Sustain. Water Resour. Manag.

DOI 10.1007/s40899-017-0127-2

ORIGINAL ARTICLE

Application of SWAT model for predicting soil erosion

and sediment yield
Subhasri Dutta1 • Dhrubajyoti Sen2

Received: 6 February 2016 / Accepted: 26 April 2017

Ó Springer International Publishing Switzerland 2017

Abstract Hirakud, the longest earthen dam in Asia on the Soil Loss Equation. The graphical as well as statistical
river Mahanadi, has been affected by sedimentation prob- results achieved in this study indicate that the model is
lems which affect reservoir storage capacities and different capable to identify critical erosion-prone areas within the
reservoir operations. This article presents the hydrological watershed on the scale of sub-watershed which helps the
modeling of the Mahanadi catchment up to Hirakud, hydrologists to introduce effective management practices
located in the states of Chhattisgarh and Odisha in India, at the lowest cost. The study shows that about 34% area of
using the modified version of Soil and Water Assessment total catchment falls under high or above soil erosion zones
Tool (SWAT) with an interface of ArcView Geographic with combination of coarse loamy type of soil and agri-
Information System software to predict soil erosion and cultural type of land use and land cover condition. The
sediment transport to the reservoir for improving its useful results of this study also identify regions of high sediment
life. The data sources used in the study were: terrain data yield and sediment delivery ratio. To enhance the SWAT
90 m resolution of Shuttle Radar Topography Mission, model performance, it is recommended to use sub-daily
Land Use and Land Cover derived from Landsat Enhanced data and to prioritize soil erosion at the Hydrological
Thematic Mapper Plus, soil map published by National Response Units level for improvement of watershed
Bureau of Soil Survey and Land Use Planning, India and management.
meteorological data collected from India Meteorological
Department, India. The study involves the analysis of Keywords SWAT
Hirakud reservoir
Erosion

sensitive parameters of the simulated data (discharge 15, Sediment yield
Sediment delivery ratio
Prioritization
sediment load 6); 5-year calibration and 3-year validation
on daily basis to predict the discharge and sediment load.
The sensitivity analysis was carried out using the Latin Introduction
Hypercube and One-factor-At-a-Time. The results also
show that the prediction of sediment yield is highly sen- Watershed is defined as a hydrological unit which produces
sitive to the sizes of different sub-basins due to the sensi- water in the form of runoff due to the interaction of pre-
tivity of topographic factors used in the Modified Universal cipitation and land surface. To reduce the peak rate of
runoff causing soil erosion, detailed understanding and
evaluation of hydrological processes, land managements,
& Subhasri Dutta soil types and climatic conditions of different parts of the
subhasri_dutta@swr.iitkgp.ernet.in
watershed are essential. Sedimentation is a major problem
Dhrubajyoti Sen in design and operation of reservoirs (Morris and Fan
djsen@civil.iitkgp.ernet.in
1998), which causes the reduction of probable water stor-
1
School of Water Resources, Indian Institute of Technology age capacity, resulting in backwater flooding, shortage in
Kharagpur, Kharagpur 721302, India water availability for irrigation purposes as well as for
2
Department of Civil Engineering, Indian Institute of industrial and domestic uses and reducing the hydropower
Technology Kharagpur, Kharagpur 721302, India generation from the multipurpose reservoir for which it

123
 Sustain. Water Resour. Manag.

was constructed across the rivers at massive cost. Previous amount of temporal and spatial data regarding sediment
studies have shown that reservoirs around the globe loose transport and do not consider impact of climate change,
their storage capacity due to the sedimentation at a rate of LULC and other natural factors (Walling 1983; de Vente
1–2% annually (Mahmood 1987), whereas in India, reser- et al. 2007). Sediment rating curve, developed from
voirs loose their capacity to the tune of 0.2–1% annually detailed sediment flow data to predict SDR, underestimates
(CWC 2001). Sediment generated due to soil erosion, is the actual loads and can be used in watersheds where no
transported by runoff and deposited into depressions, significant changes are expected in LULC, soil types and
streams and water bodies at a high deposition rate (Ver- sediment supply sources (Ferguson 1986; Bogen and
straeten and Poesen 1999) causing the degradation of water Bonsnes 2003). Dutta (2016) has presented a comprehen-
quality. Soil erosion occurs mainly due to the erodible soil, sive art of the review on soil erosion, sediment yield and
land use and land cover (LULC), high terrain slope, high sedimentation of reservoir, highlighting different methods
intensity of rainfall and human activities occurring over the for predicting these variables.
areas (Beskow et al. 2009; Li et al. 2012). It varies from Table 1 represents a brief description of empirical,
each part of the watershed to other and is also not equally physical and conceptual models for predicting soil erosion.
distributed over the watershed because of variance of nat- Empirical models, developed from the statistical analysis
ural and man-made conditions and their activities. on the information obtained from experimental measure-
All the eroded sediment particles within the catchment ment or field measurement, are useful in predicting soil
are not transported to the outlet as they are trapped and erosion, but are site specific and require long-term data
deposited in the upstream reaches. A part of the eroded (Elirehema 2001). They are also unable to analyze the
material that is transported to the watershed outlet is ter- dynamics of sediment erosion and deposition in the
med as sediment yield (Walling 1983). Sediment delivery watershed. Physically based models are based on the con-
ratio (SDR), which is defined as the ratio of sediment yield servation laws of mass and energy for describing runoff
at the watershed outlet to the gross erosion in the water- and sediment yield induced by independent rainfall events
shed, is an important parameter for sustainable manage- in the watershed in the form of partial differential equations
ment of soil erosion and sedimentation problem. Various (Bennett 1974). But the disadvantage of this type of model
methods have been used to calculate SDR such as simple is the requirement of large amount of data. Conceptual
empirical relationships and complex physically based models, which are mixture of empirical and physically
equations (Sean et al. 2013). An empirical equation using based models, provide an indication of qualitative as well
channel slope and two empirical coefficients to determine as quantitative processes within a catchment without con-
SDR was developed by Williams and Berndt (1972). But sidering process interactions. In addition, it has been well
the drawbacks of these methods are that they include huge established that Revised Universal Soil Loss

