282.
Bachèlery P., 1981. Le Piton de la Fournaise (Ile de la Réunion). Étude volcanologique,
structurale et pétrologique. Thesis manuscript, Univ. Clermont-Ferrand II, 1-215.
Billard G., 1974. Carte Géologique de l’île de la Réunion. 1 :50000.Bureau de Recherche
Géologique et Minières, Paris. pp. 1-4.
Bonneville A., Lanquette A.M., Pejoux R.1989. Reconnaissance des principales unités
géologiques du Piton de la Fournaise, La Réunion, à partir de SPOT 1. Bulletin de la
société géologique de France, n°87, pp. 1101-1110.
Bussière P., 1967. Carte Géologique du département de La Réunion, 1 :1000000. Service de la
Carte Géologique de la France.
Courtillot V., Besse J., Vandamme D., Montigny R., Jaeger J-J., Cappetta, H., 1986. Deccan
flood basalts at the Cretaceous/Tertiary boundary? Earth and Planetary Sciences
Letters, n°80 (3-4) , pp. 361-374.
De Michele M., Briole P., 2007. Deformation between 1989 and 1997 at Piton de la Fournaise
volcano retrieved from correlation of panchromatic airborne images. Geophysical
Journal International, n°169, pp. 357–364.
Despinoy M., 2000. Potentiel de la télédétection haute résolution spatiale et spectrale en
milieu intertropical. Une approche transdisciplinaire à partir d'expériences
aéroportées à la Réunion et en Guyane française. Thesis manuscript, pp.1-244.
Favalli M., Karátson D., Mazzarini F., Pareschi M.T., Boschi E., 2009, Morphometry of scoria
cones located on a volcano flank: A case study from Mt. Etna (Italy), based on high-
resolution LiDAR data. Journal of Volcanology and Geothermal Research, v. 186, p.
320-330.
Froger J.-L., Fukushima Y., Tinard P., Cayol V., Souriot T., Mora O., Staudacher T., Durand
Ph., Fruneau B., Villeneuve N.,2007. Multi Sensors InSAR Monitoring of Volcanic
Activity: The February & April 2007 Eruptions at Piton de la Fournaise, Reunion
Island, Imaged with ENVISAT-ASAR and ALOS-PALSAR data. Proc. of FRINGE
2007 Workshop, Frascati, Italy,
Habib T., Chanussot J., Inglada J., Mercier G., 2007. Abrupt change detection on
multitemporal remote sensing images: a statistical overview of methodologies
applied on real cases. Proceedings IEEE IGARSS, pp. 2593–2596.
Harris A.J.L., Dehn J., James M.R., Hamilton C., Herd R., Lodato L., Steffke A., 2007,
Pahoehoe flow cooling, discharge, and coverage rates from thermal image
chronometry. Geophysical Research Letters, n° 34, pp. 6.
Hirn B.R., Di Bartola C., Ferrucci F.,2005. Automated, Multi-Payload, High-Resolution
Temperature Mapping and Instant Lava Effusion Rate Determination at Erupting
Volcanoes. Geoscience and Remote Sensing Symposium, 2005. IGARSS '05.
Proceedings. IEEE International. n° 7, pp. 5056 – 5059.
Honda K., Nagai M., 2002. Real-time volcano activity mapping using ground-based digital
imagery. ISPRS Journal of Photogrammetry and Remote Sensing, n°57(1-2), pp.159-
168.
Inglada J., Favard J-C., Yesou H., Clandillon S., Bestault C., 2003. Lava flow mapping
during the Nyiragongo January, 2002 eruption over the city of Goma (D.R. Congo)
in the frame of the international charter space and major disasters. Proceedings of
Automatic Mapping of the Lava Flows at
Piton de la Fournaise Volcano, by Combining Thermal Data in Near and Visible Infrared 219

the IEEE International Geoscience and Remote Sensing Symposium, IGARSS ’03,
n°3, pp.1540- 1542.
James M.R., Pinkerton,H., Robson S., 2007, Image-based measurement of flux variation in
distal regions of active lava flows. Geochem. Geophys. Geosyst., n° 8, pp. Q03006.
Joshi C., De Leeuw J., Skidmore,A.K., Van Dure, I.C., Van Oosten, H., 2006. Remotely
sensed estimation of forest canopy density: a comparison of the performance of
four methods. International Journal of Applied Earth Obeservation and
Géoinformation. n°8, pp.84-95.
Kahle A.B., Abrams M.J., Abbott E.A., Mouginis-Mark P.J., Realmuto V.J., 1995. Remote
sensing of Mauna Loa. Mauna Loa Revealed: Structure, Composition, History, and
Hazards. Geophys. Monogr. N°92, AGU, pp. 145–169.
Kahle A., Gillespie A., Abbott E., Abrams M, Walker R., Hoover G., Lockwood J.P., 1988.
Mapping and relative dating of Hawaiian basalt flows using multispectral thermal
infrared images. Journal of Geophysical Research, n°93( B12), pp. 239-251
Kasdan H., 1979. Feasibility of Using Optical Power Spectrum Analysis Techniques for
Automatic Feature Classification from High Resolution Thermal, Radar, and
Panchromatic Imagery. Recognition systems inc van nuys calif, pp. 1-194
Lacroix A., 1936. Le volcan actif de l’île de la Réunion et ses Produits. Paris, Gauthier et
Villard ed., pp. 1-297
Legelay-Padovani A., Mering C., Guillande R., Huaman D., 1997. Mapping of lava flows
through SPOT images - an example of the Sabancaya volcano (Peru), International
Journal Remote Sensing, n°18(15), pp. 3111-3133.
Lénat J.-F., 1987. Structure et Dynamique internes d'un volcan basaltique intraplaque
océanique : le Piton de la Fournaise (Ile de la Réunion). Thèse de doctorat d'Etat,
Univ. Clermont II, France.
Lénat J.F., Bachèlery P. ,1987. Dynamics of Magma Transfer at Piton de La Fournaise
Volcano (Réunion Island, Indian Ocean), Earth Evolution Sciences, Special Issue
“Modeling of Volcanic Processes”, Chi-Yu and Scarpa (Eds.), pp. 57-72, Friedr.
Vieweg and Sohn, Brauschweig/Wiesbaden
Lénat J.-F., Bachèlery P. et Desmulier F., 2001 . Genèse du champ de lave de l'Enclos Fouqué
: une éruption d'envergure exceptionnelle du Piton de la Fournaise (Réunion) au 18
ème siècle. Bull. Soc. Géol Fr. 2, pp.177-188.
Letourneur L., Peltier A., Staudacher T., Gudmundsson A., 2008, The effects of rock
heterogeneities on dyke paths and asymmetric ground deformation: The example
of Piton de la Fournaise (Réunion Island). Journal of Volcanology and Geothermal
Research, n°173, pp. 289-302.
Lombardo V., Harris A.J.L., Calvari S., Buongiorno M.F., 2009, Spatial variations in lava flow
field thermal structure and effusion rate derived from very high spatial resolution
hyperspectral (MIVIS) data. Journal of Geophysical Research, n° 114, pp. B02208.

Lu Z., Rykhus R., Masterlark T., Dean K.G., 2004, Mapping recent lava flows at Westdahl
Volcano, Alaska, using radar and optical satellite imagery. Remote Sensing of
Environment, n°91(3-4), pp. 345-353.
Messar Y., Messar N., 1997. Apport de la télédétection à la cartographie de l’espace urbain
au 1: 20000 en terme d’emprise au sol du bâti : cas de la ville d’Oran (Algérie).
Télédétection des milieux urbains et périurbains. Éd. AUPELF-UREF., pp. 315-323.
220 Remote Sensing – Applications

Observatoire Volcanologique du Piton de La Fournaise internal annual report, 2009

Observatoire Volcanologique du Piton de La Fournaise internal report, janvier 2010
Paparoditis N., Souchon J-P., Martinoty G., Pierrot-Deseilligny M., 2006. High-end aerial
digital cameras and their impact on the automation and quality of the production
workflow. ISPRS Journal of Photogrammetry & Remote Sensing, n°60, pp. 400–412.
Peltier A., Bachèlery P., Staudacher T., 2009. Magma transport and storage at Piton de la
Fournaise (La Réunion) between 1972 and 2007: A review of geophysical and
geochemical data. Journal of Volcanology and Geothermal Research, n°184, pp. 93-
108.
Raciviteanu A.E., Williams M.W., Barry R.G., 2008. Optical Remote Sensing of Glacier
Characteristics: A Review with Focus on the Himalaya. Sensors , n°8, pp. 3355-3383
Saito G., Ishitsuka N., Matano Y., Kato M., 2001. Application on Terra/ASTER data on
agriculture land mapping. Paper presented at the 22nd Asian Conference on Remote
Sensing, Singapore, pp.1-6
Servadio Z., Villeneuve N., Gladys A., Staudacher T., Urai M., 2008. Preliminary results of
lava flow mapping using remote sensing in Piton de la Fournaise, La Réunion
island. Workshop on the "Use of Remote Sensing Techniques for Monitoring
Volcanoes and Seismogenic Areas", IEEE Naples. Poster session and Proceeding.
Stieltjes L., Moutou P., 1989. A statistical and Probabilitic study of the historic activity of
Piton de la Fournaise, Reunion Island, Indian Ocean. Journal of Volcanology and
Geothermal Research, n°36, pp. 67-86.
Tinard P., 2007. Caractérisation et modélisation des déplacements du sol associés à l’activité
volcanique du Piton de la Fournaise, île de La Réunion, à partir de données
interférométriques. Août 2003 – Avril 2007. Ph.D. Thesis, Univ. Blaise Pascal, pp.1-
334
Villeneuve N., 2000. Apports multi-sources à une meilleure compréhension de la mise en
place des coulées de lave et des risques associés au Piton de la Fournaise :
Géomorphologie quantitative en terrain volcanique. Thesis manuscript, pp.1-378.
Villeneuve N., Bachèlery P., 2006. Revue de la typologie des éruptions au Piton de La
Fournaise, processus et risques volcaniques associés. Cybergéo : revue européenne
de Géographie, n° 336
Yüksel A., Akay A.E., Gundogan R., 2008. Using ASTER Imagery in Land Use/cover
classification of Eastern Mediterranean Landscapes According to CORINE Land
Cover Project. Sensors. n° 8,pp. 1237-1251.
 Section 2

Climate and Atmosphere

 9

Coupled Terrestrial Carbon and Water

Dynamics in Terrestrial Ecosystems:
Contributions of Remote Sensing
Baozhang Chen
Institute of Geographic Sciences and Nature Resources Research,
Chinese Academy of Sciences, Beijing
P.R. China

1. Introduction
The Earth climate is a complex, interactive system, determined by a number of complex
connected physical, chemical and biological processes occurring in the atmosphere, land
and ocean. The terrestrial biosphere plays many pivotal roles in the coupled Earth system
providing both positive and negative feedbacks to climate change (Treut et al., 2007).
Terrestrial vegetation via photosynthesis converts solar energy into carbon that would
otherwise reside in the atmosphere as a greenhouse gas, thereby regulating climate.
Vegetation also transfers water between belowground reservoirs and the atmosphere to
maintain precipitation and surface water flows.The terrestrial carbon (C) cycle is closely
linked to hydrological and nutrient controls on vegetation (Betts et al., 2000; Cox et al., 2000).
Understanding the coupled terrestrial C and water cycle is required to gain a
comprehensive understanding of the role that terrestrial ecosystems play in the global
climate change. Much progress has been made in gaining insight of the coupling processes
between C and water cycles across a range of time and spatial scales (Pielke Sr, 2001;
Friedlingstein et al., 2003; Seneviratne et al., 2006; Betts et al, 2007a,b; Baldocchi, 2008). Since
the early 1990s, there has been an increased interest in monitoring of the CO2, water vapor
and energy exchange between the atmosphere and terrestrial ecosystems by a variety of
methods, such as the eddy-covariance techniques (EC), satellite and other airborne remote
sensing, CO2 concentration and isotope measurements. Meanwhile, there are various kinds
of models have been developed to better understanding of these processes and for large-
scale C and water budgeting.
Remote sensing (RS) from satellite and airborne platforms, along with many other sources of
land ground-based measurements (e.g., eddy covariance flux tower network, biometric
plots, radar network, etc.) is playing and will continue to play a vital role in better
understanding the coupled C and water cycle. Satellite RS allows the study of ecosystems
from a completely new vantage point, facilitating a holistic perspective like viewing the
Earth does for astronauts. Satellite-borne RS offers unique opportunities to parameterize
land surface characteristics over large spatial extents at variable spatial and temporal
resolutions. While there are challenges relating RS data recorded in radiance or backscatter
224 Remote Sensing – Applications

to variables of interest, and RS has poor temporal resolution compared to ground-based

measurement devices, RS and spatial analytical techniques and distributed biogeochemical
modeling embedded in Geographical Information Systems (GIS) have allowed us to better
understand the coupled C and hydrological dynamics across a large range of temporal and
spatial scales.
The large number of papers published since the 1980s on the terrestrial and C/water cycles
have resulted in the publication of several major reviews from different perspectives. For
example, Running et al. (2004) described a blueprint for more comprehensive coordination
of the various flux measurement and modeling activities into a global terrestrial monitoring
network by reviewing the literature published before the middle of 1990s. Baldocchi (2008)
recently provided a comprehensive review of research results associated with a global
network of C flux measurements systems. The topics discussed by this review include
history of the network, errors and issues related with the EC method, and a synopsis of how
these data are being used by ecosystem and climate modellers and the remote-sensing
community (Baldocchi, 2008). Kalma et al. (2008) reviewed satellite-based algorithms for
estimating evepotranspiration (ET) and land surface temperatures at local, regional and
continental scales, with particular emphasis on studies published since the early 1990s;
while Verstraeten et al. (2008) provided a comprehensive review of remote sensing methods
for assessing ET and soil moisture content across different scales based on the literature
published after 1990s. Marquis and Tans (2008) reviewed satellite-based instruments on CO2
concentration measurements.
In this chapter, I distil and synthesise the rapidly growing literature on C and water cycles
using remote sensing in direct or indirect ways across local to global spatial scales and over
a range of time scales. To give the reader a perspective of the growth of this literature, a
search of Web of Science produced over 1500 papers with the key words ‘ecosystem carbon,
water cycles and remote sensing’ published since 1990 which is indicative of the large
amount of research recently being undertaken on these topics. In order to filter through this
large body of literature, I concentrate on papers discussing on the coupling processes
between C and water and I extract information from a database of published results that I
have collated during the past decade (available on request). In terms of content, the report
covers the state of knowledge, monitoring and modeling of the coupled terrestrial C and
water cycles. My aim is to highlight the recent advances in this field, and propose areas of
future research based on perceived current gaps in the literature.
This is a synthesis of state-of-the-art research on how RS has informed the study of coupled
C and hydrology cycles. The review is divided into several inter-connected sections. First, I
review the scientific background of the linkage between terrestrial ecosystems and climate,
and revise the state of knowledge on terrestrial C cycling, coupling of the C and water
cycles. Second, I discuss the ground-based and satellite-based monitoring methods and
observation networks associated with measuring C and water fluxes, CO2 concentration and
C isotopes. Third, I report on the recent advances in modeling approaches associated with
the terrestrial biochemical and hydrological studies. Fourth, I discuss research gaps in C
sinks/sources estimates and finally, I discuss the current research trends and the near-future
directions in this field and propose an upscaling framework for landscape and regional C
and water fluxes estimates.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 225

2. Scientific background and state of knowledge

2.1 Overview of terrestrial ecosystems and climate
The climate system is controlled by a number of complex coupled physical, chemical and
biological processes (Figure 1). The terrestrial biosphere plays a crucial role in the climate
system, providing both positive and negative feedbacks to climate change through
biogeophysical and biogeochemical processes (Treut et al., 2007). Couplings between the
climate system and biogeochemistry are mainly through tightly linked dynamics of C and
water cycles. The importance of coupled C and water dynamics for the climate system has
been increasingly recognized (Cox et al., 2000; Pielke Sr, 2001; Friedlingstein et al., 2003;
Seneviratne et al., 2006; Betts et al, 2000, 2007a,b); however the mechanisms behind these
coupled cycles are still far from well understood.

Fig. 1. Schematic view of the components of the climate system, their processes and
interactions (Treut et al., 2007).

2.2 Terrestrial C cycling

One of the crucial issues in the prognosis of future climate change is the global budget of
atmospheric CO2. The growth rate of atmospheric CO2 is increasing rapidly. Three processes
contribute to this rapid increase: fossil fuel emission, land use change (deforestation), and
ocean and terrestrial uptake. As shown in Figure 2, terrestrial C budgets have large
uncertainties and interannual variability.
Terrestrial ecosystems mediate a large part of CO2 flux between the Earth’s surface and the
atmosphere, with ~120 Pg C yr-1 taken up by photosynthesis and roughly the same amount
226 Remote Sensing – Applications

Fig. 2. Global CO2 budget from 1959 to 2006. Upper panel: CO2 emissions to the atmosphere
(sources) as the sum of fossil fuel combustion, land-use change, and other emissions. Lower
panel: The fate of the emitted CO2, including the increase in atmospheric CO2 plus the sinks
of CO2 on land and in the ocean (Canadell et al., 2007).

released back to the atmosphere by respiration annually (Treut et al., 2007; Prentice et al.,
2001). Imbalances between gross ecosystem photosynthesis or gross primary productivity
(GPP) and ecosystem respiration (Re) lead to land surfaces being either CO2 sinks or sources.
The magnitudes of sinks and sources have fluctuated on annual and longer time scales due to
variable climate, land use change, disturbance, and changes in the age distribution and species
composition of ecosystems (Battle et al., 2000; Arain et al., 2002; Law et al., 2002; Morgenstern
et al., 2004; Humphreys et al., 2005, 2006; Urbanski et al., 2007). Terrestrial ecosystems modify
atmospheric C balance through many mechanisms. A detailed understanding of the
interactive relationships in atmosphere–biosphere exchange is relevant to ecosystem-scale
analysis and is needed to improve our knowledge of the global C cycle (Falk et al., 2008).
In recent years, scientists have learnt that terrestrial ecosystems’ vegetation, soil (Melillo et al.,
1989; Knapp et al., 1993) and animals (Naeem et al., 1995; Hattenschwiler and Bretscher, 2001)
play key roles in mediating the terrestrial C cycle. Plants being the primary producers, it is
from them that mass and energy gets transformed to other living organisms (Engel and Odum,
1999) within an ecosystem. The process of photosynthesis fixes atmospheric C into the
biosphere. Atmospheric CO2 enters the plant through stomatal opening that is controlled by a
variety of environmental factors (Jarvis, 1976; Griffis et al., 2003). These factors include ambient
temperature, atmospheric CO2 concentration, nutrient availability, soil water availability and
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 227

forest age (Schimel, 1995; Prentice et al., 2001). Changes in the atmospheric CO2 concentration
and the corresponding changes in the climate have altered the magnitudes of terrestrial C
cycling. For example, a climate change induced increases in vegetation growth due to earlier
springs and lengthened growing seasons were detected by the phase shift of seasonal
atmospheric CO2 cycle by Keeling et al. (1996) and satellite-based vegetation index analysis by
Myneni et al. (1997). Studies indicate that an increase in atmospheric CO2 enhances
photosynthesis (e.g. Woodward and Friend, 1988) and hence increases assimilation of
atmospheric CO2 by the terrestrial vegetation. Nitrogen (N) availability to plants is another
factor that can affect photosynthesis. This is because N is a primary nutrient for plant growth.
In the recent years, variations in plant N availability have also altered the trends in the
terrestrial C cycles. Variations in plant N availability occur mainly due to natural and
anthropogenic N-deposition. Based on modeling studies, e.g., researchers (Townsend et al.,
1996; Asner et al., 1997; and Holland et al., 1997) have demonstrated that N deposition is
responsible for about 0.1-2.3 PgC yr-1 fixed by terrestrial vegetation which is almost half of the
magnitude of C flux due to fossil fuel emission. Another factor that determines the nature of
terrestrial C balance of an ecosystem is the age of the vegetation. Schimel et al. (1995) have
demonstrated that forest re-growth can account for part of terrestrial C uptake as much as 0.5
± 0.5 PgC yr-1, especially in northern mid and high latitudes. This is because younger
vegetation actively grows and hence sequesters more atmospheric CO2 as opposed to mature
forest stands. There are many other processes that directly and indirectly affect photosynthesis
and thus, the C cycle. They include land use and land cover change (Caspersen et al., 2000;
Houghton and Hackler, 2006; Easter et al., 2007), reforestation (House et al., 2002; Paul et al.,
2002), agricultural and grazing activities (Cerri et al., 2005), insect attack (Chapman et al., 2003;
Throop et al., 2004) and invasive species (Szlavecz et al., 2006). Respiration is a process by
which C is added to the atmosphere from the biosphere. There are studies that indicate that
total ecosystem respiration is a major determinant of terrestrial C balances (Valentini et al.,
2000). Total ecosystem respiration includes respiration by aboveground plant parts (boles,
branches, twigs, and leaves) and soil respiration, which is the sum of the heterotrophic
respiration, and root respiration including respiration of symbiotic microorganisms. The
temporal variability of respiratory metabolism is influenced mostly by temperature and
humidity conditions (Davidson and Janssens, 2006). Although ecosystem respiration has
received considerable attention in recent decades, much less is known about the relative
contributions of its sub- components (Jassal et al., 2007), and our understanding of how they
will respond to global warming is poor. Soil respiration (root + heterotrophic respiration) is a
dominant component of C exchange in terrestrial ecosystems which accounts for more than
half of the total ecosystem respiration (Black et al. 2005). This is because soils of terrestrial
ecosystems contain more C than the atmosphere and live biomass together (Eswaran et al.,
1993). Components of respiration can have different responses to temperature and soil water
content (Boone et al., 1998; Lavigne et al., 2004), thus the effects of these environmental
controls needs to be understood in order to fully comprehend the soil C cycling mechanism.
There are many other mechanisms that can release terrestrial C to atmosphere. This includes
both natural and anthropogenic reasons. Emission of large amounts of C to the atmosphere
from vegetation can occur during forest fires (Amiro et al., 2002; Soja et al., 2004; Amiro et al.,
2004) or biomass burning (Fernandez et al., 1999; Tanaka et al., 2001). These C emissions are of
very high magnitudes although their duration is very short. Forest fires and biomass burning
also affect the nutrient status of the soil which could have positive effects on the succeeding
vegetation (Prietofernandez et al., 1993; Deluca and Sala, 2006). Another form of C flux in
228 Remote Sensing – Applications

almost all terrestrial ecosystems is the import and export of dissolved organic carbon (DOC)
(Neff and Asner 2001; Hornberger et al. 1994). DOC fluxes include C in the form of simple
amino acids to large molecules that are transported through water flows. Fluxes of DOC into
the ocean via runoff from terrestrial ecosystems are estimated to be 0.2 (Harrison et al., 2005) to
0.4 Pg C per year (IPCC, 2001). Since these fluxes are very small compared to the C fluxes due
ecosystem is a net C sink due to the presence of soil C-pools having much longer residence
times (Thompson et al., 1996; Chen et al., 2003; Canadell et al., 2007; Schulze, 2006). The
strength of the terrestrial C sink was estimated to be 0.5-2.0 Pg C yr-1 (Schimel et al., 1995). By
sequestering atmospheric C, the terrestrial ecosystems help decrease the rate of accumulation
of anthropogenic CO2 in the atmosphere, and its associated climate change (Cihlar, 2007).
Terrestrial C sinks may be responsible for taking up about one-third of all the CO2 that is
released into the atmosphere (Canadell et al., 2007). The terrestrial C sink, inferred based on
our current understanding, may not be permanent (Luo et al., 2003; Cox et al., 2000 ;
Friedlingstein et al., 2003). Over the last few years there have been several studies suggesting
that the size of this terrestrial C sink is vulnerable to global warming (Martin et al., 1998;
Nemani et al., 2002; Canadell et al., 2007). The metabolism of terrestrial ecosystems is complex
and highly dynamic because ecosystems consist of coupled, non-linear processes that possess
many positive and negative feedbacks (Levin, 2002; Ma et al. 2007). How the C budget of
major ecosystems will respond to changes in climate is not quantitatively well understood
(Baldocchi & Meyers 1998, Goulden et al., 1998; Black et al., 2000; Baldocchi et al., 2001a;
Baldocchi & Wilson, 2001; Law et al., 2002; Barr et al., 2004, 2007). A detailed understanding of
the interactive relationships in atmosphere-biosphere exchange is relevant to ecosystem-scale
analysis and is needed to improve our knowledge of the global C cycle (Falk,M et al., 2008).
The metabolism of terrestrial ecosystems is complex and highly dynamic because ecosystems
consist of coupled, non-linear processes that possess many positive and negative feedbacks
(Levin et al., 2002; Ma et al., 2007). Complex features of ecosystem metabolism are relatively
unknown and how C budget of major ecosystems will respond to changes in climate is not
quantitatively well understood (Black et al., 2000; Baldocchi et al., 2001; Baldocchi et al.,2001;
Barr et al., 2004; Law et al., 2002).

2.3 Terrestrial water cycling

Most of the Earth is covered by water, amounting to more than one billion km3. The vast
majority of that water, however, is in forms unavailable to land-based or freshwater
ecosystems. Less than 3 percent is fresh enough to drink or to irrigate crops, and of that total,
more than two-thirds is locked up in glaciers and ice caps. Freshwater lakes and rivers hold
100,000 km3 globally, less than one ten-thousandth of all water on earth (Jackson et al, 2001).
Water vapor in the atmosphere exerts an important influence on climate and on the water
cycle, even though only 15,000 km3 of water is typically held in the atmosphere at any time.
This tiny fraction, however, is vital for the biosphere. Water vapor is the most important of
the so-called greenhouse gases (others include CO2, CH4 and N2O) that warm the Earth by
trapping heat in the atmosphere. Water vapor contributes approximately two-thirds of the
total warming that greenhouse gases supply. Without these gases, the mean surface
temperature of the earth would be well below freezing, and liquid water would be absent
over much of the planet. Equally important for life, atmospheric water turns over every ten
days or so as water vapor condenses and falls as rain to the Earth and the heat of the Sun
evaporates new supplies of vapor from the liquid reservoirs on earth. Solar energy typically
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 229

evaporates about 425,000 km3 of ocean water each year. Most of this water returns back
directly to the oceans as precipitation, but approximately 10% falls on land. If this were the
only source of rainfall, average precipitation across the earth’s land surfaces would be only
25 cm a year, a value typical for deserts or semi-arid regions. Instead, a second, larger source
of water is recycled from plants and the soil through evapotranspiration. The water vapor
from this source creates a direct feedback between the land surface and regional climate.
This second source of recycled water contributes two-thirds of the 70 cm of precipitation
that falls over land each year. Taken together, these two sources account for the 110,000 km3
of renewable freshwater available each year for terrestrial, freshwater, and estuarine
ecosystems. Because the amount of rain that falls on land is greater than the amount of
water that evaporates from it, the extra 40,000 km3 of water returns to the oceans, primarily
via rivers and underground aquifers. A number of factors affect how much of this water is
available for human use on its journey to the oceans. These factors include whether the
precipitation falls as rain or snow, the timing of precipitation relative to patterns of seasonal
temperature and sunlight, and the regional topography. For example, in many mountain
regions, most precipitation falls as snow during winter, and spring snowmelt causes peak
flows that flood major river systems. In other regions, excess precipitation percolates into
the soil to recharge ground water or is stored in wetlands.

2.4 Coupling of the C and water cycles

The cycling of other materials such as C and N is strongly coupled to this water flux through
the patterns of plant growth and microbial decomposition, and this coupling creates additional
feedbacks between vegetation and climate. Thermodynamically, a terrestrial ecosystem is an
open system. Therefore, hydrological and C cycles are closely coupled at various temporal and
spatial scales (Betts, 2007; Ball et al., 1987; Levis et al., 1999; Rodriguez-Iturbe, 2000; Joos, 2001;
Arain et al., 2006; Blanken and Black, 2004; Snyder 2004). C uptake for example, is closely
coupled to water loss by ecosystems mainly through leaf stomatal pathway governed
principally through leaf conductance (Jarvis, 1976; Harris et al., 2004; Rodriguez-Iturbe, 2001).
Soil organic C decomposition is very sensitive to soil moisture content via microbial activity
and other processes (Betts, 2007; Levis 1999; Snyder et al., 2004; Parton et al., 1993; D'odorico
2004). The flux of terrestrial organic C by river runoff to the ocean and wetland discharge is an
important component of the global organic C cycle (Hedges, 1992; Wang et al., 2004). It is
estimated that 0.25×1015 g dissolved organic carbon (DOC) is discharged to the ocean by the
world rivers each year (Meybeck, 1982). The land surface hydrological processes (in particular
the terrestrial river systems) play an important role in transport of dissolved and particulate
organic C from terrestrial to marine ecosystems (Wang et al., 2004). However, the interactions
between C and water cycles and the mechanisms how these interactions will shape future
climatic and biosphere conditions are far from well understood.

3. The array of airborne and satellite sensors developed for monitoring of the
coupled C and water cycles
3.1 Satellite monitoring
RS is the observation of a phenomenon from a distance, using devices that detect
electromagnetic radiation. Satellite-borne remote sensing offers unique opportunities to
parameterize land surface characteristics over large spatial extents at variable spatial and
230 Remote Sensing – Applications

temporal resolutions. There has been a substantial increase in the number of satellite sensors
for Earth observations that cover a large range of the electromagnetic radiation spectrum
(Tables 1 and 2) since 1960s when the earlier Landsat satellites were launched into orbit,
such as the Television Infrared Observation Satellite (TIROS-1) launched in 1960. None of
these sensors have been designed exclusively for C, water or vegetation applications. For
example, the TIROS-1 was focused on weather analysis and forecasting (Natl. Res. Counc.,
2008). However, scientists were applying these observations to vegetation studies by the
next decade (Rouse et al., 1974; Tucker et al., 1979). Tuker et al. (1986) exploited the
properties of chlorophyll pigments to absorb wavelengths in the red spectral region and
structural properties of leaves to reflect near-infrared spectra based on the imagery data
obtained by the Advanced Very High Resolution Radiometer (AVHRR) sensor onboard
TIROS. This pioneer study that synoptic view of the coupled atmosphere-biosphere as C
sequestration by photosynthesis from the atmosphere in the Northern Hemisphere (Tuker et
al., 1986) opened possibilities for global perspectives in ecology. The first Landsat satellite
launched in 1972 carried the Multispectral Scanner System (MSS) sensors which were
specifically designed to map land resources with finer spatial resolution (68 m × 82 m) than
the AVHRR. The program was the first civil, non-weather satellite program and Landsat
provided observations for any place on Earth once every 18 days, offering a wide range of
studies on terrestrial vegetation and C and water cycles. The Landsat Thematic Mapper
sensors carried onboard the Landsat series of satellites, acquire images at a 30-m spatial
resolution with a 16-day interval. The acquired data have been the backbone for land-cover,
vegetation and C cycle studies. NASA’s Earth Observing System (EOS), launched in 1999
(Tilford S. 1984), brought new capabilities for monitoring vegetation productivity and other
properties with near-daily and global coverage. The multispectral sensors---Moderate
Resolution Imaging Spectroradiometer (MODIS), onboard the EOS platform, have built
invaluable global observation dataset for C and water cycles research since the early 2000s.
MODIS provides a global coverage every 1-2 days with 36 bands. The spatial resolution of
MODIS (pixel size at nadir) is 250 m for channels 1 and 2 (0.6µm - 0.9µm), 500 m for
channels 3 to 7 (0.4µm - 2.1µm) and 1000 m for channels 8 to 36 (0.4µm - 14.4µm),
respectively. Data from the satellite-borne MODIS are currently used in the calculation of
global weekly GPP and ET at 1-km spatial resolution (Running et al., 2004).
Sensors that have potential applications in C and hydrology studies fall into two groups---
optical (Table 1) and microwave (Table 2). Optical sensors cannot penetrate vegetation or
clouds. In contrast, microwave sensors are able to penetrate vegetation and can collect data
independently of cloud cover and solar illumination. This is important because it is difficult to
acquire cloud-free imagery using optical sensors. There are two types of microwave sensors:
active sensors and passive sensors. The former send and receive their own energy; while the
latter detect the microwaves emitted by the Earth's surface. The microwave bands, being
useful for vegetation and carbon and water cycles, are K, X, C, and L, ranked in increasing
wavelengths. K- and X-bands are useful for detecting surface temperature, snow density, and
rainfall rates, whereas C- and L-bands are sensitive to soil moisture (Sass and Greed, 2011).

3.2 Other airborne measurements

Besides satellite monitoring, other airborne observation techniques (e.g. aircraft, airplane
and land surface remote sensing) have been developed rapidly since the latest decade.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 231

Sensor Bands (nm) Spatial Spatial Temporal Passive Operational

Resolution Coverage Resolution /Active years
Vsible NIR SWIR Thermal
LiDAR - 0.15-1m Global No regular Active -
repeat cycle
GeoEye 450–510 780–920 - - Pan:0.41m Global Less than 3 Passive 2008-09-06 to
510–580 Multi:1.65m days now
655–690
Worldview-2 400-450 860-1040 - - 0. 5m Global 1.1day Passive 2009-10 to now
585-625
705-745
Quikbird 450-520 760-900 - - Pan:0.61m Global 1-3,5days Passive 2001-10-18 to
520-660 Multi:2.44m depending now
630-690 on latitude
Ikonos 450-530 770-880 - - Pan:1m Global 1.5-2.9days Passive 1999-09-24 to
520-610 Multi:4m now
640-720
Orbview-3 450-900 760-900 - - Pan:1m Global >3days Passive 2003-06-26 to
450-520 Multi:4m now
520-600
625-695
KOMPSAT-2 500-900 760-900 - - Pan:1m Global 3 days Passive 2006-07-28 to
450-520 Multi:4m now
Very high resolution

520-600
630-690
Resurs-DK1 580-800 - - - Pan:0.9-1.7m Global 5 days Passive 2006-06-15 to
500-600 Multi:1.5-2m now
600-700
700-800
TopSat 500-700 - - - Pan:2.5m Global 4 days Passive 2005-10-27 to
450-500 Multi:5m now
500-600
600-700
MTI 450-520 760-860 990- 1550- Global 5-8 min Passive 2000-03-12 to
520-600 860-900 1040 1750 now
620-680 910-970 1360- 3500-
1390 4100
4870-
5070
8000-
8400
8400-
8850
10200-
11500
20800-
23500
RapidEye 440-510 760-850 - - 5m Global Daily Passive 2008-8-29 to
EOC 20-590 now
630-685
Formosat-2 450–900 760-900 - - Pan:2m Global Daily Passive 2004-04-21 to
450-520 Multi:8m now
520-600
630-690
232 Remote Sensing – Applications

Sensor Bands (nm) Spatial Spatial Temporal Passive Operational

Resolution Coverage Resolution /Active years
Vsible NIR SWIR Thermal
Spot4 510-730 780–890 1580- - Pan:10m Global 2-3days Passive 1998-03 to now
500 – 1750 Multi:20m
590
610 –
680
Spot5 490-690 780– 890 1580- - Pan:2.5m,5m Global 2-3days Passive 2002-05-04 to
490-610 1750 Multi:10m now
610-680
ALOS 420-500 760-890 - - 10m Global 2days Passive 2006-01-24 to
AVNIR-2 520-600 now
610-690
Terra ASTER 520-600 760-860 1600- - 15m Global 4-16days Passive 1999-12-18 to
630-690 1700 30m now
90m
JERS-1 OPS 520-600 760-876 1600- - 18m Global 44days Passive 1992-02-11 to
630-690 1710 now
2010-
2120
2030-
2250
2270-
2400
SPOT1- 500-730 780-890 - - Pan:10m Global 26days Passive 1986-02 to now
3HRV 500-590 Multi:20m
610-680
Fine

CBERS 500-900 - 1550- 10400- 78m, Global 26days Passive 1994-10-14 to

IRMSS 1750 12500 156m now
2080-
2350
1040-
1250
Deimos-1/ 520– 770-900 - - 22m Global 3 days Passive 2009-07-29 to
UK DMC-2 600 now
630-690
IRS LISS3 20-590 770-860 1550- - 23m Global 24days Passive 2003-10-17 to
620-680 1700 now
2200-
25000
Landsat7 520-900 760-960 1550- 1040- Pan:15m Global 16days Passive 1999-04 to now
ETM 450-520 1750 1250 Muti:30m,60m
520-600 2080-
620-690 3350
Landsat4- 450-520 760-900 1550- 104000- 30m,120m Global 16days Passive 1984-03-16 to
5TM 520-600 1750 125000 now
630-690 20800-
23500
Eo-1 ALI 520-900 760-960 1550- 1040- Pan:10m Global 16days Passive 2000-11-21 to
450-520 1750 1250 Muti:30m,60m now
520-600 2080-
620-690 3350
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 233

Sensor Bands (nm) Spatial Spatial Temporal Passive Operational

Resolution Coverage Resolution /Active years
Vsible NIR SWIR Thermal
Eo-1 400-2500 30m Global 16days Passive 2000-11-21 to
Hyperion now
Meteor 3M-1 450nm-1000 1.4km,3km, Global daily Passive 2001-12-10 to
10500-12500 now
5700-7100
Mos-1,2 510-590 720-800 - 50m global 17days Passive 1987-02-19 to
MESSR 610-690 800-1100 now
-
Okean 530-590 700-800 - 104000- Visible:200m Global 17days Passive 1983-7-10 to
MSU-SK 610-690 900-1000 126000n Ir:200m now
m Thermal:600m
landsat1-5 500-600 700-800 - 80m Global 1,2,3:18 Passive 1972 -07 to now
MSS 600-700 800-1100 days;
- 4,5:16days
CBERS 500- 1550- 104000- Visible, Global 26days Passive 1999-10-14 to
IRMSS 1100 1750 125000n Ir:78m now
800-2350 m Thermal :56m
-
SAC-C 480-500 795-835 1550 - - 175m Global 16days Passive 2000-06-08 to
MMRS 540–560 1700 now
630-690
Terra B1-B36:400-144000 Global 1/4day Passive 1999-12-18 to
MODIS B1-B2:250m now
B3-B7:500m
Medium

B8-B36:1000m
Fengyun-3a 410-125000 B1-B5:250m Global 1/2 day Passive 2008-05-27 to
MERSI B6-B20:1000m now
ENVISAT 410-125000 B1-B5:250m Global 35days Passive 2002-03-1 to
MERSI B6-B20:1000m now
MOS-1,2 520-590 - 115000- Visibl:900m Global 17days Passive 1987-02-19 to
VTIR 620-680 125000 Thermal:2700m now

GEOS Shortwave:3800-4000 Shortwave:4km Global - Active 1985-04 to now

Imager Vsible:550-750 Vsible:1km
Moisture:6500-7000 Moisture:8km
IR:10200-11200 IR:4km
IR:11500-12500
Meteosat 12spectral channels in visible and 1.25km,5km Global daily Passive 2002-12-28 to
SEVIRI near infrad red region now
COARSE

NOAA 550-680 725-1100 3550-3930 1150 1.1km Global 1/4 day Passive 1978-10 to now
AVHRR - 10500-11300 0-
1250
0
GMS5 Visible:550-900 Visible:1.25km Global daily Passive 1995-03-
Water vapour:6500-7000 Water 18~2003-05
IR:10500-11500 vapour:5km
IR: 11500-12500 IR:5km
234 Remote Sensing – Applications

Sensor Bands (nm) Spatial Spatial Temporal Passive Operational

Resolution Coverage Resolution /Active years
Vsible NIR SWIR Thermal
GMS1-4 Visible:550-900 Visible:1.25km Global daily Passive 1997-7 to now
Water vapour:6500-7000 Water
IR:10500-11500 vapour:5km
IR: 11500-12500 IR:5km
Fengyun- Visible:550-1050 Visible:1.25km Global 1 hour Passive C:2004-10-19
2CD Water vapour:6200-7600 IR:5km D: 2006-12--08
IR:10500-12500
fengYun- Visible:550-1050 Visible:1.25km Global 1 hour Passive A:1997-06-10
2AB Water vapour:6200-7600 Water B:1997-07-21
IR:10500-12500 vapour:5km
IR:5km
INSAT-2E 550-750 - 10500- - Visible:2km Global daily Passive 1992-04-02 to
12500 IR:8km now
INSAT-2 550-750 - 10500- - Visible:2km Global daily Passive 1995-12-06 to
VHRR 12500 IR:8km now
Meteosat Visible:450-1000 Visible:1.25km Global daily Passive Meteosat－
MVIRI Water vapour:10500-12500 Water 7:1993-11
IR: 5700-7100 vapour:5km
IR:5km

Abbreviations: Ali, Advanced land Imager; ALOS AVNIR-2, Advanced Visible and Near Infrared
Radiometer type 2; ALOS, Advanced land observing Satellite; ASTER, Japanese Earth Resources
Satellite 1; AVHRR, The Advanced Very High Resolution Radiometer; CBERS, The China–Brazil Earth
Resources Satellite; Deimos-1, Spanish Earth imaging satellite; DMC, Disaster Monitoring Constellation;
Envisat, Environmental Satellite; EOS, Earth Observing System; Etm, Enhanced Thematic Mapper;
Formosat-2, the first and only high-resolution satellite; GMS, Geosynchronous Meteorological Satellite;
HRV, High Resolution Visible ; INSAT-2E, Indian geostationary communications and weather satellite;
KOMPSAT, Korea Multi-Purpose Satellite; IRMSS, Infra-Red Multispectral Scanner; IRMSS, Infrared
Multispectral Scanner Camera; IRS, Indian Remote Sensing; LISS-3, Linear Imaging Self-Scanning
Sensor - 3. Satellites; Lidar, Light Detection And Ranging; MTI, moving target indication radar; MOS -1,
Marine Observation Satellite 1; MESSR, Multi Spectral Electronic Self Scanning Radiometer; MSS, Multi-
spectral Scanner; Meteor 3M-1, Meteorological Satellite; 3M, Monitoring of ocean and land surfaces,
Meteorological observations, and Measurement of vertical profiles of aerosol, ozone and other
constituents in the atmosphere; MMRS, Multispectral Medium Resolution Scanner; MODIS, The
Moderate Resolution Imaging Spectroradiometer; MERSI, Medium Resolution Spectral Imager; MVIRI,
METEOSAT Visible and Infrared Imager; NOAA, National Oceanic and Atmospheric Administration;
Orbview, the satellite of Orbitally company; OPS, Optical System.
Okean MSU-SK: Multispectral Scanner - Conical Scanning; RapidEyeEOC, Electro-Optical Camera;
SEVIRI, Spinning Enhanced Visible Infra-Red Imager; Spot, systeme probatoire d’observation de laterre,
TM, Thematic Mapper; Topsat, Tactical Operational Satellite; UK-DMC 2, British Earth imaging
satellite, operated by DMC International Imaging; VHRR, Very High Resolution Radiometer; VTIR,
Visible and ThermalInfrared Radiometer.

Table 1. Optical Remote Sensing Systems

Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 235

Spatial Sensor Bands Spatial Spatial Temporal Passive/ Operational

resolution Resolution Coverage Resolution Active years
Radarsat-2 C 3m Global 24days Active 2007-12-14tonow
Radarsat-1 C 10m,25m,30m,35m,50m, Global 24days Active 1995-11-04to now
100m
Very high resolution

COSMOS X 1m Global several Passive COSMO1:08.06.2007to

Skymed times a day now
COSMO2:09.12.2007to
now
COSMO3:25.10.2008to
now
COSMO4:06.11.2010to
now
ALOS L 7-100m Global 2days Active 03.01.2002to now
PALSAR
ERS-1,2 C 30m,50km Global 2days Active ERS-1:03.17.1991to now
ERS-2:04.1995to now
TerraSAR- X 1m,3m,16m Global 11days Active 06.15.2007to now
X
7JERS-1 L 18mX18m Global 44days 02.11.1992~10.12.1998
FINE

ENVISAT X 9mX6m;30X30m;150X15 Global 35days Active 03.01.2002to now

ASAR 0m;450mX450m;1800mX
1800m
SIR-C X,C,L 50m Global - Active 04.091994~04.12.1999;
100m 30.091994~11.10.1999;
SRTM X,C 30mX30m 60°N~56°S 16days Active 02.11.2000~02.10.2000
Nimbus-7 Ka,K,Ku, 30km,60km, Global 5~6days Passive 26.10.1978~08.21.1987
medium

SMMR X,C 97.5km,156km

TRMM PR Ku, 4.3～5.0km 50°S~50°N 3h,1d, Active 12.281997 to now

180°W~180°E 3d,7d
TRMM Ka,K,C 6～50km 50°S~50°N 3h,1d, Active 12.28.1997 to now
TMI 180°W~180°E 3d,7d
FengYun 3a Ka,K,Ku, 250m Global 5d Passive 05.27.2008 to now
MWRI X
Aqua Ka,k,x 25km Global 16day Passive 05.04.2002
AMSR-E To now
COARSE

DMSP Ka,k Global 4h Passive 01.19.1965to now

SSMI
DMSP Ka k 174km Global 4h Passive 01.19.1965to mow
SSMT
MOS-1,2 Ka,K 32km,23km Global 17days Passive 02.19.1987~04.19.1996
MSR
ADEOS-1 Ku, 25km 90%global sea 2days Active 08.17.1996~06.30.1997
NSCAT 12.2002~10.24.2003
SMOS L 50km,200km Global 16days Passive 02.11.2009 to now
MIRAS
CRACE Ka,K Global 16days Passive 03.17.2002to now

Abbreviations: ADEOS, Advanced Earth Observing Satellite; ALOS, Advanced land observing Satellite;
AMSR-E, The Advanced Microwave Scanning Radiometer for EOS; ASAR, An Advanced Synthetic
Aperture Radar; COSMO, COnstellation of small Satellites for the Mediterranean basin Observation;
DMSP, Defense Meteorological Satellite Program; Envisat, Environmental Satellite; ERS, European
Remote-Sensing Satellites; GRACE, Gravity Recovery And Climate Experiment; JERS-1, Japanese Earth
236 Remote Sensing – Applications

Resources Satellite 1; MIRAS, Instrument. Synthetic. Apperture Radiometer; MSR, MicrowaveScanning

Radiometer; MWR, The Microwave Radiation Imager; NSCAT, NASA SCATterometer; PR,
Precipitation Radar; SAR, Synthetic Aperture Rada; SIR-C, Space Imaging Radar; SRTM, Shuttle Radar
Topography Mission; SMMR, Scanning Multichannel Microwave Radiometer; SMOS, Soil Moisture and
Ocean Salinity; SSMI, Special Sensor Microwave Imager; SSMT, Special Sensor Microwave
Temperature; TRMM, Tropical Rainfall Mearsuring Mission; TMI,TRMM's Microwave Imager.
Table 2. Microwave Remote Sensing Systems

Microwave wavelengths penetrate greater depths into plant canopies than optical sensors
(Kasischke et al. 1997). The potential for using RADAR (RAdio Detection And Ranging) for
studying terrestrial carbon and water cycles, particularly for assessing standing woody
biomass is promising. The sensitivity of RADAR to vegetation biomass strongly depends on
wavelength: the longer wavelengths, the greater vegetation volumes and biomass levels.
Single-band RADAR is able to detect aboveground biomass up to approximately 100 Mg per
hectare (Dobson et al. 1992, Luckman et al. 1998). In addition, multiband RADAR enables to
separate biomass into component fractions (e.g., stem and canopy) (Saatchi and
Moghaddam 2000). Synthetic aperture RADAR (SAR) is also sensitive to vegetation
structure and to the amount of biomass, including both photosynthetic (green) and
nonphotosynthetic vegetation components (Turner et al., 2004). LiDAR (Light Detection and
Ranging) is a remote sensing technology that determines distances to an object or surface
using laser pulses, which is a relatively new technology compared to optical sensors, and
has the added capability of characterizing the distribution of foliage with height in the
canopy (Lefsky et al. 2002, Treuhaft et al. 2002, 2004; Turner et al., 2004). LiDAR data have
proved to be highly effective for the determination of three dimensional forest attributes.
The suitability of airborne LiDAR for the determination of forest stand attributes including
LAI and the probability of canopy gaps within different layers of canopy has been widely
acknowledged by various studies (Coops et al., 2004; Coops et al., 2007). The interpreted
LiDAR data have been further used for landscape C modeling and scaling (Hilker et al.
2008; Chen et al., 2009). The number and types of sensors used for research on C and water
cycles have multiplied many times over since the first sensor launched into orbit. Remote
sensing provides consistency of coverage and repeat measurements through time are now
indispensable in the C and hydrological scientist’s toolbox.

3.3 Remote sensing of GPP

Satellite-based studies have used the light-use efficiency (ε) approach to estimate GPP
(Prince & Goward, 1995; Running et al., 2000, 2004; Behrenfeld et al., 2001) or net primary
production (NPP) (Field et al., 1995; Ruimy et al., 1999). Significant effort and progress have
been made in developing the satellite-based GPP algorithms (Running et al., 2004; Xiao et
al., 2004; 2005).The algorithm relies on ε approach relating GPP to the amount of absorbed
photosynthetically active radiation (APAR) (Monteith, 1966, 1972), such that,

GPP    fPAR chl  PAR , (1)

where PAR is the photosynetically active radiation (in μmol photosynthetic photon flux
density, PPFD), fPARchl is the fraction of PAR absorbed by leaf chlorophyll in the canopy,
and ε is the light use efficiency (μmol CO2/μmol PPFD). Light use efficiency (ε) is affected
by leaf phenology, temperature, and water:
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 237

   0  Pm  Wm  Tm , (2)

where ε0 is the apparent quantum yield or maximum light use efficiency (μmol CO2/μmol
PPFD) for a given land cover type or vegetation function type, and Pm, Wm and Tm are the
modifiers for the effects of leaf phenology, water and temperature on light use efficiency of
vegetation, respectively.
Different parameters and inputs for the satellite-based algorithm were estimated in different
ways: (i) the fraction of PAR absorbed by leaf chlorophyll in the canopy (fPARchl) and the
modifiers (Pm, Wm); (ii) PAR and temperature modifier (Tm) were calculated using climate
data (either from tower measurements or climate models); and (iii) the maximum light use
efficiency (ε0) was referred to the land-cover-related look-up table and then
modified/optimized using EC tower C measurements and footprint climatology.
To accurately estimate fPARchl in forests is a challenge to both radiative transfer modeling
and field measurements. Significant efforts and progress have been made in developing
advanced vegetation indices that are optimized for retrieval of fPAR from individual optical
sensors (Gobron et al., 1999; Govaerts et al., 1999). The fPARchl within the photosynthetically
active period of vegetation was estimated as a linear function of the the Enhanced
Vegetation Index (EVI),

fPAR  f  EVI  . (3)

EVI is similar in design to NDVI but uses spectral information from the blue band (ρblue).
Following Huete et al. (1997) it was computed,

= ×( − )⁄( + × − × + ), (4)
where G = 2.5, C1 = 6, C2 = 7.5, and L =1. EVI is found to be significantly correlated with the
fraction of the photosynthetically active radiation absorbed by leaf chlorophyll in the
canopy providing a good surrogate of the spatial variability index for photosynthesis rate.
The parameter Pm was estimated using the Normalized Difference Vegetation Index (NDVI)
and the Land Surface Water Index (LSWI) and was calculated at two different phases,
depending upon life expectancy of leaves (deciduous versus evergreen):

= . (5)
1
NDVI (Tucker 1979; Field et al., 1995) was calculated as,

=( − )⁄( + ), (6)
where ρnir, and ρred are the reflectance in the near infrared and red bands, respectively. NDVI is
generally related to green vegetation cover or vegetation canopy density and has been shown
to be well correlated with green LAI and biomass (e.g., Sellers, 1985; Myneni et al., 1995).
LSWI (Xiao et al. 2002) is a useful water index and was calculated as the normalized
difference between the NIR (0.78-0.89 µm) and AWIR (1.58-1.75 µm) spectral bands:

LSWI    nir – swir  /   nir  swir  , (7)

238 Remote Sensing – Applications

where ρnir and ρswir are the reflectance of near infrared bands, red bands and short infrared
bands, respectively.
The timings of bud burst and leaf full expansion can be identified using NDVI. The effect of
water on plant photosynthesis (Wm) has been estimated as a function of available soil
content in plant root zone and water vapor pressure deficit (VPD) in a number of process-
based ecosystem models (e.g. Chen et al., 2007) and remote-sensing based models (e.g.
Running et al., 2000). Soil moisture represents water supply to the leaves and canopy, and
VPD represents evaporative demand in the atmosphere. Leaf and canopy water content is
largely determined by the dynamics of both soil moisture and VPD. As the first order of
approximation, here following the alternative and simple approach that uses a satellite-
derived water index (Xiao et al., 2004), the seasonal dynamics of Wm was estimated,

Wm     1  LSWI  /  1  LSWI max  , (8)

where α is a magnifier (its default value equals 1.0) and LSWImax is the maximum LSWI
within the plant growing season for individual pixels. The temperature modifier Tm was
estimated at each time step, using the equation developed for the terrestrial ecosystem
model (Raich et al., 1991),

(T − T )(T − T ) (9)
=
(T − T )(T − T ) − (T − T )

where Tmin, Tmax and Topt are the minimum, maximum and optimal temperature for
photosynthetic activities, respectively. Their default values are respectively set to be 0, 35
and 20 °C in this study. If air temperature falls below Tmin, Tm is set to be zero.
The ε0 values vary with vegetation types, and the information about ε0 for individual
vegetation types can be obtained from a survey of the literature (Ruimy et al., 1995) and
optimized using EC tower measurements. According to the work (Zhang et al. 2006), the
default ε0 value was estimated to be 0.032 μmol CO2/μmol PPFD in this study stand in 2004.

3.4 Remote sensing of ET

We follow a drop of water traveling through a watershed from input, storage, and finally
output and assess how RS can be used to track water fluxes and reservoirs. Table 3
summarizes the potential application of RS to study of hydrology. ET, the largest
component of water loss from ecosystems, plays an important role in affecting soil moisture,
vegetation productivity, C cycle, and water budgets in terrestrial ecosystems (Dirmeyer,
1994; Hilker et al. 2008; Chen et al., 2009). In this section, I mainly discuss application of RS
to ET.
Verstraeten (Verstraeten et al., 2008) provided a comprehensive review of remote sensing
methods for assessing ET and soil moisture content across different scales and Kalma
(Kalma et al., 2008) reviewed satellite-based algorithms for estimating ET and land surface
temperatures at local, regional and continental scales, with particular emphasis on studies
published since the early 1990s.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 239

In general, water evapotranspired from ecosystems into the atmosphere will reduce the land
surface temperature (Ta). Reduction in soil moisture will decrease plant transpiration and
evaporation from soil and plant surfaces. Reduction in ET will increase Ta. Ta can be derived
from remotely-sensed thermal-infrared (TIR) band (8-14 microns) from various operational
satellites. Based on the relationship between Ta and ET, remotely sensed Ta has been used to
estimate regional ET (Gillies et al,. 1997; Kite et al., 2000; Su et al., 2000., Coops et al., 2002).
The existing thermal imaging sensors provide adequate coverage of thermal dynamics that
are useful for operational monitoring applications of ET. For example, thermal images at 15
minutes intervals and at a spatial resolution of 5 kilometers can be obtained from the NOAA
Geostationary Operational Environmental Satellites (GOES), and TIR data at a fine spatial
resolution (60 m or 120 m) with a much longer time interval (16 days) have been provided
by the Thematic Mapper (TM) and ETM+ instruments on Landsat 5 and Landsat 7.

4. Modeling of C and water dynamics in terrestrial ecosystems based on

remote sensing
The land surface of the Earth represents significant sources, sinks, and reservoirs of C, heat
and moisture to the atmosphere. C and energy fluxes and water cycles at soil-atmosphere and
plant-atmosphere interfaces are therefore important land surface processes. Due to the
complexity and non-linearity of C, N and water dynamics in terrestrial ecosystems, various
modeling tools are needed for better understanding of these biogeochemical and hydrological
processes and their feedback mechanisms with the land surface climate system (Rannik et al.,
2006). The rapidly proliferating volume of spatial data generated by RS has created a
significant challenge in terms of designing model algorithms. A spatially distributed process-
based model uses spatial data for computing ecohydrological and biophysical processes. The
model algorithms represent hypotheses that can be assessed and potentially revised after
confrontation with RS and land surface-based observations. It is well known that realistic
simulations of C and water dynamics in terrestrial ecosystems is of critical importance, not
only for the surface microclimate, but also for the large-scale physics of the atmosphere (Cox et
al., 1999; Gedney et al., 2006; Dickinson et al., 2002). Depending on the scientific objectives or
applications, C and water cycle models have been designed with varying degrees of
aggregation with respect to ecosystem processes, components, and RS data as model inputs.
Such models can be flagged by land surface, ecosystem and hydrological models based on
their objectives and emphases. The former focus on ecosystem processes and the interactions
between ecosystems and the atmosphere; while the latter place emphasis on the land surface
hydrology processes, including lateral flow resulting from catchment topography.

4.1 Land surface and ecosystem modeling

Global climate and the global carbon cycle are controlled by exchanges of water, carbon, and
energy between the terrestrial biosphere and atmosphere. Thus land surface models (LSMs)
are essential for the purpose of developing predictive capability for the Earth's climate on all
time scales (Matthews et al., 1998). Most current LSMs can be associated with three broad
types (Seth et al., 1994): soil-vegetation-atmosphere transfer schemes (SVATS), potential
vegetation models (PVMs), and terrestrial biogeochemistry models (TBMs).
The first generation of SVATS evolved from simple bucket schemes focusing on soil water
availability （Manabe et al., 1969）, through the schemes of Deardorff (Deardorff et al., 1978).
240 Remote Sensing – Applications

Marked improvements of the second generation (e.g., BATS (Seth et al., 1994), SiB (Sellers et al.,
1997; Sellers et al., 1986), and CLASS (Verseghy et al., 1999; Verseghy et al., 1993) from the first
generation are the separation of vegetation from soil and the inclusion of multiple soil layers
for dynamic heat and moisture-flow simulations (Chen et al., 2007). The second generation
SVATS firstly modeled plant physiology in an explicit manner in GCMs (General Circulation
Model or Global Climate Model) (Henderson et al., 1993). For most second-generation SVATS,
land cover was fixed, with seasonally-varying prescriptions of parameters such as reflectance,
leaf area index or rooting depth (Wang et al., 2002; Kickert et al., 1999; Kley et al., 1999;
Schwalm et al., 2001). Some SVATS incorporated satellite data to characterize more realistically
the seasonal dynamics in vegetation function (Kickert et al., 1999; Bonan et al., 1994). The latest
(third generation) SVATS used more recent theories relating photosynthesis and plant water
relations to provide a consistent description of energy exchange, ET, and C exchange by plants
(Chen et al., 2007; Sellers et al., 1996). In our effort in understanding the impact of climate
change on terrestrial ecosystems, energy, water, and C cycles need to be modelled
simultaneously (Sellers et al., 1996; Williams et al., 2001). Recently, most of SVATS have thus
been enhanced to include the CO2 flux between the land surface and the atmosphere, such as
SiB2 (Sellers, P.J et al, 1996), IBIS (Foley et al., 1996), NCAR-LSM (Bonan et al., 1995), BATS
(Dickinson et al., 2002), CLASS-C (Wang et al., 2002) and EASS (Chen et al., 2007).
The earlier generation of PVMs comprised a suite of schemes that focus on modeling
distributions of vegetation as a function of climate (Holdridge et al., 1947; Prentice et al.,
1990) without influences of anthropogenic or natural disturbance. The second generation of
PVMs included more sophisticated modules to account for factors controlling vegetation
distributions, such as competition, varying combinations of plant functional types, and
physiological and ecological constraints (Prentice et al., 1992).
TBMs developed from scaling up local ecological models, are process-based models that
simulate dynamics of energy, water, and carbon and nitrogen exchange among biospheric
pools and the atmosphere (Seth et al., 1994). Few of the existing TBMs incorporate PVMs.
These models are not applicable to transient climate change experiments without coupling
with PVMs.
In recent decades, the interactions among soil, vegetation and climate have been studied
intensively and modeled successfully on the basis of water and energy transfer in the soil-
vegetation-atmosphere system (Seth et al., 1994; Sellers et al., 1986; Verseghy et al., 1999;
Verseghy et al., 1993; Zhang et al., 2003). Also the construction and refinement of LSMs have
received increasing attention (Sellers et al., 1996; Viterbo et al., 1995; Christopher et al.,
2004). Combination of these three different LSMs and utilization of remotely sensed land
surface parameters are critical in the future LSM development, because of (1) the tight
coupling of exchanges of water, energy and carbon between the land surface and the
atmosphere; (2) the sophisticated impact/feedback mechanisms between climate change
and terrestrial ecosystems; and (3) increasingly strong anthropogenic alterations to land
cover. On-line coupling of a LSM with a GCM is needed for studying interannual to multi-
decadal climate variations.
Several model intercomparisons have focused on evaluating SVATS and TBMs with
particular objectives. For instance, the Project for Intercomparison of Land-surface
Parameterization Schemes (PILPS) was initiated to evaluate an array of LSMs existing in
GCMs (General Circulation Model or Global Climate Model) (Henderson et al., 1993); while
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 241

the AMMA (African Monsoon Multidisciplinary Analysis) Land Surface Model

Intercomparison Project (ALMIP) is being conducted to get a better understanding of the
role of soil moisture in land surface processes in West Africa (de Rosnay et al., 2009).
Coordinated land surface modeling activities have improved our understanding of land
surface processes (de Rosnay et al., 2009).

4.2 Spatially-distributed hydrological processes modeling

Hydrology and ecosystem have, for the most part, been studied independently. Most LSMs
and ecosystem models make an assumption of “flat Earth” with the absence of lateral
redistribution of soil moisture. On the other hand, hydrological models have mostly been
concerned with runoff production. Spatially-distributed models are needed, especially for
hydrological simulation objective, because of heterogeneity of land surface and non-linearity
of hydrological processes. Spatially-distributed hydrological models are not only able to
account for spatial variability of hydrological processes, but enable computation of internal
fluxes and state variables. Such kinds of models are increasingly applied to simulate spatial
variability of forcing variables (e.g. precipitation), physiographic characteristics, detailed
processes and internal fluxes within a catchment (Liang et al., 1994; Liang et al., 2004;
Beldring et al., 2003; Brath et al., 2004; Christensen et al., 2007; Reed et al., 2004).

4.3 Modeling dynamics of stable C isotopic exchange between ecosystem and the
atmosphere
It is recognized that the atmospheric measurements are still too sparse, relative to its spatial
variability, to be used for inferring the surface flux at high spatial resolution (Ciais et al.,
1995). The use of the isotope ratio as an additional constraint to identify various C sources
and sinks can contribute to a significant reduction in the uncertainty. Though available
isotopic datasets are being accumulated quickly (Griffis et al., 2005; Ponton et al., 2006; Lai et
al., 2006; Lai et al., 2005) isotope measurements are still lacking considering land surface
diversity and heterogeneity. This shortage of long-term measurements and of sampling
frequency still limits C isotopic studies.
Mechanistic ecosystem models that couple micrometeorological and eco-physiological
theories have the potential to shed light on how to extend efforts and applications of stable
isotopes of CO2 to global C budgeting, because biophysical models have the capacities of
simulating isotope discrimination in response to environmental perturbations and can
produce information on its diurnal, seasonal and interannual dynamics. Few biophysical
models, however, have been developed to assess stable C discrimination between a plant
canopy and the atmosphere (Suits et al., 2005; Oge’e et al., 2003; Baldocchi et al., 2003). Most
existing biophysical models are based on individual leaf level discrimination equations
given by Farquhar et al. (Farquhar et al., 1989; Farquhar et al., 1982) and only focus on the
land surface layer (ignoring vertical and horizontal advection effects beyond 50~100 m
above the ground (Baldocchi et al., 2003). However, in nature, the convective boundary
layer (CBL) integrates the effects of photosynthesis, respiration, and turbulent transport of
CO2 over the landscape (Lloyd et al., 1996; Pataki et al., 2003). The influence of the CBL
cannot be ignored when using isotope composition of CO2 to investigate biological
processes (Bowling et al., 1999), because the effect of atmospheric stability on turbulent
mixing/diffusion has an important impact on scalar fluxes and concentration fields within
242 Remote Sensing – Applications

and above canopies (Baldocchi et al., 1995; Leuning et al., 2000). Few such models
considering the CBL effects on isotope fractionation have been developed to date (Lloyd et
al.,1996; Lloyd et al., 2001; Chen et al,. 2006; Chen et al., 2006; Chen et al., 2007).

4.4 Modeling coupled C and water dynamics – An ecohydrological approach

C and N dynamics and hydrological processes are closely linked. The stomatal conductance
(gs) is the key linkage between C assimilation (photosynthesis) and transpiration. An
empirical equation is used in the second-generation LSMs to calculate gs, which is
hypothesized to be controlled by the environmental conditions (Jarvis et al., 1976). While
field and laboratory studies have documented that leaf photosynthesis also affects gs.
Therefore, Ball et al. (Ball et al., 1987) proposed a semi-empirical stomatal conductance
formulation (Ball-Woodrow-Berry model), in which gs is controlled by both photosynthesis
and the environmental conditions. Most of third-generation LSMs (Ecological models, e.g.
SiB2 (Sellers et al., 1997; Sellers et al., 1996); CN-CLASS (Arain et al., 2006); Ecosys (Grant et
al., 2007; Grant et al., 1999; Chen et al., 2007) fully couple photosynthesis and transpiration
processes by employing the Ball-Woodrow-Berry stamatal conductance formulation.
In addition to the coupling of hydrological condition and C assimilation through the linkage
of gs, C assimilation is also coupling with N dynamics through another biochemical
parameter, Vc25 max --- maximum carboxylation rate at 25 °C. In the photosynthesis model
proposed by Farquhar et al. (1980), the net photosynthetic rate Anet at leaf level is a function
of two tightly-correlated parameters Vc25 25
max and J c max (the maximum electron transport rate
at 25 °C), and is calculated as,

Anet  min( Ac , A j )  Rd (10)

where Ac and Aj are Rubiso-limited and light-limited gross photosynthesis rates,

respectively, and Rd is the daytime leaf dark respiration and computed as Rd = 0.015 Vc max.
Ac and Aj are expressed as,

Cc   *
Ac  Vc max (11a)
C c  K c (1  Oc / K o )

and,

Cc   *
A j  J max (11b)
4(Cc  2  * )

where Cc and Oc are the intercellular CO2 and O2 mole fractions (mol mol−1), respectively;
 * is the CO2 compensation point without dark respiration (mol mol−1); Kc and Ko are
Michaelis-Menten constants for CO2 and O2 (mol mol−1), respectively. In the nutrient-limited
stands, Anet is generally limited by Ac, while Ac is dominantly controlled by a parameter
Vc max (see Eq. 11a). Many research results showed Vc25max is very sensitive to leaf N status
(more specifically leaf Rubisco-N) (Dickinson et al., 2002; Wilson et al., 2000; Wilson et al.,
2001; Warren et al., 2001). As a result in some ecosystem models (i.e. C&N-CLASS (Arain et
al., 2006)), Vc25
max is calculated as a nonlinear function of Rubisco-N following observations
made by Warren and Adams (Dickinson et al., 2002):
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 243

Vc25max ( N )    1  exp( 1.8N r 0 

(12)
where α is the maximum value of Vc25
max and Nr0 is the leaf Rubisco-N (g N m leaf area) in
−2

the top canopy.

The coupled C, N and water processes have been carefully considered in most of the third-
generation LSMs (e.g. SiB2 (Sellers et al., 1997; Sellers et al., 1986; Sellers et al., 1996); CN-
CLASS (Arain et al., 2006) and Ecosys (Grant et al., 2007; Grant et al., 1999), the models’
grids, however, are isolated from their neighboring grids mainly due to the availability of
input data. Vertical soil hydrological processes are hard to be realistically simulated if the
lateral flows are ignored by assuming that the Earth is “flat”. However, Simulations of the
topographically-driven lateral water flows are important components in most of spatially-
distributed models, while the detailed ecophsiological processes are weakly represented
(Govind et al., 2009). Much effort to bridge these two different models has been increasingly
made (Rodriguez et al., 2001; D’odorico et al., 2004;Govind et al., 2009; Creed et al., 1998;
Band et al., 2001; Porporato et al., 2002; Porporato et al., 2003; Daly et al., 2004; Chen et al.,
2005). However, a model coupling approach --- a full combination of ecosystem model and
hydrological model, i.e. ecohydrological modeling, is still lacking.

4.5 Applications of remotely-sensed data in ecohydrological modeling

Remote sensing techniques, which inherently have the ability to provide spatially
comprehensive and temporally repeatable information of the land surface, may be the only
feasible way to obtaining data needed for land surface and ecological modeling (Sellers et
al., 1986; Gurney et al., 2003; Kite et al., 1996; Engman et al., 1996; Melesse et al., 2008). The
most common rationale for interfacing remote sensing and land surface-ecosystem models is
using remotely sensed data as model inputs (Plummer et al., 2000). These input data, corre-
sponding to forcing functions or state variables in ecological modeling, include LC, LAI,
normalized difference vegetation index (NDVI), and the fraction of photosynthetically active
radiation (fPAR) (Sellers et al., 1986; Running et al., 1998; Chiesi et al., 2002; Loiselle et al.,
2001). Another effort is the direct estimation of GPP and net primary productivity (NPP)
(Goetz et al., 1999; Seaquist et al., 2003) of biomass (Seaquist et al., 2003; Bergen et al., 1999)
and of plant growth (Maas et al., 1988; Kurth et al., 1994), by making use of fPAR and NDVI.
It has been shown that the direct estimation has lower accuracy than the integration of
remotely sensed data with process based models (Goetz et al., 1999).
Remote sensing data have also been used to parameterize hydrological models (Chen et al.,
2005; Kite et al., 1996; Boegh et al., 2004). For instance, a hydrological model (TerrainLab)
was further developed using remote sensing as inputs (Chen et al., 2005). TerrainLab is a
spatially distributed, process-oriented hydrological model using the explicit routing scheme
of Wigmosta et al. (Wigmosta et al., 2004). This model has been applied to flat areas such as
boreal and wet land region, (Govind et al., 2009; Chen et al., 2005; Govind et al., 2009), but it
has not yet been applied to mountainous areas.
Different from traditional hydrological models, which have coarse spatial resolutions, the
grid-based-distributed ecohydrological models have a high demand for spatial data (Kite et
al., 1996; Montzka et al., 2008). Some researchers highlight that the main obstacles in current
distributed ecohydrological modeling is the lack of sufficient spatially distributed data for
244 Remote Sensing – Applications

input and model validation (Stisen et al., 2008). Remote sensing can potentially fill in some
of the gaps in data availability and produce means of spatial calibration and validation of
distributed hydrological models. As a result the application of remote sensing techniques in
hydrological studies and water resources management has progressed in the past decades
(see review by (Kite et al., 1996)).
In general, the applications of remotely sensed data in ecohydrological modeling can be in
the two ways (Kite et al., 1996; Chen et al., 2005; Boegh et al., 2004; Montzka et al., 2008;
Stisen et al., 2008; Ritchie et al., 1996; Schultz et al., 1996; Melesse et al., 2007; Schmugge et
al., 2002; Jain et al., 2004; Pietroniro et al., 2005; French et al., 2006): (i) multispectral
remote sensing data are used to quantify surface parameters, such as vegetation types and
density. Although the usefulness of remote sensing data is widely recognized, there
remain few cases where remote sensing data have been actually used in ecohydrological
simulations. Difficulties still exist in choosing the most suitable spectral data for studying
hydrological processes as well as in interpreting such data to extract useful in formation
(Chen et al., 2005; Kite et al., 1996; Engman et al., 1996); and (ii) processed remote sensing
data are used to provide fields of hydrological parameters for calibration and validation
of ecohydrological models, such as precipitation (Kite et al., 1996; Wang et al., 2001), and
soil moisture (Jackson et al., 1993; Hollenbeck et al., 1996., 1996; Kim et al., 2002; Koster et
al., 2006). Koster et al. (Koster et al., 2006) pointed out that remote sensing data take the
form of emitted and reflected radiances and thus are not the type of data traditionally
used to run and calibrate models. Hence, it is important to understand and develop
relationships between the electromagnetic signals and hydrological parameters of interest
(Chen et al., 2005). Kite and Pietroniro (Kite et al., 1996) stated that the use of remote
sensing in hydrological modeling was limited. Even though a number of new sensors
have been launched since then and research has documented that remote sensing data
have promising perspectives, operational uses of satellite data in hydrological modeling
still appear to be in its infancy (Stisen et al., 2008).

5. Research gaps in C and water flux estimates and scaling approaches

A variety of methods are being used in the C and water cycles studies. As shown in Figure
3, different approaches have different temporal and spatial scales. The most direct
measurements of the terrestrial C flux are made either at the plot scale (10−2-101 m2), e.g.
using biometric methods and various forms of chamber, or at the ecosystem (patch) scale
(104 - 106 m2), using the EC technique. Ecohydrological / ecosystem modeling and remote
sensing estimations are generally available across variable spatiotemporal scales. These
estimates are normally available within a nested framework that permits a progressive
comparison of measurements made by surface instrumentation (scale: 1 to 10 m), surface
flux equipment (10 m to 1 km), airborne remote sensing equipment (100 m to several km),
satellite remote sensing (30 m to global scale) and EC tower (1-3 km),
The atmosphere integrates surface fluxes over many temporal and spatial scales and links
scalar sources and sinks with concentrations and fluxes. This principle has been successfully
used to develop inverse models to estimate annual C budgets (Tans et al., 1990; Enting et al.,
1995; Fan et al., 1998; Bousquet et al., 1999; Gurney et al., 2002; Gurney et al., 2003).
However, due to model limitations and paucity of continental CO2 observations these
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 245

studies have yielded C fluxes only at coarse resolution, over large spatial regions (Gurney et
al., 2004; Gurney et al., 2005; Gurney et al., 2008).
Progress in C balance studies has been achieved at both ends of the spatial scale spectrum,
either large continents (larger than 106 km2, e.g. global inverse modeling) or small vegetation
stands (less than 1-3 km2, e.g. EC-measurements). Methods to estimate CO2 sources and
sinks at the intermediate scale (i.e. landscape to regional scales) between continental and
local scales are less well advanced. Moreover, the C cycle in different regions can vary
markedly in response to changing climate (Friedlingstein et al., 2003). Reliable estimates of
terrestrial C sources and sinks at landscape to regional spatial scales (finer than those used
in global inversions and larger than local EC flux measurements and roughly defined as the
range between 102 and 106 km2) are required to quantitatively account for the large spatial
variability in sources and sinks in the near-field of a measurement location (Gerbig et al.,
2003), as well as fundamental to improving our understanding of the C cycle (Crevoisier et
al., 2006).

Fig. 3. Temporal and spatial scales of different approaches

It is generally considered unreliable to upscale stand-level fluxes (i.e. EC measurements) to a

region by simple spatial extrapolation and interpolation because of the heterogeneity of the
land surface and the nonlinearity inherent in ecophysiological processes (Levy et al., 1999).
It is also challenging to apply atmospheric inversion technique to regional scales for
quantifying annual C budgets because at such intermediate scales the atmosphere is often
poorly constrained (Matross et al., 2006; Gloor et al., 1999). Moreover, aggregation errors
and errors in atmospheric transport, both within the PBL and between the PBL and free
troposphere, can also be obstacles to using these approaches to obtain quantitative estimates
of regional C fluxes (Lin et al., 2004). Hence, there is a strong motivation to develop methods
246 Remote Sensing – Applications

to quantify and validate estimates of the C balance at these intermediate scales (Lin et al.,
2004; Chen et al., 2008; Bakwin et al., 2004; Matross et al., 2006). Observations of CO2 over
the continent within the PBL reflect exchange processes occurring at the surface at a regional
scale (102 – 105 km2). The flux information contained in CO2 concentration data represents
footprints of up to 105 km2 (Gloor et al., 2001; Lin et al., 2004), which are several orders of
magnitude larger than the direct EC-flux footprint. This information is therefore needed in
our effort to upscale from site to region. Moreover, the number of CO2 mixing ratio
measurements above the land surface, made by either tower or aircraft, is steadily
increasing. Previous efforts to interpret the signal of regional CO2 exchange making use of
tower concentration data have focused on simple one-dimensional PBL budgets that rely on
gradients in CO2 concentrations between the PBL and the free troposphere (Bakwin et al.,
2004; Helliker et al., 2004). These methods are limited to monthly resolution because of the
need to smooth and average over several synoptic events (Matross et al., 2006).

6. Future research directions

A synthetic research framework is needed to strength the less well researched areas as
reviewed in Section 5: bottom-up and top-down approaches integrating scalable (footprint
and ecosystem) models and a spatially nested hierarchy of observations which include
multispectral RS, inventories, existing regional clusters of eddy-covariance flux towers and
CO2 mixing ratio towers and chambers.
The current research trends and the future directions in this field include: (i) A synthesis
aggregation method --- integrating ecohydrological and isotopic models, remote sensing
and component flux data, is becoming a pragmatic approach towards a better
understanding of the coupled C, N and water dynamics at landscape/watershed scales; and
(ii) The landscape- and regional-scale C fluxes are being estimated using an integrated
approach involving direct land surface measurements, RS measurements, and ecosystem-,
footprint- and inversion- modeling.

7. Summary
After comprehensive reviewing of a variety of approaches being used in research on the
C/water cycles, the concluding remarks are summed the following:
Research gaps in this field are (i) The coupled terrestrial C and hydrological dynamics are
far from well understood, especially at landscape (watershed) and regional scales; (2) Much
progresses have been achieved at the extreme ends of the spatial-scale spectrum, either large
regions/continents or small vegetation stands. Because of the heterogeneity of the land
surface and the nonlinearity inherent in ecophysiological and ecohydrological processes in
response to their driving forces, it is difficult to upscale stand level results to regions and the
globe by extrapolation. Budgets of C and water at landscape intermediate regional scales
(102–105 km2) have large uncertainties.
A coupled spatially-explicit ecohydrological model is a powerful tool for quantitative and
predictive understanding of the coupled C and water mechanism. This modeling framework
can be used to infer aspects of the land surface system that are difficult to measure, and will
be critical to improving the accuracy of forecasts of landscape change and C dynamics in the
real world.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 247

Combining and mutually constraining the bottom-up and top-down methods to reduce
their uncertainties using data assimilation techniques is a practical and effective means to
derive regional C and water fluxes with reasonably high accuracy. In the future upscaling
framework, spatially nested hierarchy of observations, including multispectral RS,
inventories, existing regional clusters of EC flux towers and CO2 mixing ratio towers and
chambers, are able to integrated using scalable (footprint and ecosystem and
ecohydrological) models and data-model fusion techniques.

8. Acknowledgements
This research is financially supported by the National Science Foundation of China (Grant
No. 41071059), “One hundred talents” program funded by Chinese Academy of Sciences
and Alexander Graham Bell Canada Scholarship (CGS) funded by Natural Sciences and
Engineering Research Council of Canada.

