Journal of

Marine Science
and Engineering

Article
Ultra-High-Resolution Mapping of Posidonia
oceanica (L.) Delile Meadows through Acoustic,
Optical Data and Object-based Image Classification
Sante Francesco Rende 1, * , Alessandro Bosman 2 , Rossella Di Mento 1 , Fabio Bruno 3 ,
Antonio Lagudi 4 , Andrew D. Irving 5 , Luigi Dattola 6 , Luca Di Giambattista 2,7 ,
Pasquale Lanera 1 , Raffaele Proietti 1 , Luca Parlagreco 1 , Mascha Stroobant 8 and
Emilio Cellini 9
1 ISPRA, Italian National Institute for Environmental Protection and Research, Via Vitaliano Brancati 60,
00144 Rome, Italy; rossella.dimento@isprambiente.it (R.D.M.); pasquale.lanera@isprambiente.it (P.L.);
raffaele.proietti@isprambiente.it (R.P.); luca.parlagreco@isprambiente.it (L.P.)
2 Istituto di Geologia Ambientale e Geoingegneria, Consiglio Nazionale delle Ricerche (CNR-IGAG),
Via Eudossiana 18, 00184 Rome, Italy; alessandro.bosman@cnr.it (A.B.); luca.digiambattista@uniroma1.it (L.D.G.)
3 Department of Mechanical, Energy and Management Engineering—DIMEG, University of Calabria,
87036 Arcavacata di Rende, Italy; fabio.bruno@unical.it
4 3D Research Srl, 87036 Quattromiglia, Italy; antonio.lagudi@3dresearch.it
5 Coastal Marine Ecosystems Research Centre, Central Queensland University,
Norman Gardens, QLD 4701, Australia; a.irving@cqu.edu.au
6 ARPACAL—Regional Agency for the Environment—Centro di Geologia e Amianto, 88100 Catanzaro, Italy;
l.dattola@arpacal.it
7 Sapienza University of Rome—DICEA, Via Eudossiana 18, 00184 Rome, Italy
8 Research and Technological Transfer Area, Research Projects Unit, University of Florence,
50121 Florence, Italy; mascha.stroobant@unifi.it
9 ARPACAL—Regional Agency for the Environment—CRSM, Regional Marine Strategy Centre,
88100 Catanzaro, Italy; e.cellini@arpacal.it
* Correspondence: francesco.rende@isprambiente.it




Received: 20 July 2020; Accepted: 17 August 2020; Published: 22 August 2020


Abstract: In this study, we present a framework for seagrass habitat mapping in shallow (5–50 m) and
very shallow water (0–5 m) by combining acoustic, optical data and Object-based Image classification.
The combination of satellite multispectral images-acquired from 2017 to 2019, together with Unmanned
Aerial Vehicle (UAV) photomosaic maps, high-resolution multibeam bathymetry/backscatter and
underwater photogrammetry data, provided insights on the short-term characterization and
distribution of Posidonia oceanica (L.) Delile, 1813 meadows in the Calabrian Tyrrhenian Sea. We used
a supervised Object-based Image Analysis (OBIA) processing and classification technique to create
a high-resolution thematic distribution map of P. oceanica meadows from multibeam bathymetry,
backscatter data, drone photogrammetry and multispectral images that can be used as a model for
classification of marine and coastal areas. As a part of this work, within the SIC CARLIT project, a field
application was carried out in a Site of Community Importance (SCI) on Cirella Island in Calabria
(Italy); different multiscale mapping techniques have been performed and integrated: the optical and
acoustic data were processed and classified by different OBIA algorithms, i.e., k-Nearest Neighbors’
algorithm (k-NN), Random Tree algorithm (RT) and Decision Tree algorithm (DT). These acoustic
and optical data combinations were shown to be a reliable tool to obtain high-resolution thematic
maps for the preliminary characterization of seagrass habitats. These thematic maps can be used
for time-lapse comparisons aimed to quantify changes in seabed coverage, such as those caused by
anthropogenic impacts (e.g., trawl fishing activities and boat anchoring) to assess the blue carbon
sinks and might be useful for future seagrass habitats conservation strategies.

J. Mar. Sci. Eng. 2020, 8, 647; doi:10.3390/jmse8090647 www.mdpi.com/journal/jmse

J. Mar. Sci. Eng. 2020, 8, 647 2 of 25

Keywords: Pléiades-HR satellite images; multibeam bathymetry; UAV; seafloor mapping; habitat
mapping; object-based image analysis (OBIA); seagrass; Mediterranean Sea; Tyrrhenian Sea

1. Introduction

1.1. Seagrass Habitat, Importance and Knowledge

Seagrass beds are distributed over a near-global extent and play an important role in coastal
ecosystems as primary producers, providers of habitat and environmental structure, and shoreline
stabilizers [1,2]. Seagrasses are considered one of the most important coastal habitats [3], as they
support a wide range of ecologically and economically important marine species from different trophic
levels [4]. Moreover, vegetated coastal ecosystems and especially seagrass meadows, are highly
productive and have exceptional capacities to sequester carbon dioxide (CO2 ) [5]. P. oceanica is an
endemic Mediterranean species and is protected by specific European legislation frameworks such
as the EU Habitat Directive (92/43/CEE) which includes P. oceanica beds among its priority habitats
(Habitat Type 1120). Furthermore, according to the Water Framework Directive 2000/60/EC and the
Marine Strategy Framework Directive (MFSD -2008/56/EC), P. oceanica is selected as a representative
species among the angiosperm quality elements for evaluating “Good Ecological Status” and “Good
Environmental Status” in the Mediterranean marine environments. Therefore, in order to protect
this seagrass ecosystem, it is important to establish its preservation status and regularly monitor its
abundance and distribution. Furthermore, the evaluation of its ecological status should be based on a
monitoring strategy design that should be able to record accurately all its different spatial configurations
(ranging from highly fragmented to continuous meadows) and arrangements, ranging across scales of
meters to kilometers (seagrass landscapes), meters to tens of meters (patches), to tens of centimeters to
meters (rhizomes, shoot groups) [6,7].

1.2. Application of Multispectral Satellite Systems in Seagrass Habitat Mapping

Remotely sensed aerial or satellite optical imagery has been used successfully by several authors
to assess the spatial distribution and spatiotemporal changes in seagrass habitats [8–14] and has been
proven to be suitable for mapping and monitoring seagrass ecosystems [14,15]. Seagrass detection with
multispectral or hyperspectral imaging is one of the most widespread methods in this research area,
since data collection is faster than other surveying techniques. Moreover, these techniques have lower
costs, medium resolution accuracy, easier replicability and cover wider areas compared to traditional
field-based methods [5]. However, they typically provide low-resolution images and even if remote
technologies have a great potential in seabed mapping studies, the extraction of the bottom reflectance
spectrum from the orbital optical sensors data is quite complex. Several processes (i.e., absorption and
scattering) affect the electromagnetic radiation during its passage through the atmosphere and the
water column before this radiation is recorded by the satellite sensors, therefore information quality is
reduced [16,17]. Recent advances in remote sensing satellites and in acoustic sensors, computer vision,
OBIA classification algorithms and pattern recognition can overcome these limitations and encourage
new approaches to develop more accurate mapping techniques [8].

1.3. Application of Multibeam Systems in Seagrass Habitat Mapping

Side Scan Sonar (SSS) and Single-Beam Echo-Sounder (SBES), were considered as the most
common acoustic methods for seafloor mapping and analysis [18–21]. More recently (i.e., a few decades
ago) other techniques such as the Multibeam Echo-Sounder (MBES), allowed the acquisition of higher
resolution bathymetry and backscatter data especially for seagrass meadows [22–29].
Bathymetry and backscatter data and their derived products (such as Slope, Aspect, Curvature
and Terrain Ruggedness maps) can play a pivotal role in providing useful information for managing
J.J. Mar.
Mar. Sci.
Sci. Eng.
Eng. 2020,
2020, 8,
8, xx FOR
FOR PEER
PEER REVIEW
REVIEW 33 of
of 25
25
J. Mar. Sci. Eng. 2020, 8, 647 3 of 25
Bathymetry
Bathymetry and
and backscatter
backscatter data
data and
and their
their derived
derived products
products (such
(such as
as Slope,
Slope, Aspect,
Aspect, Curvature
Curvature
and Terrain Ruggedness maps) can play a pivotal role in providing useful information
and Terrain Ruggedness maps) can play a pivotal role in providing useful information for managing for managing
the world’s coasts and oceans [30]: seabed mapping with high-resolution MBES provides bathymetric
the
the world’s
world’s coasts
coasts and
and oceans
oceans [30]:
[30]: seabed
seabed mapping
mapping with
with high-resolution
high-resolution MBES
MBES provides
provides bathymetric
bathymetric
profiles that can be visualized as a high-resolution Digital Elevation Models (DEMs) along with
profiles
profiles that
that can
can be
be visualized
visualized as as aa high-resolution
high-resolution Digital
Digital Elevation
Elevation Models
Models (DEMs)
(DEMs) along
along with
with
acoustic backscatter intensity from the seafloor and, more recently, from the water column [31–34]
acoustic backscatter intensity from the seafloor and, more recently, from the water
acoustic backscatter intensity from the seafloor and, more recently, from the water column [31–34] column [31–34]
(Figure 1). Moreover, for seagrass meadow mapping, the MBES technique provides a three-dimensional
(Figure
(Figure 1).
1). Moreover,
Moreover, for for seagrass
seagrass meadow
meadow mapping,
mapping, the
the MBES
MBES technique
technique provides
provides aa three-
three-
reconstruction of the seafloor in shallow waters (from about 50 m to 10 m of depth) whereas optical
dimensional
dimensional reconstruction
reconstruction of of the
the seafloor
seafloor in
in shallow
shallow waters
waters (from
(from about
about 5050 m m to
to 10
10 mm of
of depth)
depth)
(multispectral satellites images), methods are unable to provide detailed maps after 15–25 m in
whereas
whereas optical
optical (multispectral
(multispectral satellites
satellites images),
images), methods
methods are
are unable
unable toto provide
provide detailed
detailed maps
maps after
after
depth. Indeed, it delivers a precise and accurate bathymetric data for 3D reconstruction of seafloor
15–25
15–25 mm in
in depth.
depth. Indeed,
Indeed, itit delivers
delivers aa precise
precise and
and accurate
accurate bathymetric
bathymetric data
data for
for 3D
3D reconstruction
reconstruction
morphologies (e.g., seagrass matte), by highlighting the presence of canopies of P. oceanica meadows
of
of seafloor
seafloor morphologies
morphologies (e.g.,
(e.g., seagrass
seagrass matte),
matte), by
by highlighting
highlighting the
the presence
presence ofof canopies
canopies ofof P.
P. oceanica
oceanica
(Figure 2).
meadows (Figure
meadows (Figure 2). 2).

Figure
Figure 1.
1. (a)
1. (a) Three-dimensional
Three-dimensional (3D) (3D) reconstruction
reconstruction ofof seafloor
seafloor morphology
morphology and and water
water column
column data
data
from
from multibeam
multibeam bathymetry
bathymetry (Multibeam
(Multibeam Echo-Sounder
Echo-Sounder (MBES))
(MBES)) for
for characterizing
characterizing the
the matte
matte and
and the
the
canopy
canopy height
height of
height of P.
P. oceanica.
P. oceanica. (b)
oceanica. (b) detail
detail of
of acoustic
acoustic fan
fan view
view collected
collected by
by multibeam
multibeam showing
showing the
the height
height
of
of canopy
canopy and
and the
the matter
matter of
matter of P.
of P. oceanica
P.oceanica meadows
oceanicameadows
meadowson on
onaaasandy
sandy seafloor.
sandyseafloor.
seafloor.

Figure
Figure 2.
2. Diagram
Diagram and comparison between different data sources and their resolution at the seafloor,
Figure 2. Diagram and
and comparison
comparison between
between different
different data
data sources
sources and
and their
their resolution
resolution at
at the
the seafloor,
seafloor,
for
for seabed
seabed and
and habitat
habitat mapping:
mapping: Multispectral
Multispectral Satellite
Satellite Image
Image (MSSI);
(MSSI); High-resolution
High-resolution Multibeam
Multibeam
for seabed and habitat mapping: Multispectral Satellite Image (MSSI); High-resolution Multibeam
Echo-Sounder System
Echo-Sounder System (MBES);
System(MBES); Development
(MBES);Development
Development Vehicle
Vehicle for
for Scientific
Scientific Survey
Survey (DEVSS);
(DEVSS); Unmanned
Unmanned
Echo-Sounder Vehicle for Scientific Survey (DEVSS); Unmanned Aerial
Aerial
Aerial Vehicle
Vehicle (UAV).
(UAV).
Vehicle (UAV).
J. Mar. Sci. Eng. 2020, 8, 647 4 of 25

However, the MBES surveys are limited in very shallow water areas especially at a 5 m depth
where vessel navigation might be difficult and dangerous and the swath coverage is very limited
(generally 3–4 times the depth of the seabed) and, hence, might significantly increase the surveying
time and its costs.