Table 1 Different models for predicting soil erosion (Merritt et al. 2003)
Empirical models Physically based models Conceptual models

Musgrave equation (Musgrave 1947) ANSWERS (Beasley et al. 1980) Sediment Concentration Graph (Johnson
PSIAC method CREAMS (Knisel 1980) 1943)
(Pacific Southwest Interagency Committee GUESS (Rose et al. 1983) Renard–Laursenn model (Renard and Laursen
1968) 1975)
EPIC (Williams et al. 1984)
USLE (Wischmeier and Smith 1978) Unit sediment graph (Rendon-Herrero 1978)
TOPOG (O’Loughlin 1986)
SLEMSA (Elwell 1978) Instantaneous unit sediment graph (Williams
WEPP (Nearing et al. 1989)
1978)
SEAGIS (DHI 1999) KINEROS (Woolhiser et al. 1990)
AGNPS (Young et al. 1989)
RUSLE (Renard et al. 1996) GUEST (Misra and Rose 1990)
EMSS (Vertessey et al. 2001)
SEDNET (Prosser et al. 2001c) dUSLE (Flacke et al. 1990)
HSPF (Johanson et al. 1980)
RUSLE2 (Foster et al. 2002) EROSION2D (Schmidt 1991)
IQQM (DLWC 1999)
PESERA (Kirkby et al. 2004) PERFECT (Littleboy et al. 1992)
LASCAM (Viney and Sivapalan 1999)
LISEM (de Roo et al. 1994)
SWRRB (USEPA 1994)
KINEROS2 (Smith et al. 1995)
EUROSEM (Morgan et al. 1998)
WATEM/SEDEM (Takken et al. 1999; Feng et al.
2010)
MIKE-11 (Hanley et al. 1998)

123
Sustain. Water Resour. Manag.

Equation (RUSLE) incorporating with remote sensing and sediment load (Chaplot 2005; Jha et al. 2010), impact of
Geographic Information System (GIS) can be effectively climate (Zhang et al. 2007; Mukhtar et al. 2014) and LULC
used to model the runoff and soil erosion; and also to changes (Tang et al. 2005; Cao et al. 2009), effect of
predict the impact of LULC changes on erosion rate sediment control structures in small watershed (Mishra
(Ganasri and Ramesh 2015). et al. 2007) and modeling of ungauged catchments (Prab-
To minimize the adverse effect of soil erosion on human hanjan et al. 2015). Different studies have also been carried
society and infrastructure, concerted efforts are required for out using SWAT incorporating with other models
planning and implementation of soil and water conserva- (Griensven et al. 2013; Shrestha et al. 2013; Babar and
tion practices at a sub-watershed scale. However, contin- Ramesh 2014; Song et al. 2015). Researchers (Parajuli
uous monitoring of runoff generation and sediment yield et al. 2009; Talebizadeh et al. 2009) have concluded that
contribution by the eroded land are impossible due to its SWAT model has higher model efficiency and an accept-
prohibitive expenses in cost and time (Kaur et al. 2004). able uncertainty compared with other models. The advan-
Critical erosion-prone areas need to be identified and pri- tage of using SWAT over other methods is to consider the
oritized properly to reduce the erosion problem (Tripathi watershed properties as the main source of soil erosion and
et al. 2003; Pandey et al. 2007) specially where availability sediment contribution to the reservoir. It is, therefore, that
of data is limited (Ndomba et al. 2008; Kumar et al. 2014). the SWAT model has been selected as the modeling
Therefore, using the distributed and hydrological sediment approach for this study.
yield models, problems dealing with stationarity and lin- Information about the sensitivity analysis, calibration
earity can be avoided. During the last few decades, phys- and validation is important and effective to minimize the
ically based distributed models, remote sensing techniques uncertainty and also to increase the accuracy of the output
and geographical information systems have been used of the model. However, most of the research papers on the
increasingly with the advancement to simulate hydrologi- application of the SWAT model do not include the detailed
cal processes in a watershed. However, watershed models descriptions of model calibration and validation processes
are considered as the best as they are economic and less (Griensven et al. 2013; Sean et al. 2013). Moreover, the
time consuming. In many countries these models have model is mostly run on monthly or yearly step, which
performed very well in the long-term assessment of runoff, might fail to look the detailed information of the hydro-
soil erosion and sediment yield for a wide range of soil logical processes (Singh et al. 2011; Manaswi and Thawait
types, land uses and climatic conditions. 2014; Kumar et al. 2015).
A model is selected depending on the objectives, the This study has, therefore, been carried out to overcome
availability of input data to run the model and the uncer- these gaps, considering the hydrological behavior of the
tainty in interpreting the output obtained. It has been found watershed and high temporal resolution of sediment flow
from the literature review that change of LULC, climatic data, with the following objectives: (1) detailed informa-
condition and human activities in the basin of a reservoir tion of daily calibration and validation of SWAT model,
are highly responsible for increased reservoir sedimenta- (2) identification of erosion-prone sub-watersheds within
tion (Rose et al. 2011; Schiefer et al. 2013). However, the the watershed, (3) determination of SDR of individual sub-
use of hydrological modeling to control and manage the watershed, (4) analysis of the impact of different combi-
reservoir sedimentation and to improve its life, has rarely nations of LULC and soil types on sediment erosion, (5)
been done so far, except in a recent few studies (Sardar finding the relationships between rainfall, runoff, soil ero-
et al. 2013; Hunink et al. 2013). The Soil and Water sion and sediment yield of the whole watershed, and (6)
Assessment Tool (SWAT) model (Arnold et al. 1998), a evaluate the performance of SWAT model to predict sed-
physically based spatially distributed hydrological model, imentation of Hirakud reservoir on river Mahanadi in India.
is established as a convenient tool which has been used Different studies have also been carried out on the sedi-
successfully for simulating runoff (Kannan et al. 2007; mentation of Hirakud reservoir (Rathore et al. 2006;
Zhang et al. 2008; Easton et al. 2010), sediment yield (Xu Mukherjee et al. 2007; Dutta and Sen 2016). However, this
et al. 2009; Prabhanjan et al. 2015) and water quality is a first attempt to use SWAT model to predict the soil
(Abbaspour et al. 2007; Jha et al. 2007; Geza and McCray erosion and sediment yield in the catchment of Hirakud
2008; Debele et al. 2009) in various watersheds on the reservoir. The analysis of these results is expected to assist
basis of Modified Universal Soil Loss Equation (MUSLE) practicing engineers to take necessary decisions for
having a large variation of land use, soil and management enhancing the life of a reservoir. The first segment of this
conditions over a long period of time. SWAT has also been paper focuses on the description of the area and charac-
used to predict different watershed issues such as evapo- teristics of Hirakud reservoir basin. The following sections
transpiration (Kannan et al. 2007; Licciardello et al. 2011), are dedicated to the methodology used in SWAT. The last
crop yield (Srinivasan et al. 2010), total maximum part of the paper involves the analysis of results of SWAT