9. References
Amiro, B.D.; Barr, A.G.; Black, T.A. Carbon, energy and water fluxes at mature disturbed
forest sites, Saskatchewan, Canada, Agric. For. Meteor. 2006, 136, 237-251.
Arain, M.A.; Yuan, F.M.; Black, T.A. Soil-Plant Nitrogen Cycling Modulated Carbon
Exchanges in a Western Temperate Conifer Forest in Canada. Agric. For. Meteor.
2006, 140, 171-192.
Bakwin, P.S.; Davis, K.J.; Yi, C.; Wofsy, S. C.; Munger, J. W.; Haszpra, L.; Barcza, Z. Regional
carbon dioxide fluxes from mixing ratio data. Tellus B 2004, 56, 301-311.
Baldocchi ,D.D.; Falge, E.; Gu, L. et al. FLUXNET: a new tool to study the temporal and
spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux
densities. Bull Amer. Meteorl. Soc. 2001, 82, 2415-2434.
Baldocchi, D.D. Breathing of the terrestrial biosphere: lessons learned from a global network
of carbon dioxide flux measurement systems. Aust. J. Bot. 2008, 56, 1-26.
Baldocchi, D.D., Bowling, D.R. Modeling the discrimination of 13C above and within a
temperate broad-leaved forest canopy on hourly to seasonal time scales. Plant Cell
Environ. 2003, 26, 231-244.
Baldocchi, D.D.; Harley, P.C. Scaling carbon dioxide and water vapour exchange from leaf
to canopy in a deciduous forest: model testing and application. Plant, Cell and
Environment 1995, 18, 1157–1173.
Baldocchi, D.D.; Wilson, K.B. Modeling CO2 fluxes and water vapor exchange of a temperate
broadleaved forest across hourly to decadal time scales. Ecol. Mod. 2001, 142, 155-184.
Ball, J.T.; Woodrow , I.E.; Berry J.A. A model predicting stomatal conductnance amd its
contribution to the control of photosynthesis under different environmental
conditions, In Progress in Photosynthesis Research, Martinus Nijhoff Publishers:
Dordrecht, Netherland, 1987, pp 221-224.
Band, L.E.; Tague, C.L.; Groffman, P.; Belt, K. Forest Ecosystem Processes at the Watershed
Scale: Hydrological and Ecological Controls of Nitrogen Export. Hydro. Processes
2001, 15, 2013-2028.
Barford, C. C., Wofsy, S. C., and Goulden, M. L.: Factors controlling long- and short-term
sequestration of atmospheric CO2 in a mid-latitude forest. Science, 2001, 294, 1688-
1691.
248 Remote Sensing – Applications

Barr, A.G.; Black, T.A.; Hogg, E.H. et al. Inter-annual variability in the leaf area index of a
boreal aspen-hazelnut forest in relation to net ecosystem production. Agric. For. Me-
teor. 2004, 126, 237–255.
Behrenfeld, M. J., Randerson, J. T., McClain, C. R., Feldman, G. C., Los, S. O., and Tucker, C.
J.: Biospheric primary production during an ENSO transition, Science, 2001, 291,
2594−2597.
Beldring, S.; Engeland, K.; Roald, L.A.; Sælthun, N.R.; Voksø, A. Estimation of parameters in
a distributed precipitation runoff model for Norway. Hydro. Earth System Sci., 2003,
7, 304–316.
Bergen, K.M.; Dobson, M.C. Integration of remotely sensed radar imagery in modeling and
mapping forest biomass and net primary production. Ecol. Model. 1999, 122, 257–
274.
Betts, A.K.; Desjardins, R.L.; Worth D. Impact of agriculture, forest and cloud feedback on
the surface energy budget in BOREAS. Agric. For. Meteor. 2007a, 142, 156–169.
Betts, R.A.; Boucher, O.; Collins, M.; Cox, P.M.; Falloon, P.D.; Gedney, N.; Hemming, D.L.;
Huntingford, C.; Jones, C.D.; Sexton, D.M.H.; Webb, M.J. Projected Increase in
Continental Runoff Due to Plant Responses to Increasing Carbon Dioxide, Nature
2007b, 448, 1037-U5.
Betts, R.A.; Cox, P.M.; Woodward, F.I. Simulated responses of potential vegetation to
doubled-CO2 climate change and feedbacks on near-surface temperature, Global
Ecol. Biogeogr. 2000, 9, 171-180.
Black, T.A.; Chen, W.J.; Barr, A.G. et al. Increased carbon sequestration by a boreal deciduous
forest in years with a warm spring. Geophys. Res. Lett. 2000, 27, 1271-1274.
Blanken, P.D.; Black T.A. The canopy conductance of a boreal aspen forest, Prince Albert
national park, Canada. Hydro. Proces. 2004, 18, 1561–1578.
Boegh, E.; Thorsen, M.; Butts, M.B.; Hansen, S.; Christiansen, J.S.; Abrahamsen, P.; Hasager,
C.B.; Jensen, N.O.; van der Keur, P.; Refsgaard, J.C.; Schelde, K.; Soegaard, H.;
Thomsen, A. Incorporating remote sensing data in physically based distributed
agro-hydrological modelling. J. Hydro. 2004, 287, 279–299.
Bonan, G.B. Comparison of two land surface process models using prescribed forcings, J.
Geophys. Res. 1994, 99, 25,803-25,818.
Bonan, G.B. Land-atmospheric interactions for climate system models: Coupling
biophysical, biogeochemical and ecosystem dynamical processes. Remote Sens.
Environ. 1995, 51, 57-73.
Bousquet, P.; Ciais, P.; Peylin, P.; Ramonet, M.; Monfray, P. Inverse modeling of annual
atmospheric CO2 sources and sinks 1. method and control inversion. J. Geophys. Res.
1999, 104, 26161-26178.
Bowling, D.R.; Baldocchi, D.D.; Monson, R.K. Dynamics of isotope exchange of carbon
dioxide in a Tennessee deciduous forest. Global Biogeochem. Cycles 1999, 13, 903–921.
Brath, A.; Montanari, A.; Toth, E. Analysis of the effects of different scenarios of historical
data availability on the calibration of a spatially-distributed hydrological model. J.
Hydro. 2004, 291, 232–253.
Buermann, W.; Dong, J.; Zeng, X.; Myneni, R.B.; Dickinson, R.E. Evaluation of the utility of
satellite-based vegetation leaf area index data for climate simulations, J Clim, 2001,
14, 3536-3550.
Canadell, J. G.; Que´ re´, C. Le; Raupacha, M. R.; Fielde, C. B.; Buitenhuisc, E.T.; Ciais, P.;
Conwayg, T.J.; Gillettc, N. P.; Houghtonh, R. A.; Marlandi, G. Contributions to
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 249

accelerating atmospheric CO2 growth from economic activity, carbon intensity, and
efficiency of natural sinks. PNAS 2007, 104, 18866-18870.
Chen, B.; Black, A.; Coops, N.C.; Hilker, T.; Trofymow, T.; Nesic, Z.; Morgenstern, K.
Assessing tower flux footprint climatology and scaling between remotely sensed
and eddy covariance measurements. Boundary-Layer Meteorology, 2009a , 130, 137-
167. DOI: 10.1007/s10546-008-9339-1.
Chen, B.; Black, A.; Coops, N.C.; Jassal, R.; Nesic Z. Seasonal controls on interannual
variability in carbon dioxide exchange of a Pacific Northwest Douglas-fir forest,
1997 – 2006, Global Change Biology, 2009b , 15, 1962-1981, doi: 10.1111/j.1365-
2486.2008.01832.x.
Chen, B.; Chen, J. M.; Mo, G.; Yuen, C-W.; Margolis, H.; Higuchi, K.; Chan D. Modeling and
scaling coupled energy, water, and carbon fluxes based on remote sensing: An
application to Canada’s landmass, J. Hydrometeorology, 2007, 8, 123-143,.
Chen, B.; Chen, J.M.; Mo, G.; Black, T.A.; Worthy, D.E.J. Comparison of regional carbon flux
estimates from CO2 concentration measurements and remote sensing based
footprint integration. Global Biogeochem Cycles, 2008, 22, GB2012, doi:10.1029/
2007GB003024.
Chen, B.; Chen, J.M. Diurnal, seasonal and inter-annual variability of carbon isotope
discrimination at the canopy level in response to environmental factors in a boreal
forest ecosystem. Plant, Cell and Environment 2007, 30, 1223-1239, doi:
10.1111/j.1365-3040.2007.01707.
Chen, B.; Chen, J.M.; Huang, L.; Tans, P.P. Simulating dynamics of δ13C of CO2 in the
planetary boundary layer over a boreal forest region: Covariation between surface
fluxes and atmospheric mixing. Tellus B 2006, 58,537-549.
Chen, B.; Chen, J.M.; Ju, W. Remote sensing based ecosystem-atmosphere simulation
Scheme (EASS) --- model formulation and test with multiple-year data. Ecol. Model.
2007, 209, 277-300.
Chen, B.; Chen, J.M.; Mo, G.; Black, T.A.; Worthy, D.E.J. Comparison of regional carbon flux
estimates from CO2 concentration measurements and remote sensing based
footprint integration. Global Biogeochem.Cycles, 2008, 22, GB2012, doi:10.1029/
2007GB003024
Chen, B.; Chen, J.M.; Tans, P.P.; Huang, L. Modeling dynamics of stable carbon isotopic
exchange between a boreal ecosystem and the Atmosphere. Global Change Biol.
2006, 12, 1842-1867.
Chen, J.M.; Chen, X.Y.; Ju, W.M.; Geng, X.Y Distributed Hydrological Model for Mapping
Evapotranspiration Using Remote Sensing Inputs. J. Hydrol. 2005, 305, 15-39.
Chiesi, M.; Maselli, F.; Bindi, M.; Fibbi, L.; Bonora, L.; Raschi, A.; Tognetti, R.; Cermak, J.;
Nadezhdina, N. Calibration and application of FOREST-BGC in a Mediterranean area
by the use of conventional and remote sensing data. Ecol. Model. 2002, 154, 251-262.
Christensen, N.; Lettenmaier, D.P. A multimodel ensemble approach to assessment of
climate change impacts on the hydrology and water resources of the Colorado
River basin. Hydro. Earth System Sci. 2007, 11, 1417–1434.
Christopher, S.R.; Ek, A.R., A process-based model of forest ecosystems driven by
meteorology. Ecol. Model. 2004, 179, 317-348.
Ciais, P; Tans P.P.; Trolier, M; White, J.W.C.; Francey, R.J. A large Northern Hemisphere
terrestrial CO2 sink indicated by the 13C/12C ratio of atmospheric CO2. Science 1995,
269, 1098-1102
250 Remote Sensing – Applications

Coops, N.; Wulder, M.; Culvenor, D.; St-Onge, B. Comparison of forest attributes extracted
from fine spatial resolution multispectral and lidar data. Can. J. Remot. Sens. 2004,
30, 855-866.
Coops, N.C.; Black, T.A.; Jassal, R.S.; Trofymow, J.A.; Morgenstern, K. Comparison of
MODIS, eddy covariance determined and physiologically modelled gross primary
production (GPP) in a Douglas-fir forest stand, Remote Sens Environ, 2007, 107, 385-
401.
Coops, N.C.; Hilker, T; Wulder, M.A.; St-Onge, B.; Newnham G.; Siggins, A.; Trofymow, J.A.
Estimating canopy structure of Douglas-fir forest stands from discrete-return lidar.
Trees , 2007, 21, 295-310.
Cosh, M. H.; Brutsaert, W. Microscale structural aspects of vegetation density variability, J.
Hydrol., 2003, 276,128-136.
Cox, P.M.; Betts, R.A.; Bunton, C.B.; Essery, R.L.H.; Rowntree, P.R.; Smith, J. The impact of
new land surface physics on the GCM simulation of climate and climate sensitivity.
Climate Dyn. 1999, 15, 183-203.
Cox, P.M.; Betts, R.A.; Jones, C.D.; Spall, S.A.; Totterdell, I.J. Acceleration of Global Warming
Due to Carbon-Cycle Feedbacks in a Coupled Climate Model. Nature 2000, 408, 184.
Creed, I.F.; Band, L.E. Exploring Functional Similarity in the Export of Nitrate-N From
Forested Catchments: a Mechanistic Modeling Approach. Water Resour. Res., 1998,
34, 3079- 3093.
Crevoisier, C.; Gloor, M.; Gloaguen, E.; Horowitz, L.W.; Sarmiento, J.; Sweeney, C.; Tans, P.
A direct carbon budgeting approach to infer carbon sources and sinks: Design and
synthetic application to complement the NACP observation network. Tellus B,
2006, 58, 366-375.
Daly, E.; Porporato, A.; Rodriguez-Iturbe, I. Coupled Dynamics of Photosynthesis,
Transpiration, and Soil Water Balance. Part I: Upscaling From Hourly to Daily
Level. J. Hydrometeo. 2004, 5, 546-558.
De Rosnay, P.; Drusch, M.; Boone, A.; Balsamo, G.; Decharme, B.; Harris, P.; Kerr, Y.;
Pellarin, T.; Polcher, J.; Wigneron J.-P. AMMA Land Surface Model
Intercomparison Experiment coupled to the Community Microwave Emission
Model: ALMIP-MEM, J. Geophys. Res., 2009, 114, D05108, doi:10.1029/
2008JD010724.
Deardorff, J.W. Efficient prediction of ground surface temperature and moisture, with
inclusion of a layer of vegetation. J. Geophys. Res., 1978, 83, 1889-1903.
Dickinson, R.E.; Berry, J.A.; Bonan, G.B. et al. Nitrogen controls on climate model
evapotranspiration. J. Climat. 2002, 15, 278–295.
Diego, S.; Peylin, P.; Viovy, N.; Ciais, P. Optimizing a process-based model with eddy-
covariance flux measurements: A pine forest in southern France. Global
Biogeochemical cycles, 2007, 21, GB2013, doi:1029/2006GB002834.
Dirmeyer, P.A. Vegetation stress as a feedback mechanism in mid-latitude drought. J.
Climate , 1994, 7, 1463-1483.
Dobson, M.C.; Ulaby, F.T.; Le Toan, T.; Beaudoin, A.; Kasischke, E.S.; Christensen, N.:
Dependence of radar backscatter on coniferous forest biomass. IEEE Transactions
on Geoscience and Remote Sensing, 1992, 30, 412–415.
D'odorico, P.; Porporato, A.; Laio, F.; Ridolfi, L.; Rodriguez-Iturbe, I. Probabilistic Modeling
of Nitrogen and Carbon Dynamics in Water-Limited Ecosystems. Ecol. Model. 2004,
179, 205-219.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 251

Drolet, G.G.; Middleton, E.M.; Huemmrich, K.F.; Hall, F.G.; Amiro, B.D.; Barr, A.G. Regional
mapping of gross light-use efficiency using MODIS spectral indices, Remote., Sens.
Environ., 2008, 112, 3064-3078.
Efron, B.; Tibshirani, R. J. An Introduction to the Bootstrap. Chapman & Hall/CRC, Boca
Raton, 1993.
Engman, E.T. Remote sensing applications to hydrology: future impact. Hydro. Sci. J. 1996,
41, 637-647.
Enting, I.G.; Trudinger, C.M.; Francey, R.J. A synthesis inversion of the concentration and
δ13C of atmospheric CO2. Tellus B 1995, 47, 35-52.
Falge, E.; Baldocchi, D.; Olson, R. et al. Gap filling strategies for defensible annual sums of
net ecosystem exchange, Agric. For. Meteorol., 2001, 107, 43–69.
Falk, M.; Wharton, S.; Schroeder, M. et al. Flux partitioning in an old-growth forest: seasonal
and interannual dynamics. Tree Physiol. 2008, 28, 509-520.
Fan, S.; Gloor, M.; Mahlman, J.; Pacala, S.; Sarmiento, J. et al. A large terrestrial carbon sink
in North America implied by atmospheric and oceanic carbon dioxide data and
models. Science 1998, 282, 442-446.
Farquhar, G.D.; Ehleringer, J.R.; Hubick, K.T. Carbon isotope discrimination and
photosynthesis. Annual Review of Plant Physiology and Plant Molecular Biol. 1989, 40,
503–537.
Farquhar, G.D.; O’Leary, M.H.; Berry, J.A. On the relationship between carbon isotope
discrimination and the intercellular carbon dioxide concentration in leaves. Aust. J.
Plant Physi. 1982, 9, 121–137.
Field, C. B.; Randerson, J. T.; Malmstrom, C. M. Global net primary production-combining
ecology and remote sensing. Remote Sensing of Environment, 1995, 51, 74−88,.
Finnigan, J. The footprint concept in complex terrain, Agric For Meteorol., 2004,127,117-129,.
Foken, T., Leclerc, M.Y.: Methods and limitations in validation of footprint models.
Agricultural and Forest Meteorology, 2004, 127, 223-234.
Foley, J.A.; Prentice, I.C.; Ramankutty, N.; Levis, S.; Pollard, D.; Sitch, S.; Haxeltine, A.; An
integrated biosphere model of land surface processes, terrestrial carbon balance,
and vegetation dynamics. Global Biogeochem. Cycles 1996, 10, 603-628.
French, R.H.; Miller, J.J.; Dettling, C.; Carr, J. Use of remotely sensed data to estimate the
flow of water to a playa lake. J. Hydro.2006, 325, 67–81.
Friedlingstein, P.; Dufresne, J.L.; Cox, P.M.; Rayner, P. How Positive Is the Feedback
Between Climate Change and the Carbon Cycle? Tellus B 2003, 55, 692-700.
Gao, F., Masek, J., Schwaller,M., and Hall, H: On the blending of the Landsat andMODIS
surface reflectance: Predicting daily Landsat surface reflectance. IEEE Transactions
on Geosciences and Remote Sensing, 2006, 44, 2207−2218,.
Gedney, N.; Cox, P.M.; Betts, R.A.; Boucher, O.; Huntingford, C.; Stott, P.A. Detection of a
Direct Carbon Dioxide Effect in Continental River Runoff Records. Nature, 2006,
439, 835-838.
Gerbig, C.; Lin, J. ; Wofsy, S.C.; Daube, B.C.; Andrews, A.E., et al. Towards constraining
regional scale fluxes of CO2 with atmospheric observations over a continent: 1.
Observed spatial variability from airborne platforms. J. Geophys. Res. 2003, 108,
D4756,10.1029/2002JD003018.
Gillies, R.R.; Cui, J.; Carlson, T.N.; Kustas, W.P.; Humes, K.S. Verification of a method for
obtaining surface soil water content and energy fluxes from remote measurements
of NDVI and surface radiant temperature. Inter. J. Remot. Sens., 1997, 18, 3145–3166.
252 Remote Sensing – Applications

Gloor, M.; Bakwin, P.; Hurst, D.; Lock, L.; Draxler, R.; Tans, P. What is the concentration
footprint of a tall tower? J. Geophys. Res. 2001, 106, 17,831-17,840.
Gloor, M.; Fan, S.M.; Pacala, S.; Sarmiento, J.; Ramonet, M. A model-based evaluation of
inversions of atmospheric transport, using annual mean mixing ratios, as a tool to
monitor fluxes of nonreactive trace substances like CO2 on a continental scale. J.
Geophys. Res. 1999, 104, 14245-14260.
Gobron, N.; Pinty, B.; Verstraete, M.; Govaerts, Y. The MERIS Global Vegetation Index
(MGVI): Description and preliminary application, International Journal of Remote
Sensing, 1999, 20, 1917– 1927.
Goetz, S.J.; Prince, S.D.; Goward, S.N.; Thawley, M.M.; Small, J. Satellite remote sensing of
primary production: an improved production efficiency modelling approach. Ecol.
Model. 1999, 122, 239–255.
Gökede, M.; Rebmann, C.; Foken, T. A combination of quality assessment tools for eddy co-
variance measurements with footprint modelling for the characterisation of
complex sites, Agric. For. Meteorol., 2004,127, 175–188.
Govaerts, Y. M.; Verstraete, M. M.; Pinty, B.; Gobron, N. Designing optimal spectral indices:
A feasibility and proof of concept study, International Journal of Remote Sensing,
1999, 20, 1853- 1873.
Govind, A.; Chen, J. M.; Ju, W. Spatially explicit simulation of hydrologically controlled
carbon and nitrogen cycles and associated feedback mechanisms in a boreal
ecosystem, J. Geophys. Res., 2009, 114, G02006, doi:10.1029/2008JG000728.
Govind, A.; Chen, J.M.; Margolis, H.; Ju, W.; Sonnentag, O.; Giasson, M.-A. A spatially
explicit hydro-ecological modeling framework (BEPS-TerrainLab V2.0): Model
description and test in a boreal ecosystem in Eastern North America. J. Hydro..
2009, 367, 200-216.
Grant, R.F.; Black, T.A.; den Hartog, G. et al. Diurnal and annual exchanges of mass and
energy between an aspen-hazelnut forest and the atmosphere: testing the
mathematical model Ecosys with data from the BOREAS experiment. J. Geophys.
Res. 1999, 27, 27699-27718.
Grant, R.F; Flanagan, L.B. Modeling stomatal and nonstomatal effects of water deficits on
CO2 fixation in a semiarid grassland. J. Geophys. Res. 2007, 112, G03011,
doi:10.1029/2006JG000302.
Griffis, T. J.; Black, T. A.; Morgenstern , K. Ecophysiological controls on the carbon balances
of three southern boreal forests. Agricultural and Forest Meteorology, 2003, 117, 53-
71.
Griffis, T.J.; Baker, J.M.; Zhang, J. Seasonal dynamics and partitioning of isotopic CO2
exchange in a C3/C4 managed ecosystem. Agric. For. Meteor. 2005, 132, 1-19.
Gurney, K. R. et al. TransCom 3 CO2 Inversion Intercomparison: 1. Annual mean control
results and sensitivity to transport and prior flux information. Tellus B 2003, 55,
555– 579.
Gurney, K.R.; Law, R.M.; Denning, A.S.; Rayner, P.J.; Baker, D. et al. Towards robust
regional estimates of CO2 sources and sinks using atmospheric transport models.
Nature 2002, 415, 626-630.
Hall, F. G.; Hilker, T.; Coops, N. C.; Lyapustin, A.; Huemmrich, F.; Middleton, E.; Margolis,
H.; Drolet, G.; Black, T. Multi-angle remote sensing of forest light use efficiency by
observing PRI variation with canopy shadow fraction. Remote Sensing of
Environment. , 2008, 112:3201-3211.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 253

Hall, F.; Sellers, P.; Strebel, D.; Kanemasu, E.; Kelly, R.; Blad, B.; Markham, B.; Wang, J.;
Huemmrich, F. Satellite remote sensing of surface energy and mass balance results:
Results from FIFE. Remote Sensing of Environment, 1991, 35, 187−199.
Harris, P. P.; Huntingford, C.; Cox, P. M.; Gash, J. H.C.; Malhi, Y. Effect of soil moisture on
canopy conductance of Amazonian rainforest. Agric. For. Meteor. 2004, 122, 215-227.
Hedges, J.I. Global biogeochemical cycles: progress and problems. Mar. Chem., 1992. 39, 67–
93.
Helliker, B.R.; Berry, J.A.; Betts, A.K.; Bakwin, P.S.; Davis, K.J.; Denning, A.S.; Ehleringer, J.
R.; Miller, J.B.; Butler, M.P.; Ricciuto D.M. Estimates of net CO2 flux by application
of equilibrium boundary layer concepts to CO2 and water vapor measurements
from a tall tower. J. Geophys. Res. 2004, 109, D20106, doi:10.1029/2004JD004532.
Henderson-Sellers, A.; Yang, Z.-L.; Dickinson, R.E. The Project for Intercomparison of Land-
surface Parameterization Schemes, Bull. Amer. Meteorol. Soc., 1993, 74, 1335.
Hilker, T.; Coops, N.C.; Hall, F.G.; Black, T.A.; Chen, B.; Krishnan, P.; Wulder, M.A.; Sellers,
P.J.; Middleton, E.M.; Huemmrich, K.F. A modeling approach for upscaling gross
ecosystem production to the landscape scale using remote sensing data. J. Geophys
Res.,2008 113, G03006, doi:10.1029/2007JG000666.
Holdridge, L. Determination of world plant formations from simple climatic data, Science
1947,105, 367-368.
Hollenbeck, K.J.; Schmugge, T.J.; Homberger, G.M.; Wang, J.R. Identifying soil hydraulic
heterogeneity by detection of relative change in passive microwave remote sensing
observations. Water Resour. Res. 1996, 32, 139–148.
Hollinger, D. Y.; Goltz, S.M.; Davidson, E. A. Seasonal patterns and environmental control
of carbon dioxide and water vapour exchange in an ecotonal boreal forest. Global
Change Biology, 1999, 5, 891–902.
Hollinger, D. Y.; Richardson, A. D. Uncertainty in eddy covariance measurements and its
application to physiological models. Tree Physiology, 2005, 25, 873-885.
Horst, T.W.; Weil, J.C. Footprint estimation for scalar flux measurements in the atmospheric
surface layer. Boundary-Layer Meteorol., 1992, 59, 279-296.
Huang, M.; Ji, J.; LI, K.; Liu, Y.; Yang, F.; Tao, B. The ecosystem carbon accumulation after
conversion of grasslands to pine plantations in subtropical red soil of South China,
Tellus, 2007, 59B, 439-448.
Huete, A. R.; Liu, H. Q.; Batchily, K.; vanLeeuwen, W. A comparison of vegetation indices
over a global set of TM images for EOS-MODIS, Remote Sensing of Environment,
1997, 59, 440−451.
Huete, A.; Didan, K.; Miura, T.; Rodriguez, E. P.; Gao, X.; Ferreira, L. G. Overview of the
radiometric and biophysical performance of the MODIS vegetation indices, Remote
Sensing of Environment, 2002, 83, 195– 213.
Jackson, T.J. Measuring surface soil moisture using passive microwave remote sensing.
Hydro. Processes 1993, 7, 139–152.
Jain, M.K.; Kothyari, U.C.; Raju K.G. A GIS based distributed rainfall–runoff model. J.
Hydro.2004, 299, 107–135.
Jarvis, P.G. Interpretation of variations in leaf water potential and stomatal conductance
found in canopies in field. Philos Trans R Soc Lond B Biol Sci, 1976, 273, 593-610.
Joos, F.; Prentice, I.C.; Sitch, S.; Meyer, R.; Hooss, G.; Plattner, G.K.; Gerber, S.; Hasselmann,
K. Global Warming Feedbacks on Terrestrial Carbon Uptake under the
254 Remote Sensing – Applications

Intergovernmental Panel on Climate Change (Ipcc) Emission Scenarios, Global

Biogeochem. Cycles, 2001, 15, 891-907.
Ju, J.C.; Roy, D.P. The availability of cloud-free Landsat ETM+ data over the conterminous
United States and globally. Remote Sensing of the Environment, 2007 112, 1196−1211.
Kalma, J. D.; McVicar, T. R.; McCabe, M. F. Estimating Land Surface Evaporation: A Review
of Methods Using Remotely Sensed Surface Temperature Data, Surv. Geophys.,
2008, 29, 421–469.
Kalvelage, T.; Willems, J. Supporting users through integrated retrieval, processing, and
distribution systems at the Land Processes Distributed Active Archive Center. Acta
Astronautica, 2005, 56, 681−687.
Kasischke, E.S.; Melack, J.M.; Dobson, M.C. The use of imaging radars for ecological
applications—a review. Remote Sensing of Environment, 1997, 59, 141–156
Kickert, R.N.; Tonella, G.; Simonov, A.; Krupa, S. Predictive modeling of effects under global
change. Environ. Pollut. 1999, 100, 87-132.
Kim, G.; Barros, A.P. Space-time characterization of soil moisture from passive microwave
remotely sensed imagery and ancillary data. Remote Sens. Environ. 2002, 81, 393–
403.
Kite, G.W.; Pietroniro, A. Remote sensing applications in hydrological modelling. Hydro. Sci.
J. 1996, 41, 563-591.
Kite, G.W.; Droogers, P. Comparing evapotranspiration estimates from satellites,
hydrological models and field data. J. Hydrol. 2000, 229, 3-18.
Kley, D.; Kleinmann, M.; Sandermann, H.; Krupa, S. Photochemical oxidants: state of the
science. Environ. Pollut. 1999, 100, 19-42.
Kljun, N.; Kormann, R.; Rotach, M.W. Comparison of the Langrangian footprint model
LPDM-B with an analytical footprint model. Boundary-Layer Meteorology , 2003, 106,
349-355.
Koster, R.D.; Guo, Z.C.; Dirmeyer, P.A.; Bonan, G.;Chan, E.; Cox, P.; Davies, H.; Gordon,
C.T.; Kanae, S.; Kowalczyk, E., et al. Glace: the Global Land-Atmosphere Coupling
Experiment. Part I: Overview. J. Hydrometeor. 2006, 7, 590-610.
Kurth, W. Morphological models of plant growth: possibilities and ecological relevance.
Ecol. Model. 1994, 75, 299–308
Lai C.T.; Ehleringer, J.; Schauer, A.; Tans , P.P.; Hollinger, D.; Paw, K.T.U.; Munger, J.;
Wofsy, S. Canopy-scale δ13C of photosynthetic and respiratory CO2 fluxes:
observations in forest biomes across the United States. Global Change Biol. 2005, 11,
633-643, doi: 10.1111/j.1365-2486.2005.00931.x.
Lai, C.-T.; Schauer, A.J.; Owensby, C. et al. Regional CO2 fluxes inferred from mixing ratio
measurements: estimates from flask air samples in central Kansas, USA. Tellus B
2006, 58, 523-536.
Law, B. E.; Waring, R. H.; Anthoni, P. M. Measurements of gross and net ecosystem,
productivity and water vapour exchange of a Pinus ponderosa ecosystem, and
evaluation of two generalized models, Global Change Biol., 2000, 6, 155-168.
Law, B.E.; Falge, E; Gu, L. et al. Environmental controls over carbon dioxide and water
vapor exchange of terrestrial vegetation. Agric. For. Meteor. 2002, 113, 97-120.
Lee, X.; Fuentes, J. D.; Staebler, R. M. Long-term observation of the atmospheric exchange of
CO2 with a temperature deciduous forest. Journal of Geophysical Research, 1999, 104,
15975–15984.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 255

Lefsky, M.A.; Cohen, W.B.; Parker, G.G.; Harding, D.J. Lidar remote sensing for ecosystem
studies. BioScience, 2002, 52, 19–30.
Leuning, R. Estimation of scalar source/sink distributions in plant canopies using
Lagrangian dispersion analysis: Corrections for atmospheric stability and
comparison with a multilayer canopy model. Boundary-Layer Meteor. 2000, 96, 293-
314.
Levin, S.A. Complex adaptive systems: exploring the known, unknown and unknowable.
Bull. Amer. Meteor. Soc. 2002, 40, 1-19.
Levis, S.; Foley, J.A.; Pollard D. Potential high-latitude vegetation feedbacks on CO2-induced
climate change, Geophys. Res. Lett, 1999, 26, 747–750.
Levy, P. E.; Grelle, A.; Lindroth, A. et al. Regional-scale CO2 fluxes over central Sweden by a
boundary layer budget method. Agric. For. Meteor. 1999, 98-99, 169-180.
Li, Z.; Yu, G.; Xiao, X. et al. Modeling gross primary production of alpine ecosystems in the
Tibetan Plateau using MODIS images and climate data, Remote Sensing of
Environment, 2007, 107, 510–519.
Liang, X.; Guo, J.; Leung, L.R. Assessment of the effects of spatial resolutions on daily water
flux simulations. J. Hydrol. 2004, 298, 287-310.
Liang, X.; Lettenmaier, D.P.; Wood, E.F.; Burges, S.J. A simple hydrologically based model of
land surface water and energy fluxes for GSMs. J. Geophys. Res., 1994, 99, 14415-
14428.
Lin, J.C.; Gerbig, C.; Wofsy, S.C.; Andrews, A.E.; Daube, B.C.; Grainger, C.A. Stephens, B.B.,
Bakwin, P.S.; Hollinger D.Y. Measuring fluxes of trace gases at regional scales by
Lagrangian observations: Application to the CO2 Budget and Rectification Airborne
(COBRA) study, J. Geophys. Res. 2004, 109, D15304, doi:10.1029/2004JD004754.
Liu, Y., Yu, G., Wen, X., Wang, Y., Song, X, Li, J., Sun, X., Yang, F., Chen, Y., Liu Q.: Seasonal
dynamics of CO2 fluxes from subtropical plantation coniferous ecosystem. Science
in China Series D, 2006, 49 Supp.II, 99-109.
Lloyd, J.; Francey, R.J.; Mollicone, D. et al. Vertical profiles, boundary layer budgets, and
regional flux estimates for CO2 and its 13C/12C ratio and for water vapor above a
forest/bog mosaic in central Siberia. Global Biogeochem. Cycles 2001, 15, 267-284.
Lloyd, J.; Kruijt, B.; Hollinger, D.Y., et al. Vegetation effects on the isotopic composition of
atmospheric CO2 at local and regional scales: theoretical aspects and a comparison
between rain forest in Amazonia and a boreal forest in Siberia. Austr. J. Plant Physi.
1996, 23, 371-399.
Loiselle, S.; Bracchini, L.; Bonechi, C.; Rossi, C. Modelling energy fluxes in remote wetland
ecosystems with the help of remote sensing. Ecol. Model. 2001, 145, 243–261.
Luckman, A.; Baker, J.; Honzak, M.; Lucas, R. Tropical forest biomass density estimation
using JERS-1 SAR: Seasonal variation, confidence limits and application to image
mosaics. Remote Sensing of Environment, 1998, 63, 126–139.
Ma, S.; Baldocchi, D.D.; Xu, L. et al. Interannual variability in carbon dioxide exchage of an
oak/grass sacanna and open grassland in California. Agric. For. Meteor. 2007,
147,157-171
Maas, S.J. Use of remotely sensed information in agricultural crop growth models. Ecol.
Model. 1988, 41, 247–268.
Marquis, M.; Tans, P. Carbon Crucible, Science, 2008, 320, 460-470.
Matthews, E.; Hansen, J. Eds. Land Surface Modeling: A Mini-Workshop, May 1998:
http://www.giss.nasa.gov/meetings/landsurface1998/section3.html
256 Remote Sensing – Applications

Melesse A.M.; Vijay N.; Jon H. Analysis of Energy Fluxes and Land Surface Parameters in
Grassland Ecosystem: Remote Sensing Perspective. Int. J. Remot. Sens. 2008, 29,
3325-3341.
Melesse, A.M.; Weng, Q.; Thenkabail, P.; Senay, G. Remote sensing sensors and applications
in environmental resources mapping and modeling. Sensors 2007, 7, 3209-3241
Meybeck, M. Carbon, nitrogen, and phosphorous transport by world rivers. Am. J. Sci. 1982,
282, 401–450.
Monteith, J. L. Solar radiation and productivity in tropical ecosystems, J. of Applied Ecology,
1972, 9, 747−766.
Monteith, J. L. The photosynthesis and transpiration of crops, Experimental Agriculture, 1966,
2, 1−14.
Montzka, C.; Canty, M.; Kunkel, R.; Menz, G.; Vereecken, H.; Wendland F. Modelling the
water balance of a mesoscale catchment basin using remotely sensed land cover
data. J. Hydro. 2008, 353, 322-334.
Natl. Res. Counc. Earth Observations from Space: The First 50 Years of Scientific Achievements.
2008, Washington, DC: Natl. Acad. Press. 142 pp.
Oge´e, J.; Brunet, Y.; Loustau, D.; Berbigier, P.; Delzon, S. MuSICA, a CO2, water and energy
multi-layer, multi-leaf pine forest model: Evaluation from hourly to yearly time
scales and sensitivity analysis. Global Change Biol. 2003, 9, 697-717.
Papale, D.; Valentini, R. A new assesment of European forests carbon exchanges by eddy
fluxes and artificial neural network spatialization. Global Change Biology, 2003, 9,
525-535.
Parton, W.J.; Ojima, D.S.; Cole, C.V.; Schimel, D.S. A general model for soil organic matter
dynamics: sensitivity to litter chemistry, texture and management. Soil Sci. Soc. Am.
J. 1993, 39, 147 167.
Pataki, D.E.; Ehleringer, J.R.; Flanagan, L.B.; Yakir, D.; Bowling, D.R.; Still, C.J.; Buchmann,
N.; Kaplan, J.O.; Berry, J.A. The application and interpretation of Keeling plots in
terrestrial carbon cycle research. Global Biogeochem.Cycles 2003, 17, 1022,
doi:10.1029/2001GB001850.
Pielke, R.A. Sr. Influence of the spatial distribution of vegetation and soils on the prediction
of cumulus convective rainfall. Rev. Geophys. 2001, 39, 151-177.
Pietroniro, A.; Leconte, R. A review of Canadian remote sensing and hydrology, 1999–2003.
Hydro. Processes 2005, 19, 285–301.
Plummer, S.E. Perspectives on combining ecological process models and remotely sensed
data. Ecol. Model. 2000, 129, 169-186.
Ponton ,S; Flanagan, L.B.; Alstard, K; Johnson, B.G.; Morgenstern, K; Kljun, N; Black, T.A.;
Barr, A.G. Comparison of ecosystem water-use efficiency among Douglas-fir forest,
aspen forest and grassland using eddy covariance and carbon isotope techniques.
Global Change Biol. 2006, 12, 294-310.
Porporato, A.; D'odorico, P.; Laio, F.; Rodriguez-Iturbe, I. Hydrologic Controls on Soil
Carbon and Nitrogen Cycles. I. Modeling Scheme. Adv. Water Resour. 2003, 26, 45-
58.
Prentice, I.C.; Cramer, W.; Harrison, S.P.; Leemans, R.; Monserud, R.A.; Solomon, A.M. A
global biome model based on plant physiology and dominance, soil properties, and
climate. J. Biogeog. 1992, 19, 117-134.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 257

Prentice; I.C.; Farquhar, G.D.; Fasham, M.J.R. et al. In The carbon cycle and atmospheric
carbon dioxide; Houghton, J.T., et al. Eds.; Cambridge University Press: Cambridge,
UK, 2001, pp.183-238.
Prentice, K.; Fung, I.Y. The sensitivity of terrestrial carbon storage to climate change, Nature
1990, 346(6279), 48-51.
Prince, S. D.; Goward, S. N. Global primary production: A remote sensing approach, J.
Biogeography, 1995, 22, 815–835.
Raich, J. W.; Rastetter, E. B.; Melillo, J. M. et al. Potential net primary productivity in South-
America—application of a global-model, Ecological Applications, 1991, 1, 399–429.
Rannik, U.; Kolari, P.; Vesala, T.; Hari, P. Uncertainties in Measurement and Modelling of
Net Ecosystem Exchange of a Forest. Agric. For. Meteor. 2006, 138, 244-257.
Ranson, K. J.; Kovacs, K.; Sun, G.; Kharuk, V. I. Disturbance recognition in the boreal forest
using radar and Landsat-7. Canadian Journal of Remote Sensing, 2003 ,29, 271−285.
Reed, S.; Koren, V.; Smith, M.; Zhang, Z.; Moreda, F.; Seo, D.J. Overall distributed model
intercomparison project results. J. Hydro. 2004, 298, 27–60.
Reichstein, M.; Falge, E.; Baldocchi, D. et al. On the separation of net ecosystem exchange
into assimilation and ecosystem respiration: review and improved algorithm, Glob.
Change Biol. , 2005, 11, 1424–1439.
Reichstein, M., Tenhunen, J. D., Roupsard, O., Ourcival, J. M., Rambal,S., Dore, S., and
Valentini, R.: Ecosystem respiration in two Mediterranean veergreen Holm Oak
forests: drought effects and decomposition dynamics, Funct. Ecol. , 2002, 16, 27–39.
Richardson, A. D.; Hollinger, D.Y.B.; George, G. et al. A multi-site analysis of random error
in tower-based measurements of carbon and energy fluxes, Agricultural Forest Me-
teorology, 2006, 136, 1-18.
Ritchie, J.C.; Rango, A. Remote sensing applications to hydrology: introduction. Hydro. Sci. J.
1996, 41, 429–431.
Rodriguez-Iturbe, I. Ecohydrology: A hydrologic perspective of climate-soil-vegetation
dynamics. Water Resour. Res. 2000, 36, 3–9.
Rodriguez-Iturbe, I.; Porporato, A.; Laio, F.; Ridolfi, L. Plants in Water-Controlled
Ecosystems: Active Role in Hydrologic Processes and Response to Water Stress - I.
Scope and General Outline. Adv. Water Resour. 2001, 24, 695-705.
Rouse, J.W.; Haas, R.H.; Schell, J.A. et al. Monitoring the vernal advancement and
retrogradation (greenwave effect) of natural vegetation. In NASA/GSFC Type III
Final Report. Greenbelt, 1974, MD: NASA
Roy, P.; Junchang, J.; Lewis, P. et al. Multi-temporal MODIS–Landsat data fusion for relative
radiometric normalization, gap filling, and prediction of Landsat data. Remote
Sensing of Environment, 2008, 112, 3112−3130.
Ruimy, A.; Jarvis, P. G.; Baldocchi, D. D.; Saugier, B. CO2 fluxes over plant canopies and
solar radiation: A review, Advances in Ecological Research, 1995, 1-68.
Ruimy, A.; Kergoat, L.; Bondeau, A. Comparing global models of terrestrial net primary
productivity (NPP): Analysis of differences in light absorption and light-use
efficiency, Global Change Biology, 1999, 5, 56−64.
Running, S. W.; Baldocchi, D. D.; Turner, D. P. et al. A global terrestrial monitoring network
integrating tower fluxes, flask sampling, ecosystem modeling and EOS satellite
data, Remote Sensing of Environment, 1999, 70, 108−127.
258 Remote Sensing – Applications

Running, S. W.; Nemani, R. R.; Heinsch, F. A. et al. A continuous satellite-derived measure

of global terrestrial primary productivity: Future science and applications,
Bioscience, 2004, 56, 547−560.
Running, S. W.; Thornton, P. E.; Nemani, R.; Glassy, J. M. Global terrestrial gross and net
primary productivity from the Earth Observing System. In O. E. Sala, R. B. Jackson,
& H. A. Mooney (Eds.), Methods in Ecosystem Science, (pp. 44−57). , 2000, New York:
Springer.
Running, S.W.; Coughlan, J.C. A general model of forest ecosystem processes for regional
applications I. Hydrological balance, canopy gas exchange and primary production
processes. Ecol. Model. 1998, 42, 125-154.
Running, S.W.; Nemani, R.R.; Heinsch, F.A.; Zhao, M.; Reeves, M.; Jolly, M. A continuous
satellite-derived measure of global terrestrial primary productivity: Future science
and applications. Bioscience 2004, 56, 547-560.
Saatchi, S.S.; Moghaddam, M. Estimation of crown and stem water content and biomass of
boreal forests using polarimetric SAR imagery. IEEE Transactions on Geoscience and
Remote Sensing, 2000, 38, 697–709.
Sass, G.Z.; Creed, I.F.: Bird's eye view of forest hydrology: novel approaches using remote
sensing techniques. In D. Levia, D. Carlyle-Moses, & T. Tanaka (eds), Forest
Hydrology and Biogeochemistry: Synthesis of Past Research and Future Directions,(pp.45-
68) 2011, New York: Springer.
Schmid, H. P. Experimental design for flux measurements: matching scales of observations
and fluxes, Agric For Meteorol. , 1997, 87, 179-200.
Schmugge, J.S.; Kustas, W.P.; Ritchie ,J.C.; Jackson, T.J.; Rango, A. Remote sensing in
hydrology. Adv. Water Resour. 2002, 25, 1367–1385.
Schultz, G.A. Remote sensing applications to hydrology: runoff. Hydro. Sci. J. 1996, 41, 453–
475.
Schwalm, C.R.; Ek, A.R. Climate change and site: relevant mechanisms and modeling
techniques. For. Ecol. Manage. 2001, 150, 241-258.
Seaquist, J.W.; Olsson, L.; Ard¨ O, J. A remote sensing-based primary production model for
grassland biomes. Ecol. Model. 2003, 169, 131–155.
Sellers, P.J., et al. Modeling the exchange of energy, water, and carbon between continentals
and the atmosphere. Science 1997, 275, 502-509.
Sellers, P.J.; Mintz, Y.; Sud, Y.C.; Dalcher, A. A Simple Biosphere Model (SiB) for Use Within
General-Circulation Models. J. Atmos. Sci. 1986, 43, 505-531.
Sellers, P.J.; Randall, D.A.; Collatz, G.J. et al. A revised land surface parameterization (SiB2)
for atmospheric GCMs. Part I: model formulation. J. Climat. 1996, 9, 676-05.
Seneviratne, S.I.; Luthi, D.; Litschi, M.; Schar C. Land-atmosphere coupling and climate
change in Europe. Nature 2006, 443, 205–209.
Seth, A.; Giorgi, F.; Dickinson, R.E. Simulating fluxes from heterogeneous land surfaces:
Explicit subgrid method employing the biosphere-atmosphere transfer scheme
(BATS). J. Geophys. Res., 1994, 99(D9), 18,651–18,667.
Shen, S.; Leclerc, M. Y. How large must surface layer inhomogeneities be before they
influence the convective boundary layer structure? A case study, Quart J Roy
Meteorol Soc. , 1995, 121, 1209-1228.
Snyder, P.K.; Delire, C.L.; Foley, J.A. Evaluating the influence of different vegetation biomes
on the global climate. Clim. Dyn. 2004, 23, 279-302.
Coupled Terrestrial Carbon and Water Dynamics in
Terrestrial Ecosystems: Contributions of Remote Sensing 259

Sogachev, A.; Rannik, U.; Vesala, T. Flux footprints over complex terrain covered by
heterogeneous forest, Agric. For. Meteorol. , 2004, 127, 142–158.
Stisen, S.; Jensen, K.H.; Sandholt, I.; Grimes D.I.F. A remote sensing driven distributed
hydrological model of the Senegal River basin. J. Hydro. 2008, 354, 131-148.
Stoy, P. C.; Katul, G.; Siqueira, G. Separating the effects of climate and vegetation on
evapotranspiration along a successional chronosequence in the southeastern US,
Global Change Biol. , 2006, 12, 2115-2135.
Su, Z. Remote sensing of land use and vegetation for mesoscale hydrological studies. Inter. J.
Remot. Sens. 2000, 21, 213–233.
Su, Z., The surface energy balance system (SEBS) for estimation of the turbulent heat fluxes.
Hydro. Earth Sci. 2002, 6, 85–99.
Suits, N.S.; Denning, A.S.; Berry, J.A.; Still, C.J.; Kaduk, J.; Miller, J.B.; Baker I.T. Simulation
of carbon isotope discrimination of the terrestrial biosphere. Global Biogeochem.
Cycles 2005, 19, GB1017, doi:10.1029/2003GB002141.
Tans, P.P.; Fung, I.Y.; Takahashi, T. Observation Constraints on the global atmospheric CO2
budget. Science 1990, 247, 1431-1438.
Tarantola, A. Inverse Problem: Theory Methods for Data Fitting and Parameter Estimation.
Elsevier, 1987, New York.
Tilford, S. Global habitability and Earth remote sensing. Philos. Trans. R. Soc. 1984 A
312(1519):115–18.
Treuhaft, R.N.; Asner, G.P.; Law, B.E.; Van Tuyl, S. Forest leaf area density profiles from the
quantitative fusion of radar and hyperspectral data. Journal of Geophysical Research,
2002, 107, 4568–4580.
Treut, L.; Somerville, H. R.; Cubasch, U.; Ding, Y.; Mauritzen, C.; Mokssit, A.; Peterson T.;
Prather, M. Historical Overview of Climate Change. In Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report
of the Intergovernmental Panel on Climate Change Solomon, S., D. Qin, M. Manning, Z.
Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller Eds. Cambridge
University Press, Cambridge, United Kingdom and New York, NY, USA, 2007; pp.
104-105.
Tucker, C.J.; Fung, I.Y.; Keeling, C.D.; Gammon, R. The relationship of global green leaf
biomass to atmospheric CO2 concentrations. Nature , 1986 319:159–99.
Tucker, C.J. Red and photographic infrared linear combinations for monitoring vegetation.
Remote Sens. Environ. 1979, 8(2):127–50.
Urbanski, S.; Barford, C.; Wofsy, S. Factors controlling CO2 exchange on timescales from
hourly to decadal at Harvard Forest, J. Geophys. Res., 2007, 112, G02020,
doi:10.1029/2006JG000293.
Verseghy, D.L. Class-a Canadian Land Surface Scheme for Gcms.1. Soil Model. Int. J.
Climatol. 1999, 11, 111-133.
Verseghy, D.L.; McFarlane, N.A.; Lazare, M., CLASS: A Canadian land surface scheme for
GCMs: II. Vegetation model and coupled runs. Int. J. Climatol. 1993, 13, 347-370.
Verstraeten, W. W.; Veroustraete, F.; Feyen, J. Assessment of evapotranspiration and soil
moisture content across different scales of observation, Sensors, 2008, 8, 70–117.
Viterbo, P.; Beljaars, A.C.M.; An improved land surface parameterization scheme in the
ECMWF model and its validation. J. Climat. 1995, 8, 2716-2748.
Wang, J.; Price, K.P.; Rich, P.M. Spatial patterns of NDVI in response to precipitation and
temperature in the central Great Plains. Int. J. Remote Sens. 2001, 22, 3827–3844.
260 Remote Sensing – Applications

Wang, S.; Grant, R.F.; Verseghy, D.L.; Black, T.A. Modeling carbon-coupled energy and
water dynamics of boreal aspen forest in a general circulation model land surface
scheme. Int. J. Climatol. 2002, 22, 1249-1265.
Wang, S.; Grant, R.F.; Verseghy, D.L.; Black, T.A. Modeling carbon dynamics of boreal forest
ecosystems using the Canadian land surface scheme. Clim. Chang. 2002, 55, 451-477.
Wang, X.-C.; Chen, R.F.; Gardner, G.B. Sources and transport of dissolved and particulate
organic carbon in the Mississippi River estuary and adjacent coastal waters of the
northern Gulf of Mexico, Marine Chemistry, 2004, 89, 241– 256.
Warren, C.R.; Adams, M.A. Distribution of N. Rubisco and photosynthesis in Pinus pinaster
and acclimation to light. Plant Cell Environ. 2001, 24, 597–609.
Wen, X.; Yu, G.; Sun, X.; Li, Q. et al. Soil moisture effect on the temperature dependence
ofecosystem respiration in a subtropical Pinus plantation of southeastern China,
Agric. For. Meteor. , 2006, 137, 166–175.
Williams, M.S.; Law, B.E.; Anthoni, P.M.;Unsworth, W.H. Use of a simulation model and
ecosystems flux data to examine carbon-water interactions in ponderosa pine. Tree
Physiol. 2001, 21, 287-298.
Wilson, K.B.; Baldocchi, D.D.; Hanson, P.J. Leaf age affects the seasonal pattern
photosynthetic capacity and net ecosystem exchange of carbon in a deciduous
forest. Plant Cell Environmen. 2001, 24, 571–583.
Wilson, K.B.; Baldocchi, D.D.; Hanson, P.J. Spatial and seasonal variability of photosynthetic
parameters and their relationship to leaf nitrogen in a deciduous forest. Tree Physi.
2000, 20, 565–578.
Wu, J. D.; Wang, D.; Bauer, M. E. Image-based atmospheric correction ofQuickBird imagery
of Minnesota cropland. Remote Sensing of Environment, 2005,99, 315−325.
Xiao, X. M.; Zhang, Q. Y.; Hollinger, D.; Aber, J.; Moore, B. Modeling gross primary
production of an evergreen needleleaf forest using MODIS and climate data,
Ecological Applications, 2005, 15, 954−969.
Xiao, X.; Boles, S.; Liu, J. Y.; Zhuang, D. F.; Liu, M. L. Characterization of forest types in
Northeastern China, using multi-temporal SPOT-4 VEGETATION sensor data,
Remote Sensing of Environment, 2002, 82, 335−348.
Xiao, X.; Hollinger D.; Aber, J. D.; Goltz, M.; Davidson, E.; Zhang, Q. and Moore III B.
Satellite-based Modeling of Gross Primary Production in an Evergreen Needle leaf
Forest, Remote Sensing Envir. , 2004, 89, 519 - 534.
Zhang, L. M.; Yu, G. R.; Sun, X. M. et al.; Liu, Y. F.; Xin, D.; Guan D. X.; Yan, J. H. Seasonal
variations of ecosystem apparent quantum yield (α) and maximum photosynthesis
rate (Pmax) of different forest ecosystems in China, Agric. For. Meteor. , 2006, 137,
176–187.
Zhang, Y.; Chen, W.; Cihlar, J. A process-based model for quantifying the impact of climate
change on permafrost thermal regimes. J. Geophys. Res. 2003, 108(D22), 4695,
doi:10.1029/2002JD003354.
Zhu, X.; Chen, J.; Gao, F.; Chen, X.; Masek J. An enhanced spatial and temporal adaptive
reflectance fusion model for complex heterogeneous regions, Remote Sensing of
Environment, 2010, doi:10.1016/j.rse.2010.05.032.
 10

Oceanic Evaporation: Trends and Variability

Long S. Chiu1, Si Gao1 and Chung-Lin Shie2
1Department of Atmospheric, Oceanic and Earth Sciences, George Mason University
2UMBC/JCET, NASA/GSFC

USA

1. Introduction
The global water and energy cycles are strongly coupled as two essential components of
earth system. They play important roles in altering the Earth’s climate.
Oceanic evaporation, or sea surface latent heat flux (LHF) divided by latent heat of
vaporization (Lv), is a key component of global water and energy cycle. In a bulk
aerodynamic formulation, it is determined by the transfer coefficient of evaporation, CE, and
bulk parameters such as surface wind speed (U), surface saturated and near-surface air
specific humidity (Qs and Qa),

LHF   LvC EU (Qs  Qa ) (1)

where sea surface saturated humidity is determined by sea surface temperature (SST) and
salinity, and ρ is density of moist air. The transfer coefficient is dependent on the stability of
the atmosphere and the sea state (Liu et al., 1979; Zeng et al., 1988). Historically, marine
surface observations have provided the basis for estimating these oceanic turbulent fluxes
(e.g. Bunker, 1976; Cayan, 1992; da Silva et al., 1994; Esbensen & Kushnir, 1981; Hastenrath,
1980; Hsiung, 1985; Isemer & Hasse, 1985, 1987; Josey et al., 1998; Oberhuber, 1988; Renfrew
et al., 2002; Weare et al., 1981). The advent of remote sensing techniques offers means to
retrieve a number of surface bulk variables. Microwave radiation interacts directly with
water molecules and hence is effective in providing water vapor information. The sea
surface emissivity is affected by the sea state and foam conditions, which is related to
surface wind. For instance, global microwave measurement of the Special Sensor Microwave
Imager (SSM/I) on board a series of Defense Meteorological Satellite Program (DMSP)
satellites has been used to retrieve near-surface air humidity and winds over the ocean.
At present there are several remote sensing products of global ocean surface latent heat flux.
They include the NASA/Goddard Satellite-based Surface Turbulent Flux (GSSTF) dataset
version 1 (Chou et al., 1997) and version 2 (GSSTF2, Chou et al., 2003), the Japanese Ocean
Flux utilizing Remote Sensing Observations (J-OFURO) dataset (Kubota et al., 2002) and the
Hamburg Ocean Atmosphere Parameters and Fluxes from Satellite (HOAPS) dataset (Grassl
et al., 2000). Chiu et al. (2008) examined “trends” and variations in these global oceanic
evaporation products for the period 1988–2000. They found a long-term increase in global
average LHF that started around 1990 in GSSTF2. They argued that the dominant patterns
may be related to an enhancement of Hadley circulation and El Niño-Southern Oscillation
262 Remote Sensing – Applications

(ENSO), respectively. An updated version of SSM/I version 6 (V6) data released by Remote
Sensing Systems (RSS) in 2006 [as used by Wentz et al. (2007), see http://www.ssmi.com]
that calibrates all SSM/I sensors is available in 2008. Shie et al. (2009) have reprocessed and
forward processed GSSTF2 to version 2b (GSSTF2b, Shie et al. 2010; Shie 2010) using the
SSM/I V6 data (including total precipitable water, brightness temperature, and wind speed
retrieval), covering the period July 1987–December 2008. We provide an assessment of these
data products and examine their “trends” and variability.
The data and methodology are described in Section 2. Section 3 presents the trends of these
products, compares GSSTF2 and GSSTF2b for the pre 2000 periods, assesses the post 2000
performance, and examines the GSSTF2b Set1 and Set2 differences. Summary and
discussion are presented in Section 4.

2. Data and methodology

Earlier version of these flux products have been described elsewhere (Chiu et al., 2008). The
product versions described here represent the most updated versions as of the writing of
this report.

2.1 HOAPS
Detail descriptions of the latest version of HOAPS, (version 3, or HOAPS-3) are given in
Andersson et al. (2010). Bulk variables are derived from SSM/I data except for the SST
which is derived from the Advanced Very High Resolution Radiometer (AVHRR) Oceans
Pathfinder SST product. A neural network algorithm is used to derive U. The Qa is obtained
using the linear relationship of Bentamy et al. (2003). The Qs is computed from the AVHRR
SST using the Magnus formula (Murray, 1967) with a constant salinity correction factor of
0.98. The near-surface air temperature (Ta) is estimated from the SST using the assumptions
of 80% constant relative humidity and a constant surface-air temperature difference of 1 K.
Latent and sensible heat fluxes are calculated using the Coupled Ocean–Atmosphere
Response Experiment (COARE) 2.6a bulk algorithm (Fairall et al., 1996, 2003).
The HOAPS-3 data sets cover the time period from July 1987 to December 2005. HOAPS-G
pentad and monthly data sets with 0.5-degree resolution and HOAPS-C twice daily data set
with 1-degree resolution are available at the website (http://www.hoaps.zmaw.de).

2.2 J-OFURO
The updated version of J-OFURO, (version 2, J-OFURO2) is described in Tomita et al. (2010).
Bulk variables U, Qa and SST (Qs) are determined by multi-satellite and multiple satellite
sensors. U is obtained from a combination of microwave radiometers (SSM/I, AMSR-E and
TMI) and scatterometers (ERS-1, ERS-2 and QuikSCAT). Qa is derived from SSM/I
measurements. SST is taken from the Merged satellite and in-situ data Global Daily SST
(MGDSST, Sakurai et al. 2005) analysis provided by Japanese Meteorological Agency (JMA).
Ta is obtained from NCEP/DOE reanalysis. COARE 3.0 bulk algorithm (Fairall et al., 2003) is
used to estimate LHF and SHF. The J-OFURO2 covers the time period from January 1988 to
December 2006. Daily and monthly means with 1-degree resolution are available at the
website (http://dtsv.scc.u-tokai.ac.jp/j-ofuro).
Oceanic Evaporation: Trends and Variability 263

2.3 GSSTF
The GSSTF2 product has daily and monthly fields with a 1°x1° resolution for July 1987–
December 2000 (Chou et al., 2001), based on the method of Chou et al. (1997) with some
improvements (Chou et al., 2003). The temporal and spatial resolutions of GSSTF2b are the
same as those of GSSTF2, except that GSSTF2b product covers a longer period (July 1987–
December 2008).
GSSTF2b dataset is processed using improved input datasets, namely the recently released
NCEP SST analysis, and a uniform (across satellites) surface wind and microwave
brightness temperature (TB) V6 dataset from the SSM/I produced by RSS. Table 1
summarizes characteristics of input data and parameters for HOAPS3, J-OFURO2, GSSTF2
and GSSTF2b, in that order. As we focus on LHF, only detailed descriptions and discussions
on input parameters of LHF are presented.
A major improvement in the input parameters of GSSTF2b is the use of the newly released
SSM/I V6 product (see discussions in http://www.ssmi.com). The SSM/I V6 product
removes the spurious wind speed trends found in the Wentz/RSS SSM/I V4 wind speed
retrievals. To be consistent, the SSM/I V4 total precipitable water (W) and bottom-layer
precipitable water (WB) used in GSSTF2 are replaced by the corresponding SSM/I V6
products in the production of GSSTF2b. Moreover, the weekly 1° spatial resolution
Optimum Interpolation (OI) SST version 1 (V1) dataset (Reynolds & Smith, 1994) used in
GSSTF2 is replaced by the improved OI SST version 2 (V2) dataset. The OI SST V2 has a
lower satellite bias, a new sea ice algorithm, and an improved OI analysis (Reynolds et al.,
2002) resulting in a modest reduction of the satellite bias and global residual biases of
roughly −0.03°C. The major improvement in the V2 analysis shows up at high latitudes
where local differences between the old and new analysis can exceed 1 °C due to the
application of a new sea ice algorithm. There are two GSSTF2b sets, Set1 and Set2 (Shie et
al., 2010; Shie 2010). Set1 is developed using all available DMSP SSM/I sensor data. In a
preliminary analysis, it was noted that there are large trends associated with LHF which are
mostly attributed to the DMSP F13 and F15 satellites. Set2 was produced by excluding
satellite retrievals that are judged to caused relatively large artificial trends in LHF (mostly
post 2000) from Set1. Consequently Set2 is identical to Set1 before 1997 and shows a smaller
trend than Set1 for the whole period, while Set1 has better spatial coverage (less missing
data). Hilburn & Shie (2011) further found a drift in the Earth incidence angle (EIA)
associated with the SSM/I sensors on the DMSP satellites that introduces artificial trends in
the SSM/I TB data. These artificial trends introduce large changes in the boundary water
(WB), which affects the Qa, and thus the LHF retrievals. An improved version, GSSTF2c,
incorporating the corrected SSM/I brightness temperature, has been produced as of this
writing (Shie et al., 2011). The retrieved WB, Qa and LHF have genuinely improved,
particularly in the trends post 2000 (Shie & Hilburn, 2011). An extensive study involving the
GSSTF2c will be presented in a separate paper.
In this chapter, “trend” is used to indicate results from linear regression analysis and/or
Empirical Mode Decomposition (EMD) for the period of study. Linear regression of the time
series with time is used to detect linear trends and the significance can be estimated from the
slope of the regression. EMD is based on local characteristic time scales of the data and is
therefore applicable for analyzing nonlinear and non-stationary processes (Huang et al.,
1998). It decomposes the time series into a finite and often small number of intrinsic mode
264 Remote Sensing – Applications

Datasets HOAPS3 J-OFURO2 GSSTF2 GSSTF2b

CE (transfer Fairall et al. Fairall et al.
Chou (1993) Chou (1993)
coefficient) (1996, 2003) (2003)
SSM/I V6 TB and SSM/I, AMSR-E,
U (speed) neural network TMI, ERS-1, ERS- Wentz V4 (1997) Wentz V6 (2007)
algorithm 2 and QuikSCAT
CCMP Level-2.5
U (vector) N/A N/A Atlas et al. (1996) (SSM/I, TMI, and
AMSR-E)
W/WB N/A N/A Wentz V4 (1997) Wentz V6 (2007)
NCEP/NCAR NCEP/DOE
MGDSST
Reanalysis (V1) Reanalysis (V2)
SST AVHRR Sakurai et al.
Reynolds & Reynolds et al.
(2005)
Smith (1994) (2002)
Bentamy et al. Schlussel et al. Chou et al. (1995, Chou et al. (1995,
Qa
(2003) (1995) 1997) 1997)
Estimated from
SST with
assumptions of
NCEP/DOE NCEP/NCAR NCEP/DOE
Ta 80% humidity
Reanalysis (V2) Reanalysis (V1) Reanalysis (V2)
and 1 K surface-
air temperature
difference
Spatial resolution 1°x1°, 0.5°x0.5° 1°x1° 1°x1° 1°x1°
Spatial coverage Global Oceans Global Oceans Global Oceans Global Oceans
Twice daily,
Temporal Daily and Daily and Daily and
pentad and
resolution monthly monthly monthly
monthly
Temporal Jul. 1987 – Dec. Jan. 1988 – Dec. Jul. 1987 – Dec. Jul. 1987 – Dec.
coverage 2005 2006 2000 2008
Table 1. Characteristics of input data and parameters for HOAPS3, J-OFURO2, GSSTF2 and
GSSTF2b.

functions (IMFs) of increasing time scales. The existence of a trend is dependent on the
length of the dataset. If the last IMF (one with longest time scale) is monotonically
increasing or decreasing, a trend is indicated. The EMD is a more stringent test for
significance of “trends.”
Non-seasonal variability is examined using Empirical Orthogonal Function (EOF) analysis.
Non-seasonal data are obtained by subtracting the monthly climatology of the study period
from the monthly data. EOF analysis decomposes a spatio-temporal dataset into a series of
orthogonal spatial EOF patterns and the associated time series (also called principal
components). The test proposed by North et al. (1982) is used to judge the EOFs to see if
they are significant and distinct. To examine the significance of each EOF, the logarithm of
Oceanic Evaporation: Trends and Variability 265

the variance explained by the EOF is plotted against its EOF number. The variance
N
explained by the nth EOF is given by n /  i where λn is the nth eigenvalue and N is the
i 1
number of time samples. Linear regression between the logarithms of λn vs n is computed.
EOFs above the regression line are judged to be significant.

3. Results
3.1 Global and zonal average
Figure 1 shows the area weighted global (60°N–60°S) average LHF of HOAPS3, J-OFURO2,
GSSTF2, and Set1 and Set2 of GSSTF2b. The merged OAFLUX product is included for
comparison. Visual inspection of the data products does not show large missing gaps in the
spatial distribution of the products. All satellite products-HOAPS3, J-OFURO2 and all
GSSTF datasets show increases while there is no obvious trend in OAFLUX. All products
have similar global means in the early period 1988–1991. The means of GSSTF2b (both Set1
and Set2) are generally lower than that of GSSTF2 which is the highest among all products.
All products show a dip in 1991 and an increase afterwards. The dip is clearly evident in
HOAPS3. GSSTF2b Set1 and Set2 are identical up to 1998 after which they diverge, but tend
to come close again after 2006.

Fig. 1. Time series of global (60°N–60°S) oceanic average latent heat flux derived from
HOAPS3, J-OFURO2, GSSTF2, GSSTF2b Set1 and Set2 and OAFLUX.

The zonal annual averages of LHF are depicted in Figure 2. The general features of the
zonal means are quite similar among these products: they showed maxima in the
subtropics, minima at the poles and relative minima at the equator. The subtropical
maxima in the southern hemisphere are slighter higher than those in the northern
hemisphere for the same product. While GSSTF2 is the highest among these estimates, the
266 Remote Sensing – Applications

Fig. 2. Zonal annual mean of LHF for oceanic evaporation computed from HOAPS3, J-
OFURO2, GSSTF2, GSSTF2b Set1 and Set2 and OAFLUX.

GSSTF2b Set2 is slightly lower than HOAPS3 but higher than J-OFURO2 and OAFLUX at
their maxima. Poleward of 30°, the GSSTF zonal means are generally higher than the other
products.

3.2 Trend analysis

Linear regression analyses of the time series with time were performed on the global mean
time series. The significance of the slopes of the regression (trend) is tested using a t-test.
The degree of freedom for the significance test takes into account the serial correlation of the
time series (Angell, 1981; Chiu & Newell, 1983). GSSTF2 shows the largest trend. It is
followed by GSSTF2b Set1 while the GSSTF2b Set2 trend is comparable to HOAPS3 and J-
OFURO2 for the period of overlap (1988–2005). OAFLUX exhibits the smallest trend. Table 2
summarizes our results.
To map out the geographic differences, Figure 3 compares the spatial distribution of linear
trends of GSSTF2b Set1 and Set2, HOAPS3, J-OFURO2 and OAFLUX. The linear trends are
calculated for the common time period 1988–2005. While the magnitudes of the trends are
different, the locations of maximum change are similar among HOAPS3, J-OFURO2 and
GSSTF2b Set1 and Set2 - all show increasing trends in the storm tracks in the north Atlantic
and north Pacific, the oceanic dry zones off the Inter-tropical Convergence Zone (ITCZ) in
the western south Pacific and in latitude bands between 30−40°S off the coast of Australia
and in the Indian Ocean. OAFLUX shows increasing trends in the storm tracks in both the
North Atlantic and North Pacific, and in the eastern coastal regions off South America and
Australia. There are large areas showing a decrease, notably in the south Indian Ocean,
tropical eastern North Pacific and in North Atlantic.
Oceanic Evaporation: Trends and Variability 267

GSSTF2b GSSTF2b
Period GSSTF2 HOAPS3 J-OFURO2 OAFLUX
Set1 Set2
1988–2000 10.44 9.88 5.98 7.75 6.51 3.71
1988–2005 N/A 11.69 6.34 7.35 7.62 2.41
1988–2008 N/A 10.45 7.08 N/A N/A 1.50
Table 2. Linear trends of LHF products (in W m–2 decade–1) for the different periods. All
values are significant at 99% level.

Fig. 3. Linear trends of LHF (W m−2 decade−1) over the period 1988−2005. Contours give the
trends above 95% confidence level.

3.3 EMD analysis of global average LHF

We performed EMD analyses for both GSSTF2 and GSSTF2b for the periods 1988–2000
and 1988−2008, respectively. Figure 4 shows the last IMFs from EMD analysis of GSSTF2b
global average (60°N–60°S) LHF for the period 1988–2008, along with those of HOAPS3,
J-OFURO2 and OAFLUX. EMD analysis of GSSTF2 global average LHF for the period
1988–2000 (Chiu et al., 2008; Xing, 2006) showed an increase starting around 1990 before
the Mt Pinatubo eruption event in 1991. The last IMF of GSSTF2b Set2 and HOAPS3 show
a dip around 1990 while J-OFURO2 and GSSTF2b Set1 show monotonic increases. From
EMD analysis, a “trend” cannot be ascertained for the whole period for GSSTF2b Set2 and
HOAPS3.
268 Remote Sensing – Applications

Fig. 4. The last IMFs (W m–2) from EMD analysis of global (60°N–60°S) average LHF for
GSSTF2b Set1 (black), GSSTF2b Set2 (red), and OAFLUX (blue) for the period 1988–2008,
HOAPS3 (purple)for the period 1988–2005 and J-OFURO2 (green) for the period 1988–2006.

3.4 GSSTF2b Set1 and Set2

In this subsection, we analyze the difference between GSSTF2b Set1 and Set2. The linear
trends of GSSTF2b LHF for the entire period 1988–2008 are depicted in Figure 5. The
magnitudes of the trends are reduced when compared to GSSTF2 (see Chiu et al., 2008) and
Set1 shows larger trends than Set2, as anticipated.
Trends of the zonal averages are shown in Figure 6. Both Set1 and Set2 show similar trend
patterns while the magnitude in Set2 is much reduced. Maximum increasing trends are located
in the subtropics, with maxima at 30°S and 35°N, poleward of the latitude of maximum zonal
evaporation, while the trend at the thermal equator (~5°N) shows a minimum.

Fig. 5. Linear trends of GSSTF2b Set1 and Set2 LHF (1988–2008). Unit in W m−2 decade−1.
Contours give the trends above 95% confidence level.
Oceanic Evaporation: Trends and Variability 269

Fig. 6. Linear trends of zonal mean GSSTF2b Set1 and Set2 LHF.

Figure 7 shows the spatial distribution of trends in U, DQ and Qa in Set 1 and Set2. It can be
seen that the trend pattern in U is almost identical in both Set1 and Set2. Major difference is
found in the DQ field, the magnitudes of the trends in Set1 are larger than that in Set2. In the
equatorial eastern Pacific, in portion of the South China Sea, and the Bay of Bengal, Set2
actually shows a decreasing but non-significant trend. For the Qa trend, Set1 shows large
areas of decreases. While the patterns are similar for Set2, the magnitudes are much
reduced, and in some areas, such as the eastern North Pacific and the bulk of the North
Atlantic, the decreasing trends actually reverse to increasing trends.
LHF is a product of the surface wind (U) and the humidity difference (DQ=Qs–Qa) if we
assume the variation in CE is small (Equation 1). Judging from the change pattern of U and
DQ they are essentially decoupled. Equation (1) can be integrated globally to get

 LHF U  DQ
  (2)
LHF U DQ

 DQ   Qs   Qa (3)

where δx represent the change in the quantity x, δx/x represent fractional changes in x, and
the over-bar x represents global average of x. For GSSTF2 (1988–2000), the terms in
equation (2) is approximately 17%, 6%, and 11%, in that order (Xing, 2006). Most of the
increase in DQ was attributed to increase in Qs and decrease in Qa .
Figure 8 shows the time series of global average U, DQ and Qa for GSSTF2b Set1 and Set2,
respectively. It clearly indicates the divergence of Set1 and Set2 in late 1997. The large
difference in LHF between Set1 and Set2 is mostly attributed to DQ, which is due to the
higher Qa in Set2. The difference in U between the datasets is small. There is a large decrease
in Qa at the end of 2008.
270 Remote Sensing – Applications

Fig. 7. Linear trends of surface wind speed (U, upper panel), surface humidity difference
(DQ=Qs–Qa, middle panel), and surface air humidity (Qa, lower panel) for GSSTF2b Set1
(left column) and Set2 (right column). Units are in m s−1 decade−1 for U and in g kg–1 decade–
1 for DQ and Qa. Contours give the trends above 95% confidence level.

Table 3 shows the changes in LHF in Set1 and Set2 of GSSTF2b (1988−2008) and the
associated changes in U, DQ and the changes in Qs and Qa. The change in DQ contribute
most to the change in LHF for both Set1 and Set2, while the changes in DQ is due both to an
increase in Qs and decrease in Qa. The difference in the change of LHF between Set1 (23.1%)
and Set2 (15.5%) is mostly attributed to DQ (20.0% vs. 12.3%) and changes in Qa (–0.51 g kg–1
vs. –0.24 g kg–1). It is clear that the impact of DMSP F13 is the introduction of a much lower
Qa, thus affecting DQ and ultimately LHF.

 LHF U  DQ U  DQ  DQ  Qs  Qa
 
LHF U DQ U DQ (g kg–1) (g kg–1) (g kg–1)
Set1 23.1% 3.1% 20.0% 0.73 0.22 –0.51
Set2 15.5% 3.1% 12.3% 0.46 0.22 –0.24
Table 3. Summary of changes in LHF, U and DQ and Qs and Qa for GSSTF2b Set1 and Set2.
Oceanic Evaporation: Trends and Variability 271

Fig. 8. Time series of global average surface humidity difference (DQ), surface air humidity
(Qa) and wind speed (U) for GSSTF2b Set1 and Set2.

3.5 EOF analyses and teleconnections

Empirical Orthogonal Function (EOF) analyses are performed on LHF of both GSSTF2b Set1
and Set2 for the period 1988–2008. Monthly means for the entire period are first removed
from the data to form the non-seasonal dataset. The first, second and third EOF of Set1

Fig. 9. Spatial patterns of the first (top) and second (bottom) EOF of GSSTF2b LHF for the
period 1988–2008. The variances explained are 8.6% (11.5%) and 4.3% (4.3%) for EOF1 and
EOF2 for Set2 (Set1), respectively.
272 Remote Sensing – Applications

explains 11.5%, 4.3%, and 3.4% of the total variance, and the corresponding variance
explained for Set2 are 8.6%, 4.3% and 3.4%. Figure 9 shows the spatial patterns of the first
(EOF1) and second (EOF2) EOF for Set1 and Set2, respectively. Their associated time series,
accompanied by a rescaled Southern Oscillation Index (SOI), are presented in Figure 10. The
general patterns of EOF1 of Set1 and Set2 are very similar, with large weights in subtropical
Indian Ocean, the dry zone in the eastern tropical South Pacific and South Atlantic. They
also bear striking resemblance to the first EOF pattern computed from GSSTF2 which shows

Fig. 10. Time series associated with EOF of non-seasonal GSSTF2b Set2 LHF for the period
1988–2008 and the SOI for the same period. The SOI is divided by 10 and the time series of
the first EOF is shifted by –0.4 units for clarity. The lower panel shows the same for Set1.
Oceanic Evaporation: Trends and Variability 273

weights of the same sign everywhere (Chiu et al., 2008). However, in GSSTF2b, there are
regions of opposite sign in the South China Sea area in both Set1 and Set2, with higher
weights in Set2 in the centers of highs. The EOF2 patterns for Set1 and Set2 are almost
identical. Similarity is also noted for the EOF2 pattern derived from GSSTF2 (Chiu et al.,
2008).
The time series of EOF1 also show a slight increasing trend, suggesting an increase of LHF
over most of the global ocean. The associated time series of EOF2 have a significant
correlation of 0.73 (0.72 for Set1) with SOI, reaffirming that EOF2 - ENSO events association.
The third EOF pattern (not shown) is characterized by a negative - positive - negative (– + –)
zonal changes centering around 10−20°N, 20−30°N and north of 40°N in the north Pacific
and a (+ – +) centering at 20°N, 30°N and 50°N in the North Atlantic. The pattern
correlations for EOF1, EOF2 and EOF3 between Set1 and Set2 are 0.08, 0.49, and 0.39, and
the corresponding temporal correlation for the associated time series are 0.92, 0.97, and 0.83,
in that order.
EOF3 is reminiscent of the North Atlantic Oscillation (NAO) and the North Pacific
Oscillation (NPO) patterns (Walker & Bliss, 1932; Wallace & Gultzer, 1981, Hurrell et al.,
2003). These teleconnection patterns are further discussed in terms of an atmospheric
annular mode, or Arctic Oscillation (AO), showing the opposition between subtropical highs
and the polar lows (Wallace, 2000; Deser, 2000; Aubaum et al., 2001). We compute the
correlations between an AO index with the time series of EOF3 and found no significant
correlation with Set2, however a correlation of 0.32 is found for EOF3 Set1, significant at 95%
level.

4. Summary and discussion

Four satellite based sea surface latent heat flux (LHF) products, HOSAPS3, J-OFURO2,
GSSTF2, and GSSTF2b (Set1 and Set2) and a merged analysis OAFLUX are compared.
Linear trend analysis of all satellite based products show large increasing trend, with
GSSTF2 the largest, followed by GSSTF2b Set1, J-OFURO2, HOAPS3, and GSSTF2b Set2.
OAFLUX exhibits the lowest linear increasing trend. Most of the satellite products used
SSM/I as input. Small drifts in the SSM/I brightness temperature (TB) associated with
changes in Earth incidence angle (EIA) was noted in most of the SSM/I data (Hilburn &
Shie, 2011; Shie & Hilburn, 2011). Because of the sensitivity of the boundary layer water
(WB) to the TB, these small drifts can introduce artificial trends in bulk quantities such as Qa.
A second data set, GSSTF2b Set2, which excludes satellite retrievals that were judged to
introduce these biases, was introduced. The most-excluded satellite data are SSM/I onboard
the DMSP F13 and F15 satellites. The new set, GSSTF2b Set2, was found to have a much
reduced increasing trend, the magnitude of which is comparable to HOAPS3 and J-OFURO2
for the period of overlap. To account for the drift in the EIA, a new version of GSSTF,
GSSTF2c, that takes account of the correction in EIA, has been completed as of this writing,
and will be officially released to the public via NASA/GES DISC by the end of October 2011
(Shie et al., 2011). These trend issues will be revisited after its release.
Empirical Mode Decomposition (EMD) analyses, which are designed for examining non-
stationary non-homogeneous time series, are performed on the global LHF. The last IMF of
GSSTF2b Set1 shows a monotonic increase indicating the existence of a trend in this period.
274 Remote Sensing – Applications

The corresponding IMF of the other data products do not show monotonic increases, hence
trends cannot be ascertained.
To examine the attribution of the increase in GSSTF2b Set1 and Set2, the linear trends in
both the surface wind (U) and surface humidity difference (DQ) are computed. There is no
significant difference between the trend patterns in the wind field. However, large
difference in the DQ trend is noted. The DQ trend difference is attributed to a reduction in
the negative trend in surface air humidity (Qa) in Set2.
A major difference between GSSTF2 and GSSTF2b is the use of RSS SSM/I V4 for GSSTF2
and RSS SSM/I V6 for GSSTF2b. The changes in LHF, U, and DQ are 16%, 6%, and 11% for
GSSTF2 (Xing, 2006). The corresponding changes are 23%, 3%, and 20% for GSSTF2b Set1
and 16%, 3%, and 12% for Set2 (Table 3). The use of RSS SSM/I V6 products reduces the
wind trend from 6% to 3% but increases the DQ trend from 11% to 20% for Set1. The
exclusion of F13 and F15 data reduces the LHF trend to 16%, mostly due to a reduction in
the DQ trend to 12% (from 20%) with U changes remain at 3%.
Interannual variability is examined using EOF analyses. The first three significant non-
seasonal EOF patterns are similar, and they explaining 10.5%, 4.3% and 3.4% for Set1 and
8.6%, 4.3% and 3.4% for Set2, respectively. The first EOF pattern of GSSTF2 for 1998–2000,
with opposite changes between the equatorial eastern Pacific and the subtropics in the
Pacific and Indian ocean, may be indicative of an enhance Hadley circulation (Chiu & Xing,
2004; Chiu et al., 2008). Observations also indicate large decadal variability in the Hadley
Circulation (Wielicki et al., 2002; Cess & Udelhofen, 2003; Chen et al., 2002; Mitas and
Clements, 2005).
This seesaw pattern is much reduced in the EOF1 pattern in both GSSTF2b Set1 and Set2 of
1998–2008, which may indicate a reduction, change of phase, or mixing of the signal with
the trend in GSSTF2b. The contribution to the total variance is smaller for Set2, which
excluded DMSP datasets that contains large long-term trends introduced by drifts in the
Earth incidence angle in the SSM/I sensors. The difference in the fraction of variance
explained in Set1 and Set2 is attributed to the artificial trend in F13 and F15 and the EIA
drift effect.
Examination of the trends of the zonal means show that the latitude of maximum increase,
situated in the subtropics, is found poleward of the LHF maximum in the tropic. This
pattern is consistent with the expansion of the Hadley Circulation associated with global
warming as predicted in climate models (Lu et al., 2007).
The EOF2 patterns of Set1 and Set2 are almost identical, both contributed to 4.3% of the
variance of the dataset. The association with the El Nino/Southern Oscillation phenomena is
corroborated by a high correlation between their time series and an index of the Southern
Oscillation (SOI). The patterns for EOF3 and their associated time series are also similar,
indicating that both GSSTF2b Set1 and Set2 are useful for examining interannual variability.

5. Acknowledgment
This study is supported by the MEaSUREs Program of NASA Science Mission Directorate-
Earth Science Division. The authors are especially grateful to their program manager M.
Maiden and program scientist J. Entin for their valuable supports of this research.
Oceanic Evaporation: Trends and Variability 275

6. References
Andersson, A.; Fennig, K.; Klepp, C.; Bakan, S.; Graßl, H. & Schulz, J. (2010). The Hamburg
Ocean Atmosphere Parameters and Fluxes from Satellite Data—HOAPS-3. Earth
System Science Data, Vol. 2, No. 1, pp. 215-234, ISSN 1866-3508
Angell, J. K. (1981). Comparison of Variations in Atmospheric Quantities with Sea Surface
Temperature Variations in the Equatorial Eastern Pacific. Monthly Weather Review,
Vol. 109, No. 4 , pp. 230-243, ISSN 0027-0644
Aubaum, M. H. P.; Hoskins, B. J. & Stephenson, D. B. (2001). Arctic Oscillation or North
Atlantic Oscillation. Journal of Climate, Vol. 14, No. 16, pp. 3495-3507, ISSN 0894-
8755
Atlas, R.; Hoffman, R. N.; Bloom, S. C.; Jusem, J. C. & Ardizzone, J. (1996). A multiyear
global surface wind velocity dataset using SSM/I wind observations. Bulletin of the
American Meteorological Society, Vol. 77, No. 5, pp. 869-882, ISSN 0003-0007
Bentamy, A.; Katsaros, K. B.; Mestas-Nuñez, A. M; Drennan, W. M.; Forder, E. B. & Roquet,
H. (2003). Satellite estimates of wind speed and latent heat flux over the global
oceans. Journal of Climate, Vol. 16, No. 4, pp. 637-656, ISSN 0894-8755
Bunker, A. F. (1976). Computations of surface energy flux and annual air-sea interaction
cycle of the north Atlantic Ocean, Monthly Weather Review, Vol. 104, No. 9, pp. 1122-
1140, ISSN 0027-0644
Cayan, D. R. (1992). Latent and sensible heat flux anomalies over the Northern Oceans: The
connection to monthly atmospheric circulation. Journal of Climate, Vol. 5, No. 4, pp.
354-369, ISSN 0894-8755
Cess, R. D. & Udelhofen, P. M. (2003). Climate change during 1985–1999: Cloud interactions
determined from satellite measurements. Geophysical Research Letters, Vol. 30, No. 1,
1019, ISSN 0094-8276
Chen, J., Carlson B. E., & Del Genio A. D. (2002). Evidence for strengthening of the tropical
general circulation in the 1990s. Science, Vol. 295, No. 5556, pp. 838-841, ISSN 0036-
8075
Chiu, L. S. & Xing, Y. (2004). Modes of interannual variability of oceanic evaporation
observed from GSSTF2. Gayana: International Journal of Biodiversity, Oceanology and
Conservation, Vol. 68, No. 2, pp. 115-120. ISSN 0717-652X
Chiu, L. S. & Newell, R. E. (1983). Variations of zonal mean sea surface temperature and
large-scale air-sea interaction. Quarterly Journal of the Royal Meteorological Society,
Vol. 109, No. 459, pp. 153-168, ISSN 0035-9009
Chiu, L. S.; Chokngamwong, R.; Xing, Y.; Yang, R. & Shie, C.-L. (2008). Trends and
Variations of Global Oceanic Evaporation Datasets from Remote Sensing. Acta
Oceanologica Sinica, Vol. 27, No. 3, pp. 1-12, ISSN 0253-505X
Chou, S.-H. (1993). A comparison of airborne eddy correlation and bulk aerodynamic
methods for ocean-air turbulent fluxes during cold-air outbreaks. Boundary-Layer
Meteorology, Vol. 64, No. 1-2, pp. 75-100, ISSN 0006-8314
Chou, S.-H.; Shie, C.-L.; Atlas R. M. & Ardizzone, J. (1997). Air-sea fluxes retrieved from
special sensor microwave imager data. Journal of Geophysical Research, Vol. 102, No.
C6, pp. 12706-12726, ISSN 0148-0227
Chou, S.-H., Atlas R. M., Shie C.-L. & Ardizzone J. (1995). Estimates of surface humidity and
latent heat fluxes over oceans from SSM/I data. Monthly Weather Review, Vol. 123,
No. 8, pp. 2405-2425, ISSN 0027-0644
276 Remote Sensing – Applications

Chou, S.-H.; Nelkin, E.; Ardizzone, J.; Atlas, R. & Shie, C.-L. (2001). The Goddard Satellite-
Based Surface Turbulent Fluxes Dataset Version 2 (GSSTF2.0) [global (grid of 1° x
1°) daily air-sea surface fluxes from July 1987 to December 2000]. Available from
<http://disc.gsfc.nasa.gov/precipitation/gsstf2.0.shtml>
Chou, S.-H.; Nelkin, E.; Ardizzone, J.; Atlas, R. M. & Shie, C.-L. (2003). Surface turbulent
heat and momentum fluxes over global oceans based on the Goddard satellite
retrieval, version 2 (GSSTF2). Journal of Climate, Vol. 16, No. 20, pp. 3256-3273, ISSN
0894-8755
da Silva, A. M.; Young, C. C. & Levitus, S.(1994). Atlas of Surface Marine Data. Vol. 3:
Anomalies of Heat and Momentum Fluxes. NOAA/NESDIS, Washington, D.C., USA
Deser, C. (2000). On the teleconnectivity of the “Arctic oscillation”. Geophysical Research
Letters, Vol. 27, No. 6, pp. 779-782, ISSN 0094-8276
Esbensen, S. K. & Kushnir,V. (1981). The heat budget of the global oceans: An atlas based on
estimates from marine surface observations. Climatic Research Institute, Report No. 29,
Oregon State University, 27 pp
Fairall, C. W.; Bradley, E. F.; Rogers, D. P.; Edson, J. B. & Young, G. S. (1996). Bulk
parameterization of air-sea fluxes for Tropical Ocean Global Atmosphere Coupled
Ocean-Atmosphere Response Experiment. Journal of Geophysical Research, Vol. 101,
No. C2, pp. 3747-3764, ISSN 0148-0227
Fairall, C. W.; Bradley, E. F.; Hare, J. E.; Grachev, A. A. & Edson, J. B. (2003). Bulk
parameterization of air-sea fluxes: Updates and verification for the COARE
algorithm. Journal of Climate, Vol. 16, No. 4, pp. 571-591 , ISSN 0894-8755
Grassl, H.; Jost, V.; Kumar, R.; Schulz, J.; Bauer P. & Schluessel, P. (2000). The Hamburg
Ocean-Atmosphere Parameters and Fluxes from Satellite Data (HOAPS): A climatological
atlas of satellite-derived air-sea interaction parameters over oceans. Max Planck Institute
for Meteorology, Report No. 312, ISSN 0937-1060, Hamburg, Germany
Hastenrath, S. (1980). Heat budget of tropical ocean and atmosphere. Journal of Physical
Oceanography, Vol. 10, No. 2, pp. 159-170, ISSN 0022-3670
Hilburn, K. A. & Shie, C.-L. (2011). Decadal trends and variability in Special Sensor
Microwave Imager (SSM/I) brightness temperatures and Earth incidence angle.
Report No. 092811, Remote Sensing Systems, 53 pp
Huang, N. E.; Shen, Z.; Long, S. R.; Wu, M. C.; Shih, H. H.; Zheng, Q.; Yen, N.-C.; Tung, C.
C. & Liu, H. H. (1998). The empirical mode decomposition and the Hilbert
spectrum for nonlinear and non-stationary time series analysis. Proceedings of the
Royal Society of London. Series A, Vol. 454, No. 1971, pp. 903-995, ISSN 1364-5021
Hsiung, J. (1985). Estimates of global oceanic meridional heat transport. Journal of Physical
Oceanography, Vol. 15, No. 11, pp. 1405-1413, ISSN 0022-3670
Hurrell, J. W.; Kushnir, Y.; Ottersen, G. & Visbeck, M. (Eds.). (2003). The North Atlantic
Oscillation-Climatic Significance and Environmental Impact, Geophysical Monograph
Vol. 134, American Geophysical Union, ISBN 0-875-90994-9, Washington D.C., USA
Isemer, H.-J. & Hasse, L. (1985). The Bunker Climate Atlas of the North Atlantic Ocean: 1.
Observations. Springer-Verlag, ISBN 0-387-15568-6, New York, USA
Isemer, H.-J. & Hasse, L. (1987). The Bunker climate atlas of the North Atlantic Ocean: 2. Air-sea
interactions. Springer-Verlag, ISBN 0-387-17594-6, New York, USA
Josey, S. A.; Kent,E. C. & Taylor,P. K. (1998). The Southampton Oceanography Centre (SOC)
Ocean-Atmosphere Heat, Momentum and Freshwater Flux Atlas. Southampton
Oceanography Centre, Report No.6, Southampton, United Kingdom
Oceanic Evaporation: Trends and Variability 277

Kubota, M.; Ichikawa, K.; Iwasaka, N.; Kizu, S.; Konda, M. & Kutsuwada, K. (2002).
Japanese Ocean Flux Data Sets with Use of Remote Sensing Observations (J-
OFURO). Journal of Oceanography, Vol. 58, No. 1, pp. 213-225, ISSN 0916-8370
Liu, W. T.; Katsaros, K. B. & Businger,J. A. (1979). Bulk Parameterizations of Air-Sea
Exchanges of Heat and Water Vapor Including Molecular Constraints at the
Interface. Journal of Atmospheric Science, Vol. 36, No. 9, pp. 1722-1735, ISSN 1520-
0469
Lu, J.; Vecchi, G. & Reichler, T. (2007). The expansion of the Hadley Cell under global
warming. Geophysical Research Letters, Vol. 34, L06805, ISSN 0094-8276
Murray, F. W. (1967). On the computation of saturation vapor pressure. Journal of Applied
Meteorology, Vol. 6, No. 1, pp. 203-204, ISSN 0021-8952
Mitas, C. M. & Clement, A. (2005). Has the Hadley cell been strengthening in recent
decades?. Geophysical Research Letters, Vol. 32, L03809, ISSN 0094-8276
North, G. R.; Bell, T. L.; Cahalan, R. F. & Moeng, F. J. (1982). Sampling Errors in the
Estimation of Empirical Orthogonal Functions. Monthly Weather Review, Vol. 110,
No. 7, pp. 699-706, ISSN 0027-0644
Oberhuber, J. M. (1988). An Atlas Based on the 'COADS' Data Set: The Budgets of Heat,
Buoyancy and Turbulent Kinetic Energy at the Surface of the Global Ocean. Max-Planck-
Institut für Meteorologie, Report No. 15, Hamburg, Germany
Renfrew, I. A.; Moore, G. W. K.; Guest, P. S. & Bumke, K. (2002). A comparison of surface
layer and surface turbulent flux observations over the Labrador Sea with ECMWF
analyses and NCEP reanalyses. Journal of Physical Oceanography, Vol. 32, No. 2, pp.
383-400, ISSN 0022-3670
Reynolds, R. W. & Smith, T. S. (1994). Improved global sea surface temperature analyses
using optimum interpolation. Journal of Climate, Vol. 7, No. 6, pp. 929-948, ISSN
0894-8755
Reynolds, W. R.; Rayner, N. A.; Smoth, T. M.; Stokes, D. C. & Wang, W. (2002). An improved
in situ and satellite SST analysis for climate. Journal of Climate, Vol. 15, No. 13, pp.
1609-1625, ISSN 0894-8755
Sakurai, T.; Kurihara, Y. & Kuragano, T. (2005). Merged satellite and in-situ data global daily
SST. Procceeding of International Geoscience and Remote Sensing Symposium, pp. 2606-
2608, ISBN 0-7803-9050-4, Seoul, South Korea, November 14, 2005
Schlussel, P.; Schanz, L. & Englisch, G. (1995). Retrieval of latent-heat flux and longwave
irradiance at the sea-surface from SSM/I and AVHRR measurements. Advances in
Space Research, Vol. 16, No. 10, pp. 107-116, ISSN 0273-1177
Shie, C.-L.; Chiu, L. S.; Adler, R.; Nelkin, E.; Lin, I-I; Xie, P., Wang, F-C; Chokngamwong, R.;
Olson, W. & Chu, A. D. (2009). A note on reviving the Goddard Satellite-based
Surface Turbulent Fluxes (GSSTF) dataset. Advances in Atmospheric Sciences, Vol. 26,
No. 6, pp. 1071-1080, ISSN 0256-1530
Shie, C.-L.; Chiu, L. S.; Adler, R.; Lin, I-I; Nelkin, E. & Ardizzone, J. (2010). The Goddard
Satellite-Based Surface Turbulent Fluxes Dataset --- Version 2b (GSSTF2b), In:
NASA Goddard Earth Sciences (GES) Data and Information Services Center (DISC),
October 2010, Available from
<ftp://aurapar1u.ecs.nasa.gov/data/s4pa/GSSTF/> or
<http://disc.sci.gsfc.nasa.gov/daac-
bin/DataHoldingsMEASURES.pl?PROGRAM_List=ChungLinShie>
278 Remote Sensing – Applications

Shie, C.-L. (2010). Science background for the reprocessing and Goddard Satellite-based
Surface Turbulent Fluxes (GSSTF2b) Data Set for Global Water and Energy Cycle
Research, In: NASA GES DISC, 18 pp, October 12, 2010, Available from
< http://disc.sci.gsfc.nasa.gov/measures/documentation/Science-of-the-data.pdf>
Shie, C.-L. & Hilburn, K. (2011). A satellite-based global ocean surface turbulent fluxes
dataset and the impact of the associated SSM/I brightness temperature, Proceedings
of The 2011 EUMETSAT Meteorological Satellite Conference, Oslo, Norway, September
5-9, 2011
Shie, C.-L.; Hilburn, K. A.; Chiu, L. S.; Adler, R.; Lin, I-I; Nelkin, E. & Ardizzone, J. (2011).
The Goddard Satellite-Based Surface Turbulent Fluxes Dataset --- Version 2c
(GSSTF2c). In: NASA GES DISC, by the end of October 2011.
Tomita, H.; Kubota, M.; Cronin, M. F.; Iwasaki, S.; Konda, M. & Ichikawa, H. (2010). An
assessment of surface heat fluxes from J-OFURO2 at the KEO and JKEO sites.
Journal of Geophysical Reserach, Vol. 115, C03018, ISSN 0148-0227
Walker, G. T. & Bliss, E. W. (1932). World weather V. Memoirs of the Royal Meteorological
Society, Vol. 4, No. 36, pp. 53-84, LCCN 7021-1603
Wallace, J. M. (2000). North Atlantic Oscillation/annular mode: Two paradigms—one
phenomenon. Quarterly Journal of the Royal Meteorological Society, Vol. 126, No. 564,
pp. 791-805, ISSN 0035-9009
Wallace, M. & Gutzler, D. S. (1981). Teleconnections in the geopotential height field during
the Northern Hemisphere winter. Monthly Weather Review, Vol. 109, No. 4, pp. 784-
812, ISSN 0027-0644
Weare, B. C.; Strub, P. T. & Samuel, M. D. (1981). Annual Mean Surface Heat Fluxes in the
Tropical Pacific Ocean. Journal of Physical Oceanography, Vol. 11, No. 5, pp. 705-717,
ISSN 0022-3670
Wentz, F. J. (1997). A well calibrated ocean algorithm for Special Senor Microwave/Imager.
Journal of Geophysical Research, Vol. 102, No. C4, pp. 8703-8718, ISSN 0148-0227
Wentz, F. J.; Ricciardulli, L.; Hilburn K. A. & Mears C. A. (2007). How Much More Rain Will
Global Warming Bring? Science, Vol. 317, No. 5835, pp. 233-235, Supporting
material, ISSN 0036-8075
Wielicki B.; Wong, T.; Allen, R. P.; Slingo, A.; Kiehl, J. T.; Soden, B. J.; Gordon, C. T.; Miller,
A. J.; Yang, S.-K.; Randall, D. A.; Robertson, F.; Susskind, J. & Jacobowitz, H. (2002).
Evidence for large decadal variability in the tropical mean radiative energy budget.
Science, Vol. 295, No. 5556, pp. 841-844, ISSN 0036-8075
Xing, Y. (2006). Recent changes in oceanic latent heat flux from remote sensing. Ph.D. dissertation,
George Mason University, 119 pp
Zeng, X.; Zhao, M. & Dickinson, R. E., (1998). Intercomparison of bulk aerodynamic
algorithms for the computation of sea surface fluxes using TOGA COARE and
TAO data. Journal of Climate, Vol. 11, No. 10, pp. 2628-2644, ISSN 0894-8755
 11

Multi-Wavelength and Multi-Direction Remote

Sensing of Atmospheric Aerosols and Clouds
Hiroaki Kuze
Centre for Environmental Remote Sensing (CEReS), Chiba University
Japan

1. Introduction
Aerosols are liquid and solid particles floating in the atmosphere. Aerosol particles are
originated from both natural and anthropogenic origins (Seinfeld & Pandis, 1998). In regard
to the radiation balance of the Earth’s atmosphere, aerosols reflect solar radiation back to
space (direct effect), thus reducing the influence of greenhouse gases, though some type of
aerosol causes opposite effects due to absorption of radiation. At the same time, aerosol
particles work as nuclei for cloud condensation (indirect effect). Knowledge on these
radiative effects of aerosol and cloud, however, is still insufficient so that uncertainties
remain in the prediction of future global warming trends (IPCC, 2007). In this respect,
intensive efforts are needed to evaluate the optical/physical properties of aerosols and
clouds by means of both ground- and satellite-based remote sensing observations.
In order to obtain better understanding of these particulate matters, what is obviously
needed is the monitoring technique that enables the retrieval of their optical properties. In
this chapter, we propose multi-wavelength and multi-directional remote sensing of
atmospheric aerosols and clouds. The proposed method consists of the application of
ground-based radiation measurement, lidar measurement, differential optical absorption
spectroscopy (DOAS), and satellite observations using natural as well as artificial light
sources. Such combinatory approach makes it possible to measure various aspects of
radiation transfer through the atmosphere, especially the influence of tropospheric aerosols
and clouds. Also, the data provided from the ground-based solar irradiance/sky radiance
measurement and DOAS are valuable for precisely characterizing the optical property of
aerosol particles near the ground level, including the information from both particulate
scattering and gaseous absorption. Such ground data are also indispensable for the
atmospheric correction of satellite remote sensing data in and around the visible range of the
radiation spectrum. The multi-wavelength and multi-directional observation schemes
treated in the present chapter are summarized in Table 1.