1.4. Application of Unmanned Aerial Vehicles (UAVs) and Autonomous Surface Vehicles (ASVs) Systems in
Seagrass Habitat Mapping
Marine observation techniques are usually divided into two major categories: (1) remotely
acquired data and (2) field measurements that are required for validation. The first category includes
satellite and aerial images, while the second includes techniques of ground-truth sampling like
underwater images, videos, underwater photogrammetry, in situ measurements and sampling
procedures. For seagrass habitat assessment, very fine resolution Unmanned Aerial Vehicle (UAV)
imagery has been effective for density coverage mapping and for detecting changes in small patches
and landscape features; these assessments would not have been possible with satellite or aerial
photography [10,35–38]. However, the application of UAVs for mapping and monitoring of seagrass
habitats has been limited by the optical characteristic of the water (e.g., turbidity), and the environmental
conditions (e.g., solar elevation angle, cloud cover and wind speed) during image acquisition [39].
As such, most research has been confined to clear shallow tropical waters [40,41] or small subsets
of exposed intertidal seagrass beds in temperate regions [36,37]. When it comes to Autonomous
Surface Vehicles (ASVs), they are considered promising approaches in the marine science community.
An immediate advantage is related to the collection of ground-truth data (optic and acoustic) with
single-beam echosounders and images from underwater photogrammetry cameras, which ensures
very high accuracy in very shallow waters [9,42,43]. This approach could represent, in the future, a
replacement of diver seagrass investigations (i.e., snorkeling or scuba diving [11]) especially during
either impractical or dangerous, field campaigns.

1.5. OBIA Classification Algorithms in Seagrass Habitat Mapping

Object-based image analysis (OBIA) is an advanced classification method that incorporates
spectral, weight, color, texture, shape and contextual information to identify thematic classes in optical
and acoustic data-derived images. The first step of OBIA classification is to perform a multiresolution
segmentation of the image to identify homogeneous objects (note: the term “object” in this case stands
for a contiguous group of spatial data such as pixels in a bathymetric grid). The segmentation process
is based on predefined parameters such as compactness, shape and scale, derived from real-world
knowledge of the characteristics to be identified and classified. For machine learning-based mapping,
several algorithms have been developed to improve either the classification from satellite imagery
or seagrass prediction from environmental parameters [44,45]. In a second step, the object-based
classification uses machine learning algorithms such as Support Vector Machine (SVM), Random Tree
(RT), Decision Tree (DT), Bayesian and k-Nearest Neighbor (k-NN) that have been further refined
and are accessible on a variety of data analysis software. Seafloor mapping based on multibeam
bathymetry and backscatter data using Object-based Image Analysis has been recently carried out
by Janowski [46,47]. The authors found that OBIA-based approaches led to a doubling of the overall
accuracy and Kappa measures, from 42% and 0.24–0.27, to 86% and 0.81. Some studies [45,48,49] already
employ the OBIA technique for benthic habitat mapping and automatic classification analysis of reefs
and seabed through the use the acoustic data of MBES or SSS has been described in several scientific
works [47,50]. The OBIA technique applied to map coral reefs has successfully shown improved
performance across different spatial scales [51–54]. Moreover, the OBIA technique is particularly suited
for the analysis of very high resolution (VHR) images such as QuickBird or WorldView-2, where the
increased heterogeneity of submeter pixels would otherwise confuse pixel-based classifications yielding
to an undesired “salt and pepper effect”. Finally, in the study performed by Roelfsema et al. [55],
the OBIA semiautomated approach proved to be very effective in extracting seagrass species and cover
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 5 of 25
J. Mar. Sci. Eng. 2020, 8, 647 5 of 25

images [56]. The UAV images and OBIA classifications have also been used together for the mapping
of seagrass
maps habitats
through [37,39,40,57].
the analysis of satellite images [56]. The UAV images and OBIA classifications have also
been used together for the mapping of seagrass habitats [37,39,40,57].
2. Materials and Methods
2. Materials
In this and Methods
work, an integrated approach has been followed for the detection and mapping of
seagrass. Inwork,
In this particular, multispectral
an integrated images
approach has from
been satellite,
followedaerial
for theand underwater
detection digital images
and mapping from
of seagrass.
UAV/ASV,
In particular,Underwater Towed
multispectral imagesVideo
from Camera
satellite, Systems (UTCS)
aerial and and acoustic
underwater data from
digital images fromunderwater
UAV/ASV,
sonar technology
Underwater Towed have
Videobeen
Cameraused to map
Systems the seagrass
(UTCS) meadows
and acoustic data fromand monitor sonar
underwater their technology
extent and
condition
have been (Figure
used to 2).
map the seagrass meadows and monitor their extent and condition (Figure 2).

2.1.
2.1. Study Sites and Geomorphological Characterization
Characterization
Fieldwork
Fieldwork was carried
carried out
out at
at Cirella
Cirella Island,
Island, aa Site
Site ofof Community
Community Importance
Importance in in the
the Southern
Southern
Tyrrhenian ◦ 410 54.9300 N-Long 15◦ 480 8.0100 E). Four
TyrrhenianSea Sea(Lat
(Lat39
39°41′54.93″ N-Long 15°48′8.01″ E). Four kinds of surveys in 2018 and 2019 were were
carried
carried out
out (Multibeam,
(Multibeam, UAV UAVand andDevelopment
DevelopmentVehicle VehicleforforScientific
ScientificSurvey
Survey(DEVSS)
(DEVSS)and andUTCS).
UTCS).
Cirella
Cirella Island
Island is
is located
located onon the
the continental
continental shelf
shelf at
at aa distance
distance of
of 600
600 mm from
from the
the coast
coast and
and 2.5
2.5 km
km
from
from the
the shelf
shelf break
break (Figure
(Figure3).3). The
The island
island is
is composed
composed of of an
an emerged
emerged portion
portionofofabout
about0.07
0.07kmkm22 and
and
has 2 ; both
has aagreater
greater submerged
submerged portion of about
portion 0.3 km
of about 0.3 km2; portions belong belong
both portions to a carbonate unit of limestone
to a carbonate unit of
rocks fromrocks
limestone the Cretaceous [58]. The [58].
from the Cretaceous deeper
Theseabed
deepermorphologies (from about
seabed morphologies (from27about
m to27 41mmtowater
41 m
depth) around Cirella Island are characterized by semiflat bottom with a low gradient of about 1–1.5 ◦
water depth) around Cirella Island are characterized by semiflat bottom with a low gradient of about
toward the shelf
1–1.5° toward thebreak
shelf(Figure 3a), while
break (Figure 3a),inwhile
shallow water the
in shallow morphologies
water are quite
the morphologies arerugged with a
quite rugged
local ◦ . These morphologies are produced by rocky outcrops and by an extensive
with gradient up to 70up
a local gradient to 70°. These morphologies are produced by rocky outcrops and by an
matte of P.oceanica.
extensive matte of P.oceanica.

Figure 3. (a) Study

Figure 3. Study Area
Area Site
Site of
of Community
Community Importance
Importance (SCI)
(SCI) Isola
Isola di
di Cirella
Cirella located
located in
in the
the southern
southern
Tyrrhenian
TyrrhenianSea.Sea.Track
Tracklines of Multibeam
lines bathymetry,
of Multibeam UAV, UAV,
bathymetry, Underwater Towed Towed
Underwater Video Camera
Video Systems
Camera
and Autonomous Surface Vehicles (ASVs) surveys. Ground-truth training and validation
Systems and Autonomous Surface Vehicles (ASVs) surveys. Ground-truth training and validation sampling set
for classification step: (b,c) shallow water area; (d,e) very shallow water area.
sampling set for classification step: (b,c) shallow water area; (d,e) very shallow water area.
J. Mar. Sci. Eng. 2020, 8, 647 6 of 25
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 6 of 25

P. oceanica
P. oceanica isis located
located both
both onon sediment,
sediment, represented
represented mainly
mainly by by sand,
sand, and
and onon rocky
rocky outcrops.
outcrops.
Overall, the
Overall, the coverage
coverage of of P.
P.oceanica
oceanicaisisabout
about0.54
0.54 km
km of
22 of which
which about
about 65%65% isis located
located onon rocky
rocky outcrops.
outcrops.
The matte height is quite variable due to the rocky outcrop (maximum
The matte height is quite variable due to the rocky outcrop (maximum height: 1.5 m). height: 1.5 m).
AroundCirella
Around CirellaIsland,
Island,the thedepth
depthand and distribution
distribution of of P. oceanica
P. oceanica is uneven.
is uneven. On On the western
the western side,side,
it is
it is located
located from from
a deptha depth
of 15 mof while
15 m while
on theon the eastern
eastern side it side it is located
is located from a from
depthaofdepth
3 m. of 3 m.depths
These These
depthsthe
reflect reflect the prevailing
prevailing direction direction of the waves.
of the waves. Indeed,Indeed,
most of most
theofwaves
the waves
comecomefromfromthe W theand
W and
SW
SW sectors
sectors (caused
(caused by Ponente
by Ponente and and Sirocco
Sirocco winds,
winds, respectively,
respectively, as shown
as shown in thein elaboration
the elaboration of theof the
the
the directional wave analyzed from NOAA database from 1979 to 2009).
directional wave analyzed from NOAA database from 1979 to 2009). This suggests that their effectThis suggests that their effect
in shallow
in shallow water
water is is more
more relevant
relevant inin the
the western
western sector
sector of of the
the study
study area
area which
which is is exposed
exposed to to both
both
predominantwave
predominant wavedirections.
directions.

2.2. Remote Sensing

2.2.
The Pléiades satellite
The satellite image
imageacquired
acquiredonon2828September
September 2016 waswas
2016 used for for
used the the
study area.area.
study The
Pléiades
The image
Pléiades was supplied
image was suppliedwith 4with
spectral bandsbands
4 spectral and 2 and
m of2spatial resolution
m of spatial and wasand
resolution already
was
orthorectified
already with pixel
orthorectified withvalues in reflectance
pixel values [59] (Figure
in reflectance 4a). The
[59] (Figure image
4a). was initially
The image rectified
was initially using
rectified
the Dark
using ObjectObject
the Dark Subtraction (DOS)
Subtraction technique,
(DOS) withwith
technique, the aim of removing
the aim thethe
of removing disturbance
disturbancerelated
relatedto
the return flow of the atmosphere that separates the marine
to the return flow of the atmosphere that separates the marine surface surface from the transported sensor
transported sensor
aboard the
aboard the satellite
satellite (Figure
(Figure4b).
4b).

Figure 4.
Figure 4. Pléiades
Pléiades image
image correction
correction of
of the (a) True Color
Color Image
Image (TCI);
(TCI); (b)
(b) Pléiades
Pléiades image
image after
after
atmospheric correction;
atmospheric correction;(c)
(c)Pléiades
Pléiadesimage
imageafter
afterwater
watercolumn
columncorrection.
correction.

The
The second
second processing
processing performed
performed aa correction
correction to
to compensate
compensate for for the
the effects
effects of
of light
light intensity
intensity
attenuation as depth increases. This correction was made using the procedure
attenuation as depth increases. This correction was made using the procedure proposed by [60]proposed by [60] andandby
identifying and choosing homogeneous sandy bottoms at different depths.
by identifying and choosing homogeneous sandy bottoms at different depths. The aim was to The aim was to identify,
through
identify,appropriate regressions,
through appropriate the coefficients
regressions, that allow
the coefficients thatcorrelating the reflectance
allow correlating values values
the reflectance of the
bottom pixels to the depth (Figure 4c).
of the bottom pixels to the depth (Figure 4c).
The
The water
water column
column correction
correctionwaswas carried
carried out
out with
with the
the ERDAS
ERDAS IMAGINE
IMAGINE software
software viavia the
the Spatial
Spatial
Modeler
Modeler tool
tool [61].
[61].
The
The bands
bandsofofthe thePléiades
Pléiadessatellite
satelliteimage,
image,used to obtain
used the water
to obtain correction
the water coefficients,
correction are blue,
coefficients, are
green and red:
blue, green andin red:
this specific case, three
in this specific bands
case, threeoriginated from the blue–green,
bands originated blue–red and
from the blue–green, green–red
blue–red and
pairs have been
green–red pairsobtained
have been from image from
obtained processing.
image The three resulting
processing. bands
The three were then
resulting bands combined
were then to
create a new color image that was considered more appropriate for the classification
combined to create a new color image that was considered more appropriate for the classification phase (compared
to the unprocessed
phase (compared tooriginal image). original image).
the unprocessed
J. Mar. Sci. Eng. 2020, 8, 647 7 of 25