123
 Sustain. Water Resour. Manag.

model concerning the prediction of runoff, soil erosion, reservoir is 2.67 lakh hectares, of which 1.59 lakh hectares
sediment yield, sediment delivery ratio and sedimentation is during the khariff or rainy season, and the remaining is
of Hirakud reservoir. during the rabi or post-monsoon season (Sahu and Das
2012; Panigrahi and Paul 2015). In addition, the hydro-
power generation potential is 347.5 MW, distributed at two
Study area sites (Burla 275.5 MW and Chiplima 72 MW). The total
hydropower generation in the Mahanadi basin by five
Mahanadi, one of the major rivers in East India, has a total hydroelectric projects and six powerhouses is 1184.5 MW
course of 858 km, draining into Bay of Bengal. It origi- including that of Hirakud. The flow in the Mahanadi river
nates from Sihawa mountains, Dhamtari district in the state downstream of the reservoir is mainly dependent upon
of Chhattisgarh. These mountains are an extension of the water released after hydropower generation from the
Eastern Ghats Mountain Ranges that run parallel to the reservoir. The physical quantity of water which passes
eastern coastline of India and are a source of many other through the generating turbines is utilized mostly for irri-
streams which then go on to join the Mahanadi. The river is gation in the agricultural fields of the delta. Water intake
mainly rainfed and, therefore, fluctuates widely in dis- from the reservoir through lift for peripheral area irrigation
charge over time. During eight months of the year, the river has been strictly prohibited as it would affect hydropower
assumes a narrow and shallow channel but during mon- generation from the dam.
soons, its fury causes devastations along its banks and flood The watershed of river Mahanadi up to the reservoir
plains. pertains to subtropical climate zone with hot and humid
The Mahanadi basin has varying topography with the monsoonal climate. Maximum annual rainfall over the
lowest elevation in coastal reaches and highest towards the catchment of Hirakud is 1928.42 mm and minimum annual
northern hills. The maximum elevation observed is 1321 m rainfall is 691.46 mm, 70% of which occurs from middle
above the mean sea level in the steep hilly terrain of of June to end of September with mean annual rainfall of
Mahanadi; and maximum area of the plain region of the 1122.28 mm. Daily mean relative humidity varies from 18
Mahanadi valley falls under the 200–400 m elevation zone. to 88% in the month of April and August respectively. The
Also, the river basin encompasses the Chilika lagoon, a mean annual runoff to Hirakud reservoir is estimated as
large water body connected to the Bay of Bengal. 33,130 Mm3 and the design flood discharge is 42,475
Agriculture dominates over the land use classification cumecs. For the period of 1990–2012, the hottest months
with the major crops of paddy, sugarcane and cotton as are May and June when temperature ranges between 38 and
kharif season crops (15 June to 30 October) followed by 43 °C and the coldest month is December with the tem-
paddy and wheat as rabi season crops (15 December to 30 perature between 10 and 13.7 °C (Water Resources
April). Department, Government of Odisha 2007).
The texture of the soil is mostly fine followed by med- Nowadays, the reservoir is affected strongly by the sedi-
ium types. The main soil types found in the basin are red mentation due to the soil erosion which results mainly due to
and yellow soils, mixed red and black soils, laterite soil and land degradation in the watershed (Chakrapani and Subra-
delta soils. Proterozoic sedimentary rocks like limestone, manian 1990, 1993; Mukherjee et al. 2007). These hamper the
calcareous shale and sandstone dominate the upstream part serviceability of the reservoir to a great extent. So far, there
of the river; whereas, the downstream part of the river is are no reported literature about the erosion-prone zones and
dominated by silicate rocks of metamorphic origin (Water the sediment delivery ratio for Hirakud watershed.
Resources Department, Government of Odisha 2007;
Mishra 2008).
The Hirakud reservoir was constructed on river Maha- Model description
nadi in the year of 1957 to serve multipurpose objectives
such as irrigation, flood control and hydropower genera- ARCSWAT is a user-friendly software where SWAT is
tion. The reservoir is located between latitude 21°250 to integrated with the ArcView-Geographic Information
21°550 and longitude 83°100 to 84°050 with the main System (GIS). For running SWAT model, the spatial
tributaries of Seonath, Jonk, Hasdeo, Mand and Ib (Fig. 1). datasets like topography, land use and soil data are pre-
The total catchment of Mahanadi river is 1,41,134 km2 out requisite input parameters. It has weather simulation
of which Hirakud catchment is 84,818.55 km2. The component which can predict the missing data in the
watershed of the river up to the Hirakud dam site is located observed records of weather data. SWAT delineates total
largely in the state of Chhattisgarh (91.2%) and the watershed into sub-watersheds based on Digital Elevation
remaining in Odisha (8.8%). The reservoir extends to Model (DEM) for simulation of surface runoff through the
55 km behind the dam. The irrigation potential of Hirakud channel networks within the sub-watershed and the sub-