2. Differential optical absorption spectroscopy (DOAS)

The method of differential optical absorption spectroscopy (DOAS) provides a useful tool
for monitoring atmospheric pollutants through the measurement of optical extinction (i.e.,
the sum of absorption and scattering) over a light path length of a few kilometres (Yoshii et
al., 2003, Lee et al., 2009; Si et al., 2005; Kuriyama et al., 2011). The DOAS method in the
280 Remote Sensing – Applications

Scheme Wavelength Direction Aerosol Trace gas

UV, VIS, and
Measurable Measurable (e.g.
NIR, with the Nearly horizontal
DOAS through the NO2 around 450
resolution of measurement
spectral intensity nm)
array detector
Solar direction/
UV and VIS, with Measurable Measurable (e.g.
any direction
solar/skylight the resolution of through the H2O around 720
including the
array detector spectral intensity nm)
zenith
Fundamental and
Not applicable
harmonics of Vertical and slant Profiling by
unless tunable
Lidar Nd:YAG laser path solving the lidar
lasers are
(1064, 532 and observations equation
employed
355 nm)
Evaluated and
Spectral bands
removed in the
Spectral bands in Nadir or near- are usually too
Satellite process of
UV, VIS, and NIR nadir directions wide to retrieve
atmospheric
trace gases
correction
Table 1. Various schemes of atmospheric observation discussed in this chapter.

visible spectral region is quite suitable for urban air pollution studies, since both nitrogen
dioxide (NO2) and aerosol, the most important pollutants originated from human activities,
can directly be measured using a near horizontal light path in the lower troposphere.
Although conventional approach in the DOAS measurement is to install a light source, our
group at the Centre for Environmental Remote Sensing (CEReS), Chiba University, has
established a unique DOAS approach based on aviation obstruction lights (white flashlights)
equipped at tall constructions such as smokestacks (Yoshii et al., 2003, Si et al., 2005;
Kuriyama et al., 2011). Since those xenon lamps produce flash pulses every 1.5 s during the
daytime, they can easily be recognized with the coverage of the whole visible spectral range.
Thus, a simple setup consisting of an astronomical telescope and a compact spectro-
radiometer can be employed for the measurement of NO2. Also, the stable intensity of the
light source makes it possible to retrieve aerosol, or suspended particulate matter (SPM)
concentration in the lower troposphere, since the intensity variation of the detected light is
mostly ascribable to the aerosol extinction over the light path (Yoshii et al., 2003).
As shown in Figs. 1 and 2, the principle of DOAS analysis of NO2 concentration is based on
matching high-pass filtered spectral (wavelength) features between the DOAS-observed
optical thickness () and laboratory-observed molecular absorption spectrum (). Because
of the Lambert-Beer’s law, the optical thickness, , is expressed as

   ln( I / I 0 ), (1)

where I and I0 stand for the observed and reference spectrum, respectively. The reference
spectrum can be obtained by either operating the DOAS spectrometer at a short distance
Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 281

Fig. 1. Schematic flow of the DOAS analysis. The net radiation from the pulsed light source
can be retrieved by subtracting the background due to sky radiation, and an appropriate
portion of the spectrum is compared with the molecular cross-section spectrum obtained
from laboratory measurement. Then, the “high-frequency” components of the observed
optical thickness () and cross-section data () are compared to derive the molecular
number density along the optical path length, L.

Fig. 2. An example of DOAS spectral matching, in which the correlation between the
differential optical thickness from the DOAS data and differential absorption is examined to
determine the average volume concentration of NO2 molecules.
282 Remote Sensing – Applications

from the light source, or observing the spectrum under very clear atmospheric conditions with
minimal aerosol loading. The optical thickness is generally proportional to the product of
extinction coefficient, , and the light path length, L, i.e.,=L. In the case of molecular
absorption, the extinction coefficient is equal to the absorption coefficient, which can be given as
the product of absorption cross-section, , and the molecular number density, N, i.e.,  =N.
Although molecular scattering (Rayleigh scattering) and aerosol scattering (Mie scattering) also
exist, their contribution can be eliminated by applying the high-pass filtering to both  () and 
() (where  is wavelength), since the absorption feature of NO2 is a rapidly varying function
with wavelength (see insets in Figs. 1 and 2), while the wavelength dependence of Rayleigh or
Mie scattering is much more moderate. Thus, after the high-pass filtering, one obtains

  ( NL ) . (2)

This indicates that the correlation analysis between the rapidly varying components of the
optical thickness and NO2 cross-section in an appropriate wavelength range can lead to the
determination of the molecular number density, hence the volume concentration ratio, of
NO2 along the DOAS observation light path. An example of the retrieval of NO2 in the Chiba
city area is shown in Fig. 3. In this case, the DOAS result shows the average concentration
over a light path length of 5.5 km. From Fig. 3, it is seen that the DOAS data show good
temporal correlation with the ground sampling data from nearby sampling stations. Note
that the temporal resolution (5 min) of the DOAS observation is much better than that of the
ground sampling (1 h). The observation of DOAS spectra is limited to daytime, since the
white flashlight (Xe light) is replaced with blinking red lamps during night time.

Fig. 3. Comparison of NO2 volume concentration between the DOAS and conventional
ground sampling measurements during June 19 - 27, 2011. The DOAS data are based on the
measurement at CEReS, Chiba University, using an aviation obstruction flashlamp located
around 5.5 km in the north direction. The ground sampling data are from two nearby
sampling stations (Chigusadai Elementary School and Miyanogi stations) operated by the
municipal government.
Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 283

The analysis of light intensity detected with a DOAS spectrometer can yield information
also on aerosol extinction along the light path. The wavelength dependence of each
atmospheric component is exemplified in Fig. 4(a), where it is apparent that the contribution
from aerosol extinction is much more significant than that from either NO2 or molecular
Rayleigh scattering. The optical thickness associated with aerosol extinction can generally be
given as

 ( )  B( /  0 ) A , (3)

where A=ang is called the Angstrom exponent and B the turbidity constant. The value of A
changes with the aerosol size distribution in such a way that a smaller value (~ 0.5) indicate
the dominance of relatively coarse particles (such as sea salt or dust), while a large value
(~ 2) that of relatively fine particles (such as ammonium sulfate or ammonium nitrate). The
value of B, on the other hand, is equal to the aerosol optical thickness as wavelength 0,

Fig. 4. Aerosol measurement from DOAS data: (a) comparison of contributions of gas (NO2)
absorption, aerosol extinction, and molecular extinction (Rayleigh scattering) to DOAS
optical thickness, and (b) temporal change of SPM concentration from ground sampling and
aerosol extinction coefficient from DOAS during February 1 to 7, 2011 observed in Chiba.
284 Remote Sensing – Applications

which is chosen to be 550 nm or some appropriate value within the observation wavelength
range. Figure 4(b) shows the result of analysis based on eq. (4). As seen from Fig. 4(b) the
temporal variation shows good agreement between the DOAS-derived aerosol optical
thickness and the SPM mass concentration observed from the ground sampling.

3. Solar and skylight radiation measurement

For aerosols, network observation activities have been undertaken in terms of
skyradiometer measurements (Takamura & Nakajima, 2004). Alternatively, the use of a
compact, stand-alone spectroradiometer (EKO, MS-720) enables the spectral measurements
of direct solar radiation (DSR), solar aureole (AUR) and scattered solar radiation (SSR)
(Manago & Kuze, 2010; Manago et al., 2010). Since the instrument is powered by batteries
with no PC requirement during measurements, it provides better portability compared to a
skyradiometer. The wavelength coverage between 350 and 1050 nm with a resolution of 10
nm is useful for precise evaluation of the aerosol optical properties as well as that of the
water vapor column amount. The wide dynamic-range measurement of both the direct and
scattered solar radiation is attained by means of a thick diffuser and a stable photodiode
array, in combination with the automatic exposure control equipped to the handy
spectroradiometer (MS-720) (Manago & Kuze, 2011). In order to facilitate the radiation
transfer calculation in the retrieval procedure, home-made baffle tubes are used to limit the
field of view (FOV) of the observation to 20 deg (SSR) and 5 deg (DSR).
The radiation measurements were conducted at the CEReS site (35.62ºN, 140.10ºE) under
clear-sky conditions, mostly around noon. The SSR measurements were made in 24
directions (north, east, south, and west directions, each with 6 elevation angles). The DSR
and AUR components were measured before and after the SSR measurements. The total
time required for a set of DSR, AUR, and SSR data was 30 - 40 minutes. Approximately 130
datasets were obtained during the observation period from August 2007 to March 2009.
Independent measurement of AOD was carried out with a sunphotometer (Prede, PSF-
100). This instrument has four channels centred at 368, 500, 675, and 778 nm, each having
the bandwidth of 5 nm. The wavelength dependence of AOT is analyzed with eq. (3) to
obtain the Angstrom exponent. During the daytime the AOD is retrieved from the solar
radiation intensity within a FOV of 1 deg at an interval of 10 s. From the sunphotometer
measurement, A=ang and the coefficient B (turbidity constant at the reference wavelength
0= 550 nm in this case) can be retrieved.
In our ground observation with the battery-operated spectroradiometer, the direct solar
irradiance, aureole radiance, and scattered solar radiance were measured in various
directions as mentioned above. Even with these detailed measurements, however, usually it
is not possible to determine the complete composition of aerosol particles. Thus, we rely on
the three-component aerosol model (TCAM), in which three aerosol types of water soluble,
oceanic, and soot components are considered as a basis set which is “quasi-complete” to
describe the aerosol optical parameters, namely the wavelength dependence of extinction
coefficient, single scattering albedo, asymmetry parameter, and scattering phase function.
Figure 5 shows the wavelength dependence of the real and imaginary parts of the aerosol
refractive index for the three aerosol components. It has been shown that most of the
irradiance/radiance values are well reproduced by appropriately adjusting the total and
Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 285

relative contributions of these three basis components as well as the size distribution of each
component (Manago et al, 2011). As seen from Fig. 5, the soot component shows remarkably
high value of the imaginary part of the refractive index. This indicates that the absorption
property is higher (single scattering albedo is lower) for aerosol with more contribution of
soot particles. Figure 6 shows an example of the results of the irradiance and radiance
observations. Figure 7 shows an example of aerosol optical parameters derived from the
TCAM analysis of the data: Fig. 7(a) shows the wavelength dependence of the aerosol
extinction coefficient (normalized to the value at 550 nm), (b) single scattering albedo, (c)
asymmetry parameter, and (d) scattering phase function at wavelength 550 nm. In Sec. 5
below, we describe the application of these aerosol characteristics to the atmospheric
correction of satellite remote sensing data.

Fig. 5. Real and imaginary parts of the complex refractive index of the three aerosol
components: component 1, 2 and 3 refer to the water soluble, oceanic, and soot aerosol
types, respectively.

Fig. 6. Spectra observed around noon on October 16, 2008: (a) direct solar radiation (DSR),
(b) aureole (AUR), and (c) scattered solar radiation (SSR). Acceptance angle of the
instrument is 5 deg for DSR, 5-20 deg for AUR, and 20 deg for SSR. Simulation curves based
on the TCAM best fitting are also shown with data points (circles) used for the fitting.
286 Remote Sensing – Applications

Fig. 7. Aerosol optical parameters derived from the TCAM analysis of the data shown in Fig.
6: (a) wavelength dependence of the aerosol extinction coefficient (normalized to the value
at 550 nm), (b) single scattering albedo, (c) asymmetry parameter, and (d) scattering phase
function at wavelength 550 nm.

4. Lidar measurement of aerosols and clouds

While the DOAS method and skylight/solar radiation measurement lead to the retrieval of
atmospheric information integrated over optical paths, the lidar measurement makes it
possible to measure aerosol and cloud distributions (profiles) along the optical path. Here
we report the result of multi-wavelength lidar measurement conducted at CEReS.
Conventionally lidar data have been analyzed by means of the Fernald method (Fernald,
1984), in which the lidar equation

c G( R)  R 
P( R)  P0
2
AK 2  ( R)exp
R  
 2 0  ( R) dR

(4)

is solved by starting the integration from the far-end boundary. In eq. (4), P(R) is the power
detected by the lidar system corresponding to a distance R, P0 is the power of the emitted
laser radiation, c is the speed of light, is the time duration of the laser pulse, A is the area of
the lidar telescope, G(R) is the function describing the overlap between the laser beam and
telescope field of view, R) is the backscattering coefficient, and R) is the extinction
coefficient. Since both air molecules and aerosol particles contribute to the scattering and
extinction, one needs to separate these two components in solving the lidar equation. This
can be achieved by introducing the ratio between the extinction coefficient and the
backscattering coefficient. Thus, for aerosols,

 d 
S1 ( R )   1 ( R ) /  1 ( R )   1 ( R ) /  1  (5)
 d  

is assumed, whereas for air molecules,

Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 287

S2   2 ( R ) /  2 ( R )  8.52 (sr) (6)

is used as a constant. In eqs. (5) and (6), suffix 1 and 2 refer to aerosol and air molecule,
respectively. In eq. (5), 1(R) and (d1/d)= indicate the total cross-section and backward
differential cross-section of aerosol scattering, respectively. The parameter S1 is often called
the lidar ratio. In eq. (6), the range dependence of S2 is omitted, since the composition of air
molecules is stable througout the troposphere. Under these assumptions, the lidar equation
can be analytically solved as

S 1  R S 1  R  X  R  exp I  R 
 1  R    2  R  . (7)
S2 X  Rc 
 J  R
 1  Rc   2  Rc 

S 1  Rc  S2

Here

X( R)  R 2 P( R) (8)

is the range-corrected signal, and functions I(R) and J(R) are defined as

Rc  S 1  R  
I  R  2   1  2  R dR (9)
R
 S2 

and

Rc
J  R   2  S 1  R  X  R  exp I  R  dR. (10)
R

In eqs. (7), (9) and (10), Rc denotes the range of a far-end boundary, at which each
integration is started. The reason that a far-end boundary value is assumed rather than a
near-end boundary value is the stability of the numerical evaluation of eq. (7) (Fernald
1984).
Usually signals of a vertically looking lider are analyzed assuming that the aerosol property
does not change with the altitude. Under this assumption, the range dependence in eq. (5)
can be neglected. Even in this case, however, it is necessary to determine the value of lidar
ratio as a function of wavelength [S1=S1()] in order to analyze multi- wavenength lidar data.
One way to accomplish this is to use ancillary data from a sunphotometer (Kinjo et al., 1999),
since the wavelength dependence of optical thickness provides a constraint to the
intagration of 1(R, ) from R=0 to R =Rc. Another approach is to employ the aerosol
properties measured at the ground level. In the case of Fig. 8, for example, the S1 values of
54.7, 53.0, 46.0 and 43.2 sr are used for = 355, 532, 756 and 1064 nm, respectively, as
derived from the chemical analysis of ground sampling data taken monthly at CEReS
(Fukagawa et al., 2006). In Fig. 8, panel (a) shows the temporal variation of the aerosol
extinction profile measured for 1064 nm and relative humidity (RH) at the ground level,
while panel (b) depicts that of the profile of the Angstrom exponent, ang, as derived from
the analysis of lidar data for the four wavelengths. The features in these panels indicate
288 Remote Sensing – Applications

Fig. 8. Analysis of vertical looking multi-wavelength lidar data: (a) extinction profile for
wavelength = 1064 nm observed at CEReS on 17-18 November 2005, and Angstrom
exponent derived from extinction coefficients observed for = 355, 532, 756 and 1064 nm.
The analysis is based on the Fernald method with lidar parameters S1 = 54.7, 53.0, 46.0 and
43.2 sr for each lidar wavelength (based on sampling result at CEReS) and the reference
altitude of Rc = 5.5 km.

that relatively higher extinction near the ground level is observed when RH increases, and at
the same time, smaller values of ang are observed. It is likely that both of these observations
are due to the aerosol growth associated with the increase of RH.

5. Atmospheric correction of satellite remote sensing data

Images taken from satellite sensors are affected by both the ground reflectance and
atmospheric conditions, which include the influence of scattering and absorption of air
molecules and aerosol particles. The process of atmospheric correction, in which such
atmospheric effects are precisely evaluated and removed, is indispensable for extracting the
intrinsic information of the ground reflectance from satellite imagery (Tang et al., 2005;
Kaufman et al., 1997). Although it is rather straightforward to make corrections on the
Rayleigh scattering of air molecules, aerosol particles are quite variable both temporally and
spatially. This is due to the variable origins of aerosols, consisting of relatively coarse
Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 289

particles of natural origins (such as sea-salt and soil particles) and relatively fine particles of
anthropogenic origins (such as sulphate and soot particles).
In standard atmospheric correction, it is customary to assume some representative aerosol
models such as maritime, rural, continental, or urban aerosol, to implement the radiative
transfer calculation of a satellite scene. This approach has an obvious disadvantage that if
the assumed aerosol properties are different from those of real aerosols included in the
satellite scene, the resulting information on the ground reflectance is inaccurate. To
overcome this difficulty, here we use the aerosol information derived from the ground
observation implemented nearly simultaneously with the satellite overpass. Such ancillary
information ensures better separation of the ground and atmospheric effects from satellite
imagery. Figure 9 shows the schematic drawing of radiation components considered in the
radiative transfer calculation (Kotchenova et al., 2006). In this scheme, the radiance
originated from the target area is denoted as Ltar, which consists of the ground direct (Lgd)
and ground indirect (Lgi and Lgi’) components. The environmental radiance, Lenv, is the
component associated with the surface reflection that takes place in adjacent pixels. The
atmospheric radiance, Latm, consists of two terms, namely, the path radiance due to single
scattering (Lps) and that due to multiple scattering (Lpm).

Fig. 9. Schematic drawing of radiation components considered in the radiative transfer

calculation. See text for the explanation of radiance components shown in this figure.

In the present research, the ground measurement by means of a compact spectroradiometer

was implemented in synchronous with the overpass of the satellite observation. The aerosol
optical parameters were derived by analyzing both the direct solar radiation (DSR) and
scattered solar radiation (SSR) through the Mie-scattering and radiative transfer
calculations, as explained in Sec. 3 of this chapter. When the aerosol loading is relatively
small (clear days), it is likely that the aerosol model resulting from this procedure can be
applicable to the whole region of the Moderate Resolution Imaging Spectroradiometer
(MODIS) image, and the atmospheric correction is applied to the image. Since this correction
290 Remote Sensing – Applications

is based on the aerosol model from the simultaneous measurement, the resulting
distribution of the surface reflectance is considered to be more reliable than the result that
would be obtained by assuming usually available “standard” aerosol models such as urban,
rural, or oceanic models. The surface reflectance map (clearmap) on such a “clear” day, in
turn, can be used as a standard for that particular season of the year, and the atmospheric
correction of MODIS data taken on more turbid days can be implemented on the basis of
these standard clearmaps. This process leads to the derivation of the distribution of aerosol
optical thickness (map).
For each of the visible bands of MODIS, a lookup table of the radiance at the top of the
atmosphere, Ltotal(, 550), was constructed on the basis of the aerosol optical parameters and
the geometric data describing the observational conditions of each image. Here, is the
diffuse reflectance of each pixel, and 550 is the aerosol optical thickness (AOT) at
wavelength 550 nm. The reflectance property of the surface was assumed to be Lambertian,
and the radiative transfer calculation was carried out using the 6S code (Kotchenova et al.,
2006).
The atmospheric correction was applied to channels 1 through 4 covering wavelength range
between 0.450 and 0.876 m of the Terra/MODIS and Aqua/MODIS images. The ground
resolution of the MODIS sensor is 0.5 km×0.5 km/pixel. The region of 600×600 pixels
around Chiba University was extracted from each of the MODIS images, which were taken
from the satellite data archiving system of CEReS, Chiba University. The ground
observations using the spectroradiometer were carried out at CEReS around noon on nearly
cloud-free days from 2007 to 2009 (around 130 days). In order to take the time lag of around
2 h between the satellite overpass (10:00 local time) and the spectroradiometer observation
(12:00) into account, the sunphotometer data taken at CEReS were employed to examine the
temporal stability of atmospheric conditions. If the AOT derived from the spectroradiometer
was close to the AOT value measured with the sunphotometer at the time of satellite
overpass, the data were employed in the atmospheric correction. Otherwise, the data were
not included in the clear-day analysis lest the instability in the atmospheric condition might
also degrade the regional stability (i.e. homogeneity) of the aerosol distribution. Figure 10
shows the wavelength dependence of the surface reflectance (clear map) of the pixel
including the location of Chiba University for various months in the year 2008.
Figure 11 shows the seasonal variation of the surface reflectance for the Chiba University
pixel obtained from the analysis of MODIS band 4 centred at 550 nm. For the sake of
comparison, our previous result obtained from the Landsat-5 analysis is also depicted
(Todate et al., 2004). Note that the Landsat reflectance was obtained assuming a standard
aerosol model (maritime), whereas the TCAM aerosol model is used in the present MODIS
analysis. Pixels with vegetation and soil coverage are shown for the Landsat data, since the
ground resolution associated with this sensor (30 m) is much better than the MODIS
resolution of 500 m. From Fig. 11, it is seen that the surface reflectance decreases from
November to December, due to the decrease in the vegetation coverage during winter. In
winter the reflectance shows no critical dependence on the aerosol model assumed in the
atmospheric correction because of the fact that the AOT tends to be small. In summer, on the
contrary, the AOT generally increases so that the resulting value of surface reflectance varies
in accordance with the aerosol model employed in the analysis.
Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 291

Fig. 10. Surface reflectance at Chiba University (2008)

Fig. 11. Seasonal variation of surface reflectance at the MODIS pixel including the Chiba
university campus (=550 nm).

From the present TCAM analysis of MODIS data, monthly reflectance image (monthly) is
generated for each month as a composite of pixels that exhibit the lowest reflectance. This
process ensures the removal of cloud pixels that might contaminate the resulting map.
These monthly maps, in turn, are employed in the radiative transfer analysis to derive the
aerosol distribution (map) from images taken on relatively turbid days. Examples of the
reflectance and aerosol distribution images are shown in Fig. 12.
292 Remote Sensing – Applications

Fig. 12. Analysis of MODIS data in November 2007: (a) surface reflectance map (540 - 570
nm), and (b) aerosol optical thickness at 550 nm on 24 November 2007.

6. Conclusion
Optical properties of aerosols and clouds play an important role in the consideration of the
Earth’s radiation budget. In this chapter, we have described multi-wavelength and multi-
directional remote sensing of the troposphere, putting emphasis on the visible part of the
spectrum. The DOAS approach enables the direct observation of air pollutants by
employing a nearly horizontal light path in the lowest part of troposphere, where the
highest concentrations of pollutants such as NO2 and aerosol (SPM) are found. The
observation of direct solar radiation and scattered solar radiation (sky light), on the other
hand, is useful for retrieving detailed aerosol optical properties under clear-sky conditions.
Thus, the data can be quite useful for implementing precise atmospheric correction on
satellite-observed imagery that includes the ground observation point. The multi-
wavelength lidar observation provides an efficient tool to elucidate the vertical profiles of
aerosol particles. By combining the lidar data with some appropriate ancillary data such as
the ground-level characterization of aerosol properties, it becomes possible to derive useful
information on temporal as well as spatial information on aerosol and cloud characteristics
in the atmosphere.

7. Acknowledgment
We acknowledge the financial support of the Grant-in-Aid from the Ministry of Education,
Culture, Sports, Science & Technology in Japan (#21510006). Also, contributions of a number
of graduate students who participated in various researches presented in this chapter are
gratefully acknowledged.
Multi-Wavelength and Multi-Direction Remote Sensing of Atmospheric Aerosols and Clouds 293

8. References
Fernald, F. G., Analysis of atmospheric lidar observation: some comments. Appl. Opt. Vol
23, (1984), pp. 652-653, ISSN 0003-6935
Fukagawa, S., Kuze, H., Bagtasa, G., Naito, S., Yabuki, M., Takamura, T., Takeuchi, N.,
Characterization of seasonal and long-term variation of tropospheric aerosols in
Chiba, Japan. Atmospheric Environment, Vol. 40, No.12, (2006), pp. 2160-2169,
ISSN 1352-2310
Intergovernmental Panel on Climate Change (IPCC), Climate Change 2007: The Physical
Science Basis, Cambridge Univ. Press (2007) ISBN-13: 978-0521705967, Cambridge,
U.K.
Kaufman, Y. J., Tanré, D., Gordon, H. R., Nakajima, T., Lenoble, J., Frouins, R., Grassl,
H.,Herman, B. M., King, M. D., Teillet, P. M., Passive remote sensing of
tropospheric aerosol and atmospheric correction for the aerosol effect. J.
Geophys. Res., Vol.102, No.D14, (1997), pp. 16,815-16,830, ISSN 0148-0227
Kinjo, H., Kuze, H., Takeuchi, N., Calibration of the Lidar Measurement of Tropospheric
Aerosol Extinction Coefficients. Jpn. J. Appl. Phys. Vol. 38, (1999), pp. 293-297, ISSN
0021-4922
Kotchenova, S.Y., Vermote, E.F., Matarrese, R., Klemm, Jr. F.J., Validation of a vector version
of the 6S radiative transfer code for atmospheric correction of satellite data. Part I:
Path radiance. Applied Optics Vol. 45, No. 26, No.10 (September 2006), pp.6762-
6774, ISSN 0003-6935
Kuriyama, K., Kaba, Y., Yoshii, Y., Miyazawa, S., Manago, N., Harada, I., Kuze, H., Pulsed
differential optical absorption spectroscopy applied to air pollution measurement
in urban troposphere. J. Quantitative Spectroscopy & Radiative Transfer,
Vol. 112, (2011), pp. 277-284, ISSN 0022-4073
Lee, H.; Kim, Y.J., Jung, J., Lee, C., Heue, K.P., Platt, U., Hu, M., 6 Zhu, T., Spatial and
temporal variations in NO2 distributions over Beijing, China measured by imaging
differential optical absorption spectroscopy. J. Environ. Management, Vol. 90, No.5,
(April 2009), pp. 1814-1823, ISSN 0301-4797
Manago, N., Kuze, H., Determination of tropospheric aerosol characteristics by spectral
measurements of solar radiation using a compact, stand-alone spectroradiometer.
Appl. Opt. Vol. 49, (2010), pp. 1446-1458, ISSN 1539-4522
Manago, N., Miyazawa, S., Bannu, Kuze, H., Seasonal variation of tropospheric aerosol
properties by direct and scattered solar radiation spectroscopy. J. Quantitative
Spectroscopy & Radiative Transfer, Vol. 112, Issue 2, (January 2011), pp. 285-291,
ISSN 0022-4073
Miyazawa, S., Manago, N., Kuze, H., Precice atmospheric correction of MODIS data using
aerosol optical parameters derived from simultaneous ground measurement, in
Proceedings of the 49th Annual Meeting of the Remote Sensing Society of Japan,
(2010), pp.245-246 (in Japanese)
Seinfeld, H. and Pandis, S.N. (1998). Atmospheric Chemistry and Physics, from Air Pollution to
Climate Change, John Wiley, ISBN 0-471-17815-2, New York, U.S.A.
Si, F., Kuze, H., Yoshii, Y., Nemoto, M., Takeuchi, N., Kimura, T., Umekawa, T., Yoshida, T.,
Hioki, T., Tsutsui, T., Kawasaki, M., Measurement of regional distribution of
atmospheric NO2 and aerosol particles with flashlight long-path optical
monitoring. Atmos. Environ. Vol. 39, (2005), pp. 4959-4968, ISSN 1352-2310
294 Remote Sensing – Applications

Takamura, T., Nakajima, T., Overview of SKYNET and its activities. Optica Pura y Aplicada
Vol. 37, (2004), pp. 3303–3308, ISSN 0030-3917
Tang, J., Xue, Y., Yu, T., Guan,Y. Aerosol optical thickness determination by exploiting the
synergy of TERRA and AQUA MODIS. Remote Sensing of Environment Vol. 94,
(2005), pp. 327–334, ISSN 0034-4257
Todate, Y., Minomura, M., Kuze, H., Takeuchi, N., On the retrieval of aerosol information on
the land surface from Landsat-5/TM data, in Proceedings of the 36th Annual
Meeting of the Remote Sensing Society of Japan, (2004), pp.57 - 58. (in Japanese)
Yoshii, Y., Kuze, H., Takeuchi, N., Long-path measurement of atmospheric NO2 with an
obstruction flashlight and a charge-coupled-device spectrometer. Appl. Opt., Vol.
42, (2003), pp. 4362-4368, ISSN 0003-6935
 Section 3

Oceans and Cryosphere

 12

Remote Sensing of
Submerged Aquatic Vegetation
Hyun Jung Cho1, Deepak Mishra2,3 and John Wood4
1Department of Integrated Environmental Science,
Bethune-Cookman University,
Daytona Beach, FL
2Department of Geosciences, Mississippi State University
3Northern Gulf Institute and Geosystems Research Institute,

Mississippi State University, MS State, MS

4Harte Research Institute for Gulf of Mexico Studies,

Texas A&M University-Corpus Christi,

Corpus Christi, TX
USA

1. Introduction
Remote sensing has significantly advanced spatial analyses of terrestrial vegetation for
various fields of science. The plant pigments, chlorophyll a and b, strongly absorb the energy
in the blue (centered at 450 nm) and the red (centered at 670 nm) regions of the
electromagnetic spectrum to utilize the light energy for photosynthesis. In addition, the
internal spongy mesophyll structures of the healthy leaves highly reflect the energy in the
near-infrared (NIR) (700- 1300) regions (Jensen, 2000; Lillesand et al., 2008). The distinctive
spectral characteristics of the green plants, low reflectance in the visible light and high
reflectance in NIR have have been used for mapping, monitoring and resource management
of plants; and also have been used to develop spectral indices such as Simple Vegetation
Index (SVI = NIR reflectance – red reflectance) and Normalized Difference Vegetation Index
(NDVI = (NIR reflectance – red reflectance)/(NIR reflectance + red reflectance)) (Giri et al.,
2007).
The simplicity and flexibility of vegetation indices allow comparison of data obtained under
varying light conditions (Walters et al., 2008). NDVI was first suggested by Ruose et al.
(1973) and is one of the earliest and most popular vegetation index used to date. It is usually
applied in an attempt to decrease the atmospheric and surface Bidirectional Reflectance
Distribution Function (BRDF) effects by normalizing the difference between the red and NIR
reflectance by total radiation. Index values have been associated with various plant
characteristics, including vegetation type (Geerken et al., 2005), vegetation cover (du Plessis,
1999), vegetation water content (Jackson et al., 2004), biomass and productivity (Fang et al.,
2001), chlorophyll level (Wu et al., 2008), PAR absorbed by crop canopy (Goward &
Huemmrich, 1992), and flooded biomass (Beget et al., 2007) at a broad span of scales from
individual leaf areas to global vegetation dynamics.
298 Remote Sensing – Applications

2. Remote sensing of submerged aquatic vegetation (SAV)

2.1 Submerged aquatic vegetation (SAV)
Submerged aquatic vegetation (SAV) is a group of vascular plants that grow underwater
which can grow to the surface of, but not emerge from shallow waters. SAV includes
seagrass species that are a vital component of ecological processes, dynamics, and
productivity of coastal ecosystems. Healthy beds of SAV provide nursery and foraging
habitats for juvenile and adult fish and shellfish, protect them from predators, provide food
for waterfowl and mammals, absorb wave energy and nutrients, produce oxygen and
improve water clarity, and help settle suspended sediment in water by stabilizing bottom
sediments (Jin, 2001; Findlay et al., 2006). Assessment of SAV distribution, composition, and
abundance has been of a particular interest to coastal environmental managers, scientists,
developers, and recreational users as the information serves as an excellent indicator of
aquatic environmental quality.

2.2 Remote sensing of underwater habitats

Remote sensing is a valuable tool for monitoring benthic habitats such as SAV, benthic
algae, and coral-reef ecosystems, and several researchers have tested airborne and space-
borne sensor systems for such studies (e.g., Mishra et al., 2005). Spatial resolutions of these
systems range from 30 m for the Landsat Thematic Mapper (TM) to 2.44 m for QuickBird
multispectral data and 1 m or less for airborne hyperspectral data. Those evaluating the
utility of TM have mapped subtidal coastal habitats (Khan et al., 1992), delineated sand
bottoms (Michalek et al., 1993), classified coral reef zones (Mishra et al., 2005, 2006),
evaluated the benthos (Matsunaga & Kayanne, 1997), and performed time series analyses
(Dobson & Dustan, 2000). Similarly, researchers have used IKONOS (4 m) and QuickBird
(2.44 m) imagery with radiative transfer models to map benthic habitats (Mishra et al., 2005,
2006) and apply a similar model to Airborne Imaging Spectroradiometer for Applications
(AISA) hyperspectral data to identify benthic habitats (Mishra et al., 2007).
Mapping of SAV using satellite data has focused on supervised and unsupervised
classifications based on signal variations in the multispectral bands, especially those in the
short visible wavelengths with high water penetration (Ackleson & Klemas, 1987;
Lyzenga, 1981; Marshall & Lee, 1994; Maeder et al., 2002; Ferguson & Korfmacher, 1997;
Pasqualini et al., 2005). The NIR region is seldom used due to its high attenuation in
water. When SAV beds are dense, the water is clear, and depth and sediment relatively
constant, fine-scale spectral variation is often overlooked during classification. In other
cases, the radiative transfer model is used to correct the solar angle, atmospheric
perturbation, substrate type, and depth, but requires extensive in situ measurements
(Zimmerman & Mobley, 1997).
Most of the currently available radiative transfer models or physics-based models have been
applied to map benthic features in relatively clear aquatic environments (i.e. relatively deep,
pristine coral reefs or seagrass meadows) and do not adequately correct for the strong NIR
absorption by water (Mumby et al., 1998; Holden & LeDrew, 2001; Holden & LeDrew, 2002;
Ciraolo, 2006; Brando et al., 2009). However, NIR reflectance serves as the primary cue for
discriminating vegetation type and as the critical component for the widely used vegetation
indices.
Remote Sensing of Submerged Aquatic Vegetation 299

2.3 Dilemmas in remote sensing of shallow aquatic system and SAV

Remote sensing of benthic habitats is complicated because of several factors including (1)
atmospheric interferences, (2) variability in water depth, (3) water column attenuation, and (4)
variability in bottom albedo or bottom reflectance. In the case of aquatic remote sensing, the
total signal received at satellite altitude is dominated by radiance contributed by atmospheric
scattering, and only 8-10% of the signal corresponds to the water reflectance and reflectance
from benthic features (Kirk, 1994, Mishra et al., 2005). Therefore, it is advisable to correct for
atmospheric effects to retrieve any quantitative information for surface waters or benthic
habitats from satellite images. Therefore, the lack of a rigorous absolute atmospheric correction
procedure can introduce significant errors to a satellite derived benthic habitat map. There is
also a tendency among benthic mapping researchers to use a relative atmospheric correction
procedure such as a deep-water pixel correction, especially when local aerosol data and
validation data are lacking. This often causes mediocre classification results.
Knowledge of the optical properties of the water column can help eliminate changes in
reflectance attributable to variable depth and water column attenuation effects, which often
lead to misclassification of the benthos (Mishra et al., 2005). Mishra et al (2005) proved that to
derive accurate bottom albedo or bottom reflectance using a radiative transfer model, water
depth and water column optical properties (absorption and backscattering) should be known
for the study area. Knowledge of optical bottom albedo for shallow waters is necessary to
model the underwater and above-water light field, to enhance underwater object detection or
imaging, and ultimately to determine the distribution of benthic habitats (Gordon & Brown,
1974). Mishra et al (2005, 2006, 2007) also point out that the signals measured by a sensor from
above the water surface of a shallow marine environment are highly affected by
phytoplankton abundance (chlorophyll absorption), water column interactions (absorption by
water and scattering by suspended sediments), and radiance reflected from the bottom. For
the bottom contribution to be retrieved by a sensor the water column contributions have to be
removed and the optical properties have to be known or at least be derivable. However, it is
very challenging to measure these optical properties accurately because of logistical issues and
instrumentation errors, which also leads to an inaccurate benthic mapping project.
Variability in bottom types and hence albedo gives rise to a mixed spectral response that
often reduces the classification accuracy. Specific problems such as complex benthic
combinations (e.g., sandy areas with variable amounts of algal cover; variation in color,
texture, size), and error in depth estimation can also have a considerable impact on the
classification results. Mishra et al (2005) proposed several solutions to increase the number
of elements separable by a classification scheme and the classification accuracies including
an extensive field campaign acquiring substantial samples to enable statistical evaluation for
each class and deriving detailed ecological and biological information for each in situ data
point. Close-range hyperspectral studies that may aid in discriminating between different
types of benthic features can be used to develop baseline spectra to help minimize spectral
confusion in satellite imagery.
Shallow littoral areas (generally the areas between the shoreline to a water depth 2 m) are
one of the most productive habitats, yet the most sensitive landscape to human-induced
environmental alteration and global climate changes. Modeling of optical water properties
for the littoral zone is more complicated due to rapidly changing water depth and/or
substrate and higher amounts of Colored Dissolved Organic Matter (CDOM) and/or
300 Remote Sensing – Applications

suspended particles (phytoplankton, seston, and inorganic particles) compared to the

deeper portions of oceans. In addition, bottom backscattering in the shallow areas is more
significant, which makes the NIR signals more important, especially in areas that contain
substantial amount of seagrasses, benthic algae, or phytoplankton (Kutser et al., 2009) and
that the conventional Beer-Lambert’s exponential light attenuation with depth is not
applicable (Holden & LeDrew, 2002).
Upwelling signals from water bodies contain several components including reflectance from
water surface, water column (suspended matter), and bottom backscattering (SAV and
substrate) (Spitzer & Dirks, 1987). The aforementioned conventional vegetation indices also
are not effectively used for plants that grow underwater or that are temporarily flooded
(Beget & Di Bella, 2007; Cho et al. 2008) because the water overlying the vegetation canopies
reduces the vegetation effects of ‘red absorption’ and the ‘NIR reflectance’ (Han &
Rundquist, 2003; Cho, 2007; Cho et al., 2008; Fig. 1). Differentiation of the SAV spectral
signature from bare substrate or algae is further limited in shallow coastal waters that are
more turbid than open ocean waters (Bukata, 1995) due to higher levels of phytoplankton,
suspended sediment, and dissolved color. According to our on-going study using
hyperspectral data obtained over both experimental tanks and field seagrass habitat, the
SAV signal rapidly decreases as water depth increases, and almost completely disappears
within a depth of 0.5 meter in even mildly turbid waters (turbidities > 12 NTU).

7
6
Reflectance (%)

10
5
4
3
2
1
0
400 500 600 700 800 900
Wavelength (nm)

Fig. 1. Depth-induced reflectance variation of submerged aquatic vegetation (SAV) in clear

water between 10 cm and 50 cm above the SAV canopy. The line for the highest reflectance
is at 10 cm and the reflectance continuously decreases with water depth increases.

2.4 SAV mapping using hyperspectral data

Two decades ago, only spectral remote sensing experts had access to hyperspectral images
or the software tools necessary to take advantage of such images. Over the past decade,
hyperspectral image analysis has matured into one of the most powerful and fastest
growing technologies in the field of remote sensing (Phinn et al., 2008). While multispectral
Remote Sensing of Submerged Aquatic Vegetation 301

remote sensing systems detect radiation in a small number of broad regions of the
electromagnetic spectrum, hyperspectral sensors acquire numerous very narrow,
contiguous spectral bands throughout the visible, near-infrared, mid-infrared, and thermal
infrared portions of the electromagnetic spectrum for every pixel in the image, yielding
much more detailed spectral data (Govender et al., 2009).
Collection and processing of hyperspectral imagery can be quite costly, depending on the
size of the area to be studied. In order for the imagery to be usable for sub-aquatic analysis,
the following guidelines are suggested by Finkbeiner et al. (2001):
 The best time of year for collecting hyperspectral imagery may occur in early summer,
during the season of maximum biomass, and when there is less epiphytic coverage.
 The imagery should be collected when turbidity is low; this is often during times of low
or no winds. High turbidity may also be caused by heavy rains, winds on previous
days, and localized dredging. Often, boat traffic may cause a localized but far-spreading
plume of turbidity, as sediments are re-suspended.
 Winds can also cause problems other than turbidity, such as wrack lines, debris lines,
whitecaps, and areas with unacceptable amounts of glint. As a general rule, winds less
than 8 kph are acceptable, winds between 8-15 kph may be acceptable depending on the
locality, and winds higher than 15 kph are generally unacceptable.
 Tidal stage can play an important role in the success of imagery collection. Consult local
and/or NOAA tide gauges to plan for acquisition within 2 hours of the lowest tide for
the collection area, unless the estuary drains an area of highly turbid or tannic water, in
which case, a rising tide may be desirable.
 Collection times should be planned to adjust for sun angle, to avoid both sun glint and
shadows. As a general rule, sun angles between 30º and 45º are recommended; different
sensors may allow or require more or less angle.
 Clouds and haze create areas of shadows and distortion as well as white or gray streaks
in the imagery, and should be avoided as much as possible.
 Field work should occur simultaneously with the sensor flight. Since it is virtually
impossible to collect all the field data needed for signature development and accuracy
assessment in the same time frame as the flights occur, every effort should be taken to
gather field measurements as close to the actual flight as possible, and under similar
conditions.
 Field data should include measurements of reflectivity, turbidity, empirical or
anecdotal data on epiphytic coverage, bottom type and reflectance, classification of the
field point, and precise location. Locate these field measurements within a large enough
patch that there will be no ambiguity, and consider the spatial sphere of uncertainty.
For instance, if the imagery will have a radiometric accuracy of approximately 2 m, the
location should be consistent out to a four meter radius.
The unique spectral signatures of vegetation are often used as training data for
hyperspectral imagery classification. Chlorophyll and other pigments are found in SAV as
in other photosynthesizing vegetation, however, the ratio of these to each other will differ
by species, as well as with changes in conditions and stressors (Govender et al., 2009).
While these minor differences can be detected above the surface in spite of epiphytic coverage
(Fyfe, 2003), detection of these differences below the surface may be hampered or dampened
by the effects of the water column. Depth, water clarity or turbidity, organic and inorganic
302 Remote Sensing – Applications

materials within the water column, the surface of the water itself, and physical properties such
as the absorption of energy in the NIR and beyond can all affect the ability to discriminate the
relatively small differences in ratios of accessory pigments and chlorophyll (Kutser 2004).

3. Case study in SAV mapping using hyperspectral data

3.1 Hyperspectral algorithm to correct overlying water effects
A new water-depth correction algorithm was developed to improve detection of underwater
vegetation spectra signals. The algorithm was developed conceptually, calibrated, and
validated using experimental and field data. The conceptual model was based on the idea
that the upwelling signals measured from a water surface is the sum of the energy reflected
from the water surface, the water column, and also from the water bottom surface. The
energy reflected from the water surface and the water column (the volumetric reflectance)
was combined as a single term because the surface reflectance is a constant and does not
change with water depth changes (Lu & Cho, 2011).
The effects of the overlying water column on upwelling hyperspectral signals were modeled
by empirically separating the energy absorbed and scattered by the water using data
collected through a series of controlled experiments using hypothetical bottom surfaces that
either 100% absorbs or 100% reflects (Cho & Lu, 2010). Later, the white surface (the 100%
reflecting surface) was replaced with a gray surface with a known reflectance to reduce
problems associated with enhanced multi-path scattering (Lu & Cho, 2011). The
experimental setting allowed the calculation of water absorption and scattering values for
up to a water depth of 60 cm, and the light remaining at water depths that were beyond the
experimentally measured points were estimated by establishing the mathematical
relationships between water depth and the vertical attenuation coefficient (Kd) derived from
the experimental data (Washington et al., 2011). The depth- and wavelength-dependent water
absorption and volumetric scattering factors (0-100 cm; 400 – 900 nm) were calculated and
applied to independently-measured underwater vegetation signals and airborne hyperspectral
data taken over shallow seagrass beds, to remove the effects of the overlying water.

3.2 Successful water correction in the infrared region

The empirically driven correction algorithm significantly restored the vegetation signals,
especially in the NIR region, when applied to independently measured reflectance of
underwater plants taken over indoor and outdoor tanks (Cho & Lu, 2010; Washington et al.
2011). The algorithm was also successful in restoring the NIR signals originating from
seagrass-dominated sea floors when applied to airborne hyperspectral data of Mississippi
and Texas coastal waters (Cho et al., 2009; Lu & Cho, 2011; Fig. 2). As stated earlier, NIR
reflectance serves as the primary cue for discriminating benthic vegetation from other
substrates. Due to the restored NIR reflectance, the correction algorithm increased the NDVI
values for the seagrass pixels (Lu & Cho 2011).

3.3 Seagrass classification using water corrected image

3.3.1 Ground truth data collection
Several hundred ground data points were collected over seagrass beds in Redfish Bay,
Texas, in the summer of 2008 (June – July). Seagrass species makeup, water depth,
vegetation percent coverage, and bottom substrate type were recorded at each site. Site
Remote Sensing of Submerged Aquatic Vegetation 303

Fig. 2. The original (left) and water-corrected (right) airborne AISA Eagle hyperspectral
image at 741 nm obtained over seagrass beds in Redfish Bay, TX in 2008.
location was recorded to accuracies within 1 m using a Real Time Kinetic (RTK) GPS. When
necessary, the preselected random sites were shifted to avoid dry or unreachable locations.
The field collected data were entered into a spatial database along with descriptive
attributes to help determine which class each sampling site would be assigned to. Data
points were then randomly divided into training or accuracy assessment points.

3.3.2 Image processing and vector classification

Image data were obtained in 63 bands of the AISA Eagle Hyperspectral sensor over the
seagrass beds in October 2008 and corrected for atmospheric effects. Since selection of the
proper bands for analysis helps reduce noise introduction and processing burden (Borges et
al. 2007), several selection techniques were used within this project, including Principal
Component Analysis and regression analysis. Ultimately, seven bands recommended by
Fyfe (2003) were used. To reduce noise, these were again reduced to 5 bands, as
recommended by Cho et al. (2009).
Image segmentation is performed prior to image classification. Segmentation groups like
pixels into homogenous areas. Initially, an unsupervised classification using ISODATA
(Iterative Self Organized Data Analysis Technique A) (De Alwis et al., 2007) was performed.
After the initial image classification, each segmented vector in the output was assigned to a
seagrass/substrate class (i.e. Thalassia testudinum, Halodule wrightii, Ruppia maritima, Mixed
Beds, Bare, or Unclassified). Those which contain only one type of point ('Halodule') were
considered to be finally classified. Those classified as mixed or unknown were removed
from the classified vector set, a mask created of their spatial footprint, and the entire
304 Remote Sensing – Applications

classification procedure re-run on the image for that footprint area only. When it became
impossible to further classify the image by this method, a supervised classification was
performed, using a selection of training points as training data, which produced
monospecific vectors. The same mixed-method classification procedure was performed to
the image data after the water correction algorithm was applied.

3.3.3 Improved accuracy assessment

The water correction algorithm improved the classification accuracy results in an image
subset from an overall accuracy of 28% to approximately 36%. Identification of the species
Halodule wrightii improved from 33% user's accuracy to almost 78%, and Thalassia testudimun
from 0% users accuarcy to almost 17%. Although these numbers appear to be somewhat
low, several factors must be considered: this analysis only used a subset of the imagery,
which allowed the area analysed to have less variation, but there was also less training and
accuracy assessment data to work with.
Using the full set of imagery and the combination of supervised and unsupervised
classification, results have improved considerably; we have achieved overall accuracies of
over 60%. In addition, we have calculated seagrass presence/absence using the complete
corrected imagery, achieving overall accuracies of over 95%. With the future addition of in
situ turbidity and bathymetric data, these accuracies should continue to improve.

4. Current efforts in developing water correction module and graphical user

interface
We have continued improving the algorithm by including turbidity (measured in NTU,
Nephelometric Turbidity Unit) as an additional function. In addition, the water correction
algorithm is currently being implemented as a module that can be called from the ENVI
(ENvironment for Visualizing Images, ITT Visual Information Solutions)’s programming
environment using a Graphical User Interface (GUI) (Gaye et al. 2011). Under the current
module and the GUI, users are able to select water depth (0 – 2.0 m) and turbidity (0-20
NTU) using slide bars, for which (a) given hyperspectral band(s) can be corrected. The
corrected reflectance can be generated and compared to the original one at either a given
pixel, within a small subset; and the original and corrected images can be displayed.

5. Conclusion
Remote sensing has significantly advanced spatial analyses on terrestrial vegetation for
various fields of science. However, mapping of benthic vegetation or submerged aquatic
vegetation (SAV), using remotely sensed data is complicated due to several factors
including atmospheric interferences, variability in water depth and bottom albedo, and
water column attenuation by scattering and absorption. Hence, correction for the
atmospheric and the overlying water column effects is necessary to retrieve any quantitative
information for SAV from satellite and airborne images, especially when using
hyperspectral data. Significant misclassification of the SAV often occurs due to the lack of
information on in situ water depths and water column optical properties. Most of the
currently available radiative transfer models only work well when applied to mapping of
benthic features in relatively clear aquatic environments, but they do not correct for strong
Remote Sensing of Submerged Aquatic Vegetation 305

water absorption of the near infrared energy. The fluctuating water depths, high amounts of
suspended particles and colored dissolved organic matter in shallow littoral zones make it
even more challenging to map benthic vegetation using remotely sensed data. A new water-
depth correction algorithm was developed conceptually, and calibrated and validated using
experimental and field data. The effects of the overlying water column on upwelling
hyperspectral signals were modeled by empirically separating the energy absorbed and
scattered by the water using data collected through a series of controlled experiments. The
empirically driven algorithm significantly restored the vegetation signals, especially in the
NIR region. Due to the restored NIR reflectance, which serves as the primary cue for
discriminating SAV from other substrates, use of the water corrected airborne data
increased the NDVI values for the SAV pixels and also improved the seagrass classification
accuracy. Our continuing efforts to incorporate turbidity and CDOM into the algorithm, in
developing a graphical user interface, and in implementing the algorithm into a module that
can be called from commercially available image processing software promise a user-
friendly application and wide use of the algorithm in the near future.

6. Acknowledgment
The work was supported by grants from the National Geospatial-Intelligence Agency (Grant
No. HM1582-07-1-2005 and HM1582-08-1-0049) and National Oceanic and Atmospheric
Administration-Environmental Cooperative Sciences Center (Grant No.NA17AE1626).

7. References
Ackleson, S. G., & Klemas, V. (1987). Remote Sensing of Submerged Aquatic Vegetation in
Lower Chesapeake Bay: A Comparison of Landsat MSS to TM imagery. Remote
Sensing of Environment, Vol.22, No.2, pp. 235−248, ISSN 0034-4257
Beget, M. & Di Bella, C. (2007). Flooding: The Effects of Water Depth on Spectral Response of
Grass Canopies. Journal of Hydrology, Vol.335, No.3-4, pp. 285-294. ISSN 0022-1694
Borges, J.; Marçal, A. & Dias, J. (2007). Evaluation of feature extraction and reduction
methods for hyperspectral images. In: New Developments and Challenges in Remote
Sensing, Bochenek, Z., pp. 265-284, Millpress, ISBN 978-90-5966-053-3, Rotterdam
Brando, V.E.; Anstee, J.M.; Wettle, M.; Dekker, A.G.; Phinn, S.R. & Roelfsema, C. (2009). A
Physics Based Retrieval and Quality Assessment of Bathymetry from Suboptimal
Hyperspectral Data. Remote Sensing of Environment, Vol.113, No.4, pp. 755-770, ISSN
0034-4257
Bukata, R.; Jerome, J.; Kondratyev, K. & Pozndyakov, D. (1995). Optical Properties and Remote
Sensing of Inland Coastal Waters. New York: CRC Press.ISBN 978-0849347542 Cho,
H.J. & Lu, D. (2010). A Water-depth Correction Algorithm for Submerged Vegetation
Spectra. Remote Sensing Letters, Vol.1, No.1, pp. 29-35, ISSN 2150-7058
Cho, H.J., & Lu, D. (2010). A Water-depth Correction Algorithm for Submerged Vegetation
Spectra. Remote Sensing Letters, Vol.1, No.1, pp. 29-35
Cho, H.J.; Lu, D. & Washington, M. (2009). Water Correction Algorithm Application for
Underwater Vegetation Signal Improvement. Proceedings of the 2009 MS Water
Resource Conference, pp. 154-157
Cho, H.J.; Kirui, P. & Natarajan, H. (2008). Test of Multi-spectral Vegetation Index for
Floating and Canopy-forming Submerged Plants, International Journal of
Envirnmental research and Public Health, Vol.5, No.5, pp. 447-483, ISSN 1661-7827
306 Remote Sensing – Applications

Cho, H. (2007). Depth-variant Spectral Characteristics of Submerged Aquatic Vegetation

(SAV) detected by Landsat 7 ETM+. International Journal of Remote Sensing, Vo.28,
No.7, pp. 1455-1467, ISSN 1366-5901
Ciraolo, G.; Cox, E.; La Loggia, G. & Maltese, A. (2006). The Classification of Submerged
Vegetation Using Hyperspectral MIVIS Data. Annals of Geophysics, Vol. 49, No.1,
pp. 287-294, ISSN 2037-416X
De Alwis, D.; Easton, Z., Dahlke, H., Philpot, W., & Steenhuis, T. (2007). Unsupervised
Classification of Saturated Areas using a Time Series of Remotely Sensed Images.
Hydrology and Earth System Sciences Vol 11, No. 511, pp. 1609-1620, ISSN 1027-5606
Dobson, E. & Dustan, P. (2000). The Use of Satellite Imagery for Detection of Shifts in Coral
Reef Communities. Proceedings, American Society for Photogrammetry and Remote
Sensing, Washington, DC, USA, May 22-26, 2002, (CD-ROM)
du Plessis, W.P. (1999). Linear Regression Relationships between NDVI, Vegetation and
Rainfall in Etosha National Park, Namibia. Journal of Arid Environments, Vol.42,
No.4, pp. 235-260, ISSN 0140-1963
Fang J.Y.; Piao S.L.; Tang Z.Y.; & Peng J.W. (2001). Interannual Variability in Net Primary
Production and Precipitation. Science, Vol.29, No.5526, pp. 293:1723, ISSN 0036-8075
Ferguson, R.L.; & Korfmacher, K (1997). Remote Sensing and GIS Analysis of Seagrass
Meadows in North Carolina, USA. Aquatic Botany, Vol.58, No. 3-4, pp. 241-258,
ISSN 0304-3770
Findlay, S.E.G.; Nieder, W.C.; Blair, E.A. & Fischer, D.T. (2006). Multi-scale Controls on
Water Quality Effects of Submerged Aquatic Vegetation in the Tidal Freshwater
Hudson River. Ecosystems, Vol.9, No.1, pp. 84-96, ISSN 1432-9840
Finkbeiner, M.; Stevenson, B., & Seaman, R. (March 2001). Guidance for Benthic Habitat
Mapping: An Aerial Photographic Approach. U.S. NOAA Coastal Services Center.
http://www.csc.noaa.gov/benthic/mapping/pdf/bhmguide.pdf
Fyfe, S. (2003). Spatial and Temporal Variation in Spectral Reflectance: Are Seagrass Species
Spectrally Distinct? Limnology and Oceanography Vol. 48, No. 1, pp. 464-479, ISSN
1939-5590
Gaye, G.; Kim, H.; Cho, H.J. (2011). Utilizing the Significant Spectral Bands for Image
Outputs for SAV from the Water Effect Correction Module. Proceeding of the 2011
WORLDCOMP (CD-ROM)
Geerken, R.; Zaitchik, B.; & Evans, J.P. (2005). Classifying Rangeland Vegetation Type and
Fractional Cover of Semi-arid and Arid Vegetation Covers from NDVI Time-series.
International Journal of Remote Sensing, Vol.26 No.24, pp. 5535−5554, ISSN 1366-5901
Giri, C.; Pengra, B.; Zhu, Z.; Singh, A. & Tieszen, L. (2007). Monitoring Mangrove Forest
Dynamics of the Sundarbans in Bangladesh and India Using Multitemporal
Satellite Data from 1973-2000. Estuarine, Coastal and Shelf Science, Vol. 73, No.1-2, pp.
91-100. ISSN 0272-7714
Gordon, H.R. & Brown, O.B. (1974). Influence of Bottom Albedo on the Diffuse Reflectance
of a Flat Homogeneous Ocean. Applied Optics, Vol.13, No.9, pp. 2153-2159, ISSN
1559-128X
Govender, M.; Chetty, K. & Bulcock, H. (2009). A Review of Hyperspectral Remote Sensing
and Its Application in Vegetation and Water Resource Studies. Water SA Vol. 33,
pp. 145-152, ISSN 1816-7950
Goward, S.N.; & Huemmrich, K.F. (1992). Vegetation Canopy PAR Absorptance and the
Normalized Difference Vegetation Index: An Asssessment Using the SAIL
model. Remote Sensing of Environment, Vol.39, No.2, pp. 119-140, ISSN 0034-4257
Remote Sensing of Submerged Aquatic Vegetation 307

Han, L. & Rundquist, D.C. (2003). The Spectral Responses of Ceratophyllum demersum at
Varying Depths in an Experimental Tank. International Journal of Remote Sensing,
Vol. 24, pp. 859-864, ISSN 1366-5901
Holden, H. & LeDrew, E. (2001). Effects of the Water Column on Hyperspectral Reflectance
of Submerged Coral Reef Features. Bulletin of Marine Science, Vol.69, No.2, pp. 685-
699, ISSN 0007-4977
Holden, H. & LeDrew, E. (2002). Measuring and Modeling Water Column Effects on
Hyperspectral Reflectance in a Coral Reef Environment. Remote Sensing of
Environment, Vol. 81, No. 2-3, pp. 300-308, ISSN 0034-4257
Jackson, T.J.; Chen, D.; Cosh, M.; Li, F.; Anderson, M.; Walthall, C.; Doriaswamy, P. & Hunt,
E.R. (2004). Vegetation Water Content Mapping Using Landsat Data Derived
Normalized Difference Water Index for Corn and Soybeans. Remote Sensing Of
Environment, Vol.92, No..3 pp. 475−482, ISSN 0034-4257
Jensen, J.R. (2000). Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice
Hall, Inc. Upper Saddle River, NJ 07458, ISBN-13: 978-0131453610, Upper Saddle
River, NJ
Jin, X. (2001). Technologies of Lake Eutrophication Control management. Beijing: Chemistry
Industry Press.
Khan, M.A.; Fadlallah, Y.H. & AL-Hinai, K.G. (1992). Thematic Mapping of Subtidal Coastal
Habitats in the Western Arabian Gulf Using Landsat TM Data - Abu Ali Bay, Saudi
Arabia. International Journal of Remote Sensing. Vol.13, No.4, pp. 605 – 614, ISSN
1366-5901
Kirk, J.T.O. (1994). Light and Photosynthesis in Aquatic Ecosystems. Cambridge University
Press, ISBN 978-0-521-15175-7, Cambridge, UK
Kutser, T. (2004). Quantitative Detection of Chlorophyll in Cyanobacterial Blooms by
Satellite Remote Sensing. Limnology and Oceanography Vol.49, No. 6, pp. 2179-2189,
ISSN 1939-5590
Kutser, T.; Vahtmae, E. & Praks, J. (2009). A Sun Glint Correction Method for Hyperspectral
Imagery Containing Areas with Non-negligible Water Leaving NIR Signal. Remote
Sensing of Environment, Vol.113, No.110, pp. 2267-2274, ISSN 0034-4257
Lillesand, T.M., Kiefer, R.W., and Chipman, J.W. (2008). Remote Sensing and Image
Interpretation. John Wiley and Sons, Inc., 111 River Street, Hoboken: NJ , ISBN-13
978-0470052457
Lu, D. & Cho, H.J. (2011). An Improved Water-depth Correction Algorithm for Seagrass
Mapping Using Hyperspectral Data. Remote Sensing Letters, Vol.2, No.2, pp. 91-97,
ISSN 2150-7058
Lyzenga, D. (1981). Remote Sensing of Bottom Reflectance and Water Attenuation
Parameters in Shallow Water Using Aircraft and Landsat data. International Journal
of Remote Sensing, Vol.2, No.1, pp. 71-82, ISSN 1366-5901
Maeder, J.; Narumalani, S.; Rundquist, D.C.; Perk, R.L.; Schalles, J.; Hutchins, K. & Keck. J.
(2002). Classifying and Mapping General Coral-reef Structure Using Ikonos Data.
Photogrammetric Engineering & Remote Sensing, Vol.68, No.12, pp. 1297-1305, ISSN
0099-1112
Marshall, T., & Lee, P. 1994. An Inexpensive and Lightweight Sampler for the Rapid
Collection of Aquatic Macrophytes. Journal of Aquatic Plant Management, Vol.32, pp.
77-79, ISSN 0146-6623
Matsunaga, T. & Kayanne, H. (1997). Observation of Coral Reefs on Ishigaki Island, Japan,
Using Landsat TM Images and Aerial Photographs, Proceedings of the Fourth
308 Remote Sensing – Applications

International Conference on Remote Sensing for Marine and Coastal Environments, Vol. I,
pp. 657-666, ISSN 1066-3711, Orlando, Florida, USA, March 17-19, 1997
Michalek, J.; Wagner, T.; Luczkovich, J. & Stoffle, R. (1993). Multispectral Change Vector
Analysis for Monitoring Coastal Marine Environments. Photogrammetric Engineering
and Remote Sensing, Vol. 59, No.3, pp. 381-384, ISNN 0099-1112
Mishra, D., Narumalani, S.; Rundquist, D. & Lawson, M. (2005). High Resolution Ocean
Color Remote Sensing of Benthic Habitats: A Case Study at the Roatan Island,
Honduras. IEEE Transactions in Geosciences and Remote Sensing, Vol. 43, No.7, pp.
1592-1604, ISSN 0196-2892
Mishra, D., Narumalani, S.; Rundquist, D. & Lawson, M. (2006). Benthic Habitat Mapping in
Tropical Marine Environments Using QuickBird Imagery. Photogrammetric
Engineering and Remote Sensing, Vol.72, No.9, pp. 1037-1048, ISSN 0099-1112
Mishra, D., Narumalani, S.; Rundquist, D.; Lawson, M. & Perk, R. (2007). Enhancing the
Detection and Classification of Coral Reef and Associated Benthic Habitats: A
Hyperspectral Remote Sensing Approach. Journal of Geophysical Research, Vol.112,
C08014, doi:10.1029/2006JC003892
Mumby, P.; Clark, C.; Green, E. & Edwards, A. (1998). Benefits of Water Column Correction
and Contextual Editing for Mapping Coral Reefs. International Journal of Remote
Sensing, Vol.19, No.1, pp. 203-210, ISSN 1366-6901
Pasqualini, V.; Pergent-Martini, C.; Pergent, G.; Agreil, M.; Skoufas, G.; Sourbes, L. & Tsirika,
A. (2005). Use of SPOT 5 for Mapping Seagrasses: An Application to Posidonia
oceanica. Remote Sensing of Environment, Vol.94, No.1, pp. 39-45, ISSN 0034-4257
Phinn, S.; Roelfsema, C.; Dekker, A.; Brando, V. & Anstee, J. (2008). Mapping Seagrass Species,
Cover and Biomass in Shallow Waters: An Assessment of Satellite Multi-spectral and
Airborne Hyper-spectral Imaging Systems in Moreton Bay (Australia). Remote Sensing
of Environment 2008, Vol.112, No.8, pp. 3413-3425, ISSN ISSN 0034-4257
Rouse, J.; Haas, R.; Schell, J. & Deering, D. (1973). Monitoring Vegetation Systems in the
Great Plains with ERTS. 3rd. ERTS Symposium, NASA SP-351 I, 309-317
Spitzer, D. & Dirks, R. (1987). Solar Radiance Distribution in Deep Nataural Waters Including
Fluorescence Effects. Applied Optics, Vol. 26, Issue 12, pp. 2427-2430, ISSN 1559-128X
Walters, B.; Ronnback, P., Kovacs, J., Crona, B., Hussain, S., Badola, R., Primavera, J.H.,
Barbier, E. & Dahdouh-Guebas, F. (2008). Ethnobiology, Socio-economics and
Adaptive Management of Mangroves: A Review. Aquatic Botany, Vol. 89, No.2, pp.
220–236, ISSN 0304-3770
Washington, M.; Kirui, P.; Cho, H. & Wafo-Soh, C. (2011). Data-driven Correction for Light
Attenuation in Shallow Waters. Remote Sensing Letters, Vol.3, No.4, pp. 335-342,
ISSN 2150-7058
Wu, C.; Niu, Z.; Tang, Q. & Huang, W. (2008). Estimating Chlorophyll Content from
Hyperspectral Vegetation Indices: Modeling and Validation. Agricultural and Forest
Meteorology, Vol. 48, No. 8-9, pp. 1230-1241, ISSN 0168-1923
Zimmerman, R. & Mobley, C. (1997). Radiative Transfer within Seagrass Canopies: Impact
on Carbon Budgets and Light Requirements. Proceedings of SPIE Ocean optics XIII,
2963: 331-336
 13

Remote Sensing and Environmental

Sensitivity for Oil Spill in the Amazon, Brazil
Milena Andrade¹ and Claudio Szlafsztein²
¹Federal University of Pará, Amazon Advance Studies (NAEA)
²Federal University of Pará, Center of Environment (NUMA)
Brazil

1. Introduction
The use of remote sensing has become a fundamental tool for the identification and analysis
of different types of risks in coastal zones. The numerous and, in some cases, recent
incidents of oil spills have encouraged companies and government agencies to improve
methods, both anticipatory and corrective, to minimize damages. The term ‘risk’ denotes the
possibility that adverse effects may occur as a result of natural events or human activities
(Kates et al., 1985). Risk is defined as an association between the hazard´s characteristics
(e.g. frequency, magnitude and location) and the vulnerability of affected human
populations, environment and infrastructure (Wisner et al., 2004). Risk can be classified by
their origin, such as natural, social, or technological (Renn, 2008). Oil spills are an example
of the last category, and the coastal areas are one of the most impacted. Environmental
sensitivity to oil impacts can be defined through the coastal Environmental Sensitivity Index
(ESI), which considers: (i) the geomorphologic aspects such as type and slope of coastline
and the degree of exposure to the energy of waves and tides; (ii) oil sensitive biological
resources; and (iii) the socio-economic activities that can be affected by oil spills (Gundlach
& Hayes, 1978; Dutrieux et al., 2000).
In Brazil, environmental sensitivity mapping has been carried out under the law 9966/2000,
which gave the Ministry of the Environment (Climate Change and Environmental Quality
Secretary) responsibility to identify, locate and define the boundaries of ecologically
sensitive areas with respect to the spill of oil and other dangerous substances in waters
within national jurisdiction. This way, based on PETROBRAS (2002) and NOAA (2002), the
specifications and technical standards for preparing environmental sensitivity maps for oil
spills in coastal and marine zones was elaborated upon (MMA, 2002). Such environmental
sensitivity maps provide information in an easy format being useful to determine priorities
to impact protection and mitigation. Identification and mapping is developed at three levels:
(i) Strategic (1:500,000 for the entire area of a hydrographical basin); (ii) Tactical (1:150,000
for the entire coastline mapped); and (iii) Operational (up to 1:50,000 for a high-
risk/sensitivity areas). Each of these mapping scales uses specific tools for remote sensing
and GIS tools.
The Amazonian coastal zone extends along ~2250 km, not including the several inlets,
islands and small estuaries, which punctuate the coastline (Souza Filho et al., 2005a). This
310 Remote Sensing – Applications

coastal zone is placed in the context of the tropical humid regions, in a low-lying area with
active processes of erosion, sedimentation and neotectonics. Also, it is marked by a great
hydrologic influence; in a meso- to macrotidal area (Souza Filho, 2005). It is a high-density
drainage network, in which the Amazon River discharges a volume of water of 6.3 trillion
m³/year and of sediment estimated at 1.2 billion tons/year (Meade et al., 1985).
Such environmental characteristics are responsible for the development of an extensive mud
plain and mangrove area which is located in three States (Amapá, Pará and Maranhão), is
approximately 8,386 km² wide, and contains 80% of all mangroves in Brazil (Herz, 1991).
Where macrotides are present, the area of a flooded mangrove may extend for up to 30 km
inland, and the estuaries themselves as much as 80 km (Souza Filho, 2005) (Figure 1). These
extensive mud and mangroves plains are considered to be one of the most sensitive areas to
oil spills. Also, these mangroves are along national and international ships routes.
Transportation and storage are mainly responsible for oil spills in Amazonian coastal zone,
since there is no expressive exploration. In 2001, in the state of Pará, approximately 1900
tons of oil sank near the Port of Vila do Conde (Berredo et al., 2001).

Fig. 1. Amazonian coastal zone in radar SRTM representation (source: modified from Souza
Filho et al., 2005a)

In this sense, researches from Federal University of Pará have been working on several
projects since 2001 aiming to study the Amazonian coastline and the impact of oil spills on
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 311

the environment. Therefore, from 2004 to 2010, a large group of scientists were grouped in
PIATAM-Mar project “Potential Environmental Impacts and Risks of the Oil and Gas
Industry”, financially supported by PETROBRAS1, to map and analyse the vulnerability of
the Amazonian coastline oil related disasters. Since 2012, the project “Elaboration of
Environmental Sensitivity Maps (SAO maps) for oil spills in Pará-Maranhão and
Barreirinhas Basins”, founded by the National Research Council of Brazil (CNPQ) has been
developed with similar objectives.
Remote sensing and GIS are principal tools aimed to enhance basic socio and environmental
knowledge about Amazonian coast. Maps were prepared in strategic and tactical scales
through the use of digital elevation models derived from the SRTM (Shuttle Radar
Topographic Mission) and optical sensors data (Cohen & Lara, 2003; Souza Filho & El
Robrini, 2000; Souza Filho, 2005; Szlafsztein & Sterr, 2007; Silva et al., 2009), synthetic
aperture radar (SAR) data (Souza Filho & Paradella, 2002 and 2005; Costa 2004; Souza Filho
et al., 2005b; Silva et al., 2009), and the combination of some of them (Souza Filho &
Paradella, 2005; Gonçalves et al., 2009; Rodrigues & Souza Filho, 2011).
Oil spill environmental sensitivity maps, adapted to the peculiarities of the Amazonian region
(Souza Filho et al., 2004) were drawn at tactical scales through the use of Radarsat and Landsat
sensors (Gonçalves et al., 2009; Teixeira & Souza Filho 2009; Boulhosa & Souza Filho, 2009),
and operational scale through the use of High Resolution remote sensing (Andrade et al., 2010;
Rodrigues & Szlafsztein 2010; Andrade et al., 2009). Over the past decade were reached
advances in identification and assessment of sensitivity through spatial maps, the impacts to
oil spill analyses, and oil spill risk in Amazonian coastal. The goal of this book chapter is to
present a review of the oil spills environmental sensitivity mapping activities using remote
sensing and GIS tools in the Amazonian coastal zone of Brazil.

2. Remote sensing and coastal environmental sensitivity for oil spill

2.1 Remote sensing
Remote sensing tools are essential for the construction of maps. These tools help in the
precise delimitation of coastlines and specific landforms. The selection of appropriate
remote sensing data and applicable digital image processing techniques involves a
compromise between costs and mapping capabilities, including coverage area, and spatial
resolution (Green, 2000).
For risk maps, remote sensing are fundamental. Risk appears in a broader context in
humans transform of the natural into a cultural environment, with the aim of improving
living conditions and serving human wants and needs (Turner et al., 1990). There are several
sources of hazards to the environment and to society, some of them originated in human
activities (Smith & Petley, 2008).
Oil spills are an example of this technological risk. Information and detection about oil spills
can be collected through remote sensing tools for prevention planning, as well as
river/ocean pollution monitoring and restoration. Some reviews of the use of remote
sensing and oil spills including Brekke & Solberg (2005) and Fingas & Brown (2000).

1 PETROBRAS is the large oil company in Brazil

312 Remote Sensing – Applications

For a coastal environment, remote sensors can provide information about the physical
characteristics of the shoreline, coastal ecosystems dynamics, water quality, and land
use/occupation. This information could be mapped at different scales generating
cartographic products using all types of sensors and specific digital image processing
(Jensen, 1996). Sensors can provide timely and valuable information about oil spills,
including the location and extent, thickness distribution, and oil type in order to estimate
environmental damage, take appropriate response activities, and to assist in clean-up
operations for oil spill contingency planning (Grüner, 1991).
The most common sensors utilized to detect oil spills and to map coastal environments are:
optical (visible, infrared sensors and ultraviolet sensors) or radar. Both types of sensor may
be acquired at terrestrial, sub-orbital or orbital levels. At terrestrial level, both still and video
cameras are commonly used. At the sub-orbital level (or airborne remote sensing), airplanes
is the most commonly utilized platform. At the orbital level, satellites are usually used as a
platform for sensors. Satellite differs from airborne remote sensing due to timing and
frequency of the data collection, the demand of good climate conditions and the time
required for processing the dataset (Jha et al., 2008). Aiming to compare sensors, a brief
description is given in Table 1.

2.1.1 Optical
Optical sensors can be composed by three bands in the electromagnetic spectrum. These
sensors are usually composed by multispectral bands in visible and infrared intervals from
the electromagnetic spectrum. In the visible region (350 to 750 nm), oil has a higher surface
reflectance than water, but also shows limited nonspecific absorption tendencies (Jha et al.,
2008). Instruments such as cameras, films and spectrometers are optical techniques for
remote sensing with the benefit of low cost. Normally, visible sensors cannot operate at
night as they depend on the reflectance of sunlight, but, in the case of oil spills they can be
used to create environmental and logistic maps of the coast to subsidize field trips and first
risk management decisions. The infrared sensors are at the 0,7-14 μm intervals in the
electromagnetic spectrum. Solar radiation is partially absorbed and emitted as thermal
energy by oil. This is thermal energy concentrated in the thermal infrared region with a
distinct spectral signature; water has a higher emissivity (Salisbury et al., 1993). Infrared
sensors can provide information about the relative thickness of oil slicks, but these sensors
are unable to detect emulsions of oil in water when oil is diluted to 70% water (Fingas &
Brown, 1997). Infrared is reasonably inexpensive, but has limitations related to false positive
results generated by weeds and shorelines (Fingas & Brown, 2000).
Ultraviolet sensor scanners capture ultraviolet radiation (0,003 – 0,38 μm) reflected by the
sea surface for detecting oil spills. Oil is more reflective than water in the ultraviolet region.
Limitations of this sensor are related to undetected information greater than 10 microns and
false images produced by such hindrances as wind slicks, sun glints, and biogenic material
(Grüner, 1991).

2.1.2 Radar
Radar is an active sensor (not dependent on electromagnetic radiation from the sun) and
operates in a radio wave region (1m – 104m). Radar sensors can have two principal
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 313

Spatial Over-pass
Imagery
Sensors Platform resolution Frequency Application
area
(m) (days)

Spaceborne 3, 35 and Identify large offshore

SAR ERS-2 30 100km
176 days spills and coastal
environments – Strategic
SAR RADARSAT-1 8-100 24 days 45-500km planning and monitoring
RADAR

SLAR Airplane 10-50 As required 60-80km Detect and identify the

polluter, the extent and
type of oil spill and the
Airbone

cleaning necessity;
SAR Airplane 1-10 As required - Environmental mapping –
Strategic and tactical
planning
MSS,TM, Landsat 5 Detect oil spill if the
183-
ETM, Landsat 6 15-120 16 days weather conditions are
185km
ETM + Landsat 7 good; can discriminate false
positives; identify and
Spot-2 60x60km mapping environments –
HRV 10-20 26 days Strategic and tactical
Spot-3 /100km
planning
Detect oil spill if the
weather conditions are
Cbers-1;
CCD 20 26 days 113 km good; identify and mapping
Cbers-2
Spaceborne

environments – Strategic
and tactical planning
Detect oil spill if the
OPTICAL

weather conditions are

Cbers-1; good; capable to detect
IRMSS 80-160 26 days 120 km
Cbers-2 thermal surface
differentiations - Strategic
and tactical planning
Detect oil spill if the
weather conditions are
Cbers-1;
WFI 260 5 days 890 km good; monitoring; identify
Cbers-2
and mapping environments
– Strategic planning
Video Altitude As Oil spill and coastal
Airplane -
camera Dependent required environmental
Airbone

documentation. The
Still Altitude As infrared sensor for measure
Airplane - the thickness of oil slicks –
camera Dependent required
Operational planning

Table 1. Characteristics of some existing sensors for oil spill management applications.
314 Remote Sensing – Applications

instruments: Side-Looking Airborne Radar (SLAR) and Synthetic Aperture Radar (SAR).
Radar is a very powerful and useful sensor for searching large areas, observing oceans at
night, and capturing images during cloudy weather conditions. The presence of an oil spill
can be detected without thickness estimation or oil type recognition. In the radar image, the
leak appears as a dark area in contrast to the bright image of the ocean because radar waves
are reflected by capillary waves on the ocean (Brown et al., 2003). For a coastal environment,
mapping SAR is already considered to be a powerful tool for geomorphologic mapping,
providing relevant information about the emergence and submergence of the coast (Souza
Filho et al., 2009a).
SLAR is an old technology predominantly used for airborne remote sensing (Fingas &
Brown, 2000). Airborne surveillance is limited by high costs and is less efficient for wide
area observation due to its limited coverage. SAR has greater spatial range and resolution
than the SLAR because it uses the forward motion of the aircraft to synthesize a very long
antenna, thereby achieving very good spatial resolution, at the expense of sophisticated
electronic processing (Mastin et al., 1994). SAR can be used to provide an initial warning
because aircrafts are more suitable to identify the polluter, the extent, and the type of
spill.
For large scale oil spill detection, satellite platforms, including ERS-1 and -2, Radarsat, and
JERS-1, are commonly used for large scales oil spills (Fingas & Brown, 2005). Radar
satellites, including ERS-1 and -2, Radarsat, and JERS-1, have been useful for mapping
known large offshore spills (Biegert et al., 1997). On the other hand, optical satellite imagery
does not offer much potential for oil spill detection (Fingas & Brown, 2000). However, to
map coastal environments, geomorphology and its sensitivity, multi-sensor data fusion such
as optical and radar has proved to be a successful tool (Souza Filho et al., 2009b).