2.3. Multibeam Bathymetry

During 2018, in the SIC CARLIT project, high-resolution multibeam survey campaigns (bathymetry,
backscatter and water column data), were performed in the Tyrrhenian sea (Figure 3). During this
project, 535 nautical miles of multibeam bathymetry data in shallow water, covering 81 km2 of the
seafloor, were collected. However, the study area of Cirella Island described in this work was examined
in just one day. The multibeam survey was carried out in September 2018 collecting data between
10 m and 40 m water depth with a Kongsberg EM2040 Multibeam Echo-Sounder System installed on
“Astrea”, a 24 m long boat with a 3 m draft. The multibeam was set to use the equidistant mode with a
frequency of 300 kHz and 256 beams per swath. A positioning system was installed: a Kongsberg
Seapath 300 with the correction of a Fugro HP Differential Global Positioning System (horizontal
accuracy: 0.2 m). A Kongsberg Seatex IMU (MRU 5) Motion Reference Unit recorded the attitude
and rotation about the three orthogonal axes centered on the vehicle’s center of gravity (pitch, roll,
heave and yaw movements). A Valeport mini sound velocity sensor (SVS), mounted in proximity
to the transducers, measured continuously the sound velocity for the forming beam. Three sound
velocity profiles were collected with a Valeport sound velocity profiler. Data were logged, displayed
and checked in real-time with the Kongsberg SIS software (Seafloor Information System). Bathymetric
multibeam data were processed using Caris HIPS and SIPS hydrographic software. The processing
workflow consisted of a patch test and in the application of statistical and geometrical filters to
remove coherent/incoherent noise [27,34]. Since the investigated area is a microtidal environment,
particular attention was given to tidal corrections to make sure that the DEMs were corrected. The local
hydrometric level was related to the permanent tide station of Palinuro, belonging to the National
Tide gauge Network (http://www.mareografico.it). The backscatter intensity (BS) maps were created
by using Caris HIPS and SIPS hydrographic software after applying the Angle Varying Gain (AVG)
beam pattern correction to remove the angular dependence artifacts from the seabed backscatter.
The soundings profiles were merged and gridded for the generation of DEMs at 0.3 m cell size resolution
and the backscatter intensity map at 0.2 m cell size resolution. All data (DEMs, backscatter intensity
map, UAV data and multispectral images), were merged through a GIS software (i.e., Global Mapper
20.0) that allowed to control the overlapping datasets positioning [62]. Furthermore, about 717 control
points were used to validate, by direct observation, the Multibeam morphology, the ground-truth
reference data for training, and the classification algorithm, Table 1.

Table 1. Number of ground-truth data derived by the acoustic data MBES, and collected through the
Underwater Towed Camera Systems (UTCs), ASVs DEVSS and UAV.

Class MBES UTCS ASV UAV

P. oceanica (P) 214 41 26 13
Rock (R) 95 \ 6 50
Mobile Fine sediment (FS) 197 13 \ \
Coarse sediment (CS) 211 5 \ \
Cystoseira (Cy) \ \ \ 49
Total 717 59 32 112

2.4. UAV Survey and Processing for Digital Terrain and Marine Model Generation
The images were collected across Cirella Island in July 2019 via a Parrot Anafi Work UAV.
This model has an onboard camera with a 1/2.4-inch CMOS sensor which captures 21-megapixel images
(.jpeg format) and an f/2.4 lens with an 84◦ field of view. Automated flights were carried out with the
Pix4D Capture free drone flight planning mobile application with an 80% overlap on both axes, and a
flight altitude that ranged between 60–75 m depending on the total size of the surveyed site. In total,
4 photogrammetric flights were performed in order to cover the north/south and east/west side of the
island. A total of 360 frames were captured during a continuous flight (Figure 5). The surveyed images
were processed with Pix4D Mapper software [63] by using 11 ground control points located around
J.J. Mar.
Mar. Sci.
Sci. Eng.
Eng. 2020,
2020, 8, x FOR
FOR PEER
8, x647 PEER REVIEW
REVIEW 88 of
of 25
25

around the
around the island
island between
between 00 and
and 1010 m
m altitude.
altitude. The
The mosaic
mosaic image
image and
and the
the DEM
DEM were
were processed
processed atat
the island
0.02 m
0.02 m and between
and 0.3
0.3 m 0 and 10
m resolution m altitude.
resolution respectively. The
respectively. In mosaic image
In particular,
particular, 112 and the
112 frames DEM
frames were were
were used
used to processed
to collect, at
collect, by 0.02
by UAV m and
UAV direct
direct
0.3 m resolution
observation,
observation, therespectively.
the ground-truthInreference
ground-truth particular,
reference 112for
data
data frames
for wereand
training
training used
and to collect,the
validating
validating byclassification
the UAV direct observation,
classification algorithm,
algorithm,
the ground-truth
Table
Table 1.
1. reference data for training and validating the classification algorithm, Table 1.

Figure 5.
Figure 5. (a)
(a)Unmanned
UnmannedAerial
Unmanned AerialVehicle
Aerial Vehicle
Vehicle (Anafi
(Anafi
(Anafi Parrot
Parrot
Parrot WorkWork
Work
UAV) UAV)
UAV) survey
survey
survey performed
performed
performed from
fromfrom aa small
a smallsmall
boat;
boat;
boat;
(b) (b) GNSS
(b)
GNSS GNSS surveys
surveys
surveys carried
carried
carried out along
out
out along along
the theof
the
coast coast
coast of Cirella
of
CirellaCirella Island
Island Island to identify
to
to identifyidentify the 11
the 11 ground
the 11 ground ground control
controlcontrol
points;
points;
points;
(c) (c)
UAV(c) UAV Mosaic
UAV
Mosaic Mosaicofimage
image image ofIsland.
of
Cirella Cirella Island.
Cirella Island.

2.5.
2.5. Image Ground-Truth
2.5. Image
Image Ground-Truth Data
Ground-Truth Data
Data
The
The ASVrobotic
The ASV
ASV roboticplatform
robotic platformDEVSS
platform DEVSS(DEvelopment
DEVSS (DEvelopment
(DEvelopment Vehicle
Vehicle
Vehicle forfor
for Scientific
Scientific
Scientific Survey)
Survey)
Survey) developed
developed
developed by
by the
by the
the 3D Research
3D Research
3D Research private private
private research research
research companycompany
company [64] [64]
[64] and and
and the the Underwater
the Underwater
Underwater Towed Towed
Towed CameraCamera Systems
Camera Systems (UTCS)
Systems (UTCS)
(UTCS)
were used,
were used,
were in
used, in the
in the very
the very shallow
very shallow water
shallow water area,
water area,
area, toto collect
to collect
collect thethe ground-truth
the ground-truth reference
ground-truth reference data
reference data for
data for training
for training and
training and
and
validating the
validating the
validating classification
the classification algorithm,
classification algorithm,
algorithm, TableTable 1 [65].
Table 11 [65]. The
[65]. The ASVs
The ASVs
ASVs waswas equipped
was equipped
equipped with with a GoPro
with aa GoPro Hero
GoPro HeroHero 444
Black model,
Black model,
Black which
model, which
which is is a consumer-brand
is aa consumer-brand high-definition
consumer-brand high-definition
high-definition sports sports camera
sports camera
camera withwith a
with aa 1212
12 MPMP
MP HD HD CMOS
HD CMOS
CMOS
sensor, 00 size [64]. The GoPro Hero 4 Black records at different video and photo resolutions and
sensor, 1/2.5
sensor, 1/2.5″ size
1/2.5″ size [64].
[64]. The
The GoPro
GoPro Hero
Hero 44 Black
Black records
records at at different
different video
video andand photo
photo resolutions
resolutions andand
Field-Of-View (FOV).
Field-Of-View (FOV).
Field-Of-View In
(FOV). In this work,
In this
this work,we
work, weused
we usedthe camera
used the (set
the camera in time-lapse
camera (set
(set in mode)
in time-lapse with
time-lapse mode) 12 MP
mode) with widescreen
with 12 12 MP
MP
1080 p and a1080
widescreen
widescreen FOVppofand
1080 108◦aa. FOV
and The camera
FOV 108°.was
of 108°.
of Thepositioned
The camera was
camera face-down
was positioned
positionedin order to obtain
face-down
face-down vertical
in order
in order toimages
to obtain
obtain
at the same
vertical
vertical height
images
images at from
at the
the same
same theheight
bottom.
height from
from Inthe
thebottom.
the shallowIn
bottom. Inwater area, time-lapse
the shallow
the shallow water area,
water photos
area, were photos
time-lapse
time-lapse recorded
photos by
were
were
using: an
recorded by
recorded UTCs
by using:Platform,
using: an an UTCs equipped
UTCs Platform, with a
Platform, equipped caudal
equipped with fin in order
with aa caudal to reduce
caudal fin
fin in pitch
in order
order toand roll
to reduce movements
reduce pitch
pitch and and
and roll
roll
stabilize
movements
movements video and
andacquisition,
stabilizea video
stabilize SeaViewer’s
video acquisition,Sea-Drop™
acquisition, 6000 high-definition
aa SeaViewer’s
SeaViewer’s Sea-Drop™
Sea-Drop™ underwater video camera,
6000 high-definition
6000 high-definition
with a surface
underwater
underwater console
video
video and two
camera,
camera, withGoPro
with Hero
aa surface
surface 3+ cameras
console
console and two
and (Figure
two GoPro
GoPro 6b),Hero
with3+
Hero a 12
3+ MP HD
cameras
cameras CMOS6b),
(Figure
(Figure sensor
6b), withf
with
2.8–170 ◦ [66] (Figure 6b).
aa 12
12 MP
MP HDHD CMOS
CMOS sensorsensor ff 2.8–170°
2.8–170° [66]
[66] (Figure
(Figure 6b).6b).

Figure 6.
Figure 6. (a)
(a) Autonomous
Autonomous Surface
Surface Vehicle
SurfaceVehicle called DEvelopment
Vehicle called DEvelopment Vehicle
Vehicle for Scientific Survey
for Scientific Survey (DEVSS);
(DEVSS);
(b) Underwater
(b) Underwater Towed
TowedVideo
Towed VideoCamera
Video CameraSystems
Camera Systems(UTCS).
Systems (UTCS).
(UTCS).

The combined
The combined sampling
sampling of
of acoustic
acoustic and
and optical
optical data
data was
was carried
carried out
out in
in the
the shallow
shallow water
water coastal
coastal
area (depth
area (depth <10
<10 m)
m) that
that surrounds
surrounds Cirella
Cirella Island.
Island. In
In order
order to
to acquire
acquire overlapping
overlapping pictures,
pictures, ensuring
ensuring
 J. Mar. Sci. Eng. 2020, 8, 647 9 of 25

J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 9 of 25

The combined sampling of acoustic and optical data was carried out in the shallow water
about 75% area
coastal of shared < 10 m) that
(depth coverage between
surrounds two consecutive
Cirella Island. Inphotos,
order toaacquire
speed overlapping
of about 1 pictures,
knot/h was
ensuringA
maintained. about
total75% of shared
of about 2500coverage between two
georeferenced consecutive
images photos, a Before
were collected. speed ofperforming
about 1 knot/h
the 3D
was maintained. A total of about 2500 georeferenced images were collected.
processing, an underwater image enhancement technique [67] was performed to minimize the effect Before performing
of thethe 3D processing,
water column onanthe underwater
underwater image enhancement
images. After thetechnique [67] was performed
image enhancement step, a to minimize
Structure-from-
the effect of the water column on the underwater images. After the image enhancement step,
motion (SfM) 3D reconstruction was performed using the commercial software Agisoft Metashape
a Structure-from-motion (SfM) 3D reconstruction was performed using the commercial software
Pro and Pix4D mapper [68]. The average root-mean-square error (RMSE) achieved at this step was
Agisoft Metashape Pro and Pix4D mapper [68]. The average root-mean-square error (RMSE) achieved
0.022atmthis
forstep
GPSwascoordinates. Finally, a Multiview Stereo (MVS) algorithm was used by the Metashape
0.022 m for GPS coordinates. Finally, a Multiview Stereo (MVS) algorithm was used by
software to produce a dense
the Metashape software to produce 3D point cloud
a dense from cloud
3D point the refined
from theintrinsic orientation
refined intrinsic and ground-
orientation and
referenced camera exterior orientation (Figure 7). In particular, in the very shallow
ground-referenced camera exterior orientation (Figure 7). In particular, in the very shallow water area, water area, 32
sampling areas were
32 sampling areasacquired with Autonomous
were acquired with Autonomous Surface Vehicle
Surface DEVSS
Vehicle DEVSS and 5959
and sampling
samplingareas
areaswere
acquired with UTCS
were acquired withinUTCS
the shallow waterwater
in the shallow area,area,
Table 1 [64].
Table The
1 [64]. The3D3Dreconstructions generated by
reconstructions generated
DEVSSby DEVSS
and UTCS and UTCS were used
were used in order
in order to: interpret
to: (a) (a) interpret
thethe bathymetricDEM
bathymetric DEMand and backscatter
backscatter and
and (b)
(b) identify further ground-truth
identify further ground-truth data. data.