123
Sustain. Water Resour. Manag.

Fig. 1 Location of study area

watersheds are further divided into multiple homogeneous Penman–Monteith (Allen et al. 1989), (2) Hargreaves
Hydrological Response Units (HRUs), each of which is (Hargreaves and Samani 1985) or (3) Priestley-Taylor
characterized by a unique combination of land use, soil (Priestley and Taylor 1972). The hydrological cycle sim-
type and average slope. The climate parameters consist of ulated by SWAT is based on the well known water balance
rainfall, temperature (maximum and minimum), relative equation (Setegn et al. 2009):
humidity, wind speed and sunshine hours. SWAT com-
t 
putes the surface runoff and sediment yield at the HRU X 
SWt ¼ SW0 þ Rday Qsurf Ea wseep Qgw ;
level which are aggregated to the sub-watershed level and i¼1
routed to the watershed outlet through the stream network.
ð1Þ
Figure 2 represents the flowchart showing the entire
methodology of SWAT model. where SWt is the final soil water content (mm), SW0 is
In the SWAT model, daily surface runoff is calculated initial soil water content (mm), t is time (days), Rday is
using the Soil Conservation Service (SCS) curve number amount of rainfall on day i (mm), Qsurf is amount of surface
(CN) technique, which is a function of soil permeability, runoff on day i (mm), Ea is amount of evapotranspiration
land use and 5-day antecedent soil moisture content. Low on day i (mm), wseep is amount of water entering the vadose
value of CN indicates low runoff potential whereas large zone from the soil profile on day i (mm), Qgw is amount of
CN value indicates increasing runoff potential. For ground water on day i (mm).
streamflow routing, either variable storage coefficient The Universal Soil Loss Equation (USLE) was used in
(Williams 1969) or Muskingum method (Overton 1966) is SWAT at HRU level to predict soil erosion but it is not
used. Peak rate of runoff is simulated by employing considered in sediment routing (Neitsch et al. 2011). The
rational method. Potential evapotranspiration (PET) can be following equation represents USLE (Wischmeier and
calculated either by using following three methods: (1) Smith 1965):

123
 Sustain. Water Resour. Manag.

Fig. 2 Flowchart showing the

entire methodology of SWAT
model

ero ¼ 1:292  EIUSLE  KUSLE  CUSLE  PUSLE runoff rate (m3/s), areahru is area of the HRU (ha), KUSLE is
 LSUSLE  CFRG, ð2Þ USLE soil erodibility factor [0.013 metric ton m2 h/(m3-
metric ton cm)], CUSLE is USLE cover and management
where ero is the quantity of eroded soil in a given day factor, PUSLE is USLE support practice factor, LSUSLE is
(metric tons/ha), EIUSLE is rainfall erosion index [0.017 m- USLE topographic factor and CFRG is coarse fragment
metric ton cm/(m2h)], KUSLE is USLE soil erodibility factor factor. MUSLE is implemented in SWAT by assuming a
[0.013 metric ton m2h/(m3-metric ton cm)], CUSLE is USLE simple hydrograph shape to estimate daily runoff volume
cover and management factor, PUSLE is USLE support with peak flow rate within the sub-watershed area, which is
practice factor, LSUSLE is USLE topographic factor and further used to predict the variation of runoff erosive
CFRG is the coarse fragment factor. energy. SWAT simulates the sediment yield in terms of
Modified Universal Soil Loss Equation (MUSLE), total sediment loadings and also as the fractions of sand,
which is a function of runoff factor, was used to predict silt and clay from individual sub-watershed.
sediment yield on a given day (Wischmeier and Smith
1965). The equation is as follows:
0:56
Model input

sed ¼ 11:8  Qsurf  qpeak  areahru KUSLE
ð3Þ
 CUSLE  PUSLE  LSUSLE  CFRG,
In this study, ARCSWAT 2005 was used to simulate sed-
where sed is the sediment yield on a given day (metric iment yield at sub-watershed scale for a period of
tons), Qsurf is surface runoff volume (mm/ha), qpeak is peak 2000–2012. The area of interest was delineated by