2.2 Coastal environmental sensitivity to oil spills

The inter-relationships involving natural resources and human societies have led to a
concentration of human activities, services and survival strategies in the coastal
environment (Viles & Spencer, 1995; Muehe & Neves, 1995; Pernetta & Elder, 1992). The
unique natural geodynamics, the highly productive and extremely diverse biological
systems extending from coastal lands to deep water regions (Malthus & Mumby, 2003), the
growing land use changes and the pressure on natural resources (MEA, 2005) transform the
coastal zone into a conflict area. Oil exploration, transportation and storage have increased
the technological risk in this zone.
Areas neighboring major ports (environmental and human populated) may be affected by
oil transportation, tank cleaning and oil storage procedures in a port area (Noernberg &
Lana, 2002). One of the initial concerns about oil spills result in a necessity for the
construction of maps that indicate which type of environment and human resources will be
affected. In the mid 1970s, scientists from the National Oceanic and Atmospheric
Administration (NOAA) and the American Coast Guard of the United States began to study
and classify the sensitivity of coastal environments to oil spill.
This classification was based, initially, on the vulnerability index to oil spills proposed by
Gundlach & Hayes (1978). Coastal area is segmented considering environmental and
geomorphologic characteristics and then classified using the Vulnerability Index, scaled
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 315

from 1 (low) to 10 (high). This Vulnerability Index became the standard for coastal
management, planning and research about the effects of oil spills on different types of
coastline. Over time, this index evolved and was modified, leading eventually to the
development of the Environmental Sensitivity Index (ESI).
The ESI should be represented cartographically as maps in different scales for different
goals. The first ESI map was produced in 1979, in response to the advance toward the coast
of oil resulting from the blowout of the IXTOC 1 oil-well in the Gulf of Mexico. In the 1980s,
ARPEL produced an innovative ESI atlas for the whole coast of the United States, including
Alaska and the Great Lakes, to be used for the planning of contingency measures in
response to oil spills (NOAA, 2002). From this moment on, ESI maps have been an integral
component of response and contingency planning for oil spills, looking for the protection of
life, the reduction of environmental impacts, and facilitation of the response efforts. These
atlases were integrated by color printed maps on a two dimensional representation of a
three-dimensional world and high production costs.
After the 1990’s, NOAA (2002) standardized output formats and symbols for ESI maps
construction. The basic necessary information is 1) shoreline classification; 2) biological
resources; 3) human-use resources. The shoreline classification scheme is based on an
understanding of the physical and biological characteristics of the shoreline environment.
Relationships among physical processes, exposure to wave and tidal energy, slope,
substrate type (i.e. grain size, mobility, penetration and/or burial, and mobility), and
associated biota produce specific geomorphic/ecologic shoreline types. Shoreline
classification helps to identify oil spill origin and impacts and the best cleanup methods
for a specific shoreline type. The sensitivity ranking was developed for the estuarine
settings and is slightly modified for lakes and rivers. The human use resources relate to
specific, valuable specific areas because of their use, such as beaches, parks and protected
marine areas, water intakes, fisheries, tourism, economic sectors, and archaeological sites.
The biological resources include the study and maps of oil-sensitive biological and
ecological resources.

3. Remote sensing and coastal environmental sensitivity in Brazil

Brazil has an expansive coastline through the equatorial region to the subtropical latitude of
the south hemisphere. The length is approximately 8.500 km with 17 of the 26 states of the
country lying on the coast of the Atlantic Ocean. The Brazilian coast is defined by the
National Plan for Coastal Management (law 7661/1988), as the geographic space where
there are air, sea and land interacts, which includes renewable and non-renewable resources
along a maritime and terrestrial border.
A diversity of coastal environments and population densities are found along the Brazilian
coast. Population is higher in state capitals than in the other coastal municipalities.
Environments vary from very productive, such as mangroves, to rocky and artificial man-
made structures. Man-made structures, such as ports are established along the entire
coastline of Brazil (Figure 2).
Ports are high-risk areas, and oil spill monitoring is clearly important there. In 2000, two
large oil spills occurred at Baía de Guanabara (Rio de Janeiro) and Paraná, both resulting
from pipeline ruptures. After these accidents, fundamental changes have been made to
316 Remote Sensing – Applications

Fig. 2. Main ports of the Brazilian coast (IBAMA, 2011).

the environmental policies of Brazil and PETROBRAS (Souza Filho et al., 2009b), in order to
give priority to prevention and mitigation activities.
After these accidents, the prevention and mitigation oil spill impacts became a priority.
The law 9966/2000 was established to regulate the activities of prevention, control and
supervision of pollution caused by oil and other dangerous substances in Brazilian
waters. The Ministry of the Environment has the responsibility of identifying, locating
and defining the boundaries of the ecologically sensitive areas to oil spills. Ecologically
sensitive areas are defined as regions where special attention is needed in order to protect
and preserve the environment from pollution by regulatory and preventive measures
(MMA, 2002). In this context, Araújo et al. (2002) published the document “Basic Manual
for the Elaboration of Maps of Environmental Sensitivity to Oil Spills in the Petrobras
System: Coastal and Estuarine Environments” edited by the Ministry of the Environment
(Climate Change and Quality Environmental Secretary) and based on PETROBRAS (2002)
and NOAA (2002).
This change in attitude was reflected in a large production of oil spill environmental
sensitivity maps for the Brazilian coast (Araújo et al., 2006) (Table 2). These maps are
produced in order to support environmental management and the elaboration and
implementation of contingency plans. The ESI preparation was intensified through Decrees
4136/2002 and 4871/2003.
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 317

Spatial
Author Sensor Method Scale Case study (State)
Resolution
Cananéia-Iguape
Romero, et Aerial Literature review,
High Operational Estuary
al. (2010) photography Field data collection
(São Paulo)
SRTM, Highly precision field
Low Potiguar
Radarsat data collection with
Silva et al. Tactical, Sedimentary Basin
Landsat 7 DGPS, Previous
(2010) Operational (Rio Grande do
Moderate ETM+, Cbers- database research,
Norte)
2, Cbers-2B visual classification
Cantagalo Aerial Visual classification, Tactical, Santos Estuary
High
et al. (2008) photography field data collection Operational (São Paulo)
Automatic Potiguar
Carvalho &
classification, visual Tactical, Sedimentary Basin
Gherardi Moderate Landsat 7 ETM
interpretation, Field Operational (Ceará and Rio
(2008)
data collection Grande do Norte)
Visual interpretation,
Bellotto & Municipality of
Landsat 7 Previous database
Sarolli Moderate Operational Imbatuba
ETM+ research, Field data
(2008) (Santa Catarina)
collection
Moderate Cbers Visual interpretation,
Poletto & Municipality of
Previous database Tactical,
Batista Aerial Ubatuba
High research, Field data Operational
(2008) photography (São Paulo)
collection
Rocha- Southeast and
Visual interpretation,
Oliveira et Moderate Landsat 7 TM+ Operational south area
Field data collection
al. (2008) (Santa Catarina)
Santa Catarina
Visual interpretation,
Silva et al. Island and
Moderate Landsat 7 TM+ field works, literature Operational
(2008) surrounding areas
review
(Santa Catarina)
Landsat 7 Municipalities of
Moderate
ETM+ Itapoa, Barra
Araújo et Visual interpretation, Vellha, Piçarras,
Operational
al. (2007) Aerial Field data collection Itajaí, Balneário
High Camboriú (Santa
photography
Catarina)
Chacaltana Visual interpretation, Vitória Bay
High Ikonos Operational
(2007) Field data collection (Espírito Santo)
Lima et al. Aerial Visual interpretation, São Sebastião
High Operational
(2008) photography Field data collection Island (São Paulo)
Cananéia-Iguape
Wieczorek Aerial Visual interpretation,
High Operational Estuary
et al. (2007) photography Field data collection
(São Paulo)
318 Remote Sensing – Applications

Spatial
Author Sensor Method Scale Case study (State)
Resolution
Landsat 5 TM, Database
Moderate Landsat 7 development,
São Bento,
ETM+ geomorphology;
Galinho
Castro et al. hydrodynamic, waves
Operational Municipalities
(2006) energy, currents
Aerial (Rio Grande do
High direction; slope and
photography Norte)
grain size of profile
beach
Landsat 5 TM, Normalized Difference
Moderate Landsat 7 Vegetation Index,
Souto et al. ETM+ Automatic Ponta Macau (Rio
Operational
(2006) classification, Visual Grande do Norte)
Aerial interpretation, Field
High
photography data collection
Landsat 5 TM,
Landsat 7 Database utilization; Northern coast
Souza, et al. Moderate ETM+, SPOT,
visual interpretation, Tactical (Rio Grande do
(2005) Cbers-2 Field data collection Norte)
High Ikonos
Noernberg Paranaguá
Database access,
& Lana Moderate Landsat TM Operational Estuary (Santa
digital processing
(2002) Catarina)

Table 2. Principal studies of oil spill coastal sensitivity using remote sensing techniques in
order to generate ESI maps in Brazil organized by date (Amazon Region are not included).

In Brazil, the ESI maps were also developed in a cartographic plan at strategic, tactical and
operational scale for the role country. As an initial step, the tools of remote sensing and GIS
are necessary to ESI maps construction and to comprehend differential spread of the
technological risk for the country’s coasts. Mostly moderate and high resolution images
were used to produce these maps.
Moderate resolution images (e.g. RADARSAT-1 and Landsat TM/ETM/ETM+) and SRTM
derived digital elevation models have been used to map the Brazilian coastal zone at
strategic and tactical scales. Studying an oil spill emergency due to a pipeline rupture in
Guanabara Bay (Rio de Janeiro), Bentz and Miranda (2001) found that RADARSAT-1
provided suitable temporal coverage. Once cloud cover, haze and the eight-day revisit
schedule (using both Landsat-5 and -7) prevented Landsat from being used systematically
for oil spill monitoring. In the same case Thematic Mapper (TM) sensor was used to capture
images after the oil spill emergency where a pipeline ruptured (Bentz & Miranda, 2001).
Carvalho & Gherardi (2008) used Landsat 7 ETM+ images to generate land use and land
cover maps, as well as ESI maps in Northeast Brazil, aiming for oil spill contingency
planning and emergency responses. A fusion of multispectral and panchromatic ETM
images via IHS (Intensity-Hue-Saturation) transformation was used. Then socioeconomic
information was inserted using automated and visual image interpretation.
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 319

High resolution images have mostly been used for operational ESI maps production in the
states of São Paulo and Rio Grande do Norte using aerial photography and Ikonos. Visual
interpretation, together with field data collection, has been the principal methodological
procedure. Most areas have mangroves, conservation units and are surrounding by
intensive technological activities.
The methodology, standards and technical specifications for determining coastal sensitivity
follow Araújo et al. (2002). The principal steps for shoreline identification are: 1) Analysis of
the available literature, aerial photographs, maps of the entire area; 2) Aerial reconnaissance
of the entire area and selection of detailed study areas; 3) Mapping of major features in
representative areas; 4) Collection of sediment from the intertidal zone and biologic floral
and faunal groups samples; construction of beach topographic profile; 6) Analysis of the
sediment sample sizes; 7) Data compilation and classification; and 8) Construction of
detailed sensitivity maps.
Colors are used indicate the ESI and symbols to the human and biological resources. Each
number is represented by a color index. Two environments may occur at the same coastal
segment; in that case, both colors of the separated lines should be displayed, one inside and
the other outside. In the case of intertidal zones, for example, the intertidal plain should
display colors according to the differences of sediment sizes to the high tidal line and the
low tidal line. Table 3 compares ESI specification defined by the Ministry of the
Environment (MMA, 2002) with the original defined by NOAA (2002).

ESI
Color NOAA (2002) MMA (2002)
number
Exposed rocky shores and man-made Exposed rocky shores; exposed
1 structures; rocky cliffs with boulder rocky sedimentary cliffs; exposed
talus base solid man-made structures
Exposed medium to high
Exposed wave-cut platforms in
declivity rocky shores; exposed
2 bedrock, scarps and steep slopes in
sandy substrate with medium
clay
declivity
Fine to medium grained sand in
Fine to medium grained sand beaches; dissipative beaches; continuous
3 Scarps and steep slopes in sand; and multiple beach strings;
Tundra cliffs Scarps and steep slopes in sand;
exposed dune field
Coarse-grained sand beaches;
exposed; exposed fine to
medium grained sand
4 Coarse-grained sand beaches
intermediary beaches; sheltered
fine- to medium- grained sand
beaches
Mixed sand and gravel beaches,
5 Mixed sand and gravel beaches coral reefs fragments; vegetated
abrasion platform; sandy reefs
320 Remote Sensing – Applications

ESI
Color NOAA (2002) MMA (2002)
number
Gravel beaches; dendritic
Gravel beaches; Riprap gravel beaches
6 limestone coast; platform with
(cobbles and boulders)
lateritic concretion
Exposed sandy tidal flats; low
7 Exposed tidal flats
tide platform
Sheltered: scarps in bedrock, mud or Sheltered scarps in bedrock
clay, rocky shores (permeable and non permeable);
8 (impermeable/permeable), solid man- Scarps and steep slopes in sand;
made structures, riprap, rocky rubble permeable sheltered man-made
shores; Peat shoreline structures (riprap)
Sheltered tidal flats; Vegetated low Sand tidal flats; sheltered mud
9
banks; hypersaline tidal flats tidal flats; coral reefs
Salt and brackish water marshes;
Delta and vegetated sand bars;
Freshwater marshes; Swamps; Scrub-
10 sheltered wetlands; salt marshes
shrub wetlands: mangroves; Inundated
saline wetlands; mangroves
low-lying tundra
Table 3. ESI comparison between NOAA (2002) and Ministry of the Environment (MMA,
2002) classification.

4. Remote sensing and coastal environmental sensitivity in Amazon

The coastal zone of the Brazilian Amazon is composed by tree states: Amapá, Pará and
Maranhão. According to the IBGE (2011), the total population of Amapá State is 669,526
distributed in 16 municipalities; the state capital, Macapá, concentrates 59% of this
population. The state of Pará has a total population of 7,581,051 distributed in 143
municipalities; Belém comprises 18%. Maranhão state has a total population of 6,574,789
distributed in 217 municipalities; São Luis comprises 15% of this population. The population
density in capital cities is over 100 hab/km², while in other coastal municipalities vary from
10 to 50 hab/km².
Until 21st century most of the coastal zone of the north of Brazil had sectors virtually devoid
of information, or where data was available, it was non-systematized and both temporally
and spatially non-continuous. The most important environmental dataset is related to the
large continuous and well-developed mangroves - Rhizophora mangle, Avicennia germinans
and Laguncularia racemosa (Szlafsztein, 2000). The mangroves have ecological and
socioeconomic importance due to communities’ livelihoods, and they are considered a
protected ecosystem. The main activities are fishing, collecting shrimp and crabs (Andrade
et al., 2010; Andrade et al., 2009).
However, port complexes and industries have been established alongside residential,
protected areas and fishing grounds. For example, in Piatam Mar context the principal ports
chosen to develop oil mapping were “Santana” (State of Amapá); “Itaqui” (State of Maranhão);
“Outeiro”, “Miramar” and “Vila do Conde” (State of Pará). The biological information was
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 321

intensively identified in “Lago Piratuba biological reserve” (Amapá); “Soure extractive

reserve” (Pará) and “Ilha dos Caranguejos Environmental Protection Area” (Maranhão).
According to Souza Filho et al., (2009a) these conservation units work as control areas, given
both their well-preserved conditions and their proximity to transportation routes due to
proximity to protected areas along the ports mentioned above (Figure 3).

Fig. 3. Principal ports and environmental protected areas in the Amazon coast (source:
Souza Filho et al., 2009a).

Oil spills are a potential risk around these port areas which can affect the environment, human
population infrastructure and livelihood, resulting from the transportation process, as well as
tank cleaning and oil storage procedures within the area of the port (Noernberg & Lana, 2002).
To comprehend the oil impact, it is necessary to analyze the coastal Amazonian environment
as a whole. PETROBRAS established and financed nine projects to deal with this subject,
among them, the “Environmental Sensitivity Map to Oil Spill in Guajará Bay (PA)” (2001 –
2003), the “JERS-1, RADARSAT-1 and ALOS/PALSAR application in monitoring and
mapping Amazon coastal environments: an approach for multi-temporal environmental
sensitivity maps to oil spill” (2004 - 2006), PIATAM MAR (2004 - 2010) and currently
“Elaboration of ESI maps for Pará-Maranhão and Barreirinhas basin” (2012 until 2014).
The PIATAM MAR project was implemented in Northern Brazil and was led by the Federal
University of Pará, the Federal University of Rio de Janeiro and PETROBRAS. The general
aims proposed are: the consolidation of a multidisciplinary researcher’s network that are
active in the Amazonian coastal zone; the development of technological tools and
infrastructure to support local monitoring and environmental management; and ESI maps
construction (Souza Filho et al., 2009a).
322 Remote Sensing – Applications

Initially, the researchers of PIATAM MAR project compiled environmental data and other
information available on the Amazon coastal zone. The results are integrated in the book,
“Bibliography of the Amazon Coastal Zone: Brazil” (Souza Filho et al., 2005a). Meanwhile, a
computational database system using the MYSQL language was developed and used as a
basis for the development of a geographic information system called SIGmar.
All of these initial steps support the subsequent aim of PIATAM MAR: the construction of
ESI maps. From 2006 to 2010 socio-economic and environmental data were integrated in ESI
maps. These maps have been developed through the SIGmar and the extensive use of
remote sensing. ESI maps create an operational alternative for the monitoring and mapping
of the Amazonian coastal zone and provide guidelines for the use of the InfoPAE
(Computerized Emergency Action Plan Support) System on the Amazonian coast (Souza
Filho et al., 2009a).
Two considerations should be taken into account when mapping and monitoring oil spills in
the Amazonian coastal environments. First, the unique complex environmental dynamics of
the Amazon basin have demanded an adaptation of ESI classification with values from 1
(low) to 10 (high) sensitivity (Souza Filho et al., 2004) (Table 4). Second, the Amazonian
coast is situated in the intertropical convergence zone (ITCZ) that is located near the equator
and has a broad area of low atmospheric pressure. Therefore, there is a huge cloud cover
between December and May which limits the use of some kinds of sensors.
Coastal ESI mapping for the Amazon uses remote sensing as an indispensable and very
powerful tool. Oil spill and environmental sensitivity to oil spills in the Amazon were
mostly mapped during the PIATAM MAR project. Table 5 shows the most important
scientific results in this study area. The perspectives of the ESI adaptation proposed by
Souza Filho et al. (2004) were extensively used.
The whole Amazon coastal zone was mapped with spatial resolution of 90 m based on the
processing and images mosaics of SRTM images and 30 m of RADARSAT-1 Wide 1 images and
mosaics of JERS-1 SAR. This sensor was chosen given the six months of unfavorable climatic
conditions; radar sensors (Synthetic Aperture Radar – SAR) are used for strategic scale.
On a tactical scale, multi-sensor data fusion between microwaves sensors and optical
sensors are considered to be the most important source of spatial data for geomorphologic
recognition and basic coastline characteristics. The commons sensors fusion are made in
general with low resolution data from RADARSAT-1 Wide 1 and JERS-1 SAR mosaic,
together with moderate spatial resolution data (10–30 km) from Landsat series (MSS, TM
and ETM+) and Cbers-2 images (20m). The multi-fusion of optical (Landsat 5 TM) and radar
(RADARSAT-1) sensors had a particular emphasis on the evaluation of the new hybrid
sensor product combining PCA (Principal Component Analysis) and IHS components. In
areas with little or no data, this fusion method from multi sensors to orbital images, together
with field data are economically efficient and provide a good environment sensitivity
characterization (Rodrigues & Souza Filho, 2011).
Hydrological dynamics with flood area delimitation could be differentiated by the use of
JERS-1, L band (Santos et al., 2009), which is important in a region dominated by different
tidal regimes that can amplify the area affected by oil spills. Methods include visual and
automatic classification leading to good results in identifying widespread occurrence of
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 323

ESI Amazon Coastal Environment

1A Exposed rocky shores
1B Exposed, solid man-made structures
1C Exposed rocky cliffs with boulder talus base
2 Exposed scarps and steep slopes in clay
3A Fine to medium grained sand beaches
3B Scarps and steep slopes in sand
4 Coarse-grained sand beaches
5 Mixed sand and gravel banks and beaches
6 Riprap
7 Exposed tidal flats
8A Sheltered scarps in bedrock, mud, or clay
8B Sheltered, solid man-made structures
8C Sheltered riprap
8D Peat shorelines
9A Sheltered tidal flats
9B Vegetated low banks
9C Hypersaline tidal flats
10A Salt, and brackish-water marshes
10B Freshwater marches, aquatic vegetation
10C Intertidal mangrove
10D Supratidal mangrove
Table 4. ESI shoreline classification for the Amazon Coast, modified by Souza Filho et al.
(2004) based on the proposals of NOAA (2002) and Araújo et al. (2002).

flooded mangrove forests. This environment is considered to be the most oil-sensitive

habitat described in Table 4 - ESI Ranking specification = 10c and 10d (Souza Filho et al.,
2005a).
High resolution images, such as Ikonos, were used for operational scale mapping. The
resolution of 1 m provides a detailed geomorphic map, and it’s also possible to map the
potentially hazardous industrial structures stratified by type of hazard. However, the use of
Ikonos images is limited when cloud cover is higher than 25%. As a result the images are
mostly inadequate between March and June (Andrade et al., 2010). On the other hand,
biological and socioeconomic resources, risk areas and oil spill hazard zones of storage and
platform transportation can be better identified and delimitated (Rodrigues & Szlafsztein,
2010; Andrade et al., 2009). This location contributes to planning and management strategies
and cleaning efforts.
324 Remote Sensing – Applications

Spatial
Author Map Type Sensor Method Scale Study case
Resolution

Visual
Municipality
Andrade et Oil spill interpretation,
High Ikonos Operational of São Luis
al. (2010) Vulnerability Field data
(Maranhão)
collection

Rodrigues Visual
Municipality
& interpretation,
Oil spill risk High Ikonos Operational Barcarena
Szlafsztein Field data
(Pará)
(2010) collection

Oil spill
hazard
Visual
representation Municipality
Andrade et interpretation,
and High Ikonos Operational of São Luis
al. (2009) Field data
susceptible (Maranhão)
collection
socioeconomic
resources

Visual
Boulhosa Municipality
interpretation,
& Mendes ESI Moderate Spot-5 Operational of Barcarena
Field data
(2009) (Pará)
collection

Municipalities
of Maracanã,
Santarém
Landsat 7 Automatic Novo,
Boulhosa
ETM+, classification, Salinópolis,
& Souza
ESI Moderate SRTM multi-fusion Tactical Cuiarana, São
Filho
High, aerial sensors, Field João de
(2009)
photography data collection PIrabas, Santa
Luzia and
Primavera
(Pará)

Automatic
Landsat 7 classification, Municipality
Gonçalves
ESI Moderate ETM, multi-sensor Tactical of Belém
et al. (2009)
Radarsat 1 fusion, field (Pará)
data collection

Souza Landsat 5 Municipality

Visual
Filho et al. ESI Moderate TM, Tactical of Bragança
classification
(2009) Radarsat 1 (Pará)

Visual
Municipality
Novaes et Landsat 5 interpretation,
ESI Moderate Operational of São Luis,
al. (2007) TM Field data
(Maranhão)
collection

Table 5. Results published in a scientific paper related to remote sensing use and sensitivity
environment to oil spill in Amazon coast.
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 325

5. Conclusions
Remote sensing techniques are used for risk identification, assessment and analysis. The
technological risk of oil spills needs continuous planning and monitoring actions. The
availability of airborne and satellite remote sensing provides a diversity of resolution and
sensors required to construct environmental sensitivity maps, using basic information about
socioeconomic and biological resources and geomorphic characteristics.
Remote sensing and ground confirmation provide accurate information about this basic
information. In particular, the coastline is usually mapped in detail with both optical and
radar sensors. The multi-sensor data fusion of an optical moderate sensor with radar has
been extensively used in the Amazon region to provide basic information about coastal
environments. Radar is a very powerful tool once it can operate in difficult weather
conditions. It provides detailed information about shoreline irregularities and geomorphic
units if the texture and the altitude of this type of images are precise. Optical sensors are
used for environmental differentiation once land cover and land use have different spectral
responses.
Studies in Brazil regarding oil spills have increased after 2000, and ESI maps have been
generated at different scales for different areas along the coast. Remote sensing tools were
essential to achieve initial and advanced cartographic information in a context of the
diversity of the environment, information and cartographic background. Particularly in the
Amazon, little background information about the coastline existed before the PIATAM
MAR project. In the context of this project, the Amazon coast was previewed at strategic
scale with the use of a moderate sensor. ESI maps were produced at the tactical and
operational scales and it was possible to map the coastal environment and organize
information about socioeconomic and biological resources. A large, extensive mangrove
system coexists with industrial port areas on the Amazon coast with a high sensitivity to oil
spills, which should to be continuously monitored with remote sensing techniques.

6. References
Andrade, M.; Souza Filho, P. & Szlafsztein, C. (2009). High Resolution Images for
Recognition of the Susceptibility of Social Economic Resources to Oil Spill in the
Itaqui-Bacanga Port Complex, Maranhão, Brazil. Journal of Integrated Coastal Zone
Management, vol. 9, pp. 127-133.
Andrade, M; Szlafsztein, C., Souza Filho, P., Araújo, A. & Gomes, M. (2010). A
socioeconomic and natural vulnerability index for oil spills in an Amazonian
harbor: A case study using GIS and remote sensing. Journal of Environmental
Management, vol 91, pp. 1972-1980.
Araújo, S., Silva G. & Muehe, D. (2002). Manual básico para elaboração de mapas de
sensibilidade ambiental a derrames de óleo no sistema Petrobras: Ambientes
Costeiros e Estuarinos. CENPES/Petrobras, Rio de Janeiro, 134 pp.
Araújo, S.; Silva, G. & Muehe, D. (2006). Mapas de sensibilidade ambiental a derrames de óleo -
Ambiente costeiros, estuarinos e fluviais. Rio de Janeiro Rio de Janeiro.
CENPES/Petrobras, pp.168 Ed. 2ª, ISBN: 8599891014
Araújo, R.; Petermann, R.; Klein, A.; Menezes, J.; Sperb, R. & Gherardi, D. (2007).
Determinação do ìnndice de sensibilidade do litoral (ISL) ao derramamento de óleo
326 Remote Sensing – Applications

para regiões norte e centro-norte da costa de Santa Catarina (SC), Gravel, N°5, pp.
45-73.
Belloto, V. & Sarolli, V. (2008). Mapeamento da sensibilidade ambiental ao derramamento
de óleo e ações de resposta para a região costeira e área portuária de Imbituba, SC,
Brasil. Brazilian Journal for Aquatic Sciences and Technology, vol. 12, N°2, pp. 115-125.
Bentz, C. & Miranda, F. Application of remote sensing data for oil spill monitoring in the
Guanabara Bay, Rio de Janeiro, Brazil. In: Proceedings of the International Geosciences
and Remote Sensing Symposium (IGARSS), Sydney, 2001.
Berredo J.; Mendes A.; Sales M. & Sarmento J. (2001). Nível de Contaminação por óleo nos
sedimentos de Fundo e na água do Rio Pará, decorrente do acidente com a balsa
Miss Rondônia. In: Prost M. & Mendes A. (Org.). Ecossistemas Costeiros: Impactos e
Gestão Ambiental. Museu Paraense Emílio Goeldi, Belém, pp. 153–165.
Biegert, E., Baker, R., Berry, J., Mott, S. & Scantland, S. (1997). Gulf Offshore Satellite
Applications Project Detects Oil Slicks Using Radarsat, In: Proceedings of
International Symposium: Geomatics in the Era of Radarsat, Ottawa, Canada.
Boulhosa, M. & Mendes, A. (2009). Mapeamento dos Índices de Sensibilidade Ambiental ao
derramamento de óleo através de Imagens SPOT 5, na região portuária de vila do
conde - Barcarena - PA. XIV Simpósio Brasileiro de Sensoriamento Remoto, Natal,
Brasil, pp. 3597-3603.
Boulhosa, M. & Souza Filho, P. (2009). Reconhecimento e mapeamento de ambientes
costeiros para geração de mapas de índice de sensibilidade ambiental ao
derramamento de óleo, Amazônia Oriental. Revista Brasileira de Geofísica, vol. 27,
pp. 23-37.
Brekke, C. & Solberg, A. (2005). Oil spill detection by satellite remote sensing. Remote Sensing
of Environment, vol. 95, pp. 1–13.
Brown, C. & Fingas, M. (2003). Development of airborne oil thickness measurements. Marine
Pollution Bulletin, vol. 47, pp. 485–492.
Cantagallo, C.; Milanelli, J. & Dias-Brito, D. (2007). Limpeza de Ambientes Costeiros
Brasileiros Contaminados pó Petróleo: uma revisão. Pan-American Journal of Aquatic
Sciences, vol 2, pp. 1-12.
Castro, A.; Grigio, A.; Souto, M.; Amaro, V. & Vital, H. (2006). Modeling and development
of a geographic database: application to the elaboration of oil-spill environmental
sensitivity maps in coastal areas of the Rio Grande do Norte state. Journal of Coastal
Research, vol. 39, pp. 1436-1440.
Carvalho, M. & Gherardi, D. (2008). Mapping the environmental sensitivity to oil spill and
land use/land cover using spectrally transformed Landsat 7 ETM data. Brazilian
Journal for Aquatic Sciences and Technology, vol. 12, N° 2, pp. 1-9.
Chacaltana, T. (2007). Mapeamento de áreas de sensibilidade ambiental ao derrame de óleo
na Baía de Vitória, ES. Master dissertation. Federal University of Espírito Santo.
p.140.
Cohen, M. & Lara R. (2003). Temporal changes of mangrove vegetation boundaries in
Amazonia: Application of GIS and remote sensing techniques. Wetlands Ecology and
Management, vol 11, pp. 223 -231.
Costa, M. (2004). Use of SAR satellite for mapping zonation of vegetation communities in
the Amazon floodplain (10): 1817-1835. International Journal of Remote Sensing, vol
25, pp. 1817-1835.
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 327

Dutrieux, E.; Canovas, S.; Denis, J.; Hénocque, Y.; Quod, J.; Bigot, L. (2000). Guide
méthodologique pour l’élaboration de cartes de vulnérabilité dês zones côitères de
Océan Indien Réalisé par Créocéan, I fremer et Arvan pour le compte de
l’UNESCO/IOC et lê PRE-COI/UE. IOC Manuals and Guides, No 38.
Fingas, M., & Brown, C., 1997, Remote sensing of oil spill, Sea Technology, vol. 38, pp. 37-46.
Fingas, M. & Brown, C. (2000), "Review of Oil Spill Remote Sensing”, In Proceedings of the
Fifth International Conference on Remote Sensing for Marine and Coastal Environments,
Environmental Research Institute of Michigan, Ann Arbor, Michigan, pp. I211-218.
Fingas, M. & Brown, C. (2005). An update on oil spill remote sensors. In: Proc. 28th Arctic and
Marine Oil Spill Program (AMOP) Tech. Seminar, Calgary, Canada, pp. 825-860.
Gonçalves F.; Souza Filho, P.; Paradella, W. & Miranda, F. (2009). Fusão de dados
multisensor para a identificação e o mapeamento de ambientes flúvio-estuarinos da
Amazônia. Revista Brasileira de Geofísica, vol. 27, pp. 57-67.
Gundlach, E. & Hayes, M. (1978). Classification of coastal environments in terms of potential
vulnerability to oil spill impact. Marine Technology Society Journal, vol. 12, pp. 18-27.
Green, E. 2000. Satellite and airborne sensors useful in coastal applications. In: Edwards A.J.
(Ed.) Remote Sensing Handbook for Tropical Coastal Management, Costal Management
Sourcebooks 3, France: UNESCO Publishing, 41-56pp. ISBN: 9231037366
Grüner, K.; Reuter, R. & Smid, H. (1991). A new sensor system for airborne measurements of
maritime pollution and of hydrographic parameters. Geojournal, vol. 24, n° 1, pp.
103-117.
Herz, R. (1991). Manguezais do Brasil. IOUSP, São Paulo, São Paulo. p.227.
IBAMA. 2011. Informações vetorial sobre porto no Brasil. Avaiable in: siscom.ibama.gov.br
IBGE. 2011. Sinopse do censo demográfico. Ministério do Planejamento, Orçamento e Gestão,
Rio de Janeiro, pp. 261.
Jha, M.N.; Levy, J. & Gao, Y. (2008). Advances in remote sensing for oil spill disaster
management: state-of-the-art sensors technology for oil spill surveillance, Sensors,
vol. 8, pp. 236-255.
Jensen, J. (1996). Introductory digital image processing, 2 ed., Prentice Hall, London, 318p.
ISBN: 0131453610.
Kates, R., Hohenemser, C. & Kasperson, J. (1985) Perilous Progress: Managing the Hazards of
Technology, Westview Press, Boulder, CO, ISBN: 0813370256
Lima, M.; Dias-Brito, D. & Milanelli, J. (2008). Environmental sensitivity mapping for oil
spills in Ilhabela, São Paulo, Revista Brasileira de Cartografia, N° 60, pp. 145-154.
Malthus, T. & Mumby, P. (2003). Remote sensing of the coastal zone: an overview and
priorities for future research, International Journal of Remote Sensing, vol. 24, N°13,
pp. 2805–2815.
Mastin, G.; Mason, J.;. Bradley, J.; Axline, R. & Hover, G.(1994). Comparative Evaluation of
SAR and SLAR, In: Proceedings of the Second Thematic Conference on Remote Sensing
for Marine and Coastal Environments: Needs, Solutions and Applications, ERIM
Conferences, Ann Arbor, Michigan, pp. I-7-17.
MEA, Millenium Ecossystem Assessment. (2005). Ecossystem and human well-being. Island
Press, Washington, DC, ISBN 1-59726-040-1.
Meade, R.; Dunne, T.; Richey, J.; Santos, U. & Salati, E. (1985). Storage and remobilization of
suspended sediment in the lower Amazon River of Brazil. Science, 228, pp. 488-490.
328 Remote Sensing – Applications

MMA (2002) Especificações e normas técnicas para elaboração de cartas de sensibilidade

ambiental para derramamentos de óleo. MMA, Brasília.
Muehe, D. & Neves, C. (1995). The implications of sea level rise in the Brazilian coast: A
preliminary assessment. Journal of Coastal Research, vol. 14, pp.54-78.
NOAA. (2002). National Oceanic and Atmospheric Administration. Environmental Sensitivity
Index Guidelines. Office of Response and Restoration Technical Memorandum No
11, Seattle, USA.
Novaes, R.; Tarouco, J.; Rangel, M. & Dias, L.(2007). XIV Simpósio Brasileiro de Sensoriamento
Remoto, Florianópolis, Brasil, pp. 4089-4096.
Noernberg, M. & Lana, P. (2002). A sensibilidade de manguezais e marismas a impacto por
óleo: fato ou mito? Uma ferramenta para avaliação da vulnerabilidade de sistemas
costeiros a derrames de óleo. Geografares, vol.3, pp. 109-122.
Pernetta, J. & Elder, D. (1992). Climate, sea level rise and the coastal zone: Management and
planning for global changes. Ocean & Coastal Management, vol. 18, N°1, pp. 113-160.
PETROBRAS. (2002). Manual Básico para Elaboração de Mapas de Sensibilidade Ambiental a
Derrames de Óleo no Sistema Petrobras: Ambientes Costeiros e estuarinos. PETROBRAS,
Rio de Janeiro.
Poletto, C. & Batista, G. (2008). Sensibilidade ambiental das ilhas costeiras de Ubatuba, SP,
Brasil. Revista Ambiente & Água – An interdisciplinary Journal of Applied Science, vol. 3,
N° 2, pp. 106-121.
Renn, O. (2008). Risk governance: coping with uncertainty in a complex world. 1°ed, Earthscan,
ISBN 978-1-84407-292-7, USA.
Rocha-Oliveira, T.; Klein, A.; Petermann, R.; Menezes, J. & Sperb, R. (2008). Determinação do
índice de sensibilidade do litoral (ISL) ao derramamento de óleo, para região
sudeste e sul do estado de Santa Catarina. Brazilian Journal for Aquatic Sciences and
Technology, vol. 12, N°2, pp. 91-114.
Rodrigues, S. & Souza Filho, P. (2011). Use of Multi-Sensor Data to Identify and Map
Tropical Coastal Wetlands in the Amazon of Northern Brazil, Wetlands, vol. 31, 11-
23pp.
Rodrigues, J. & Szlafsztein, C. (2010). Análise de vulnerabilidade do núcleo urbano da Vila
do Conde frente a vazamento de óleo. Olam: Ciência & Tecnologia, vol 10, pp. 125-
142.
Romero, A.; Riedel, P. & Milanelli, J. (2010). Environmental Oil Spills Sensitivity Mapping of
the Cananéia-Iguape Estuarine System, Southern Coast of São Paulo, Revista
Brasileira de Cartografia, Nº 62, special edition 01, 299-238.
Salisbury, J.; D'aria, D. & Sabins, F. (1993). Thermal Infrared Remote Sensing of Crude Oil
Slicks. Remote Sensing of Environment, vol. 45, pp. 225-231.
Santos, V.; Polidori, L.; Silveira, O. & Figueiredo Jr., A. (2009). Aplicação de dados
multisensor (SAR e ETM+) no reconhecimento de padrões de uso e ocupação do
solo em costas tropicais – costa Amazônica, Amapá, Brasil. Brazilian Journal of
Geophysics, vol. 27, supl. 1, pp. 39-55.
Silva, D.; Amaro, V.; Souto, M.; Nascimento, M. & Pereira, B. (2010). Geomorfology of a
High Sensitive Area on Potiguar Basin (NE Brazil). Journal of Integrated Coastal Zone
Management, vol. 10, N°4, pp. 545-566.
Silva, A.; Klein, A.; Petermann, R.; Menezes, J.; Sperb, R. & Gherardi, D. (2008). Índice de
sensibilidade do litoral (ISL) ao derramamento de óleo, para a ilha de Santa
Remote Sensing and Environmental Sensitivity for Oil Spill in the Amazon, Brazil 329

Catarina e áreas do entorno. Brazilian Journal for Aquatic Sciences and Technology, vol.
12, N°2, pp. 73-89.
Smith, K. & Petley, D. (2008). Environmental Hazards Assessing risk and Reducing disaster.
Routledge: Londres. ISBN: 0-203-88480-9.
Souto, M.; Castro, A.; Grigio, A.; Amaro; V. & Vital, H. (2006). Multitemporal analysis of
geoenvironmental elements of the coastal dynamics of the region of the Ponta do
Tubarão, city of Macau/RN, on the basis of remote sensing products and
integration in GIS. Journal of Coastal Research, vol. 39, pp. 1618-1621.
Souza, C.; Castro, A. ; Amaro, V. & Vital, H. (2005). Sistema de Informações Geográficas
para o monitoramento de derrames de óleo no litoral norte do estado do Rio
Grande do Norte. In: XII Simpósio Brasileiro de Sensoriamento Remoto, Goiânia, Brasil,
vol. 1, pp. 2383-2388.
Souza Filho, P. & Elrobrini, M. (2000). M. Coastal Zone Geomorphology of the Bragança
Area, Northeast of Amazon Region, Brazil. Revista Brasileira de Geociências, vol. 30,
No 3, pp. 518-522.
Souza Filho, P. & Paradella, W. (2002). Recognition of main geobotanical features in the in
the Bragança mangrove coast (Brazilian Amazon Region) from Thematic Mapper
and RADARSAT-1 Data. Wetlands Ecology and Management, vol 10, pp.123-132.
Souza Filho, P.; Miranda, F.; Beisl, C.; Almeida, E. & Gonçalves, F. (2004). Environmental
sensitivity mapping for oil spill in the Amazon coast using remote sensing and GIS
technology. In: International Geoscience and Remote Sensing Symposium, IEEE
Geocience and Remote Sensing, Ancorage,Alaska.
Souza Filho P. & Paradella, W. (2005). Use of RADARSAT-1 fine mode and Landsat 5 TM
selective principal component analysis for geomorphological mapping in a
macrotidal mangrove coast in the Amazon Region. Canadian Journal Remote Sensing,
vol 31, pp. 214–224.
Souza Filho, P. (2005). Costa de Manguezais de Macromaré da Amazônia: Cenários
Morfológicos, Mapeamento e Quantificação a partir de Dados de Sensores
Remotos. Revista Brasileira de Geofísica, Rio de Janeiro, vol. 23, No 4, pp. 427-435.
Souza Filho P.; Gonçalves F.D.; Beisl C.; Miranda F.; Almeida E. & Cunha E. (2005a). Sistema
de Observação Costeira e o Papel dos Sensores Remotos no Monitoramento da
Costa Norte Brasileira, Amazônia. Revista Brasileira de Cartografia, 57, pp. 79–86.
Souza Filho, P.; Paradella, W. & Silveira, O. (2005b). Synthetic Aperture Radar for
Recognition of Coastal Features in the Wet Tropics: Applications in the Brazilian
Amazon Coast. Boletim do Museu Paraense Emilio Goeldi. Zoologia, Belém, vol. 1, No
1, pp. 149-154.
Souza Filho, P.; Prost, M.; Miranda, F.; Sales, M.; Borges, H.; Costa, F.; Almeida, E.; &
Nascimento Jr.; W.(2009a). Environmental sensitivity index (esi) mapping of oil
spill in the amazon coastal zone: the piatam mar project. Brazilian Journal of
Geophysics, vol. 27, supl. 1, pp. 7-22.
Souza Filho P.; Gonçalves, S.; Rodrigues, F. Costa, R. & Miranda, F. (2009b). Multi-sensor
data fusion for geomorphological and environmental sensitivity index mapping in
the Amazonian mangrove coast, Brazil. Journal of Coastal Research, vol. 56, pp. 1592-
1596.
330 Remote Sensing – Applications

Silva, C.; Souza Filho, P. & Rodrigues, S. (2009). Morphology and modern sedimentary
deposits of the macrotidal Marapanim Estuary (Amazon, Brazil). Continental Shelf
Research, vol. 29, pp. 619-631.
Szlafsztein, C. (2000). Coastal management on the state of Pará and the MADAM project
research. In: German-Brazilian Workshop on Neotropical Ecosystems – Achievements and
Prospects of Cooperative Research, Hamburg, Germany.
Szlafsztein, C. & Sterr H. (2007). A GIS-based vulnerability assessment of coastal natural
hazards, state of Pará, Brazil. Journal of Coast Conservation. vol. 11, pp. 53-66.
Teixeira, S. & Souza Filho, P. (2009). Mapeamento de ambientes costeiros tropicais (Golfão
Maranhense, Brasil) utilizando imagens de sensores remotos orbitais. Revista
Brasileira de Geofísica, vol. 27, pp. 69-82.
Turner, B., Clark, W., Kates, R., Richards, J., Mathews, J. & Meyer, W. (1990) The Earth as
Transformed by Human Action, Cambridge University Press, Cambridge, MA.
ISBN: 978521446303
Viles, H. & Spencer, T. (1995). Coastal Problems Geomorphology, Ecology and Society at the Coast.
Edward Arnold, London. ISBN: 0340625406
Wieczorek, A.; Dias-Brito, D. & Milanelli, J. (2007). Mapping oil spill environmental
sensitivity in Cardos Island State Park and surroundings areas, São Paulo, Brazil.
Ocean & Coastal Management. Vol.50, 872-886 pp.
Wisner, B.; Blaikie, P.; Cannon, T. & Davis, I. (2004). At risk: natural hazards, people´s
vulnerability and disasters, 2° ed., Routledge, London, ISBN 0415252164.
 14

Satellite Remote Sensing of

Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro
Reefs, México: A Classification Approach
Ameris Ixchel Contreras-Silva1, Alejandra A. López-Caloca1,
F. Omar Tapia-Silva1,2 and Sergio Cerdeira-Estrada3
1Centro
de Investigación en Geografía y Geomática
“Jorge L. Tamayo” A.C., CentroGeo
2Universidad Autónoma Metropolitana, Unidad Iztapalapa
3Comisión Nacional para el Conocimiento y

Uso de la Biodiversidad, CONABIO

Mexico

1. Introduction
Interest in protecting nature has arisen in contemporary society as awareness has developed of
the serious environmental crisis confronting us. One of the ecosystems most impacted is the
coral reefs, which while offering a great wealth of habitats, diversity of species and limitless
environmental services, have also been terribly damaged by anthropogenic causes. One
example of this is the oil spill from petroleum platforms (in the recent case of the Gulf of
Mexico). The effects of global warming—such as the increase in the incidence and intensity of
hurricanes and drastic changes in ocean temperature—have caused dramatic damage, such as
the bleaching and decrease of coral colonies. In light of this devastating situation, scientific
studies are needed of coral reef communities and the negative effects they are undergoing.
The case study presented in this work takes place in the Chinchorro Bank coral reefs in
Mexico. These are part of the great reef belt of the western Atlantic, with a biological
richness that inherently provides environmental, economic and cultural services at the local
scale as well as worldwide. Nevertheless, these services have been weakened for decades
due to overexploitation, inducing imbalances and problems in the zone. Over recent
decades, numerous biological communities that house constellations of species—whose
natural evolutionary process dates back million of years (Primack et al., 1998)—have been
alarmingly degraded. If this trend continues, the entire evolution that is sustained by the life
of these communities will disappear in a relatively short period of time.
This study clearly demonstrates the application of state-of-art Remote Sensing (RS) in coral
ecosystems. It includes an analysis based on the application of Iterative Self Organizing Data
Analysis (ISODATA) as a classifier for generating classes of benthic ecosystems present in a
coral reef system, using satellite images (Landsat 7-ETM+).
332 Remote Sensing – Applications

2. Use of remote sensing in coral reef ecosystems

The observation of the earth using remote sensors is a most complete method for monitoring
the most significant natural risks (Xin et al., 2007). In general, RS has proven to be a
powerful tool in the overall understanding of natural and anthropogenic phenomena. It is
particularly appreciated as a non-invasive, non-destructive technique with global coverage.
Thus, satellite, airborne and in-situ radiometry have become useful tools for tasks such as
characterization, monitoring and the continuous prospecting of natural resources.
Research using RS has been strengthened in recent decades as a result of the growing
concern worldwide for the preservation of coral reef systems as natural reservoirs. This has
been observed to be an excellent method for analysis, which aids in the holistic study of this
complex ecosystem. In order to develop an approach that helps to safeguard these
ecosystems, it is necessary to understand the physical, chemical, biological and geological
dynamics that occur therein (Brock et al., 2006). Andréfouët & Riegl (2004) refer to RS as a
technology that is now virtually mandatory for research where spatial and temporal
precision is required. RS has gone from being a tool with no application to coral reef systems
to one that is per se indispensable. Andréfouët & Riegl (2004) discuss four reasons why this
change has occurred:
 The proliferation of new sensors for acquiring direct and indirect data for monitoring
coral reefs,
 The proliferation and improvement of analytical, statistical and empirical approaches,
 Recognition of global climate change due to anthropogenic human impacts that are
lethal to coral reefs and
 Improved integration of technology for the conceptual design of coral reef research.
RS techniques offer an option for marine habitat mapping to determine not only the location
and amount of different benthic habitats (Kirk, 1994) but also how these habitats are
distributed and the degree of connectivity among them (Rivera et al., 2006). Nevertheless,
the study of coral reefs using RS presents several important limitations. For example, intense
cloud cover in optical images, optical similarities among spectral signatures of benthic
communities, attenuation of the deep component (specific to each coral reef ecosystem) as
well as the spatial and spectral resolution of remote sensors. In spite of these limitations,
satellite sensors are highly useful for mapping the benthic bottom (Mumby et al., 1997),
monitoring changes in its ecology (Krupa, 1999) and defining management strategies (Green
et al., 1996).

2.1 Determination of ecological characteristics of coral reefs using remote sensors

Some of the characteristics of coral reefs that can be calculated using RS are temperature,
wave height, sea level, turbidity, amount of chlorophyll and concentration of dissolved
organic matter. In the case of atmospheric variables, it is possible to determine cloud cover,
amount of seasonal rainfall, presence of contaminants and incidental solar energy
(Andréfouët et al., 2003). All these factors directly and indirectly influence coral reefs and
determine their health status (Andréfouët & Riegl 2004). In addition, it is possible to
determine the different benthic ecosystems present in the coral reefs, such as seagrass, type
of bottom, algae communities and different types of coral. If the reef is near a tourist or
vacation area, anthropogenic impacts can be determined by calculating the growth of the
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 333

urban stain, vegetation coverage, the structure of the hydrographic basins, etc. Intrinsic
conditions of coral reefs can be described, which are largely defined by the inflows and
outflow and their transport of sediments and export of dissolved organic matter. This
enables us to understand the patterns involved in coral whitening, among other events
(Brock et al., 2006).
The coral reefs—located in relatively clear water—allow us to use passive optic sensors
(Benfield et al., 2007). The more common satellite sensors that have been used to study this
are SPOT, Landsat TM and ETM+ (Andréfouët & Riegl 2004; Benfield et al., 2007; Mumby
2006; Mumby et al., 2004; Mumby and Harborne 1998). Studies previously conducted
(Green, 2000; Mumby et al., 1999) have observed that Landsat and SPOT images are
suitable for mapping corals, sands, and seagrass, depending on their resolution.
Nevertheless, it is important to note that various types of habitats can be represented in
one Landsat image pixel (or others with less spatial resolution), which may limit
classification abilities (Benfield et al., 2007). Previous studies conducted (Green, 2000;
Mumby et al., 1999) have observed that according to the resolution of Landsat images, they
are suitable for mapping sea corals, sands and seagrass. Based on this assumption, the data
obtained from Landsat and SPOT are adequate for simple complexity mapping (3-6 classes,
such as seagrass, sand, dead corals and some species of corals) but for more complex
targets (7-13 classes) they are limited by their spatial and spectral resolution. (Mumby,
1997; Andréfouët et al., 2003; Capolsini et al., 2003). To a lesser extent, SeaWiFS (sea-
viewing wide field of view sensors) have also been used, as well as IKONOS with higher
spatial resolution, LIDAR and SONAR, among others (Andréfouët & Riegl 2004;
Andréfouët et al., 2003; Brock et al., 2006; Elvidge et al., 2004; Liceaga-Correa & Euan-
Avila, 2002; Hsu et al., 2008; Lesser and Mobley, 2007). It is important to note that
analytical methods as well as spatial modeling, statistics and empirical methods at
different scales and for different applications have been used in direct relation to ecological
processes of reefs (Andréfouët & Riegl 2004). The use of airborne remote sensors, such as
CASI (Compact Airborne Spectrographic Imager) with a high spectral or hyperspectral
resolution, has gradually been increasing in this type of studies, to the extent that the
specialists mention that mapping reefs using air or satellite sensors have proven to be more
effective than fieldwork (Mumby, 1999). Nevertheless, field measurements cannot be
discarded, since they provide us with the basis for corroborating the information obtained
from satellite images. In addition, images from satellite sensors provide the opportunity to
conduct multi-temporal monitoring (Helge et al., 2005) in order to identify the status of an
ecosystem and predict possible future changes.
According to the above, it can be stated that studies applying RS in coastal ecosystems and,
specifically, in coral reef ecosystems provide information and knowledge that can
successfully be applied to define management strategies for these important ecosystems, as
well as to design viable alternatives for their conservation.

3. Spectral reflectance of coral

To make observations, we move vertically and gradually from the coral surface to the water
surface, measuring the changes in the quantity of light in the water column that falls directly
on the coral. The quantity of light present obviously affects the amount that is reflected by
the coral, and is therefore a crucially important parameter for mapping it.
334 Remote Sensing – Applications

Spectral reflectance (ρ) is a key parameter for conducting studies of coral reefs using RS
(Hochberg et al., 2004). Two factors clearly and concisely explain this. First, ρ represents the
boundary of radiative transference in the water surface optics. Therefore, taking into
account ρ can resolve the problem of inverse radiative transference presented by passive
remote sensors when applied in this field. Second, ρ is the function that denotes the object,
the composition of the material and its structure. Therefore, it serves as a bridge between the
optics of the object and the shape of the sea bottom (Hochberg et al., 2004).
In the process of classifying images and generating thematic maps, large differences have
been noted in spectral reflectance among the coral reefs’ benthic communities (Brock et al.,
2006). Variability in the vertical relief, or rugosity, is a significant aspect of the complexity of
a habitat, a factor that both reflects and governs the spatial distribution and density of many
reef organisms (McCormick 1994). These factors, which respond to these evaluations, vary
according to the differences among sediments, the presence of different algae species and
the coverage of atypical algae in surface water in some reef zones. Thus, Hochberg et al.
(2004) mention the importance of creating a specific approach using RS to study the surface
water mass presented by atypical algae, since it has been shown that the mere presence of
these organisms indicates classes that are spectrally distinct from other reef communities,
even when they represent the same species.
Differences among the spectral signatures of corals provide a high likelihood of
satisfactorily delineating and defining their different features in a satellite image. The
problem with the above process is that the ρ of the corals is a function of pigmentation,
structure, the orientation of their branches and their internal characteristics (Newman et al.,
2006). In addition, though the interactions between light and the atmosphere are well-
studied, the challenge is to establish controls for the effects of the water column in which the
coral is found that influence these factors. Taking into account the curvatures in order to
correct the acquired data provides more valuable information about the conditions and
health of the living communities sheltered by the coral. Newman et al. (2006) indicate that
two categories have been defined by recent studies which were conducted to measure in situ
the spectral signatures of the coral environment:
i. The spectral signatures are examined according to the variation in the pigment density,
which characterizes the sensorial color of the different coral species (Newman et al., 2006).
Some studies have analyzed the contribution of color to the measurement of radiance
(R), in particular, by comparisons with unpigmented coral structures. These
observations resulted in the spectrum of coral whitening and structures saturated with
zooxanthellae (Newman et al., 2006), which provide a measure of the health status of
the complex reef system. Color has been used as a comparison measurement among
three coral species, five algae species and three benthic communities (Hochberg and
Atkinson, 2000), and as a means to differentiate between dead coral in different stages
and algae colonization (Clark et al., 2000).
ii. Spectral signatures were examined according to morphological characteristics
(Newman et al., 2006).
Corals exhibit distinct and complex structural morphologies, partially due to
environmental conditions such as light availability, water motion and suspended
sediment (Joyce & Phinn, 2002). Reflectance values measured over varying angles and
azimuths were examined to determine the bidirectional reflectance distribution
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 335

function of coral species and the inter-species variation between rounded and
branching types (Joyce and Phinn, 2002; Newman et al., 2006).

4. Mapping coral reefs using remote sensors

The worldwide importance of coral reefs in light of current threats has generated interest in
developing methods to study this type of ecosystems at global scales (Kuhn 2006). The use
of remote sensing to map underwater habitats is increasing substantially. This enables using
the derived information to determine the status of these natural resources as a basis for
planning, management, monitoring, conservation and evaluating their potential.

Fig. 1. Components of Remote Sensing in mapping coral reefs.

As was mentioned previously, high resolution spectral sensors exist that have elements that
enable specific analysis with an excellent capacity for modelling environmental and
structural variables in the coral reefs (Holden and LeDrew, 1998). The data produced by this
type of sensors provide products that can be combined with models to photosynthetically
calculate the radiation available through the photic zone and the surface of benthic
substrates. Established models for calculating incident solar radiation are developed and
evaluated based on routine satellite and meteorological observations (Brock et al., 2006). The
spectral differences among corals, seagrass and algae are nearly imperceptible and not easy
to detect with the three bands (blue, green and red) of the sensors that can penetrate the
336 Remote Sensing – Applications

water column (Holden and LeDrew 1998; Hedley and Mumby 2002; Karpouzli et al., 2004).
This is why RS studies applied to the mapping of submerged benthic ecosystems requires
the generation of new processing methodologies. In addition, coral habitats present a
heterogeneity that is inherent of their complexity, and therefore the task of discerning
among the different spectral signatures is more complicated. That is, the pre-processing of
images applied to this type of environments should not only incorporate the elimination of
noise in the atmospheric and batimetric portions, but should also take into account the
components of the water column, as shown in Figure 1.

5. Pre-processing of satellite images

All satellite images must undergo an initial processing of crude data to correct radiometric
and geometric distortions of the image and eliminate noise. It must be taken into account
that the energy captured by the sensor goes through a series of interactions with the
atmosphere before reaching the sensor. As a result, the radiance registered by the sensor is
not an exact representation of the actual radiance emitted by the covering. This means that
the image acquired in a numerical form presents a series of anomalies with respect to the
real scene being detected. These anomalies are located in the pixels and digital levels of the
pixels that make up the data matrix. The purpose of correction operations is to minimize
these alterations. The corrections are made during pre-processing operations, since they are
carried out before performing the procedures to extract quantitative information. The
product obtained is a corrected image that is as close as possible, geometrically and
radiometrically, to the true radiant energy and spatial characteristics of the study area at the
time the data are collected. Atmospheric correction is a process used to reduce or eliminate
the effects of the atmosphere and allow for more precisely seeing the reflectance values of
the surface being studied or analyzed.
Nevertheless, when attempting to map or derive quantitative information from subaquatic
habitats, the depth of the water significantly affects the measurements taken by remote
sensors, making it possible to generate confusion about spectral signatures. Therefore,
atmospheric and geometric corrections are not sufficient when the objective is to extract
features of the covering of the bottom of the water. That could be considered a characteristic
and, in some cases, a limitation of passive sensors in remote sensor applications in marine
environments. Thus, in this type of studies, a water column correction is performed to
improve reliability when analyzing the results of the image and to eliminate the noise
resulting from the variation in the ground’s reflectance (Holden 2002; Holden and LeDrew,
1998; Mumby, 1998).

5.1 Correction of remotely sensed imagery

5.1.1 Radiometric correction
The radiance from the sensor (L) is calculated as:

L=c0+c1*ND (1)
Where c0 and c1 are the offset and gain, respectively, of the radiometric calibration and ND
is the digital number recorded in a particular spectral band. The process of obtaining L is
called radiometric correction.
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 337

The total signal captured by the sensor consists of three parts: atmospheric scattering of
radiation, radiation reflected by the pixel and radiation reflected by the vicinity of the pixel
and scattered in different (adjacent) directions.

5.1.2 Atmospheric correction

The atmospheric conditions (water vapor, aerosols and visibility) in a scene can be
calculated using algorithms that are performed using a database based on atmospheric
functions. The surface spectral reflectance of an interaction target in a scene can thereby be
seen as a function of the atmospheric parameters. ¶(6pt)

5.1.3 Geometric correction

The geometric correction consists of distinguishing the other types of radiation and only
considering that which is reflected by the pixel. The objective is to remove geometric
distortion; that is, to locate each pixel in its corresponding planimetric position. This enables
associating the information obtained from a satellite image with thematic information from
other sources.

5.2 Water column correction

The coral reefs generally develop in transparent or clear water, which facilitates study and
analysis with passive optic, multispectral or hyperspectral sensors (Mumby et al., 1999).
When light penetrates the water column, its intensity exponentially decreases as the depth
increases. This process is known as attenuation, and it has an important effect on data
obtained by remote sensors in aquatic environments (Green, 2000). The attenuation process
is shown in Figure 2.

Fig. 2. Processes of light attenuation in the water column (SERC, 2011).

338 Remote Sensing – Applications

There are two reasons for this phenomenon:

 Absorption: light energy is converted into another type of energy, generally heat or
chemical energy. This absorption is produced by the algae, which utilize the light as a
source of energy, by suspended organic and inorganic particulate matter (OPM and
IPM), dissolved inorganic compounds and the water itself.
 Scattering: This phenomenon results from the collision of light rays and suspended
particles, causing multiple reflections. The more turbid the water (more suspended
particles) the greater the scattering effect, making it difficult for light to penetrate.
The attenuation varies according to the wavelength of the electromagnetic radiation (EMR).
For example, in the region of visible light, the red portion of the spectrum attenuates more
quickly than the short wavelength, such as blue.
Figure 3 shows, for 4 spectral bands (blue, green, red and infrared), how the spectrum in a
particular habitat (seagrass or macroalgae) can change as the depth increases. The spectral
radiance registered by a sensor is dependent on the reflectance of the substrates and the
depth. As the depth increases, the possibility to discriminate spectrums or spectral
signatures of the habitats decreases. In practice, the spectrum of sand at a depth of 2 meters
is very different than that at 20 meters. According to Mumby and Edwards (2000), the
spectral signature of sand at 20 meters could be similar to that of seagrass at 3 meters. All
these factors influence the signal and can create a good deal of confusion when using visual
inspection or spectral classification to classify these habitats. Therefore, the influence of the
variability in depth must be eliminated, which is known as water column correction or
depth correction (Mumby and Edwards 2000).

Fig. 3. Spectral differences for a habitat (seagrass or macroalgae) at different depths (Mumby
and Edwards, 2000).

A variety of models exist that can be used to compensate for the effect of the water
column. Nevertheless, many require optical measurements of the optical properties of the
water itself, as well as information about the depth of water per pixel (Gordon, 1978;
Philpot, 1989; Mobley et al., 1993; Lee et al., 1999; Maritorena et al., 1994; Maritorena 1996;
Lee et al., 1999). Thus, the method proposed by Lyzenga (1981) is applied, which has been
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 339

used and described by other authors, such as Mumby et al., 1997, 1998, Mumby and
Edwards 2002, Andréfouët et al., 2003, etc. This approach has the advantage of taking into
account the majority of the spectral information and not requiring data for the
components of the water surrounding the reef. Instead of deriving the spectra of the
different types of sea bottoms and water properties, this method transforms the spectral
values into “depth-invariant indices.” The primary limitation of this method, among
others, is that it must be applied to clear water (i.e. type 1 or type 2); the study area meets
this requirement.
To eliminate the influence of depth on sea bottom reflectance, the following need to be taken
into account: the identification of the characteristics of attenuation of the water column and
having digital models of the depth; although these are not very common, particularly for
coral reef systems (Clark et al., 2000). This work used a bathymetric model provided by
SEMAR (2008) that makes possible a good deal of reliability and precision to the
measurements.
The procedure is divided into various steps:
1. Elimination of the atmospheric scattering and the external reflection from the water
surface (atmospheric correction). This can be carried out using a variety of methods,
such as dark pixel subtraction (Maritorena, 1996) and ATCOR (Richter, 1996, 1998).
2. Selection of pixel samples with the same substrate and different depths.
3. Selection of a spectral band pair, with good penetration of the water column (that is,
bands found in the visible light spectrum—Landsat TM and ETM+ 1/2, 2/3 and 1/3).
4. Linearization of the relationship between depth and radiance, Xi = ln (Li), where Xi is
the transformed radiance of the pixel in band i (band 1) and Li is the radiance of the
pixel in band j (band 2). When the intensity of the light (radiance) is transformed using
the natural logarithm (ln), this relationship becomes linear with the depth. Therefore,
the transformed radiance values will decrease linearly as depth increases:

Xi  Ln  Li  (2)

5. Determination of the attenuation coefficient (quotient) using a biplot of the transformed

radiance of the 2 bands (Li and Lj). The biplot contains data for one type of uniform
bottom (sand) and variable depth. It is created using the following equations:

Ki K j  a  a 2
1  (3)

 jj   ii
a and  ij  Xi X j  Xi X j (4)
2 ij

where  ii is the variance in band i and a is the covariance between bands i and j.

6. Lastly, the depth-invariant index is generated using the equation by Lyzenga (1981):

 k  
IIPij  ln  Li    i  ln  L j
 k j  
  (5)
340 Remote Sensing – Applications

The result of this operation generates a new band—the image with water column correction
for a band pair (depth-invariant index). Since the values of this band are whole numbers
with decimals and can be negative, in order to visualize them they need to be converted into
an 8-bit format, that is, gray values between 0 and 255. To this end, minimum and maximum
values for the resulting image must be found and linearly distributed between the values 1
and 255 (0 is not included because it is assigned to the masked surface area). The depth-
invariant index is essential when the objective of the study is to extract spectral data for
submerged aquatic environments.

6. Review of classification methodologies

The classification of a satellite image consists of assigning a group of pixels to specific
thematic classes based on their spectral properties. The spatial classification of
underwater coastal ecosystems is one of the most complex processes in thematic
cartography using satellite images. As previously mentioned, this can be attributed
primarily to the influence of the atmosphere and the ocean water column, through which
electromagnetic radiation passes. In addition, it is worth mentioning that these
ecosystems undergo constant variation, especially after significant events such as strong
hurricanes. Nevertheless, different authors (Mumby et al., 1997; Andréfouët & Payri 2000;
Mumby and Edwards 2002; Andréfouët et al., 2003; Pahlevan et al., 2006; Call et al., 2003,
etc.) have been using remote sensing to develop different classification methods for these
ecosystems and, in particular, for coral reefs.
The maximum likelihood classifier is the most common method, and has been used by
authors such as Mumby et al. (1997), Andréfouët et al. (2000), Mumby and Edwards (2002),
Andréfouët et al. (2003), Pahlevan et al. (2006), and Benfield et al. (2007). Its primary
advantage is that it offers a greater margin for accounting for the variations in classes
through the use of statistical analysis of data, such as the mean, variance and covariance.
The results of the method can be improved with the incorporation of additional spatial
information during the post-classification process, since this helps to spectrally separate the
classes that had been mixed.
Another method also used by Mumby et al. (1997) is agglomerative hierarchical
classification with group-average sorting. An alternative proposal is object-oriented
classification, which consists of two steps, segmentation and classification. Segmentation
creates image-objects and is used to build blocks for further classifications based on fuzzy
logic. Another method that has been used is ISODATA (iterative self-organizing data
analysis), which uses a combination of Euclidian squared distance and the reclassification of
the centroid (Call et al., 2003). In this study, ISODATA was used to perform the
classification.

6.1 ISODATA (Iterative Self Organizing Data Analysis)

ISODATA is an unsupervised classification method as well as a way to group pixels, and
uses the minimum spectral distance formula. It begins with groups that have arbitrary
means and each time the pixels in each of the iterations are regrouped and the means of the
groups change. The new means are then used for the next iterations.
The algorithm for obtaining the classification is based on the following parameters:
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 341

a. The user decides on the number N of clusters to be used. For the first calculation, it is
recommended to use a high number, which is then reduced by interpreting the image.
b. A set of N clusters in the space between the bands is selected. The initial location is in
the zones with the highest reflectance.
c. The pixels are assigned to the closest cluster.
d. The clusters are associated, dispersed or eliminated depending on the maximum
distance of the class or the minimum number of pixels in a class.
e. The grouping of pixels in the image is repeated until the maximum number of iterations
has been reached, or a maximum percentage of pixels are left unchanged after two
iterations. Both parameters can be specified

7. Case study
The Chinchorro Biosphere Reserve (Fig. 4) is located in the open Caribbean Sea, 30.8 km east
of the coastal city of Mahahual, which is the closest continental point. The coral reef of
Chinchorro Bank, Mexico, is part of the great reef belt in the western Atlantic, the second
largest in the world, and is the biggest oceanic reef in Mexico. With a reef lagoon area of 864
km2, it is considered a pseudo-atoll or reef platform (Camarena, 2003). Chinchorro Bank is a
reef complex that contains an extensive coral formation with a vast wealth and diversity of
species and high ecological, social and cultural value. It inherently provides certain services,
including the protection of the coast from battering by storms and hurricanes. The area has
been exploited by fishing and tourist-related scuba diving over the past decades. The
Chinchorro Bank supports pristine reefs, coral patches, extensive areas of seagrass,
microalgae beds and sand beds. The reserve’s ecosystems are marked by mangroves and
reef zones. The composition of the taxocenosis of coral is known to contain hexacorals,

Fig. 4. Study Area: Chinchorro Bank, Mexico.

342 Remote Sensing – Applications

octocorals and hydrozoas and a reported 95 different species (Camarena, 2003). The
diversity of the fauna in the Chinchorro Bank is very high and includes several phyla,
families, genres and species, with at least 145 macro invertebrate and 211 vertebrate species,
in addition to the corals (Bezaury et al., 1997).
The biogeographic region of Chinchorro Bank is delimited on the north by the Caribbean
Province which extends along Central and South America. This province begins in Cabo
Rojo, in southern Tampico, and extends into eastern Venezuela and the northern Orinoco
delta. The land biota is greatly similar to that of the continent and is therefore considered
to be part of the Yucateca Province. It is located in the Mexican Caribbean, across from the
southeastern coast of the state of Quintana Roo, between the 18º47’-18º23’ N and 87º14’-
87º27’ W parallels. It is 30.8 km from the continent and separated from it by a wide canal
1000 m deep. The shape of Chinchorro Bank is elliptic, with a reef lagoon that includes a
sandy bank 46 m long (north-south) and 18 km wide (east-west) at its broadest part. The
total area is 144360 ha. The periphery of the bank is bordered by active coral growth on
the eastern (windward) margin, which forms a coral reef, or breaker, while along the
western margin (leeward) the breaker disappears and the coral growth is semicontinuous
and diffuse (Camarena, 2003). There are four emerged zones within the bank—known as
“Cayo Norte” (two islands), “Cayo Centro” and “Cayo Lobos”—whose ecological value is
very high because of their diverse species of land and water flora and fauna (Camarena,
2003).

8. Information resources
The geospatial database used in this study includes a Landsat 7-ETM+ image (Table 1),
bathymetric information and in situ data for sand (Figure 5). The digital data were projected
to UTM (Universal Transverse Mercator) zone 16 north with WGS-84 datum. ERDAS,
GEOMATIC 10.2 and ArcMap 9.3 were used to process the data.
The importance of choosing the type of image with which to work is well-known,
particularly because the users will need to make sure to use images that are suitable to the
purpose of the study. The nature of a platform-sensor system determines the characteristics
of the image’s data (Green, 2000). The Landsat 7-ETM+ (Table 1) image obtained had no
cloud cover. It is worth noting that this type of images provides adequate coverage of the
area for regional and temporal monitoring studies.

Date 2000-03-29
Scan time 16:03:05
Path/Row 18/47
Spatial resolution (m) 30
Spectral bands used 3
Spectral range (μm) 0.5-0.69
Azimuth 116.29
Solar angle 59.43
Table 1. Characteristics of the Landsat 7-ETM+ image used
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 343

It is also very important to note that bathymetry is one of the most relevant factors in the
dynamic ecology of coral reefs. Numerous reef studies show that coral species diversity tends
to increase as a function of depth, reaching its maximum between 20–30 m and diminishing
with greater depth (Huston, 1985). This depth effect results in a marked zonation of the reef
community (Aguilar-Perera and Aguilar-Dávila, 1993). While the upper depth limits of corals
are controlled by various physical and biological factors, their maximum depth depends
largely upon light availability (García-Ureña, 2004). The bathymetric soundings for
Chinchorro Bank used by this study were done in 2008 by the Mexican Navy (SEMAR, 2008).
The depth of the interior of the bank varies. The northern portion is shallower, between 1 and
2 m, the depth of the central portion ranges between 3 and 7 m, and the southern portion is
deepest, varying between 8 and 15 m (SEMAR, 2008). There are 4 emerged zones within the
bank, known as keys, which have high ecological value because of their diverse species of flora
and aquatic and land fauna (Camarena, 2003). Figure 5b shows bathymetry data for the
Chinchorro Bank, where the depths of the zone can be seen.
In situ sampling data were provided by SEMAR. Data from Carricart-Ganivet et al. (2002)
were also used. Based on these data, 4 of the most representative classes were determined:
1) coral mass, 2) coral patches, 3) seagrass and algae and 4) sand. The ocean and keys, or
emerged areas, are not part of the classification criteria, though they are also represented.
Unfortunately, the databases for the in situ sampling have disadvantages—such as mixing
classes in the same point and lack of definition of the benthic bottom, among others—that
prevent their being used for validation purposes. Only data for sand provided by SEMAR
do not present these disadvantages and could be used for water column correction, as
explained further below.

Fig. 5. Information resources. a) Landsat 7-ETM+ image and b) depth of the Chinchorro Bank
344 Remote Sensing – Applications

9. Results and discussion

9.1 Image processing
A Landsat 7-ETM+ image from March 29, 2000 was processed. Before conducting the
quantitative analysis of the data, a post-calibration was performed of the constant gain and
offset to convert the image ND to spectral radiance. The spectral radiance was also corrected
for atmospheric effects to obtain the surface reflectance values. A geometric correction was
not performed because the level of the processing of the Landsat images includes this
correction. Only 3 of the 8 bands contained by Landsat were used (blue, green and red). The
depth correction was developed with the Lyzenga (1981) method, which has been used and
described by other authors (Mumby et al., 1997, 1998; Mumby and Edwards 2002;
Andréfouët et al., 2003).

9.2 Water column correction

Lyzenga (1981) shows that when drawing a scatterplot of 2 of the logarithmically
transformed bands in the visible spectrum (one on each axis), the pixels for the same type of
bottom (i.e. sand at different depths) follow a linear trend. Repeating this process for
different types of bottoms produces a series of parallel lines and the intersection of those
lines generate a unique depth-invariant index which is independent of the type of bottom;
all the pixels for a particular bottom have the same value as the index regardless of the
depth at which they are found (Andréfouët et al., 2003). A group of pixels representative of
the depth of the water column was selected for this study, therefore pixels very close to the
surface (< 1m) were eliminated. Sand was the only substrate used since it is the most
homogenous bottom in coral environments, and is the one most used by various authors
(Mumby and Edwards 2002; Lyzenga, 1981) and the most easily recognizable for
interpretation purposes. For the specific case of the Chinchorro Bank, 100 points of sand
between 1 and 10m of depth were used to determine the attenuation coefficient (quotient)
for the band pair ½, 99 points were used for bands 1/3 and 96 for bands 2/3. The data for
point radiance to a type of bottom were extracted from the image and transferred to a
spreadsheet. Figure 6a shows the graphic spectral radiance of bands 1 and 2
(atmospherically corrected) with respect to the depth for one specific type of bottom (sand)
and variable depth.
Figure 6b shows the linearization of the exponential attenuation of the radiance for bands 1
and 2 using natural logarithms, since in practice it is virtually impossible for the points to
adhere to a perfect line given the natural heterogeneity of the different types of bottoms,
variations in the water quality, surface roughness of the water, etc. Figure 6c shows the
biplot of bands 1 and 2 for a single substrate (sand) at different depths. To this end, the
variance of band 1 and the covariance of bands 1 and 2 are evaluated (Table 2 and 3). Table 3
shows the different values for obtaining the attenuation coefficient, according to spectral
band.

Band 1 Band 2 Band 3

Variance (  ii ) 0.2628 0.6334 0.2761

Table 2. Variance of the radiance of each band

Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 345

Fig. 6. Steps for water column correction: (a) spectral radiance of bands 1 and 2
(atmospherically corrected), (b) exponential decay of the radiance for bands 1 and 2 using
natural logarithms and (c) biplot of bands 1 and 2 for a single bottom (sand) at different
depths.

Ratio 1/2 Ratio 1/3 Ratio 2/3

Covariance (  ij ) 0.3200 0.1178 0.2327
aij -0.0593 -0.0031 0.0184
ki/kj 0.94 0.99 1.00
Table 3. Calculation of ratio of attenuation coefficients

Figure 6c shows the biplot of the logarithmically transformed bands 1 and 2, representing the
attenuation coefficient (ki/kj) for bands 1 and 2. It is important to mention that if different
types of bottoms are represented in a biplot, they would theoretically represent a line with a
similar behavior, varying in position only due to differences in spectral reflectance. The
gradient of the line would be identical since ki/kj does not depend on the type of bottom. The
intersection of the line with the y-axis represents the depth-invariant index, since each type of
bottom has a unique y-intersect regardless of depth. Each pixel is assigned an index depending
on the type of bottom, which is obtained using the natural logarithm transformation for each
band and the connection of the coordinate to the origin of the y-axis through gradient line
ki/kj. The pixels are thus classified for different types of bottoms.
346 Remote Sensing – Applications

As mentioned before, the depth-invariant index is generated according to band pairs—1/2,

1/3 and 2/3, correponding to bands 1 (blue), 2 (green) and 3 (red) (Figure 7). The image

Fig. 7. Visualization of the Landsat 7-ETM+ image before and after water column correction.
a) image of band 1 (blue, 450-520 nm), b) band 2 (green, 530-610 nm), c) band 3 (red, 630-655
nm), d) depth-invariant index combination of bands 1/2, e) 2/3 and f) 1/3.
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 347

resulting from the depth-invariant index was significantly different than the image without
correction, since it reveals more details of the structures of the benthic bottom, especially in
zones with greater depths.

9.3 ISODATA classification

As an initial approach to the classification of submerged benthic ecosystems in the
Chinchorro Bank, ISODATA was used as a classification method, since not much needs to
be known about that data beforehand. A little user effort is required to identify spectral
clusters in data. The results of the benthic classification in the Chinchorro Bank were
visually evaluated according to the quality of the segmentation using the classification by
Aguilar-Perera & Aguilar-Dávila (1993), and with bathymetric data that greatly determine
the ecology of the corals, as explained next.
Figure 8a shows the Landsat image with atmospheric correction for the RGB (1,2,3)
combination and Figure 8b shows the image resulting from the depth-invariant index by
bottom type. At the bottom of the figure, two images classified using ISODATA are
included, both with the same type and number of classes. Figure 8c presents the
classification performed without water column correction; that is, using the image from 8a
as input. Figure 8d includes the classification performed based on the depth-invariant index
(shown in 8b); that is, taking into account water column correction. To identify the
categories resulting from the ISODATA process, benthic bottoms in the Chinchorro Bank as
defined by Aguilar-Perera & Aguilar Dávila (1993) were used as a basis. It can be seen (8c)
that the classification without water column correction produced a substantial mix of classes
throughout the image, unlike the classification obtained by applying water column
correction (8d). According to authors such as Aguilar-Perera & Aguilar Dávila (1993),
Chávez and Hidalgo (1984) and Jordán (1979), the periphery of the Chinchorro Bank is
surrounded by abundant coral growth on the eastern margin. A barrier reef is thereby
formed that disappears along the western margin where the coral growth is semi-
continuous and diffuse. This spatial distribution of the corals can be clearly seen in the
results of the classification with water column correction (Figure 8d), unlike classification
without correction (Figure 8c).
One known ecological characteristic of reef systems is that the zonation of the reef bottom
and its ecological dynamics are strongly influenced by the depth (Huston, 1985; Loya, 1972;
Gonzáles et al., 2003). The seagrasses constitute a type of benthic bottom normally present in
shallower zones. These observations and the use of bathymetry enable corroboration of the
validity of the spatial distribution of seagrasses obtained by classification with water
column correction. The shallower zones are located in the northern (1-2m) and central (3 and
4 m) portions; these two zones best correspond to the zone with seagrass generated in the
image shown in 8d, as opposed to the image in 8c where it can be seen that the seagrass
class is distributed throughout the bank. In addition, 8c shows a mix between seagrass and
corals, a result that is not justifiable since the corals normally develop at depths between 5
and 30m. Using the depth criterion again in order to define the zonation, it is possible to
state that the classification with water column correction produces good results for
identifying coral patches, since they are found at depths between 7 and 12 m, as can be seen
in Figure 8d. As a general observation, we can state that the results of the classification with
water column correction generate data that are consistent with the theory regarding the
348 Remote Sensing – Applications

Fig. 8. a) Landsat 7-ETM+ image, RGB (1, 2, 3), b) image resulting from the depth-invariant
index by bottom type using bands 1 and 2, and classification of the benthic bottom in the
Chinchorro Bank using ISODATA, c)without water column correction and d) with water
column correction.
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 349

influence of depth in defining the zonation of benthic bottoms, as well as observation of

other authors regarding the spatial distribution of sea-bottoms.
Figure 9 shows a close-up to facilitate the visual analysis of the differences between the
classes obtained using ISODATA, implemented with and without water column correction.

Fig. 9. Comparison among a) Landsat 7-ETM+ image, RGB (1, 2, 3), b) depth-invariant index
by bottom type for bands 1/2, c) ISODATA without water column correction and c)
ISODATA with water column correction.
350 Remote Sensing – Applications

In this figure, it can be seen that thanks to the water column correction, the classes are better
defined, with mixing among them—caused by interference by the depth of the water
column—avoided to whatever extent possible. The ISODATA algorithm more accurately
selects and groups clusters, eliminating this problem. This visualization again confirms the
advantage of performing water column corrections to obtain better results for the processes
to classify benthic bottoms.

10. Conclusions
The study shows that the application of new remote sensing methods is crucial to the pre-
processing of images in order to identify submerged aquatic ecosystems. This is because
when quantitative information is mapped or derived from satellite images of aquatic
environments, the depth of the water causes spectral confusion and therefore significantly
affects the measurements of submerged habitats. Water column correction minimizes this
effect, which enables distinguishing the classes of benthic ecosystems present in the
Chinchorro Bank and demonstrates improvement especially in zones representing more
variation in depth. Thus, water column correction is an indispensible pre-processing method
in the cartography of submerged aquatic ecosystems.
The water column correction method used in this study uses the majority of the spectral
information while disregarding the characteristics of the water surrounding the reef, such
that the spectral values are transformed from a band pair into a depth-invariant index. This
should be applied in relatively clear water (type 1 or type 2), as is the case of the Chinchorro
Bank. Using this process, the attenuation effect of the water column was minimized, which
is one of the primary problems with the segmentation of images of submerged ecosystems.
Traditional, unsupervised classification methods, such as ISODATA, have difficulty
detecting subclasses, that is, this type of classifier makes it complicated to detect pixels
between very close classes with distributions that share an overlapping zone. When
classifying benthonic habitats in the Chinchorro Bank, it was possible to observe that the
classes with less concentration of pixels were masked by those with greater amounts. This
may be because standard methods, such as ISODATA, use moving mass center techniques
to locate the classes and, thus, what are called subclasses become undetectable.
In general, the data from remote sensors are used for mapping reef habitats. Although the
classification presented here was quite general—only 4 classes were determined—the results
show that the Landsat 7-ETM+ images are able to identify different classes in submerged
benthonic environments. Although the classification resulted in visually optimal results, the
need to incorporate statistical validation of the data is important, so as to determine the
accuracy of the classification performed in comparison to the reality; this was not possible
for this study because an adequate database of in situ sampling was not available.
Nevertheless, because of the visual comparison with classes identified by studies such as
those by Aguilar-Perera & Aguilar Dávila (1993), Chávez and Hidalgo (1984) and Jordán
(1979) and the consistency with the theory of the zonation of benthic bottoms based on
depth, it can be concluded that the classifications obtained by ISODATA successfully
determined the majority of the benthonic cases defined in this study of the Chinchorro Bank.
Coral reefs are being threatened worldwide by a combination of natural and anthropogenic
impacts. Although the natural impacts are intense, there are intermediate time lapses that
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 351

can contribute to maintaining biodiversity. On the other hand, the human impacts—which
may seem to be less intense because they are not as perceivable to the eye—are chronic and
can unleash a chain of negative effects. This sequence of negative effects normally does not
give ecosystems the opportunity to recover and maintain their characteristic function and
structure.
The search for new methodologies to process satellite images is indispensable to identifying
the current trend in the degradation of marine habitats; methodologies that generate new
and improved classifications that are highly reliable and with a level of detail that is
adequate for mapping these ecosystems. Through this type of study, it is possible to
organize, relate and manage information from satellite images in order to propose agreed-
upon strategies to conserve natural resources, as part of comprehensive environmental
policies to properly solve the problems. Thus, these data can be used as a basis to plan the
monitoring of reefs in order to create scientific methods to generate knowledge and
environmental awareness in the society and to contribute to the mitigation of the loss of
reefs due to impacts from current global warming and other anthropogenic and global
changes.

11. Acknowledgments
The authors would like to thank the Mexican Navy (SEMAR), Deputy Department of
Oceanography, Hydrography and Meteorology (Dirección General Adjunta de
Oceanografía, Hidrografía y Meteorología) for the information provided regarding
bathymetry and the field sampling of sand data. We also thank Dr. Juan Pablo Carricart
Ganivet and Janneth Padilla Saldívar for the information and geographic basis from the
Comprehensive Management of the Chinchorro Bank: Geographic survey and
geomorphologic characterization of the reef.