Figure 7. (a)
Figure 7. Sample
(a) Sampleimage
image before
beforeandand (b) after
afterthe
theapplication
application of the
of the imageimage enhancement
enhancement algorithm
algorithm on
a singleframe
on a single frameof of the
the underwater
underwater photogrammetric
photogrammetric survey. Orthogonal
survey. Orthogonalview view
of theof
3Dthepoint
3Dclouds of
point clouds
three sample areas: (c) area size: 35 m × 7 m, 13 million 3D points; (d) area size: 40 m
of three sample areas: (c) area size: 35 m × 7 m, 13 million 3D points; (d) area size: 40 m × 7 m, 15 × 7 m, 15 million
3D points;
million (e) area
3D points; (e)size:
areaarea size:
size: area m × 933
33size: m,m 14×million
9 m, 143Dmillion
points. 3D
Thepoints.
map between (c) and
The map (d) shows
between (c) and
the location of the submerged transects with respect to Cirella Island.
(d) shows the location of the submerged transects with respect to Cirella Island.
In general, all the ground-truth data used in this work were carefully checked for possible overlaps
In general,
between all the and
the training ground-truth data used
validation samples. in this work
The minimum were
distance, carefully
calculated checked
between any for possible
training
overlaps between samples,
and validation the training and validation
was about samples.
25 m and this The 10%
affects only minimum distance,
of the total dataset.calculated between
any training and validation samples, was about 25 m and this affects only 10% of the total dataset.
2.6. OBIA Segmentation and Classification
2.6. OBIAAll
Segmentation and
collected data, Classification
i.e., DEMs, backscatter intensity map, UAV data and multispectral images data,
were processed in the OBIA process [69] by eCognition Essentials 1.3. software, using a classification
All collected data, i.e., DEMs, backscatter intensity map, UAV data and multispectral images
algorithm [70] (Figure 8). The Multibeam bathymetry data were converted into secondary features:
data, were processed in the OBIA process [69] by eCognition Essentials 1.3. software, using a
Slope, Northness, Eastness, Curvature and Terrain Roughness Index (TRI) using the Morphometry
classification algorithm [70] (Figure 8). The Multibeam bathymetry data were converted into
secondary features: Slope, Northness, Eastness, Curvature and Terrain Roughness Index (TRI) using
the Morphometry Tool in SAGA (SAGA (System for Automated Geoscientific Analyses) Version) [71]
as shown in (Table 2 and Figure 9).
 J. Mar. Sci. Eng. 2020, 8, 647 10 of 25

Tool in SAGA (SAGA (System for Automated Geoscientific Analyses) Version) [71] as shown in (Table 2
J. Mar. Sci.and
Eng.Figure 9).x FOR PEER REVIEW
2020, 8, 10 of 25

Figure 8.Figure 8. Workflow

Workflow of ofthe
the processing
processing steps developed
steps in this study
developed in to create
this the seabed
study map classification
to create the seabed map
of Cirella Island marine habitats (see Figure 10).
classification of Cirella Island marine habitats (see Figure 10).
Table 2. Multidata source variable and list of extracted secondary features of the bathymetric and
Table 2.backscatter
Multidata data.
source variable and list of extracted secondary features of the bathymetric and
backscatterSource
data. Features Resolution Software Area

Source Multibeam Features

Backscatter/Bathymetry 0.3Resolution
m
Caris HIPSSoftware
and
shallow Area
EM2040 SIPS
Multibeam
Bathymetry Backscatter/Bathymetry
Curvature General 0.5 m 0.3 m Caris HIPS and
SAGA-GIS shallow shallow
EM2040Bathymetry Curvature Total 0.5 m SAGA-GIS SIPS shallow
Bathymetry Curvature
Bathymetry SlopeGeneral 0.5 m 0.5 m SAGA-GIS
SAGA-GISshallow shallow
Bathymetry Aspect 0.5 m SAGA-GIS shallow
Bathymetry
Bathymetry
Curvature
Northness
Total 0.5 m
0.5 m SAGA-GISshallow
SAGA-GIS
shallow
Bathymetry
Bathymetry Slope
Eastness 0.5 m 0.5 m SAGA-GISshallow
SAGA-GIS shallow
Bathymetry Terrain
Aspect 0.5 m SAGA-GIS shallow
Bathymetry Ruggedness Index 0.5 m SAGA-GIS shallow
Bathymetry Northness
(TRI)
0.5 m SAGA-GIS shallow
BathymetryPléiades Eastness
Satellite image 2 m 0.5 m ERDAS IMAGINE
SAGA-GISshallow shallow
Terrain Ruggedness AGISOFT
BathymetryUTCS Image Truth data Index 0.001 m0.5 m SAGA-GISshallow shallow
METASHAPE
Parrot Anafi
(TRI) very
Orthophoto 0.02 m
Pléiades Work Satellite image 2 m PIX4D Mapper
ERDAS IMAGINE
shallow shallow
AGISOFT AGISOFT very
UTCSASVs (DEVSS) ImageImage Truth data
Truth data 0.001 m
0.001 m METASHAPE shallow shallow
METASHAPE
Parrot Anafi very
Orthophoto 0.02 m PIX4D Mapper
Work shallow
AGISOFT very
ASVs (DEVSS) Image Truth data 0.001 m
METASHAPE shallow
J. Mar. Sci. Eng. 2020, 8, 647 11 of 25
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 11 of 25

Figure 9.
Figure 9. Bathymetry
Bathymetry data
data products
products used
used to
to classify
classify the seabed morphologies
the seabed morphologies and acoustic facies:
and acoustic facies:
Multibeam Digital
Multibeam Digital Elevation
Elevation Model
Model (DEM),
(DEM), backscatter
backscatter intensity
intensity map
map and
and secondary
secondary features:
features: Terrain
Terrain
RoughnessIndex
Roughness Index(TRI),
(TRI),Aspect,
Aspect, Curvature
Curvature and
and Slope
Slope obtained
obtained fromfrom postprocessing
postprocessing of bathymetric
of bathymetric data.
data.
The multiresolution segmentation algorithm, performed by the eCognition Essential software,
The multiresolution
was used to identify homogeneoussegmentation objects.algorithm,
The process performed by the eCognition
of multiresolution Essential
segmentation was software,
carried
wasby
out used to identifythe
considering homogeneous objects. The Scale
following parameters: process of multiresolution
Factor, Shape, Smoothness segmentation was carried
and Compactness.
outshallow
In by considering
water, thethe following parameters:
multiresolution segmentation Scalealgorithm
Factor, Shape,
was used Smoothness
to generate and Compactness.
objects with similar In
shallow water,
information by the multiresolution
using segmentation
only the backscatter intensity algorithm
and TRIwas used to
features. The generate objects
bathymetry, with Aspect
Slope, similar
information
and Curvature by features
using only theexcluded
were backscatter fromintensity
this first and TRI features.procedure.
segmentation The bathymetry, Slope, Aspect
The multiresolution
and Curvaturealgorithm
segmentation features were excluded
was used from this
to generate first segmentation
objects procedure. The
with similar information by multiresolution
using only the
segmentation
most important algorithm
featureswas used toThis
selected. generate objects
decision waswith similar
taken since information
the use ofby allusing only(primary
features the most
important
and secondary)featureshadselected.
created an This decisiondisturbance
excessive was takeneffect sinceand the the
usesegmentation
of all features (primary
results were and not
secondary) had created an excessive disturbance effect and the segmentation
adapted to the real shapes of the objects. In order to identify and remove nonimportant features results were not adapted
to theall
from real
the shapes
inputoflayers,
the objects. In order
a feature to identify
selection algorithm and (i.e.,
remove the nonimportant
R package Boruta features from allwas
algorithm) the
inputtolayers,
used assessatheir
feature selection
relevance algorithm (i.e.,
in determining thethe R package
thematic Boruta
classes. Thealgorithm)
Boruta algorithmwas used to assess
is built on a
their relevance
“random” forest in determining
classification the thematic
algorithm. Withclasses.
the wrapper The Boruta
algorithm,algorithm
presentisinbuilt on a “random”
the Boruta package,
forest
we classification
selected algorithm.
all the relevant With the
features wrapper algorithm,
by comparing present of
the importance in the Boruta
originalpackage,
attributes wewith
selected
the
all the relevant features by comparing the importance of the original
importance reached in a random way, estimated using permutations (i.e., shadows). Only the features attributes with the importance
reachedimportance
whose in a random wasway,
higher estimated
than those using permutations
of the randomized(i.e., shadows).
features Only the features
were considered important whose
[72].
importance
The was higherresult
Boruta algorithm thangave
thosea oflistthe
of randomized
attributes accordingfeatures to were
theirconsidered
importance, important
separating [72].them
The
Boruta algorithm result gave a list of attributes according to their importance,
into “confirmed” or “rejected” features. The features without decision at the end of the analysis were separating them into
“confirmed”
classified or “rejected” In
as “provisional”. features. The features
the following without
classification step,decision
we usedatthe the10 end
mostofimportant
the analysis were
attributes
classified asas
considered “provisional”.
“confirmed” In bythethefollowing classification
Boruta algorithm. step, wethe
Therefore, used the 10
Boruta most
test important
allows attributes
identification of
considered
the as “confirmed”
most important by the Boruta
and nonredundant algorithm.
features to train Therefore,
a model the Boruta test
improving allows identification
the learning process timing of
the most
and accuracyimportant and classification
of the final nonredundantmap features
(Figure to 8).
trainThea orthophotos
model improving the learning
generated from UAV process
data
timing and accuracy
surrounding of the final
the area around classification
the Cirella Island map
(seabed (Figure
depth8).from The0orthophotos
to 12.5 m) were generated
treated from UAV
separately
data
by thesurrounding the area around
eCognition software due to the thelackCirella Island
of data (seabed
acquired fromdepth from 0 surveys
multibeam to 12.5 m) wereshallow
in very treated
separately
water. by thetwo
Therefore, eCognition
different software due tothe
projects within theeCognition
lack of data acquired
Essential from multibeam
software were created. surveys in
The first
very shallow water. Therefore, two different projects within the
included all data acquired by multibeam in shallow water areas (8–30 m) and the second included eCognition Essential software were
created.
only The first included
the orthophotos acquiredall data
by UAV acquired
in very byshallow
multibeam waterinareas
shallow(1–12.5waterm; areas
Table (8–30
1). Wem) and the
tested
second included
performance only
of the the orthophotos
three acquiredclassification
different supervised by UAV in very shallow water
algorithms areas
available in(1–12.5 m; Table
the eCognition
1). We tested
Essential the performance
software: (k-NN), (RT) of and
the three
(DT).different
The k-NN supervised
algorithmclassification
is a method algorithms
for classifying available
objectsinby the
a
eCognition
majority Essential
ranking software:
of its (k-NN), (RT)
neighbors—with theand
object(DT).
beingTheassigned
k-NN algorithm
to the class is a most
method for classifying
common among
objects
its by a majority
k-Nearest ranking
neighbor. of its neighbors—with
DT learning is a method commonly the object being usedassigned to the class
in data mining wheremost a common
series of
among its k-Nearest neighbor. DT learning is a method commonly used in data mining where a series
of decisions are made to segment the data into homogeneous subgroups through a tree system with
J. Mar. Sci. Eng. 2020, 8, 647 12 of 25

decisions are made to segment the data into homogeneous subgroups through a tree system with
branches. Random tree is a combination of tree predictors where each tree depends on the values of a
random vector, sampled independently and with the same distribution for all trees in the forest [73].
All trees are trained with the same features but on different training sets, which are generated from the
original training set. It aggregates the ranks from different decision trees in order to assign the final
class of the test object. The ground-truth data acquired with the direct interpretation of multibeam data,
by means of the ASV DEVSS (both images and 3D point clouds) and the UTCS platform were used to
assign specific class values to the segmented objects. In the sampled area, five different classes were
identified: Fine sediment-FS, Coarse sediment-CS, Rock-R, P. oceanica meadows-P and Cystoseira-Cy.
Some samples were identified through other techniques in order to have homogeneous ground-truth
data coverage. Along the perimeter of the island, in the very shallow water areas, ground-truth data
were taken directly via visual identification during UAV survey, while for deeper areas that could not
be reached by the ASV DEVSS, some ground-truth data were selected based on the study of multibeam
bathymetry, backscatter and water column data acquired during the multibeam survey and by direct
observation with the UTCS platform. Overall, 920 ground-truth data were used in this study and
divided into two different groups (Training set and Validation set): one to perform the training step and
the second to evaluate the classification accuracy and Kappa coefficients [74–76] (Table 3 and Figure 3).

Table 3. Main thematic classes of seabed and number of ground-truth data collected through the ASV
DEVVS, UTCS, UAVs and derived by the acoustic data.