123
Sustain. Water Resour. Manag.

ARCSWAT, using 90 m resolution DEM provided by the basin boundary. The types are close to the following val-
Shuttle Radar Topography Mission (SRTM), which was ues: agriculture (52.58%), deciduous forest (34.67%),
then subdivided into 23 sub-watersheds. The delineated mixed forest (8.18%), pasture (0.27%), urban area (0.55%),
area of Hirakud watershed was obtained as 84,818.55 km2. urban area of very low density (0.02%) and water body
The topography of the watershed area varies widely from (3.73%) as shown in Fig. 4.
137 to 1148 m (Fig. 3) above mean sea level. Soil map of the study area in the scale of 1:2,50,000 was
Land use–land cover (LULC) was prepared using ima- collected from National Bureau of Soil Survey and Land
ges of the Landsat Enhanced Thematic Mapper Plus Use Planning (NBSS and LUP), Nagpur, India, with the
(ETM?) (date 14.10.2009 and 23.11.2012). The image detailed information of its texture profile and then it was
dates were selected as they were free from cloud cover and digitalized for further reclassification (Fig. 5). Soils are
coinciding with the flow simulation period. The images mostly classified as coarse loamy (3.84%), clayey
were enhanced after geometric correction, which resulted (46.14%), fine (1.79%) and fine loamy (48.23%). Lime-
in Root Mean Square Error (RMSE) of 0.64 and 0.82 of a stone, calcareous shale and sandstone form the major
pixel for the two images respectively, considered within the sedimentary rocks in the catchment of Hirakud. Soils are
acceptable range of RMSE (Clark Labs 2001). Since the classified as group A (high infiltration rate and low runoff
individual images were of different dates, it was not fea- potential), group B (moderate infiltration rate and runoff
sible to classify the mosaic image into land cover types. potential), group C (low infiltration rate and high runoff
Henceforth, each image was classified individually using potential) and group D (very low infiltration and high
unsupervised classification and then they were merged runoff potential). Based on the analysis, it was found that
based on their spectral signatures into seven land use and most of the soil falls in the hydrological group of C with
land cover types. The LULC classification was found to be texture fine loamy followed by group D with texture
similar to that of Dadhwala et al. (2010) who carried out clayey. SWAT model requires different soil textural and
their study in Mahanadi river basin. The individual clas- physico-chemical properties in the form of soil texture,
sified images were then mosaiced to merge them together available water content, hydraulic conductivity, bulk den-
and finally the study area was extracted as defined by the sity and organic carbon content for different layers of each

Fig. 3 Elevation of study area Fig. 4 Land use classification of study area

123
 Sustain. Water Resour. Manag.

Pune, India, for 1990–2012 were provided to the SWAT

model. The daily discharge and sediment load data were
collected from gauge-discharge stations on the Mahanadi
river maintained by Central Water Commission (CWC),
Government of India, for the period of 2000–2012. In the
present study, in spite of using Green–Ampt infiltration
method (Green and Ampt 1911) due to the absence of
subdaily rainfall data for the study area, the authors used
the SCS-CN method for simulating surface runoff. Pen-
man–Monteith method was used in this study to estimate
PET because it was reported (Allen et al. 1989) that it is the
best method when a full detailed set of weather data is
available. The other advantage of using this method over
two other methods available in SWAT is that it yields good
result under wide range of variety of climate scenarios
(Droogers and Bastiaanssen 2002; Saghravani et al. 2009).
The Muskingum routing method was employed for routing
stream flow and the Bagnold stream power equation
(Bagnold 1977) was used for routing sediment through the
channels. For all unavailable parameters default values
were used. After setting all the input parameters, the
SWAT model was run on daily basis for simulating various
hydrological components.

Sensitivity analysis, calibration and validation

Fig. 5 Soil class of study area
An initial simulation of the model using default parameters
group of soil. The database obtained from soil map was did not give satisfactory results so far as the output
updated with the SOIL-FAO database (FAO: Food and parameters of the model are concerned. Therefore, the
Agriculture Organization of the United Nations 1995). The sensitivity analysis of the simulated data, calibration and
soil data of Hirakud catchment were then added into the validation were carried out. In SWAT, 21 sensitive
SWAT soil database using the manual option provided in parameters (discharge 15, sediment load 6) were identified
the SWAT model. based on literature review and ranked accordingly from the
After reclassifying both the LULC map and soil maps most sensitive to the least. Sensitivity analysis was per-
with respect to their databases using look up tables, they formed using the Latin Hypercube and One-factor-At-a-
were overlaid along with the slope layers with the delin- Time (LH-OAT) sampling method developed by Van
eated watershed to subdivide the sub-watersheds into 167 Griensven and Meixner (2006), in which the model was run
HRUs (land use threshold 20%, soil threshold 10% and (p ? 1) 9 10 times, where p is the number of parameters
slope threshold 20%) which are the smallest units in being simulated and 10 is the number of loops (Nearing
SWAT to reflect the evapotranspiration and the hydrolog- et al. 1989). This procedure combines the advantages of
ical conditions for different combinations of land use, soil strength of global and local sensitivity analysis methods
and slope. Reducing the number of HRUs, to decrease the and can provide efficiently the ranking order of parameters
computation time of the model simulation, will cause an according to their sensitivity. Sum of Squared Residuals
error in HRU aggregation by increasing an error in input (SSR) was used as an objective function for the sensitivity
data. Similarly, defining more number of HRUs will pre- analysis in this study. Tables 2 and 3 represent the sensitive
sent more accurate results but will take a longer simulation parameters according to their sensitivity, their corre-
time. Therefore, the above-said adequate threshold values sponding optimum values and percentage of change in the
were used as recommended by Winchell et al. (2007) to result of simulating discharge and sediment load respec-
remove minor land use, soil and slope types so that a tively, which was supported by different studies (Setegn
simplified HRU definition could be achieved. et al. 2010; Pandey et al. 2016). Increasing the curve
The daily grided (1° 9 1°) meteorological data, number for moisture condition II (CN2) to a value of 76.26
obtained from India Meteorological Department (IMD), caused the runoff to increase and infiltration to decrease.