12. References
Aguilar-Perera A. & Aguilar-Dávila W., (1993). Banco Chinchorro: Arrecife Coralino en el
Caribe. Pp. 807-816. In: Biodiversidad Marina y Costera de México, Salazar-Vallejo, S.I.
& González, N.E. pp. 1-35 Com. Nal. Biodiversidad y CIQRO, México
Andréfouët, S. y Riegl, B. (2004). Remote sensing: a key tool for interdisciplinary assessment
of coral reef precesses. Coral reef, Vol. 23, No. 1, (April 2004), pp 1-4, ISSN: 0722-
4028
Andréfouët, S. ; Kramer, P. ; Torres-Pulliza, D. ; Joyce, K. ; Hochberg, E. ; Garza-Perez, R. ;
Mumby, P. ; Riegl, B. ; Yamano, H. ; White, W. ; Zubia, M. ; Brock, J. ; Phinn, S. ;
Naseer, A. ; Hatcher, B. & Muller-Karger, F. (2003). Multi-site evaluation of
IKONOS data for classification of tropical coral reef environments. Remote sensing of
environment, Vol. 88, No. 1-2. (November 2003), pp. 128-143, ISSN 0034-4257
Benfield, S., H. Guzman, J. Mair & A. Young (2007). Mapping the distribution of coral reefs
and associated sublittoral habitats in Pacific Panama: a comparison of optical
satellite sensors and classification methodologies. Internatinal joural of remote
sensing, Vol. 28, No. 22, (November), pp. 5047-5070, ISSN 5047 5070
352 Remote Sensing – Applications

Brock, J., K. Yates & R. Halley (2006). Integration of coral reef ecosystem process studies and
remote sensing, In: Remote sensing of aquatic coastal ecosystem processes, Richardson,
L.L & LeDrew, E. F., 324 pp. Springer, ISBN-13 978-1-4020-3968-3, Netherlands
Call, A. K., T. J. Hardy & D. O. Wallin (2003). Coral reef habitat discrimination using
multivariate spectral analysis and satellite remote sensing. International Journal of
remote sensing, Vol. 24, No. 13, pp. 2627-2639
Camarena-Luhrs, T., (2003). Ficha Informativa de los Humedales de Ramsar: Reserva de la
Biosfera Banco Chinchorro. Cancún, México.
Capolsini P, Andréfouët S, Cedric R, Payri C (2003) A comparison of Landsat ETM+, SPOT
HRV, Ikonos, ASTER, and air- borne MASTER data for coral reef habitat mapping
in South Pacific islands, Canadian Journal of Remote Sensing, Vol. 29, No. 2, pp. 187-
200
Capolsini, P., B. Stoll, S. Andréfouët (2003). A comparison of classification algorithms for
coral reef habitat mapping in South Pacific islands, IGARSS
Carricart-Ganivet J.P., E. Arias-González, G. García-Gil, G. Acosta-González, A. Beltrán-
Torres, J. Castro-Pérez, N. Membrillo-Venegas y J. Padilla-Saldívar, (2002). Manejo
integral de Banco Chinchorro: levantamiento geográfico y caracterización
geomorfológica del arrecife, con especial énfasis en las comunidades coralina e
íctica. SISIERRA, Mexico
Chávez, E. & Hidalgo, E. (1984). Spatial structure of benthic communities of Banco
Chinchorro, México, In: Advances in Reef Science, A Joint meeting of the Atlantic Reef
Committee and the International Society for Reef Studies, Miami
Clark, C. D., P.J. Mumby, J. R. M. Chisholm, J. Jaubert (2000). Spectral discrimination of
coral mortality states following a severe bleaching. International Journal of Remote
Sensing, Vol. 21, No. 11, (January 2000), pp. 2321-2327, ISSN 0143-1161
Elvidge, C.; Dietz J.; Berkelmans, R.; Andréfouët, S.; Skirving, William.; Strong, Alan. &
Tuttle, B. (2004). Satellite observation of Keppel Islands (Great Barrier Reef) 2002
coral bleaching using IKONOS data. Coral Reefs, Vol. 23, No.1, pp 123-132, ISSN
07224028
García-Ureña, R. P., (2004). Dinámica de crecimiento de tres especies de coral en relación a las
propiedades ópticas del agua, 24 September 2011, Available from:
<http://grad.uprm.edu/tesis/garciaurena.pdf>
González, A.; Torruco, D.; Liceaga, A. & Ordaz, J. (2003). The shallow and deep bathymetry
of the banco chinchorro reef in the mexican caribbean. Bulletin of Marine Science,
Vol. 73, No. 1, pp. 15-22, ISSN 00074977
Gordon, H. (1978). Removal of atmospheric effects from satellite imagery of the oceans.
Applied Optics, Vol. 17, No. 10, (May 1978) pp. 1631-1636
Green, E. P. ; Mumby, P. J. ; Edwards, A. J. & C. D., Clark, (1996). A review of remote
sensing for the assessment and management of tropical coastal resources. Coastal
Management, Vol. 24, No. 1, (January 1996), pp 1-40
Green, E., (2000). Satellite and airbone sensors useful in coastal applications. In: Remote
sensing handbook for tropical coastal management, Green, E., P., Mumby, A. Edwards &
C. Clark, Pp 41-56, UNESCO. ISBN 92-3-103736-6, France
Hedley, J.D. & Mumby, P.J. (2002). Biological and remote sensing perspectives on
pigmentation in coral reefs. Advances in Marine Biology, Vol. 43, pp. 277-317,
Satellite Remote Sensing of Coral Reef Habitats Mapping in
Shallow Waters at Banco Chinchorro Reefs, México: A Classification Approach 353

Helge, J. ; Lindberg, B. ; Christensen, O. ; Lundälv, T. ; Svellingen, I. ; Mortensen, P. B. &

Alvsvåg, J. (2005), Mapping of Lophelia reefs in Norway: experiences and survey
methods, In: Cold-water Corals and Ecosystems, Freiwald, A. & Roberts J.M., pp 359-
391, Springer-Verlag, ISBN 978-3-540-27673-9, Berlin Heidelberg
Hochberg, E. ; Atkinson, M. ; Apprill, A. & Andréfouët, S. (2004). Spectral reflectance of
coral. Coral Reefs, Vol. 23. No. 1, (Abril 2004), Pp. 84-95, ISSN 0722-4028
Holden, H. (2002). Characterization of Optical Water Quality in Bunaken National Marine
Park Indonesia. Singapore Journal of Tropical Geography, Vol. 23, No. 1, pp. 23-36,
ISSN: 01297619
Holden, H. & LeDrew, E. (1998). Hyperespectral identification of coral reef features.
International Journal of Remote Sensing, Vol. 20, No. 13, pp. 2545-2563
Hsu, M. K.; Liu, A. K.; Zhao, Y. & Hotta, K. (2008). Satellite remote sensing of Spratly Islands
using SAR, International Journal of Remote Sensing, Vol. 29, No. 21, pp. 6427-6436,
ISSN 01431161
Huston, M. (1985). Variation in coral growth rates with depth at Discovery Bay, Jamaica.
Coral Reefs, Vol. 4, No. 1, (December 1985), pp. 19-25, ISSN 07224028
Jordán, E. (1979). Estructura y composición de los arrecifes coralinos en la región noreste de
la península de Yucatán, México. An. Centro de ciencias del mar y limnología, UNAM,
Vol. 16 No. 1, pp. 69-86
Karpouzli, E. ; Malthus, T. J. ; & Place, C. J. (2004). Hyperspectral discrimination of coral reef
benthic communities in the western Caribbean, Coral Reefs, Vol. 23, (February 2004),
pp. 141–151, ISBN 00338-003-0363-9
Kirk, J.T.O. (1994). Light & photosynthesis in aquatic ecosystems (2nd edition), Cambridge
University Press, ISBN 0521 45966 4, New York
Krupa, S. (1999). Polución, población y plantas, Colegio de Postgraduados, México.
Kuhn, T., (2006). La estructura de las revoluciones científicas, Fondo de cultura económica,
México
Lee, Z., K. L. Carder, C. D. Mobley, R. G. Steward & J. S. Patch (1999). Hyperspectral remote
sensing for shallow waters: 2.Deriving bottom depths and water properties by
optimization. Applied Optics, Vol. 38, No. 18, (June 1999), pp. 3831-3843
Lesser, M. P. & Mobley, C. D. (2007). Bathymetry, water optical properties, and benthic
classification of coral reefs using hyperspectral remote sensing imagery. Coral Reefs,
Vol. 26, No. 4, (January 2007) pp. 819-829, ISSN 07224028
Liceaga-Correa, M. A. & Euan-Avila, J. I. (2002). Assessment of coral reef bathymetric
mapping using visible Landsat Thematic Mapper data. International Journal of
Remote Sensing, Vol. 23, No. 1 pp 3-14, ISSN: 01431161
Loya, Y. (1972). Community structure and species diversity of hermatypic corals at Eilat,
Red Sea. Marine Biology, Vol. 13, No. 2 pp. 100-123, ISSN 00253162
Lyzenga, D.R., (1981). Remote sensing of bottom reflectance and water attenuation
parameters in shallow water using aircraft and Landsat data. International Journal of
Remote Sensing, Vol. 2, No. 1, (Enero 1981), pp. 71-82. ISSN 0143-1161
Maritorena, S., A. Morel & B. Gentili (1994). Diffuse Reflectance of Oceanic Shallow Waters:
Influence of Water Depth and Bottom Albedo. Limnology and Oceanography, Vol. 39,
No. 7 (Nov., 1994), pp. 1689-1703
Mobley, C.D., B. Gentili, H. R. Gordon, Z. Jin, G. W. Kattawar, A. Morel, P. Reinersman, K.
Stamnes & R. H. Stavn (1993). Comparison of numerical models for computing
354 Remote Sensing – Applications

underwater light fields. Applied Optics, Vol. 32, No. 36, (December 1993) pp. 7484-
7504
Mumby, P.J, E.P. Green, A.J. Edwards & C.D. Clark (1997). Coral reef habitat mapping: how
much detail can remote sensing provide? Marine Biology, Vol. 130, No.4 (December
1997), pp. 193-202. ISSN: 0025-3162
Mumby, P.J. & Harborne, A.R. (1999). Development of a systematic classification scheme of
marine habitats to facilitate regional management of Caribbean coral reefs.
Biological Conservation, Vol. 88, No. 2, (May 1999), pp155-163, ISSN 0006-3207
Mumby, P.J., W. Skirving, A.E. Strong, J.T. Hardy, E.F. LeDrew, E.J. Hochberg, R.P. Stumpf
& L.T. David. (2004). Remote sensing of coral reefs and their physical environment.
Marine Pollution Bulletin, Vol.48, No. 3-4, (February 2004), pp. 219-228, ISSN 0025-
326
Mumby, P.J., A.R. Harborne, J.D.Hedley, K. Zychualuk & P. Blackwell (2006). The spatial
ecology of Caribbean coral reefs: revisiting the catastrophic die-off of the urchin
Diadema antillarum. Ecological Modelling, Vol. 196, pp. 131-148
Newman, C., E. LeDrew & A. Lim. (2006). Mapping of coral reef for management of marine
protected areas in developing nations using remote sensing. In: Remote sensing of
aquatic coastal ecosystem processes, Richardson, L. y LeDrew, E., pp. 251-278,
Springer, ISBN 978-1-4020-3968-3, Netherlands
Pahlevan, N.; Valadanzouj, M. J. & Alimohamadi, A. (2006). A quantitative comparison to
water column correction techniques for benthic mapping using high spatial
resolution data. ISPRS Commission VII Mid-term Symposium "Remote Sensing: From
Pixels to processes", Enschede, Netherlands, 8-11 May 2006
Philpot, W. D. (1989). Bathymetric mapping with passive multispectral imagery. Applied
Optics, Vol. 28, No. 8 (April 1989), pp. 1569-1578
Primack, R., R. Rozzi, P. Feinsiger, R. Dirzo y F. Massardo. (1998). Fundamentos de
conservación biológica. Fondo de cultura económica. México.
Richter, R. (1996). Atmospheric correction of satellite data with haze removal including a
haze/clear transition region. Computers & Geosciences, Vol. 22, No. 6, pp. 675-681,
ISSN 00983004
Richter, R. (1998). Correction of satellite imagery over mountainous terrain. Applied Optics,
Vol. 37, No. 16, (June 1998), pp. 4004-4015
Rivera, J. A. ; Prada, M. C. ; Arsenault, J.L. ; Moody, G. & Benoit, N. (2006). Detecting fish
aggregations from reef habitats mapped with high resolution side scan sonar
imager, In: Emerging technologies for reef fisheries research and management, NOAA,
pp. 88-104. NOAA Professional Paper NMFS, 5 Seattle, WA.
Secretaría de Marina, Dirección General adjunta de Oceanografía, Hidrografía y
Meteorología (SEMAR) (2008). Carta náutica SM-932 Majahual to Banco Chinchorro,
Scale 1: 100000
SERC (September 2011). Phytoplankton lab, 10.09.2011 Available from
http://www.serc.si.edu/labs/phytoplankton/primer/hydrops.aspx
Xin, Y., J. Li y Q. Cheng (2007). Automatic Generation of Remote Sensing Image Mosaics for
Mapping Large Natural Hazards Areas, In: Geomatics solutions for disaster
management, Li, J., Zlatanova, S. y Fabbri, A. 451 pp. Springer, ISBN 13 978-3-540-
72106-2, Springer Berlin Heidelberg New York.
 15

Predictability of Water Sources

Using Snow Maps Extracted from
the Modis Imagery in Central Alborz, Iran
Seyed Kazem Alavipanah*, Somayeh Talebi and Farshad Amiraslani
Faculty of Geography, Department of Cartography, University of Tehran, Tehran
Iran

1. Introduction
1.1 Snow reserves and remote sensing
Snow reserves in mountainous basins are important and reliable water resources in Iran.
Identification of their quality is necessary because of an increasing value of freshwater and
utilization of water recourses. About 60 percent of surface water and 57 percent of ground
water sources in Iran flows in snowy regions (Rayegani, 2005). The water produced from
snowmelt process provides soil water, ground water reserves and water in lakes and rivers.
Since snow cover is one of the most important sources of provided water, an accurate
prediction and timing of snow runoff is necessary for the efficient management and
decision- making in water supply.
The science of snow hydrology, compared to other branches of hydrology science, has a
relatively shorter history due to difficulties accompanied with snow measurement. The
correct analysis of snow issues needs a set of observations and statistics in snow-gauging.
Currently, however, there are no regular and comprehensive snow measurement
procedures in most parts of Iran. Measurements are only limited to those snowy basins
recharging important dams; even these measurements are carried out in scattered points
rather than an entire dam catchment area.
The measurement range of these stations is limited to 2000-3000 m asl heights. Thus, in
mountainous Iran, current distribution of stations would not seem to be adequate. In such
conditions, study of snow reserves and identification of snow melting trend in most basins
would be accompanied with limitations. Consequently, measuring snow cover using
ground methods will be difficult and costly. Remote sensing technology has many
applications in various environmental and earth resources studies including ice and snow
research. These applications have been increased recently as a result of unique technical
advantages such as multi-temporal imagery acquired in various wavelengths, extent of
spatial coverage, and improvement of computer hardwares for interpretation and extraction
of information. Regarding snow research, remote sensing technology can provide

* Corresponding Author
356 Remote Sensing – Applications

continuous information layers with higher accuracy and lower cost compared to the ground
survey, so it can fill the information gaps in snow hydrological statistics. However, using
ground data can increase the efficiency of remotely-sensed measurement of snow-gauging.
Satellites are appropriate tools for gauging snow coverage, because of high reflection of
snow that creates proper contrast to most of natural surfaces (with the exception of clouds).
Therefore, using satellite imagery and GIS modeling one can produce snow-cover maps,
assess the changes in snow cover area with various time series, discriminate snow from
other features, and model it in a catchment area. These simplify decision-making process for
engineers and hydrology managers.
One of the important issues in remotely-sensed snow-gauging is the selection of sensor.
Some of optical sensors that have ever used in snow-measuring include sensors mounted to
satellites namely TIROS-1 (1960), ESSA_3, NOAA (1996), LANDSAT (MSS and ETM), and
MODIS (2000). Since each sensor has unique properties, a sensor with appropriate spectral,
temporal and spatial resolution for snow-gauging must be selected. Since snow is a
phenomenon with noticeable surface changes over time, it is necessary to select a sensor that
produces proper multi-temporal series. Snow-gauging is done in vast areas, and snow
surface is generally even; therefore, MODIS is an appropriate imagery for this purpose.
From the view point of spectral resolution, MODIS is one of the best optical sensors for
studying snow and discrimination of snow from phenomena such as cloud which has
similar spectral reflectance.
One of the purposes of designing of MODIS is a global identification of various types of
clouds; hence, several bands have been considered for it to identify various types of cloud
cover, optical thickness, effective radius and thermal phase (King et al., 2004).
NASA (National Aeronautics and space administration) launched TERRA satellite to space
on December 18th 1999, and MODIS as one of the five sensors mounted on TERRA
transferred the first information to Earth on February 24th 2000. MODIS has 36 various
bands in visible, infra-red and thermal parts of electromagnetic spectrum including 2 visible
bands with 250 m resolution, 5 infra-red bands with 500 m resolution, and 29 thermal bands
with 1000 m resolution (Hall et al., 2000).

1.2 A review on remotely-sensed snow measurement

Various methods have been used to estimate snow surface such as classification methods,
threshold limit, decision-based methods, etc. One of the most applicable algorithms used to
estimate snow surface is MODIS snow map algorithm. It was introduced in 1998 as a
decision-based algorithm which uses group tests of threshold limit for detection of snow.
This algorithm has very small volume from the calculation viewpoint and simple from the
conceptual viewpoint, thus user can track how product has been created. In addition, this
algorithm has an appropriate efficiency with global application (Hall et al., 1998).
Totally the properties of this algorithm include:
1. The precision of this method for various types of snow-covered surfaces for
identification of snow surface is higher than other methods such as supervised
classification, unsupervised classification and sub-pixel methods provided that
atmospheric correction is considered (Dadashi Khaneghah, 2008).
Predictability of Water Sources Using Snow
Maps Extracted from the Modis Imagery in Central Alborz, Iran 357

2. This method is a completely automatic algorithm.

3. This algorithm is applicable for all regions in the globe.
4. This method is simple, accurate and easy to understand.
Snow Map algorithm uses normalized subtractive index (NDSI). Lee et al. (2001) compared
MODIS snow maps created with NDSI index with maps prepared by National Operational
Hydrologic Remote Sensing Center (NOHRSC, prepared automatically by GOES and
NOAA images) in upper region of Rio Grande reservoir. In NOHRSC, the teta algorithm is
used. In this algorithm, two classified images are subtracted to identify snow surface. In teta
algorithm, two separate threshold limit is introduced for each image. Lee et al. (2001)
concluded that both images are affected by cloudy condition and the main error is cloud
coverage. They also mentioned that maps produced from MODIS were more accurate than
the above-mentioned maps.
Ault et al. (2006) concluded that MOD10-L2 snow surface product, MODIS sensor, in clear
sky condition had the highest accuracy. They showed that the highest error was associated
with those conditions that snow depth was lower than 1 cm; thus the higher was the snow
depth, the higher was the accuracy. Hall et al. (2000) also showed a low accuracy of low-
mass and patchy snows in New England. Klein and Barnett (2003) carried out a snow cover
study using MODIS in Rio Grande reservoir during the 2000-2001 period and compared
their results to the ground-measuring methods such as snowpack telemetry (SNOTEL) and
NOHRSC models. They ultimately concluded that the highest error associated with maps
prepared by MODIS was related to the beginning and end of snowfall period. They showed
that when the surface was completely covered by snow with no mixed cloud, ground survey
or SNOTEL had the highest accuracy.
It can be mentioned that MODIS sensor and NDSI index are appropriate in snow map
preparation, although cloud coverage and classification are regarded as constraints (Klein
and Barnett, 2003; Zhou et al., 2005). In fact, in spite of various advantages, Snow Map
algorithm has some limitations due to inseparability of snow cover from cloud and
similarity of cloud behaviours to snow cover. This algorithm cannot completely distinguish
clouds from snow (of course, this problem is relatively removable by using Cloud Mask
algorithm). Also, this algorithm cannot detect coastal terrains which are similar to snow
from viewpoint of whiteness and brightness. However, temperature can act as factor to
discriminate snow from these terrains using MODIS bands 31 and 32. Since Cloud Mask and
thermal Mask are used before applying algorithm, some error sources in snow map
algorithm will be removed (Taghvakish, 2005; Adhami, 2005).
In snow-gauging using satellite imagery, the existence of cloud is problematic due to the
following reasons (Riggs and Hall, 2002): first, clouds conceal Earth information; second,
clouds create shades on area and change reflectance. Indeed, if clouds cannot be detected
well, they will reduce the accuracy of snow map.
Clouds and snow have generally similar spectral reflectional properties in range of visible and
infra-red spectra, so thermal properties is not enough for distinguishing them from each other
as clouds may be cooler or hotter than snow (Singh and Singh, 2001). In order to detect clouds,
a procedure called Cloud Mask algorithm is being used. Akerman et al. (1998) introduced
MOD35 Cloud Mask algorithm. MOD35 algorithm is based on obstruction of Earth surface
affected by cloud or dust particles that identifies water body, land and atmosphere (Strabala,
358 Remote Sensing – Applications

2003). In this process, based on land type, geographical position and available data, Cloud
Mask algorithm uses 14 bands amongst 36 bands of MODIS to test 18 spectral and spatial
features (Hall and Riggs, 2002). However, this procedure was modified by Hall and Riggs
(2002) who presented a new version of Cloud Mask algorithm (Liberal). This algorithm can
analyse the pixels located under thin and transparent clouds (Zhou et al., 2005; Ault et. al.,
2006). This procedure identifies the darkness and if it faces to such darkness, it means that sun
angle is higher than 85º. This algorithm is called Liberal Cloud Mask algorithm. In fact, Liberal
Cloud Mask algorithm functions as subset of spectral tests of old Cloud Mask algorithm
(MOD35) and uses 7 bands of MODIS and set 4 criteria (Hall and Riggs, 2002, 2004).

2. Material and methods

2.1 Study area
The Central Alborz mountainous range extends from 49º 5´- 53º 5´ longitude to 35º 5´- 37º 2´
latitude. Its area is about 40,000 km2 and covers 64 sub-basins. The lowest, highest and
average altitude of the basin is 48 m, 5671 m and 1870 m, respectively. The minimum,
maximum and average slope is 5%, 25.4% and 23.56%, respectively. The main slope aspect
of this region is directing towards north and south. Climatically, the region is classified to
three classes as temperate in north, cold in center, and semi-arid in south. Geological
structures mostly consist of mild outcrops that are inconsistent with general trend of east-
west. In western part of Alborz, the structures have northwest-southeast trend, but in
eastern part, the structures have northeast-southwest trend. These two inconsistent trends
cross each other in Central Alborz.

2.2 Data
2.2.1 MODIS Data, TERRA satellite
MODIS encompasses noticeable number of spectral and thermal bands with narrow width,
high radiometric resolution, proper width and collecting time, powerful and accurate
calibration, and diverse land resolution (MODIS Home page) (http://modis.gsfc.nasa.gov).
In many cases, MODIS provides satellite snow-gauging requirements and therefore these
data were used. In this research, images were provided from website
(https://wist.echo.nasa.gov/wist) according to Table 1.

Year 2006 2006 2006 2006 2006 2006

Month February February march march march march

Day 21 25 4 8 13 15
Table 1. Temporal table applied images from MODIS sensor data

The imagery used in this research include MOD02 and MOD09. The MOD02 imagery
include 36 bands while MOD09 include 6 bands. In MOD09GA imagery, atmospheric
corrections have been done based on 6sv model, as one of the best models in atmospheric
corrections with minimum error, suitable for measuring snow surface and detecting cloud
from snow (Vermote and Kotchenova, 2008). In MOD09 imagery, corrections have been
Predictability of Water Sources Using Snow
Maps Extracted from the Modis Imagery in Central Alborz, Iran 359

implemented in way that atmospheric diffusion and reflection were minimum. Since data
with higher wavelength are being less influenced by aerosols, suspended particles and non-
selective diffusion phenomenon, thermal bands of MOD02 imagery were used.

2.2.2 Digital Elevation Model

Digital Elevation Model (DEM) obtained from SRTM Shuttle was used. These data that was
in format of GeoTiff had Lambert image system. For transformation of this data to UTM
coordinates system, PCI Geomatica software was used.

2.2.3 Ground stations data

Snow-gauge station data were obtained from Water Organization- Department of Surface
Water. Snow-gauge stations of Central Alborz are located in five basins namely Lattian, Lar,
Taleghan, Karaj and Golpayegan basins. In some cases, ground data survey time was not
consistent with the time of image acquisition. For solving this problem, in those dates that
no ground statistics were available, previous and next day’s information were interpolated. Of
course interpolation was carried out for those stations where sampling time was close to image
acquisition date, and the station had snow cover during the period (February and March).
Among sampled stations, 18 stations with above-mentioned conditions were selected.
It is necessary to mention that snow depth data has been used to examine the presence or
absence of snow cover. Those gauging stations located far from human interfering features
(e.g. buildings) were selected and their snow depth measured. The snow surface was
defined an area where surface is regular and even, with the minimum wind blowing effect
to increase measurement accuracy (Pfister and Schneebli, 1999). The output spatial resolution
of snow map algorithm and Cloud Mask Algorithm is one kilometre. Around each station, up
to two kilometres was regarded to include 9 to 13 pixels covering snow. The most repeated
pixel shows snowy or not snowy condition. Finally, snow surface obtained from Snow Map
Algorithm with and without Liberal Cloud Mask was compared to ground data. Figure 1
illustrates distribution pattern of snow-gauge stations within five basins in Central Alborz.

Fig. 1. The position Of snow gauge stations in Alborz-e-Markazi

360 Remote Sensing – Applications

2.3 Research method

Figure 2 illustrates overall flowchart of this research methodology.

MODIS Data

Atmospheric
correction

MOD09 MOD02

Corrections Corrections

Snow Map Cloud mask Thermal mask

Snow Map

Snow cover image Snow cover image

Ground data

Discussion and
conclusion

Fig. 2. The flowchart of this research methodology

2.3.1 Snow map algorithm

Snow map algorithm benefits from Normalized Difference Snow Index (NDSI). Because of
low reflection of snow in infrared bands and high reflection in visible bands, NDSI can be
useful for discrimination of snow from other phenomena. NDSI is calculated by equation
below (Hall et al., 1998):
NDSI = (band4-band6) / (band4+ band6)
Predictability of Water Sources Using Snow
Maps Extracted from the Modis Imagery in Central Alborz, Iran 361

Snow map algorithm includes following thresholds:

If NDSI ≥ 0.4; and MODIS band2 > 11%; and MODIS band 4  0.4.
NDSI index is used to recognize snow and ice and also to differentiate between cumulus
and ice or snow. In fact this index represents relative differential reflectance value of visible
and short wave-infrared channels emitted from snow. Pure snow has a high NDSI value but
other materials such as soil, smoke, etc cause a NDSI reduction. The mentioned threshold
for band 4 of MODIS is a key tool to prevent identifying pixels with low reflection, for
example dark Cypress, instead of snow. Water and cloud are separable using mentioned
threshold for band 2 of MODIS and finally, NDSI has the key role in investigating snow
(Hall et al., 1998).
Necessary eligibilities of pixels for applying snow map algorithm are as follows (Riggs et al.,
2003):
 Pixels should have level1B reflection (geo-reference process and radiometric correction
should be done),
 They should belong to terrestrial region or water bodies surrounded by lands,
 Imagery should be taken in day light,
 Imagery should not be covered by cloud (applying cloud mask),
 Their approximate temperature should be less than 2830 K (applying temperature mask)

2.3.2 Cloud isolation

Liberal cloud masking just uses 7 out of 36 MODIS bands as well as 4 out of 18 old cloud
masking algorithm criteria. Before performing snow map as one of the preprocessing steps,
liberal cloud masking was applied. With regards to spectral resemblances between snow
and cloud, applying mask on image is inevitable. In liberal cloud masking, a pixel will be
considered as cloud provided that it covers one of the following criteria:
1. High cloud index introduces it as cloud
2. Heat difference index consider it as cloud
3. Visual bands reflection index proves the existence of cloud when reflection of band
(1/625, 1/628, µm) is more than 20 percent and visual band threshold is applied.
4. NDSI ≥ 0.4 and reflection of band 6 is more than 20 percent (Riggs and Hall, 2002).
In this research, since the study area is located on terrestrial area and consequently
discrimination of snow from cloud is very important and also all MODIS imagery were
taken in day time, thresholds related to terrestrial region in day time were applied.
Furthermore, water bodies were eliminated before image processing.

2.3.3 Heat masking

Heat masking is the final step before using snow map algorithm. This method was
introduced on 3rd October 2001 and resulted in eliminating many of incorrect land cover
classified as snow. In MODIS version 3, a threshold of 2770 K was used whereas in version 4
this value increased to 2830 K. Every individual pixel of band 31 with a temperature more
than threshold of version 4 is not classified as snow (Kamanpoon, 2004). Heat masking is
used to remove ambiguity between snow and other phenomena such as water bodies, sand
362 Remote Sensing – Applications

and cloud (Zhou et al., 2005). In this part using calculated apparent temperature for band 31
and applying 2830K threshold, heat masking is performed after new cloud masking
algorithm and before snow map algorithm.

3. Results
Images resulted from snow map algorithm before and after applying Liberal cloud mask
related to February and March are illustrated in Figure 3. Right column shows images
before liberal cloud masking and middle column show them after masking. Left column
shows false color images which are made by combining visual and infrared channels of
MODIS according to method introduced by Miller et al. (2004). Lands without snow cover,
with snow cover, low height clouds and higher clouds appear as green, white, yellow and
violet tones, respectively. False color image help to recognize cloudy regions on image as
well as cloud height.

Fig. 3. Snow area before applying the liberal cloud mask (right column) and after applying
the Liberal cloud mask (middle column) and false color composite (left column).
Predictability of Water Sources Using Snow
Maps Extracted from the Modis Imagery in Central Alborz, Iran 363

4. Discussion
In this part, using NDSI, topographic data and data gathered from snow measurement
stations, snow map algorithm alone and together with Liberal cloud masking were
separately interpreted. Ground-based snow measurement data and their corresponding
points on images resulted from snow map algorithm as well as images resulted from snow
map algorithm together with Liberal cloud masking in different dates are illustrated in table
2. In fact error matrix is drawn for each image and results have been surveyed.

Date of Snow Snow No Total Accuracy Snow Snow No Total Accuracy

acquisition map snow % map snow %
(masking)
13/3/2006 Snow 12 2 14 77 Snow 12 0 12 100
No snow 0 4 4 100 No snow 0 6 6 100
Total 12 6 18 Total 12 6 18
Accuracy 100 66 88 Accuracy 100 100 100
% %
15/3/2006 Snow 12 2 14 85 Snow 12 0 12 100
No snow 0 4 4 100 No snow 0 6 6 100
Total 12 6 18 Total 12 6 18
Accuracy 100 66 88 Accuracy 100 100 100
% %
Table 2. Evaluation of accuracy of snow gauge obtained from snow map algorithm and
snow map algorithm with attending the Liberal cloud mask using earth data in 6th of
February, 13th and 15th of March

Results demonstrate that in both images of 13th and 15th March in which snow map was
applied, the number of points classified as snow is more than the time when applying snow
map algorithm; adding cloud masking to snow map algorithm reduces this number. It
means that regions which are incorrectly classified as snow by snow map algorithm can be
categorized as cloud after adding cloud masking. Furthermore, no snow regions identified
as snow in snow-gauging station and snow map algorithm with Liberal cloud masking are
more than those no snow regions that are not classified as snow without applying Liberal
cloud masking. So it can be concluded that snow map algorithm shows some regions as
snow despite the fact that they are clouds. However, cloud masking can detect them and
classify as cloud. Error matrix demonstrates that accuracy of snow map algorithm increases
by applying cloud mask (Riggs and Hall, 2002; Ault et al., 2006; Hall and Riggs, 2007).
Overall, results from snow map algorithm together with Liberal cloud masking are more
compatible with data gathered from ground-based stations.
One of the factors affecting accuracy of snow detection is clouds which cover snow surface.
These clouds are distinguishable by Liberal cloud masking provided that they are
transparent and thin (Riggs and Hall, 2002; Ault et al., 2006). In images related to 21st
February and 8th March, the observed cloud is thick and far from the Earth. False color
images show that clouds are far from the Earth surface in both mentioned images so they
can be detected and classified correctly by Liberal cloud masking. However, there is snow
under these clouds and should be considered as snow. Data from ground-based snow
364 Remote Sensing – Applications

measurement and their corresponding points on images related to 21st February and 8th
March resulted from snow map algorithm before and after applying Liberal cloud masking
is shown in Table 3. As is shown in Table 3, field survey data are different from results
obtained as a result of snow map algorithm together with Liberal cloud masking. In this
situation, considering neighborhood effect, topographic factors and false color images,
clouds over snow can be distinguishable and classify them as snow. Of course,
neighborhood and topographic factors can be helpful when the cloud is smaller that total
area of snow.

Date of Snow Snow No Total Accuracy Snow Snow No Total Accuracy

acquisition map snow % map snow %
(masking)
21/2/2006 Snow 15 2 17 77 Snow 12 0 12 100
No snow 0 1 1 100 No snow 3 3 6 50
Total 15 3 18 Total 15 3 18
Accuracy 100 66 88 Accuracy 83 16 83
% %
8/3/2006 Snow 12 2 14 85 Snow 8 0 8 100
No snow 0 4 4 100 No snow 4 6 10 55
Total 12 6 18 Total 1 6 18
Accuracy 100 66 88 Accuracy 66 100 77
% %
Table 3. Evaluation of accuracy of snow gauge obtained from snow map algorithm

There is a negligible difference before and after applying Liberal masking images covered
by low height clouds (e.g. image of 4th March) (Table 4). It means that in this situation snow
map algorithm with and without liberal cloud masking has the same result. So it can be
concluded that snow map algorithm is able to detect low height clouds because the spectral
diagram of low height clouds are different from that of snow in visual and infrared
spectrum range. Data gathered from ground-based snow measurement and its
corresponding points on images resulted from snow map algorithm as well as images
resulted from snow map algorithm together with liberal cloud masking is shown in Table 4.

Date of Snow Snow No Total Accuracy Snow Snow No Total Accuracy

acquisition map snow % map snow %
(masking)
4/3/2006 Snow 11 0 11 100 Snow 11 0 11 100
No snow 1 6 7 85 No snow 1 6 7 85
Total 12 6 18 Total 12 6 18
Accuracy 91 100 94 Accuracy 91 100 94
% %
Table 4. Evaluation of accuracy of snow gauge obtained from snow map algorithm and
snow map algorithm with attending the Liberal cloud mask using earth data in 4th of March
Predictability of Water Sources Using Snow
Maps Extracted from the Modis Imagery in Central Alborz, Iran 365

In order to show NDSI ability in isolation of cloud from snow, variation range of NDSI in
regions which are identified as cloud using new cloud masking is compared with variation
in regions where classified as snow by use of snow map algorithm. As it can be found in
diagram NDSI variation in regions where identified as cloud and regions where classified as
snow has some overlaps so NDSI cannot distinguish between snow and cloud in these
regions (Figure 4).

Fig. 4. The variation range of DSI in cloudy and snowy area on images of 25th February and
8th March

Figure 5 shows comparison of NDSI variation for image taken in 4th March which has only
low height clouds in regions which are identified as cloud using new cloud masking with
variation in regions where classified as snow using snow map algorithm. As it can be seen
in diagram NDSI variation in regions where identified as cloud and regions where classified
as snow are absolutely separable so NDSI can distinguish between snow and cloud in these
regions and act similar to new cloud mask (Figure 5).
As a general rule, the amount of snow increases in higher elevations so if classification of
snow and cloud is done perfectly, percentage of pixels related to snow should increase in
higher elevation. This rule can be used to evaluate the accuracy of outputs resulted from
snow map algorithm alone and together with Liberal mask algorithm. Figure 6 shows the
relative frequency of snow pixels in each altitudinal zone. As it can be seen in this Figure,
ascending trend occur whenever new cloud mask is applied together with snow map
algorithm. In fact, cloud masking leads to the better identification of cloud pixels and
prevents these pixels to be classified as snow. However ascending trend will not happen in
mentioned diagram when snow map algorithm is used alone because some cloud pixels are
categorized as snow incorrectly.
366 Remote Sensing – Applications

Fig. 5. The variation range of NDSI in cloudy and snowy area on images of 4th March

Fig. 6. Relative frequency percentage of snow pixels in each altitude class; right: the
obtained images of snow map algorithm; left: the obtained images of snow map algorithm
accompanying the Liberal clod mask
Predictability of Water Sources Using Snow
Maps Extracted from the Modis Imagery in Central Alborz, Iran 367

5. Conclusion
Reviewing data resulted from ground-based snow measurements in addition to results from
snow map algorithm and Liberal cloud mask, it can be concluded that snow map algorithm
cannot detect some types of cloud and classify them as snow (Zhou et al., 2005; Riggs and
Hall, 2002; Ault et al., 2006; Hall and Riggs, 2007), reducing the accuracy of maps produced
for snow detection. Clouds which are not detected by snow map algorithm are those include
ice particles in high elevations (Taghvakish, 2005). Using Liberal cloud masking can largely
solve this problem and prevent some types of clouds to be categorized as snow. The
accuracy of maps is increased approximately 10% in comparison with other methods. In
images including only low elevation clouds, cloud masking cannot make better results;
therefore it can be concluded that these kinds of clouds can be detected by snow map
algorithm alone. Also, results from applying NDSI shows that some types of clouds are
categorized in the same class as snow, so NDSI cannot distinguish between snow and cloud.
However, those clouds in low elevation can be detected from snow.
Altitudinal parameter is another tool in order to evaluate the accuracy of snow map
algorithm and Liberal cloud masking. An ascending trend in frequency of snow pixels is
evident whenever cloud masking is used in addition to the snow map algorithm.
In summary, it can be said that although low height clouds are separable by snow map
algorithm, some types of clouds cannot be detected by snow map algorithm alone and thus,
application of cloud masking is inevitable. These are clouds which are in high elevation and
include ice particles (Taghvakish, 2005). Finally, in some cases even cloud masking cannot
distinguish between snow and ice particles (Ault et al., 2006, Riggs and Hall, 2002 ,
Taghvakish, 2005).

6. References
Ackerman S A, Strabala K I, Menzel P W P, Frey R A, Moeller C C and Gumley L E, 1998:
Discriminating clear sky from clouds with MODIS, Journal of Geophysical Research,
103(D24):32,141-32,157.
Adhami S, 2005. Application of remote sensing and geographic information system in snow
cover (Agichay). Unpublished MSc Thesis, University of Tabriz (In Persian).
Ault T W, Czajkowski K P, Benko T, et al., 2006. Validation of the MODIS snow product and
cloud mask using student and NWS cooperative station observations in the Lower
Great Lakes Region. Remote sensing of Environment 105: 341-353.
Dadashi Khanegha S, 2008. Appointment of snow cover using image processing techniques.
Unpublished MSc Thesis. University of Shahid Beheshti (In Persian).
Hall D K, Tait A B, Riggs G A, Salomonson V V, Chien J, Andrew Y L, and Klein G. 1998.
Algorithm Theoretical Basis Document (ATBD) for the MODIS Snow-, Lake Ice-
and Sea Ice-Mapping Algorithms, MODIS Algorithm Theoretical Basis Document
Number ATBD-MOD-10, NASA Goddard Space Flight Center, 1998.
Hall D K and Riggs G, 2007. Accuracy assessment of the MODIS snow products. Hydrological
Processes 21:1534-154.
Hall D K, Tait A B, Foster J L, Change A T C, and Allen M, 2000. Intercomparison of
satellite-derived snow-cover maps. Annals of Glaciology 31:396-376.
368 Remote Sensing – Applications

Kamanpoon S, 2004. Hydrological modelling using MODIS data for snow covered area in
the Northern Boreal Forest of Manitoba, MSc Thesis.
King M D, Platnick S, Yang P, Arnold G T, Gray M A, Riedi J C, Ackerman S A, Liou K,
2004. Remote sensing of liquid water and ice cloud optical thickness and effective
radius in the Arctic: Application of airborne multispectral MAS Data. Journal of
Atmospheric and Oceanic Technology 21: 857-875.
Klein A G and Barnett A C, 2003. Validation of daily MODIS snow cover maps of the Upper
Rio Grande River Basin for the 2000-2001 snow year. Remote Sensing of Environment
86:162-176.
Lee S, Klein A G and Over T M, 2001. A Comparison of MODIS and NOHRSC snow cover
products for simulating streamflow using the Snowmelt RunOff Model in:
http://www.modissnowice.gsfc.nasa.gov
Miller S D and Lee T F, 2004. Satellite-Based imagery techniques for daytime cloud/snow
delineation from MODIS. Journal of applied meteorology. volume 44.
Pfister R and Schneebeli M, 1999. Snow accumulation on boards of different sizes and
shapes. Hydrological Processes 13:2345-2355.
Hall D k, Foster J L, Robinson D A, Riggs G A, 2004. Merging the MODIS and RUCL
monthly snow cover records. In Proceeding of IGARSS 04, Anchorage, AK,
September 2004.
Rayegani B, 2005. Investigation on snow cover changes and estimation snowmelt runoff
using MODIS data. Unpublished MSc Thesis. Isfahan University of Technology (In
Persian).
Riggs G and Hall D K, 2002. Reduction of cloud obscuration in the MODIS snow data
product, 59th Eastern Snow Conference, Stowe, Vermonte, USA.
Riggs G A , Haii D K and Salomonson V V, 2003. MODIS snow products users guide.
Available at: http://www.modis-snow-ice.gsfc.nasa.gov/sug.pdf.
Strabala K, 1999. MODIS Cloud Mask User’s Guide’, User’s Guide.
Singh P and Singh V P, 2001. Snow and Glacier Hydrology, Kluwer Academic Publishers,
p742.
Taghvakish S, 2005. Supervision of snow cover using satellite images. Unpublished MSc
Thesis, University Of Tehran (In Persian).
Vermote E F and Kotchenova S, 2008. Atmospheric correction for the monitoring of land
surfaces. . Journal of Geophysical Research -ATMOSPHERES, 113(D23), D23S90.
Zhou X, Xie H and Hendrick J M H, 2005. Statistical evaluation of remotely sensed snow
cover products with constraints from streamflow and SNOTEL measurements,
Remote Sensing of Environment 94: 214–231.
 16

Remote Sensing of Cryosphere

Shrinidhi Ambinakudige and Kabindra Joshi
Mississippi State University
USA

1. Introduction
The cryosphere is the frozen water part of the Earth’s system. The word is derived from the
Greek “kryos,” meaning cold. Snow and ice are the main ingredients of the cryosphere and
may be found in many forms, including snow cover, sea ice, freshwater ice, permafrost, and
continental ice masses such as glaciers and ice sheets. Snow is precipitation made up of ice
particles formed mainly by sublimation (NSIDC, 2011). Ice is the key element in glaciers, ice
sheets, ice shelves and frozen ground. Sea ice forms when the ocean water temperature falls
below freezing. Permafrost occurs when the ground is frozen for a long period of time, at
least two years below 00 C, and varies in thickness from several meters to thousands of
meters (NSIDC, 2011). Glaciers are thick masses of ice on land that are caused by many
seasons of snowfall. Glaciers move under their own weight, the external effect of gravity,
and physical and chemical changes. The cryosphere lowers the earth’s surface temperature
by reflecting a large amount of sunlight, stores fresh water for millions of people, and
provides habitat for many plants and animals.
Apart from the Arctic and Antarctic regions, the cryosphere is mainly a high altitude
phenomenon. It is found on Mount Kilimanjaro in Africa, the Himalayan mountain range,
high mountains of United States, and in Canada, Russia, Japan, and China. Researchers in
the cryosphere are often hindered by the lack of accessibility due to the rugged terrain. In
such cases, remote sensing technologies play an important role in cryosphere research.
These techniques are imperative for researchers studying glacial retreat and mass balance
change in relation to global climate change.
The cryosphere has a significant influence on global climate and human livelihoods. Change
in spatial and temporal distribution of the cryosphere influences the water flow in the
world’s major rivers. Among the various parts of the cryosphere, glaciers play the most
important role in climate change studies since glacier recessions are indicators of global
climate change (Oerlemans et al., 1998; Wessels et al., 2002; Ambinakudige, 2010). Retreating
glaciers can pose significant hazards to people (Kaab et al., 2002). Glacier retreat often lead
to the formation of glacial lakes at high altitudes, the expansion of existing lakes, and the
potential for glacial lake outburst floods (GLOFs) (Fujita et al., 2001; Bajracharya et al., 2007).
A GLOF is the sudden discharge of a huge volume of water stored in a glacial lake due to
huge ice falls, earthquakes, avalanches, rock fall or failure of a moraine dam (Grabs &
Hanisch, 1993). There are more than 15,000 glaciers and 9,000 glacial lakes in the Himalayan
mountain ranges of Bhutan, Nepal, Pakistan, China and India (Bajracharya et al., 2007). All
370 Remote Sensing – Applications

these countries within the Himalayan region have at some time or another suffered a flood
from a glacial lake outburst causing loss of property and lives, and these floods can be
disastrous for the downstream riparian area (Richardson & Reynolds, 2000; Bajracharya et
al., 2007). The significance of the glaciers as fresh water resources for millions of people is
another reason to justify the continuous monitoring of these glaciers (Shiyin et al., 2003).
Therefore, monitoring glaciers has significant importance both in understanding global
climate change and in sustaining the livelihoods of the people downstream of the glaciers.
This chapter explores the use of remote sensing technologies in studies of the cryosphere
and particularly in glaciers. First, we will discuss remote sensing sensors that are effective in
monitoring glaciers. Then we will discuss the global effort to create glacier data, using
remote sensing tools to delineate glacier areas, estimate volume and mass balance.

2. Remote sensors for monitoring glaciers

Remote sensing methods are more convenient than field methods to measure changes in
glaciers. Studies have used Landsat (Ambinakudige, 2010), SPOT (Berthier et al., 2007),
Terra ASTER (Rivera & Casassa, 1999; Kaab, 2007), IRS (Kulkarni et al. 2011), and ALOS
(Narama et al., 2007) sensors successfully to measure glacial parameters. High-resolution
satellite data such as IKONOS (Huggel et al., 2004) and Quickbird (Schmidt, 2009) have also
been used in studies on glaciers. ASTER, SPOT5, IRS-1C, Resourcesat1 and 2, CORONA
KH-4, KH 4A and KH 4B satellites have the capability to acquire stereoscopic images from
which elevation data can be extracted for monitoring glacial surfaces in three dimensions
(Racoviteanu, et al. 2008). Digital elevation models developed from these stereoscopic
images can be used in measuring volume and mass balance change in glaciers.

2.1 Glacial mapping

Glacial mapping using remote sensing initially involved manual digitization of glacier
boundaries on a false color composite (FCC) of Landsat MSS and TM images in Iceland
(Williams, 1987) and Austria (Hall et al., 1992). Figure 1, is a picture taken by the astronauts
in the international space station shows the snow and ice in Colonia glacier and its
surroundings in Chile (International Space Station, 2000).
Snow has high reflectivity in the visible wavelength region and relatively less reflectivity in
middle and shortwave infrared regions (Pellika and Rees, 2010). Freshly fallen snow has the
highest reflectance in the visible and near-infrared wavelengths. Firn (partially compacted
snow) has 25-30 % less reflectance than snow. Ice in glaciers has high reflectance in blue
(400- 500 nm) and green (500 – 600 nm) wavelengths. However, in red (600-700 nm) the
reflectance of ice is near zero. Debris in glaciers will significantly reduce the reflectance
(Pellika and Rees, 2010). The spectral reflectance properties of snow also depend on time
and season of the year. Sharp changes in the reflectance of snow can be seen when the
melted snow recrystallizes to form firn and the density of snow changes. Similarly, albedo,
which is the ratio of radiation reflected from a surface to the radiation incident on that
surface, also varies among different cryospheric surfaces. Snow has high albedo (0.8 – 0.97)
while dirty ice has low albedo (0.15-0.25). A high albedo value helps to reflect a huge
amount of sunlight, which otherwise would have heated the earth’s surface. Snow and ice
reflectance is the main characteristic that is measured using remote sensing techniques. This
Remote Sensing of Cryosphere 371

Fig. 1. Astronaut photograph of Colonia Glacier, Chile. Photographer: International Space

Station (2000).

characteristic assists in delineating glacier boundaries and classifying various cryospheric

surface types (Pellika and Rees, 2010).
In figure 2, the Landsat TM bands 1 to 6, acquired on 25 April 2010, are shown to compare
spectral characteristics of glaciers. This figure presents the area around the Imja glacier in
the Sagarmatha National Park in the Himalayas of Nepal. TM1 (0.45 – 0.52 µm) is useful to
distinguish snow/ice in cast shadow, and also in mapping glacier lakes. Snow and firn areas
get saturated in TM1. TM2 (0.52 – 0.60 µm) and TM3 (0.63 – 0.69 µm) have very similar
spectral reflectance. TM2 is also useful in distinguishing snow and ice in cast shadow. TM3
is used in band ratio such as the normalized difference vegetation index (NDVI), which
helps in classification of ice and snow in areas of dense vegetation. TM4 (0.76-0.90 µm) in the
near infrared wavelength region has less reflectance from snow than TM2 and TM3. The
clean ice region looks darker in near infrared band, indicating lower reflectance due to the
presence of water at the surface (Hall et al., 1988). In TM5, the snow-covered area absorbs
nearly all radiation and appears almost dark. This band is also useful in identifying clouds.
The thermal band TM6 (2.08-2.35 µm) registers thermal emissions from the surface. Debris
has a higher temperature and thus brighter pixels. Thick debris on ice can be easily
distinguished as it will have a higher temperature (Pellika and Rees, 2010).
The high reflectance of the snow compared to the ice makes it easy to separate snow and ice.
Snow and clouds are often difficult to distinguish when single imagery is used. Clouds and
snow have similar reflectance at wavelengths below 1 µm in the near infrared region. The
distinction between snow and ice is clearer near 1.55 and 1.75 µm. Therefore a ratio of two
372 Remote Sensing – Applications

Fig. 2. Six Landsat TM bands showing Imja and surrounding glaciers in the Himalayas.
Remote Sensing of Cryosphere 373

spectral bands in the visible (VIS) and shortwave near infrared (SWIR) regions are
commonly used in automated mapping of snow and ice. A Normalized Difference Snow
Index (NDSI) calculated as (VIS-SWIR)/(VIS+ SWIR) helps to separate snow and ice from
darker areas such as rocks and soils. Whereas the visual spectrum band covers a wavelength
of 0.57µm, the shortwave infrared band covers 1.65µm. If the NDSI value exceeds 0.4, it is
assumed there is snow cover on the ground (Dozier 1984). However, seasonal variation in
NDSI value for snow has also been observed. A threshold of 0.48 in July and 0.6 in
September was observed as an optimal threshold during the field investigation in Abisko,
Sweden (Vogel 2002). The Normalized Difference Water Index (NDWI) calculated as (NIR-
VIS)/(NIR+VIS) is useful to differentiate water from snow, ice and other physical features.
NDWI is very useful in detecting formation of new, as well expansion of existing, glacial
lakes (Huggel et al., 2002).
Band ratios such as Landsat TM3/TM5 or TM4/TM5 are also helpful in mapping glacial
areas. The TM4/TM5 ratio is more appropriate for clean-ice glacier mapping (Paul and
Kaab, 2001), whereas TM3/TM5 works better in areas of dark shadow and thin debris cover
(Andreassen et al., 2002). Both NDSI and ratio methods have similar robust outcomes in
glacier mapping and are recommended. NDSI and band ratio methods often misclassify
debris-covered glacier ice because of the similarity in spectral signature to the surrounding
debris Band ratio VIS/NIR also often misclassifies proglacial lakes. The band ratio
NIR/SWIR is good for classifying only clean glacier ice (Bhambri and Bolch, 2009).
Therefore, manual corrections have to be made after classifying glaciers using either NDSI
or ratio methods. High resolution panchromatic images used in manual corrections can help
to delineate the precise boundaries of glaciers. Similarly, figure 3 also indicates spectral
responses of different types of snow and water.

(Data Source: ENVI Spectral Library derived from John Hopkins University
http://asterweb.jpl.nasa.gov/speclib/)
Fig. 3. Spectral responses of different types of snow and water
374 Remote Sensing – Applications

3. Digital elevation models in glacier mapping

Debris-covered glaciers can be mapped effectively by using morphometric parameters
derived from DEM and thermal bands (Ranzi et al., 2004). The surface temperatures of the
debris on glacier ice are cooler than the debris outside the ice, which can be identified by a
thermal image. AVHRR and MODIS satellites have coarse resolution thermal bands; the
Landsat TM/ETM+ satellite, on the other hand, has a higher resolution thermal band.
Using morphometric characteristics like slope can help to separate a debris-covered
glacier from its surrounding moraines (Bishop et al., 2001). DEMs help in estimating
morphometric characteristics such as slope, glacier profile curvature, and aspect. DEMs
are most important for the estimation of volume change for inaccessible glacier regions
(Bolch et al., 2008).
DEMs are generated from topographic maps, spaceborne optical stereo data, synthetic
aperture radar (SAR) imagery, spaceborne radar and laser altimetry, such as LiDAR data.
Terra ASTER optical stereo satellite data with a spatial resolution of 15 m has been used
successfully in generating DEM to study glaciers (Kargel et al., 2005). SPOT-5, ALOS
PRISM, CartoSat I and II, IKONOS, Quickbird and GeoEye-1 satellites also have stereo
capability and can be used in creating DEMs. ASTER and ALOS PRISM produce along-
track stereo images and are capable of simultaneous image acquisition. Other sensors
provide across-track acquisition with a time lag, which causes problems in certain
situations, such as under different atmospheric conditions (Bhambri & Boch, 2009).
Measuring surface changes using DEMs, known as indirect geodetic methods, is a useful
method to calculate a glacier’s volume and mass balance changes (Etzelmüller & Sollid,
1997).
Another technology used to study the cryosphere is radar, which is an active microwave
system composed of a transmitter and a receiver. Radar imagery is not confined only to
daylight, cloud-free conditions, and thus has advantages in estimating glacial change. The
first spaceborne SAR was the SEASAT satellite, launched in 1978. ERS-1/2, Envisat – ASAR,
Radarsat-1, ALOS PALSAR, Radarsat-2 and TerraSAR-X are some of the satellites with
spaceborne SAR launched after the initial success of SEASAT. Due to its all-day, all-weather
imaging capabilities, large spatial coverage, and ability to measure minute changes on the
earth’s surface, SAR has a major advantage in glacier studies. Digital elevation models
(DEM) created using SAR images are useful in measuring glacier mass balance, glacier
velocity, and snow density.
Light Detection and Ranging (LiDAR) is also used in remote sensing of glaciers. High
density measurement with a high vertical accuracy (10 cm) allows a very detailed
representation of terrain in LiDAR imagery. LiDAR imagery also comes with accurate
positional information because of the global positioning system receiver in the LiDAR
instruments on aircraft.

4. The geodetic method of measuring glacier mass balance

The mass balance of a glacier is referred to as the total loss or gain in glacier mass at the end
of a hydrological year (Cuffey and Paterson, 2010). Glacier mass balance is the link between
climate and glacier dynamics (Kaser et al., 2002). Mass balance change is a direct reaction to
Remote Sensing of Cryosphere 375

atmospheric conditions. The specific mass balance can be compared directly between
different glaciers. This makes it easier than using length changes to establish a link to
climate data (UNEP, 2008). Glacier mass losses affect local hydrology and are thus important
for regional water supplies and assessing global sea level rise.
In the geodetic method, changes of glacier volume are measured from maps and elevation
models, unlike the direct or glaciological methods that derive mass changes from ground-
based spot measurements. With the development of remote sensing techniques, digital
elevation models (DEM) are commonly used in the geodetic method. The difference in
DEM values in two years is multiplied by the glacier accumulation area to obtain volume
change (dV). The difference in volume (dV) multiplied by the density of ice, generally
considered to be 850-900 Kg/m3, would provide the mass balance (Cuffey & Paterson,
2010).
The accumulation and ablation area in a typical glacier is separated by an equilibrium line
altitude (ELA). This line divides the accumulation zone (the higher reaches of a glacier
where there is a net mass gain) and the ablation zone (the lower reaches where mass is
lost). ELA is the elevation at which mass is neither gained nor lost through the course of a
hydrological year. At the end of the season, the elevation at which there was no net gain
or loss is identified as the ELA. The snow line altitude (SLA) divides the ice in the ablation
zone from the snow in the accumulation zone. Since ice and firn have lower albedo, SLA
can be determined using remote sensing images (Khalsa et al., 2004; Racoviteanu et al.,
2007).
The geodetic approach has been used in several studies based on historical topographic
maps and DEMs derived from SPOT imagery (Berthier et al., 2007), SRTM (Racoviteanu et
al., 2007), ASTER (Rivera & Cassassa, 1999; Kaab, 2007). Studies have also used high
resolution DEMs derived from ALOS PRISM and Corona (Narama et al., 2007) to estimate
mass balances with the geodetic method.
The two bands in ASTER VNIR, 3N and 3B, generate an along-track stereo pair with a base-
to-height (B/H) ratio of about 0.6. Studies have found that the DEM accuracy has a linear
relation with terrain slopes (Toutin, 2008; Bolch et al., 2004; Racoviteanu et al., 2007). DEMs
created using ASTER images on Mt. Fuji, Japan; in the Andes Mountains, Chile-Bolivia; at
San Bernardino, CA and Huntsville, AL resulted an accuracy of ±5 m, ±10m, ± 6m and ±
1.5m respectively (Hirano, 2003). Therefore, accuracy, number and distribution of GCPs are
required to create an accurate DEM.
The ALOS data have been available since January 2006. ALOS has three remote-sensing
instruments. The Panchromatic Remote-Sensing Instrument for Stereo Mapping (PRISM)
sensor of ALOS data is suitable for digital elevation mapping. It consists of three
independent telescopes for forward, nadir and backward view, and each telescope
provides 2.5m spatial resolution. The Advanced Visible and Near Infrared Radiometer
type 2 (AVNIR-2) sensor collects data with 10m resolution and is suitable for glacier
mapping. The accuracy of elevation extracted from ALOS PRISM is 5 m (Racoviteanu et
al., 2008).
The three images (forward, nadir and backward) that ALOS provides are required for
creation of DEM (Ye, 2010). The base-to-height (B/H) of PRISM on ALOS is set to 1.0
376 Remote Sensing – Applications

(forward view + backward view) and 0.5 (sidelong view + nadir view). ALOS images have
been used to produce highly accurate DEMS. The DEMs created for the northern slope of
Qomolangma in the Mt. Everest region had a mean elevation difference of 1.7m with a DEM
created using topographic maps in non-glaciated areas. The mean difference between Aster
and ALOS images was found to be about 45m (Ye, 2010).
Once the outlines of the glaciers are delineated, they can be combined with DEM to derive
glacier parameters such as length, termini elevations, and volume. DEMs derived from
SPOT5, ASTER, CORONA or ALOS PRISM can be used in mass balance studies.

5. Global Land Ice Measurements from Space (GLIMS)

In an effort to analyze the glacial change throughout the world, a global-level consortium,
the Global Land Ice Measurements (GLIMS), has established a database at the National
Snow and Ice Data Center (NSIDC) in Boulder, Colorado (Raup et al., 2007). Under GLIMS,
12 regional centers are working to acquire satellite images, analyze them for glacial extent
and changes, and assess change data for causes and implications for people and the
environment.
GLIMS is an international consortium established to monitor the world’s glaciers. Although
GLIMS is making use of multiple remote-sensing systems, ASTER (Advanced Spaceborne
Thermal Emission and Reflection Radiometer) satellite images are the major data input in
GLIMS database. The GLIMS team has put together a network of international collaborators
who analyze imagery of glaciers in their regions of expertise. Collaborators provide digital
glacier outlines and metadata. Data also include snow lines, center flow lines, hypsometry
data, surface velocity fields, and literature references. The National Snow and Ice Data
Center archives the data provided by the regional centers.
The GLIMS team also developed tools to aid in glacier mapping, such as GLIMSView, which
is an open-source, cross-platform application designed to support and standardize the
glacier digitization process. It allows regional centers to transfer data to the National Snow
and Ice Data Center for incorporation into the GLIMS glacier database. Users can view
various types of satellite imagery, digitize glacier outlines and other material units within
the images, attach GLIMS-specific attributes to segments of these outlines, and save the
outlines to ESRI shapefiles. GLIMSView is free and available at
http://www.glims.org/glimsview/.

6. Conclusions
Satellite remote sensing of the cryosphere has progressed over the last five decades. It has
helped us to understand the global distribution of the cryosphere, variation and trends in
snow cover, sea ice, and glaciers. We have a pretty decent map of the cryosphere. Remote
sensing has helped in rapid assessment of glaciers in hostile ground conditions in areas such
as Antarctica, the Artic and alpine glaciers.
There are several challenges in remote sensing of the cryosphere. Acquiring cloud-free
satellite imagery is still challenging. Synthetic aperture radar (SAR) imagery has received a
great deal of attention in recent years as it can provide cloud-free data. SAR interferometry
has been used successfully in areas such as glacier motion and topographical mapping. The
Remote Sensing of Cryosphere 377

use of radar has been gaining more attention recently. Ground penetrating radars are being
used to study the internal structure and bedrock configuration of glaciers.
For most part, in situ measurement of the cryosphere is often not a viable option, so the
focus of cryosphere study remains on the use of remote sensing techniques. The World
Glacier Monitoring Service (WGMS) coordinates the global glacier observation strategy with
the help of the Global Land Ice Measurement from Space project and the European Space
Agency’s Global Glacier Project.
Spaceborne remote sensing techniques in the last five decades have shown tremendous
advancement. From Landsat to InSAR imagery, the remote sensing technology has helped
in understanding and mapping the cryosphere. Many of these data are available free or for
low cost; some of them are very expensive, and using them requires specialized skill. With
the increase in computer processing power, the potential for the collection, storage,
transmission and processing of remotely sensed data on the cryosphere has improved.
Changes in glaciers provide evidence of climate change, and therefore glaciers play a key
role in early detection of global climate-related observations (WGMS, 2011). Glacier change
will impact global sea level fluctuations and other natural hazards. These environmental
changes require international glacier monitoring efforts to make use of remote sensing and
geo-informatics along with the more traditional field observations.

7. References
Ambinakudige, S. (2010). A study of the Gangotri glacier retreat in the Himalayas using
Landsat satellite images. International Journal of Geoinformatics 6 (3), pp. 7-12.
Andreassen, L.M., Elvehoy, H. & Kjollmoen, B. (2002). Using aerial photography to study
glacier changes in Norway. Annals of Glaciology 34: pp. 343-348.
Bajracharya, S.R., Mool, P.K., & Shrestha, B. (2007). Impact of climate change on Himalayan
glaciers and glacial lakes: case studies on GLOF and associated hazards in Nepal and
Bhutan. Kathmandu, International Centre for Integrated Mountain Development
and United Nations Environmental Programme Regional Office Asia and the
Pacific. (ICIMOD Publication 169).
Berthier, E., Arnaud, Y., Kumar, R., Ahmad, S., Wangnon, P., & Chevallier, P. (2007). Remote
sensing estimates of glacier mass balances in the Himachal Pradesh (Western
Himalaya, India). Remote Sensing of Environment 108:327-338.
Bhambri, R., & Bolch, T. (2009). Glacier Mapping: A Review with special reference to the
Indian Himalayas, Progress in Physical Geography 33(5), 672–704.
Bolch, T., Buchroithner, M. F., Pieczonka, T., & Kunert, A. (2008). Planimetric and
volumetric glacier changes in Khumbu Himalaya since 1962 using Corona, Landsat
TM and ASTER data, Journal of Glaciology, 54, pp. 592–600.
Bishop, M.P., Bonk, R., Kamp, U. & Shroder, J.F. (2001): Topographic analysis and modeling
for alpine glacier mapping. Polar Geography, 25: 182-201.
Cuffey, K.M & Paterson, W.S.B. (2010). The Physics of Glaciers. 4th ed. Academic Press.
Dozier, J., (1984). Snow reflectance from Landsat-4 Thematic Mapper. IEEE Transactions on
Geoscience and Remote Sensing. 22(3), pp. 323-328.
378 Remote Sensing – Applications

Etzelmüller, B. & Sollid, J.L. (1997). Glacier geomorphometry - an approach for analysing
long-term glacier surface changes using grid-based digital elevation models. Annals
of Glaciology, 24, pp. 135-141.
Fujita, K. Kadota, T. Rana, B., Kayastha, R.B., & Ageta, Y. (2001). Shrinkage of glacier Ax010
in Shorong region, Nepal Himalayas in the 1990s. Bulletin of Glacier Research 17, pp.
51-54.
Grabs, W. & Hanisch, J. (1993). Objectives and prevention methods for glacier lake outburst
floods (GLOFs). Snow and Glacier Hydrology : Proceedings of the Kathmandu
symposium, November 1992. IAHS Publication, 218.
Hall, D.K., Williams, R.S., & Bayr, K.J. (1992). Glacier récession in Iceland and Austria. EOS,
73(12), 129.
Hirano, A. (2003). Mapping from ASTER stereo image data: DEM validation and accuracy
assessment. ISPRS Journal of Photogrammetry and Remote Sensing, 57(5-6), pp. 356-
370.
Huggel, C., Kääb, A., Haeberli, W., Teysseire, P. & Paul, F. (2002). Remote sensing based
assessment of hazards from glacier lake outbursts: a case study in the Swiss Alps.
Canadian Geotechnical Journal, 39, pp.316-330.
Huggel, C., Kääb, A., & Salzmann, N. (2004). GIS‐based modeling of glacial hazards and
their interactions using Landsat‐TM and IKONOS imagery. Norsk Geografisk
Tidsskrift - Norwegian Journal of Geography, 58(2), pp 61-73.
International Space Station. (2000). Colonia glacier: from the glacier photograph collection.
Boulder, Colorado USA: National Snow and Ice Data Center/World Data Center
for Glaciology. Digital media.
http://eol.jsc.nasa.gov/scripts/sseop/photo.pl?mission=ISS001&roll=E&frame=51
07
Kargel, J.S, Abrams, M., Bishop, M., Bush, A., Hamilton, G., Jiskoot, H., Kaab, A., Kieffer,
H.H., Lee, E., Frank, P., Rau, F., Raup, B., Shroder, J.F., Soltesz, D., Stainforth,
D., Leigh, S, & Wessels, R. (2005). Multispectral imaging contributions to global
land ice measurements from space. Remote Sensing of Environment, 99(1-2),
pp.187-219.
Kaab, A. (2007). Glacier volume changes using ASTER optical stereo. A test study in Eastern
Svalbard. IEEE Transactions on Geosciences and Remote Sensing 10:3994 -3996.
Kaab, A, Paul, M. Maisch, M., Hoelzle, M. & Haeberli, W. (2002). The new remote sensing
derived Swiss glacier inventory: II. first results. Annals of Glaciology, 34.
Kaser, G., Fountain, A., & Jansson, P. (2002). A manual for monitoring the mass balance of
mountain glaciers. International Hydrological Programme, 69.
Khalsa, S.J.S., Dyurgerov, M.B., Khromova, T., Raup, B. H., & Barry, R.G. (2004). Space-
based mapping of glacier changes using aster and gis tools. IEEE transactions on
geosciences and remote sensing, 42, pp. 2177-2183.
Kulkarni, A.V., Rathore, B.P., Singh, S. K., & Bahuguna, I. M. (2011). Understanding changes
in the Himalayan cryosphere using remote sensing techniques. International journal
of remote sensing, 32: 3, 601-615.
Remote Sensing of Cryosphere 379

Narama, C., Kaab, A., Kajiura, T., & Abdrkhmatov, K. (2007). Spatial variability of recent
glacier area and volume changes in central Asia using corona, Landsat, Aster and
ALOS optical satellite data. Geophysics Research Abstract, 9.
NSIDC (2011). Education resources. http://nsidc.org/pubs/education_resources/
Oerlemans, J., Anderson, B., Hubbard, A., Huybrechts, P., Jóhannesson, T., Knap, W. H.,
Schmeits, M., Stroeven, A. P., van de Wal S.R.W., & Wallinga, J. (1998).
Modelling the response of glaciers to climate warming. Climate Dynamics 14,
pp.267-274.
Paul, F & Kaab, A. (2005). Perspectives on the production of a glacier inventory from
multispectral satellite data in the Canadian Arctic : Cumberland Peninsula, Baffin
Island. Annals of Glaciology. 42, pp. 59-66.
Pellika P. & Rees, W.G., (2010). Remote Sensing of Glaciers. CRC Press. London.
Ranzi, R., Grossi, R., Iacovelli, l. & Taschner, T. (2004). Use of multispectral ASTER images
for mapping debris-covered glaciers within the GLIMS Project. IEEE International
Geoscience and Remote Sensing Symposium 2, 1144-1147.
Racoviteanu, A., Manley, W.F., Arnaud, Y. & Williams, M. (2007). Evaluating digital
elevation models for glaciologic applications: An example from Nevado Coropuna,
Peruvian Andes. Global and Planetary Change, 59(1-4), pp.110-125.
Racoviteanu, A.E., Williams, M.W., & Barry, R.G. (2008). Optical Remote Sensing of
Glacier Characteristics: A Review with Focus on the Himalaya. Sensors, 8, pp
3355-3383.
Raup, B., Racoviteanu, A., Khalsa, S. J. S., Helm, C., Armstrong, R. & Arnaud, Y. (2007). The
GLIMS geospatial glacier database: A new tool for studying glacier change. Global
and Planetary Change, 56, pp. 101-110.
Richardson, S.D. & Reynolds. J.M. (2000). An overview of glacial hazards in the Himalayas.
Quaternary International, 65-66, pp.31-47.
Rivera, A. & Casassa, G. (1999). Volume changes on Pio Xi Glacier, Patagonia. Global Planet
Change 22:233-244.
Schmidt, S. (2009). Fluctuations of Raikot Glacier during the past 70 years: a case study from
the Nanga Parbat massif, northern Pakistan. Journal of Glaciology, 55 (194), pp. 949-
959.
Shiyin, L., Wenxin, S., Yongping, S., & Gang, L. (2003). Glacier changes since the little ice age
maximum in the western Qilian Shan, Northwest China, and consequences of
glacier runoff for water supply. Journal of Glaciology, 49, pp.117-124.
Toutin, T. (2008). ASTER DEMs for geomatic and geoscientific applications: a review.
International Journal of Remote Sensing, 29(7), pp. 1855-1875.
UNEP. (2008). World Heritage Site: Sagarmatha National Park, Nepal. United Nations
Environment Programme.
Vogel, S.W. (2002). Usage of high-resolution Landsat-7 band 8 for single band snow cover
classification. Annals of Glaciology, 34, pp.53-57.
WGMS (2011). World glacier monitoring service. http://www.wgms.ch/
Wessels, R., Kargel, J. S. & Kiefffer. H.H. (2002). Aster measurement of supraglacial lakes
in the Mount Everest region of the Himalaya. Annals of Glaciology, 34, pp. 399-
408.
380 Remote Sensing – Applications

Williams, R. (1987). Satellite remote sensing of Vatnajökull, Iceland, Annals of Glaciology, 9,

pp. 119-125.
Ye, Q. (2010). The generation of DEM from ALOS/PRISM and ice volume change in Mt.
Qomolangma region. Geophysical Research Abstracts, 12.
 17

Remote Sensing Application in

the Maritime Search and Rescue
Jing Peng and Chaojian Shi
Shanghai Maritime University
P.R. China

1. Introduction
Maritime search and rescue (MSR- In the maritime publications, the abbreviation for search
and rescue is also SAR. Here we use MSR to distinguish it from the abbreviation for
Synthetic Aperture Radar.) became an enormous task with the vast growth of marine
transportation and other marine activities. In the year of 2006, the MSR centers and maritime
authorities in China organized and coordinated 1620 MSR operations, which involved 5322
vessels and 17498 human lives. The past few years have witnessed tremendous changes in
the organizations of maritime rescue. A large part of this evolution stems from the
involvement on an international scope and the contribution of the advanced technology.
However, current maritime search operation, especially searching people over board,
depends mostly on human eyes.
SOLAS (International convention for safety of life at sea) convention prescribes that ships must
be equipped with GMDSS (Global maritime distress and safety system) equipments, which
have improved the search and rescue. However, for many non SOLAS convention ships, such
as fishing boats and small crafts, the detection results are not very much satisfied. With the
complex sea environment, the searching of distress vessel becomes a nail-biting task. Because
of the physiological characteristics of human eyes, it is difficult for the rescuer to find small
target in the adverse background lighting, night or dark condition, wave or clustered seas.
Continuous long time observation also causes fatigue of human eyes, resulting poor sensitivity
of detection. All those factors decay the results of searching operation.
In order to improve the effect of MSR operations during the dark hours or in adverse
lighting or sea conditions, remote sensing technique is a potential approach to overcome the
limitation of human eyes in MSR, and thereby may hopefully improve the searching
performance in complex environment or in a fatigued state of human being. Regarding ship
monitoring, compared with shore-base, shipboard or airborne detecting devices, and other
visible visible or infrared monitoring methods, the Synthetic Aperture Radar (SAR) remote
sensing system possesses the capability of all-time, all weather, extensive and high
resolution for detecting ships on the sea. Especially due to its working characteristics of not
being limited by the sea surface, weather or human factors, it can detect the sea areas with
geographical remote positions and hostile environment which cannot be entered directly.
In this chapter, some remote sensing techniques and algorithms concerned with the MSR are
introduced. A Remote Sensing Monitoring System for Maritime Search and Rescue (RS-MSR)
382 Remote Sensing – Applications

is presented. This work is a part of our project—Vision Enhancement System for Maritime
Search and Rescue. The main task for the RS-MSR is to acquire general information in a wider
scale. The distress ship is detected and located for guiding the search operation. Surrounding
ships are also distinguished to coordinate the MSR operation. Some important data such as
current and sea state are retrieved to help decision-making of the operation. Section 2 proposes
the outline of the remote sensing methods for maritime search and rescue; Guided by the
systematic functions and structures of the RS-MSR described in Section 2, Section 3 introduces
the related algorithms used in RS-MSR. Section 4 describes the architecture of the remote
sensing aided system for maritime search and rescue. The experiment design and the
implementation performance are given in Section 5. Finally, Section 6 concludes the paper.

2. Outline
The primary role of the remote sensing is to provide a secondary source of information for
the MSR operations. A remote sensing monitoring system can, to some extent, overcome the
shortcomings and inadequacies of human eyes. It can also improve the searching speed and
accuracy, and is of significance in promoting rescue success rate and efficiency. It can aid the
rescuers to fulfill the task of search and rescue, especially for small targets, such as persons in
distress and life boats, and could provide a good detection and identification result.

2.1 Introduction to the system functions

According to the requirement of search and rescue, a Remote Sensing Monitoring System for
Maritime Search and Rescue (RS-MSR) is designed. Table 1 illustrates the main functions of
a RS-MSR system.

RS-MSR System demand Function

Satellite transit inquiry Satellite transit inquiry
Ship detection 1) Ship position detection
2) Ship type identification/classification
3) Ship size estimation
4) Ship heading/direction estimation
Sea state analysis Wave direction and estimation
Integrated processing 1. Non-remote sensing data fusion
- VTS (Vessel Traffic Services) report
- AIS (Automatic Identification System) report
2. Performance analysis
- Recognition rate
- Identification/classification rate
- Position error
3. Rescue position prediction
- Time prediction
- Ship speed and velocity
- Predict the rescue area according to the heading/
direction of the distress ship and the sea state
Table 1. The main functions of a RS-MSR system required
Remote Sensing Application in the Maritime Search and Rescue 383

2.2 Algorithm introduction

According to the requirement of the maritime search and rescue, the Synthetic Aperture
Radar (SAR) imageries are used for this purpose. The algorithms concerned with the system
demand are introduced as follows.

2.2.1 Distributed target detection method in the Gaussian scale-space

In the SAR images with high resolution, each target occupies several resolution units to
form area target. So detecting the ship target in the SAR images with high resolution
should regard the target as distributive target, and the assumption of point target under
the traditional radar is not suitable any more. The project here proposes a distributed
target detection method in the Gaussian scale-space. The distance relationship among the
detected objects is adopted to identify the distributed target. In the situation of hardly
estimating the background's scattering distribution or of low SNR (signal-to-noise ratio),
this method can realize the distributive target detection more effectively than CFAR
method.

2.2.2 Ship size category estimation model

In the SAR images with high resolution, the ship target can be divided into three categories
according to their dimension, among which small ships(≤50m) are represented as point
target, while the middle size ship(≤100m) as distributive target of single corner and big size
ship(≥100m) as double corner distributive target. According to the relative positions of the
corner and combining the orbit information (resolution ratio and incident angle) of the SAR
image, the length, height and the direction of the ship will be worked out.

2.2.3 Ship location correction and ship direction estimation method

The imaging geometry of the SAR imagery is slant-range projection. So, due to the
geometric distortions, such as layover, foreshortening, and shadows, there exits
measurement error between the observed position and its actual location. In this method,
using the directional texture of the wake of ship, the convergent point of the wake pattern
can be calculated, which is the actual location of the stern. Then the position correcting
parameters can be worked out.
A bow wave is the wave that forms at the bow of a ship when it moves through the water.
As the bow wave spreads out, it defines the outer limits of a ship's wake. Theoretically, the
convergent point of the bow wave’ outline must at the extended line of the ship’s heading.
The vanishing point can be calculated by the Hough transform, and the heading direction of
the ship can be calculated according to the coordinates of the bow and the stern.

2.2.4 Wave direction estimation based on the partial energy direction

Wave direction estimation can be used to analyze the sea state of the target area and supply
basis for search and rescue area decision. This method is set up on the basis of steerable
filter, which is a filter set composed of an even-symmetric filter and an odd-symmetric filter.
When the orthogonal filter set is rotated to the same orientation of the local texture, the
oriented energy reaches its maximum. The orientation corresponding to the maximum
384 Remote Sensing – Applications

oriented energy is defined as the dominant orientation of the local energy at that point. And
the main direction of the wave can be estimated.

2.2.5 The registration algorithm of SAR image and nautical chart based on Gaussian
principle curve
To ensure precise detection and location of the distressed ship in the MSR, the navigational
chart and the remote sensing image should be matched beforehand. Because the SAR image
and the electronic chart are data from different sensors, the content and intensities of these
images are much different from each other. Coastline is a stable and reliable feature for
navigation in coast area. However, the deformation between edges extracted from different
signals may produce position errors, and the noise in radar signals may greatly influence the
edge extraction result. And how to obtain reliable control-points and how to obtain the
correct correspondence are the key issues in the registration algorithm. In this chapter, a
multi-scaled registration algorithm for SAR image and electronic chart is presented. Based
on the scale-space theory, coastlines from the two images are matched in both frequency
domain and image domain with continuous scale level.

3. Algorithms and methods in the maritime search and rescue

3.1 Distributed target detection method in the Gaussian scale-space
In the high resolution SAR images, the large-scale ship and super large-scale ship (>100
meters) are presented as distributed targets. In this algorithm, these targets are detected by
the distance relationship between the echo intensity of the masthead light and the ship hull.
The detection of distributed target based on location-dependent information can be
completed by two-step detection. In the first step, the ship target is characterized in the
Gaussian scale-space. This transforms the signal range value into binary. Then, the detection
of the singular objects is implemented using constant false alarm rate(CFAR), and record the
location of the pixel point whose value is 1. The second step aims at finding the distributed
target from the result of the first step detection by applying location-dependent information.

3.1.1 Ship characteristic description

Under the ideal condition, in the z(x) of the image which has its background gray value as 0,
there is a maxima value h and strong scattered point fb(x) with width as w. Considering the
edge effect of the image, we build the mathematical model as follows:

h(1  ( x w )2 ), |x| w
fb (x)   (1)
0, |x| w

The responses of these spot-like targets in Gaussian scale space are presented as rb(x,,w,h):

rb ( x , , w , h )  g ( x )  f b ( x )
h
 2 [( w 2  x 2   2 )( ( x  w )   ( x  w )) (2)
w
2 2 x( g ( x  w )  g ( x  w ))   4 ( g ( x  w )  g ( x  w ))]
Remote Sensing Application in the Maritime Search and Rescue 385

x x2

Among them,  ( x )   e 2 2 dt .


In order to departure the ship target from the grey scale space, we define a Gaussian
comparison function eb(x). If the responses of f(x) in Gaussian scale-space is r(x,, w, h), then
we have:

1, r ( x , , w , h )  f ( x ), and x  w
eb ( x )   (3)
0, other

Then, the detection of the singular target is implemented by way of constant false alarm
rate(CFAR).

3.1.2 Distributed target detection based on location-dependent

Set the SAR image as a N R  N A matrix, where N R and NA represent the dimension of
range and azimuth respectively, and use i R1 ,i R2 , ,i RK  as range coordinate and
iA1 ,i A2 , ,iAK  as azimuth coordinate separately for convenience.
The detected relative distance among different scattered units can be defined.

d( j , k )  i j  jk , k  j , j  1, 2, , K  1 (4)

where means norm. Equation(4) represents the relative distance between two scattered
units j and k, and it expresses the location relations among each target pixel point after the
Gaussian scale-space detection.
When the range and the azimuth are considered separately, the distance can be defined as


d( j , k )  iRk  iRj , i Ak  i Aj  (5)

Due to the geometric distortions of the SAR image, the ship hull and the mast will occupy
several resolution units in the image. Set a candidate ship target M0 ( M0 includes the ship
hull and the mast), where M0R and M0 A represent the size of M 0 in the range direction and
the azimuth direction respectively. Set a bounding box with size of M 0R  M 0A , which is the
smallest rectangle containing the detecting ship. Assume that the size of the detecting target
S S
is S R  S A and the radar resolution is R  A , then M0 R  M0 A  R  A .
R A
Regarding to the distributed targets, at least M scattering points will be detected after the
second detection step, and then M can be defined as the distance threshold dth within the
distributed target.

dth  M   M0 R  M0 A (6)

Here   (0,1) is the confident coefficient and it is determined by the empirical data of the
radar echo.
386 Remote Sensing – Applications

Define the size of the distributed target Ts as:

Ts  d( M0 ,(0,0)) (7)

Equation (7) defines the distance from the position of the bounding box M 0 to the origin
point (0,0). The distance describes the size of the detecting target, which is related to the
form of the distance definition.
According to the size of the bounding box M0 and Equation (6), the number of scattered
points (marked as u) of each target reference window and their locations (i.e. target location)
can be calculated as follow.

u  dth , target 
 (8)
u  dth , non  target 

Under the definition in Equation (7), d(Tposition , ik ) and Ts are all two-dimension data, we
need define its output. Set Ts  (TsR , TsA ) , d  (Tposition , ik )  ( dR , d A ) , and then the export
definition of d(Tposition , ik )  Ts is:

1, other
d  (Tposition ,i k )  Ts   (9)
 0, d R  TsR and dA  TsA

3.1.3 Algorithm realization

According to the above description of the algorithm, to design the integrated detection
algorithm based on two-dimension location-dependent information, the procedure is shown
as follows:
1. Initially set k  1 , u  1 , and the position of the target centre is Tposition  ik ;
2. If d  (Tposition , ik )  Ts , the number of pixels contained in the current target is u  u  1 ,
(u  1)Tposition  ik
and renew the central position of the current target Tposition  ;
u
3. If d  (Tposition , ik )  Ts , then create a new target with its central pixel location as
Tposition  ik , u  1 ;
4. Traversal searches the image.

3.2 Ship size category estimation model

In the SAR images with high resolution, the ship target can be divided into two categories
according to their dimension. The small ships are represented as single corner target while
big ships as double corner distributed targets. According to the relative plane location of the
double corner and combining the orbit information (resolution ratio and incident angle) of
SAR images, the parameters of the ship’s status, such as the length, height and the heading
direction can be calculated. Table 2 shows the corresponding incident angle of ENVISAT
ASAR.
Remote Sensing Application in the Maritime Search and Rescue 387

Distance to the
Image Position Breadth (km) Incident Angle (º)
Substellar Point (km)
IS1 104.8 187.2—292.0 15.0—22.9
IS2 104.8 242.0—346.9 19.2—26.7
IS3 81.5 337.2—418.7 26.0—31.4
IS4 88.1 412.0—500.1 31.0—36.3
IS5 64.2 490.4—554.6 35.8—39.4
IS6 70.1 549.7—619.8 39.1—42.8
IS7 56.5 614.7—671.1 42.5—45.2
Table 2. Corresponding incident angle range of ENVISAT ASAR

Build a space coordinates as Fig.1 with the true north (Y axis) as the reference direction. The
ship length is l, and the width is w, height is h, azimuthal angle of the bow is  (the angle
between the bow and the North), and then a ship model can be descripted by a set of
geometric parameters as P  [ , l , w , h ] . For example, if the satellite is descending, for
ENVISAT ASAR it is a right view image. Set radar incident angle as θ, the azimuthal angle
as β (the angle between the satellite and the true north) and the height as H, which is shown
as in Fig.1. According to SAR imaging mechanism, the shadow in the figure shows the ship's
scattering image, which is determined by the ship’s ground position, shape and the radar’s
scattering orientation. Assume that the ship hull scattering length is s and the width is d. The
range is

w  b cos(    )

b  d tan 2  (10)

w  d tan 2  cos(    )

Here w is the ship width, b is the width of the ship at the radar range, d is the scattering
width of hull target，θ is the radar incident angle, β is the radar azimuthal angle and  is the
bow azimuthal angle. The azimuth is

s  l cos(    )

s
l (11)
cos(    )

w  d tan 2  cos(    )

Here l is the ship length and s is the scattering length of the ship target images at the radar
range, β is the radar azimuthal angle and  is the bow azimuthal angle.
According to the length of the ship ships can be divided into small-sized (<50 meters), mid-
sized (50-100 meters), big-sized (100-200 meters) and extra big-sized ship (>200 meters).
388 Remote Sensing – Applications

Fig. 1. The ship geometric projection model in the SAR image

3.3 Ship positioning correction and direction estimation

3.3.1 Ship positioning correction
Due to the geometric distortions of the SAR imaging, there exits measurement error between
the observed position and its actual location. With this method here we can get the real
position of the stern by way of detecting the ship’s wake region with directional texture and
calculating the convergent point of the ship's wake. Thereby, the position correcting parameter
will be worked out. This method enables us to get the information of wave-making of the
distressed ship when the wake profile is invisible or in violent sea status, which improves the
flexibility and automaticity for understanding the marine remote sensor images.
3.3.1.1 Wake region determination
A wake is the directional texture formed on the water surface immediately behind the ship.
Therefore, the mean direction of the wake is consistent with the ship heading. In this
algorithm, regarding the ship as the center, we divide the sea surface of the ship's
neighborhood into several partially overlapped sectors. Define the angle between the edge
of the sector and the positive direction of X axis as the edge direction angle of the sector, and
the two edge direction angles formed by two edges of sectors are defined the direction range
of the sector areas, and the median of the two edge angles is the main direction of the sector
area. Calculate the textural energy of the main direction in each sector area of the ship, and
set the sector area with the most textural energy as the wake areas. The direction range of
the sector is regarded as the wake direction range, and the target area close to the wake
range is the stern in the image.
Remote Sensing Application in the Maritime Search and Rescue 389

3.3.1.2 Course calculation

In the wake area, calculate the textural direction pixel by pixel, and set the pixels whose
textural directions belong to the wake direction interval as the wake points. Use the least-
square method to calculate the mean direction of the wake, and the result is defined as the
direction of the ship’s wake, i.e. the course.
3.3.1.3 Actual lactation of the stern
Theoretically, a wake is the region of disturbed flow immediately behind the ship, and the
texture of the wake pattern should converge to a point which is the real position of the stern.
Collect the wake points in the wake region. Use the Hough transform to calculate the
vanishing point of the wake texture. And this point is regarded as the actual location of the
stern.
3.3.1.4 Projection offset calculation
According to the actual position of the stern and its image position, the projecting offset can
be calculated. Use the stern offset, the projecting model is built. After calibration, the
original detected result can be modified and the actual location of the ship target can be
obtained.

3.3.2 Heading direction estimation method

A bow wave is the wave that forms at the bow of a ship when it moves through the water.
As the bow wave spreads out, it defines the outer limits of a ship's wake. Theoretically, the
convergent point of the bow wave’ outline must be on the extending line of the ship’s
heading. The convergent point can be obtained by the Hough transform, and the heading
direction of the ship can be calculated by the positions of the bow and the stern. Using the
orbit information (resolution ratio and incident angle) of SAR images, the length, height and
the direction of the ship can be worked out. The heading direction vector is defined in
accordance with the bow wave’s outline, which points from the actual location of the stern
to the convergent point of the bow wave. Then, the heading direction is defined as the
direction between the heading direction vector and the true north in anti-clockwise.

3.4 The registration algorithm of SAR image and nautical chart based on Gaussian
principle curve
To implement registration to remote sensing images with navigational radar image and the
chart, the detection results will be directly showed both on the remote sensing image and
the chart, and then do contrast verification among the remote sensing detection results and
the data of radar and AIS. A multi-scale matching algorithm of radar image and chart is
proposed in this project, transforming the coastline into a set of smooth curves in the
Gaussian scale space, and making coarse to fine image registration to radar image and the
coastline in the chart separately in the frequency field and the spatial field.