Shallow Water Very Shallow Water

Class Training Set Validation Set Training Set Validation Set
P. oceanica (P) 123 122 23 26
Rock (R) 80 15 35 21
Fine sediment (FS) 207 170 \ \
Coarse sediment (CS) 44 5 \ \
Cystoseira (Cy) \ \ 31 18
Total 454 312 89 65

3. Results
Multibeam and photogrammetric (UAV) DEMs were merged via Global Mapper software using
the acoustic data and point cloud raw data respectively. The cloud points obtained by Pix4D software
were georeferenced by comparing the position of the Ground Control Points detected along the
coastline and on the island, while the bathymetric data collected with multibeam were used in order to
correct the altitude in the marine area. This operation was carried out through the tools for Quality
Control of LiDAR point cloud data LiDAR module of the Global Mapper software allowing comparison
and/or 3D correction of the height of point cloud data to known control points, and to report statistics
about subsets of points.
The acoustic profiles and point clouds were gridded (0.3 m) generating an integrated digital land
and sea model, overall the merged soundings and cloud points highlighted a vertical subdecimetrical
accuracy, see Figure 10.
J. Mar. Sci. Eng. 2020, 8, 647 13 of 25
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 13 of 25

Figure 10. Digital

Digital Terrain and Marine Model (DTMM) generated
generated by
by the
the optical
optical (UAV) and acoustic
(multibeam) data fusion; (a) Area of Interest (AOI) in shallow water; (b) Area of Interest (AOI) in very
shallow water. The white boxes are shown in Figure 11.

In the shallow water areas, the most relevant results were obtained by using the following
multiresolution segmentation
multiresolution segmentation parameters:
parameters: scale
scale segmentation
segmentation 20;
20; Shape
Shape 0.1,
0.1, compactness
compactness 0.5 and
scale slider min. value
value 10, scale slider max value 100. These parameters
parameters were selected after several
tests carried out with different settings comparing the segmentation objects created by the algorithm
with those of real morphology.
morphology. Figure 10 shows the two Areas of Interest (AOI) from which we have
extracted the two examples of segmentation and classification processes. The The results
results are shown in
Figure 11.
The acquired ground-truth data were used to assign specific class values to determine segmented
objects (Figure 11).
J. Mar. Sci. Eng. 2020, 8, 647 14 of 25
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 14 of 25

Figure 11.
Figure 11. Shallow
Shallow water
water classification
classification process
process from
from multibeam
multibeam data:
data: (a)
(a) backscatter
backscatter intensity
intensity map;
map;
(a’) multiresolution segmented objects image and training and validation set; (a’’) RT classified
(a’) multiresolution segmented objects image and training and validation set; (a”) RT classified map: map:
(green: P.
(green: oceanica, light
P. oceanica, light brown:
brown: Fine
Fine sediment,
sediment, gray:
gray: Coarse
Coarse sediment);
sediment); (a’’’)
(a”’) RT final classified
RT final classified map
map
after merging and smoothing objects. Very shallow water classification process
after merging and smoothing objects. Very shallow water classification process from UAV aerial from UAV aerial
photogrammetry: (b)
photogrammetry: (b) Orthomosaic
Orthomosaic UAV image; (b’)
UAV image; (b’) Orthomosaic
Orthomosaic segmented
segmented into
into image
image objects
objects and
and
training and
training and validation
validation set;
set; (b”)
(b’’) k-NN
k-NN classified
classified map:
map: green:
green: P.
P. oceanica,
oceanica, blue:
blue: Cystoseira,
Cystoseira, pale
pale brown:
brown:
Rock); (b”’)
Rock); (b’’’) k-NN
k-NN final
final classified
classified map
map after
after merging
merging andand smoothing
smoothing objects.
objects. See
See Figure
Figure 1010 for
for location.
location.

Thestatistical
All acquiredinformation
ground-truth aboutdata were used
the features to assign
(primary specific class
and secondary), valueswith
associated to the
determine
objects
segmentedfrom
classified objects
766(Figure 11). data (training and validation set) were extracted from eCognition
ground-truth
All statistical
Essential software and information
analyzedabout featuresR (primary
with the Boruta and secondary),
package (version associated
6.0.0) in order withtheir
to assess the
objects classified from 766 ground-truth data (training and validation set) were
importance in the assignment of the corresponding class (Figure 12) [77]. The classification tests extracted from
eCognitionwith
performed Essential
a highersoftware
than 10and analyzed
number with theused
of attributes Boruta R package
in the (version
training phase did6.0.0) in order to
not significantly
assess their
increase importance
the accuracy of theinclassification.
the assignment of the
The most corresponding
relevant class (Figure
features emerging from the12)Boruta
[77]. test,
The
classification
are testsinperformed
the following with a higher
order of importance: the than
mean10 number ofintensity,
backscatter attributes
theused
meaninAspect,
the training phase
the general
did not significantly
Curvature increase themean
standard deviation, accuracy of the
Pléiades Bluclassification.
band, the meanThe most relevant
of Total features
Curvature, theemerging
mean of
from the
Slope, theBoruta
Aspect test, are the
standard following
deviation, theinSlope
order of importance:
standard deviation,theTRI
mean backscatter
standard intensity,
deviation; the
the mean
mean
of Aspect,
Eastness. the general
Figure 12 showsCurvature standard
the results deviation,
of the Boruta test.mean Pléiades Blu band, the mean of Total
Curvature, the mean of Slope, the Aspect standard deviation, the Slope standard deviation, TRI
standard deviation; the mean of Eastness. Figure 12 shows the results of the Boruta test.
J. Mar. Sci. Eng. 2020, 8, 647 15 of 25
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 15 of 25

Figure 12.
Figure 12. Results
Results of
of the
the Boruta
Boruta feature
feature selection
selection algorithm.
algorithm. Blue
Blue boxplots
boxplots correspond
correspond to minimal,
to minimal,
and maximum
average and maximum Z Z score
score of
of aa shadow
shadow attribute.
attribute. M: mean; SD:
SD: standard
standard deviation.
deviation.

The comparison between the different different classification

classification algorithms
algorithms available in the the eCognition
eCognition
Essential
Essential software:
software:(k-NN),
(k-NN), (RT)(RT)
and (DT)
and was(DT)performed considering
was performed three levelsthree
considering of data combination:
levels of data
(A) Pléiades image;
combination: (B) Pléiades
(A) Pléiades image
image; (B) combination
Pléiades image with backscatter with
combination and bathymetry
backscatterdata;
and (C) Pléiades
bathymetry
image combination
data; (C) with backscatter,
Pléiades image combinationbathymetry data and
with backscatter, secondary data
bathymetry features,
and (Table 4C). The
secondary best
features,
classification
(Table 4C). The performance was obtained
best classification with the
performance was multilayer combination
obtained with (Table combination
the multilayer 4C). The algorithm
(Table
4C). generally
(RT) The algorithm (RT) the
provided generally provided
best overall the best
accuracy overall
results in twoaccuracy
levels ofresults in two levels of
data combinations (B data
and
combinations
C). (B and
In particular, C). In particular,
the algorithm the algorithm
(RT), compared (RT),
to the k-NNcompared
and RTto the k-NN and
algorithms, RT algorithms,
showed an overall
showed an
accuracy overall and
of 99.63% accuracy of 99.63%
the Kappa and the
coefficient 0.99,Kappa coefficient
the Producer’s 0.99, the
accuracy Producer’s
and accuracy
User’s accuracy and
values
User’s
with accuracy
a low values with
gap between thema(Table
low gap 4C).between
The k-NN them (Table returned
algorithm 4C). The overall
k-NN algorithm returned
accuracy values of
overall and
86.94% accuracy valuescoefficient
the Kappa of 86.94%0,80,and the
whileKappa
the DT coefficient
learning,0,80, whileanthe
showed DT learning,
overall accuracyshowed
of 88.57%an
overall
and the accuracy of 88.57%
Kappa coefficient 0.77and the 4C).
(Table Kappa Thecoefficient 0.77 (Table
k-NN algorithm 4C).
did not Thethe
show k-NN
bestalgorithm
response in did not
terms
show
of the best response
classification (Table 4),inwhereas
terms of theclassification
DT learning (Table 4), whereas
showed an overalltheaccuracy
DT learning showed
always loweranthan
overall
RT
accuracy
(Table always
4). As shown lower than4,RT
in Table in (Table
the RT 4). As shown
all classes had in Table 4, in
a relatively the RT all
consistent classes as
accuracy, had a relatively
shown by the
consistent
small accuracy,
differences as shown
between User’sbyandthe Producer’s
small differences between
Accuracy for eachUser’s and Producer’s
thematic map class. Accuracy for
Results from
eachDT
the thematic
approach map class. the
showed Results from the
divergence ofDTtheapproach
Producer’s showed the divergence
and User’s of the Producer’s
accuracy increases and
considerably
User’s 4).
(Table accuracy increases considerably (Table 4).
J. Mar. Sci. Eng. 2020, 8, 647 16 of 25

Table 4. Accuracy assessment for the DT, k-NN and RT supervised classification algorithm (shallow
water area) for 3 different combinations: (A) Pléiades image; (B) Pléiades image-Backscatter-Bathymetry;
(C) Pléiades image-Backscatter-Bathymetry-Secondary features.

Combinations
Decision Tree (DT) Random Tree (RT) k-NN
(Data Source)
Overall accuracy: Overall accuracy: Overall accuracy:
67.83% 66.78% 71.33%
K: 0.48 K: 0.47 K: 0.48
A User’s Producer’s User’s Producer’s User’s Producer’s
Class
Pléiades image accuracy accuracy accuracy accuracy accuracy accuracy
(P) 75.36% 85.25% 72.67% 89.34% 70.31% 73.77%
(R) 83.33% 33.33% 83.33% 33.33% 100.00% 33.33%
(FS) 84.21% 55.56% 87.80% 50.00% 75.35% 74.31%
(CS) 10.64% 100.00% 10.42% 100.00% 18.18% 40.00%
Overall accuracy: Overall accuracy: Overall accuracy:
83.61% 91.80% 82.38%
K: 0.73 K: 0.85 K: 0.69
B
User’s Producer’s User’s Producer’s User’s Producer’s
Pléiades-Backscatter Class
accuracy accuracy accuracy accuracy accuracy accuracy
Bathymetry
(P) 95.45% 77.78% 96.97% 88.89% 90.22% 76.85%
(R) 28.57% 80.00% 42.11% 80.00% 42.11% 80.00%
(FS) 100.00% 88.43% 100.00% 95.04% 89.92% 88.43%
(CS) 23.81% 100.00% 45.45% 100.00% 21.43% 60.00%
Overall accuracy: Overall accuracy: Overall accuracy:
88.57% 99.63% 86.94%
C K: 0.80 K: 0.99 K: 0.77
Pléiades-Backscatter User’s Producer’s User’s Producer’s User’s Producer’s
Class
Bathymetry accuracy accuracy accuracy accuracy accuracy accuracy
Secondary features (P) 94.95% 86.24% 100.00% 99.07% 94.95% 86.24%
(R) 43.75% 70.00% 100.00% 100.00% 43.75% 70.00%
(FS) 94.74% 89.26% 99.31% 100.00% 94.74% 89.26%
(CS) 25.00% 80.00% 100.00% 100.00% 25.00% 80.00%

In the very shallow water area, the same workflow was performed to generate the classified habitat
map (Figure 8). The UAV orthophoto was imported via the eCognition software and a multiresolution
segmentation algorithm in order to identify objects. The most relevant segmentation results were
obtained by using the following parameter settings: scale 300; Shape 0.1, compactness 0.5 and scale
slider min. value: 50, scale slider max. value: 600 (Figure 11). These parameters were selected after
several tests carried out with different settings comparing the shapes created by the algorithm with
those of the real morphologies of the seabed.
A greater scale was used for the orthophotos due to the higher image resolution (0.03 m).
Data available for the OBIA classification derived only from the RGB orthophoto image, therefore the
Boruta algorithm was not used. In the very shallow water area, 3 different classes were identified
(Cystoseira (Cy), Rock (R), P. oceanica meadows (P); Table 2). The supervised classification algorithms
(k-NN, RT and DT) were tested, but in this case, the best classification result in terms of accuracy
was obtained when using the k-NN algorithm, with an overall accuracy equal to 95.24% and Kappa
coefficient 0.92 (Table 5). With regard to the k-NN classification algorithm, all classes (Cystoseira,
Rock and P. oceanica) showed small differences between the Producer’s and User’s accuracy values
(Table 5).
 mapping, has nevertheless obtained an excellent result compared to the DT and RT algorithms. The
OBIA object classification, as highlighted in existing literature [45], represents an effective tool to
obtain robust thematic maps.
The Boruta feature selection algorithm showed promising results on seagrass habitat mapping
J.[23,66].
Mar. Sci.The
Eng. 2020, 8, 647
results confirmed the usefulness of applying this feature selection method in seagrass 17 of 25

habitats mapping [47]. The multiresolution segmentation scale requires careful adjustment of the
value and5.represents
Table a very important
Accuracy assessment setting
for the DT, for correct
k-Nearest classification
Neighbors’ algorithmby OBIA
(k-NN) and
and RThas a significant
supervised
impact on the results of seagrass habitat classification
classification algorithms (very shallow water area). [48]. Multiresolution segmentation represents
a very important step in the whole process, so defining the wide range of parameters, first of all, the
scale factor, the shape,Decision
the Tree (DT)
Smoothness Random Tree
and Compactness, is(RT)
of paramount importance k-NNsince they can
Overall Accuracy: 74.6% Overall Accuracy: 77.78% Overall Accuracy: 95.24%
determine the good progress of the classification. In this work, even if only five thematic classes have
K: 0.61 K: 0.65 K: 0.92
been used (Fine sediment (FS), Coarse sediment (CS), Rock (R) P. oceanica meadows (P) and
User’s Producer’s User’s Producer’s User’s Producer’s
Cystoseira
Class (Cy), Figure 13), it has been necessary to increase the number of ground-truth data for
accuracy accuracy accuracy accuracy accuracy accuracy
training(P) and validation
71% in order to improve the
65.38% answers of the
66.70% 100%classification algorithms.
100.00% In order to
100.00%
obtain(R)a relatively 62.50%high number 75%of ground-truth
88% data, different
35.00% sampling 87.00% platforms have been
100.00%
(Cy) and 100.00%
integrated used, such as88.24% 100% ASV, UAV,
acquisitions with 94.12%UTCS and 100.00% 82%
direct observation and
interpretation of the backscatter and DEM data, obtained by the Multibeam acquisition (Figure 9).
ThisTheintegrated technique could
two best-classified be useful
habitat to reduce
maps (very the time
shallow and and costsmaps)
shallow for theselected
acquisition
on theof ground-
basis of
truth data.
accuracy were merged into one very high-resolution classified map (Figures 13 and 14).