123
Sustain. Water Resour. Manag.

Table 2 Sensitive parameters for discharge simulation and their calibrated values
Sensitive parameters Range Ranks of the parameters Calibrated value Change in result (%)

CN2 35 to 98 1 76.26 29.63

SURLAG 0.05 to 24 2 0.07 27.87
CH_N -0.01 to 0.3 3 0.26 26.24
ESCO 0.01 to 1 4 0.59 24.54
SLSUBBSN (m) 10 to 150 5 129.42 20.20
GWQMN (mm) 0 to 5000 6 4245.00 18.12
EPCO 0.01 to 1 7 0.62 16.114
GW_REVAP 0.02 to 0.2 8 0.12 15.87
SOL_AWC (mm H2O/mm soil) 0.5 to 1 9 0.97 14.65
SOL_ALB 0 to 0.25 10 0.11 10.21
GW_DELAY (days) 0 to 500 11 71.50 8.47
ALPHA_BF (days) 0.1 to 0.3 12 0.12 2.12
REVAPMN (mm) 0 to 500 13 365.50 2.04
SOL_K (mm/h) 0 to 2000 14 1354.00 0.78
TLAPS (°C/km) -10 to 10 15 -0.18 0.90

Table 3 Sensitive parameters for sediment load simulation and their in the shallow aquifer, required for return flow to occur
calibrated values (GWQMN) and groundwater ‘revap’ coefficient
Sensitive Range Ranks Calibrated Change in (GW_REVAP) are set to the values of 4245.00 mm and
parameters of the value result (%) 0.12 respectively as they control the movement of
parameters groundwater from the shallow aquifer to the overlying
USLE_P 0 to 1 1 AGRL 0.2 6.25 unsaturated zone. For accurately predicting surface runoff,
FRSD 0.03 the moist soil albedo coefficient (SOL_ALB) was set to
FRST 0.03 0.11, which controls the amount of reflection of solar
PAST 0.003 radiation by the watershed body. Groundwater delay time
USLE_C 0.001 to 0.5 2 0.49 5.29 (GW_DELAY) was adjusted to the value of 71.50 days as
SPCON 0.0001 to 0.01 3 0.002 5.10 it controls the flow of water to pass through the lowest
CH_COV -0.001 to 1 4 0.20 4.75 depth of the soil profile and through the vadose zone before
CH_EROD -0.05 to 0.6 5 0.53 3.78 recharging the shallow aquifer. Baseflow alpha factor
SPEXP 1 to 1.5 6 1.27 1.25
(ALPHA_BF) was adjusted to 0.12 days as it makes stee-
per the simulated discharge recession curve to represent the
faster drainage behavior of the watershed. Threshold depth
Surface runoff lag coefficient (SURLAG) was set to a of water (365.50 mm) in the shallow aquifer required for
value of 0.07 as it controls the fraction of total available percolation to occur (REVAPMN), was adjusted to move
water to enter into the reach in one day. The value of the water from the shallow aquifer to the unsaturated zone.
Manning’s friction coefficient n for the main channel Available water capacity of the soil layer (SOL_AWC) was
(CH_N) was increased to 0.26 as it affects indirectly the increased to a value of 0.97 mm of water per mm of soil to
peak discharge in the channel by affecting the time of decrease the flow of water within the soil profiles. Negative
concentration of the tributaries. Soil evaporation compen- value (-0.18 °C/km) of temperature lapse rate (TLAPS)
sation factor (ESCO) and plant uptake compensation factor signifies that an increase of elevation causes a decrease in
(EPCO) were set to the values of 0.59 and 0.62 respec- temperature. It is used to adjust temperature for elevation
tively, to modify the depth distribution of water in the soil bands in the sub-watershed. The parameters listed above
layers to meet the soil evaporative and plant uptake were found to be the most sensitive parameters that affect
demands. Average slope length (SLSUBBSN) was the hydrological responses of the catchment. The values of
increased to a value of 129.42 m as it adjusts the move- support practice factor in USLE equation (USLE_P) were
ment of lateral flow. Saturated hydraulic conductivity adjusted according to LULC as they influence the human
(SOL_K) was increased to 1354.00 mm/h to increase the activity on the physical processes in the catchment. The
water flow rate through the soil. Threshold depth of water value of crop management factor in USLE equation