3.4.1 Curve feature representation

In the extracting coastlines, many near-shore objects such as ships and navigation marks
that also have strong echoes may be merged by mistakes, leading to Ω-shaped spurs of the
390 Remote Sensing – Applications

coastlines. In order to reduce the influence caused by this kind of noises, a geometric
criterion is proposed to avoid selecting initial seeds on spurs. To find proper seed, each
candidate seed on the rough coastline is considered by means of judging the angle between
the candidate seed and its adjacent selected seed from a certain point on the land, which is
the mirror of the radar image center, i.e., the own-ship position. This procedure is illustrated
in Fig.2. Search for the follow-up seed to seed vi in the counter clockwise direction along
the initial coastline, where O is the own-ship position, and O is its mirror point
perpendicular to the course. Judge the angle i between vi and the candidate point v i  1
from O . Spur may occur when i is small or even negative. Bypass those kinds of points
v i  1 until meet a point vi  1 whose angle  i with vi is larger than a predefined threshold.

Fig. 2. Dispose the burr noise of radar coastline

Then, a family of smoothed coastlines is derived in the Gaussian scale-space, as shown in

Fig.3. Scale-space is a special type of multi-scale representation that comprises a continuous
scale parameter and preserves the same spatial sampling at all scales.

L(.;  )  g(.; )  f (12)

where the Gaussian kernel is g ( x )  [exp( x 2 2 2 )] 2 . Because the coastline in the
electronic chart is rather smooth, the scale-space derivation is only done for the SAR image.

Fig. 3. Continuous smooth coastline in different scale spaces

Remote Sensing Application in the Maritime Search and Rescue 391

3.4.2 Registration
3.4.2.1 Coarse registration in the phase domain
The image registration technique based on Fourier-Mellin transform finds its applications in
many different fields thanks to its high accuracy, robustness and low computational cost. It
can be used to register images which are misaligned due to rotation, scaling and translation.
The basic theory for translation estimation is the Fourier shift theorem. Denote


F  f ( x , y )  F( wx , wy ) (13)

which is the Fourier transform of f ( x , y ) . Then

 j ( wx x  wy y0 )
F  f ( x  x , y  y )  F( wx , wy )e (14)

And the image translation can be estimated by the cross-spectrum of the two images.

F1 (u , v )F2* (u , v ) j ( wx x  wy y0 )
e (15)
F1 (u , v )F2* (u , v )

Assuming s( x , y ) is transformed image of r ( x , y ) after translation ( x , y ) , rotation  and

scaling  (in both x and y directions).

s( x , y )  r[ ( x cos   y sin  )  x , (  x sin   y cos  )  y ] (16)

And s( x , y ) will gain a two-dimensional pulse at the position of ( x , y ) in the ( x , y ) space.

Then, the relation between the corresponding Fourier transform of s( x , y ) and r ( x , y ) is:

s(u , v )  e  js ( u , v ) R[ 1 (u cos   v sin  ), 1 ( u sin   v cos  )] (17)

And the corresponding amplitude spectrum is:

s(u , v )   2 R[ 1 (u cos   v sin  ), 1 ( u sin   v cos  )] (18)

Then, the rotation angle and the scaling factor can be calculated in the log-polar coordinates.

sp 1 ( ,log  )  rp1 (   ,log   log  ) (19)

And evidently, sp 1 will gain a 2-D pulse at ( , ) in the Hemi-Polar-Log

( ,log  ) coordinates. The phase-correlation method computes the transformation
parameters by taking the curve as a whole, which takes the advantage of low computation
cost and a good ability of noise immunity. This procedure is repeated in the Gaussian scale-
space with a set of decreasing observing scales, and the two images are registered from
rough to precise. And the transformation parameters are evaluated by clustering based on
the evidence theory.
392 Remote Sensing – Applications

3.4.2.2 The selection of control point and registration seed

The derived curve is transformed into graph, and the weight of each node is represented by
the energy defined by,

 
E  wi 2 ( vi , vi  1 )  wi   (C ( x )  C ( x ))2  (20)
 xvi vi1 

where C is the Gaussian curvature under scale  defined by the coined product of the
largest and the smallest curvatures of vi vi  1 .

C ( vi , vi  1 )  LvvLw  Lxx Ly  2Lx Ly Lxy  Lyy Lx (21)

where Li is the Laplacian operator. C turns out to be a good corner detector, which is an
important invariant feature to describe the structure of a derived curve in certain scale-
space. And scaled energy E is a three-order vector, which describes the variance of
curvature.
The nodes with big E are selected as control points. On the local straight line points the
Gaussian curvature is zero, and the connections of these points form a parabolic line. Then,
every two adjacent parabolic lines construct a registration curve fragment. This method
assures each seed curve contain the typical topology of the local region.

3.4.3 Precise registration based on the principle curve graph

The Hausdoff distance is adopted as the comparability metric, and the best matching feature
curve fragment is obtained by using the minimum distance classifier. This procedure is
repeated in the Gaussian scale-space with a set of decreasing observing scales, and the two
images are registered from coarse to fine. The Haussdoff distance between the registration
curve N 1 and the reference curve M1 is defined by Equation(22).


DHaussdoff ( N 1 , M1 )  max( dF ( N 1 , M1 ), dB ( N 1 , M1 ))  max min Li  L j
L j N 1 Li M1
 (22)

If DHaussdoff ( N 1 , M2 )   , then the two curves are matched.  is a given threshold. The
matching metric is shown as Fig.4.

C1 C1

C2 C2

Fig. 4. The Haussdoff distance metric

Remote Sensing Application in the Maritime Search and Rescue 393

3.4.4 Experiment analysis

The experimental images are obtained at a narrow channel in Yangzi River, China, May 1st,
2007. The own ship’s position (OS POSN) is [32°13.525 N, 119°40.368 E], at speed of 13.8
knots and on course of 235.0°. The electronic chart of this area is a version in 2002. Many
new docks are built, and moreover, the inland electronic chart in China uses the Gauss-
Kruger coordinates, while the radar image uses the projected polar coordinates. Different
coordinate systems also add extra deformation between the two images.
We choose a series of image sections from the electronic chart as the reference image, and
take the radar image as image to be matched. The chart sections are selected along the
coastline with a half size of the chart. And the registration is done between the chart section
and the radar image. Because the two images are from different sensors, the coarse
registration in single scale cannot carry out a prominent pulse in the ( ,log  ) space, as
shown in Fig. 5. The registration procedure is repeated at scale levels   2 i , i  1, ,6 .
Then, the estimated transformation parameters are clustered as
(ˆ , sˆ , tˆx , tˆy )  [ 1.9, 0.344, 353, 725] .

Fig. 5. The IFFT of spectrum in single scale

Twenty-one pairs of control points are selected from both the derived SAR image and the
chart at the scale level of   16 . Using the Housdoff distance metric, the transformation
parameters of second registration are obtained as (ˆ , sˆ , tˆx , tˆy )  [ 1.9, 0.357, 371, 660] .
The registration results are shown in Fig.6. The registration performance is evaluated by
manually registering a remote sensing image from the Google Earth with the nautical chart.
The registered image is at [32°13.369 N, 119°40.279 E], 7m distance from its true position,
and the rotation bias is -1.1°. The result proves that our method is feasible. Errors come
from the strong echo of various objects near the shore.

3.5 Wave direction estimation based on local energy orientation

This method is based on Gabor filter. According to Morrone and Owens theories, local
energy is the image mean square response of filter set formed by an even symmetry filter
M e and an odd symmetry filter M o , and it gets the biggest local energy value at singular
points, such as edges and corners.
394 Remote Sensing – Applications

Fig. 6. Registered image pairs. (left) radar image and nautical chart, (right) remote sensing
image and nautical chart

E( x , y )  ( M e  f ( x , y ))2  ( Mo * f ( x , y ))2 (23)

The steerable filter is the linear combination of a set of base filters, which are partially
overlapped in the frequency domain, and can be rotated. An orthogonal filter pair is a
combination of a steerable filter and its Hilbert transformation, which is designed to detect
precisely the features of the edge, texture and singular point of the target. To obtain the 2-D
local energy in continuous frequency space, the Wavelet Transform is used to decompose
the signal into a series of sub-band signals with particular frequencies. Here we use the
Mexico-hat wavelet G2 to build the steerable filter G2 :

G2  k1 ( )G20  k2 ( )G2 /3  k3 ( )G22 /3 (24)

The Mexican-hat wavelet G2 ( x , y )   2 exp[ ( x 2  y 2 )] x 2 is a symmetric filter with sharp

narrow bandwidth, thus it can effectively restrain noise and enhance the signal in particular
frequency, and it is common used in multi-scale edge detection. The fundamental filters G20 ,
G2 /3 , G22 /3 represent the forms of G2 rotating to 0,  / 3 , 2 / 3 , respectively. ki ( ) is
the interpolation function corresponding to the fundamental filters. Then, the form of G2 in
any orientation is represented by the linear combination of G20 ， G2 /3 ， G22 /3 . We can get
the direction energy of arbitrary pixel ( x , y ) of the image in an arbitrary direction  by using
the orthogonal filter bank formed by steerable filter G2 and its Hilbert transformation H 2 .

E ( x , y )  (G2  f ( x , y ))2  ( H 2 * f ( x , y ))2 (25)

As for the singular characteristics, e.g. the edge, when the orthogonal filter moves to the
same direction with this characteristic, the direction energy reaches maximum value. The
corresponding direction of the local orientation energy is called the principal direction of the
pixel's local energy.
Remote Sensing Application in the Maritime Search and Rescue 395

The wave image is filtered in this algorithm to eliminate speckles by way of Lee filter, and
on this basis the principal energy direction of the wave can be estimated.
The experiment uses the satellite ENVISAT-1 ASAR data of 30th Sep. to 19th Oct, and the
experimental area covers 30°48′N ~ 31°20′N, 122°10′E ~ 122°47′E. We use the wave direction
estimation based on local energy direction to calculate the wave direction for AP
polarization data. The calculation results are compared with the JMH wave analysis chart
from Japan Meteorological Agency. Table 3 shows the experiment result of this wave

JMH
Acquisition Polarization Wave Direction Incident Estimated
Image Wave
Time (UTC) Mode Direction Energy Angle Direction
Analysis
0° 124.3043
30 117.9377
60 132.5371
VV
90 135.8590
120 131.9720
2008-09-30 150 118.6072
41.1016 75.1857
13:53 0 111.2229
30 116.8233
60 133.8221
VH
90 136.7364
120 132.8555
150 116.2534
0 78.0928
30 97.9058
60 130.5053
HH
90 145.1742
120 130.5707
20081008 150 99.4210
33.9364 104.2056
01:50 0 77.5663
30 99.3430
60 133.5955
VV
90 146.8708
120 133.2834
150 99.9747
0 14.0134
30 33.8942
20081010 60 133.1682
HH 19.2636 99.9323
13:39 90 268.4664
120 132.8741
150 34.1923
396 Remote Sensing – Applications

JMH
Acquisition Polarization Wave Direction Incident Estimated
Image Wave
Time (UTC) Mode Direction Energy Angle Direction
Analysis
0 7.7873
30 19.1000
60 74.9717
HV
90 153.2269
120 75.1100
150 18.9022
0 120.3329
30 115.1276
60 120.3069
HH
90 116.8428
120 120.2216
20081019 150 118.1973
44.0092 104.8161
13:56 0 66.6512
30 69.6033
60 73.6283
HV
90 71.8446
120 73.2198
150 69.0945

Table 3. The experiment result of the wave direction estimation algorithm

direction estimation algorithm. The experimental result analysis shows that VV polarization
mode is the best way for wave analysis, and the following is HH, while cross polarization
VH and HV mode are not ideal.

4. The architecture of the remote sensing aided maritime search and rescue
system
The Remote Sensing Monitoring System for Maritime Search and Rescue (RS-MSR) consists
of four modules including satellite transit inquiry module, vessel detection module, sea state
analysis module and integrated processing module. Ship detection module has three
functions and they are ship location, ship type identification/classification and ship
movement direction estimation. Sea state analysis mainly estimates the wave direction. The
integrated processing module receives the detection results from ship detection module and
sea state analysis module. According to the distressd ownship' position, heading and the
wave direction, combining the time used for data receiving, it estimates the position of the
distressed ship, and combining the satellite parameter, it can revise the result obtained
through ship detection. The analytic result by way of integrated processing module can be
transmitted to the Maritime Safety Administration (MSA) and the rescue vessel on the
working field, providing assisting decisions of areas for the rescue work. Fig. 7 describes the
architecture of RS-MSR.
Remote Sensing Application in the Maritime Search and Rescue 397

SAR mission subscription module MSA

Vessel detection module

SAR data
Vessel positioning Integrated processing
module
Type classification
Rescue area
Course heading prediction&estimation
estimation
Error analysis
RS-MSR Sea state analysis module Rescue vessel on
the working field
Wave direction
estimation VTS data
Vessel’s report

AIS data

Fig. 7. The architecture of the Remote Sensing Monitoring System for Maritime Search and
Rescue (RS-MSR)

4.1 Satellite transit inquiry module

With the development of astronavigation, the number of satellites installed with SAR
sensors is increasing. Facing with so many satellites with different purposes, it has become a
tough problem for clients to judge and select what they want quickly. RS-MSR sets up a
real-time satellite coverage inquiry system including the commonly used satellites around
the world, such as RadarSat, Envisat, ERS, CosmoSAR, TerraSAR, helping clients to retrieve
quickly the crossing time and the orbit data of these satellites at specific area.

4.2 Ship detection module

Ship detection module is the core unit of the whole RS-MSR system. Using the micro-area
images of distressed areas supplied by way of satellite, it can detect and monitor the ships
and the accidental areas, supplying clue of the distressed ship for search and rescue and
helping to determine the areas quickly. This module consists three parts including: (1)
ship detection; (2) ship classification/identification; and (3) ship direction and course
estimation.

4.3 Sea state analysis module

Sea state analysis module can perform initial analysis on the situation of the distressed area
by estimating the wave direction and supply foundation for search and rescue decision,
which is useful for estimating the floating direction and location of the distressed ships.
398 Remote Sensing – Applications

4.4 Integrated processing module

The integrated processing module is mainly used for follow up processes of the detection
result and search and rescue assisting forecast. It has two main functions. Firstly, search and
rescue range estimation. According to the current location, navigation direction and the
wave direction, estimate the potential searching areas of the distressed ship under the
settled speed of the ship and velocity of flow. Secondly, ship location correction. Combine
the satellite parameter to revise the error caused by slant-range projection imaging of the
SAR images. The analysis data obtained by way of integrated processing module can be
saved as *.dat or *.mat form and transmitted to marine department and the rescue spot,
supplying assistant for the rescue areas determination.

5. System experimental performance

5.1 Experiment design
In our experiment, the Yangtze River estuary (Changjiangkou) precautionary (30°48′N ~
31°20′N, 122°10′E ~ 122°47′E) is selected as the experimental working zone (Fig. 8). This
region includes the inward and outward fairway of the Yangtze River estuary and the 1# and
2# anchorages. The Yangtze River estuary is a tide-coordinated region, and the tide rise and
fall twice per day. The tidal range is up to 4 meters. And the flow is rapid. Both anchored
ships and underway ships are aggregated at this place. Casualties happen often, and it is a
key region for the monitoring of maritime search and rescue. Therefore, this place is an
excellent experimental region.

Fig. 8. The experimental working zone in the nautical chart

5.2 Experiment data

From 2008.09.30 to 2008.10.19, we took four times of experiments. EnviSat-1 ASAR data is
used in this experiment. Data acquisition is implemented considering the three tidal
conditions: rise, fall and still. And the data includes: VTS maritime Radar, AIS, weather
information, tide and flow, etc. The details are listed in Table 4. Here we present the
experiment on 2008.10.19 as an example.
- Experiment region: 30°48′N~31°20′N, 122°10′E~122°47′E
- Experiment time: from 2008.09.30 to 2008.10.19
Remote Sensing Application in the Maritime Search and Rescue 399

- Tide reference: Jigujiao tide-gauge station (31°10′24〞N, 122°22′54〞E)

- Flow reference: Xinkaihe tide-gauge station (31°14′36〞N, 121°29′12〞E)

Experiment Time
No. Tide reference (Jigujiao) Flow reference (Xinkaihe)
(UTC)
1 2008-09-30, 13:53:24 14:00 L, 431cm 10:00 L, 105cm/s
2 2008-10-08, 01:50:45 02:00 L, 244cm 10:00 L, 105cm/s
3 2008-10-10, 13:39:00 14:00 L, 303cm 22:00 L, -076cm/s
4 2008-10-19, 13:56:17 14:00 L, 246cm 22:00 L, 077cm/s
Table 4. The hydrological information in the experiment

5.2.1 The remote sensing data

Data: ASA_APP_1PNBEI20081019_135613_000000202073_00082_34705_4038.N1
Polarization Mode: HH/HV
The detail information is listed in Table 5.

Lower Upper Upper Lower

Passing Start Date
Mode Track Frame Left Right Left Right Swath
direction /Time (UTC)
longitude longitude latitude latitude
APP_H 2008-10-19
82 596 122.167 122.921 31.6478 30.7285 S7 Descending
H/HV 13:56:18.802

Table 5. The detail information of the remote sensing data

5.2.2 The VTS reference data

The radar data and AIS data are received from the VTS center of Shanghai port. Fig. 9 is the
VTS shore-based radar detection picture simultaneously at the acquisition time of the
remote sensing data. The VTS report shows the name and position of all the vessels
equipped with AIS in the Yangtze River estuary. The Radar data and AIS data provide a
reference for the performance analysis of the system experiment.

Fig. 9. The VTS report at the passing time of the satellite (2008-10-19, 13:56:17 UTC)
400 Remote Sensing – Applications

5.2.3 The weather information

The weather information includes the JMH weather chart, wave height, flow rate, and the
tidal data. Fig. 10 shows the track of the No.15 typhoon (30 September, 2008).

Fig. 10. The track of the No.15 typhoon (30 September,2008)

5.3 System experiment

Here we set detecting zone as follow:
- UpperLeft longitude: 122.338E
- UpperLeft latitude: 30.9261N
- Range in longitude: 9.8276 nmile
- Range in latitude: 8.96046 nmile
Fig. 11 presents the working interface of the Remote Sensing Monitoring System for
Maritime Search & Rescue (RS-MSR). And Fig. 12 shows the original SAR image.

5.3.1 Vessel detection

Vessel detection includes: ship position detection, type identification, length estimation and
heading estimation. The detection results are shown in Fig. 13(middle) and Table 6. The ship
type classified into four categories: small, middle, large, and extra-large. The ship heading
direction is the angle between the ship heading and the real north in clockwise.
In Fig. 13(right), the red circle represents small ship (point target) and the green arrow
indicates the estimated heading direction of big ship (distributed corners).
Remote Sensing Application in the Maritime Search and Rescue 401

Fig. 11. The working interface of the Remote Sensing Monitoring System for Maritime
Search & Rescue (RS-MSR)

Fig. 12. The original SAR image (2008-10-19, Changjiangkou precautionary, Envisat-1 ASAR
HH)
402 Remote Sensing – Applications

Fig. 13. (left) The selected detecting zone (original image), (middle) The vessel detection
results (the red cross), (right) Ship size identification and the heading direction estimation
result

No. Position Type Length(m) Heading angle (degree)

1 123.5118E 38.4035N M 99.3905 124.2916

2 123.8202E 36.0825N M 80.3638 176.8479

3 124.2371E 31.8074N M 98.8651 0.4757

4 124.2207E 38.8641N M 92.0447 173.3593

5 124.441E 32.1691N M 108.5735 89.9972

6 129.429E 35.3644N M 95.7427 233.4018

7 129.5839E 33.1727N L 140.0645 10.0805

8 129.8325E 36.7182N L 116.6667 179.5634

9 130.0705E 35.3129N XL 219.6904 0.0432

10 129.9275E 38.2278N M 73.48 7.3131

Note: There are totally 44 vessels are detected, and only 10 are listed here.
Table 6. The vessel detection results

5.3.2 Performance analysis

- VTS vessel report: 51
- Detected vessel: 44
- Detecting rate: 86%
- Position error: 0.06 nmile
Among the detection, six small vessels and one middle sized vessel were missed. The
performance is better in the detection of big and extra big vessels.
Remote Sensing Application in the Maritime Search and Rescue 403

5.3.3 Sea state estimation

The wave direction estimation result is shown in Table 7.

Acquisition Polarization Incident Estimated JMH

Time (UTC) Mode Angle Direction Wave Analysis

20081019 13:56 HH/HV 44.0092 104.8161

Table 7. The wave direction estimation result

5.3.4 Ship position prediction

- Ship number: 1＃
- Predicting time: 2 hours After
- Ship speed: 15 knot
- Flow speed: 3 knot
- MSR Searching region prediction:
 Northwest: 123.1202E, 36.7825N;
 Southeast: 124.5202E, 35.3825N.

5.4 The system performance

The complete maritime search and rescue system supplies the function of ship detection,
identification and location algorithm. It can be also used to inquire the satellite transit time
and its orbit data. The precision of ship size estimation reached over 80% and the position
estimate the position detecting error is within 0.5 nautical miles.

6. Conclusion
In this chapter, a remote sensing monitoring system for maritime search and rescue (RS-MSR)
is presented. Some related algorithms are introduced. The satellite remote sensing imageries of
large scale water area are acquired to detect and locate distress ships for guiding the search
operation. Some important data such as current and sea state are retrieved to assist decision-
making of the operation. System experiment design and test are presented, and the
performance shows that this system can effectively improve the searching speed and accuracy,
and is of significance in promoting rescue success rate and efficiency.

7. Acknowledgment
The research work in this paper is partially sponsored by the Shanghai Leading Academic
Discipline Project (grant number: S30602), and the Natural Science Foundation of China
(grant number: 40801174), and the Program of Shanghai Subject Chief Scientist (grant
number: 10QA1403100).
404 Remote Sensing – Applications

8. References
Dare P., Dowman I. (2001). An Improved Model for Automatic Feature-based Registration
of SAR and SPOT Images, ISPRS Journal of Photogrammetry & Remote Sensing, 2001,
Vol.56, pp.13-28
Gerkacg J. (1999). Spatially Distributed Target Detection in Non-Gaussian Clutter. IEEE
Trans. on AES, 1999, Vol.35, No.3, pp.926-934
Huang D.S & Han Y.Q. (1997). A Detection Method of High Resolution Radar Targets Based
on Position Correction. Journal of Electronics, 1997, Vol.19, No.5, pp.584-590.
Kapoor R.; Banerjee A.; Tsihrintzis G.A. & et al. (1999). UWB Radar Detection of Target in
Foliage using Alpha-stable Clutter Models. IEEE Trans. on AES, 1999, Vol.35, No.3,
pp.819-833
Kuttikkad S. & Chllappa R. (1994). Non-Gaussian CFAR Techniques for Target Detection in
High Resolution SAR Images. IEEE, International Radar Conference 1994, pp.910-914,
Austin, TX, USA, 1994
Morrone, M.C. & Owens, R.A. (1987). Feature Detection from Local Energy, Pattern
Recognition Letters, Vol. 6, pp. 303-313
Novak L.M. & Hesse S.R. (1993). Optimal Polarization for Radar Detection and Recognition
of Targets in Clutter, IEEE, National Radar Conference, pp.79-83, Lynnfield, MA,
USA, 1993
Steger, C. (1996). An Unbiased Detector of Curvilinear Structures, Technical Report FGBV-96-
03, Forschungsgruppe Bildverstehen (FG BV), Informatik IX, Technische
Universität München
Uratsuka S. & et al. (2002). High-resolution Dual-bands Interferometric and Polarimetric
Airborne SAR (Pi-SAR) and Its Applications, Proceedings of IGARSS’02, Vol.3,
pp.24-28
 Section 4

Human Activity Assessment

 18

Object-Based Image Analysis of VHR

Satellite Imagery for Population Estimation
in Informal Settlement Kibera-Nairobi, Kenya
Tatjana Veljanovski1,2, Urša Kanjir1, Peter Pehani1,2,
Krištof Oštir1,2 and Primož Kovačič3
1Scientific Research Centre of the Slovenian Academy of Sciences and Arts
2Space-SI – Centre of Excellence for Space Science and Technologies
3Map Kibera Trust
1,2Slovenia
3Kenya

1. Introduction
Cities in Africa and developing countries in general are having a difficult time coping with
the influx of people arriving every day. Informal settlements are growing, and governments
are struggling to provide even the most fundamental services to their urban populations.
Kibera (edge region within the Nairobi) is the biggest informal settlement in Kenya, and one
of the biggest in Africa. The population estimates vary between 170,000 and 1 million and
are highly debatable. What is certain is that the area is large (roughly 2.5 km2), host at least
hundreds of thousands people, is informal and self-organized, stricken by poverty, disease,
population increase, environmental degradation, corruption, lack of security and - often
overlooked but extremely important – lack of information which all contribute to lack of
basic services such as access to safe water, sanitation, health care and formal education.
In Africa, but also in other continents, urban growth has reached alarming figures. Informal
settlements formation has been associated with the rapid growth of urban population
caused by rural immigration, triggered by difficult livelihood, civil wars and internal
disturbances. The result of this very rapid and unplanned urban growth is that 30% to 60%
of residents of most large cities in developing countries live in informal settlements
(UNHSP, 2005). Nowadays, informal residential environments (slums) are an important
component reflecting fast urban expansion in poor living conditions.
Densely populated urban areas in developing countries often lack any kind of data that
would enable the monitoring systems. Monitoring systems joining spatial (location) and
social data can be used for the monitoring, planning and management purposes. New
methods of monitoring are required to generate adequate data to help link the location
and socioeconomic data in urban systems to local policies and controlling actions. In the
past, rapid urban growth was quite difficult to manage and regulate when processes were
in progress. Available census data barely accounts for the reality, as in most cases, they
408 Remote Sensing – Applications

are based on figures extrapolated from old census, carried out in the 1970s or, if recent,
they are obtained with poor accuracy, as informal settlements are difficult to survey
(Sartori et al., 2002). More can now be done at least to monitor the extent and
consequences of rapid urban growth. Where accurate maps of informal settlements and
relevant census data completely lack, answers can be found using independent survey,
derived from satellite or aerial technologies. Usage of satellite imagery nowadays enables
rather quick answers to questions such as: where informal settlements are, what was the
dynamics of their growth, how many people potentially live there, what basic services
inhabitants need. Among the main issues to be addressed in informal settlements are the
needs for potable water, waste evacuation, energy, education and health care facilities,
and crime control. It is believed these actions can be planned based on quality mapping of
the phenomena.
The spatial resolution of space-borne remote sensing has improved to such extent that their
products are comparable with the ones provided by aerial photography. Satellite images
taken with very high resolution (VHR) sensors, i.e. resolution around and below 1 m, enable
skilled user to identify and extract buildings, trees, narrow paths and other objects of
comparable size. A side effect of higher resolution is larger quantity of data which require
more storage capacities and processing costs. Detection of informal residential settlements
from satellite imagery is especially challenging task due to the microstructure,
merged/overlapping rooftops and irregular shapes of buildings in slum-like areas. High
spatial resolution is essential to facilitate extraction of individual buildings that are
characterized by small, densely packed shanties and other structures. Informal settlement
Kibera is composed of varying sizes of houses, where roofs can be a combination of many
different materials, and mainly unpaved road and path network. Typically this can produce
a spectral response on satellite imagery that is difficult to interpret and makes it difficult for
traditional classification strategies to differentiate across object class type.
Various approaches enable to extract data from imagery in urban environments.
Simultaneously with expansion of VHR satellite systems an object-based image analysis
(OBIA) was developed to answer new technological opportunities. OBIA approach works
in similar way as human brain perceives nature/environment, namely (high detailed)
image is segmented into homogeneous regions called segments or “image objects” (Benz
et al., 2004), which are then classified into meaningful classes, following the specific
context of the study.

1.1 Objectives of the research

Objective of the work perform was to help Map Kibera Trust initiative with satellite data
processing. Studies on Kibera informal settlement had two aims: first, to derive detailed
land use/cover map that can further supply population estimation, and second, to analyse
the potential of VHR imagery for detecting changes and settlement growth in recent past.
Since object-based classification of VHR satellite data has been argued as the most
appropriate method to obtain information from urban remote sensing applications, this
approach was used to derive accurate land cover map. The study involved GeoEye and
QuickBird satellite images acquired between 2006 and 2009. Object-based approach was
used to determine detailed urban structure in informal settlements area. Urban expansion
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 409

was analyzed through comparison of images taken on different dates, using contextual
multi-level pixel based approach. The results of object-based analysis based on morphology
attributes were further explored to estimate the potential population. There is a big
discrepancy among estimations on Kibera population, thus different density parameters
were tested to approach the potential population scenario.
The first, introductory chapter sets the informal residential settlement issue in the wider
context of the remote sensing possibilities framework, highlighting the methodology of the
study. Chapter 2 gives an overview of research and applications of informal residential
environments monitoring. Chapter 3 reviews existing conditions in Kibera, Nairobi’s
informal residential settlement, bringing into perspective the historical development of the
slum, and its current characteristics. Chapter 4 consists of a set of specific procedures
performed at two spatial extents, to attain both aims of the study. Entire Kibera settlement
was being reviewed, to map the general state and dynamics of housing (change detection)
between years 2006 and 2009. Raila village was studied in detail using object-based analysis
to derive precise map of the village land cover/use to derive population estimation models
in a given situation. Chapter 5 collects the results of mapping and population estimations.
Chapter 6 discusses the data and analyses involved in managing monitoring aspects of the
slums. The last chapter concludes the study with some suggestions for future work.

2. Informal residential environments monitoring

Although there is a strong need to obtain spatial information about informal settlements in
order to increase living conditions for its residents and regarding the fact that remote
sensing images offer a well suited data source, studies on informal settlements with VHR
data are not frequent. Nevertheless, in Hoffman (2001), first results of detecting informal
settlements from IKONOS data in Cape Town showed the principle feasibilities using
object-oriented approach. The results were promising but seemed to be very dependent on
the data. Later on Hoffman et al. (2006) showed that several adaptations were necessary to
OBIA algorithm improvement when applying their extraction methods to the QuickBird
scene. Automatic image analysis procedures for a rapid and reliable identification of refugee
tents from IKONOS imagery over the Lukole refugee camp in Tanzania was made by Giada
et al. (2002). Sliuzas and Kuffer (2008) analyzed the spatial heterogeneity of informal
settlements using selected high resolution remote sensing based spatial indicators such as
roof coverage densities and a lack of proper road network characterized by the irregular
layout of settlements. Cooperation between KeyObs, UNOSAT, OCHA and Metria resulted
in digitalization of VHR GeoEye satellite image of Afgooye corridor (Somalia) from 2009,
where all temporary shelters were identified (UNHCR, 2010). Different methods to detect
and monitor spatial behaviour of informal settlements were presented also by Lemma et al.
(2005), Radnaabazar et al. (2004), Kuffer (2003), Sartori et al. (2002), Dare & Fraser (2001) and
Mason et al. (1998).

3. Study area description

Kibera is a division of Nairobi area, Kenya, within Langata constituency. Located southwest
of the city centre of Nairobi, Kibera encompasses an area of 2.5 km2, accounting for less than
percent of Nairobi’s total area while containing more than 25% of its population. It is the
410 Remote Sensing – Applications

largest informal settlement in Nairobi, and the second largest urban slum in Africa, with
population number varying with the season. The settlement is divided into a number of
villages, including Kianda, Soweto West, Raila, Gatwekera, Kisumu Ndogo, Lindi, Laini
Saba, Siranga, Kamdi Muru, Makina, Mashimoni and Soweto East (Fig. 1).

Fig. 1. Kibera settlement is divided into three formal and 12 informal villages.

3.1 General background of Kibera, Nairobi

Kibera emerged in 1912 when the British East African army, known as the King’s African
Rifles, granted temporary rights to a group of 300 former soldiers from the Nubian
community, who had served in the army, to settle on a small piece of land near Nairobi’s
city centre. Temporary structures were put in place but as the Nubian soldiers grew older
and became unable to continue their military service, they began to set up more permanent
residence on the land (A history of Kibera, 2011). Turbulent years after the independence
combined with socioeconomic factors brought a dramatic increase in the population of
Kibera’s residents.
Today Kibera consists of 15 villages out of which just 3 are formal and thus connected to the
city’s utility grids (water, sewage, electricity, waste collection etc.), however the rest (12) are
informal and “disconnected” from the rest of the city. Apart from lacking basic services and
adequate infrastructure it is also affected by population growth, the illegal construction of
infrastructure, and the increasing degradation of the environment. Unclear land-tenure
arrangements in informal settlements discourage investments in proper infrastructure and
repair; structures are often owned and rented by people, who mostly do not have any rights
to the land on which the structures stand. This leads to the lack of legal security of tenure for
most of the residents.
Because of this lack of the legal security of tenure and neglectfulness from the city and the
government there’s little initiative from the residents to improve their living conditions.
That is why most of the structures in Kibera are temporary, wooden, mud houses covered
with corrugated iron sheets (Fig. 2) and most of the service providers are self-organized
groups or cartels which drive up the prices of service delivery – in some cases residents pay
10 times as much as those in the rest of the city.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 411

Fig. 2. View over Kibera informal settlement (Photo: Primož Kovačič).

All these reasons lead to, as one resident of Kibera put it, “survival tactics”. These “survival
tactics” engulf communities, the provincial administration and the government, leading
them into a vicious cycle of under the table dealings, vandalism, lack of engagement,
threats, and price controls with no clear perspective or solutions.

3.2 Map Kibera Project and Map Kibera Trust

Kibera is likely one of the most photographed, researched, and well-known slums in the
world but the complete and mapped information was not shared and not easily (if at all)
accessible. Before October 2009, Kibera did not even appear on any of the online maps. Map
Kibera Project (MKP) was first initiated in response to the lack of available data. The
initiative wanted to produce reliable data and maps showing the actual physical and socio-
demographic features of the Kibera informal settlement, making them publicly available
through a digital geo-referenced data base (MKP, 2011).
Map Kibera Project trained 13 youth from the slum in GPS system and basic GIS techniques
to map points of interest in their community: clinics, schools, water sources, toilets, street
lights, hot spots, businesses and other landmarks. The youth uploaded the data themselves
to OpenStreetMap (www.openstreetmap.org), a volunteer-built map of the world.
The Map Kibera Trust (MKT) offers organizational support for the mapping work, as well as
other youth driven programmes such as video production and other new media tools (blog,
twitter, SMS platforms). The mission of the MKT is to contribute to a culture where digital
story-telling, open data and geographic information lead to greater influence and
representation for marginalized communities in Kenya. MKT has since grown into a
platform specializing in community-driven data for informal settlements and on community
based development.
American Association for the Advancement of Science supports the operation of MKT and
other NGO activities and has donated several satellite images of the area. MKT activities
412 Remote Sensing – Applications

include various Kibera specific phenomena mappings (www.mappingnobigdeal.com),

though the assessment of potential of VHR satellite imagery for mapping purposes presents
one of the recent examinations of their use for Kibera community.

3.3 Available VHR satellite data

Six VHR satellite images were available for our research (Table 1): one GeoEye image and
five QuickBird images. Satellite images were partly (pre)processed. This means images were
roughly georeferenced and corrected for sensor radiometry, also pan-sharpened, and
provided as a stack of three visible bands only.

Cloud Analysis
Date Sensor Bands used Spatial resolution
coverage performed
2006-03-27 QuickBird R-G-B 0.6 (pansharpened) minor Change detection
2006-07-31 QuickBird R-G-B 0.6 (pansharpened) free
2007-01-22 QuickBird R-G-B 0.6 (pansharpened) present
2008-01-07 QuickBird R-G-B 0.6 (pansharpened) free
2008-08-10 QuickBird R-G-B 0.6 (pansharpened) present Change detection
Land use/cover
2009-07-25 GeoEye R-G-B 0.5 (pansharpened) free
Change detection

Table 1. List of available satellite images and their main characteristics.

Besides different inherent spatial resolution the main differences among GeoEye and
QuickBird images were sensor viewing angles, causing higher objects roof prints and
shadows to have different positions among images. As Kibera informal settlement lies in a
hilly terrain, the positional accuracy fit of geographical entities among images was not
reached because much of distortion comes from the terrain as well. For the study no digital
elevation model was available, thus ortorectification was not possible. However, GPS field
walks tracks were available for the main roads and path-network in the area.

4. Methods
Study of Kibera informal settlement has two main aims: to derive detailed land use/land
cover map that can supply population estimation, and to analyse the settlement growth and
changes between 2006 and 2009.
Extracting data of urban land use structure from remote sensing imagery require methods
that are able to provide appropriate level of details observed. Object based classification has
been successfully implemented to obtain land cover information from urban VHR remote
sensing applications. Thus, this approach was selected in the land use/cover classification of
Kibera settlement with main aim to delimitate well residential objects from open areas, and
potentially to obtain informal settlement structure in the microstructure level (distinguish
individual houses). In addition to determination of detailed urban structures we were also
interested in locating the step-wise expansion of informal residential areas, which was
analyzed through comparison of images taken in different time using pixel-based multi-
level image differencing approach.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 413

Object based classification of the Kibera informal settlement was performed on GeoEye
image since its characteristics (close to nadir viewing angle, good spatial resolution, and
fine contrast) were most promising to obtain adequate details on object recognition within
the informal settlement area. Rooftops are covered with different materials, ranging from
new to rusty sheets, bricks and other materials, each of them having specific reflectance
characteristics (spectral representation) on satellite image (Fig. 3, Fig. 6a). For population
estimation study we need to differentiate well rooftops, unpaved roads and non-build
land and therefore discriminate residential areas from open soils, respectively. Object
based segmentation automatically delimits satellite image into homogeneous elements
(segments), where close correspondence to the real (geographical) objects on the Earth’s
surface is expected. Usage of thus obtained image elements (segments) has a number of
benefits, one of them is ability to incorporate spatial and contextual information such as
size, shape, texture and topological relationships (Blaschke et al., 2004; Benz et al., 2004) in
contextual classification. In the stage of classification all these segments are classified
according to their attributes into most appropriate classes (representing various
geographical objects under study consideration), while obtaining detailed classification of
urban area land cover/use.

Fig. 3. Examples of rooftops, rooftops renovations and buildings constructions on VHR

satellite imagery from three different dates.

With object-based analysis on rooftops morphology attributes we expected to improve the

assessment of the potential population in slum areas. Since no complete and relevant field
survey (official census) was recently performed, different density parameters were tested
to approach the potential population and compared to other available population
assessments.

4.1 Data pre-processing and preparation

Data preprocessing is important procedure in remote sensing technology. It meets issues
that have to be carefully understood and solved before any data analysis process starts. In
order to be able to compare satellite images taken for the same scene at different acquisition
dates they have to be co-registered and radiometrically adjusted. Recent automatic
registration algorithms can accomplish the task well when similar acquisition geometry
among sensor systems is provided. Global geometric transformations are mostly
appropriate for positional corrections in such cases. However, this was not the case with the
imagery obtained for Kibera study (see section 3.3).
Obtained images were rectified but not precisely aligned one to another. Due to agitated
terrain in Kibera and lack of any digital elevation model, semi-automated rigorous
414 Remote Sensing – Applications

procedures of co-registration could not be applied. The non‐linear rubber‐sheeting method

was the only possibility to obtain mutually aligned images. This procedure is effective, but
very time consuming, since it demands manual selection of hundreds of control points for
each image.

Fig. 4. Selecting the control points for rubber-sheeting method for image geo-referencing
(AutoSync module of ERDAS Imagine).

GeoEye image taken on 25th of July 2009 was selected as the reference image considering its
highest spatial resolution, good matching with GPS path-network tracks and the fact it is
most recent. Then QuickBird images selected for analysis were manually registered to the
reference, based on cca. 1,400 manually selected control points per image and using a
piecewise transformation based on triangles formed from the tie points (Fig. 4). Resampling
was nearest neighbour. An average RMSE is not reported as this is local approximation
technique. Geo-corrected images were evaluated through detailed visual control.
Geometrically matched GeoEye and QuickBird images were then used for OBIA. Finally
three images were selected for change detection due to their best results in geo-correction
phase: GE2009-07-25, QB2008-08-10 and QB2006-03-27. For change detection analysis
images were resampled to uniform 1 m resolution, to be prepared for radiometric
standardisation.
After geometric adjustments there are still differences amongst the spectral properties of
satellite images (spectral bands from the same or different images are not adjusted to each
other). Hence, before pixel-based image comparisons (image differencing) the radiometric
standardisation is needed. Most standardisation procedures derive from adjustments of
invariant objects or from the least squares method (linear regression). The problem with
the first method is that invariant areas should be verified with field measurements.
Furthermore, the generally recognised invariant objects, such as deserts or light sandy
beaches, do not come into play, since they cannot be found in Kibera area. The principle of
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 415

the second approach is that it tries to globally adjust the given (to-be-adjusted) image or a
chosen area with the reference image or a subset through a statistical approach. We
applied linear regression for the relative adjustment of spectral bands between the
images. Relative radiometric normalisation was done through local adjustments of
QuickBird images onto GeoEye reference image, for Kibera settlement with 30 m buffer
subset only.

4.2 Land cover classification

Since GeoEye image of the whole Kibera informal settlement contains lots of information
and the analysis of the total area would be too demanding in terms of computer processing,
we divided image of Kibera into 12 smaller parts (according to 12 informal villages). The
complete process of segmentation and classification was applied systematically for each
informal village separately, with same parameter settings at each phase. This was possible
due to relatively homogenous landscape over Kibera settlement. Thus we obtained 12
regional classification outputs, which were then merged in the final stage. To avoid
erroneous classification on the edges of the splitted images, we applied 30 meter buffer
when masking the village fragments out from the whole GeoEye image (Fig. 5).

Fig. 5. A buffer of 30 meters around the village border.

Object based classification consists of two stages. Image is first segmented into a set of
segments (regions) that are considered to be homogeneous in terms of one or more spectral
or spatial properties. Then follows classification where each segment is classified into
belonging object class.
Supervised segmentation within software used (ENVI EX, Feature Extraction module) is
defined by two segmentation parameters that influence an average size of segments:
segmentation and merging. Setting different values for these two parameters causes
change of size of segments, allowing for an image to be segmented at many different
scales, so both parameter values influence classification results. Since structure of the
Earh’s surface is similar throughout the whole settlement, same general segmentation
parameters were used for each of the 12 villages. Visual example of segment structures
are shown on Fig.6.
While classifying Raila village we adapted segmentation parameters to best extract shapes
of individual buildings inside informal settlement. Since segmentation parameters were
adequate for one particular land use only (i.e. buildings), others were expected to be under-
or over-segmented. There is no single “optimal” scale for analysis of remote sensing images,
rather there are many optimal scales that are specific to the image-objects that exist within a
scale (Hay et al., 2003) and this is why using a multi-scale approach may often be preferable
(Johnson and Xie, 2011). All spectral bands were used and given equal weight for image
segmentation and all available attributes were calculated for all segments.
416 Remote Sensing – Applications

Fig. 6. Original GeoEye image (a), objects segmented (b), objects merged (c) (Feature
Extraction module of ENVI EX).

The objects extracted during the segmentation were then classified using Support Vector
Machine (SVM) classification algorithm in an object-oriented framework along with
training sets, selected by experienced user. Nine land cover classes were used all together
(Table 2).

Land use / cover classes Description

Buildings_blue Residential houses with blue spectral reflectance on image.
Buildings_light Residential houses with white or bright spectral reflectance.
Buildings_brown Residential houses with brown or dark spectral reflectance.
Buildings_red Residential houses with red spectral reflectance.
Roads Traffic connection between villages, usually unpaved.
Shadowed areas around high objects (high vegetation and
Shadows
buildings).
Soil Areas of unvegetated soil, mudded ground.
Vegetation 1 Green vegetated areas, low vegetation (grass).
Vegetation 2 Green vegetated areas, high vegetation (trees).
Table 2. Land cover classes anticipated with object-based classification.

These nine urban land use/cover classes included four types of residential housing. Sub-
classes were choosed because of their different spectral signature inside the same land cover
class (e.g. instead of selecting only class “buildings” we selected four subclasses
“buildings_blue”, “buildings_light”, „buildings_brown“ and “buildings_red”, Fig. 7). This
way we obtained better results than we would have using one general class only. More
detailed classification of residential housing was made only in Raila village.
Classification results were obtained as a raster image and a vector file. Vectors were
exported to a single layer and later processed for the need of post-classification in ESRI
ArcMap software (all polygons smaller than 2 m2 were merged with neighbouring larger
polygons).
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 417

Fig. 7. Selecting training samples for supervised classification (Feature Extraction module of
ENVI EX).

4.3 Change detection

Satellite data offers unique utility for monitoring and quantifying land cover change over
time. Consequently, change detection has become a significant part of the remote sensing
research over the last decades. The goal of remote sensing change detection is to detect the
geographic location of changes, identify their type (if possible) and quantify their amount. If
long term imagery time series are handled, trends can be recognised. A large number of
change detection methods have been evolved and they differ in their refinement, robustness
and complexity (Hall and Hay, 2003). Nowadays a three level systematisation system is
proposed that differentiates change detection methods by introducing the notion of pixel,
feature and object level image processing (Deer, 1998). In general change detection
techniques can be grouped into two major types (Jianya et al., 2008; Coppin et al., 2004; Lu et
al. 2004; Singh 1989): image differencing techniques and post-classification comparison
techniques. The main difference between the two types is that image differencing methods
can identify the location and the magnitude of change but can not identify the type of land
use or surface changes taken place in the area. Post-classification techniques can identify the
location and provide the change character. Recent advances in change detection mostly
involve high resolution data and consequently object-oriented and/or multi-scale
approaches, with a range of techniques to approach contextual modelling (Lang et al., 2006;
Blaschke et al., 2008; Addink & Van Coillie, 2010).

4.3.1 Change detection of Kibera informal settlement

The use of VHR satellite image time series may provide a reliable approach to detect dense
urban growth in detail (Hofmann et al., 2006). A generically applicable and rapid
operational land cover mapping of these settlements has generally proven difficult
(Netzband & Rahman, 2010). Object-based classification and land use mapping of Kibera
settlement from GeoEye image (section 4.2 and 5.2) highlighted some typical problems for
object delineation in slum-like areas that can be corrected only with a lot of manual work.
Main difficulties are associated with informal area outer-homogenity but inner-heterogenity
due to the microstructure of urban agglomeration. Object-based classification is thus very
demanding in terms of methodology adaptation to informal residential areas specifics,
418 Remote Sensing – Applications

especially when accounting for their direct relation to representation on different satellite
data sources. Thus within the limited framework of this case study pixel-based approach to
identify outline of urban growth was preferred. The procedure was implemented on
radiometrically adjusted time series (section 4.1). GeoEye 2009-07-25 and QuickBird 2006-03-
27 images were compared for the changes over whole Kibera, and GeoEye 2009-07-25,
QucikBird 2008-08-10 and 2006-03-27 images were analysed for the observation of sequential
urban growth of Raila village.
The simple thresholding of difference images is a well-known method that leads to the
delimitation of changes and no-changes. The advantage of this method is that it can be fully
automatic. However main disadvantage is that it heavily depends on consistency of
datasets. Regardless of the carefully performed data preparations certain unwanted effects
remains, which may drastically burden the imagery comparisons. This data variability
behaves as a detected change and may well be enclosed within identified pattern of changes,
although not all of the identified differences belong to real changes (false, non-intrinsic
changes). Such false effects result in over-estimation of change pattern and can cause the
quantitative evaluation fail. Since this data noise originates from the pre-processing
algorithms as well as the natural and technological conditions during data acquisition, it can
not be completely removed with data radiometric corrections (Veljanovski & Oštir, 2011).
To overcome this drawback a contextual multi-level change detection approach was applied
that can efficiently treat most of the unwanted differences and suppress sensor related noise
(Veljanovski, 2008). Taking into account the neighbourhood and change information by
joining two spatial scales (Fig. 8), approach reduces amount of small size false differences.

Fig. 8. Change detection approach takes into account the neighbourhood and change
information by joining different spatial scales.

The model is based on focal information logic that gives averaged change information in a
slightly reduced spatial scale – within the specified neighbourhood. It is based on the fact
that in a larger geographic area (e.g. a 3 x 3 pixel window or 3 m spatial resolution for
resampled VHR data) the information of the changes will tend to level abrupt change
information if a small spatial scale change is present, and will show the averaged difference
if the majority of pixels in the observed window are subjected to change. Computing the
piecewise change information between two time-successive data sets provides valuable
information regarding the location and numeric change value derived from contextual
information within the specified neighbourhood.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 419

The procedure was implemented as follows. First, spectral information for a slightly coarser
scale (i.e. 3 m spatial resolution) was computed for images or areas of interest. This may be
accomplished with a specified neighbourhood mean value annotation. Second, change
differentiation between images is performed on a coarser resolution scale and change
magnitude categorisation is applied (see below for categorisation classes and their
definition). Third, upper positive and negative changes are reclassified so that the mask of
important changes based on the neighbourhood context characteristics is prepared. Fourth,
change differentiation is calculated on the original data scale (i.e. GeoEye and QuickBird
data in 1 m spatial resolution), then categorisation is applied, and finally a mask (or a mask
with a buffer) of an arbitrary specified magnitude of changes (obtained in the previous step)
is overlaid in order to restrict merely the contextually supported changes.
Normally, if there is no substantial unwanted effects (noise) due to meteorological or sensor
influences in images, changes obtained from the difference image (image differentiation) are
distributed normally and symmetrically, with the average at 0. Abrupt changes (objects or
land cover transformations) can then be defined by thresholding the distribution tails,
giving the pattern of positive and negative changes in reflectance (spectral) space. We have
calculated change magnitude (transition class category) intervals for every 0.5 standard
deviation. Then the criterion of 2.5 standard deviations for negative changes and 2.0 for
positive changes was applied to enclose the majority of detected transformations in both real
and spectral world situations. Result of such categorisation of image difference is a pattern
of abrupt changes (locations of appearance, disappearance of objects), with no association to
change characterisation (type of change, from-to).
Each output from the above automatic threshold procedure is finally refined to an arbitrary
degree, depending on case study objective. For entire Kibera informal settlement changes
were obtained from comparison of GeoEye 2009-07-25 and QuickBird 2006-03-27 images.
Change patches smaller than 5 m2 were eliminated for the purpose of this example, mainly
to reduce the impact of change artefacts belonging to small patches of rooftop renovations
(Fig. 3). False changes identified due to differences in viewing angle (location of buildings
and trees shadows, buildings boundaries due to different original resolution of imagery)
were also removed using the results obtained from object based classification (state of land
use in 2009, section 4.2). Where shadow or vegetation object class were present, change
pattern was corrected in the given context. Described and implemented step-wise change
pattern refinement is shown in Fig. 9, for a subset of Raila village and northern
neighbourhood. In other words, with simple generalisation we could control many aspects
of the change pattern for study’s specific aims.
Results were visually examined and evaluation of change pattern characteristics (over- and
under-estimations) was done throughout the area using complementary comparison of
satellite images involved.

4.3.2 Raila village change detection

Raila village is known to have undergone extensive development during recent years
(Kibera Wikipedia, 2011; MKP, 2011). Thus temporally detailed examination was performed
observing its urbanisation. Change detection procedure described in section 4.3.1 was in
addition implemented for Raila village only, but for two time sequences: 2006-2008 and
2008-2009. QB 2006-03-27, 2008-08-10 and GE 2009-07-25 images were used for this example.
420 Remote Sensing – Applications

2006-2009 3-band difference image for Kibera subset Change pattern over 2006-2009 3-band difference
(Raila village). image.

Change pattern after automatic Change pattern after eliminating Change pattern after eliminating
multi-level threshold procedure. false changes due to shadows small patches due to rooftop
differences. renovations.
Fig. 9. Overview of change pattern intermediate results through implemented processing
steps.

4.4 Population estimation

Population statistics gives very important information for understanding of modern society.
Demographic research is one of the main research directions of social science for a better
understanding of the interactions between population growth and social, economic and
environmental conditions. The collection of population data depends mainly on the census,
which is labour-intensive, time-consuming and demands high financial resources. The 2009
Kenya Population and Housing Census reported Kibera’s population to be 170,070 (Karanja,
2010). This report was far from the belief of that time that Kibera slum was of the biggest
informal urban settlements in the world. Several actors had provided and published over the
years increasing estimations of the size of its population, most of them stating that it was the
largest slum in Africa with the population exceeding 1 million. According to Davis (2006), a
well known expert on urban slums, Kibera had a population of about 800,000 people.
International Housing Coalition (IHC, 2007) talked about more than half a million people. UN-
Habitat (2004) had released several estimations ranging between 350,000 and 1 million people.
These statistics mainly come from analyses of aerial images of the area. IRIN (2006) estimated a
population density of 2000 residents per hectare. In 2008 an independent team of researchers
began a door-by-door survey named Map Kibera Project (MKP, 2011). A trained team of locals,
after having developed an ad-hoc surveying methodology, has so far gathered census data of
over 15,000 people and completed the mapping of 5000 structures, services (public toilets,
schools), and infrastructures (drainage system, water and electricity supply) in the Kianda
village. Considering data collected for Kianda village, the population of the whole Kibera slum
can be estimated between 235,000 and 270,000 people.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 421

Nonetheless, no estimation so far guessed by the MKP, or the UN, or the Government of
Kenya or by other actors can be taken for granted and does not represent the real dimension
of the population of Kibera. In general, no estimation can be proved nor refuted until an
exhaustive census will be taken throughout the whole slum (Kibera Wikipedia, 2011).
Because population is not directly related to land cover surface reflectance, population
estimation is still a challenging task based purely on remote sensing spectral signatures.
Although population is not directly measurable on the remote sensing images this
technology may provide good approximation of population estimation by measurement
of visible variables, e.g. the number of residential buildings and/or the area of build-up
zone (Zhang, 2003). There exist many studies using different approaches on remote
sensing data for population estimation. Studies date from the early 1970s onward, where
air photos were utilized for manual counts of dwelling units. There are three most used
methods of population estimation by remote sensing: residence count method, area
(density) method and regression model method (Zhang, 2003). Residence count method
was mostly done in first period of studies on this topic on the western urban environment
(Horton, 1974, Barrett & Curtis, 1986). Area density method was used by H.H. Wang
(1990), F.Z. Wang (1990), P. Sautton (1998), Langford et al. (1994), Z.J. Lin (2001) and
others. Regression model method is currently also often used (Galeon, 2010, Dengsheng et
al., 2006, Zhang, 2003). With each type of method some ancillary field survey data are
needed.
Considering the above situation and the fact that for Kibera we lack other potential socio-
geographic data (elsewhere applied to predict population with regression technique), we
decided to assess the population on residential land cover class information obtained from
object-based classification with density per area method solely. For each village a total area
of buildings was calculated and different occupation scenarios (i.e. persons/living area)
were tested to observe the range of possible population fluctuation.

5. Results
With object-based (contextual) classification performed on GeoEye image with Feature
Extraction module of ENVI EX, it was possible to obtain accurate land cover map and
following this, total residential area of Kibera slum and its divisions (villages) with very
high accuracy. From this data, those related to build-up areas, were used for population
estimation. With multi-level contextual change detection implemented in Erdas Imagine,
it was possible to obtain representative change pattern reflecting where in informal
settlement intensive urbanisation processes have taken place. Results are presented in the
following order: land cover mapping, change pattern identification and population
estimation, for Kibera informal settlement (section 5.1) and Raila village (section 5.2),
respectively.

5.1 Kibera
5.1.1 Kibera land cover map 2009
Object based classification and post-classification on GeoEye 2009-07-25 image was
performed for each village in Kibera informal settlement separately. Finally individual
results were joined in a land cover map of Kibera informal settlement (Fig. 10).
422 Remote Sensing – Applications

Fig. 10. Merged final classification results of GeoEye image where all 6 selected land cover
classes are shown for all 12 Kibera villages (ESRI ArcGIS).

Vectorized classes of entire Kibera were merged together in order to be able to calculate area
of total land use/land cover. As it is seen from Table 3, residential areas cover 2/3 (66%) of
the whole Kibera area and are prevailing when compared to other land uses. This can be
well confirmed from the visual examination of the satellite images of the discussed area.
Accuracy assessment was done by comparing results of supervised classification with
manually digitalized objects. Comparison was done on the area 200 x 300 m in the village

Land use [m2] Soils (roads,

Residential Trees Green areas Shadows Total
/ village bare ground)
Kianda 117,710 8,144 9,511 16,775 18,005 170,145
Soweto West 49,620 4,089 6,455 15,001 6,177 81,343
Raila 45,657 8,835 17,066 30,886 6,139 108,583
Gatwekera 234,609 17,634 11,718 19,682 23,878 307,520
Kisumi Ndogo 111,472 6,979 6,944 10,891 28,432 164,717
Makina 303,599 53,844 11,992 40,474 34,801 444,710
Kamdi Muru 51,248 1,829 7,541 12,123 8,670 81,412
Mashimoni 96,287 5,520 2,018 7,628 16,900 128,355
Laini Saba 181,211 17,400 13,141 45,545 20,490 277,786
Lindi 159,112 2,5826 11,885 48,215 26,630 271,668
Silanga 150,058 21,569 19,975 34,833 17,298 243,733
Soweto East 174,200 6,238 15,618 29,302 22,823 248,181
SUM [m2] 1.674,784 177,906 133,863 311,354 230,245 2.528,152
SUM (%) 66.25 7.04 5.29 12.32 9.11 100,00

Table 3. Area of different land use types for 12 informal villages and the total sum in Kibera.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 423

Lindi. Only residential segments were estimated. Since with ENVI EX classification the
outline of individual residential objects could not be extracted, we compared only the total
sums of areas classified as residential. Results are shown in Table 4. The best result (error of
3%) was obtained when choosing parameter values for segmentation/merge: 85/85.
All (semi)automatic classification methods display some errors, but as an approximate
solution object based classification on VHR data yielded very good results upon selection of
proper segmentation parameter values. Different shapes and colours in informal settlements
determine a complex urban formation, which is difficult to differentiate from other land
cover types, especially from bare soils and unpaved streets. With the object-based
classification dwelling zones were found with high accuracy all over the image, in spite of
their spectral similarities with streets and other urban features, especially bare soils.

Area Coverage
[m2] [%]
Testing rectangle (200m x 300m) 60,000
Manual digitalization 37,913
Supervised classification ENVI EX (segmentation 85, merge 85) 36,663 97
Supervised classification ENVI EX (segmentation 85, merge 65) 42,172 111
Supervised classification ENVI EX (segmentation 85, merge 45) 41,174 109
Table 4. Comparison of results of ENVI EX supervised classification using different
segmentation parameters with manual digitalization. The total areas were compared for
residential classes only.

5.1.2 Kibera change detection

The multi-level contextual change detection method we used in this case study was first
developed for monitoring various greater and intensive processes on the Earth’s surface
with middle resolution imagery (i.e. Landsat, SPOT). The result on Kibera proved same
contextual logic is efficient also when implemented on very high spatial resolution
imagery.
Change detection applied to Kibera informal settlements aimed to obtain an outline of the
distribution and extent of major urbanisation processes. Comparison was made for QB
2006-03-27 and GE 2009-07-25 images. Method has proven to be suitable for monitoring
changes related to various processes (buildings construction, buildings collapse or
disappearance, rooftop renovation, increase/decrease in vegetation) and/or of the
coincident description of their trends (see section 5.2.2 for identified informal settlement
growth in Raila village area). Although quantitative assessment of change occurrences
was not performed due to lack of independent reference data, a detailed visual control
was made through comparison of before and after images and, nevertheless, several
conclusions can be derived.
Identified pattern of changes (Fig. 11) clearly draw attention to spots where urbanisation
between years 2006 and 2009 was most intensive: Kianda north-east, Raila southern border,
Mashimoni and Laini Saba northern border and Soweto East eastern tail. According to the
424 Remote Sensing – Applications

density of change pattern elements, in addition to the edges of Kibera informal settlement,
several changes occurred in the eastern part of Kibera. These are mainly due to larger new
buildings constructions or complete rooftops renovations. Additional socio-economic data
or bigger events information (like flooding) would be welcomed to associate the rate of
more intensive rooftops renovations at some of Kibera villages and compared to the other
parts of the settlements.
With external data, for example land cover for a reference point in time, a fairly reliable
“from-to” change statistics could be extracted. However in rather homogenous land use
areas like Kibera informal settlement is, where most of the land use is residential (section
5.1.1) and where urbanisation direction limits are well known due to formal residential
settlement boundaries, such information would not additionally characterize the change
pattern. On the example of Raila village change detection in section 5.2.2 we comment
some difficulties observed related to detection of change in slum-like areas in a more
detail.

Fig. 11. Identified changes between years 2006 and 2009 (Erdas Imagine). Change pattern
enclose changes due to new buildings construction, buildings disappearance and larger
rooftop renovations.

5.1.3 Kibera population estimation in 2009

Total surface of Kibera was calculated as a sum of housing areas of all 12 villages
throughout whole Kibera from vector results obtained with object-based classification. All
the vectorized structures have been computed using ArcGIS software and we assumed that
all the structures are used for habitat purposes.
Sum of the residential areas show that total residential area of Kibera is around 1.646,883 m2,
which is 358,706 m2 more than total surface from year 1993 given by Sartori et al. (2002), where
residential area was 1.333,834 m2. We can see that informal settlement population density has
increased from year 1993. Since Kibera is surrounded by well-defined residential boundaries
and its drastical expansion therefore was not possible, we can assume that the increase of
residential area can attribute to denser housing inside the informal settlements through years.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 425

As we can see from Table 5, population estimation of Kibera can vary from 150 thousand up
to 650 thousand people living in the informal settlements according to different population
density sources. Nevertheless, both limits are high taking into account Kibera is spread on
2.5 km2 only.

Source Density [people/m2] People living in Kibera

MapKibera Project (2008) 0.0951 156,652
IRIN (2006) 0.2000 329,377
AHI US (2005) 0.3000 494,065
Sartori et al. 1 (2002) 0.3300 543,471
Sartori et al. 2 (2002) 0.3900 642,284

Table 5. The estimation of population according to the density acquired from different
sources. Housing area of Kibera was calculated from results obtained with object
classification of GeoEye image from 2009.

5.2 Raila village

5.2.1 Detailed land cover map of Raila village 2009
We made a detailed classification focused on residential housing only on image subset
covering the area of Raila village. The main idea was to automatically obtain polygon shapes
of each individual residential object or settlement. Although on VHR image individual objects
in general can be distinguished, due to the complexity and variety of dense roof surface there
was impossible to extract the shape of every individual residential object or each group of
objects (settlements). Roofs and their elements themselves present heterogeneity, resulting in
distinct spectral variation within areas of homogeneous land cover. Class „buildings_bright“
were possible to detect since they had a very distinct morphological pattern that was
contrasted by surrounding (built-up) land cover. Segments of class „buildings_brown“ were
sometimes integrated with their surroundings (soil) so they could not be readily distinguished.
Some urban elements (like rooftops) are a combination of many different surface materials;
this produced a spectral response that was difficult to interpret with routine procedures.
Automatic shape detection of each individual residential object would enable good total
population estimation. Two approaches to get classified vector shapes closer to individual
buildings were considered:
 Manual correction of the shapes of obtained polygons: Small test showed that this is
extremely time consuming. Correction took more time than it would take if one would
manually digitize the whole image.
 Use of additional attributes that are automatically created in the phase of ENVI EX
segmentation: Investigation showed that there is no useful relation within additional
attributes of all sub-classes of buildings. Therefore additional attributes could not assist
to semi-automatic reclassification of some segments in order to obtain better shapes of
individual object in build-up zones.
Finally, focused to residential build-up objects in Raila area four major object classes were
modelled and mapped (Fig. 12).
426 Remote Sensing – Applications

Fig. 12. Detailed object based classification results for residential areas in Raila village.

5.2.2 Raila village urban growth

Multi-level contextual change detection approach was applied to Raila village to highlight
the extent of major urbanisation processes that can take action even in a short time span. As
expected the development on southern border of the village is well recognised (Fig. 13) and
different time sequences analysed outlined where and when the growth has taken place.
Method incorporates spatial neighbourhood dependence to control the false change
information (i.e. inherent changes due to locally based variability in data). As a result it
gives a change pattern of positive and negative changes in imagery spectral space. Positive
changes correspond to an increase in digital number values at the same location and
negative changes to a decrease in spectral values. Method was developed for middle
resolution imagery, but its application to Kibera informal settlement data proved this
concept to be efficient on VHR imagery as well. In general, majority of objects
transformations regardless of their size (small patches to entire buildings) was identified.
The results were evaluated with visual control or before-after imagery comparison.
Although quantitative assessment was not possible as more detailed independent data were
not available, we estimate that more than 90% of changes associated to buildings
appearance and disappearance are captured.
Hence, detailed inspection of identified changes outlined several difficulties related to detect
change in slum-like areas. First of them is preservation of shape. Because of the use of multi-
level information the objects edges may be shrinked to some degree, causing that change is
not recognised at entire object extent but only at part of it. Second, the level of change
recognition (detection rate) is different in positive and negative spectral space. New
materials (i.e. metal plates) used to cover houses have very strong reflection, so difference
from low reflecting brown soil to new buildings with bright cover is obvious and
unproblematic. In contrast, for example from rusty rooftops to bare soil this difference is not
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 427

QB2006-03-27 QB2008-08-10 GE2009-07-25

Change pattern from years 2006-2008. Change pattern from years 2008-2009.

Urban growth pattern through years 2006, 2008 and 2009 for Raila village.
Dark red corresponds to 2006-2008 and light red to 2008-2009 period.
Fig. 13. Results of change detecion for urban growth pattern through years 2006, 2008 and
2009 for Raila village, Kibera.

great and consequently buildings disappearance is more difficult to detect. Same applies if
new building has dark colour rooftop covers (for example, often used blue colour plates).
Main change missing occurrences therefore apply to the just described situations, making
difficulties for automatic approaches. To some extent this problem was solved so that the
threshold for negative changes was reduced during transformation categorisation step.

5.2.3 Raila village population estimation in 2009

According to the total residential area in Raila village obtained from object based
classification (section 5.1.1, Table 3), population estimation was calculated based on different
density per area parameters (Table 6). Additionaly we illustrate this information also in
terms of typical size of houses in the Raila village (Fig. 14).
428 Remote Sensing – Applications

Density Density Estimated population of

Source
[people/m2] [people/32m2] Raila
MapKibera Project (2008) 0.0951 3.04 4,343
IRIN (2006) 0.2000 6.40 9,131
AHI US (2005) 0.3000 9.60 13,697
Sartori et al. 1 (2002) 0.3300 10.56 15,067
Sartori et al. 2 (2002) 0.3900 12.48 17,806

Table 6. Density of people living in Raila village according to different sources and total
residential area obtained from object-based classification of GeoEye 2009-07-25 image.

Fig. 14. Density of people according to different sources per typical size of houses in Raila
village.

In Table 6 we presented also the density on 32 m2 unit since this size of house was measured
to be a most typical and frequent (medium range size) housing unit in Raila village. It was
also observed that size of small buildings is approximately 16 m2, while large buildings
reach on average 65 m2. Fig. 14 summarise in addition the above relationship.
When perfoming such population estimation it is assumed that all of the build-up area is
used for dwelling, which is not allways true as houses may be also used for other activities.
For the possible calibration of the above population densities field survey sampling would
be required. Nevertheless, Table 6 clearly outlines how population estimation based on
critera of simple density per area based modelling can propose up to 4-times higher
(different) population estimation.

6. Discussion
Informal settlements are a very dynamic phenomenon in space and time and the number
of people living in these areas is growing worldwide. The reasons for this are many-sided
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 429

and were not under detailed examination of this study. Informal settlements represent a
particular housing and living conditions which is from a humanitarian point of view in
most cases below acceptable standards (UN-HABITAT, 2004; MKP, 2011; Sartory et al.
2002). Due to informal character and low governmental management services in the past,
reliable and accurate data about informal settlements and their population is rarely
available. On the other side there is a strong need to transform informal into formal
settlements and to gain more control about the actual urbanisation progress. Thus,
obtaining spatially and temporally accurate information is a first step to establish proper
actions in terms of local or regional planning. For these tasks, conventional data sources,
such as maps, statistics or even GIS data are usually obsolete, not available, not as
accurate as needed or do not hold the information needed (MKP, 2011; Sartory et al. 2002).
The case study presented demonstrates how informal settlements can be approached from
VHR satellite image data. Using an object based approach of image analysis detailed land
cover/use within informal settlements can be obtained to facilitate GIS-based
management tasks and population modelling. The application of automatic, even if
simple pixel-based, change detection proved to support real-time observation of informal
settlement areas whenever appropriate VHR satellite data are available with relatively
low processing costs.
Merits of object-based image analysis in dense informal settlements analysis with VHR
remote sensing data have been confirmed in several studies (section 2). However some
drawbacks still resist. In case of residential land cover/use map derivation main unsolved
difficulty is automatic detection and separation of individual houses. Although small
differences in heights of rooftops create visually well distinguished boundaries of objects,
the heterogeneity of rooftop material and its small scale changeability often overrule the
value of neighbourhood houses boundaries relationship. Current limit of object based
analysis is also that still requires substantial post-classification routines and check-over
that can be done mainly manually through visual control. What characterise this
procedure as time demanding whenever geometrically and semantically correct
information is aimed.
Change detection applied revealed great potential for long-term monitoring and informal
settlements urbanisation growth analysis. Hence more research is needed to provide
sufficient detection rate of spectrally lower magnitude changes that are typical for informal
settlements specifics and its reflectance intensity representation on satellite imagery. Due to
rooftop material used bright (metal) materials are unproblematic to distinguish from soils,
however dark rooftop materials (blue, brown colour) are spectrally closer to bare, unpaved
(brown) or vegetated (green) soils. Here object based approaches would prove better option
as sub-object attributes could be explored and used.
Valuable population estimation can be made with a relatively low cost if residential area if
accurately estimated from high-resolution images, although some considerations exist.
Area based population estimation model can be used for the informal settlements in other
images of similar resolution knowing the number of people living per surface unit. Zhang
(2002) exposed some problems of selection of the scales of remote sensing imagery,
reduction of influence of plant cover on remote sensing data, stability of the correlation
between population and remote sensing indicator variables and correction of building
count. In this research we met all mentioned problems in order to accurately estimate
430 Remote Sensing – Applications

population out of VHR satellite data. Nevertheless, if a clear understanding of mentioned

issues is considered, reliable population model of population estimation by remote
sensing data can be created. Thus, the application of satellite data information (such as
accurate information on land use extent and other measures of surface or environmental
characteristics) along with socio-economic data may well facilitate complex modelling to
estimate population trends.
In terms of remote sensing technology contribution it is necessary to continue to develop
new techniques for complex densely packed urban environments such as informal
settlements. Emphasis on spectral properties should be considered but also emphasis on the
characteristics of the shape, texture, context, and relationship with neighbouring pixels
(and/or objects) information needs to be enhanced; as well as integration of the knowledge
on corresponding socio-economic drivers should not be neglected.

7. Conclusion
Effective methods of monitoring informal settlements are required to generate appropriate
data fast enough to assist to local policies and their controlling actions. Remote sensing data
are especially powerful in that respect since, apart they are up‐to date, they assist to link the
geographic location with the accurate socio-economic data.
The results of change detection confirmed that VHR imagery is very promising for
immediate monitoring of dense informal residences in the areas where much information is
lacking. The results of object-based (contextual) classification of the land use in informal
settlements of Kibera were highly accurate, especially if taking into consideration that
informal settlements are difficult to be interpreted with automatic or semi-automatic
routines. On the other side, the results indicate the problem of the ratio between spectral
and spatial heterogeneity of objects in slum-like areas when viewed only from the above
(satellite) perspective. Overall, the use of the object-based image analysis holds great
promise for dense urban environments and was proved useful for studies of urban change
structure and corresponding population estimation.
Satellite derived information can greatly complement the information that is traditionally
collected by field observations (UNHCR, 2000). Quantitative information that can be
derived from it should not be underestimated. The production of maps with geometrical
shapes of settlements can contribute to recover the management of informal settlements,
especially when interfaced with database that has information collected on the field.
Although several challenges have not been yet solved adequately, e.g. delimitation of
individual objects in slum-like areas, we can notice that applications are being developed.
Thus (automatic) analysis of objects enables tremendous opportunities for population
estimation in informal settlements.

8. Acknowledgment
The Slovenian Research Agency has sponsored the post-doctorate research project (Z2-2127).
The Centre of Excellence for Space Sciences and Technologies Space-SI is an operation partly
financed by the European Union, European Regional Development Fund, and Republic of
Slovenia, Ministry of Higher Education, Science and Technology. American Association for
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 431

the Advancement of Science supported the operation of Map Kibera Trust with donated
satellite images of the area.

9. References
A History of Kibera (2011). Available from www.brianekdale.com/?p=230.
Addink, E.A., Van Coillie, F.M.B. (Eds.) (2010). GEOBIA 2010: Geographic Object-Based Image
Analysis, Ghent, Belgium, 29. June - 2. July 2010. ISPRS Vol.No. XXXVIII-4/C7,
Archives ISSN No 1682-1777. Available from
http://geobia.ugent.be/proceedings/html/papers.html.
AHI US (2005). Affordable housing institute. Available from
http://affordablehousinginstitute.org/blogs/us/2005/07/kibera_africas.html.
Barrett, E., Curtis, L. (1986). Introduction to Environmental Remote Sensing, pp. 238-241.
London: Chapman and Hall.
Benz, U.C., Hofmann, P., Willhauch, G., Lingenfelder, I., Heynen, M. (2004). Multi-
resolution, object-oriented fuzzy analysis of remote sensing data for GIS-ready
information. ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 58, pp. 239-
258.
Blaschke, T., Burnett, C., Pekkarinen, A. (2004). New contextual approaches using image
segmentation for object-based classification. In: Remote Sensing Image Analysis:
Including the spatial domain, De Meer, F. and de Jong, S. (Eds.), Kluver Academic
Publishers, Dordrecht, pp. 211-236.
Blaschke, T., Lang, S., Hay G.J. (eds.) (2008). Object-Based Image Analysis: Spatial Concepts for
Knowledge-Driven Remote Sensing Applications (Lecture Notes in Geoinformation and
Cartography). Springer.
Dare, P.M., Fraser, C.S. (2001). Mapping informal settlements using high resolution satellite
imagery. In: Int. J. of Remote Sensomg, Vol. 22, No. 8 pp. 1399-1301.
Davis, M. (2006). The Planet of Slums.
Deer, P. (1998). Digital change detection in remotely sensed imagery using fuzzy set theory. Ph.D.
thesis, Departments of Geography and Department of Computer Scoence,
University of Adelaide, Australia.
Dengsheng, L., Qihao, W., Guiying, L. (2006). Residential population estimation using a
remote sensing derived impervious surface approach. International Journal of Remote
Sensing, Vol. 27, No. 16, pp. 3553-3570.
Galeon F. (2008). Estimation of population in informal settlement communities using high
resolution satellite image. The International Archives of the Photogrammetry: remote
sensing and spatial information sciences, Vol. 37, Beijing.
Giada, S., De Groeve, T., Ehrlich, D. (2003). Information extraction from very high resolution
satellite imagery over Lukole refugee camp, Tanzania. International Journal of
Remote Sensing, Vol. 24, pp. 4251-4266.
Hay, G. J., Blaschke, T., Marceau, D.J., Bouchard, A. (2003). A comparison of three image-
object methods for the multiscale analysis of landscape structure. ISPRS Journal of
Photogrammetry & Remote Sensing 57, pp. 327-345.
432 Remote Sensing – Applications

Hoffman, P., Strobl, J., Blaschke, T., Kux, H., (2006). Detecting informal settlements from
Quickbird data in Rio de Janeiro using an object based approach. Available from
http://ispace.researchstudio.at/downloads/2006/165_Hofmann_et_al.pdf
Hofmann, P., 2001. Detecting urban features from IKONOS data using an object oriented
approach. Fist Annual Conference of the Remote Sensing & Photogrammetry Society, pp.
28-33.
Horton, F. (1974). Remote sensing techniques and urban acquisition. In: Remote Sensing
Techniques for Environmental Analysis, Estes, J., Senger L. (Eds.) , pp. 243-276. Santa
Barbara: Hamilton Publishing Co.
IHC (2007). International Housing Coalition. Urban Investments and Rates of Return:
Assessing MCC’s Approach to Project Evaluation. Available from
http://www.intlhc.org/docs/urban-investment.pdf.
IRIN (2006). Humanitarian news and analysis. A project of the UN Office for the
Coordination of Humanitarian Affairs (09/13/2006), "KENYA: Kibera, The
Forgotten City". Available from
http://www.irinnews.org/Report.aspx?ReportId=62409.
Jianya, G., Haigang, S., Guorui, M., & Qiming, Z. (2008). A review of multi-temporal remote
sensing data change detection algorithms. The International Archives of the
Photogrammetry, Remote Sensing and Spatial Information Sciences.
Johnson, B., Xie, Z. (2011). Unsupervised image segmentation evaluation and refinement
using a multi-scale approach. ISPRS Journal of Photogrammetry & Remote Sensing 66,
pp. 473-483.
Karanja, M. (2010). Myth shattered: Kibera numbers fail to add up. Daily Nation. Available
from
http://www.nation.co.ke/News/Kibera%20numbers%20fail%20to%20add%
20up/-/1056/1003404/-/13ga38xz/-/index.html
Kibera Wikipedia (2011). Available from http://en.wikipedia.org/wiki/Kibera.
Kuffer, M. (2003). Monitoring the Dynamics of Informal Settlements in Dar Es Salaam by
Remote Sensing: Exploring the Use of SPOT, ERS and Small Format Aerial
Photography. In: Schrenk, M. (Ed.): Proceedings CORP 2003, pp. 473-483.
Lang, S., Blaschke, T., Schöpfer, E. (2006). 1st International Conference on Object-based Image
Analysis.
http://www.isprs.org/proceedings/XXXVI/4-C42/papers.htm.
Langford, M., Unwin, D.J. (1994). Generating and mapping population density surfaces
within a geographical information system. The Cartographic Journal, Vol. 31, No.1,
pp.21-26.
Lemma, T. et al. (2005). A Participatory Approach to Monitoring Slum Conditions. Available
from
http://www.itc.nl/library/papers_2005/conf/sliuzas_par.pdf
Lin, Z., Jin, Y., Li, C. (2001). Urban population geographical information systems. Proceedings
of the 20th International Cartographic Conference, pp. 1279-1282, Beijing, China, August
6-10, 2001.
Lu, D., Mausel, P., Brondizio, E., Moran E. (2004). Change detection techniques. International
Journal of Remote Sensing 20 (25), 2365–2407.
Object-Based Image Analysis of VHR Satellite Imagery for
Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 433

Mason, O.S.., Fraser, C. (1998). Image sources for informal settlement management. In:
Photogrammetric Record, Vol. 92 (1998), No. 16, pp. 313-330.
MKP (2011). Map Kibera Project. Available from http://mapkibera.org,
http://www.mapkiberaproject.org, http://mapkiberaproject.yolasite.com
Netzband, M., Rahman, A. (2010). Remote Sensing for the mapping of urban poverty and
slum areas. Panel contribution to the Population-Environment Research Network
Cyberseminar, “What are the remote sensing data needs of the population-
environment research community?”, May 2010.
Radnaabazar, G. et al. (2004). Monitoring the development of informal settlements in
Ulanbaatar, Mongolia.
http://www.schrenk.at/CORP_CD_2004/archiv/papers/
CORP2004_radnaabazar_kuffer_hofstee.pdf (accessed 2006).
Sartori, G., Nembrini G. & Stauffer, F. (2002). Monitoring of Urban Growth of Informal
Settlements and Population Estimation from Aerial Photography and Satellite
Imaging.
http://www.thirstycitiesinwar.com/files/informalsettlements.pdf
Singh, A. (1989). Digital change detection techniques using remotely-sensed data.
International Journal of Remote Sensing 10, pp. 989–1003.
Sliuzas, R.V. and Kuffer, M. (2008). Analysing the spatial heterogeneity of poverty using
remote sensing : typology of poverty areas using selected RS based indicators. In:
Remote sensing : new challenges of high resolution, EARSeL, joint workshop, 5-7
March 2008 Bochum, Germany / ed. by C. Jürgens. Bochum : EARSeL, 2008. pp.
158-167.
Sutton, P. (1997). Modeling population density with night-time satellite imagery and GIS.
Computers, Environment and Urban Systems, Vol. 21, No.3-4, pp. 227-244.
UN-Habitat (2004). Africa on the Move. An urban crisis in the making. Available from
http://www.unhabitat.org/downloads/docs/4626_83992_GC%2021%20Africa%2
0on%20the%20Move.pdf
UNHCR (2000). Population estimation and registration. In: Handbook for Emergencies, 2nd
edition, pp. 118–131, Geneva: UNHCR Publications.
UNHCR (2010). IDP Population Assessment of the Afgooye Corridor – January 2010.
UNHSP (2005). United Nations Human Settlements Programme (UN-HABITAT), Nairobi.
Situation analysis of informal settlements in KISUMU, Kenya slum upgrading
programme. Available from
http://www.unhabitat.org,
http://www.unhabitat.org/pmss/listItemDetails.aspx?publicationID=2084
Veljanovski, T. (2008). The problem of false (non-intrinsic) changes in pixel-based change
detection on Landsat imagery. Geodetski vestnik, Vol. 52, No. 3, pp. 457-474.
Veljanovski, T., Oštir, K. (2011). Influence of radiometric pre-processing on image properties
and comparability consistency. International Conference on Sensors and Models in
Photogrammetry and Remote Sensing: SMPR2011, May 18-19, Tehran-Iran.
Wang, F.Z. (1990). Urban population estimation – multispectral imagery analysis. Urban
Environment and Urban Ecology, pp. 34-42.
434 Remote Sensing – Applications

Wang H.H. (1990). City population estimation method with satellite imagery. Remote Sensing
technology, pp. 48-54.
Zhang, B. (2003). Application of remote sensing technology to population estimation.
Chinese geographical science, Vol. 13, No.3, pp.267-271.
 19

Remote Sensing Applications

in Archaeological Research
Dimitrios D. Alexakis1, Athos Agapiou1,
Diofantos G. Hadjimitsis1 and Apostolos Sarris2
1Cyprus University of Technology, Department of Civil Engineering and Geomatics
2Foundation for Research and Technology, Institute for Mediterranean Studies,
Laboratory of Geophysical, Satellite Remote Sensing and Archaeoenvironment
1Cyprus
2Greece

1. Introduction
The spectral capability of early satellite sensors opened new perspectives in the field of
archaeological research. The recent availability of hyperspectral and multispectral satellite
imageries has established a valid and low cost alternative to aerial imagery in the field of
archaeological remote sensing. The high spatial resolution and spectral capability can make
the VHR satellite images a valuable data source for archaeological investigation, ranging
from synoptic views to small details. Since the beginning of the 20th century, aerial
photography has been used in archaeology primarily to view features on the earth’s surface,
which are difficult if not impossible to visualize from the ground level (Rowland and Sarris,
2006 ; Vermeulen, F. and Verhoeven, G., 2004). Archaeology is a recent application area of
satellite remote sensing and features such as ancient settlements can be detected with
remote sensing procedures, provided that the spatial resolution of the sensor is adequate
enough to detect the features (Menze et al., 2006). A number of different satellite sensors
have been employed in a variety of archaeological applications to the mapping of
subsurface remains and the management and protection of archaeological sites (Liu et al.,
2003). The advantage of satellite imagery over aerial photography is the greater spectral
range, due to the capabilities of the various on-board sensors.
Most satellite multi-spectral sensors have the ability to capture data within the visible and non-
visible spectrum, encompassing a portion of the ultraviolet region, the visible, and the IR
region, enabling a more comprehensive analysis (Paulidis, L., 2005). Multispectral imagery
such as Landsat or ASTER is considered to be a standard means for the classification of ground
cover and soil types (Fowler M.J.F., 2002). Concerning the detection of settlement mounds the
above sensors have been proved to be helpful for the identification of un-vegetated and
eroded sites. In recent years the high spatial resolution imageries of IKONOS and Quickbird
have been used for the detection of settlements and shallow depth monuments (De Laet et al.,
2007; 36 Massini et al., 2007; Sarris, A., 2005). Hyperspectral imagery (both airborne and
satellite) has been also applied in archaeological investigations on an experimental basis and
need further investigation (Cavalli et al., 2008; Merola et al., 2006).
436 Remote Sensing – Applications

The record of electromagnetic radiation can be achieved using special sensors. Such kind of
sensors are used to record the electromagnetic radiation from satellites while handheld
sensors can be used for field measurements. The ground radiometry and spectroscopy
involves the study of the spectral characteristics of objects according to their physical
properties (Milton, 1987). Indeed, data from portable radiometer are often refer in the
literature as “ground truth data”, due to the fact that measurements are collected in a
relatively short distance from the object so that any noise is minimized (Jonhson, 2006).
However, as Curran and Williamson (1986) emphasizes even these ground “true” data are
subject to errors, which researchers should take into account.
The spectral signature diagram, from different materials or objects, is an easy way to plot
radiation against wavelength, in a graphical form. Curves of spectral signature (reflectance
curves) and the so-called critical spectral bands (critical spectral regions) are used in many
applications of Remote Sensing (e.g. vegetation indices). The way of how measurements are
collected by radiometers can be explained through physical laws. Already, by the 1970s
Nicodemous et al. (1977) have proposed the basis for the model of "bidirectional reflectance
distribution function" (BRDF), which describes the relationship of the incident radiation
from a given address in the reflected radiation in another direction. Nevertheless, the use of
the results of the Nicodemous et al. (1977) study was not appreciated and understood by the
scientific community (Schaepman-Strub et al., 2006; Milton et al., 2009). Their study has been
used several years later by Martonchik et al. (2000) and Schaepman-Strub et al. (2006). The
original classification proposed by Nicodemous et al. (1977), depending on the geometry of
the radiation which included nine categories was reduced to only four which are actually
encountered (Martonchik et al., 2000; Milton et al., 2009).
Milton et al. (2009) stated that all spectroscopy measurements in a strict physical sense can
be categorized within the "hemispherical-conical reflectance function" (HCRF case). It
should also be noted that natural materials do not follow the rules of a diffuse Lambertian
surface, since the intensity of the reflected radiation varies regarding the angle of refraction.
Ground spectroradiometer may be used to provide calibrated measurements, since these
instruments are often accompanied by special Lambertian targets. Milton et al. (2009)
emphasizes that a critical factor for good results is the calibration of a specific target. The
only disadvantage, apart from the price of handheld spectroradiometer is that it is difficult
to cover a large area (such as an archaeological site) (Atkinson et al. 1992; Milton et al., 2009).
Apart from the purchase of the ground radiometry, there is still an important limitation that
should be taken into account. Most spectroradiometers which are found in the market are
"single-beam": the same instrument used to measure the radiation to a specific target
(reference panel) used to measure the targets of interest (target). In the interval of these
measurements the atmospheric conditions are assumed to be the same.
Spectroradiometer may be used for archaeological research in order to retrieve
characteristics of vegetation and to calculate vegetation indices. Such indices are
quantitative measures, based on vegetation spectral properties that attempt to measure
biomass or vegetative vigor. Theoretical analyses and field studies have shown that VIs are
near-linearly related to photosynthetically active radiation absorbed by plant canopy, and
therefore to light-dependent physiological processes, such as photosynthesis, occurring in
the upper canopy (Glenn et al. 2008).
Remote Sensing Applications in Archaeological Research 437

Concerning geophysical techniques they offer a non-invasive way of providing valuable

information regarding the subsurface context of the archaeological sites and contribute
significantly in leading archaeological excavations, reconstructing the past landscapes,
suggesting directives for the cultural heritage management and preservation of sites and
historical buildings and providing a prior strategy for large construction works. Employing
a suite of methods measuring the physical properties of the soils with high efficiency,
reliability and resolution, geophysical prospection has been designated as a valuable tool in
the domain of archaeological research, especially in the study of the ancient landscapes. A
very detailed review of the physical properties of each method and the fundamentals of the
operation of the corresponding instrumentation is provided by Linford (2006) and Scollar et
al. (1990).
Methods that involved measurements of the soil's electrical resistance and of the local
intensity of the magnetic field of the earth have been the earliest that were applied in the
field of archaeological prospection. Making use of Ohm's law, soil resistance meters,
acting as active prospection methods, introduce a current within the ground through the
use of metal electrodes and measure the electrical resistance of the soil, as this is
influenced by the features located within the ground and below the electrode array
(Clark, 1990). Soil resistance measurements can be carried out in two main ways, either by
moving a fixed electrode spacing (corresponding to a specific penetration depth) array
along profiles or through vertical electrical soundings (VES) (Sarris, 2008). In the latter,
the current electrode spacing is increased with respect to a fixed centre of the electrode
array, providing plots of the apparent resistivity versus electrode spacing that are
ultimately compared to theoretical curves to provide information about the layering of the
subsoil's strata. Soil resistance values are modified depending of the resistivity contrast of
the targets (eg. a high resistance wall structure or a high conductive ditch) with the
surrounding soil matrix. The measured apparent resistivity is also affected by the type of
the configuration of electrodes (e.g. Twin probe, Square, Wenner, pole-pole, dipole-pole)
and the distance between the electrodes, which is relatively proportional to the
penetration depth.
This chapter seeks to address applications of remote sensing and GIS in archaeological
research in a three-fold way. Initially, potential of satellite remote sensing is highlighted
through a multi – sensor case study in Thessaly, Greece, where different satellite image
processing techniques contributed to the detection of Neolithic tells (the so called
‘magoules’) that are found in the Thessaly plain. Four satellite remote sensing images with
different spatial resolutions (ASTER, Landsat, HYPERION, IKONOS) were examined in
order to search their potential for automatic extraction of Neolithic settlements, by means of
pixel – based (RGB composites, spatial and radiometric enhancement, vegetation indices,
data fusion, classification methods, data fusion etc). The satellite data were statistically
analyzed, together with other environmental parameters, to examine any kind of correlation
between environmental, archaeological and satellite data. Moreover, different methods were
compared and integrated methodologies for the detection of Neolithic settlements were
extracted.
Concerning ground spectroradiometer contribution to archaeological research, new
innovative tools and methodologies are also presented in this chapter. Specifically, ground
“truth” data, presented as spectral signatures libraries, were provided from different
438 Remote Sensing – Applications

spectro-radiometric campaigns at archaeological environments (e.g. Tombs of the Kings and

Sikyon archaeological site, C. Greece). Moreover field spectroscopy was used to detect
buried archaeological sites in similar ways as applied in remote sensing applications in
Neolithic tells in Thessaly. In addition the comparison of the phenological cycle 5 profile of
similar crops - under same meteorological and soil conditions - is also searched over
archaeological and non archaeological sites concerning for different case studies in Cyprus.
At the end, in order to highlight the potential of geophysical remote sensing in
archaeological research certain case studies from surveys held in Greece and Europe are
presented such as magnetic surveys in the Neolithic settlement of Veszto –Bikeri in Hungary
and the Byzantine walls of ancient Nicopolis in Greece, the ground penetrating radar
methods in the ancient Agora of Feres (Velestino) in Thessaly, in Agora of Sikyon at NE
Peloponesse - Greece and in the area of the hypothesized amphitheatre of Ierapetra (SE
Crete and finally the electrical resistivity tomographies from the Nemea, Peloponesse.