Figure
Figure13.13. Comparison
Comparisonbetween
betweenseafloor
seafloormapping
mapping techniques
techniques using
usingoptical GoPro
optical images
GoPro collected
images by
collected
(UTCS) and Autonomous Surface Vehicle (DEVSS) and: (a–d) high-resolution
by (UTCS) and Autonomous Surface Vehicle (DEVSS) and: (a,b,c,d) high-resolution multibeam multibeam bathymetry,
(a’,b’,c’,d’)
bathymetry, backscatter intensity
(a’,b’,c’,d’) data, (a”,b”,c”,d”)
backscatter intensity data,multispectral reflectance
(a’’,b’’,c’’,d’’) of Pléiades
multispectral images.
reflectance ofThe data
Pléiades
collected in the
images. The datastudy area show
collected in thedifferent types
study area of seafloor:
show differentsediments (from fine
types of seafloor: to coarse),
sediments rocks
(from and
fine to
P.coarse),
oceanica.rocks
The GoPro
and P.images wereThe
oceanica. usedGoPro
to interpret
images andwere
calibrate
usedthe tomorphological
interpret andfeatures
calibratewhile
the
using the eCognition
morphological software.
features See Figure
while using 14 for location.
the eCognition software. See Figure 14 for location.

More specifically,
The accuracy of about hectares of P.oceanica
habitat51classification might bewere mapped,
affected by while 119 factors,
multiple hectaresincluding
and 1.70 hectares
habitat
were, respectively, identified for the Fine and Coarse sediment, and 0.45 hectares
heterogeneity, bathymetry and water column attenuation [88]. This study highlights for Cystoseira.
how the
integration of satellite imagery with ultrahigh spatial resolution UAV aerial imagery and bathymetric
4. Discussion
The main objective of this work was to perform benthic habitat mapping by using several data
acquisition platforms and OBIA algorithms and to develop a high-resolution seagrass mapping digital
elevation model. OBIA algorithms are now stable and powerful approaches to use for classifying
J. Mar. Sci. Eng. 2020, 8, 647 18 of 25

benthic habitats and to produce accurate maps [78–82]. In this work, an integrated methodological
approach has been followed for multisensor data fusion (acoustic and optical) with different degrees
of resolution. This approach might be useful, if necessary, for mapping the P. oceanica meadows,
but also seabed geomorphological features and furthermore, to estimate the carbon sequestration
and storage of the seagrass ecosystem [83]. Generally, surveying techniques are used individually,
and show several limitations such as spatial coverage in very shallow waters (e.g., Multibeam) or poor
resolving capacity (e.g., satellite images) that might not allow a complete characterization of the spatial
and temporal extent of P. oceanica meadows in deeper areas. Most benthic seagrass habitat mapping
studies examine a single data source and quite a few attempts have been made to combine multiple
data sources [84,85]. High spatial resolution satellite imagery (2 m or smaller) alone has proven to
be unable to produce adequate accuracy for fine descriptive detailed maps [51] (a fact that is also
confirmed by this study). High-resolution multispectral and hyperspectral imagery can be useful
in discriminating habitat community size to a not-so-fine detail, but its moderate spatial resolution
might limit its broader application especially as the depth of the seabed increases, and reflectance
and radiance are absorbed [86]. The present study highlights how the OBIA classification did not
provide a satisfactory result in terms of thematic accuracy by using only the Pléiades satellite images.
Indeed, satellite images can be effective for the mapping of P. oceanica meadows, only in conditions of
high water transparency and in the presence of excellent spatial and spectral resolution. However,
the combination of acoustic bathymetric data (DEM and backscatter) combined with optical data (e.g.,
multispectral satellite) has proven to improve the final classification performance. The set of optical and
bathymetric acoustic data combined with secondary features showed the best classification results and
this has also been confirmed by other studies [14,35,87]. The present work, specific on high-resolution
seagrass mapping, highlights how RT seems to be the OBIA algorithm with the best classification
performance. Similar answers on RT have been highlighted also by Janowski [47] on the automatic
classification of benthic habitats [78,80,81]. As far as the K-NN classifier is concerned, it has a lower
case history of application in marine habitat mapping studies [79], and in general, the performances of
the K-NN classifier were almost always moderate or fair. The RT algorithm in this study proved to be
very effective in generating accurate classification, thus showing fair performance. Instead, the DT
classification algorithm has always shown the lowest accuracy. The classification of the orthophoto
produced with the UAV showed instead the best accuracy with the k-NN algorithm, which, although
not much used for marine habitat mapping, has nevertheless obtained an excellent result compared to
the DT and RT algorithms. The OBIA object classification, as highlighted in existing literature [45],
represents an effective tool to obtain robust thematic maps.
The Boruta feature selection algorithm showed promising results on seagrass habitat
mapping [23,66]. The results confirmed the usefulness of applying this feature selection method in
seagrass habitats mapping [47]. The multiresolution segmentation scale requires careful adjustment of
the value and represents a very important setting for correct classification by OBIA and has a significant
impact on the results of seagrass habitat classification [48]. Multiresolution segmentation represents
a very important step in the whole process, so defining the wide range of parameters, first of all,
the scale factor, the shape, the Smoothness and Compactness, is of paramount importance since they
can determine the good progress of the classification. In this work, even if only five thematic classes
have been used (Fine sediment (FS), Coarse sediment (CS), Rock (R) P. oceanica meadows (P) and
Cystoseira (Cy), Figure 13), it has been necessary to increase the number of ground-truth data for
training and validation in order to improve the answers of the classification algorithms. In order to
obtain a relatively high number of ground-truth data, different sampling platforms have been integrated
and used, such as acquisitions with ASV, UAV, UTCS and direct observation and interpretation of
the backscatter and DEM data, obtained by the Multibeam acquisition (Figure 9). This integrated
technique could be useful to reduce the time and costs for the acquisition of ground-truth data.
The accuracy of habitat classification might be affected by multiple factors, including habitat
heterogeneity, bathymetry and water column attenuation [88]. This study highlights how the integration
J. Mar. Sci. Eng. 2020, 8, 647 19 of 25
J. Mar. Sci. Eng. 2020, 8, x FOR PEER REVIEW 19 of 25

of satellite
acoustic imagery
data showswith ultrahigh
a good spatial resolution
performance UAV aerial imagery
in habitat classification and
[24]. The bathymetric
integrated acoustic
multisource
data shows a good performance in habitat classification [24]. The integrated multisource
technique represents, indeed, an improved solution to map benthic habitats with high degrees of technique
represents, indeed,Therefore
spatial accuracy. an improvedthe solution
higher theto map benthic
quality and habitats with high degrees
spatial resolution of spatial
of the data accuracy.
the better the
Therefore
performance of the segmentation algorithm and resulting OBIA classification [45,83,88]ofand
the higher the quality and spatial resolution of the data the better the performance the
segmentation
consequently the algorithm and resulting
generation of thematicOBIA
maps.classification [45,83,88]
A high-resolution and consequently
seabed classificationthemapgeneration
has been
of thematic maps. A high-resolution seabed classification map has been obtained
obtained for producing high-quality habitat-classified maps (Multilayer data and UAVs data). Two for producing
high-quality
maps (shallow habitat-classified
and deep watermaps (Multilayer
classified databeen
area) have and UAVs data).
selected basedTwoonmaps (shallow
an accuracy and deep
assessment
water classified
(shallow and deeparea) haveclassified
water been selected
area) based
and haveon an accuracy
been mergedassessment
into a unique(shallow and deephabitat
and complete water
classified area) and have
map (Figures 13 and 14). been merged into a unique and complete habitat map (Figures 13 and 14).

Figure 14.
Figure 14. Seabed
Seabed map classification of
map classification Cirella Island
of Cirella Island overlapped
overlapped over
over high-resolution
high-resolution multibeam
multibeam
bathymetry. The
bathymetry. The map
map shows
shows part
part of
of the
the analyzed areas in
analyzed areas in order
order to
to provide an example
provide an of the
example of the mapping
mapping
outcomes. The
outcomes. The inset
inset at
at the
the bottom
bottom right
right shows
shows the
the circular
circular histogram
histogram of of the wave directions
the wave directions and
and
significant wave height (Hs) plotted using the directional wave time series recorded
significant wave height (Hs) plotted using the directional wave time series recorded from 1979 to from 1979 to
2009, from NOAA wave watch III model [89] for the Mediterranean Sea at the Lat
2009, from NOAA wave watch III model [89] for the Mediterranean Sea at the Lat 41.33 and Long ◦
41.33° and Long
12.50°◦ coordinates.
12.50

At this stage, in the shallow classification, the integration of Pléiades satellite image, bathymetric
and backscatter data has shown that the contribution of the optical component of the satellite data
alone did not confer anyany clear
clear improvements
improvements in in the
the classification
classification procedure.
procedure. Therefore, for future
implementations method, the
implementations of the method, the team
team will evaluate
evaluate how
how the
the different types of satellite sensors,
sensors,
the different
different characteristics
characteristics of
of satellite
satellite images,
images, in terms of spatial and spectral resolution, can improve
the results of the multilevel classification.
classification.
Finally, the
the adopted
adopted methodological
methodological approach has proven its ability to extract information of
geomorphological
geomorphologicaldata datarelated to marine
related ecosystems
to marine from the
ecosystems coastthe
from to the deep
coast toareas,
the in a georeferenced
deep areas, in a
3D environment.
georeferenced 3DThis provides important
environment. geoinformation
This provides important on the extension of
geoinformation onthe
thethree-dimensional
extension of the
three-dimensional
coverage coverage
(area extensions and(area extensions
volumes) and volumes)
of the benthic of the benthic
macrohabitats macrohabitats
necessary necessary
for the analysis of data
for the
such asanalysis of data such as blue carbon.
blue carbon.
J. Mar. Sci. Eng. 2020, 8, 647 20 of 25

5. Conclusions
The integration of different methodological techniques (Multibeam, UAV, DEVS, UTCS and
multispectral image) represents a rapid and effective method for high-resolution seafloor and habitat
mapping from the coastline to deeper water. The geophysical and optical techniques, if correctly
processed, allow generation of high resolution integrated terrestrial and marine digital elevation
models that can also be used for the analysis of the physical environment both in the geological context
and in oceanographic modeling. Furthermore, these processed data can be used for time-lapse analysis
aimed to verify seabed changes, such as, the loss of seagrass habitats, migration of sediments (erosional
and depositional processes), as well as the anthropogenic impacts caused by fishing activities or
leisure boating.
The processing carried out from the multisensor (optical–acoustic) data fusion has significantly
improved the resolution of the mapping of the P. oceanica meadows mainly along the upper limit,
especially in shallower areas where data acquisition, performed with orthophoto UAVs image, are more
likely to be valid. The best results of the Object-based Image classification were achieved with combined
processing of DEM bathymetry, backscatter, secondary data and optical satellite data. The worst and
most inaccurate results of the Object-based classification were obtained when processing relied only
on Pléiades satellite image.
Based on the increasing use of thematic maps for habitat and the current interest in using
seagrass extension as a monitoring indicator, this digital cartographic method improves the quality
(limits of benthic facies) of the final maps, returning high-resolution products with high spatial and
thematic accuracy.
The integration of multiple acquisition methods (Satellite sensor, UAV, DEVS, UTCS and
Multibeam) allows to map the full extension of the P. oceanica meadows starting from very shallow
waters up to the lower limit; improves the performance of the cost-efficiency of the monitoring according
to the quality response of every single sensor (acoustic and/or optical), reduces the monitoring execution
time of the acoustic surveys with the Multibeam in very shallow water areas, which generally require
higher costs and more time to perform the surveys. This mapping technique may represent, within the
Marine Strategy Framework Directive (MFSD-2008/56/EC) and the EU Habitat Directive (92/43/CEE),
a valid methodology to determine the habitat extend and condition of P. oceanica meadows and to
quantify also the carbon sinks and capture rates of seagrass meadows.