123
 Sustain. Water Resour. Manag.

(USLE_C) was set to the value of 0.49 to control the for- sim
observed data, Ymean is mean value of the simulated
mation of sediment load. Linear parameter for calculating data and n is the total number of observations.
sediment reentrained in channel sediment routing (SPCON) 2. Observations standard deviation ratio (RSR): it is the
and exponent parameter for calculating sediment reen- ratio of root mean square error (RMSE) and standard
trained in channel, sediment routing (SPEXP) were set to deviation of observed data (Singh et al. 2007), and
0.002 and 1.27 as they increase the maximum amount of written in the following form:
sediment to be transported through the channel and to qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn obs 2
decrease amount of sediment that could be deposited. i¼1 ðYi Yisim Þ
RMSE
Channel cover (CH_COV) and erodibility (CH_EROD) RSR ¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi :
STDEVObs Pn  obs obs 2

factor linearly influence the soil loss from channels (Easton i¼1 Yi Ymean
et al. 2010). So they were set to the values of 0.20 and 0.53
respectively. These sensitive parameters are used to control ð5Þ
the amount of erosion from the channel and its catchment RSR varies from 0 to any positive value. Lower value
as they affect the rate of runoff, sediment and soil nutrient of RSR indicates better performance of the model
loss to the maximum extent. simulation. The advantage of using RSR over RMSE is
Calibration, which is a process of modifying sensitive that the resulting statistic values can be applied to any
input parameters, one at a time, to get a good agreement constituents as it incorporates the benefits of error
between observed and predicted values, is carried out for index statistics by including a scaling or normalization
the simulation of discharge and sediment load of the factor (Moriasi et al. 2007).
model. The hydrological component of the model was first 3. Nash-Sutcliffe efficiency (NSE): it is a statistical
calibrated manually to optimize the simulated discharge on measure that determines the relative magnitude of the
the basis of daily observed data for the period of residual variance compared to the measured data
2000–2005 on three sub-watershed areas: Rampur, Bam- variance (Nash and Sutcliffe 1970). The following
nidih and Baronda. The validation was performed without equation presents NSE:
changing these values of calibrated parameters to verify the "P  2 #
n obs
model’s ability to simulate the discharge at the same sub- i¼1 Yi Yisim
NSE ¼ 1 ð6Þ
obs 2
 :
watersheds during the period of 2006–2008. Same time Pn  obs
Y Y
i¼1 i mean
period of calibration and validation were selected for sed-
iment load simulation at daily time-steps by SWAT model
NSE ranges from -? to 1. Value of NSE between 0
on four sub-watershed areas: Basantpur, Baronda, Jondhra
and 1 is an acceptable level of performance, whereas value
and Bamnidih. It was performed after getting satisfactory
B0 is an unacceptable performance as the mean observed
result of calibration and validation of simulated discharge.
value is a better predictor than the simulated value.
Calibration and validation period were selected according
As recommended by Moriasi et al. (2007), discharge and
to the availability of observed data of the watershed. The
sediment load were calibrated and validated with respect to
model was run for several times to get the optimum cali-
their statistical indices. It was assumed that throughout the
brated values of the input parameters. Following three
simulation period there was no spatial variation of land use
statistical measures were employed to evaluate the per-
and soil properties.
formance of SWAT in terms of the accuracy and consis-
The analysis of the results is mainly quantitative
tency on the prediction of discharge and sediment load:
without any significance tests on quality of water. The
1. Pearson’s correlation coefficient (R): it is an index of data were used to determine the soil erosion and sedi-
degree of relationship between observed and simulated ment load transported in Hirakud reservoir from its
values, ranging from -1 to ?1. R = 0 means that watershed.
there is no correlation, whereas R = ?1 or -1 The comparison of statistical evaluations between model
represents a positive or negative correlation between simulated and observed data of daily discharge was plotted
two sets of data. The formula for R is as follows: in Fig. 6 which shows that the model performed well to
Pn  obs obs
 sim sim
 represent the dynamics of the basin. In Fig. 6, both the
i¼1 Yi Ymean Yi Ymean
R ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 qP 2ffi ;
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rising and falling limbs of the hydrographs are well rep-
obs 2
Pn  obs n  sim
i¼1 Yi Ymean i¼1 yi ysim
mean resented and the peak flow, responsible for predicting soil
ð4Þ erosion and sediment transport are well expressed. How-
ever, the simulated discharge was slightly higher than the
where Yiobs is the ith value of observed data, Yisim is ith observed discharge corresponding to the same concentrated
obs value of rainfall.
value of simulated data, Ymean is mean value of the

123
Sustain. Water Resour. Manag.

It is shown in Fig. 6 that the discharge varies both

temporally over time and spatially over the Hirakud
watershed. This is due to the spatial distribution of the
rainfall which has pronounced effects on the spatial dis-
tribution of soil water content, temperature and generated
runoff. Hence, it is important to pay more attention on the
spatial distribution of the rainfall data than the land use and
soil map data. Figure 7 shows satisfactory agreement
between simulated and observed daily sediment load data
in four sub-watershed areas.
In Figs. 8 and 9 statistical results show good model
prediction for the discharge and sediment load at the cal-
ibrated sub-watersheds during the validation period. Dur-
ing high storm events, daily simulated runoff is
overestimated (Fig. 8) by SWAT model as SCS-CN
method is unable to predict accurate runoff for a day that
experiences several storm events. This is because of the
variation of the soil moisture level and corresponding CN
value from storm to storm which occur successively on a
single day (Kim and Lee 2008). It is clear from Fig. 9 that
the sediment load increases with increasing rainfall value
and vice-versa but the rate of increasing sediment load is
not fixed.
The results of statistical measures show that SWAT
Fig. 7 Observed and simulated daily sediment load during the
performed very well in predicting discharge and sediment
calibration period for the sub-watershed area a Basantpur, b Baronda,
load in the Hirakud watershed with slight variation of peak c Jondhra and d Bamnidih
value of sediment load. The inconsistency between
observed and simulated data is found due to the non
availability of some discharge and bed load data in the
corresponding sub-watersheds.
The scattergrams of the observed and simulated dis-
charge and sediment load data were plotted during the
calibration and validation in Figs. 10 and 11, respectively.
The statistical measures (R and NSE) show accept-
able correlation between the predicted and observed vari-
ables. In Fig. 10, the over estimation of discharge is due to
the difference in the soil water conditions. Some points are
also under estimated as seen with respect to the 1:1 line in
Fig. 11. This is attributed to the fact that in practise, high
intensity and short duration rainfall events may generate
more sediment load than the simulated one on the basis of
daily rainfall data (Xu et al. 2009).