2. Application of remote sensing, GIS and geomorphology to the

reconstruction of habitation in Neolithic Thessaly
2.1 Introduction
The aim of this study is to highlight the contribution of different approaches such as Remote
Sensing, GIS and geomorphology analysis for the detection of Neolithic settlements and the
modeling of habitation in the area of Thessaly - Greece.
The Neolithic settlement mud mounds in the area of Thessaly, Greece are called Magoules.
They are low hills of 1-5 meters height and mean diameter 300 meters. The vast majority of
Magoules are laid on Larisa basin and a smaller number is distributed in west Thessaly
(Karditsa basin) (Fig.1). Both of these plains consist of Quaternary alluvial deposits
(Alexakis et al. 2008). In order to achieve the goals of the research it was necessary to
proceed with the topographic mapping of the settlements through the use of GPS, digitize
1:50,000 scale topographic and geological maps and construct a detailed archaeological and
environmental database in SQL environment.

Fig. 1. Color composite RGB→3,2,1 of the mosaic of the 4 ASTER images used to cover the
whole area of interest. The dots represent the location of the Neolithic magoules (left).
Magoula Aerino (right).
Remote Sensing Applications in Archaeological Research 439

2.2 Landscape reconstruction

Prior to the landscape reconstruction of Thessaly during different Neolithic periods, the
reliability of the existing Digital Elevation Model (DEM) was evaluated compared to other
digital elevation models, such as the 90m pixel size DEM (from the Shuttle Radar
Topography Mission (SRTM) and the 30m pixel size DEMs provided by ASTER images or
constructed by the L1-A stereoscopic products (3N and 3B) of ASTER. The results indicated
that the RMSE for the DEM created through the digitization of the contour lines of the
topographic maps was the lowest. The two major plains of Thessaly contain 181 out of the
342 known registered “magoules”, stressing the important role of reconstruction of the relief
of each basin during each Neolithic sub-period. Both geological (stratigraphic data from
boreholes and past geomorphologic studies) and archaeological data were placed under
consideration for achieving this task and a reconstructed DEM for each basin and each
Neolithic period was created.

2.3 Satellite image processing

2.3.1 Data and preprocessing of satellite images
Concerning this study different multispectral images were used. Specifically, 4 IKONOS
images of 1m pixel size, 1 Landsat ETM+ 30m pixel size image, one 30 m pixel size
Hyperion image and 4 ASTER images (15, 30 and 90m pixel size). Masking of the sea, clouds
and snow areas in all images preceded the processing of the images in order to focus to the
mainland and the areas that provided useful information. Image mosaics were created
accordingly depending on the types of sensors and both image mosaics and isolated images
were rectified to a common projection system (EGSA87/HGSR87). Digital numbers of
images were also converted to reflectance values according to specific conversion equations.
The last step was necessary in order to have a uniformity in the values of images originating
from different sensors.

2.3.2 Spectral enhancement techniques

Several RGB composites were constructed in an effort to examine their efficiency in the
detection of the Neolithic settlements. For the ASTER image with acquisition date 19-03-
2003, where most of the magoules were registered, the RGB→1,2,3, RGB→3,2,5 and
RGB→2,3,7 composites were the most successful for the visual detection of the Neolithic
settlements (39 out of 239 settlements were highly visible, 49 average visible and 151 poorly
visible). All these composites appeared to have the highest Optimum Index Factor (Alexakis
et al, 2009). Similarly, RGB composites of IKONOS images were able to detect 27 out of 48
settlements within their area of coverage. It is worth mentioning that 19 of the detectable
magoules, namely the highest of all corresponding to an average altitude of 4.6m, were
highly visible in all RGB composites. On the other hand, RGB composites of Landsat and
HYPERION images were not very promising (for HYPERION composites only 5 out of 21
settlements were detected). Due to their high spatial resolution, all the 5 settlements that felt
within the spatial limits of the aerial photo mosaic were easily detectable. As a general
conclusion however, the acquisition date of the images proved to be the most crucial factor
for the detection of magoules mainly due to the intensive cultivation (mainly soft and
shallow cultivation) of the landscape both on the top and the surroundings of magoules.
Principal Component Analysis (PCA) was applied to ASTER, Landsat and Hyperion
440 Remote Sensing – Applications

images, being especially effective for ASTER images where 39 and 47 out of 247 settlements
were highly or medium discriminated correspondingly. Image fusion techniques through
the combination of high spatial resolution images such as IKONOS (1m) and high spectral
resolution images such as Hyperion (30m) concluded to very promising results (Fig 2).
Finally, a spectral mixer utility (Erdas Imagine 9.1 software) contributed to exploit the
dynamic range of all the multispectral information of the Hyperion image by combining
more than three bands to an RGB composite. Using the specific utility and assigning a
weighting coefficient for each band, a RGB composite of 23 bands (38, 42, 48, 49, 50, 51, 52),
(85, 86, 87, 88, 89, 90, 91, 92,) & (93, 94, 108, 109, 110, 111, 113, 114) was constructed that
enhanced the visual appearance of the magoules.

2.3.3 Spectral enhancement techniques

Radiometric enhancement was vital for the appearance of the images. After applying
radiometric enhancement to ASTER images (acquisition date of 19-03-2003) we managed to
detect 57 settlements (Fig. 2). A non-linear radiometric enhancement of the HYPERION PCA
image, followed by an inversion of brightness was able to highlight 8 settlements from a
total of 9. (Melia 1, Melia 2, Anagennisi 2, Moshohori 3, Kipseli 2, Prodromos 1 of Larisa,
Nikaia 17 and Kuparissia 2). Similar type of non-linear radiometric enhancement of the high
resolution IKONOS images through the modification of the histogram outlined the round
shape of known magoules, as well as outlined 10 more targets of similar geometry that need
to be verified by the ground truthing activities that will follow (Fig. 2).

Fig. 2. Appearance of 3 settlements in the original IKONOS image (left) and the
radiometrically enhanced image where three Neolithic settlements are highlighted (right).
To the north of Galini-3 settlement, shown at the lower right of the image), another smaller
potential magoula is suggested.
Remote Sensing Applications in Archaeological Research 441

2.3.4 Land classification, vegetation indices, spatial enhancement

A spectral signature database was constructed to provide the basic spectral information
about tells, especially at the plain areas of Thessaly. Several classification methods were
applied to Landsat and ASTER images in order to investigate the land use regime around
the magoules. Examination of the overall accuracy of the various algorithms tested (based
on the error matrix), proved that the Mahalanobis algorithm was the most efficient for the
exact classification of the images (Fig. 3a). Additionally, object based segmentation
techniques were applied to ASTER images and 15 settlements in total of 234 were detected
easily (Alexakis, 2009). The computation of the Normalized Difference Vegetation Index
(NDVI) was used to highlight the vegetation differences during different periods of time, in
an effort to pinpoint any vague indications for the detection of magoules. As expected, the
NDVI of the “spring” ASTER image was higher than the “summer” Landsat image, but still
the vegetation differences of the spring time favored the detection of magoules mainly due
to the differentiations in the soil’s humidity (Fig. 3b).
Application of certain spatial high pass filters contributed further to the spatial
enhancement of smaller features such as the magoules. The most reliable of them proved to
be Sobel Right Diagonal 3x3 and Laplace 3x3, both of which outlined clearly the limits of the
most prominent of them (Fig. 3c).

Fig. 3. (a) Land classification of ASTER image through the use of Mahalanobis distance
(fuzzy) algorithm. (b) Detail from the application of NDVI to ASTER image. (c) ASTER
image around Halki area after the application of Sobel Right Diagonal filter. Neolithic
magoules are indicated within the ellipses.

2.4 Analysis in GIS environment

An extensive spatial analysis of the magoules distribution was carried out in GIS
environment using the reconstructed DEMs. Besides the extraction of statistics regarding the
relation of settlements to the aspect, slope and relief height, the distance of settlements from
natural resources was calculated by applying buffer zones around the quarries and the
water springs (mainly springs existing on the mountainous areas). Watersheds were
constructed and the distance of each settlement from its neighbor watershed was calculated.
Density maps of the settlements were created for each Neolithic period. The calculation of
the density of the settlements was accomplished through the use of a non-parametric Kernel
technique. The spatial territorial limits of the settlements were explored using the Thiessen
polygons analysis. The site catchment of Neolithic settlements was studied through least
cost surface analysis. Cost surfaces contributed also to the exploration of communication
routes between the different settlements (Alexakis, 2011).
442 Remote Sensing – Applications

Finally, GIS tools were employed to construct predictive habitation models for each phase of
the Neolithic period in an effort to locate areas that could possibly host similar type of
settlements. The specific predictive models were based on the use of a multi-parametric
spatial analysis method of geographic elements and other information (statistical,
archaeological, a.o.). All the environmental factors (height, aspect, slope, distance from
watersheds, distance from water springs, distance from quarries, geology, viewshed,
distance from chert sources, least cost paths, a.o.) that could affected the choice of habitation
in Neolithic Thessaly were statistically examined and certain weight factors were applied to
each one of them. At a final stage, a fuzzy logic algorithm and a normalization equation
were also applied a more efficient tuning of the results and for rating the final probability
from 0 to 1.

2.5 Application of sophisticated fields to the Digital Elevation Models

The final approach of the particular project involved the detection of Neolithic settlements
through the analysis of DEMs with the use of three different semi–automated
methodologies. Three different DEMs (90m pixel size SRTM DEM, 30m ASTER DEM and a
20m DEM from the digitization of contours of topographic maps) were tested in all
procedures to attest for their potential in the detection of the magoules.
The first methodology involved the estimation of the index of convexity (CI) to the three
different DEMs according to Fry et al. (2004):

CI = (x – xmed) / (xmax – xmed), (1)

where x is the initial DEM, xmed is the DEM after the application of median 7x7 filter and xmax
is the DEM after the application of maximum 7x7 filter.
Although the index of convexity seemed to be ideal for the detection of low hills such as
magoules, only 35 (28%) of them at Larisa plain and 28 (47%) at Karditsa plain were
detected by this method.
The second methodology is related with the design and application of customized filters
similar to those used by Menze and Sherratt (2006). The optimal filter for the detection of a
signal with a well known shape is the matched filter (Fig. 4). For the construction of the
matched filter, an area of 5x5 pixel was cut around the DEM of each settlement for a total of
50 settlements. Then, the value of the central pixel was subtracted by each pixel, followed by
stacking of all the 50 local DEMs (through the layer stack utility of Erdas Imagine software)
to form a final multilayer image. The particular image was imposed to Principal Component
Analysis and the first 5 principal components were summed. The negative sum replaced the
value of the central pixel of each of the above 5 filters, which they were then applied
individually to Larisa and Karditsa basins as a detection filter (Fig. 14). The statistics for
these filters proved that the specific methodology is really promising especially for the
SRTM DEM in the area of Larisa. More specifically, about 60% of the settlements were
detected through the application of the first and second filters.
The third methodology followed the approach of Iwahashi and Kamiya (1995) for the
estimation of the geometric signatures of DEM through a combined study of slope gradient,
surface convexity and texture (Fig. 4). Specifically, binary files were formed through
estimation of the mean value of slope gradient: all pixels with value above the mean value
Remote Sensing Applications in Archaeological Research 443

took a value 1 and all the rest took a value of 0. The same binary archives were created after
the subtraction of the initial DEM from the one that has been processed through the
application of a median filter. The last binary image was created after the application of a
Laplace 3x3 filter to the initial DEM. In the end, the three binary archives were summed and
the final map highlighted the areas of high local convexity. Fuzzy logic algorithms were
applied to the final results of the filtered DEM in order to produce a better classification
scheme (Pixel values equal to 0 formed the first group, values from 0 to 3 formed the second
group and pixels with value equal to 3 formed the third group) (Fig. 15). The application of
this methodology to SRTM resulted to the detection of 35% of the magoules in Larisa plain
and only 15% of the magoules in Karditsa plain.
The results obtained through the Menze & Sherratt (2006) approach to the SRTM DEM were
also implemented to the predictive modelling, together with other subproducts of the
satellite image analysis, such as the NDVI map, land use classification and spectral
signatures library of magoules. A similar methodology of significant weights and factors
was considered and results were subjected to fuzzy logic and normalization techniques.
Still, the results of predictive modelling did not alter significantly from the previous
approach, signifying a state of saturation for the parameters considered.

Fig. 4. Magoula Turnavos 6 after the application of matched filter (left). Application of
geometric signatures methodology to SRTM DEM in the area of Larisa. With the red color
are indicated the areas of higher height where magoules could be established (right).

3. Application of field spectroscopy and satellite remote sensing to

archaeology
3.1 Introduction
This chapter aims to introduce the capabilities and the potentials of field spectroscopy to
archaeological research (Fig. 5). Field spectroscopy involves not only the acquisition of
accurate measurements (e.g. spectral signatures profiles) but also the study of the
interrelationships between the spectral characteristics of objects and their biophysical
attributes in their field environment. Therefore, field spectroscopy can provide valuable
information for an area if we consider the fact that human eye senses only a small part of the
electromagnetic spectrum, from approximately 0.4 to 0.7 nm, whereas field spectroscopy in
444 Remote Sensing – Applications

support of remote sensing operates in a wider spectrum range including near infrared as
well. In this section, ground “truth” data, presented as spectral signatures libraries, are
provided from different spectroradiometric campaigns at archaeological environments (e.g.
Tombs of the Kings and Nea Paphos at SW Cyprus, Sikyon archaeological site, C. Greece).
Furthermore, spectral libraries include vegetation profiles, mainly over barley crops (from
the Palaepaphos – Cyprus archaeological site and from Neolithic tells at Thessaly - Greece).
Such libraries are used in order to examine either the seasonal changes of vegetation, or the
anomalies of vegetation profiles due to buried archaeological remains. Moreover, spectral
libraries are used for the atmospheric correction of satellite imagery. Finally, the theoretical
background of scaling up ground narrow bands taken from handheld spectroradiometers to
bandwidths satellite imagery, using the Relative Response Filters, is presented.
Ground spectroscopy may be used as a fast detection method in order to evaluate positive
or negative crop marks. In this case sections over archaeological areas are taken and
evaluated in terms of vegetation indices. Different Neolithic sites at Thessaly (central
Greece) are examined with the use of this approach. Finally, an alternative method for the
detection of archaeological remains is presented in this chapter. This method is based on the
comparison of the phenological cycle profile of similar crops -under same meteorological
and soil conditions -over archaeological and non archaeological sites. The case of
Palaepaphos site in Cyprus is presented with the use of medium resolution images (Landsat
TM/ETM+) and the support of ground spectroradiometric measurements.

Fig. 5. Potentials of ground field spectroscopy for archaeological purposes

3.2 Collection of field measurements

Field measurements were carried out in different archaeological areas in Cyprus and
Greece. The areas investigated and presented in this chapter were fully vegetated with
crops. Moreover ground spectroradiometric measurements were taken: a) over visible
monuments in order to develop an archaeological spectral signature database and b) at
dams (inland clear water) for applying atmospheric correction to satellite images.
The spectroradiometric instrument that was used to register the spectral signature was GER
1500 (Fig. 6). This instrument may record electromagnetic radiation from a range of 350 nm
up to 1050 nm. It includes more than 500 different channels and each channel cover a range
of about 1.5 nm. The field of view (FOV) of the instrument was set to 4o (≈ 0,02 m2).
Remote Sensing Applications in Archaeological Research 445

Fig. 6. GER 1500 used in this study with its calibration target (Agapiou et al. 2010)

A reference spectralon panel was used to measure the incoming solar radiation. The
Labertian spectralon panel (≈100% reflectance) measurement was used as references while
the measurement over vegetated areas or archaeological sites as a target. Therefore
reflectance for each measurement can be calculated using the following equation (2):

Reflectance = (Target Radiance / Panel Radiance) x Calibration of the panel (2)

In order to examine the use of broadband vegetation indices such as NDVI, narrow band
reflectance (from the spectroradiometer) needed to be recalculated according to the spectral
characteristics of a specific satellite sensor. The authors selected to simulate these data to
Landsat TM /ETM+ satellite imagery based on Relative Spectral Response (RSR) filters. RSR
filters describe the instrument relative sensitivity to radiance at various part of the
electromagnetic spectrum (Wu et al. 2010). These spectral responses have a value of 0 to 1
and have no units since they are relative to the peak response (Fig. 7, left). Bandpass filters
are used in the same way in spectroradiometers in order to transmit a certain wavelength
band, and block others. The reflectance from the spectroradiometer was calculated based on
the wavelength of each sensor and the RSR filter as follows:

Rband = Σ (Ri * RSRi)/ ΣRSRi (3)

Where: Rband = reflectance at a range of wavelength (e.g. Band 1)
Ri = reflectance at a specific wavelength (e.g R 450 nm)
RSRi = Relative Response value at the specific wavelength
To avoid any errors due to significant changes in the prevailing atmospheric conditions, the
measurements over the panel and the target are taken in a short time. In this case it is
assumed that irradiance had not significant change which is true for non hazy days (Milton
et al. 2009). Finally the measurements were carried out between 10:00 and 14:00 (local time)
in order to minimize the impact of illumination changes on the spectral responses (Milton,
1987) at a height of 1.2 m (Fig. 7, right)

3.3 Spectral libraries

Spectral signature diagram is an easy way to plot target reflectance against wavelength, in a
graphical form. Therefore ground field measurements from archaeological sites may be used
in order to create an “archaeological” digital spectral signature library. Even though several
remote sensing applications investigate the correlation of the spectral signature of an object,
446 Remote Sensing – Applications

Fig. 7. Relative Response filters for Bands 1-4 of Landat TM sensor (left) and typical diagram
of the in-situ spectroradiometric measurements.

in the majority these applications the aim is exactly the opposite: the study and
identification of "unknown" targets through the spectral signature. Therefore
“archaeological” spectral libraries may be used for identification or correlation of different
archaeological sites remotely.
Different spectral signatures from the archaeological site “Tombs of the Kings” (Agapiou et
al. 2011a) and “Sikyon” archaeological sites were taken. Spectral profiles indicate that there
is great potential for detecting archaeological remains in the spectral range between 550 – to
850 nm (from the green visible part of spectrum to near infrared) because of the extremely
different spectral response of the archaeological material compared to sand and local
marl/carbonate sandstone in the archaeological site “Tombs of the Kings”(Fig. 8). Similar
results were found for “Sikyona” site also (Fig. 9).

Fig. 8. Spectral signatures profiles from different targets at the archaeological site “Tombs of
the Kings”.

Spectral signatures libraries proved to be really efficient for any potential researcher that
may use satellite imagery in order to detect archaeological relics in the area because it
highlights the high correlation of spectral response of archaeological material, sand and
local geological formations in the area of red visible band.
Remote Sensing Applications in Archaeological Research 447

Fig. 9. Spectral signatures from the archaeological site of Sikyon.

3.4 Atmospheric correction of satellite images based on ground data

Atmospheric correction of satellite images is a necessary pre-processing step before any
image analysis. Earth's surface radiation, undergoes significant interaction with the
atmosphere before it reaches the satellite sensor. This interaction is stronger when the target
surfaces consist of non-bright objects, such as vegetated areas examined in many
archaeological studies. This problem is especially significant when using multi-temporal
satellite data for monitoring purposes (Hadjimitsis et al. 2010). As Lillesand et al. (2004)
argue satellite images need to be atmospherically corrected before being subjected to any
post-processing techniques.
Atmospheric effects are a result of molecular scattering and absorption of the incoming
radiation and influence the quality of the information extracted from remote sensing images.
Such errors occurred by atmospheric effects can increase the uncertainty up to 10%,
depending on the spectral channel (Che and Price, 1992). Hadjimitsis et al. (2010) have also
highlighted the importance of considering atmospheric effects when several vegetation
indices, such as NDVI were applied to Landsat TM/ETM+ images for agricultural
applications. In their study a mean difference of 18% for the NDVI was recorded before and
after the application of darkest pixel method. Therefore atmospheric correction is an
important pre-processing step required in many remote sensing applications since is needed
to convert the at-satellite spectral radiances of satellite imagery to their at-surface
counterparts.
The modified Darkest Pixel (DP) atmospheric correction method (Hadjimitsis et al., 2004)
was applied to multi-series Landsat images (Agapiou et al., 2011b). The surface radiance of
the dark targets is assumed to have approximated zero surface radiance or reflectance.
Instead of assuming Ldarkest target to be zero value, the modified DP considers the ‘true’
ground radiance or reflectance value over dark targets as the Ldarkest target.
For this reason ground spectroradiometric measurements were taken in inland clear water
target (Asprokremmos Dam in Paphos). The GER-1500 field spectroradiometer was
equipped with a fibre optic probe was used in order to retrieve the spectral signatures from
the dam. After using the RSR filters for Landsat TM/ETM+ images the spectral reflectance
448 Remote Sensing – Applications

after the atmospheric correction was calcualated. The results have shown that satellite
images were slightly improved after the removal of atmospheric effects. Indeed crop and
soil marks from archaeological areas were enhanced. Photo interpretation quality was
enchanced at images with low water vapour optical thickness and in general for images
with water vapour optical thickness less than 0.05, the quality of the images after the
atmospheric correction was improved. In the case of higher values the quality was not
improved sufficiently (Agapiou, 2011a). Fig. 10 shows some typical histograms before and
after the atmospheric correction. As it is shown the initial histogram of the image is
stretched and therefore interpretation is improved.

Fig. 10. Histogram for Band 3 before (left) and after (right) atmospheric correction (image:
Landsat TM, 25-09-2009).

Generally the interpretation showed that in cloud-free image with low water vapour optical
thickness (≈ < 0.05) atmospheric correction can increase the quality of the satellite image and
therefore improve the interpretation.

3.5 Verification of buried archaeological sites

Field spectroscopy can be also used for detection of buried archaeological remains. The
advantage of using ground spectroscopy against satellite remote sensing, is based on the
fact that the researcher may repeatable use such methodology in contrast to the temporal
resolution of satellite images (e.g. 16 days for Landsat images). Although spatial resolution
is increased (few cm) the extent and scale of spectroscopy is limited compared to the area
coverage of a single satellite image.
For the verification of known archaeological sites using ground spectroscopy, GER 1500
spectroradiometer was used in several vegetated archaeological sites. In this chapter results
from field campaigns over Neolithic tells in Thessaly (central Greece) and buried remains in
the Palaepaphos area (SW Cyprus) are presented (Agapiou et al., 2010; Agapiou and
Hadjimitsis,2011 ; Agapiou et al., 2012a). In each archaeological site several sections were
carried out. For the first site, along each section, more than 50 ground spectroradiometric
measurements were taken while in each consecutive 5th measurement the calibration
spectralon panel was used in order to minimize sun changes illuminations. At the second case
study measurements were taken over known geophysical anomalies (potential subsurface
monuments). To avoid differences due to variations in cultivation techniques, all
measurements were carried out within the same parcel. As it shown in Fig. 11, vegetation
indices such NDVI and Simple Ratio tend to give higher values at the highest peak of the tell,
Remote Sensing Applications in Archaeological Research 449

similar to other flat – healthy crops of the area, while the slope of the tell gives lowest NDVI
and SR values. This is due to the fact that top of the tell seems to have similar hydrological
behaviour as the flat healthy region (e.g. same level of water surface run off and similar
inclination ≈ 0%) in contrast to the slope of the tell. The sloping part of the tell seems to behave
differently due to rainfall erosion processes. All these results denote the correlation between
the morphology and the spectral response of canopy on the magoules (Agapiou et al., 2012a).
Moreover ground spectral signatures at Palaepaphos area (Fig. 11) indicated a stress condition
for crops over the geophysical anomaly in contrast to the rest of the measurements. This stress
condition was detected from ground spectroradiometric measurements as shown in Fig. 11.

Fig. 11. NDVI (Top left) and Simple Ratio (Top right) profile over archaeological site at
Thessaly. Spectral signature profiles over geophysical anomaly (indicated with arrow) at
Palaepaphos (Bottom).

The results of Thessaly were able to be confirmed using Landsat TM/ETM+ images. Indeed
as it was found the similar characteristics were observed and in satellite images (Fig. 12).
Therefore using this experience of the spectroradiometer, where ground hyperspectral data
were collected, a researcher focusing in satellite imagery can seek and search for similarly
spectral characteristics as those in the spectroradiometric campaign.

3.6 Monitoring phenological cycle of crops

Monitoring the phenological cycle of crops for archaeological sites has been very limited
discussed in the literature. Nevertheless as Agapiou and Hadjimitsis (2011) argue, that this
approach may be used –under some assumptions- for the detection of buried archaeological
remains. This methodology may be used in cases were spatial resolution of satellite imagery
is very low or the cost of high multispectral satellite imagery is forbidden for an
archaeological research. The basic theory of the applied method is based on the different
spectral signature characteristics of ‘stressed’ (negative crop mark due to buried walls) and
‘non-stressed’ (i.e. healthy) vegetation based on the following two criteria: (a) similar soil
characteristics & (b) similar climatic characteristics. The determination of spectral signatures
of barley can be also verified using field spectro-radiometric measurements.
450 Remote Sensing – Applications

Fig. 12. NDVI (Top left) and Simple Ratio (Top Right) images over archaeological site at
Thessaly using Landsat TM image. Characteristics sections of NDVI (Bottom left) and
Simple Ratio (Botom right) over the tell Nikaia 6.

In their study Agapiou and Hadjimitsis (2011) and Agapiou et al. (2012b) have used fifteen
Landsat TM and ETM+ images all freely available from USGS Glovis database. After
applying the necessary pre-processing steps, such as geometric and radiometric corrections,
the NDVI algorithm was applied in three selected case studies where barley crop was
cultivated. The whole phenological cycle of barley crops was examined for a period of one
year, from June 2009 until June 2010, using Landsat TM/ ETM+ images, in order to detect
areas of “possible” archaeological remains indicated as spectral signatures anomalies. Site 1
was an archaeological area excavated in July 2010 by the Department of Antiquities, while
sites 2 and 3 were healthy sites. Moreover site 3 was in close proximity to site 1 in order to
minimize errors due to different climatic or soil characteristics.
At the same time meteorological data have not shown any significant variations over these
sites (temperature, precipitation and humidity). Fig.13 shows the red and near infrared
values during the phenological cycle. As it is expected in a healthy situation (similar to the
Tasseled Cap algorithm, see Kauth and Thomas 1976) after the first rains the vegetation starts
to grow until its reach to its highest peak (see Site 2, Fig.13). However this is not the case for
stress crops as in the case of the archaeological site (site 1). A stress condition is indicated as it
is shown in Fig.13 (Point D) which may be related to the presence of archaeological remains
in the area. Fig.13 presents the phenological cycle of the three sites as examined by Agapiou
and Hadjimitsis (2011). As indicated in 07/01/2010 an immediate drop of NDVI value was
found for Site 1 (archaeological site). The low NDVI value could be explained as a result of
the presence of areas of potential archaeological site, which affected the growth of the crop.
The agricultural barley crop in Site 1 can be characterized as a “stressed” vegetation (negative
crop mark predominantly found above walls). The excavations carried out in the area have
Remote Sensing Applications in Archaeological Research 451

Fig. 13. Red band against NIR band during the phenological cycle of crops (Agapiou and
Hadjimitsis, 2011) (left). NDVI for archaeological and non archaeological areas. In situ
Spectroradiometric measurements are plotted as dots (right).

verified that this localized crop stress was due to the presence of archaeological remains
(walls). Positive crop marks, due to ditches (crop vigour), were not found either in the
methodology applied or in the excavated area. In the non-stress area of Site 3, crops were still
growing this period indicating a peak of the NDVI value.Crops in Site 2 have not yet grown
at this time. From 07/01/2010 until 31/05/2010, crops in Site 2 start to grow gradually until
the harvest period. The chronological shift (X – axis) occurred in NDVI peaks for Sites 1, 3 (at
22/12/2009) and Site 2 (at 19/03/2010), as recorded from the satellite images, were due to
different land management of the fields in these two areas.

4. Ground based geophysical techniques

Electrical resistivity tomography (ERT) uses various combinations of distances among
equally spaced electrodes (of a specific array) in order to extract information about the
lateral and vertical variations of the apparent resistivity.
Magnetic methods on the other hand are passive, measuring the magnetic properties of the
soils or the magnitude of the local magnetic field of the earth as it is modified by the
magnetic properties of the underlying features. The local magnetic field of the earth is
modified by either the enhancement of iron oxides due to past activities or due to the
burning of features, which upon cooling they keep a permanent thermoremanent
magnetization, quite distinct from the current magnetic field (in intensity and direction)
(Aitken, 1974 ; Sarris & Jones, 2000). Based on this property, magnetometers measure either
the total magnetic field (through proton precession or cesium magnetometers) or one of its
components, for detecting variations of the magnetic field (magnetic anomalies) that can be
caused by anthropogenic buried agents (Nishimura, 2001). In order to account for the
diurnal variations of the magnetic field and counterbalance effects resulting from geological
trends, these instruments were also used in a gradient mode. Similarly, the use of fluxgate
gradiometers (measuring the vertical magnetic gradient) was essential for increasing the
sensitivity (although not comparable to that of the cesium or optically pumped alkali vapor
magnetometers) and sampling pace of the magnetic surveys (Fig. 14). Penetration depth of
the magnetic gradient surveys depends on the distance between the two vertical magnetic
sensors.
452 Remote Sensing – Applications

Fig. 14. Results of the usage of Geoscan FM256 and Bartington G601 fluxgate gradiometers
for the mapping of the Early Copper Age (ca. 4,500-3,900 BC) settlement of Veszto-Bikeri, in
the Great Hungarian Plain (left image) and the urban center of ancient Nikopolis (Epiros,
NW Greece) (right image). The magnetic survey on the Hungarian Bronze Age tell revealed
three circular ditches encircling the settlement, which consists of a dense cluster of structural
remains (rectangular houses, pits, kilns, etc) (Sarris et al. 2004). In the case of ancient
Nikopolis, a large complex building (45x70m) with a number of internal divisions appeared
right at the edge of the SW corner of the Byzantine Paleochristian walls (Sarris et al., 2010).

Electromagnetic methods (EM), including the ground penetrating radar (GPR) and the soil
conductivity techniques (SC), have been also employed for the prospection of archaeological
sites and the reconstruction of the ancient terrains. The Slingram type of soil conductivity
devices (such as Geonics EM31 or EM38) operate at low frequencies (usually at the range of 50-
300kHz), make use of the electromagnetic induction and are capable of providing
measurements of both soil's apparent electrical conductivity (quadrature component) and
apparent magnetic susceptibility (in-phase component), with various penetration depths
depending on the orientation (vertical coplanar (VCP) and horizontal coplanar (HCP)
orientations), the frequency of operation and separation of the transmitter and receiver coils
(Dalan, 2006; Gaffney & Gater, 2003; Cheethman, 2010). The strength of EM signals that are
registered by the receiver system depends on the conductivity of the soils, the magnetic
permeability and the dielectric permittivity (especially for the GPR). Operating within the
range of radio frequencies, GPR systems consist of a transmitter antenna that sends a signal
(~30-1000MHz) which propagates through the different strata or features (reflectors) of the
subsurface and a receiver antenna that registers all the secondary reflections (with a modified
amplitude) that arrive to it after a time delay which is converted to the depth. Signal
attenuation and penetration depth of the GPR decreases with the increase of the frequency of
the antennas and the conductivity of the soils. GPR signals are collected with high sampling
rate along transects and the resulting reflection sections (radargrams) represent the variation of
the amplitude of the reflected signal with depth and thus they depict an image of the
stratigraphy of the subsurface. GPR parallel transects are usually combined to created 3D
volumetric maps of the subsurface and through the isolation of specific time or depth slices it
is possible to allocate the horizontal extent of archaeological features at different depths, which
is of importance especially in cases that one wants to have information regarding the vertical
extent of the features or to construct 3D models of them. In this sense, the GPR survey can be
valuable in mapping different occupation strata and resolving features that are located at
various depths (Fig. 15) (Conyers, 2004; Conyers & Goodman, 1997).
Remote Sensing Applications in Archaeological Research 453

(a) (b) (c) (d)

(a) Section of a Stoa from the ancient Agora of Feres (Velestino) in Thessaly. The (image dimensions:
ExN=16x40m).
(b) Monumental structural remains SE of the Zeus temple in the archaeological site of Nemea,
Peloponesse. (image dimensions: ExN=7x15m) (Papadopoulos et al. 2011).
(c) Architectural complex at a depth of 90-100cm below the surface located to the east of the Agora of
Sikyon at NE Peloponesse (image dimensions: ExN=30x50m).
(d) The 2.5-3m GPR depth slice from the area of the hypothesized amphitheatre of Ierapetra (SE
Crete). The survey was carried out in the area that was suggested through the rectification of the
map of the British Vice-Admiral Thomas Spratt (1811-1888) (which was depicting the approximate
location of the amphitheatre) on the satellite (Quickbird) image of the region. The deeper GPR slices
in combination to ERT measurements provided evidence for the underlying relics of the
amphitheatre (Sarris et al., 2011).
Fig. 15. Examples of GPR time slices obtained by Nogin Plus (Sensors & Software) GPR
system using a 250MHz antenna.

Even microgravity measurements have been carried out for the detection of features that
have a substantial mass density contrast with respect to their surrounding geological
domain, creating a difference at the local gravitational acceleration. Measurements of earth's
gravitational acceleration are carried out though the use of gravimeters that measure the
acceleration of gravity within a hundredth of a mGal (1Gal=1cm/sec2) or less. As such, the
resolution of the method is dependent on the size and volume of the targets and requires
tedious corrections and processing (such as drift correction, latitude correction, free air
correction, Bouguer correction, etc) as measurements are influenced by the regional or even
local trends. A recent review of archaeological, environmental and geological microgravity
applications has been provided recently by Eppelbaum (2011).
Mainly used for landscape reconstruction, large monumental structure detection and deep
prospection surveys, seismic techniques exploit acoustical waves generated either by a
sledge hammer or an explosion discharge. In seismic refraction, the acoustic wave sensors
(geophones) are laid along specific distances and record the refracted signals with respect to
their arrival times and in this way their velocity of propagation (increasing with depth) is
measured. Seismic reflection techniques require a smaller distance between the source and
the geophones and through the examination of the arrival times, amplitude and shape of the
reflected waves, we can conclude on the types of the subsurface interfaces (Metwaly et al,
2005 ; Scott & Markiewich, 1990).
If the above techniques are capable of providing a mean of detection and localization of
architectural features within an archaeological site, magnetic susceptibility (MS)
measurements and chemical analyses can contribute in providing a further tool for
investigating the land use patterns at a specific area. Magnetic susceptibility provides not
454 Remote Sensing – Applications

only a measure of the effectiveness of the potential application of magnetic surveys (through
the estimation of the normalized Le Borgne Contrast, namely the variation of the magnetic
susceptibility with depth), but also an index of the past workshop activities in an area.
Measurements of the magnetic susceptibility and the frequency dependent susceptibility
(namely the variation of MS with the frequency of an induced magnetic field) are capable in
distinguishing soils enriched in single domain magnetic particles (from the geological origin
multidomain particles) which are indicative of the intensity of the occupation of a site
(Clark, 1990; Mullins & Tite, 1973; Thompson & Oldfield,1986). Coupled with results of
chemical properties of soils (especially those dealing with phosphate analysis or heavy
metals tracing) it is possible to characterize the type of workshop activities (eg. increase of
manganese content can be associated to glass workshop activities) or differentiate areas
used for animal husbandry, midden deposits, foundation trenches, cultivation, cooking, etc.
Even chemical stability of certain organic chemical compounds (e.g. coprastanol) may act as
a biomarker of the human presence at a particular locatio (Sarris, 2008).
The choice of the technique depends mainly on a number of factors: the type of the targets,
their lateral and vertical dimensions, their deposition depth and type/properties of the
surrounding soils (to be able to create a significant signal, contrast or "anomaly"). Architectural
features such as stone/brick structures, roads, walls, built/chamber or rock-cut tombs, can be
relatively easily resolved through soil resistance or GPR surveys. Brick structures or
architectural features that are either burnt or contain residues of heating/burning, kilns,
workshop facilities, slag deposits, metal concentrations, and sometimes roads, walls and
fortifications can be detected through magnetic and electromagnetic techniques. The use of
ERT, GPR and microgravity is especially useful for the identification of vaults, caves, chamber
tombs and fissures. Shallow depth surveys usually employ magnetic, soil resistance
techniques and the use of GPR. In cases where deeper penetration is required GPR, ERT and
seismic approaches are more appropriate (Fig. 3) (Sarris 2008, Linford 2006).
Ground based prospection techniques are not only limited to the survey of archaeological
sites in an open/rural context (Fig. 16). They can also be applied within an urbanized
environment, but in such a case only specific techniques can be used (such as ERT and GPR)
that are influenced as less as possible by the modern interventions and structures that exist
in the urban matrix. A lot of these applications do not only involve the mapping of the
subsurface (eg. below asphalt roads, pavements, concrete blocks, etc), but sometimes they
are oriented towards the stability or structural damage assessment of monuments or
historical structures aiming towards their architectural restoration (Bertroli et al., 2011;
Pettinelli et al., 2011; Utsi, 2010; Masini et al., 2010).
Currently there are two different tendencies in archaeological prospection: the integration of
different geophysical techniques for maximizing the information content and the
employment of multi-sensor methods for the rapid coverage of sites. In most cases, the
integrated use of various techniques is employed to extract more information about an
archaeological site, allowing the interpretation of various measurements that are dealing
with different properties of the soil. The fusion of this information permits a more holistic
approach as the data can complement each other and provide a more integral plan of the
subsurface features. One the other hand, the recent development of new multi-sensor (for
magnetics), multi-antennas (for GPR) or multi-electrode (for soil resistivity) motorized
systems carrying DGPS allow the fast and detailed assessment of large regions, although
Remote Sensing Applications in Archaeological Research 455

Fig. 16. Example of the application of ERT in the area of the assumed hippodrome in the
archaeological site of Nemea, Peloponesse. A number of ERT transects was materialized
reaching the depth of about 9m below the surface. Upon the synthesis of all the 2D
resistivity inverted sections (a), it was possible to create a scatter plot of all the sections (b)
and through interpolation techniques the 3D volumetric resistivity map. The isolation of the
various strata of the subsurface was based on the range of their resistivity values (c). In the
particular case, the iso-resistivity surfaces that resulted from the ERT transects did not
identify any specific leveling of the subsurface at the west side of the archaeological site
where the hippodrome was expected, suggesting that the original hypothesis of the
archaeologists has to be rejected (Papadopoulos et al., 2011).

they are restricted by the surface coverage and the terrain morphology (Linford et al., 2011;
Doneus et al., 2011). Although the particular systems offer increased sampling density and
rates of coverage, they often suffer from positioning errors due to high measuring velocities
and the introduction of noise due to the non uniform balancing of the sensors or
multichannel GPR systems (Zollner et al., 2011; Verdonck & Vermeulen , 2011).
Finally, image processing techniques play a significant role in the visualization of the results
of the geophysical surveys as the ultimate goal is to provide images that they depict the
underlying features at their exact location and horizontal/vertical extent in a way that can
approach the results of an after-the-excavation plan. This objective can be achieved through
the use of a number of filtering/convolution processes, the employment of synthetic models
or inversion algorithms, or other image processing functions (Papadopoulos & Sarris, 2011;
Sarris, 2008; Loke & Barker, 1996; Scott & Markiewich, 1990). In cases that multiple datasets
are available for the same region, composites can be made using visualization techniques
similar to those used in satellite remote sensing (Böniger & Tronicke, 2010). Even more
impressive visualization can be created through the fusion of geophysical data with satellite
remote sensing or aero-photogrammetric data and lidar or terrestrial 3D laser scanning
(Bem et al., 2011). Indeed, the continuous improvement of high resolution satellite remote
sensing sensors has made possible their simultaneous utilization with conventional
geophysical data affecting their resolution and potential in the detection and mapping of
underground features (Crespi et al., 2011).

5. Conclusions
The various approaches applied on different satellite images for the detection of Neolithic
settlements in Thessaly illustrated the benefits that satellite remote sensing can provide in
456 Remote Sensing – Applications

archaeological investigation. It was proven that an integration of images from different

satellite sensors can contribute to a faster and more accurate and qualitative detection of
archaeological sites. In addition, the GIS spatial analysis and DEM processing contributed
substantially to the detection and monitoring of settlements and modeling of Neolithic
habitation partners in Thessaly.
Moreover, it was proved that spectroradiometric measurements can be used as an
alternative approach in order to identify buried archaeological remains, since they can
provide accurate spectral signatures for a wide spectral region. Anomalies of the crop
spectral signatures due to buried archaeological remains can be recorded in detail and
contribute to the construction of a predictive archaeological model in the future. However,
the real benefit of this instrument is when it is used in conjunction with satellite images.
Moreover the spectroradiometric measurements highlight the high correlation of spectral
response of archaeological material, sand and local geological formations in the area of red
visible band. Finally it was proved that the monitoring of the phenological cycle of crops can
be used for the detection of buried archaeological sites.
At the end geophysical survey, ground and space remote sensing methods have been
gradually adopted in archaeological research in an effort to capture the residues of the
past anthropogenic activity underlying below the current surface of the ground and to
provide a more synthetic and holistic image of the archaeological landscapes. Magnetic
and resistivity techniques, electromagnetic, gravity and seismic methods, measurements
of the chemical and magnetic properties of the ground, have been all mobilized to
produce an accurate picture of the underlying monuments, contributing to a variety of
applications in both urban and rural environmental settings. In this way all these different
methodologies Have been applied to various areas of potential archaeological interest
throughout Europe. The tuning of the methods and the corresponding instrumentation,
together with the development of specific processing algorithms, are necessary in order to
enhance the shallow depth signals that are registered within the increased noise levels of
upper horizons of the soil. In this way, shallow depth prospection techniques have been
used to map architectural relics, to guide excavations, and to identify craft, workshop,
agricultural or animal husbandry activities.

6. Acknowledgments
Part of this chapter is part of Dr. Dimitrios Alexakis Phd thesis. This research project was
co-financed by INSTAP (Institute for Aegean Prehistory) and PENED (E.U.-European
Social Fund (75%) and the Greek Ministry of Development-GSRT (25%))."Field
Spectroradiometer and Archaeology" chapter results are part of the PhD thesis of Mr.
Athos Agapiou. The authors would like to express their appreciation to the Alexander
Onassis Foundation for funding the PhD study. Thanks are given to the archaeologists Dr.
K. Vouzaxakis and Dr. J. Lolos, for their valuable assistance during field campaigns at the
Neolithic tells (Thessaly) and the archaeological site of Sikyona, Corinthos and to the
Department of Antiquities of Cyprus for their permission to carry out filed measurements
at different archaeological sites of Cyprus. The whole project has also been co-financed by
the internal programs “Integration” and “Monitoring”of Cyprus University of
Technology. Finally thanks are given to the Remote Sensing Laboratory of the Department
Remote Sensing Applications in Archaeological Research 457

of Civil Engineering & Geomatics at the Cyprus University of Technology for the support
(http://www.cut.ac.cy).

7. References
Agapiou, A., Hadjimitsis, D.G., Alexakis, D., Sarris, A. (2012a). Observatory validation of
Neolithic tells ("Magoules") in the Thessalian plain, central Greece, using
hyperspectral spectroradiometric data. Journal of Archaeological Science, 39 (5): 1499-
1512, 10.1016/j.jas.2012.01.001.
Agapiou, A., Hadjimitsis, D.G., Alexakis, D., Papadavid G. (2012b). Examining the
Phenological Cycle of Barley (Hordeum vulgare) for the detection of buried
archaeological remains: the case studies of the Thessalian plain (Greece) and the
Alampra (Cyprus) archaeological test fields. GIScience and Remote Sensing (in
press).
Agapiou A.and Hadjimitsis D. G.,(2011). Vegetation indices and field spectro-radiometric
measurements for validation of buried architectural remains: verification under
area surveyed with geophysical campaigns, Journal of Applied Remote Sensing, 5:
053554-1.
Agapiou A.; Alexakis D.; Hadjimitsis D.G. & Themistocleous K. (2011a). Earth Observations
and Ground Measurements for Cultural Heritage Protection: the Case Study of
Cyprus, 1st International Conference on Safety and Crisis Management in the
Construction, SME and Tourism Sectors, June 24th – 28th 2011, Nicosia, Cyprus (in
press).
Agapiou A., Hadjimitsis D. G., Papoutsa C., Alexakis D. D., Papadavid, G. (2011b). The
Importance of Accounting for Atmospheric Effects in the Application of NDVI and
Interpretation of Satellite Imagery Supporting Archaeological Research: The Case
Studies of Palaepaphos and Nea Paphos Sites in Cyprus. Remote Sensing, 3: 2605-
2629, doi: 10.3390/rs3122605.
Agapiou A.; Hadjimitsis G. D.; Sarris A.; Themistocleous K. & Papadavid G. (2010).
Hyperspectral ground truth data for the detection of buried architectural remains,
Lecture Notes in Computer Science 6436, pp. 318–331, 2010.
Aitken, M. (1974). Physics and Archaeology, 2nd ed., Oxford: Clarendon Press.
Alexakis, D., Astaras, Th., Sarris, A., Vouzaxakis, K., Karimali, L., (2008). Reconstructing the
Neolithic Landscape of Thessaly through a GIS and Geological Approach. In:
Polsluchn, K. Lambers and I. Herzog (eds.), Layers of Perception. Proceedings of the
35th International Conference on Computer Applications and QuantitativeMethods in
Archaeology/(CAA), Berlin, Germany.
Alexakis, D., Sarris, A., Astaras, T., Albanakis, K., (2011). Integrated GIS, remote sensing and
geomorphologic approaches for the reconstruction of the landscape habitation of
Thessaly during the neolithic period. Journal of Archaeological Science 38, 89-100.
Alexakis, D.; Sarris, A.; Astaras, T.;Albanakis, K. (2009). Detection of Neolithic settlements in
Thessaly (Greece) through multispectral and Hyperspectral satellite Imagery.
Sensors 9 (2), 1167- 1187. doi:10.3390/s90201167.
Atkinson P. M.; Webster R. & Curran P. J., (1992). Cokriging with Ground-Based
Radiometry, Remote Sensing of Environment, 41, pp. 45−60.
458 Remote Sensing – Applications

Bartroli, S. R., Garsia, G. E. & Tamba, R. (2011). GPR Imaging of Structural Elements. Case
Study of the Restoration Project of the Modernist Historic Site of Sant Pau, in
Archaeological Prospection - Extended Abstracts, ed. by M. G. Drahor & M. A.
Berge, 9th International Conference of Archaeological Prospection, Izmir, Turkey, pp.
149-151.
Bem, C.; Bem, C; Asandulesei, A.; Venedict, B. & Cotiuga, V., (2011). Identity in Diversity, in
Archaeological Prospection - Extended Abstracts, ed. by M. G. Drahor & M. A.
Berge, 9th International Conference of Archaeological Prospection, Izmir, Turkey, pp. 25-
28.
Böniger, U. & Tronicke, J. (2010). Integrated data analysis at an archaeological site: A case
study using 3D GPR, magnetic, and high-resolution topographic data, Geophysics,
Vol. 75, No. 4, pp. B169-B176.
Cavalli, R.M.; Colosi, F.; Palombo, A; Pignatti, S. &Poscolieri, M. (2007).Remote
Hyperspectral Imageryas a support to Archaeological Prospection, J. Cult. Herit., 8,
272-283.
Che, N. & Price, J. C. (1992) Survey of Radiometric calibration results and methods for
visible and near infrared channels of NOAA-7, -9, and -11 AVHRRs. Remote Sensing
of Environment, 41, 19 – 27.
Cheetham, P. (2010). An Empirical Reassessment Of The Utility Of The Geonics Em38B,
Together With Suggested Methodologies For Its Application In Archaeological
Investigations, Meeting of the New Surface Geophysics Group - Recent Work in
Archaeological Geophysics, London.
Clark, A. J. (1990). Seeing Beneath the Soil. Prospecting Methods in Archaeology. London: B.T.
Batsford Ltd.
Conghe, S. & Woodcock, E. C. (2003) Monitoring Forest Succession With Multitemporal
Landsat Images: Factors of Uncertainty, IEEE Transactions on Geoscience and Remote
Sensing, 41 (11), 2557 -2567.
Conyers, B. L. & Goodman, D. (1997). Ground Penetrating Radar: An Introduction for
Archaeologists, Walnut Creek, CA., Altamira Press.
Conyers, L B. (2004). Ground Penetrating Radar for Archaeology, Walnut Creek, CA: AltaMira
Press.
Crespi, M.; Dore, N.; Partuno, J.; Piro, S. & Zamuner, D. (2011). Comparison of SAR data,
Optical Satellite Images and GPR Investigations for Archaeological Site Detection,
in Archaeological Prospection - Extended Abstracts, ed. by M. G. Drahor & M. A.
Berge, 9th International Conference of Archaeological Prospection, Izmir, Turkey, pp.
169-173.
Curran P. J. & Williamson H. D., (1986). Sample Size for Ground and Remotely Sensed Data,
Remote Sensing of Environment, 20, pp. 31−41.
Dalan, R.A. (2006). Magnetic Susceptibility in Johnson, J.K. (ed.) Remote Sensing in
Archaeology: An Explicitly North American Perspective, Alabama Press: Tuscaloosa,
162-203.
De Laet, V.; Paulissen, E. & Waelkens, M. (2007). Methods for the extraction of
archaeological featuresfrom very high-resolution IKONOS-2 remote sensing
imagery, Hisar (southwest Turkey). J.Archeol. Sci., 34, 830–841.
Remote Sensing Applications in Archaeological Research 459

Demitrack, A. (1986). The Late Quaternary Geologic History of the Larisa Plain,Thessaly,
Greece: Tectonic, Climatic and Human Impact on the Landscape, Ph.D.thesis,
Stanford University.
Doneus, N.; Flory, S.; Hinterleitner, A.; Kastowsky. K.; Kucera, M.; Nau, E.; Neubauer, W.;
Scherzer, D.; Schreg, R.; Trinks, I.; Wallner, M. & Sitz, T. (2011). Integrative
Archaeological Prospection - Case Study Stubersheimer Alb. Bridging the Gap
Between Geophysical Prospection and Archaeological Interpretation, in
Archaeological Prospection - Extended Abstracts, ed. by M. G. Drahor & M. A.
Berge, 9th International Conference of Archaeological Prospection, Izmir, Turkey, pp.
166-168.
Eppelbaum, V. L. 2011. Review of Environmental and Geological Microgravity Applications
and Feasibility of Its Employment at Archaeological Sites in Israel, International
Journal of Geophysics, vol. 2011, issue 1, pp. 1-9.
Fowler, M.J.F. (2002). Satellite Remote Sensing and Archaeology: a Comparative Study of
Satellite Imagery of the Environs of Figsbury Ring, Wiltshire. Archaeol.Prospec., 9,
55-69.
Fry, G.L.A.; Skar, B.; Jerpansen, G.; Bakkestuen, V. & Erikstad, L., (2004). Locating
archaeological sites in the landscape: a hierarchical approach based on landscape
indicators. Landscape and Urban Planning 67, 97-107.
Gaffney, C. & Gater, J. (2003). Revealing the Buried Past, Tempus: Stroud.
Glenn E. P.; Huete A. R. ; Nagler P. L. & Nelson S. G., (2008). Relationship Between
Remotely-sensed Vegetation Indices, Canopy Attributes and Plant Physiological
Processes: What Vegetation Indices Can and Cannot Tell Us About the Landscape.
Sensors, 8, 2136-2160.
Hadjimitsis, D.G.; Papadavid G.; Agapiou A..; Themistocleous K.; Hadjimitsis M. G.; Retalis
A.; Michaelides S.; Chrysoulakis N.; Toulios L. & Clayton C. R. I. (2010).
Atmospheric correction for satellite remotely sensed data intended for agricultural
applications: impact on vegetation indices, Nat. Hazards Earth Syst. Sci., 10, 89-95 .
Hadjimitsis, D.G.; Clayton, C.R.I. & Hope, V.S. (2004). An assessment of the effectiveness of
atmospheric correction algorithms through the remote sensing of some reservoirs.
International Journal of Remote Sensing., 25, 3651-3674.
Iwahashi, J., & Kamiya, I., (1995). Landform classification using digital elevation model by
the skills of image procesingdmainly using the Digital National Land Information.
Geoinformatics 6 (2), 97 - 108 (in Japanese with English abstract).
Jonhson J. K., (2006). Remote Sensing in Archaeology, The University of Alabama Press,
Tuscaloosa.
Kauth R. J. & Thomas G. S. (1976). The tasseled Cap - A Graphic Description of the Spectral-
Temporal Development of Agricultural Crops as Seen by LANDSAT”. In
Proceedings of the Symposium on Machine Processing of Remotely Sensed Data,
Purdue University of West Lafayette, Indiana, 4B,44-51.
Lillesand, T. M.; Kiefer, R. W. & Chipman, J. W. (2004). Remote Sensing and image
interpretation, Wiley International Edition.
Linford, N. (2006). The Application Of Geophysical Methods To Archaeological Prospection.
Reports on Progress in Physics, v.69, pp. 2205–2257.
460 Remote Sensing – Applications

Linford, N.; Linford, P.; Payne. A.; David, A.; Martin, L. & Sala, J. (2011). Stonehenge: Recent
Results from a Ground Penetrating Radar Survey of the Monument, in
Archaeological Prospection - Extended Abstracts, ed. by M. G. Drahor & M. A.
Berge, 9th International Conference of Archaeological Prospection, Izmir, Turkey, pp. 86-
89.
Loke, M. H. & Barker, R. D. (1996). Rapid Least-Squares Inversion of Apparent Resistivity
Pseudo-Sections using Quasi-Newton method. Geophysical Prospecting, v.48, pp.
181-152.
Martonchik, J. V.; Bruegge, C. J. & Strahler, A. H. (2000). A review of reflectance
nomenclature used in remote sensing. Remote Sensing Reviews, 19, pp. 9−20.
Masini, N. & Lasaponara, R. (2007). Investigating the spectral capability of Quickbird data to
detectarchaeological remains buried under vegetated and not vegetated areas, J.
Cult. Herit., 8, 53-60.
Masini, N.; Persico, R. & Rizzo, E. (2010). Some Examples Of GPR Prospecting For
Monitoring Of The Monumental Heritage, Journal Of Geophysics And Engineering,
vol. 7, pp. 190–199.
Menze, B.H. & Sherratt, A.G. (2006). Detection of Ancient Settlement Mounds:
Archaeological SurveyBased on the SRTM Terrain Model. Photogramm. Eng. Remote
Sens., 72, 321-327.
Merola, P.; Allegrini, A.; Guglierra, D. & Sampieri, S. (2006). Buried Archaeological
Structures Detection Using MIVIS Hyperspectral Airborne Data, In Proceedings of
SPIE, the International Society for Optical Engineering, 2006; pp. 62970Z.1-62970Z.12.
Metwaly, M.; Green, A. G.; Horstmeyer, H.; Maurer, H.; Abbas, A. M. & Hassaneen, A. G.
(2005). Combined Seismic Tomographic And Ultrashallow Seismic Reflection Study
Of An Early Dynastic Mastaba, Saqqara, Egypt. Archaeological Prospection, 12, pp.
245–56.
Milton E. J. & Rollin Ε. Μ. (2006). Estimating the irradiance spectrum from measurments in a
limited number of spectral bands, Remote Sensing of Environment 100, pp. 348-355.
Milton, E. J. (1987). Principles of Field Spectroscopy. Remote Sensing of Environment 8 (12),
1807--1827.
Milton, E. J.; Schaepman, M. E.; Anderson, K.; Kneubühler, M & Fox, N. (2009). Progress in
Field Spectroscopy. Remote Sensing of Environment 113, 92--109.
Mullins, C. E. & Tite, M. S. (1973). Magnetic Viscocity, Quadrature Susceptibility and
Frequency dependence of Susceptibility in Single-Domain Assemblies of Magnetite
and Maghaemite, Journal of Geophysical Research, no. 78, pp. 804-809.
Nicodemus, F. F.; Richmond, J. C.; Hsia, J. J.; Ginsberg, I.W. & Limperis, T. L. (1977).
Geometrical considerations and nomenclature for reflectance. National Bureau of
Standards Monograph, 160, Washington, D.C U.S. Govt. Printing Office.
Nishimura, Y. (2001). Geophysical prospection in Archaeology in D. R. Brothwell and A. M.
Pollard (eds.), Handbook of Archaeological Sciences, pp.543-553.
Papadopoulos N.G.; Sarris, A.; Michalopoulou, S. & Salvi, M.C. (2011). Integrated
Geophysical Investigations in Nemea and Tsoungiza, 9th International Conference of
Archaeological Prospection, September 19 – 24, Izmir, Turkey.
Remote Sensing Applications in Archaeological Research 461

Papadopoulos, N. & Sarris, A. (2011). An Algorithm for the fast 3-D Invesrion of Direct
Current Resistivity Data Using LSMR, in Archaeological Prospection - Extended
Abstracts, ed. by M. G. Drahor & M. A. Berge, 9th International Conference of
Archaeological Prospection, Izmir, Turkey, pp. 197-200.
Pavlidis, L. (2005). High resolution satellite imagery for archaeological application.
www.fungis.org/images/newsletter/205-1.pdf
Pettinelli, E.; Barone, P.M.; Mattei, E. & Lauro, S.E. (2011). Radio Waves Technique for
Nondestructive Archaeological Investigations. Contemporary Physics, Taylor &
Francis Group, London, Volume 52, Issue 2, pp. 121-130.
Rowlands, A. & Sarris, A. (2006). Detection of exposed and subsurface archaeological
remains using multi– sensor remote sensing. J. Archaeol. Sci., 34, 795-803.
Sarris, A.; Papadopoulos N.G.; Salvi, M.C. & Dederix, S. (2011). Preservation Assessment of
Ancient Theatres through Integrated Geophysical Technologies, XVIth Congress of
the International Union of Prehistoric and Protohistoric Sciences (UISPP), Florianopolis,
Brazil.
Sarris, A.; Teichmann, M.; Seferou, P. & Kokkinou, E. (2010). Investigation of the Urban-
Suburban Center of Ancient Nikopolis (Greece) through the Employment of
Geophysical Prospection Techniques, Computer Applications and Quantitative
Methods in Archeology “Fusion of Cultures” CAA’2010, Fco. Javier Melero & Pedro
Cano (Editors), Granada, Spain.
Sarris, A. (2008). Remote Sensing Approaches / Geophysical, in Encyclopedia of Archaeology,
ed. By Deborah M. Rearsall, Academic Press, New York, vol. 3, pp. 1912-1921.
Sarris, A. (2005). Use of remote sensing for archaeology: state of the art. Presented at the
International Conference on the Use of Space Technologies for the Conservation of Natural
and Cultural Heritage, Campeche, Mexico.
Sarris, A., Galaty, M. L.; Yerkes,R. W.; Parkinson, W. A.; Gyucha, A.; Billingsley, D. M. &
Tate, R. (2004). Geophysical prospection and soil chemistry at the Early Copper
Age settlement of Vésztó-Bikeri, Southeastern Hungary , Journal of Archaeological
Science, volume 31, Issue 7, pp. 927-939.
Sarris, A. & Jones, R. E. (2000). Geophysical and Related Techniques Applied to
Archaeological Survey in the Mediterranean: A Review, Journal of Mediterranean
Archaeology (JMA), v.13, no.1, pp.3-75.
Schaepman-Strub, G.; Schaepman, M. E.; Painter, T. H.; Dangel, S., & Martonchik, J. V. (
2006). Reflectance quantities in optical remote sensing-Definitions and case studies.
Remote Sensing of Environment, 103, pp. 27−42.
Scott, J. H. & Markiewich, R. D. (1990). Dips and chips-PC programs for analyzing seismic
refraction data: Proceedings, SAGEEP 1990, Golden, Colorado, pp. 175-200.
Thompson, R. & Oldfield, F. (1986). Environmental Magnetism. London: Allen and Unwin.
Utsi, E. (2010). Reflections From Westminster Abbey, Meeting of the New Surface Geophysics
Group - Recent Work in Archaeological Geophysics, London.
Verdonck, L & Vermeulen, F. (2011). 3-D GPR Survey with a Modular System: Reducing
Positioning Inaccuracies and Linear Noise, in Archaeological Prospection -
Extended Abstracts, ed. by M. G. Drahor & M. A. Berge, 9th International Conference
of Archaeological Prospection, Izmir, Turkey, pp. 204-212.
462 Remote Sensing – Applications

Vermeulen, F. & Verhoeven, G. (2004). The contribution of aerial photography and field
survey to thestudy of urbanization in the Potenza valley (Picenum). J. Roman
Archaeol., 17, 57-82.
Wu X., Sullivan T. J. & Heidinger K. A. (2010). Operational calibration of the Advanced Very
High Resolution Radiometer (AVHRR) visible and near-infrared channels.
Canadian Journal of Remote Sensing, 36 (5), 602–616.
Zollner, H.; Kniess, R. & Meyer, C. (2011). Efficient Large-scale Magnetic Prospection Using
Multichannel Fluxgate Arrays and the New Digitizer LEA D2, in Archaeological
Prospection - Extended Abstracts, ed. by M. G. Drahor & M. A. Berge, 9th
International Conference of Archaeological Prospection, Izmir, Turkey, pp. 201-203.
 20

The Mapping of the Urban Growth of

Kinshasa (DRC) Through High Resolution
Remote Sensing Between 1995 and 2005
Kayembe wa Kayembe Matthieu1,
Mathieu De Maeyer2 and Eléonore Wolff2
1University of Lubumbashi
2Université Libre de Bruxelles
1D.R. Congo
2Belgium

1. Introduction
The study of urban growth in Kinshasa is not a new topic, as shown by the work on the
dynamics of housing in the 1970s (Flouriot et al., 1975; Pain, 1978). These authors followed
the spatial extension of Kinshasa by collecting old cartographic documents and comparing
them. Flouriot (1975) combined a cartographic approach with household surveys to follow
the long-term housing growth.
The advent and development of remote sensing and Geographic Information Systems (GIS)
have changed the methods, making it now possible to map and quantify urban growth
quickly and easily. More recently in Kinshasa, Tshibangu et al. (1997) have integrated into a
GIS a vegetation map drawn by Compere in 1960. This was possible thanks to the
interpretation of aerial photographs and Landsat and SPOT images conducted respectively
in 1982 by Wilmet and 1987 by Nsekera to quantify the urban sprawl. Delbart and Wolff
(2002) evaluated the extension of the city of Kinshasa from an old map (1969) and the
delineation of the city in 1995 observed on a SPOT image (from 1995). The current extension
of the city (between 1995 and 2005) is not precisely known, but the figures are around
600km² (Lelo Nzuzi, 2008). The purpose of this chapter is to map and quantify urban growth
between 1995 and 2005 using a time series of high resolution satellite images.

2. Study area
The city of Kinshasa province, located between 4 ° and 5 ° south and between 15 ° and 17 °
east, is the largest city in the Democratic Republic of Congo. It covers an area of 9965 km2
(De Saint Moulin, 2005), about 600 km2 being only urbanized. The city had 400,000
inhabitants in 1960 and reached more than six million in 2008, the average annual growth
rate between 1960 and 2003 would therefore be about 6.80% (Lelo Nzuzi, 2008).
Kinshasa has grown in the plains bordering the Congo River. The plain 300 metres above
sea level covers about 200 km2. This is the most industrialized area and formerly the most
464 Remote Sensing – Applications

densely inhabited, commonly called the “ville basse“ (low city). After independence in 1960,
the city has spread into the complex hills surrounding the city and low peaks around 600 m
above sea level. This area is mainly occupied by slums, called the "high".

3. Data
Two SPOT images dating respectively from March 31, 1995 and July 1, 2005 were used. They
are recorded in panchromatic and multispectral modes. Their radiometric quality is
variable. The 1995 images have a cloud cover of 7% in multispectral and 5% in
panchromatic, while the 2005 images have 6% of cloud cover in multispectral mode and 10%
in panchromatic mode. The presence of these clouds is evidence of the difficulty of
obtaining cloud-free images for areas located in the sub-equatorial climate. To make the
different images comparable, a radiometric correction was performed. Unfortunately, due to
the low correlation between the red and the green bands, it did not yield good results and
was abandoned.
Other data were collected, digitized and georeferenced, if necessary, to analyse urban
growth of the city. This entailed using the old cards to map the growth of Kinshasa over the
long-term, population data and the relief and major roads.
In addition, to map the dynamics of the habitat of the Atlas of Kinshasa (Flouriot, 1975), the
map "District Urban Leopoldville 1/60 000" presents the urban area in 1920. Plan Leopoldville
(map 1/15 500 published by the bookseller Congo Leopoldville) gives the limit of the city in
1954. The map "Plan of Commons of Kinshasa and its Environs" to 1/20 000 published in 1959
by the Geographic Institute of Zaire is the drawing of municipal boundaries of the urbanized
area in 1959. The map "City of Kinshasa-health zones" (Card 1/20 000 published in 1969 and
revised in 1997 from the bottom of the base map of Kinshasa), provisional edition, published
by the Geographical Institute of Zaire has the delineation of municipal boundaries of the
urbanized area in 1969. All these documents are completely overwhelmed by the current
situation (Delbart et al., 2002; Fox et al., 1997) and require updating.
The population data used suffers from both a paucity of quality and reliability in a country
where the offices of the civil state are characterized by operating failure and where the
general census of the population is not regularly organized. With the exception of the 1984
population numbers from the 1984 census, the others are mere projections of the National
Institute of Statistics.
Coverage maps scale 1 / 10 000 by the Geographical Institute of the Belgian Congo (IGCB)
dating from 1958 covering the city of Kinshasa have been scanned. The contours at a contour
interval of 5 metres were digitized by students from MA1 geography at the university,
corrected and interpolated by Mathieu De Maeyer (IGEAT / ULB) by the spline technique to
produce a digital terrain model and derive the slope.
Some roads (in the north of the city and the far east, after the airport) were digitized from
the SPOT panchromatic band (of 10 April 2000) and a plan of the city of Kinshasa (1 / 10
000) of March 1970 created by the Geographical Institute of the Congo. The roads in the west
and south were measured and corrected by DGPS Pathfinder software. The railway was also
digitized from the map of the city of Kinshasa. The roads of the southern part were digitized
using only the SPOT panchromatic band of the 10 April 2000.
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 465

4. Methodology
Two approaches for change detection exist. "Image-image" comparison methods imply a
radiometric normalization; this standardization is difficult to implement on data from
different seasons and radiometric quality is also variable (Singh, 1986; Alphan, 2003; Coppin
et al., 2004; Yuan et al., 2005). In addition, they do not identify the nature of change.
Comparison methods compare the post-classification classifications of land produced
independently at different dates (Gupta et al., 1985). The other group of methods is less
sensitive to differences in season and they identify the nature of change but are susceptible
to misclassification. To detect changes, classifications are compared in pairs. From this
comparison, a map where the changes can be located and a change matrix that summarizes
the amount and the nature of these changes are derived.

4.1 Geometric correction and cutting recovery images

To detect changes, it is essential that the SPOT images are properly stowed from the
geometrical point of view.
This is why the latest panchromatic image has been corrected from an image of higher
resolution. This is a panchromatic IKONOS image from 2002 of a resolution of 1 m corrected
itself with control points measured in absolute mode with a Garmin GP60 GPS. Root mean
square errors of 9.46 m on the hilly part and 4.14 m on the plain were obtained.
Then all the other images SPOT (panchromatic and multispectral mode) were corrected on
the panchromatic SPOT image, corrected with a polynomial function of first order and the
nearest neighbour method. All are projected onto the ellipsoid WGS 84 UTM coordinates,
zone 33 south.
Geometric corrections lead to RMS errors smaller than the size of a pixel with 29 to 35
control points (Table 1), which is acceptable according to Moller-Jensen (1990) and is
suitable for a detect changes study.

Control points XRMS XRMS YRMS YRMS

Image Cell (m)
number (cell) (m) (cell (m)
Spot
panchromatic 10 30 0.44 4.4 0.57 5.7
1995
Spot multispectral
20 34 0.49 9.8 0.54 10.8
1995
Spot multispectral
20 29 0.53 10.6 0.44 8.8
2005

Table 1. RMS errors after geometric correction

Not all SPOT images have the same spatial extension. In addition, their size being 60 km on
each side, is wider than the extension of the city of Kinshasa. The images of 1995 and 2005
were cut to the same extension.
466 Remote Sensing – Applications

4.2 Land use classification

Given the uneven quality of SPOT images and the strong texture of the buildings, they were
classified by a supervised method and object-oriented software using eCognition.

4.2.1 Legend
The legend distinguishes four categories: the built-up, the non built-up (vegetation and bare
soil), water and clouds.
Some classes are difficult to discriminate using only spectral characteristics, especially so in
countries in sub-Saharan Africa. The spectral confusions are numerous, for example, the
fields are easily confused with the built-up. Production facilities and services, and the
buildings for residential use in some places have the same spectral signature as the sand and
burned areas. To overcome these problems, we have enriched the description of spectral
regions of texture parameters (see 4.2.3).

4.2.2 Selection of training and validation areas

Training and validation areas were selected based on a visual interpretation of SPOT images
supported by a consultation of Google Earth and the plan of the city of Kinshasa, and edited
by Aquaterra Kin Art in 1997, ensuring changes due to differences between dates of these
documents. 68 areas were selected in common areas of the SPOT images. To ensure an
equivalent content of classes on each date, only areas unchanged between 1995 and 2005
were selected. The sample was divided into two, 34 areas for training and 34 for validation.

4.2.3 Choice of attributes

The attributes used in the classifications were chosen on the basis of visual interpretation.
The regions are described in terms of spectral averages in each spectral band and the NDVI
and the textural point of view, by the standard deviations on the green and red bands, and
two textural parameters of Haralick (1973), such as homogeneity and entropy of the
panchromatic band.

4.2.4 Segmentation and classification

eCognition was used to perform segmentation and classification. This software can
simultaneously use a variety of data, panchromatic and multispectral images or vector data
bases, and can create multiple levels of segmentation using a hierarchical approach.
The segmentation algorithm is the "multiresolution segmentation." According to the
"Definiens Developer 7 User Guide" (2007), this algorithm merges the pixels into segments
of image by minimizing the average heterogeneity and maximizing their respective
homogeneities. It can do the same with image segments from a previous segment. The
procedure iteratively merges the pixels or segments, as long as the maximum threshold of
heterogeneity is not exceeded. Homogeneity is defined as a combination of spectral
properties and form. The spectral homogeneity is based on the standard deviation of the
distribution of the colour and consistency of form is based on the deviation from a compact
or smooth (Cantou et al., 2006). The procedure can be influenced by the scale factor that
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 467

limits the size of the resulting segments. The segmentation was performed on the image of
2005 spectral bands of green, red and near infrared respectively, giving them a weight of 2, 1
and 1. The scale parameter was chosen by trial and ,error and set at 20 with the shape
parameter 0.1 (0.5 for compactness and 0.5 for smoothing).
The algorithm for supervised classification of the nearest neighbour was used. It ranks the
regions according to their proximity to areas of statistical training.

4.2.5 Validation
The classifications are evaluated by comparing 34 areas of validation within the matrix of
confusion. Indices are calculated to assess the quality of results (Richards, 1993):
 The overall accuracy,
 The overall Kappa,
 The Kappa class.
The overall accuracy is good (> 80%) obtained for the different classifications (Table 2). The
Kappa coefficient is only acceptable for the classification of 1995 (85%) and 2005 (92%). The
classification of 2000 has a poor Kappa (64%) caused by the fog that covers the southwest of
the city. This result will therefore not be used subsequently.

Years Overall accuracy (%) Kappa Coefficient (%)

1995 93 85
2005 96 92
Table 2. Classification accuracy

Extensive field visits conducted in late January 2009 to the end March 2009 in the extension
zones of Kinshasa to understand the factors of urbanization has revealed the existence of
different confusions and omissions in the class "building". For example, here are some for
the image of 1995 and 2005. They are located in Figure 1 and identified in Table 3.

Confusions errors in 1995

Zone 1 Canoes to the east of Industrial Limete
Zone 2 Island Mimosa
Zone 3 Fields and sand pit in the southwest in the commune of Mont Ngafula
Confusions errors in 2005
Zone 1 Sand pit area behind the camp CETA and fields of vegetable crops
Zone 2 Field burned to the east of the city
Zone 4 Island Mimosa with large rocks and a mining company in building materials
Zone 5 Sand bank to the west
Errors of omission in 2005
Zone 3 Residences in the area of the general staff of the Congolese armed forces
(abundance of vegetation)
Table 3. Confusion and omission errors for the class built-up in 1995 and 2005
468 Remote Sensing – Applications

Fig. 1. (a and b): Location of misclassification (Source: Delbart and Wolff, 2002 for the map of
municipal boundaries)
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 469

These errors will not be included in the analysis.

The results obtained at different dates are generalized by removing polygons classified as "built"
with an area less than 1km² and the inclusion of less than 1 km² within the urban sprawl.
The superposition of classes “built” on two successive dates can map the evolution of the
building when the matrix changes can be quantified.

5. Results
5.1 Urban growth
The location change was analysed using:
 Field visits conducted in-depth from the end of January 2009 through to March 2009 in
the extension zones of Kinshasa to understand the factors of urbanization,
 The layout of the lines of roads and railways digitized,
 Maps and plans of the city of Kinshasa,
 Population data,
 The digital elevation model and slope map.
The map resulting from the comparison of land use classifications in 1995 and 2005 shows that
the extensions of the city is concentrated in the southwest and northeast of Kinshasa (Figure 2).

Fig. 2. Evolution of urban sprawl between 1995 and 2005 (Sources: Images Spot KJ 3 096-358
March 31, 1995 and KJ 4 096-358 July 01, 2005)
470 Remote Sensing – Applications

The spatial extension in the southwest took place mainly along the roads of Matadi and
Lutendele (Zone 1). This process concerns the cities of Benseke, Kimbondo, Sans Fil and
Matadi Mayo on the road to Matadi, and the cities of Lutendele, Kimbala, Zamba and
Mazanza on the road to Lutendele. Cities such as Benseke and Kimbondo or Kimbala and
Zamba have even joined in 2005.
To the east, there is a filling of interstitial spaces (Zone 2) and extension (Zone 3). Indeed, in
neighbourhoods Mpasa I, II and III, Mikonga and the Badara camp, to the east of the River
Ndjili, the blanks were filled. While in the far east, across the river Ndjili (Kinkole), the built-
up was extended.
Urban growth can be explained by a population growth (5.1.1). Its spatial location can be
explained by two main geographical factors beyond the simple distance to downtown, also
an employment centre, the relief (5.1.2) and lines of communication (5.1.3).

5.1.1 Urban growth and population growth

Table 4 shows the evolution of the population of the extent of the city of Kinshasa and its
density. In 45 years, the population rose from 400,000 to 7.5 million inhabitants in 2005,
while the building area covered 6800 ha in 1960 against 43,400 in 2005. The population
density tripled between 1960 and 2005 from about 60 inhabitants / ha in 1960 to 170
inhabitants / ha in 2005, on the whole, the city has expanded and become denser.

Surface Density
Years Population
(ha) (hab/ha)
1960 400000 6800 59
1967 901520 9470 95
1969 1051000 12903 81
1973 1323039 14600 91
1975 1679091 17992 93
1981 2567166 20160 127
1984 2653558 26000 102
1995 4719862 31007 152
2000 6000000 39518 151
2005 7500000 43414 173
Table 4. Evolution of the population, the extent of Kinshasa and its density (Sources: Lelo
Nzuzi, 2008; Yebe Musieme, 2004; Delbart et al., 2002; Mbuila Matot, 2001)

When reporting on data on population and built-up areas in 1960, one can compare the
growth in urban population. Figure 3 shows that the extension of buildings characterized by
an index of 600 in 2005, while the population has an index of nearly 1900.
To compare growth rates, there is data on a semi-logarithmic graph (Figure 4).
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 471

2000

1800

1600

1400

1200

1000 Pop(1960)=100
Surf(1960)=100
800

600

400

200

0
1940 1960 1980 2000 2020

Fig. 3. Urban growth and population growth (Sources: Lelo Nzuzi, 2008; Yebe Musieme
Beni, 2004; Delbart et al., 2002; Mbuila Matot, 2001)

10000000

1000000

Population
100000
"Surfaces bâties
(ha)"
10000

1000
1940 1960 1980 2000 2020

Fig. 4. Semi-logarithmic graph of urban growth and population growth (Sources: Lelo
Nzuzi, 2008; Yebe Musieme Beni, 2004; Delbart et al., 2002; Mbuila Matot, 2001)
472 Remote Sensing – Applications

Figure 4 shows that the average growth rate of the population is less than the extension of
the city. The average growth rate of the population over the period 1960-2005 is 6.73%, while
that of the built area is 4.21%. Applying this growth rate to the built area of 2005 to calculate
the extension of the city in 2009, we do not get the 600 km2 regularly cited, but only about
510 km2.
By analysing the slopes, one can compare the growth rates, they both appear to decline in
1981 (Table 5). This result confirms the observations of Bruneau (1994).

Growth rate Population Urban

1960-1981 9.26% 5.31%
1981-2005 5.07% 3.25%
Table 5. Population growth rate

The first period covers the 20 years after independence (until 1981). It is characterized by a
very high population growth and rapid expansion space. During the second period, the
population growth rate slows sharply, although it remains high, from 9.26% between 1960
and 1981 to 7.5% between 1981 and 2005. The spatial extension grew at a slower pace and
passes from 5.31% between 1960 and 1981 to 3.25% between 1981 and 2005.

5.1.2 Urban growth and relief

According to the observation of existing maps and plans (Figure 5), the extension to Kinkole
phases that have characterized the spatial development of Kinshasa since its inception are
confirmed.
Born in the west to the Bay of Ngaliema, the city had its first developments to the east with
the birth of the ancient cities (Kinshasa, Barumbu, Lingwala, Kintambo) and Gombe
(formerly Kalina) in the late 1920s. From then the city grew to the south with the birth of
new cities (Kasavubu and Ngiri-Ngiri) between 1930 and 1940. It was during the 1950s that
the city took over the management of the east with the merger of Kalina (Leo West) and Leo.
Compared to the town of 1959, we find that the city is much more extensive in the south,
southwest and east (beyond the communes of Kimbanseke and Masina). Comparing the
growth of map altitudes and landforms, we observe that Kinshasa was first extended in the
plain corresponding to the extension of the Malebo pool and until independence in 1960
(Figure 6), the colonial authorities strictly prohibited constructions on the hills in the
absence of a particular development.
After independence, the city expanded to the southwest on the plateau to the east and the
plains. For the period 1995-2005, the growth has continued in the same directions.
To the east, it extends the plain due to the narrow width of flat land in the east (Biyeye, 1997).
Indeed, the extension behind the neighbourhoods Mpasa I, II, III and Mikonga did not take
place because of steep slopes; this is how urban sprawl has moved beyond the city Kinkole.
To the west of the river Ndjili, areas of flat plains to the south are being built upon,
urbanization covers steep slopes (Figure 7), but they are unfit for human settlement in the
absence of appropriate management. Indeed, these areas of steep slopes are subject to
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 473

Fig. 5. Urban growth in Kinshasa from 1889 to 2005 (Source: historical maps collected by
Johan Lagae, Department of Architecture and Urban Planning, Ghent University)
474 Remote Sensing – Applications

Fig. 6. Urbanization and altitude (Source: Mathieu De MaeyerIGEAT / ULB to the DTM,
unpublished)
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 475

Fig. 7. Urbanization and slope (Source: Mathieu De Maeyer, IGEAT / ULB for the slope
map, unpublished)
476 Remote Sensing – Applications

significant risk of erosion as they are laid bare (Van Caillie, 1990, 1997). In addition, the
plains downstream of these steep slopes are affected by floods because of silting. These
areas contain steep slopes occupied by the poor. In the future, urbanization will continue to
locate in areas of high slope, as is already happening in many places on the hillsides
(Camping areas, Kindele, etc.).

5.1.3 Urban growth and major roads

Figure 2 shows that urban growth is more influenced by the roads along the railway. If
before independence the railway played a role in the location of industrial areas, the urban
railway had not developed, now it has not facilitated access to downtown as a centre of
employment and therefore has not polarized urban growth. The roads in Matadi and
Bandundu, as well as routes to the drop in Lukaya and to Lake Ma-Vallée, correspond to
areas of urban growth today.
Despite the role of highways in urban growth, there are areas that develop latest far from
downtown and away from these axes. Indeed, an urban extension area has been observed
since the 1990s, south and east of the camp which houses the faculty of the University of
Kinshasa. This is the area south Cogelos and neighbourhoods Tchad, Mandela and
Department Plateau to the east. These areas develop in the absence of urban amenities.
Indeed, they are connected to almost none of the service water supply of REGIDESO. The
population is supplied fitted to the sources (Mayi ya Libanga, Mayi ya Niwa, Mayi ya
Zamba) or the fountain. This is the case in the district of Mbiti. Where the water is high,
people dig a well. The lack of urban amenities does not limit urban growth in Kinshasa. The
bottom line for people is to have a home.
The quantitative analysis of urban growth compared to major communication axes, made in
the GRASS software, shows that 47.5% of the growth took place at more than 1 km of main
roads. Beyond this 1 kilometre threshold, the advantage of proximity to the main roads can
be neglected and the neighbourhood effect becomes more important in the sense that people
settled near existing neighbourhoods, but at a greater distance from the road.
This helps to highlight the fact that the major communication axes are not always, or are no
longer, a major factor in urban growth.