Author Contributions: Conceptualization, S.F.R., A.B.; methodology, S.F.R., A.B. and R.D.M.; software, A.B.,
R.D.M. and L.D.; formal analysis, A.B., S.F.R., R.D.M., A.L. and L.D.; investigation, A.B. and R.D.M.; resources, P.L.;
data curation, A.B., S.F.R. and R.D.M.; writing—original draft preparation, S.F.R., A.B., and L.D.G.; writing—review
and editing, S.F.R.; A.B., F.B., A.D.I., M.S. and L.P., visualization, S.F.R, A.B. and R.P.; supervision, S.F.R., A.B. and
L.D.G.; project administration, S.F.R. and A.B.; funding acquisition, E.C. All authors have read and agreed to the
published version of the manuscript.
Funding: This research has been carried out within SIC CARLIT (“I SITI DI IMPORTANZA COMUNITARIA
DELLA CALABRIA “SIC MARINI”), a project financed by the ROP Calabria 2014-2020 ERDF-ESF funds by
the Department of Environment of the Calabria Region and co-financed at 50% with ARPACAL funds with the
management-organizational coordination of the CRSM—Regional Marine Strategy Centre. This research has also
been used for the development of the methodology for quantification of seagrass carbon sink in the SeaForest
LIFE 17CCM/IT 000121 project.
Acknowledgments: The Authors would like to thank Lucia Gigante and Stefano Papa (ISPRA), Salvatore Barresi
and Alfredo Amoruso (CRSM—ARPACAL) for the administrative support. We gratefully acknowledge the
support of the management board of the ASTREA Research Vessel: Giuseppe Cosentino, Luigi Manzueto,
Zeno Amicucci (ISPRA), and Giuseppe Coppola (ARGO). Last but not least the Authors appreciate the unknown
referee’s valuable and profound comments.
Conflicts of Interest: The authors declare no conflict of interest.
J. Mar. Sci. Eng. 2020, 8, 647 21 of 25

References
1. Green, E.P.; Short, F.T.; Frederick, T. World Atlas of Seagrasses; University of California Press: Berkeley, CA,
USA, 2003.
2. Den Hartog, C.; Kuo, J. Taxonomy and Biogeography of Seagrasses. In Seagrasses: Biology, Ecologyand
Conservation; Springer: Dordrecht, The Netherlands, 2007; pp. 1–23. [CrossRef]
3. EEC. Council Directive 92/43/EEC of 21 May 1992 on the conservation of natural habitats and of wild fauna
and flora. Off. J. Eur. Commun. 1992, 206, 7–50.
4. Serrano, O.; Kelleway, J.J.; Lovelock, C.; Lavery, P.S. Conservation of Blue Carbon Ecosystems for Climate
Change Mitigation and Adaptation. In Coastal Wetlands, 2nd ed.; Elsevier: Amsterdam, The Netherlands;
Oxford, UK; Cambridge, MA, USA, 2019; pp. 965–996. [CrossRef]
5. Orth, R.J.; Carruthers, T.J.B.; Dennison, W.C.; Duarte, C.M.; Fourqurean, J.W.; Heck, K.L.; Randall Hughes, A.;
Kendrick, G.A.; Kenworthy, W.J.; Olyarnik, S.; et al. A global crisis for seagrass ecosystems. Bioscience 2006,
56, 987–996. [CrossRef]
6. Duarte, L.D.S.; Machado, R.E.; Hartz, S.M.; Pillar, V.D. What saplings can tell us about forest expansion over
natural grasslands. J. Veg. Sci. 2006, 17, 799–808. [CrossRef]
7. Turner, S.J.; Hewitt, J.E.; Wilkinson, M.R.; Morrisey, D.J.; Thrush, S.F.; Cummings, V.J.; Funnell, G. Seagrass
patches and landscapes: The influence of wind-wave dynamics and hierarchical arrangements of spatial
structure on macrofaunal seagrass communities. Estuaries 1999, 22, 1016–1032. [CrossRef]
8. Lathrop, R.G.; Montesano, P.; Haag, S. A multi scale segmentation approach to mapping seagrass habitats
using airborne digital camera imagery. Photogramm. Eng. Remote Sens. 2006, 72, 665–675. [CrossRef]
9. O’Neill, J.D.; Costa, M. Mapping eelgrass (Zostera marina) in the Gulf Islands National Park Reserve of
Canada using high spatial resolution satellite and airborne imagery. Remote Sens. Environ. 2013, 133, 152–167.
[CrossRef]
10. Hogrefe, K.; Ward, D.; Donnelly, T.; Dau, N. Establishing a baseline for regional scale monitoring of eelgrass
(Zostera marina) habitat on the lower Alaska Peninsula. Remote Sens. 2014, 6, 12447–12477. [CrossRef]
11. Reshitnyk, L.; Robinson, C.L.K.; Dearden, P. Evaluation of WorldView-2 and acoustic remote sensing for
mapping benthic habitats in temperate coastal Pacific waters. Remote Sens. Environ. 2014, 153, 7–23.
[CrossRef]
12. Traganos, D.; Aggarwal, B.; Poursanidis, D.; Topouzelis, K.; Chrysoulakis, N.; Reinartz, P. Towards
Global-Scale Seagrass Mapping and Monitoring Using Sentinel-2 on Google Earth Engine: The Case Study
of the Aegean and Ionian Seas. Remote Sens. 2018, 10, 1227. [CrossRef]
13. Finkl, C.W.; Makowski, C. The Biophysical Cross-shore Classification System (BCCS): Defining Coastal
Ecological Sequences with Catena Codification to Classify Cross-shore Successions Based on Interpretation
of Satellite Imagery. J. Coast. Res. 2020, 36, 1–29. [CrossRef]
14. Hossain, M.S.; Bujang, J.S.; Zakaria, M.H.; Hashim, M. The application of remote sensing to seagrass
ecosystems: An overview and future research prospects. Int. J. Remote Sens. 2015, 36, 61–114. [CrossRef]
15. Pham, T.D.; Yokoya, N.; Bui, D.T.; Yoshino, K.; Friess, D.A. Remote Sensing Approaches for Monitoring
Mangrove Species, Structure, and Biomass: Opportunities and Challenges. Remote Sens. 2019, 11, 230.
[CrossRef]
16. Green, E.P.; Mumby, P.J.; Edwards, A.J.; Clark, C.D. Remote Sensing Handbook for Tropical Coastal Management;
Unesco: Paris, France, 2000; pp. 1–316.
17. Zoffoli, M.L.; Frouin, R.; Kampel, M. Water Column Correction for Coral Reef Studies by Remote Sensing.
Sensors 2014, 14, 16881–16931. [CrossRef] [PubMed]
18. Kenny, A.; Cato, I.; Desprez, M.; Fader, G.; Schüttenhelm, R.; Side, J. An overview of seabed-mapping
technologies in the context of marine habitat classification. ICES J. Mar. Sci. 2003, 60, 411–418. [CrossRef]
19. Brown, C.; Blondel, P. Developments in the application of multibeam sonar backscatter for seafloor habitat
mapping. Appl. Acoust. 2009, 70, 1242–1247. [CrossRef]
20. Pergent, G.; Monnier, B.; Clabaut, P.; Gascon, G.; Pergent-Martini, C.; Valette-Sansevin, A. Innovative method
for optimizing Side-Scan Sonar mapping: The blind band unveiled. Estuar. Coast. Shelf Sci. 2017, 194, 77–83.
[CrossRef]
21. Le Bas, T.; Huvenne, V. Acquisition and processing of backscatter data for habitat mapping–comparison of
multibeam and sidescan systems. Appl. Acoust. 2009, 70, 1248–1257. [CrossRef]
J. Mar. Sci. Eng. 2020, 8, 647 22 of 25

22. De Falco, G.; Tonielli, R.; Di Martino, G.; Innangi, S.; Parnum, S.; Iain, I.M. Relationships between multibeam
backscatter, sediment grain size and Posidonia oceanica seagrass distribution. Cont. Shelf Res. 2010, 30,
1941–1950. [CrossRef]
23. Lacharité, M.; Brown, C.; Gazzola, V. Multisource multibeam backscatter data: Developing a strategy for the
production of benthic habitat maps using semi-automated seafloor classification methods. Mar. Geophys. Res.
2018, 39, 307–322. [CrossRef]
24. Gumusay, M.U.; Bakirman, T.; Tuney Kizilkaya, I.; Aykut, N.O. A review of seagrass detection, mapping and
monitoring applications using acoustic systems. Eur. J. Remote Sens. 2019, 52, 1–29. [CrossRef]
25. Micallef, A.; Le Bas, T.P.; Huvenne, V.A.; Blondel, P.; Hühnerbach, V.; Deidun, A. A multi-method approach
for benthic habitat mapping of shallow coastal areas with high-resolution multibeam data. Cont. Shelf Res.
2012, 39, 14–26. [CrossRef]
26. Held, P.; Schneider von Deimling, J. New Feature Classes for Acoustic Habitat Mapping—A Multibeam
Echosounder Point Cloud Analysis for Mapping Submerged Aquatic Vegetation (SAV). Geosciences 2019,
9, 235. [CrossRef]
27. Bosman, A.; Casalbore, D.; Anzidei, M.; Muccini, F.; Carmisciano, C. The first ultra-high resolution
Marine Digital Terrain Model of the shallow-water sector around Lipari Island (Aeolian archipelago, Italy).
Ann. Geophys. 2015, 58, 1–11. [CrossRef]
28. Bosman, A.; Casalbore, D.; Romagnoli, C.; Chiocci, F. Formation of an ‘a’ā lava delta: Insights from time-lapse
multibeam bathymetry and direct observations during the Stromboli 2007 eruption. Bull. Volcanol. 2014, 76,
1–12. [CrossRef]
29. Tecchiato, S.; Collins, L.; Parnumb, I.; Stevens, A. The influence of geomorphology and sedimentary processes
on benthic habitat distribution and littoral sediment dynamics: Geraldton, Western Australia. Mar. Geol.
2015, 359, 148–162. [CrossRef]
30. Wölfl, A.C.; Snaith, H.; Amirebrahimi, S.; Devey, C.W.; Dorschel, B.; Ferrini, V.; Huvenne, V.A.I.; Jakobsson, M.;
Jencks, J.; Johnston, G.; et al. Seafloor Mapping—The Challenge of a Truly Global Ocean Bathymetry.
Front. Mar. Sci. 2019, 6, 283. [CrossRef]
31. Clarke, J.H.; Lamplugh, M.; Czotter, K. Multibeam water column imaging: Improved wreck least-depth
determination. In Proceedings of the Canadian Hydrographic Conference, Halifax, NS, Canada, 6–9 June
2006; pp. 5–9.
32. Colbo, K.; Ross, T.; Brown, C.; Weber, T. A review of oceanographic applications of water column data from
multibeam echosounders. Estuar. Coast. Shelf Sci. 2014, 145, 41–56. [CrossRef]
33. Dupré, S.; Scalabrin, C.; Grall, C.; Augustin, J.; Henry, P.; Celal Şengör, A.; Görür, N.; Namık Çağatay, M.;
Géli, L. Tectonic and sedimentary controls on widespread gas emissions in the Sea of Marmara: Results from
systematic, shipborne multibeam echo sounder water column imaging. J. Geophys. Res. Solid Earth 2015, 120,
2891–2912. [CrossRef]
34. Bosman, A.; Romagnoli, C.; Madricardo, F.; Correggiari, A.; Fogarin, S.; Trincardi, F. Short-term evolution of
Po della Pila delta lobe from high-resolution multibeam bathymetry (2013–2016). Estuar. Coast. Shelf Sci.
2020, 233, 106533. [CrossRef]
35. Doukari, M.; Batsaris, M.; Papakonstantinou, A.; Topouzelis, K. A Protocol for Aerial Survey in Coastal
Areas Using UAS. Remote Sens. 2019, 11, 1913. [CrossRef]
36. Barrell, J.; Grant, J. High-resolution, low altitude aerial photography in physical geography: A case study
characterizing eelgrass (Zostera marina L.) and blue mussel (Mytilus edulis L.) landscape mosaic structure.
Prog. Phys. Geogr. 2015, 39, 440–459. [CrossRef]
37. Duffy, J.P.; Pratt, L.; Anderson, K.; Land, P.E.; Shutler, J.D. Spatial assessment of intertidal seagrass meadows
using optical imaging systems and a lightweight drone. Estuar. Coast. Shelf Sci. 2018, 200, 169–180. [CrossRef]
38. Makri, D.; Stamatis, P.; Doukari, M.; Papakonstantinou, A.; Vasilakos, C.; Topouzelis, K. Multi-scale seagrass
mapping in satellite data and the use of UAS in accuracy assessment. In Proceedings of the Sixth International
Conference on Remote Sensing and Geoinformation of the Environment, Proc. SPIE 10773, Paphos, Cyprus,
6 August 2018. [CrossRef]
39. Nahirnick, N.K.; Reshitnyk, L.; Campbell, M.; Hessing-Lewis, M.; Costa, M.; Yakimishyn, J.; Lee, L. Mapping
with confidence; delineating seagrass habitats using Unoccupied Aerial Systems (UAS). Remote Sens. Ecol.
Conserv. 2019, 5, 121–135. [CrossRef]
J. Mar. Sci. Eng. 2020, 8, 647 23 of 25

40. Ventura, D.; Bonifazi, A.; Gravina, M.F.; Belluscio, A.; Ardizzone, G. Mapping and Classification of
Ecologically Sensitive Marine Habitats Using Unmanned Aerial Vehicle (UAV) Imagery and Object-Based
Image Analysis (OBIA). Remote Sens. 2018, 10, 1331. [CrossRef]
41. Casella, E.; Collin, A.; Harris, D.; Ferse, S.; Bejarano, S.; Parravicini, V.; Hench, J.L.; Rovere, A. Mapping coral
reefs using consumer-grade drones and structure from motion photogrammetry techniques. Coral Reefs 2017,
36, 269–275. [CrossRef]
42. Carlson, D.; Fürsterling, A.; Vesterled, L.; Skovby, M.; Pedersen, S.; Melvad, C.; Rysgaard, S. An affordable
and portable autonomous surface vehicle with obstacle avoidance for coastal ocean monitoring. Hardwarex
2019, 5, e00059. [CrossRef]
43. Alvsvåg, D. Mapping of a Seagrass Habitat in Hopavågen, Sør-Trøndelag, with the Use of an Autonomous
Surface Vehicle Combined with Optical Techniques. Master’s Thesis, NTNU, Gjøvik, Norway, 2017.
44. Blaschke, T. Object based image analysis for remote sensing. ISPRS J. Photogramm. Remote Sens. 2010, 65,
2–16. [CrossRef]
45. Diesing, M.; Mitchell, P.; Stephens, D. Image-based seabed classification: What can we learn from terrestrial
remote sensing? ICES J. Mar. Sci. 2016, 73, 2425–2441. [CrossRef]
46. Janowski, L.; T˛egowski, J.; Nowak, J. Seafloor mapping based on multibeam echosounder bathymetry and
backscatter data using Object-Based Image Analysis: A case study from the Rewal site, the Southern Baltic.
Oceanol. Hydrobiol. Stud. 2018, 47, 248–259. [CrossRef]
47. Janowski, L.; Trzcinska, K.; Tegowski, J.; Kruss, A.; Rucinska-Zjadacz, M.; Pocwiardowski, P. Nearshore
benthic habitat mapping based on multi-frequency, multibeam echosounder data using a combined
object-based approach: A case study from the Rowy Site in the Southern Baltic Sea. Remote Sens. 2018,
10, 1983. [CrossRef]
48. Wicaksono, P.; Aryaguna, P.A.; Lazuardi, W. Benthic Habitat Mapping Model and Cross Validation Using
Machine-Learning Classification Algorithms. Remote Sens. 2019, 11, 1279. [CrossRef]
49. Mohamed, H.; Nadaoka, K.; Nakamura, T. Assessment of machine learning algorithms for automatic benthic
cover monitoring and mapping using towed underwater video camera and high-resolution satellite images.
Remote Sens. 2018, 10, 773. [CrossRef]
50. Menandro, P.S.; Bastos, A.C.; Boni, G.; Ferreira, L.C.; Vieira, F.V.; Lavagnino, A.C.; Moura, R.; Diesing, M.
Reef Mapping Using Different Seabed Automatic Classification Tools. Geosciences 2020, 10, 72. [CrossRef]
51. Benfield, S.L.; Guzman, H.M.; Mair, J.M.; Young, J.A.T. Mapping the distribution of coral reefs and
associated sublittoral habitats in Pacific Panama: A comparison of optical satellite Sensors and classification
methodologies. Int. J. Remote Sens. 2007, 28, 5047–5070. [CrossRef]
52. Leon, J.; Woodroffe, C.D. Improving the synoptic mapping of coral reef geomorphology using object-based
image analysis. Int. J. Geogr. Inf. Sci. 2011, 25, 949–969. [CrossRef]
53. Phinn, S.R.; Roelfsema, C.M.; Mumby, P.J. Multi-scale, object-based image analysis for mapping geomorphic
and ecological zones on coral reefs. Int. J. Remote Sens. 2012, 33, 3768–3797. [CrossRef]
54. Wahidin, N.; Siregar, V.P.; Nababan, B.; Jaya, I.; Wouthuyzen, S. Object-based image analysis for coral reef
benthic habitat mapping with several classification algorithms. Procedia Environ. Sci. 2015, 24, 222–227.
[CrossRef]
55. Roelfsema, C.M.; Lyons, M.; Kovacs, E.M.; Maxwell, P.; Saunders, M.I.; Samper-Villarreal, J.; Phinn, S.R.
Multi-temporal mapping of seagrass cover, species and biomass: A semi-automated object based image
analysis approach. Remote Sens. Environ. 2014, 150, 172–187. [CrossRef]
56. Siregar, V.P.; Agus, S.B.; Subarno, T.; Prabowo, N.W. Mapping Shallow Waters Habitats Using OBIA
by Applying Several Approaches of Depth Invariant Index in North Kepulauan seribu. In Proceedings
of the IOP Conference Series: Earth and Environmental Science, The 4th International Symposium on
LAPAN-IPB Satellite for Food Security and Environmental Monitoring, Bogor, Indonesia, 9–11 October 2017;
IOP Publishing: Bristol, UK, 2018; Volume 149, p. 012052. [CrossRef]
57. Papakonstantinou, A.; Stamati, C.; Topouzelis, K. Comparison of True-Color and Multispectral Unmanned
Aerial Systems Imagery for Marine Habitat Mapping Using Object-Based Image Analysis. Remote Sens. 2020,
12, 554. [CrossRef]
58. Amodio-Morelli, L.; Bonardi, G.; Colonna, V.; Dietrich, D.; Giunta, G.; Ippolito, F.; Liguori, V.; Lorenzoni, S.;
Paglionico, A.; Perrone, V.; et al. L’arco Calabro-peloritano nell’Orogene Appeninico-Maghrebide.
Mem. Soc. Geol. Ital. 1976, 17, 1–60.
J. Mar. Sci. Eng. 2020, 8, 647 24 of 25

59. Pléiades Images. Available online: https://www.intelligence-airbusds.com/en/8692-pleiades (accessed on

15 January 2020).
60. Lyzenga, D.R. Passive Remote Sens. techniques for mapping water depth and bottom features. Appl. Opt.
1978, 17, 379–383. [CrossRef] [PubMed]
61. Erdas Imagine. Available online: https://www.hexagongeospatial.com/products/power-portfolio/erdas-
imagine (accessed on 15 January 2020).
62. Global Mapper 20.1. Available online: https://www.bluemarblegeo.com/products/global-mapper.php
(accessed on 22 January 2020).
63. Pix4DMapper. Available online: https://www.pix4d.com/product/pix4dmapper-photogrammetry-software
(accessed on 22 January 2020).
64. 3D Research Srl. Available online: http://www.3dresearch.it/en/ (accessed on 20 January 2020).
65. Rende, F.S.; Irving, A.D.; Lagudi, A.; Bruno, F.; Scalise, S.; Cappa, P.; Di Mento, R. Pilot application of 3D
underwater imaging techniques for mapping Posidonia oceanica (L.) Delile meadows. Int. Arch. Photogramm.
Remote Sens. Spat. Inf. Sci. 2015, 5, W5. [CrossRef]
66. Rende, F.S.; Irving, A.D.; Bacci, T.; Parlagreco, L.; Bruno, F.; De Filippo, F.; Montefalcone, M.; Penna, M.;
Trabbucco, B.; Di Mento, R.; et al. Advances in micro-cartography: A two-dimensional photo mosaicing
technique for seagrass monitoring. Estuar. Coast. Shelf Sci. 2015, 167, 475–486. [CrossRef]
67. Mangeruga, M.; Cozza, M.; Bruno, F. Evaluation of underwater image enhancement algorithms under
different environmental conditions. J. Mar. Sci. Eng. 2018, 6, 10. [CrossRef]
68. Agisoft. Available online: https://www.agisoft.com/ (accessed on 20 January 2020).
69. Borra, S.; Thanki, R.; Dey, N. Satellite Image Analysis: Clustering and Classification; Springer: Dordrecht,
The Netherlands, 2019. [CrossRef]
70. eCognition Essential. Available online: http://www.ecognition.com/essentials (accessed on 23 January 2020).
71. SAGA. Available online: http://www.saga-gis.org/en/index.html (accessed on 23 January 2020).
72. Kursa, M.; Rudnicki, W. Feature selection with the Boruta Package. J. Stat. Softw. 2010, 36, 1–13. [CrossRef]
73. Breiman, L. Random Forests. Mach. Learn. 2001, 45, 5–32. [CrossRef]
74. Cohen, J. A Coefficient of Agreement for Nominal Scales. Educ. Psychol. Meas. 1960, 20, 37–46. [CrossRef]
75. Congalton, R. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sens.
Environ. 1991, 37, 35–46. [CrossRef]
76. Congalton, R.G.; Green, K. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices, 3rd ed.;
CRC Press: Boca Raton, FL, USA, 2019.
77. BORUTA Package. Available online: https://cran.r-project.org/web/packages/Boruta/index.html (accessed on
23 January 2020).
78. Ierodiaconou, D.; Schimel, A.; Kennedy, D.; Monk, J.; Gaylard, G.; Young, M.; Diesing, M.; Rattray, A.
Combining pixel and object based image analysis of ultra-high resolution multibeam bathymetry and
backscatter for habitat mapping in shallow marine waters. Mar. Geophys. Res. 2018, 39, 271–288. [CrossRef]
79. Montereale Gavazzi, G.; Madricardo, F.; Janowski, L.; Kruss, A.; Blondel, P.; Sigovini, M.; Foglini, F. Evaluation
of seabed mapping methods for fine-scale classification of extremely shallow benthic habitats—Application
to the Venice Lagoon, Italy. Estuar. Coast. Shelf Sci. 2016, 170, 45–60. [CrossRef]
80. Hasan, R.; Ierodiaconou, D.; Monk, J. Evaluation of Four Supervised Learning Methods for Benthic Habitat
Mapping Using Backscatter from Multi-Beam Sonar. Remote Sens. 2012, 4, 3427–3443. [CrossRef]
81. Stephens, D.; Diesing, M. A comparison of supervised classification methods for the prediction of substrate
type using multibeam acoustic and legacy grain-size data. PLoS ONE 2014, 9, e93950. [CrossRef] [PubMed]
82. Diesing, M.; Stephens, D. A multi-model ensemble approach to seabed mapping. J. Sea Res. 2015, 100, 62–69.
[CrossRef]
83. Moniruzzaman, M.; Islam, S.; Lavery, P.; Bennamoun, M.; Lam, C.P. Imaging and classification techniques
for seagrass mapping and monitoring: A comprehensive survey. arXiv 2019, arXiv:1902.11114.
84. Zhang, C. Applying data fusion techniques for benthic habitat mapping and monitoring in a coral reef
ecosystem. ISPRS J.Photogramm. Remote Sens. 2015, 104, 213–223. [CrossRef]
85. Veettil, B.; Ward, R.; Lima, M.; Stankovic, M.; Hoai, P.; Quang, N. Opportunities for seagrass research derived
from remote sensing: A review of current methods. Ecol. Indic. 2020, 117, 106560. [CrossRef]
J. Mar. Sci. Eng. 2020, 8, 647 25 of 25

86. Dattola, L.; Rende, S.; Dominici, R.; Lanera, P.; Di Mento, R.; Scalise, S.; Cappa, P.; Oranges, T.; Aramini, G.
Comparison of Sentinel-2 and Landsat-8 OLI satellite images vs. high spatial resolution images (MIVIS and
WorldView-2) for mapping Posidonia oceanica meadows. In Proceedings of the Remote Sensing of the Ocean,
Sea Ice, Coastal Waters, and Large Water Regions, Proc. SPIE 10784, Berlin, Germany, 10 October 2018;
Volume 10784. [CrossRef]
87. Pham, T.D.; Xia, J.; Ha, N.T.; Bui, D.T.; Le, N.N.; Tekeuchi, W. A Review of Remote Sens. Approaches for
Monitoring Blue Carbon Ecosystems: Mangroves, Seagrassesand Salt Marshes during 2010–2018. Sensors
2019, 19, 1933. [CrossRef]
88. Li, J.; Schill, S.R.; Knapp, D.E.; Asner, G.P. Object-Based Mapping of Coral Reef Habitats Using Planet Dove
Satellites. Remote Sens. 2019, 11, 1445. [CrossRef]
89. Ardhuin, F.; Rogers, E.; Babanin, A.; Filipot, J.F.; Magne, R.; Roland, A.; Van der Westhuysen, A.; Queffeulou, P.;
Lefevre, J.; Aouf, L.; et al. Semiempirical dissipation source functions for ocean waves. Part I: Definition,
calibration, and validation. J. Phys. Oceanogr. 2010, 40, 1917–1941. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).