6. Conclusions
Two high-resolution satellite images (SPOT) of 1995 and 2005 were used to map and
quantify the urban growth in Kinshasa relatively quickly and with an acceptable reliability.
The city spreads very quickly on its margins primarily to the east and southwest along the
road to Matadi and Bandundu allowing access by public transport to the city centre which
polarizes the bulk of urban employment. However, since the early 1990s, neighbourhoods
are growing away from the city centre and transport routes (e.g. district Cogelo, Tchad,
Mandela, Department, Plateau), yet they do not benefit from any urban convenience. The
extension of the city after 1960 did not spare areas of steep slopes unfit for human settlement
in the absence of a particular development. These areas are home to the poor.
By measuring the average growth rate of the population and the extension of the city over
45 years (from 1960 to 2005), it was found that it is 6.73% versus 4.21% for that of the built-
The Mapping of the Urban Growth of Kinshasa (DRC)
Through High Resolution Remote Sensing Between 1995 and 2005 477

up area. The average growth rate of built surface applied to the surface, built in 2005 to
calculate the area built in 2009 revealed some significant errors with the figures regularly
quoted.
In the future, a study could be carried out to understand the logic which pushes people to
occupy the steeply sloping zones where the problem of gully erosion is acute.

7. References
Alphan H. (2003). Land use change and urbanization in Adana, Turkey, Land Degradation
and Development, Vol.14, No.6, pp. 575–586, Online ISSN 1099-145X.
Biyeye Unzola E. (1997). Urbanisation de Kinshasa: la politique de la partie extrême
orientale, Mémoire en Urbanisme et Aménagement, Université Libre de Bruxelles,
85 p.
Bruneau J-C. (1995). Crise et déclin de la croissance des villes au Zaïre – Une image
Actualisée, Revue Belge de Géographie, Belgeo, 119e année, N° spécial offert au
professeur H. Nicolaï, pp. 103-114, ISBN 1377-2368.
Cantou J. P., Maillet G., Flamanc D. & Buissart H. (2006). Preparing the use of pleiades
images for mapping purposes: preliminary assessments at ign-france»,
Topographic Mapping from Space, ISPRS Workshop Commission I, WGI/5 and
WG I/6, page 6.
Coppin P., Jonckheere I., Nackaerts K., Muys B. & Lambin E. (2004). Digital change detection
methods in ecosystem monitoring: A review, International Journal of Remote Sensing,
Vol.25, No.9, pp.1565–1596, ISSN 0143-1161.
Delbart, V. & Wolff, E. (2002). Extension urbaine et densité de la population à Kinshasa:
contribution de la télédétection satellitaire, Revue Belge de Géographie, BELGEO,
Vol.2, No.1, pp. 45-59, ISSN 1377-2368.
Flouriot, J., De Maximy, R. & Pain, M. (1975). Atlas de Kinshasa, Institut Géographique
National du Zaïre.
Guptad, N. & Munshi, M.K. (1985).Urban change detection and land-use mapping of Delhi,
International Journal of Remote Sensing, Vol.6, No.3-4, pp. 529-534, ISSN 0143-1161.
Haralick R.M., (1979). Statistical and structural approaches to texture, Proceedings of the IEEE,
Vol. 67, No. 5, pp. 786-804, ISSN 0018-9219.
Lelo Nzuzi, F. (2008). Kinshasa, ville et environnement, L’Harmattan, 275p, ISBN 978-2-296-
06080-7, Paris.
Moller-Jensen, L. (1990). Knowledge based classification of an urban area using texture an
context information in LandSat-TM imagery. Photogrammetric Engineering and
Remote Sensing, Vol.56, No.6, pp. 899-904, ISSN 0099-1112.
Pain, M. (1978). Kinshasa : Ecologie et organisation urbaines. Thèse de doctorat, Université
de Toulouse le Mirail, 470 p+ Annexes.
Singh, A. (1986). Digital change detection techniques using remotely sensed data.
International Journal of Remote Sensing, Vol.10, pp. 989-1003, ISSN 0143-1161.
Tshibangu, K.W.T., Engels, P. & Malaisse, F. (1997). Evolution du couvert végétal de la
région de Kinshasa (1960-1987), Geo-Eco-Trop, Vol.21, No.1-4, pp.95-103, Liège.
Van Caillie X.D. (1990). Erodabilité des terrains sableux du Zaïre et contrôle de l’érosion,
Cahier ORSTOM, série Pédologie, vol. 25, No.1-2, pp. 197-208.
478 Remote Sensing – Applications

Van Caillie X.D. (1997). La carte des pentes (1/20 000) de la région des collines à Kinshasa.
Cahier ORSTOM, réseau érosion, bulletin 17: Erosion en montagnes semi-arides et
méditerranéennes, pp. 198-204.
Yebe Musieme B. (2004). L’impact des érosions sur l’habitat à Kisenso et les travaux de lutte
anti-érosive par la population locale. Travail de fin de cycle, Unikin, Faculté des
Sciences, Département des Sciences de la terre, Géographie, 63p, unpublished.
Yuan F., Sawaya K.E., Loeffelholz B.C. & Bauer M.E. (2005). Land cover classification and
change analysis of the Twin Cities (Minnesota) Metropolitan Area by
multitemporal Landsat remote sensing, Remote Sensing of Environnement, Vol.98,
No.2 & 3, pp. 317-328, ISSN 0034-4257.
 21

Remote Sensing for Medical

and Health Care Applications
Satoshi Suzuki and Takemi Matsui
Kansai University, Tokyo Metropolitan University
Japan

1. Introduction
Radar-based remote sensing techniques are typically employed to determine the velocities
and positions of targets such as aircraft, ships, and land vehicles. In particular, X- and K-
band microwave devices, including oscillators and antennas, have been used to measure
the velocity of automobiles and other moving objects in recent years. Microwave devices
that are compact, accurate, reliable, and inexpensive are currently commercially available.
Over the past few years, there have been increasing attempts to apply such techniques to
biomedical measurements. Although some studies have applied these devices to medicine
and health care, such research is still in its infancy. This chapter focuses on the
mechanisms of and the recent research trends in microwave remote sensing techniques
that are used to detect minute vibrations on the body surface induced by heartbeat and
respiration.

1.1 Background
The increasing proportion of elderly in the population represents an appreciable problem in
developed countries due to social concerns such as increased medical and social welfare
costs and a shortage of manpower. Such concerns are expected to worsen in the future. It is
thus necessary to focus on preventing illnesses and to promote healthy lifestyles.
Consequently, simple equipment that can be used to self-monitor medical conditions and to
acquire related data is required for homes as well as medical facilities.
Vital signs are parameters of physiological functions that are used to express the physical
condition. They are used by medical professionals for making initial diagnoses. There are
four primary vital signs: heart rate, respiratory rate, body temperature, and blood pressure.
Thermometers for home use are commercially available and are generally approved by
medical bodies. In addition, heart rate and respiratory rate can be easily confirmed by visual
and palpation methods. However, there is currently still not spread to home device capable
of accurately measuring and recording vital sign data that can be used to make detailed
diagnoses. Monitoring cardiac function can be used for diagnosing arrhythmia and mental
stress (Akselrod et al., 1981, Singh et al., 1996, Carney et al., 2001). Recently, monitoring
mental condition has attracted more attention than monitoring physiological parameters.
And also obesity and aging are thought to contribute to the risk of developing sleep apnea
480 Remote Sensing – Applications

syndrome (SAS). Airway obstruction due to fat deposition in the neck is one cause of SAS
and it is related to reduced alertness during the daytime (Morriset al., 2008). A simple device
that can monitor respiratory activity throughout the night is thus required. These examples
show the necessity for monitoring of the vital signs in daily life. Moreover, these sensing
technique are presently being studied in the research area on human–machine interfaces
that can be applied anywhere (for example, in a car or at the workplace) (Sirevaag et al.,
1993, Gould et al., 2009).
In addition, patients who have been exposed to toxic chemicals or infectious diseases are
often treated in isolation chambers to prevent secondary exposure to health-care workers. In
such cases, a doctor must often make a diagnosis without touching the patient, which is
difficult as the vital signs are of primary importance for emergency medical treatment. With
the exception of body temperature (which can be measured by infrared radiation), it is
difficult to measure vital sign parameters without contact. Consequently, remote sensing of
vital signs has attracted much attention.
In this way, several fields require remote sensing of vital signs and various remote sensing
methods have been proposed. However, such methods should perform biomedical
measurements described as non-invasive, non-restrictive, or non-contact means that can be
used without the user being conscious of them. The use of physically attaching sensors
should be minimized to reduce the burden on users.

1.2 Purpose and requirements of remote sensing in medicine and health care
Monitoring cardiac and respiratory parameters is useful for health-care management as
users go about their everyday lives. However, such daily monitoring needs to overcome
many problems. For example, users must have sufficient technical and medical competence
to set electrodes to themselves and they must not feel physically restricted by the electrodes
and leads. To overcome such problems, research is increasingly being conducted on
developing non-invasive and non-restrictive sensing techniques for acquiring vital signs
(Jacobs et al., 2004, Wang et al., 2006, Ciaccio et al., 2007). This kind of sensing technique
aims to detect and measure vibrations on the body surface induced by cardiac and
respiratory activity. In the case of respiratory activity, a person’s abdomen expands and
contracts during the breathing cycle and this movement can be detected by sensing
techniques. Similarly, for cardiac activity, the body surface moves in response to the
heartbeat in minute scales. Although the vibration is slight and its amplitude depends on
the individual and the part of the body, it has been observed from all parts of the body with
an average amplitude of about 0.1–0.2 mm by a high-resolution laser distance meter (Suzuki
et al., 2011).
Some studies have measured heart rate by placing a pressure sensor (Jacobs et al., 2004) or
polyvinylidene fluoride piezoelectric sensors (Wang et al., 2006) between the user and the
mattress on which they sleep. This kind of measurement method measures responses to
pressure changes. Other trials have used strain gauges to measure the heart rate (Ciaccio
et al., 2007). The size of minute changes due to pressure changes on the body surface
induced by the heartbeat and information relating to heartbeat and respiration were
obtained. A similar procedure was employed in studies using air mattresses (Watanabe et
al., 2005).
Remote Sensing for Medical and Health Care Applications 481

Such sensing techniques have the advantages of being inexpensive because of their simple
structure and of enabling stable relatively stable data acquisition because they employ direct
contact with the body. Some of these sensors are already commercially available. However,
they suffer from one drawback: measurement is not possible when the sensor is separated
from the body by moving their bodies. This raises the question: “Is a remote sensing method
available?”

1.3 Biomedical measurement using microwaves

Radio-frequency sensing techniques were originally developed for military applications and
they were used to determine the location and velocity of aircraft and ships. The same
technology was then applied to search and rescue; for example, they have been used to
locate survivors buried under earthquake rubble (Chen et al., 1986, 2000, Lin et al., 1992).
Radar can remotely acquire information on the motion of targets. Additionally, depending
on the frequency of the electromagnetic wave used, radar can penetrate barriers. These
characteristics of radar have been employed to detect body motion of survivors under
earthquake rubble. Such devices initially had very limited effectiveness because of their
poor resolution by using low-frequency waves to penetrate rubble; they could only detect
relatively large body motion (at best, the abdominal motion due to breathing). However, the
permeability is not a problem for everyday applications since microwaves can readily
penetrate materials such as clothing, bedding, and mattresses.
A cheap, small unit that is stable and can oscillate at high frequencies has recently been
developed and ongoing development is being conducted to produce safer, more flexible
devices. As a consequence, higher frequency electromagnetic waves were contributed to
enhance the resolution of measurement. At the same time, the output power was reduced to
reduce its effect on humans, allowing microwaves to be used in everyday applications.
Gradual progress, therfore, has made it possible to detect even human heartbeat.

2. Theory and methods

Here, we describe a system that employs microwaves to remotely measure vital signs by
detecting vibrations on the body surface induced by cardiac and respiratory activity.
Vibrations induced by heartbeat are particularly small with amplitudes of about 0.1–0.2 mm
on average. This section discusses approaches using continuous-wave (CW) Doppler radar
and ultra-wideband (UWB) pulse radar, which are generally used for measuring vital signs,
and their mechanisms.

2.1 Mechanisms of measurement

While frequency-modulated continuous wave (FMCW) radar is used to identify the exact
location of a subject in some reports, UWB or CW Doppler radar are generally used for
monitoring vital signs. (Saunders, 1990, Immoreev & Tao, 2008, Li & Lin, 2010)
In a UWB pulse radar, the transmitter sends very short electromagnetic pulses toward the
target. A pulse duration of about 200–300 ps and a pulse repetition frequency in the range of
1–10 MHz are typically used for vital sign detection. When the transmitted pulse reaches the
chest wall, some of the energy is reflected and captured by the receiver. The nominal round-
482 Remote Sensing – Applications

trip travel time of the pulse is defined as t  2d C , where d is the nominal detection distance
and C is the speed of the electromagnetic wave. If a local replica of the transmitted pulse
with a delay close to the nominal round-trip travel time correlates with the received echo,
the output correlation function will have the same frequency as the physiological
movement.
On the other hand, the CW Doppler radar mechanism is based on following (1);

T  t   cos 2 ft    t   (1)

where an unmodulated signal T(t) with a carrier frequency f and a residual phase   t  , is
transmitted toward a human body where it is phase-modulated by the physiological
movement x(t). The reflected signal R(t) detected by the radar receiver is given by following
(2);

 4 d0 4 x  t   2d  
R  t   cos 2 ft      t  0  (2)
    c 

where 4 d0  is a constant phase shift due to the nominal detection distance d0 and the
  t  2d0 c  is phase noise. Using the same transmitted signal T(t) as the local oscillator
signal, the radar receiver down-converts the received signal R(t) to the baseband signal B(t)
as following (3);

 4 d0 4 x  t  
B  t   cos    0    (3)
   

where  is determined by the nominal detection distance and the oscillator phase noise.
Since the delay corresponds to the signal round-trip travel time, the detection range of a
UWB radar can be varied by controlling the delay between the two inputs of the correlation
function block. This makes it possible to eliminate interference caused by reflection from
other objects (clutter) and multipath reflection. However, one disadvantage of UWB radar is
that the delay needs to be recalibrated when the detection distance is changed; this increases
the system complexity and cost. Furthermore, since the correlation function is nonlinear, it is
not simple to recover the original movement pattern, even though frequency information
can be easily obtained. On the other hand, CW Doppler radar has a low power consumption
and a simple radio architecture. These characteristics make it suitable for home-based
systems. Moreover, proper adjustment of the radio front-end architecture of a CW radar can
cancel clutter (Li & Lin, 2008a, 2008b). In addition, single-input multi-output and multi-
input multi-output techniques can be easily implemented with CW radar, enabling the
movements of multiple targets to be detected (Boric-Lubecke et al., 2005, Zhou et al., 2006).

2.2 Carrier frequency and output power

The carrier frequency and output power employed must be safe for use on people. Carrier
frequencies ranging from hundreds of megahertz to millimeter wave frequencies have been
tested for remote vital sign detection using a variety of physiological movements. The
carrier frequency should be carefully selected to ensure suitable sensitivities and
Remote Sensing for Medical and Health Care Applications 483

characteristic response for vital sign measurement. Some studies used extremely high-
frequency waves (228 GHz (Petkie et al., 2009)), which have shorter wavelengths and are
more sensitive to small displacements. Moreover, a 228 GHz frequency prototype has been
extended to perform heart rate and respiration measurements at a distance of 50 m.
However, such high frequency waves are not realistic for monitoring vital signs in everyday
applications. In many cases, a carrier frequency that does not require a license is often
chosen. However, carrier frequencies that do not require a license vary from country to
country and some frequency bands are allocated to amateur radio stations. For example, the
laws regulating radio frequency use in Japan allow band frequencies 10.525 and 24.15 GHz
to be used for detecting moving objects. These devices are marketed as sensors for
measuring the speeds of vehicles. Although there are limitations on how they are used (e.g.,
limited to indoor use), these frequency bands can be used by low-power radio stations
without a license provided the output is less than 10 mW. They have been increasing studies
on frequency bands for vital sign monitoring. Regarding safety, different countries have
different guidelines regarding radio-frequency electromagnetic fields.
The World Health Organization (WHO) and the Scientific Committee on Emerging and
Newly Identified Health Risks (SCENIHR) define exposure as the subjection of a person to
electric, magnetic, or electromagnetic fields or to contact currents other than those originating from
physiological processes in the body and other natural phenomena (WHO, 2003, SCENIHR, 2006).
The intended frequency band of electromagnetic field intensity differ slightly in different
guidelines. For example, the International Commission on Non-Ionizing Radiation
Protection (ICNIRP) guidelines (ICNIRP, 1998) specify the frequency band from 300 Hz to
300 GHz, SCENIHR specifies 100 Hz to 300 GHz (SCENIHR, 2006), and the IEEE Standard is
from 3 kHz to 300 GHz (IEEE Standard Committee, 1998). Each country employs different
methods for determining their criteria. Consequently, it is important consider device
development and intended usage.
The carrier frequency of medical applications of ultrasound is a low frequency of about 3 to
10 MHz. Such applications acquire information by penetrating the human body. In
comparison, monitoring using microwave frequencies in the range 10.525 to 24.15 GHz is
considered to be less invasive and safer. It is not easy to make simple comparisons, but
wireless local area networks (WLANs) use 2.4 GHz radio waves and microwaves in the
range 10.525 and 24.15 GHz are considered to be safer. Moreover, safety can be further
increased by using a lower power than a WLAN.
It seems appropriate to use high frequencies for sensing to ensure a high resolution while
considering invasiveness. However, high frequencies are not necessarily ideal for actual
applications as increased sensitivity results in increased susceptibility to artifacts; the target
motion induced by heartbeat on the body surface is much smaller than the artifacts
generated by general movement of the body and arm. Furthermore, people being monitored
move freely as they conduct everyday activities, which makes artifacts a significant
problem.

3. Examples of applications
While remote sensing is not currently used for medical and health care applications in
everyday life, several studies have been conducted. This section discusses the following
484 Remote Sensing – Applications

typical examples of remote sensing in medical and health care settings: (1) monitoring daily
health and mental stress by estimating changes in the autonomic nervous system (ANS); (2)
welfare and health care for the elderly; and (3) medical diagnosis such as screening of
patients with infections.

3.1 Measurement of ANS

To determine stress levels when driving or operating equipment, a 24 GHz compact
microwave radar was used to perform remote measurements of the heart rate variability
(HRV) under autonomic activation induced by a stressful sound and foods (Suzuki et al.,
2008, Gotoh et al., 2009). Changes in the ANS (sympathovagal balance) can be monitored by
measuring the HRV. This variability is used as an index of mental stress.
For examples, when an animal is attacked, its sympathetic nervous system spontaneously
prepares for fight or flight by elevating the heart rate and increasing the blood pressure and
body temperature. In contrast, when it is in a relaxed state, the parasympathetic nervous
system is activated, and the opposite phenomena occur. Such processes are constantly in
equilibrium in daily life (Akselrod et al., 1981, Derrick,1988).
Sympathovagal balance can be monitored by measuring variations in the heartbeat interval,
and the degree of mental stress can be determined. This index is used for diagnosing
psychological disorders and as well as the condition of circulatory organs. Moreover, many
approaches for measuring the HRV are used in psychology (Vincent et al., 1996),
occupational health (Miyake, 2001, Princi et al., 2005), and ergonomics research (Sirevaag et
al., 1993, Gould et al., 2009). An estimation technique has been medically established. Thus,
if electrocardiograms (ECG) can be replaced by a remote sensing method that allows simple
and accurate monitoring in everyday life, mental stress can be assessed at any time and in
any place.

Fig. 1. Prototype chair equipped with a non-contact microwave radar system to monitor
mental stress in workers (Suzuki et al., 2008).
Remote Sensing for Medical and Health Care Applications 485

Fig. 2. Sample data of a compact microwave radar output (A) showing a cyclic oscillation
that corresponds to cardiac oscillations measured by ECG (B) (Suzuki et al., 2008).

Audio Stimuli Contact

6000 Non-contact
1
Cross-Correlation

5000
0.5
4000
LF (msec )
2

3000 0
2000 -150 -100 -50 0 50 100 150
-0.5
1000

0
-1
0 60 120 180 240 300 360 Time lag (sec)
Time (sec)

(A) (B)
Fig. 3. Example of remote sensing in health care. (A) In both non-contact and contact (ECG)
measurements, the LF of a subject (reflecting sympathetic activation) exhibits a peak during
audio stimulation. (B) Cross-correlation of the LF between non-contact and contact
measurements of the same subject (Suzuki et al., 2008).

The prototype device in this example consists of a compact 24 GHz microwave Doppler
radar (8  5  3 cm) attached to the back of a chair. The low-frequency/high-frequency
(LF/HF) components of the HRV (which reflects the sympathovagal balance) of eight
volunteers were determined by a prototype system using the maximum entropy method
(MEM). The change in the LF and LF/HF components measured by a remote sensor was
consistent with that measured by a contact electrocardiography sensor. It is very difficult to
extract R-R intervals of heartbeats sufficiently accurately to calculate the HRV; this is
thought to be because inaccuracies were introduced due to the microwave radar being
susceptible to noise. However, measuring HRV by adopting MEM, which enables HRV to
be estimated rapidly, stably, and accurately, was demonstrated to be successful for the first
time.
486 Remote Sensing – Applications

3.2 Elderly care

The next examples were monitoring of the elderly, infants and also animals using a
microwave radar, especially focused respiratory monitoring (Li et al., 2009, Suzuki et
al.,2009).
There is a real need to reduce the physical and mental work load of care workers and also to
immediately detect sudden changes in the condition of a bedridden elderly person, both at
home and in hospitals, because nighttime activity of persons with dementia increases the
risk of injury and disrupts the sleep patterns of caregivers. Moreover, care of the elderly will
increasingly be undertaken by family members at home as the population continues to age.
Therefore, devices for long-term monitoring of the elderly that do not interfere with
activities of daily life are required.
Various approaches have been proposed in recent years. Rowe (Rowe et al., 2009) proposed
a night monitoring system that alerts caregivers when care recipients leave their beds and
that tracks them as they move about the house during the night. In addition, another study
compared the effects of a self-care and medication compliance device, which was linked to a
web-based monitoring system, with the effects of conventional care alone on compliance
with recommended self-care behavior (Artinian et al., 2003). Demiris (Demiris et al., 2009)
investigated elderly behavior using a video and image processing system while carefully
addressing privacy concerns. Another study proposed a non-restrictive, non-invasive vital
sign measurement system for measuring heartbeat and respiration to monitor health status
at home or in hospitals and nursing facilities (Tanaka et al., 2002). Systems for nursing care
should meet the following requirements: (1) monitoring of activity and vital signs must not
be a burden for the elderly; (2) accurately monitor changes in physical condition of the
patient; and (3) when the physical condition of a patient rapidly changes, the system should
promptly notify a doctor or nursing caregiver.
A prototype system using microwaves has been developed and has been applied for
monitoring elderly in a nursing home. Although still a trial, the system can effectively
perform real-time monitoring and it can acquire SAS respiratory data.
In addition, a baby also monitor using same technology has recently been demonstrated
(Li et al., 2009). The baby monitor integrates a low-power Doppler radar that can detect
minute movements induced by breathing. If no movement is detected within 20 s, an
alarm goes off.
Using same type of sensing devices, respiratory activity of a hibernating black bear has been
monitored at a Zoo (Suzuki et al., 2009). Ueno Zoological Gardens in Tokyo made plans to
assist a Japanese black bear to enter hibernation, because the bear showed extremely slow
movements in winter. Moreover, the staff wanted visitors to understand this instinctive
behavior of bears as occurs in the wild. The bear’s condition during hibernation must be
carefully observed to avoid the risk of long-term fasting. Therefore, to observe the
physiological condition of the bear during hibernation, about 3 months, a microwave radar
system was set up in the hibernation booth. As a result, the respiratory rate decreased while
the bear was entering hibernation, and became extremely low at approximately 2 bpm and
showed almost no change. Additionally, a trend similar to a circadian rhythm in the changes
of respiratory rate was observed.
Remote Sensing for Medical and Health Care Applications 487

Night-vision Microwave Doppler

camera radar antenna
in the ceiling attached TV monitor

Respiratory
Network
Signal
1.68m

router
Controller LAN

PC/
Analyzing
System
HDD1
HDD2
A/D Board
Cardiac Backup
Signal

Microwave Doppler
radar antenna
under the floor embedded

Fig. 4. Example of remote sensing using non-contact vital sign sensors to monitor a
hibernating black bear at Ueno Zoo, Tokyo (Suzuki et al., 2009).

3.3 Screening and medical examination of influenza patients

The advantage of monitoring without direct touch and without removing clothing is useful
and appealing in the medical field as it is not burdensome to patients and it can reduce the
risk of secondary infection. There is also growing interest in the fields of health, life science,
and engineering.
In another trial, a new screening system was developed to conduct rapid screening (<5 s)
of passengers who may have infectious diseases such as severe acute respiratory
syndrome (SARS) or pandemic influenza at quarantine stations. This system enables
medical inspection by measuring heart and respiratory rates, as well as body temperature
by infrared thermography (Matsui et al., 2009, 2010). A similar system is already being
tested at the quarantine station at Naha Airport for domestic flights and at Narita
International Airport in Tokyo. The results demonstrate the efficacy of the concept and
the system. The system even detected a patient with influenza whose fever was reduced
by antifebrile medication. This finding is important as it demonstrates that the parameters
488 Remote Sensing – Applications

for monitoring cardiac and respiratory activity by microwaves are effective. In the future,
remote sensing using microwaves is anticipated to attract more interest in the medical
field.

Infrared thermograhy
for measuring
200～ body temparature
300cm

10cm

Microwave radar
for monitoring cardiac activity
30～ Microwave radar
40cm for monitoring respiratory activity
100cm

Fig. 5. Example of remote sensing for medical diagnosis; non-contact monitoring for
screening systems at airport quarantine stations (Matsui et al., 2009).

4. Topics for future investigation

Remote sensing of biomedical parameters has been described by focusing on monitoring
using microwave radar and examples of medical and health care applications have been
presented. Microwave-based remote sensing offers the following advantages: movement of
an object can be detected from a distance, it means “remotely”, and microwaves can pass
through many materials (notable exceptions are metals and water). Thus, such systems will
enable health care workers to measure motion of a body surface through clothing from a
distance.
The most serious obstacle to practical applications of this technology is the need to reduce
the effects of noise and artifacts. Vibrations on the body surface caused by heartbeat are
quite small (with amplitudes of about 0.1–0.2 mm), whereas motions of arms and the
abdomen are much larger. Additionally, movement of the person being monitored while
conducting every day activities and data from other body motion will generate noise. The
effect of multipass reflection should also be considered. One study performed simultaneous
measurements of the vital signs of two people, but it is difficult to apply this technology to
real-world applications.
Remote Sensing for Medical and Health Care Applications 489

To use this technique in medical fields and daily life, aspects such as non-contact
measurement and the non–removal of clothing need to be addressed. Data obtained by
remote sensing must be more reliable and accurate than those measured by conventional
electrode-based methods. Furthermore, a new index for diagnosing disease is required.
Most studies of remote sensing using microwaves only measure the heart rate. However, the
heart rate alone has limited diagnostic value; the heart rate only provides information about
the tachycardia or bradycardia. It is hoped that such a system will be able to measure
parameters for diagnosing cardiovascular disease.
In conclusion, research on remote sensing using microwaves is still in its infancy, but it
appears promising not only for medical and health care, but also for other fields such as
ergonomics research.

5. References
Akselrod, S. et al. (1981) Power spectrum analysis of heart rate fluctuation: a quantitative
probe of beat-to-beat cardiovascular control. Science, 213(4504), pp.220-222.
Artinian, N.T., Harden, J.K., Kronenberg, M.W., Vander Wal, J.S., Daher, E., Stephens, Q.,
Bazzi, R.I. (2003) Pilot study of a Web-based compliance monitoring device for
patients with congestive heart failure. Heart Lung, 32(4), pp.226-233.
Boric-Lubecke, O., Lubecke, V., Host-Madsen, A., Samardzija, D., Cheung, K. (2005) Doppler
radar sensing of multiple subjects in single and multiple antenna systems. 7th
International Conf. on Telecom. In Modern Satellite, Cable and Broadcasting
Services, 1, pp. 7-11.
Carney, R.M., Blumenthal, J.A., Stein, P.K., Watkins, L., Catellier, D., Berkman, L.F.,
Czajkowski, S.M., O'Connor, C., Stone, P.H., Freedland, K.E. (2001) Depression,
heart rate variability, and acute myocardial infarction. Circulation, 104, pp.2024-
2028.
Chen, K.M., Misra, D., Wang, H., Chuang, H.R., Postow, E. (1986) An X-band microwave
life-detection system. IEEE Trans Biomed Eng, 33, pp.697-702.
Chen, K.M., Huang, Y., Zhang, J. (2000) Microwave Life-Detection Systems for Searching
Human Subjects Under Earthquake Rubble or Behind Barrier. IEEE Trans Biomed
Eng, 27, pp.105-113.
Ciaccio, E.J., Hiatt, M., Hegyi, T., Drzewiecki, G.M. (2007) Measurement and monitoring of
electrocardiogram belt tension in premature infants for assessment of respiratory
function. Biomed Eng Online 6, pp.1-11
Demiris, G., Oliver, D.P., Giger, J., Skubic, M., Rantz, M. (2009) Older adults' privacy
considerations for vision based recognition methods of eldercare applications.
Technol Health Care, 17(1), pp.41-48.
Derrick, W.L. (1988) Dimensions of operator workload. Human Factors, 30, pp.95-110.
Gotoh, S., Suzuki, S., Kagawa, M., Badarch, Z., Matsui, T. (2009) Non-contact determination
of parasympathetic activation induced by a full stomach using microwave radar,
Medical & Biological Engineering & Computing, 47, pp.1017–1019.
490 Remote Sensing – Applications

Gould, K.S., Roed, B.K., Saus, E.R., Koefoed, V.F., Bridger, R.S., Moen, B.E. (2009) Effects of
navigation method on workload and performance in simulated high-speed ship
navigation. Appl Ergon 40, pp.103-114.
ICNIRP Guidelines (1998) For Limiting Exposure to Time‐Varying Electric, Magnetic and
electromagnetic Fields (UP TO 300 GHZ), Health Physics, 74(4), pp.494‐522.
IEEE Standards Coordinating Committee 28 on Non-Ionizing Radiation Hazards (1998)
IEEE Standard for Safety Levels With Respect to Human Exposure to Radio
Frequency Electromagnetic Fields, 3 kHz to 300 GHz. C95.1-1991 and C95.1a-1998.
Immoreev, I., Tao, T.-H. (2008) UWB radar for patient monitoring. IEEE Aerospace and
Electronic Systems Magazine, 23(11), pp.11-18.
International EMF Project of World Health Organization (2003) Framework for Developing
EMF Standards.
Jacobs, J., Embree, P., Glei, M., Christensen, S., Sullivan, P. (2004) Characterization of a novel
heart and respiratory rate sensor. Conf Proc IEEE Eng Med Biol Soc, 3, pp.2223-
2226.
Li, C., Lin, J. (2008) Complex Signal Demodulation and Random Body Movement
Cancellation Techniques for Non-contact Vital Sign Detection. IEEE MTT-S
International Microwave Symposium Digest, pp.567-570.
Li, C., Lin, J. (2008) Random Body Movement Cancellation in Doppler Radar Vital Sign
Detection. IEEE Transactions on Microwave Theory and Techniques, 56(12),
pp.3143-3152.
Li, C., Cummings, J., Lam, J., Graves, E., Wu, W. (2009) Radar remote monitoring of vital
signs. IEEE Microwave Magazine, 10(1), pp. 47-56.
Li, C., Lin, J. (2010) Recent Advances in Doppler Radar Sensors for Pervasive Healthcare
Monitoring. Proceedings of Asia-Pacific Microwave Conference 2010, pp.283-290.
Lin, J. C. (1992) Microwave sensing of physiological movement and volume change: A
review. Bioelectromagnetics, 13, pp.557-565.
Matsui, T., Hakozaki, Y., Suzuki, S., Usui, T., Kato, T., Hasegawa, K., Sugiyama, Y.,
Sugamata, M., Abe, S. (2010) A novel screening method for influenza patients using
a newly developed non-contact screening system, Journal of Infection, 60(4),
pp.271-277.
Matsui, T., Suzuki, S., Ujikawa, K., Usui, T., Gotoh, S., Sugamata, M., Abe, S. (2009)
Development of a non-contact screening system for rapid medical inspection at a
quarantine depot using a laser Doppler blood-flow meter, microwave radar, and
infrared thermography, Journal of Medical Engineering & Technology, 33(6),
pp.481-487.
Miyake, S. (2001) Multivariate workload evaluation combining physiological and subjective
measures. Int J Psychophysiol 40, pp.233-238
Morris, L.G., Kleinberger, A., Lee, K.C., Liberatore, L.A., Burschtin, O. (2008) Rapid risk
stratification for obstructive sleep apnea, based on snoring severity and body mass
index.Otolaryngol Head Neck Surg, 139(5), pp.615-618.
Obeid, D., Issa, G., Sadek, S., Zaharia, G., El Zein, G. (2008) Low power microwave systems
for heartbeat rate detection at 2.4, 5.8, 10 and 16 GHz. First International
Remote Sensing for Medical and Health Care Applications 491

Symposium on Applied Sciences on Biomedical and Communication Technologies,

pp.1-5.
Petkie, D.T., Benton, C., Bryan, E. (2009) Millimeter wave radar for remote measurement of
vital signs. IEEE Radar Conference, pp.1-3.
Princi, T., Parco, S., Accardo, A., Radillo, O., DeSeta, F., Guaschino, S. (2005) Parametric
evaluation of heart rate variability during the menstrual cycle in young women.
Biomed Sci Instrum 41, pp.340-345.
Rowe, M.A., Kelly, A., Horne, C., Lane, S., Campbell, J., Lehman, B., Phipps, C., Keller, M.,
Benito, A.P. (2009) Reducing dangerous nighttime events in persons with dementia
by using a nighttime monitoring system. Alzheimers Dement, 5(5), pp.419-426.
Saunders, W.K. (1990) CW and FM Radar. in Radar Handbook, 2nd ed. M.I.Skolnik) San
Francisco, MaGraw-Hill, Inc. pp.14.1-14.45
Scientfic Committee on Emerging and Newly Identified Health Risk (SCENIHR) (2006)
Possible effects of Electromagnetic Fields (EMF) on Human Health.
Singh, N., Mironov, D., Armstrong, P.W., Ross, A.M., Langer, A. (1996) Heart rate variability
assessment early after acute myocardial infarction. Pathophysiological and
prognostic correlates. Circulation, 93, pp. 1388-1395.
Sirevaag, E.J., Kramer, A.F., Wickens, C.D., Reisweber, M., Strayer, D.L., Grenell, J.F. (1993)
Assessment of pilot performance and mental workload in rotary wing aircraft.
Ergonomics 36, pp.1121-1140.
Suzuki, S., Matsui, T., Imuta, H., Uenoyama, M. (2008) A novel autonomic activation
measurement method for stress monitoring: non-contact measurement of heart rate
variability using a compact microwave radar. Medical & Biological Engineering &
Computing, 46(7), pp. 709-714.
Suzuki, S., Matsui, T., Kawahara, H., Gotoh, S. (2009) Development of a Noncontact and
Long-Term Respiration Monitoring System Using Microwave Radar for
Hibernating Black Bear. Zoo Biology, 28(3), pp.259–270.
Suzuki, S., Matsui, T., Kawahara, H., Ichiki, H., Shimizu, J., Kondo, Y., Yura, H., Gotoh, S.,
Takase, B., Ishihara, M. (2009) A non-contact vital sign monitoring system for
ambulances using dual-frequency microwave radars, Medical & Biological
Engineering & Computing, 47(1), pp.101-105.
Suzuki, S., Matsui, T., Sugawara, K., Asao, T., Kotani, K. (2011) An approach to remote
monitoring of heart rate variability (HRV) using microwave radar during a
calculation task, Journal of Physiological Anthropology, 30(6), pp.241-249.
Tanaka, S., Matsumoto, Y., Wakimoto, K. (2002) Unconstrained and non-invasive
measurement of heart-beat and respiration periods using a phonocardiographic
sensor. Med Biol Eng Comput, 40(2), pp.246-252.
Vincent, A., Craik, F.I., Furedy, J.J. (1996) Relations among memory performance, mental
workload and cardiovascular responses. Int J Psychophysiol 23, pp.181-98.
Wang, F., Tanaka, M., Chonan, S. (2006) Development of a wearable mental stress
evaluation system using PVDF film sensor. J Adv Sci, 18, pp.170-173.
Watanabe, K., Watanabe, T., Watanabe, H., Ando, H., Ishikawam, T., Kobayashi, K. (2005)
Noninvasive measurement of heartbeat, respiration, snoring and body movements
492 Remote Sensing – Applications

of a subject in bed via a pneumatic method.IEEE Trans Biomed Eng, 52(12),

pp.2100-2107.
Zhou, Q., Liu, J., Host-Madsen, A., Boric-Lubecke, O., Lubecke, V. (2006) Detection of
multiple heartbeats using Doppler radar,” IEEE ICASSP 2006 Proceedings, 2,
pp.1160-1163.
 22

Demonstration of Hyperspectral Image

Exploitation for Military Applications
Jean-Pierre Ardouin1, Josée Lévesque1,
Vincent Roy1, Yves Van Chestein1 and Anthony Faust2
1Defence R&D Canada, Valcartier
2Defence R&D Canada, Suffield
Canada

1. Introduction
Airborne hyperspectral imagers have been available from various providers for many years
and their performance keeps improving. On the other hand, space-based hyperspectral
sensors have only been available from few exploratory missions such as NASA Hyperion on
EO-1 (Pearlman et al, 2003) and ESA CHRIS on Proba (Cutter et al, 2003). In recent years,
there have been many civilian space missions being planned in different countries
(Buckingham & Staenz, 2008), as well as military space demonstrations (Cooley et al, 2006).
Given the increase in potential space-based hyperspectral sensors, Defence R&D Canada
(DRDC), which is part of the Canadian department of National defence, began in 2005 a
project to demonstrate the military utility of space-based reflective hyperspectral imagery
(0.4-2.5 microns) to the Canadian Forces (CF). The project is called HYperspectral iMage
EXploitation (HYMEX) and ended its activities in 2010 (Ardouin et al, 2007).
Before the HYMEX project, DRDC had been conducting and sponsoring R&D in the area of
hyperspectral image exploitation for a number of years to explore its various possibilities
(Davenport & Ressl, 1999; Sentlinger et al, 2003; Webster et al, 2006). The focus of this work
was on military target detection applications. In parallel with these activities, the Canadian
remote sensing community has also been active in developing hyperspectral applications for
various civilian applications related to forestry, agriculture, fisheries, mineral exploration
and environmental monitoring (Buckingham et al, 2002). Many hyperspectral techniques
developed for civilian applications can be applied to military applications such as terrain
characterization.
Building on previous efforts at DRDC and with support from Canadian industry, academic
institutions and other government departments, the HYMEX project identified a set of
applications and related algorithms to be demonstrated to the Canadian Forces.
This chapter presents an overview of the project, beginning with a description of its main
activities (Section 2.0), including field trials, data analysis and algorithms evaluation and the
development of an image exploitation software. Then, for each application areas, target
detection (Section 3.0), land mapping (Section 4.0) and marine mapping (Section 5.0), we
494 Remote Sensing – Applications

discuss some of the most promising algorithms and show examples of application of these
algorithms.

2. HYMEX applications and activities

The HYMEX project addressed applications divided in 3 categories:
1. Target detection and identification: This includes targets such as military vehicles,
camouflages and various man-made materials.
2. Land mapping applications: This includes the characterization of soil and vegetation.
3. Marine mapping applications: This includes beach characterization and near-shore
bathymetry, as well as water color mapping.
As mentioned in the introduction, DRDC developed expertise and advanced exploitation
techniques for target detection over the years but had limited expertise for terrain analysis
and water mapping. To identify available techniques for the last 2 categories, HYMEX
conducted a survey of about 60 different groups from industry, academia and other
government departments forming the Canadian remote sensing community. Most
algorithms from the groups that responded have been included in the project.
The project delivered a suite of near-operational hyperspectral image exploitation tools
called HOST (Hyperspectral Operational Support Tools). HOST is developed in the IDL
language as an add-on to a commercial-off-the-shelf package called ENVI available from ITT
Visual Information Solutions. ENVI is already a widely used platform for hyperspectral
image analysis. HOST builds onto ENVI to provide tools specifically oriented for the needs
of the Canadian Forces. The HYMEX System Integrator responsible for the development of
HOST was MacDonald Dettwiler and Associates (Richmond, BC).
HYMEX adopted an algorithms validation strategy that led to the selection of the best
algorithms to be integrated into HOST and explore the performance limits of these
algorithms. Knowing these limits is an important element of information for transitioning
the tools into operational use by the Canadian Forces. The validation strategy included the
following components:
Field Trials: The HYMEX project conducted several field trials at different locations across
Canada to acquire remotely sensed hyperspectral imagery and ground truth for testing the
various algorithms. Each location had specific characteristics likely to affect the performance
of the algorithms. Marine trials were conducted in two different locations (East and West
coast). Land trials were conducted in an Acadian forest (CFB Gagetown, NB), a boreal forest
(CFB Valcartier, QC), an aspen dominant forest (CFB Wainwright, AB) and in prairie
grassland (CFB Suffield, AB). One trial was conducted in winter under snow conditions.
While some space-based hyperspectral imagery was acquired, the project main data type
came from airborne hyperspectral sensors. Airborne data contains similar information as
space-based sensors, and can be used in the interim to evaluate algorithms. In fact, even
though HYMEX focused on demonstrating the utility of space-based imagery, the results
can easily be transposed into the utility of airborne sensors.
Data analysis: All HYMEX algorithm providers analyzed trial data using performance
metrics approved by DRDC. In addition to providing quantitative performance evaluation
and comparisons of the algorithms, the overall data analysis allowed the demonstration in a
Demonstration of Hyperspectral Image Exploitation for Military Applications 495

limited way of a) how the performance varies in different environments, b) the advantage of
hyperspectral (HSI) over multispectral (MSI) imagery and c) the effect of varying ground
sampling distance (GSD) via the analysis of airborne data at different altitudes or of
different sensors.

Fig. 1. HOST two main windows: the visualization window on the left and the control
window on the right

HOST demonstrations: We developed the HOST software in three successive iterations. Each
iteration ended with a live or hands-on demonstration of the tools to CF stakeholders and
image analysts. Feedback from the demonstration participants were integrated in the
following iterations and the selection of the algorithms to be integrated for the iteration was
based on interim data analysis results. As explained above, HOST is an add-on to ENVI.
While ENVI is a powerful exploitation package for advanced hyperspectral imagery users,
HOST is oriented towards military end-users with introductory knowledge of hyperspectral
image exploitation. In order to present a simplified and more uniform user interface than
ENVI, HOST is organized in two main windows as illustrated in Figure 1: the visualization
window and the control window. The HOST visualization window regroups, in a single
window, many of the familiar visualization tools offered by ENVI (image display, plot
display, available bands list, region of interest tool and the vector layer manipulation tool).
By regrouping these tools in a single window, the user can more easily keep track of these
functions as they are applied to specific hyperspectral images. The control window
provides, in a single window, an interface to the parameters of the different advanced
exploitation algorithms offered by HOST. The user would typically use the control window
to setup batch scripts to process many hyperspectral images without user intervention. The
HOST control window is organized in different logical categories of algorithms such as: pre-
496 Remote Sensing – Applications

processing, atmospheric correction, exploitation and interactive tools. A Navigation Tool

also allows loading customized task descriptions with links to the HOST user interface. This
guides the user through the algorithms needed to accomplish a task and the selection of the
parameters for those algorithms.

3. Target detection algorithms

Throughout the HYMEX project, DRDC gained experience in applying algorithms for target
detection applications. In this section, we describe a typical processing chain (atmospheric
correction, detection and target abundance estimation (Roy, 2010)) used in the project and
present results from an experiment aimed at evaluating the performance of the target
abundance estimation part of the processing chain using data collected in difficult
illumination and atmospheric conditions.
In late October and early November 2009, DRDC collected airborne hyperspectral imagery
near Suffield, Alberta (50°13’N, 110°10’W) using an Itres SASI-600 SWIR pushbroom
imaging system. The sensor was flown at various altitudes ranging from 330m to 1700m
above ground level in order to acquire imagery at across-track ground sampling distance
(GSD) of 0.4m, 1.0m, and 2.0m, while along-track GSD remained constant at 1.0m. Imagery
was collected between 13h00 and 15h00 local time, which resulted in sun elevation between
17 and 25 degrees. Furthermore, thin altostratus clouds and an overcast of altocumulus
clouds on the first (29 Oct) and second day (03 Nov) of collect respectively degraded the
illumination conditions considerably, as illustrated in Figure 2. Compared to typical
reflective hyperspectral field trials usually conducted under clear skies and high solar
elevation, this data collection was conducted under significantly adverse environmental
conditions not often considered in the hyperspectral literature.

Fig. 2. Typical sky conditions on 29 Oct 2009 (left) and 03 Nov 2009 (right).

One objective of this field trial was the evaluation of the constrained energy minimisation
(CEM) algorithm (Settle, 2002) sub-pixel abundance estimation accuracy. For this purpose,
we designed targets of known abundances made of thin strips of painted metal, as
illustrated in Figure 3 below. The design allowed changes to the abundance level by varying
the distance between the strips of metals while their overall size (5m x 5m) ensured that they
filled completely at least one pixel in the imagery, as showed in Figure 4. We used two
different types of paint to vary the contrast between the target and background, one beige
(see Figure 3) and one green (not shown). The base color was mixed with small quantities (2
to 10% per volume) of black feature-less paint to control the spectral features depth and
overall signature albedo. A total of 6 targets were used in this experiment.
Demonstration of Hyperspectral Image Exploitation for Military Applications 497

Fig. 3. Left) Example of controlled abundance target. Right) Spectral signatures of the beige
paint at different albedo levels, as measured in field conditions using an ASD FieldSpec Pro
spectrometer.

Fig. 4. Left) SWIR 3-colors composite of targets, imaged at a 2.4m GSD (coarsest resolution).
Right) Background (red) and target (blue) signatures as measured by the airborne sensor,
compared to the target library signature (black).

DRDC favours automated and adaptive approaches to hyperspectral target detection,

minimizing user interaction and processing time as much as possible. In this context, we
used the following processing chain for this dataset exploitation:
1. We first manually identified and removed 12 out of the original 100 spectral bands due
to their low signal to noise level. We also removed unusable part of the imagery due to
sensor vignetting. This is typically done only once for a given sensor.
2. We then converted the imagery from at-aperture radiance to apparent reflectance units
using the empirical line method (Smith & Milton, 1999), with five very large targets
(greater than 10x10 pixels) of known reflectance. This was possible due to the controlled
environment of this imagery collection; else we would typically have used a semi-
empirical technique such as the QUick Atmospheric Correction (QUAC) method
(Bernstein et al, 2005).
3. The CEM algorithm requires a description of the background 2nd order statistics (mean
and covariance). For this calculation, we identified a subset of background pixels from
the complete image using the RX anomaly detection algorithm (Reed & Yu, 1990) by
keeping only the first 90% lowest scoring pixels. Second order statistics were then
evaluated using this limited set of pixels.
498 Remote Sensing – Applications

4. Finally, we calculated the score for all pixels of the image using the CEM algorithm. As
described in Settle (2002), when properly normalised, the CEM output is an estimation
of the searched target signature abundance in the pixel under test.
The controlled ground targets were collected in 45 different images. On 29 Oct, the abundance
was set to 50% while on 03 Nov it was reduced to 33%. All images were manually interpreted
to delineate the area of the targets in the images. The CEM scores were averaged over each
target area to derive an “average abundance”, as shown in Figure 5. This was necessary
because as imaged, the targets had inhomogeneous abundance over their physical extent,
particularly at the finest GSDs. This suggests that the target design could be improved for
future experiments by using thinner strips of material more closely spaced together.

Along-track
direction

Fig. 5. Example of manual delineation of target area. Target is 5m x 5m, GSD is 0.4m (across-
track) by 1.0m (along track).

Estimated abundance error was calculated using the root mean square error (RMSE) and the
estimation bias, both normalized by the true abundance in order to get a relative error in
percent:

  evaluated   true 
2
100%
Relative RMSE [%]  * (1)
 true N

100% 1
Relative bias 
 true
*
N
  evaluated   true  (2)

RMSE results are presented in Table 1. Overall, the root mean square error for this experiment
is between 11.7% and 30%. In absolute terms, the overall RMSE translates to 0.064 and 0.078
for the 0.5 and 0.33 abundance targets respectively. The retrieved abundances were slightly
underestimated, with bias of -1.9% and -14.7% again on the 0.5 and 0.33 abundance targets
respectively. Since atmospheric conditions degraded between the two collects, it is unclear if
the observed increase in error is related to the change in illumination conditions, to the lower
abundance level considered, or to a combination of both.
The results achieved are encouraging and show that target abundance can be retrieved at the
subpixel level using the CEM algorithm with a high accuracy. The fact that the estimated
abundances are generally lower than the true abundances which is consistent with an error
that could have been introduced during the manual delineation of targets area, by assigning
larger areas to targets than their true area. Also, the imaging system true point spread function
has not been characterized and taken into account in this analysis; non-uniform sampling over
the GSD could lead to an underestimation of the sub-pixel abundances (Settle, 2004).
Demonstration of Hyperspectral Image Exploitation for Military Applications 499

Target type Target configuration

Contrast type Albedo type  = 0.5 (29 Oct 2009)  = 0.33 (03 Nov 2009)
High (beige) High 11.7 30.0
High (beige) Medium 13.4 24.3
High (beige) Low 12.8 19.6
Low (green) High 11.9 20.8
Low (green) Medium 13.2 24.5
Low (green) Low 13.2 21.7
Average over all targets: 12.7 23.5

Table 1. Relative RMS errors of the retrieved abundances using the CEM algorithm; 
denotes the target abundance.

The results demonstrate the robustness of the processing chain; with minimal user
interaction and using a simple processing chain suitable for near real-time exploitation,
targets can be characterized at the sub-pixel level even under adverse illumination
conditions. This demonstrates the processing chain’s military utility, and indicates that it
could be adapted to the detection and characterization of spectral signatures of interest in a
military operational context.

4. Land mapping applications

Land mapping applications were studied in collaboration with the University of New-
Brunswick, the University of Alberta, the University of Lethbridge, York University and
Laval University. The work was oriented towards soil and vegetation characterization and
mapping for trafficability and environmental applications. In Section 4.1, algorithms for
classification and the extraction of vegetation canopy attributes (density, structure) were
evaluated using airborne hyperspectral data acquired over three Canadian Forces bases
(CFB). The resulting validated hyperspectral products were then used to improve a
trafficability model developed by the University of New Brunswick for Gagetown military
base as well as promote environmentally sustainable training on military bases. Winter
airborne images were also acquired over the Montmorency experimental forest (near
Quebec City) to investigate the potential of winter imagery to better derive forest
information. In Section 4.2 we show that among the classification algorithms that were
evaluated, the Mercury algorithm (an evidential-reasoning-based supervised classification
algorithm developed by the University of Lethbridge (Peddle & Ferguson, 2002)) achieved
the best performance. Finally, Section 4.3 shows results from a laboratory study conducted
by the University of Alberta demonstrating how hyperspectral techniques can be used to
discriminate between vegetation stresses caused by exposure to different toxic industrial
chemicals (Rogge et al, 2008).

4.1 Trafficability and the monitoring of military training areas

This section presents results obtained for the two main land applications of HYMEX,
trafficability and the monitoring of training ranges to promote environmentally sustainable
500 Remote Sensing – Applications

training. Table 2 provides a summary of the trials and the objectives sought for each
application. The primary objective of each trial was the validation of algorithms used to
derive vegetation cover information such as type, density and height, the presence of
wetlands and the determination of soil type. These surface features are easily derived from
hyperspectral imagery and can contribute to improve knowledge of the terrain for the
purposes of trafficability and environmental applications. Each trial was conducted in a
different vegetation background ranging from various forest biomes (deciduous, mixte,
boreal) to prairie grassland. Details regarding each trial, the available ground truth and the
algorithms used to analyse the various datasets can be found in (Ardouin et al, 2007).

Trial location Date Sensor GSD Background Objectives Application

CFB Gagetown, Sep 2005 Probe-1 15m Acadian forest Forest parameter trafficability &
NB algorithms validation sustainable
(mixed deciduous) & training
CFB Gagetown
trafficability model
CFB Sep 2006 AISA 4m Boreal/grassland Forest parameter trafficability
Wainwright, AB algorithms validation
(single deciduous
species)
Montmorency Jun 2004 4m Boreal/summer Forest parameter trafficability
Experimental Feb 2007 AISA 4m Boreal/winter algorithms validation
Forest, QC (mixed conifers),
wetland mapping &
summer/winter
dataset investigation
CFB Suffield, AB Sep 2006 AISA 4m Prairie grassland Map invasive species, sustainable
burnt areas, soil training
disturbance

Table 2. HYMEX land mapping application trials

CFB Gagetown trial. One of the objectives of this trial was to improve the trafficability
model used by the Army Meteorological Center (AMC) at CFB Gagetown to plan training
exercises, avoid erosion by vehicles and promote environmentally sustainable training. The
model currently use as input, the soil moisture content simulated by the University of New
Brunswick (UNB) Forest Hydrology Model (ForHyM2) which is based on air and ground
temperatures, soil type, the amount of precipitation and the wind speed and direction.
Improvement of the trafficability model was achieved by the addition of above ground
restrictions such as the forest type (hard/softwood), density and height which can be
readily derived from hyperspectral remote sensing. Figure 6b shows a vegetation species
classification derived from 15m GSD imagery collected by the Probe-1 sensor (Figure 6a).
The overall accuracy (81.8%) and Kappa coefficient (0.78) are based on 533 pixels. These
results were obtained with the University of Lethbridge Mercury classification algorithm.
Figure 7 shows the shadow fraction of the forest canopy which was derived from spectral
mixture analysis (SMA) (Peddle & Smith, 2005) along with the sunlit deciduous fraction, the
sunlit conifer fraction and the background fraction. The image shadow fraction was found to
correlate the best with LAI, as measured on the ground on 29 plots with hemispherical
Demonstration of Hyperspectral Image Exploitation for Military Applications 501

pictures (R2 = 0.55) with an average difference between the SMA LAI and the ground LAI of
less than 0.5 LAI. Figures 8a and 8b show two forest canopy structures, stem density and
stand height, as output from the University of Lethbridge Multiple Forward Mode 3-D
Canopy Reflectance Model (MFM-3D) applied to the modified geometric optical mutual
shadowing model (GOMS) (Peddle et al, 2003). MFM-3D uses a Look-up-tables (LUT)
approach based on various ranges and increments of forest structure parameters (density,
horizontal & vertical crown radius, crown height and height distribution) as input. The
ranges and increments can be determined either from field data or automatically without
prior knowledge. Inversion of MFM-3D model produces results when image reflectance
values match the modelled reflectance. Field and MFM-3D stand height produced less than
2 m average height difference with the under-estimation of the MFM-3D model attributed to
the difficulty in locating neighbouring pixels with similarity to the center pixel.
The addition of above ground restrictions to UNB trafficability model, as determined by the
vegetation layers described above, helps produce more refined trafficability classes as
illustrated in Figure 9. Figure 10a shows the graphical user interface (GUI) of the route
planning tool. Once all the available layers are loaded into the Input Dialog, the user can
select from the Interactive Parameters which restrictions to apply for a particular vehicle
type. Examples of route planning for four types of military vehicles are shown in Figure 10b
with a low environmental concern (not avoiding areas with a high rutting index), and 10c
with a high environmental concern (avoiding areas prone to produce ruts).

(a)

0 1km

R:672nm, G:549nm, B:488nm

Unclassified
Conifer dominant
Deciduous dominant
(b) Mixed forest
Grassland/shrubland
Cutblock

Exposed soil/road
Water

Fig. 6. (a) Sub-image of Probe-1 airborne imagery of CFB Gagetown and (b) landcover
classification derived from an evidential reasoning classifier with classes of vegetation and
exposed soil and roads. Overall accuracy: 81.8%. Kappa coefficient: 0.78 (D. Peddle, U. of
Lethbridge).
502 Remote Sensing – Applications

Fig. 7. Sub-image of Probe-1 airborne imagery of CFB Gagetown showing the shadow
fraction of the forest canopy derived from spectral mixture analysis (D. Peddle, U. of
Lethbridge).

5000

a)

0
Stem/ha

6.5

(b)

0m

Fig. 8. Sub-image of Probe-1 airborne imagery of CFB Gagetown showing forest structural
parameters (a) stand density and (b) stand height as derived from the MFM-3D model (D.
Peddle, U. of Lethbridge).
Demonstration of Hyperspectral Image Exploitation for Military Applications 503

Current model Improved model

Avoid Poor Highly No Go

(too wet) restricted zone

Poor
Moderate Highly

Soil moisture
Moderate

Soil moisture
good restricted
Caution

Moderate Poor
Good Excellent
risky (too dense)
mobility

Vegetation density

Fig. 9. Off-road trafficability classes for the current (left) and improved models (right).

(a) (b)

0 1 km

(c)

0 1 km

Fig. 10. (a) Route planning for four military vehicles with (b) low and (c) high environmental
concern.

CFB Wainwright trial. A similar trafficability model was also constructed by UNB for CFB
Wainwright with the objective to demonstrate its portability to different soil and vegetation
ecosystems. The above ground trafficability is determined by a semi-arid prairie landscape
with well defined dry and wet seasons. The vegetation is composed of grassland with areas
of deciduous trees (aspen, balsam poplar and willow) and the topography is gentle. The
hyperspectral vegetation products made available to UNB are the land cover classes and the
leaf area index (LAI) (Figures 11b and 11c) provided by York University under contract to
HYMEX. Figures 11d to 11f show the optimal route planning between point A and point B
using a Wolf and a LAV vehicles when constrained by wet areas and vegetation during the
wet season (Figure 11d), when constrained by wet areas and vegetation during the dry
season (Figure 11e) and when constrained by wet areas and vegetation during the wet
season and the requirement to move along tree lines as closely as possible (Figure 11f).
504 Remote Sensing – Applications

Grass
Shrubs
R: 800 nm
Low-density trees
G: 68 0nm High-density trees
B: 550 nm 0 1km Road
0 5

(a) (b) (c)

(d) (e) (f)

Fig. 11. (a) AISA color composite, (b) land cover classes and (c) leaf area index map of CFB
Wainwright (York University). Route between A and B for a Wolf and a LAV vehicles: (d)
when constrained by wet season conditions, (e) when constrained by dry season conditions
and (f) when constrained by wet season conditions and the requirement to move along tree
lines (UNB).

Montmorency Experiment Forest (MEF) trials. Two airborne hyperspectral datasets are
available for the MEF site, one from Jun 2004 (summer) and one from Feb 2007 (winter). The
objective for imaging this site was to test algorithms for vegetation mapping in a coniferous
dominant forest ecosystem. The winter 2007 dataset was acquired for the purpose of
investigating the usefulness of summer/winter data to extract relevant terrain information
for trafficability in the boreal forest. One first attempt to address this later objective is to map
tamarack trees (Larix laricina) which can be used as an indicator species for the location of
peatlands areas dominated by trees, a wetland type being of interest for trafficability.
Demonstration of Hyperspectral Image Exploitation for Military Applications 505

Tamarack trees thrive in open areas because of their intolerance to shade and their
resentment to compete with other species (Beeftink, 1951). They also adapt very well to
poorly drained soil. Their presence is generally associated with peatlands although their
absence do not indicate that there is no wetlands. Spectrally, tamaracks are similar to other
conifers in summer and to deciduous trees in winter because they loose their leaves (needles)
before winter. A combination of summer and winter data allows the exploitation of this
unique characteristic of the tamarack trees to locate and map treed wetlands. Figure 12b
shows a RGB of AISA data acquired in winter 2007 at the MEF. Figure 12c is a moisture stress
vegetation index in which the red color represents exposed bark. When looking closer at the
single tree where the red arrow is pointing in Figure 12a, it is easy to recognise from the
shape of its shadow on the snow that it is a defoliated coniferous tree. Not having this
information on hand it would be difficult to determine whether the trees in the red color class
in Figure 12c are deciduous, dead spruce or fir, or dormant tamarack.
Under contract to DRDC, Laval University applied four filters, each made of a band ratio
index and a predefined threshold, to classify the tamarack trees in the AISA image with 95%
of the tamarack pixels correctly classified and only 1.2% of the remaining pixels
misclassified as hardwood trees. Each filter discriminates tamarack trees from other forest
features such as other softwood and deciduous. The effects of the application of the first 3
and the first 4 filters are shown in Figure 13a and 13b.

(a) (b) (c)

Fig. 12. (a) high resolution color image, (b) AISA color composite (R:1290 nm , G:1655 nm,
B:2189 nm), and (c) a moisture stress vegetation index from the AISA winter image of an
area at the Montmorency experimental forest. The red arrow points at a tamarack tree.

CFB Suffield trial. The objectives of this trial were primarily environmentally oriented.
Despite a semi-arid climate, prairie grasslands are very sensitive to the introduction of
invasive species which are often dispersed during military training and along the
maintenance roads of pipelines and gas wells. Leafy spurge and crested wheat are the main
invasive species and can easily spread in windward direction into preserved native prairie
areas. Moreover, there is a need for monitoring training areas for an environmentally
sustainable training. This is to ensure that excessive training does not over stress the soil and
506 Remote Sensing – Applications

(a) (b)
Fig. 13. (a) result from applying the first 3 filters in which several hardwood pixels are
misclassified as tamarack, and (b) result from applying the first 4 filters (red dots are
correctly classified tamarack pixels). (Prof. Sylvie Daniel and Gaël Briant, Laval University).

therefore the vegetation capacity to recover. The imaged areas include a wide range of soil
and vegetation species including invasive species, burned areas, cultivation and grazing
areas, wetlands, and various levels of disturbances by vehicle pathways. At the time of the
airborne hyperspectral survey (Sep 2006), the prairie landscape was dry and with the
exception of the low lands and around wetlands, the vegetation exhibited a low
photosynthetic activity which resulted in less pigment absorption in the visible and more
apparent absorption features in the short wave infrared by other plant cell constituents such
as lignin and cellulose. The following results obtained by the University of Alberta (under
contract for HYMEX) demonstrate the potential of this dataset for mapping soil and
vegetation at CFB Suffield to help the environmentally sound planning of military exercises.
Soil was determined to have a high clay content. Thus, exposed soil was mapped using the
spatial distribution of the depth of the clay absorption feature in the vicinity of 2200 nm after
removing the vegetation effect using an orthogonal subspace projection and known green
and dry vegetation endmember spectra (Figure 14b). Band depth was measured using the
continuum removal between 2210nm and 2230 nm. The band depth was classified into four
classes (Figure 14c) defined as (1) low clay absorption depth (green) corresponding to natural
undisturbed terrain, (2) slightly (yellow) and strongly (blue) disturbed soils areas and (4)
high clay absorption (red) which correspond to bare soils, active roads, non-vegetated dry
wetlands and burnt areas. In undisturbed grassland areas the soil is covered with dry grass,
old grass residue and a layer of moss. When the surface is disturbed, some of the soil
becomes exposed and the amount of moss and old residues decreases. Thus, a good indicator
of vegetation recovery following exercises would be a dominance of dry grass. Figure 14d
shows an RGB of the clay band depth (red), the most dominant grass endmember (green) and
an endmember associated with spectra of moss covered soil measured with a field
spectrometer (blue). These three classes of endmembers can easily be associated with the
following three conditions: (1) permanently disturbed areas such as roads and areas
surrounding gas wells (red), (2) recently disturbed areas where the moss and old residues are
removed (green) and (3) undisturbed areas covered by moss, old residues and grass (blue).
Invasive species could not be spectrally identified due to the overall dryness of the
vegetation cover. The RGB composite of Figure 15b displays the most abundant green
Demonstration of Hyperspectral Image Exploitation for Military Applications 507

RGB: R: 8 60 n m, G: 6 50 n m, B: 55 0 nm

(a) (b) (c) (d)

Fig. 14. (a) AISA imagery RGB color composite, (b) soil clay band depth, (c) classification of
soil clay band depth and (d) RGB of the clay band depth (red), the most dominant grass
endmember (green) and moss covered soil spectra (blue). (B. Rivard, U. of Alberta)

vegetation endmember in red and the two dominant dry vegetation endmembers in green
and blue. The green vegetation (Red color) is located in low land areas which are often
located in the vicinity of wetlands. Local cattle grazing is allowed in some area of the
military base. The dry vegetation shown in blue represents overgrazed areas which can be
compared to impacted areas from training exercise in other area of the military base, thus
showing the potential for environmental monitoring for sustainable training. The black area
in the northern part represents a recently burnt area where vegetation hasn’t started to grow
back.

4.2 Classification algorithms comparison

As indicated in the previous section, the Mercury supervised classification algorithm
(Peddle & Ferguson, 2002) performed well during HYMEX field trials and as a result was
integrated into the HOST software. In order to evaluate its performance in more details, we
compared Mercury to all the supervised classification algorithms offered by the ITT’s ENVI
4.8 software (Van Chestein, 2011). Two data sets were used, a 15 m spatial resolution Probe-
1 hyperspectral image of CFB Gagetown and a 4 m resolution AISA hyperspectral imagery
of CFB Wainwright. The classes defined for each dataset are listed in Table 3.
508 Remote Sensing – Applications

(a) (b)
Fig. 15. (a) RGB “true color” (red: 640nm, green: 550nm, blue: 460nm) AISA image. (b) RGB
composite image of the most widely spread green vegetation endmember (Red) and two
dominant dry vegetation endmembers (green, blue). (B. Rivard, U. of Alberta).

(a) (b)
Table 3. Number of training and test pixels for each class: (a) classes for CFB Gagetown data
(Probe-1, 15m GSD) and (b) classes for CFB Wainwright data, (AISA, 4m GSD).

During the comparison, it was found that the tested algorithms behaved differently as the
number of bands used in the classification process increases. Some see classification
accuracy increase, others prove unaffected by the number of bands while a third group see
the accuracy decrease, albeit slightly.
An immediate advantage in using fewer bands is that processing times are shorter, which is
very convenient when analyzing large files. It can also be useful to identify which
Demonstration of Hyperspectral Image Exploitation for Military Applications 509

algorithms are most consistent in accuracy as the number of bands is changed. This way, by
using a classification algorithm with known consistency, the optimal band-set can be
selected quickly after performing a few tests.
The major finding was that the Mercury algorithm consistently provides very high overall
classification accuracy values as illustrated in Table 4. It proves stable and offers the advantage
of not requiring that the number of training pixels for each class be at least equal to the number
of bands used plus one as is the case with the Maximum Likelihood and Mahalanobis Distance
techniques. Mercury’s accuracy increased with the number of bands and it offered the highest
individual accuracy values in both datasets. Using Mercury on the principal components
yielded lower accuracy than with the original dataset. With the Maximum Likelihood
algorithm applied to the principal components, results were almost identical to those obtained
with the original data. The following table illustrates the findings.

Number of bands → 4 7 20 50 Average

Mercury 82.30% 83.60% 88.20% 91.30% 86.4%
Support Vector Machine 84.40% 83.10% 84.40% 85.60% 84.4%
Mahalanobis Distance 77.60% 75.20% 86.20% 89.50% 82.1%
Neural Network 81.60% 84.20% 88.30% 65.20% 79.8%
Maximum Likelihood 87.60% 73.00% 77.40% 75.20% 78.3%
Minimum Distance 80.30% 76.70% 76.60% 76.50% 77.5%
Parallelepiped 75.60% 74.00% 77.40% 77.00% 76.0%
Spectral Angle 67.20% 67.60% 68.80% 69.00% 68.2%
Spectral Information
68.70% 66.60% 68.40% 67.60% 67.8%
Divergence
Binary Encoding 52.00% 51.40% 56.60% 66.70% 56.7%
Table 4. Comparison of Mercury and ENVI supervised classification algorithms accuracy.
Green is for the algorithm that ranked 1st in classification accuracy and yellow is the
algorithm that ranked second.

The study also highlighted the fact that class accuracy varies greatly with the choice of
bands in most algorithms. Figures ranging from 0% to 100% accuracy were observed in
some algorithms but Mercury came out with very consistent global figures.
In summary, Mercury compares very favourably with ITT’s offering for global and class
accuracy and for all algorithms, one would be well advised to run a few tests as to the
number and choice of bands to ensure optimal feature accuracy.

4.3 Chemical effects on vegetation

4.3.1 Plants as chemical detectors
The Canadian Centre for Mine Action Technologies initiated a study in 2003 to investigate
the possibility of exploiting advances in genetic engineering and plant biotechnology to
design a process by which plants, local to a region of interest, could be genetically modified
510 Remote Sensing – Applications

(GM) to be sensitive to the compounds known to permeate the soil around emplaced
landmines. In this case it was envisioned that the plant's genes would also be designed to
include a reporting mechanism, signalling the presence of these compounds through a
change in the plant's structure, appearance or some other physical characteristic. The
Deyholos group at the University of Alberta was funded to conduct the initial study
(Deyholos et al, 2006).
At the same time, the United States' Defense Advanced Research Projects Agency (DARPA)
inititated the Biological Input Output Systems (BIOS) program. The BIOS program's
objective was to produce basic biochemical modules for future use in plant or microbial-
based detectors of chemical and biological compounds of strategic interest. Collaboration
between the two projects advanced efforts in developing a human-readable biological
signalling event (Deyholos et al, 2007; Antunes et al, 2006)
The DARPA-funded team at Colorado State University went on to develop the first
generation plant-based sensor capable of detecting 2,4,6-TNT in the low ppt (parts per
trillion) range. The Canadian effort made significant progress in the development of a root-
to-shoot transducer system and an effective visual reporter system (Deyholos, 2009).
This effort clearly demonstrated that plants’ natural responses to chemicals in their
environment could be harnessed, exploited and enhanced to provide an in situ chemical
detection capability of remarkable sensitivity. This observation, amongst others, led to a
study to investigate whether it might be possible to detect, through optical means, the
naturally occurring effects of exposure to various chemical agents on vegetation, by which
in situ vegetation may provide a highly sensitive stand-off detection capability to chemical
exposures occurring at ground level. These agents cause stress and damage to surrounding
vegetation the extent of which is dependent on dosage and time of exposure.

4.3.2 Passive detection

It is well recognized that reflective hyperspectral imagery (400-2500nm) is well suited to
analyze vegetation. Under the Canadian Space Agency HERO program, a feasibility study
was conducted (Peddle et al, 2008) to determine whether a space-based system such as
HERO can be used to detect toxic industrial chemicals indirectly by detecting the stress that
these chemical cause on vegetation. Recognizing that this could have a potential military
application, we pursued this project under HYMEX by conducting a laboratory evaluation
of the stresses caused by various chemicals.
The aim of this investigation was to provide information that would help quantify the
potential of reflective hyperspectral imagery for chemical and biological surveillance,
reconnaissance involving plants exposed to Toxic Industrial Chemicals and Materials
(known as TICs and TIMs, such as Ammonia, Sulphur Dioxide, Chlorine, Hydrogen
Sulphide, Hydrogen Cyanide, Cyanides, Phosgene).
The two objectives of this study were to determine if: 1) vegetation subjected to TICs could
be distinguished from background vegetation during varying growth stages (new growth to
senescence) and environmental stresses; and, 2) different TICs could be distinguished based
on the vegetation spectral response. This work was conducted by teams at the University of
Alberta (Rivard et al, 2008).
Demonstration of Hyperspectral Image Exploitation for Military Applications 511

This research team examined the spectral response of individual leaves of three common
Canadian plant species (poplar (Populus deltoides, Populus trichocarpa), wheat (Triticum
aestivum), canola (Brassica napus)), which were subjected to fumigation with gaseous phase
toxic industrial chemicals and chemicals precursor to chemical warfare agents (e.g.
ammonia and sulphur dioxide) (TICs). Treatments were designed to allow quantification of
the variation in spectra that might be expected due to environmental, developmental, and
stochastic effects on the physiological state of individual plants within each species.
The test plants were grown in controlled environment chambers at the University of
Alberta, using standardized conditions. Each spectral measurement collected with the
ASD® FR spectrometer, as shown in Figure 16, consisted of an average of 10 scans. Multiple
scans were taken per leaf location to reduce the effects of noise. For each leaf, three different
locations were measured located approximately halfway between the main leaf vein and the
leaf edge, precluding overlap of areas measured. The measurements from each leaf were
then averaged accounting for spectral variability across the leaf. For smaller leaves (e.g. new
growth) only one or two measurements were possible.

Fig. 16. Basic set-up for spectral measurements. Inset is an image of the ASD® Leaf Clip, the
field spectrometer used to collect plant data.

The study was broken into two phases: 1) to capture the spectral variability of the various
leaf growth stages (new to senescing leaves) observed in each of the three plant types; and 2)
subjecting the plants to environmental stresses (e.g. drought) and the following five
industrially relevant gaseous phase TICs: ammonia (NH3), sulphur dioxide (SO2), hydrogen
sulphide (H2S), chlorine (Cl2), and hydrogen cyanide (HCN). The experimental data were
analyzed to determine if the various treatments resulted in specific leaf spectral features
related to TICs. Figure 17 illustrates typical effects of the chemicals on plants and Figure 18
depicts representative spectra collected, in this case for canola exposed to Cl2 and SO2. Here
one can see key absorption features observed in endmember spectra, which were exploited
in subsequent analysis.
Observations showed that both environmental stress and TIC treatments induce similar
spectral features inherent to plants, which can be related primarily to chlorophyll and water
loss. These include pigments in the visible and cellulose, lignin, lipids, starches, and sugars
in the short wave infrared. Although no specific spectral features could be tied to individual
TICs, an analysis of the data using vegetation indices, which focus on key spectral bands
associated with chlorophyll, pigments and water content, showed that the TICs and
512 Remote Sensing – Applications

Fig. 17. Examples of environmental (H2O, NaCl) and chemicals (NH3) stresses on plants.

Fig. 18. Representative spectra for canola experiments using Cl2 and SO2. Mean spectra of
control G1 (black line, mature healthy) and G4C (pink line, high senescence) are included for
reference. Arrows denote key absorption features observed in endmember spectra
compared with G1. Black dotted box denotes smoothing of red edge. Chemical exposure
differences is apparent.

environmental stresses result in diagnostic light reflectance data trends from healthy mature
to highly stressed leaves.
Comparison of relevant vegetation indices, such as that depicted in Figure 19, showed that
specific combinations could be used to distinguish NH3, SO2, Cl2 consistently across all three
species (Rogge et al, 2008). The trends result from the variable leaf response within plants,
between plants and between species and it is expected much of the variability observed within
species would be preserved or even enhanced in nature. As such it is encouraging for the
possible detection of TIC effects on natural vegetation using airborne/spaceborne imagery.
As the detection methodology was developed from leaf-level observations, it is important to
note that field trials remain to be conducted in order to test if the findings of this study can
be extended to the detection of TICs in the natural environment. The principal unknown is
the effect of varying vegetation canopy structural parameters (e.g. canopy gaps, leaf area)
and background properties (litter and soil reflectance) on the specific data trends that were
identified.
Demonstration of Hyperspectral Image Exploitation for Military Applications 513

Fig. 19. A selection of vegetation indices across all species for treatments with NaCl, NH3,
SO2, Cl2, HCN, and, dehydration (H2O), senescence and controls plants. The existence of
species-specific responses of vegetation to TICs presents both a challenge and an
opportunity for regional remote sensing.

While the exact physiological response to each stressor remains to be understood, the
existence of species-specific responses of vegetation to TICs presents both a challenge and
an opportunity for regional remote sensing.

5. Marine algorithms
The HYMEX project studied several potential marine applications in collaboration with
Borstad Associates and the Dept. of Fisheries and Ocean (Institut Maurice Lamontagne and
Bedford Institute of Oceanography). We conducted airborne hyperspectral surveys on East
and West coasts of Canada to evaluate algorithms for near-shore bathymetry, beach
trafficability, near-shore bottom type mapping as well as retrieval of chlorophyll and
suspended matter concentrations as indicator of water clarity. More details are provided in
(Ardouin, 2007). Through this work it was realized that most of these algorithms can be
applied to multispectral imagery and that their experimental validation is difficult. The later
is particularly true for products that vary with time (or current) and thus would require
many measurement stations (for validation) that would need to operate coincidentally with
the airborne survey and be distributed over the area of the survey.
More recently, we tasked OEA Technologies to provide an operational assessment of
HYMEX marine algorithms. In this assessment, a distinction was made between dynamic
(e.g. water color) and static (e.g. bathymetry) products. It was pointed out that the Canadian
Forces needs for off-shore dynamic products (e.g. water colour) is already fulfilled by
marine multispectral sensors (MERIS, MODIS) with pixel size > 250m (Williams, 2009).
There might however be a niche for hyperspectral sensors (airborne and spaceborne) which
514 Remote Sensing – Applications

typically have better spatial resolution (e.g. from submeter to tens of meter) for near-shore
static and dynamic products. The better spatial resolution and increased number of bands of
hyperspectral sensors might provide an ability to handle the more complex near-shore
environment. Potential static products to consider include target detection and near-shore
bottom characterization in support of mine countermeasures and battlespace mapping and
possibly submarine operations. To this we can also add near-shore bathymetry in support of
route survey, battlespace mapping, anti-submarine warfare and submarine operations.
While not requiring hyperspectral sensing, HSI could keep playing a role (e.g. selection of
optimal bands) in the development of new dynamic products for both near-shore and off-
shore applications. Overall, this assessment point to possible follow-up for marine
applications development with hyperspectral sensors.

6. Conclusion
In this chapter, we discussed a wide variety of military applications resulting from the
exploitation of reflective hyperspectral imagery. These applications were demonstrated in
the DRDC HYMEX project, allowing DRDC and the Canadian Forces stakeholders to get
more familiar with the military utility of hyperspectral imagery. While some of these
applications such as target detection are relatively mature and are near to operational
deployment, others still require further development but are representative of the unique
capability of hyperspectral remote sensing. The many datasets that were acquired and the
algorithms and exploitation tools that were developed in the project are being used to
continue the development of hyperspectral technology at DRDC. One avenue that is being
pursued is the development of an airborne hyperspectral real-time target detection
demonstration system. We are also looking at opportunities to further develop the land
mapping and marine applications areas as well as potential space-based demonstration with
international partners.

7. References
Antunes, M.S.; Ha, S.B.; Tewari-Singh, N.; Morey, K.J.; Trofka, A.M.; Kugrens, P.; Deyholos,
M. & Medford J.I. (2006) A synthetic de-greening gene circuit provides a reporting
system that is remotely detectable and has a re-set capacity, Plant Biotechnology J.,
Vol. 4, No. 6, (November 2006), pp. 605-622
Ardouin, J.-P.; Lévesque, J. & Rea, T.A. (2007) A Demonstration of Hyperspectral Image
Exploitation for Military Applications, Proc. of the 10th International Conference on
Information Fusion (FUSION 2007), Quebec, Canada, (9-12 July 2007), pp. 1-8., ISBN
978-0-662-45804-3
Beeftink, H.H. (1951) Some observations on tamarack or eastern larch. Forestry Chronicle,
Vol. 27, No. 1, (1951), pp. 38-39
Bernstein, L.S.; Adler-Goldstein, S.M.; Sundberg, R.L., Levine, R.Y.; Perkins, T.C.; Berk, A.;
Ratkowski, A.J.; Felde, G. & Hoke, M.L. (2005) A new method for atmospheric
correction and aerosol optical property retrieval for VIS-SWIR multi- and
hyperspectral imaging sensors : QUAC (QUick Atmospheric Correction), IEEE
International Geoscience and Remote Sensing Symposium, (25-27 July 2005), pp. 3549-
3552, ISBN 0-7803-9050-4
Demonstration of Hyperspectral Image Exploitation for Military Applications 515

Buckingham, R.; Staenz, K. & Hollinger (2002) A., Review of Canadian Airborne and Space
Activities in Hyperspcetral Remote Sensing, Canadian Aeronautics and Space Journal,
Vol. 48, No. 1, (2002), pp. 115-121
Buckingham, R. & Staenz, K. (2008) Review of current and planned civilian space
hyperspectral sensors for EO, Can. J. Remote Sensing, Vol. 34, Suppl. 1, (2008), pp.
S187-S197
Cooley T.; Davis, M. & Straight, S. (2006) ARTEMIS-Advanced Tactically-Effective Military
Imaging Spectrometer: Tactical Satellite 3 for Responsive Space Missions,
International Symposium on Spectral Sensing Research, Bar Harbor, MN, (May 2006)
Cutter, M.A.; Johns, L.S.; Lobb, D.R.; Williams, T.L. & Settle, J.J. (2003) Flight Experience of
the Compact High Resolution Imaging Spectrometer (CHRIS), Proc. of SPIE Imaging
Spectrometry IX. Vol. 5159, (2003) pp. 392-405
Davenport, M. & Ressl, W. (1999) Shadow Removal in Hyperspectral Imagery, International
Symposium on Spectral Sensing Research (ISSSR 99), Las Vegas (November 1999)
Deyholos, M.; Faust, A.A.; Miao, M.; Montoya, R. & Donahue, D.A. (2006) Feasibility of
landmine detection using transgenic plants, in Detection and Remediation
Technologies for Mines and Minelike Targets XI, J. Thomas Broach, Russell S. Harmon,
John H. Holloway, Jr, eds., Proc. SPIE, Vol. 6217, (May 2006), pp. 700-711
Deyholos, M.K.; Rogge, D.; Rivard, B. & Faust, AA. (2007) Plants as sensors for toxic
industrial chemicals and munitions: A feasibility analysis, In Vitro Cellular &
Developmental Biology-Animal, Vol. 43, pp. S7-S7.
Deyholos, M. (2009) Demonstration of Plant-based Explosives Detection, Contract Report
Defence R&D Canada – Suffield, DRDC Suffield CR-2010-010 (2009).
Pearlman, J.S.; Barry, P.S; Segal, C.C.; Shepanski, J.; Beiso & D. & Carman, S.L. (2003)
Hyperion, a Space-Based Imaging Spectrometer, IEEE Trans. on Geoscience and
Remote Sensing, Vol. 41, No. 6, (June 2003), pp. 1160- 1173, ISSN 0196-2892
Peddle, D.R. & Ferguson, D.T. (2002) Optimization of Multisource Data Analysis using
Evidential Reasoning for GIS Data Classification, Computers & Geosciences, Vol. 28,
No 1, (2002), pp. 45-52
Peddle, D.R.; Franklin, S.E.; Johnson, R.L.; Lavigne, M.B. & Wulder, M.A. (2003) Structural
Change Detection in a Disturbed Conifer Forest Using a Geometric Optical
Reflectance Model in Multiple-Forward Mode. IEEE Trans. on Geoscience and Remote
Sensing Vol. 41, No. 1, (Jan 2003), pp. 163-166, ISSN 0196-2892
Peddle, D.R. & Smith, A.M. (2005) Spectral Mixture Analysis of Agricultural Crops:
Endmember Validation and Biophysical Estimation in Potato Plots, International
Journal of Remote Sensing, Vol. 26, No. 22, (2005), pp. 4959–4979
Peddle, D.R.; Boulton, R.B.; Pilger, N.; Bergeron, M. & Hollinger, A. (2008) Hyperspectral
detection of chemical vegetation stress: evaluation for the Canadian HERO satellite
mission, Can. J. Remote Sensing, Vol. 34, Suppl. 1, (2008), pp. S198-S216
Reed, I. S. & Yu, X. (1990) Adaptive multiple-band CFAR detection of an optical pattern
with unknown spectral distribution, IEEE Trans. on Acoustics, Speech, and Signal
Processing, Vol. 38, No. 10, (October 1990), pp. 1760-1770, ISSN 0096-3518
Rivard, B.; Deyholos, M. & Rogge, D. (2008) Chemical Effects on Vegetation Detectable in optical
bands 350-2500 nm, Contract Report Defence R&D Canada - Suffield , DRDC
Suffield CR-2008-234, (2008)
516 Remote Sensing – Applications

Rogge, D.; Rivard, B.; Deyholos, M.; Lévesque, J. & Faust, A. (2008) Toxic Industrial
Chemical Effects on Poplar, Canola, and Wheat detectable over the 450-2500 nm
Spectral Range, Intl. Geoscience & Remote Sensing Symposium (IGARSS 2008), Boston,
MA, (July 2008)
Roy, V. (2010) Hybrid algorithm for hyperspectral target detection, Proc of SPIE Algorithms
and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XVI, Vol.
7695, May 2010, pp. 1-10
Sentlinger, G.; Davenport M., & Ardouin, J.-P. (2003) Automated Target Recognition in
Hyperspectral Imagery using Subpixel Spatial Information, Proc. SPIE Aerosense –
Automatic Target Recognition XIII, Vol 5094, (April 2003)
Settle, J. (2002) On constrained energy minimization and the partial unmixing of
multispectral images, IEEE Trans. on Geoscience and Remote Sensing, Vol. 40, No. 3,
(March 2002), pp. 718-721, ISSN 0196-2892
Settle, J. (2004) On the Use of Remotely Sensed Data to Estimate Spatially Averaged
Geophysical Variables, IEEE Trans. on Geoscience and Remote Sensing, Vol. 42, No. 3,
(March 2004), pp. 620-631, ISSN 0196-2892
Smith, G. M. & Milton, E.J. (1999) The use of the empirical line method to calibrate remotely
sensed data to reflectance, International Journal of Remote Sensing, Vol. 20, (1999), pp.
2653-2662
Van Chestein, Y. (2011) Comparative evaluation of the Mercury classification algorithm: On the
influence of the number of bands on classification accuracy using hyperspectral data,
Defence R&D Canada - Valcartier Technical Memorandun, TM 2010-385, (March
2011)
Webster, A.H.; Davenport, M.R. & Ardouin, J.-P. (2006) 3D Deconvolution of Vibration
Corrupted Hyperspectral Images, International Symposium on Spectral Sensing
Research 2006, Bar Harbor, MA., (May 2006).
Williams, D.; Vachon, P.W; Wolfe, J.; Robson, M.; Renaud, W.; Perrie, W.; Osler, J.; Isenor,
A.W.; Larouche, P.; Jones, C. (2009) Spaceborne Ocean Intelligence Network – SOIN –
fiscal year 08/09 year-end summary, Defence R&D Canada - Ottawa External Client
Report, DRDC Ottawa ECR 2009-139, (September 2009)