Analysis of results obtained from SWAT

simulation

After successful calibration and validation, SWAT model

is used in Hirakud watershed for further analysis of soil
erosion and sediment yield.
The simulated sediment yield depends on different
Fig. 6 Observed and simulated daily discharge during the calibration
period for the sub-watershed area a Rampur, b Bamnidih and factors such as physiographical characteristics of sub-wa-
c Baronda tersheds, contribution of these sub-watersheds in the stream

123
 Sustain. Water Resour. Manag.

flow reaching the outlet, rainfall intensity, quantity and

peak rate of surface runoff and sediment concentration.
Sub-watersheds having larger value of CN, PUSLE and
LSUSLE result in higher amount of runoff and sediment
yield than the sub-watersheds having lower values of these
parameters. Rainfall and runoff are the main factors for
causing detachment, transport and deposition of sediment
particles. A previous study (Qiu et al. 2012) has shown that
during the daily simulation, SWAT model computes the
same amount of runoff to the given amount of rainfall,
regardless of event duration or the distribution of rainfall
intensity and therefore, the simulated soil erosion and
sediment yield are also not sensitive to the rainfall inten-
sity. Easton et al. (2010) also concluded that on the basis of
daily simulation by SWAT model, rainfall intensity is
assumed to be inconsequential. Thus, Fig. 12a, b have been
plotted to show the seasonal evolution of the rainfall depth,
runoff and sediment yield and the contribution of various
water pathways in the total stream flow. It is observed from
the figure that the average peak rainfall in this watershed
occurs in July whereas the average peak runoff is in
August. Most of the sediment yield occurs from the surface
Fig. 8 Observed and simulated daily discharge during the validation runoff due to the intense rainfall during July to October,
period for the sub-watershed area a Rampur, b Bamnidih and whereas the peak of sediment yield occurs in the month of
c Baronda
September after the cessation of rainfall. A study by Turner
(2013) indicates that an increase or decrease in total rain-
fall over India due to possible changes to the Indian
monsoon due to climate change may be in the range of
5–10%, although some climate models suggest more and
some less. Hence, a 10% variation (increase or decrease) in
rainfall has been chosen to assess its impact on the simu-
lated data of sediment yield as it is a function of the sim-
ulated surface runoff and its peak value. The fluctuation of
simulated value of runoff and sediment yield with rainfall
variation shows that an increase of 10% in rainfall value
causes an increase of 38 and 68% in simulated runoff and
sediment yield respectively whereas 10% decrease in
rainfall results 5% decrease in simulated runoff and 7%
decrease in sediment yield.
Surface flow is defined as the overland flow whereas the
base flow is defined as the amount of water in a stream that
comes from groundwater discharge or seepage. Lateral
flow refers to the amount of water that infiltrates the soil
surface and moves laterally downslope. From Fig. 12b, it is
seen that most of the streamflow originates due to the
surface and lateral flow. At the beginning of rainy season
(June), lateral flow is more than the surface flow as less
amount of water is present in the soil initially. After con-
tinuation of rainfall, soil is saturated and surface flow
becomes higher than sub-surface flow.
Temporal variation of sediment yield and sediment
Fig. 9 Observed and simulated daily sediment load during the
validation period for the sub-watershed area a Basantpur, b Baronda, delivery ratio (SDR) in the entire watershed is shown in
c Jondhra and d Bamnidih Fig. 13a on monthly basis. The sediment yield varies from

123
Sustain. Water Resour. Manag.

Fig. 10 Scattergrams of the

observed and simulated daily
discharge during the calibration
and validation period for the
sub-watershed area a Rampur,
b Bamnidih and c Baronda

0.13 t/ha (May) to 4.37 t/ha (September) whereas SDR some runoff events are also observed due to the limited
ranges from 0.03 (July) to 0.1 (May). It shows that during local extent of certain precipitation of 27.79 mm which
the month of May, the value of sediment yield is minimum causes minimum sediment yield in the study area. Previous
whereas SDR value approaches to maximum. This inverse studies (Wu and Chen 2012; Shao et al. 2013) have also
relationship between the sediment yield and SDR can be shown that not just surface runoff affects sediment yield,
explained by the fact that the maximum sediment transport different topography factors (slope length and slope
capacity is reached by the stream and a greater proportion steepness) as well as practice factor within the HRUs also
of sediment deposition occurs. This aspect is also sup- affect the movement of sediment particles which may
ported by the inferences of Sean et al. (2013) who conclude cause minimum sediment yield in the dry period. It is
that the above condition may also occur when a greater apparent from Fig. 13b that low soil erosion occurs due to
proportion of eroded sediment is transported to the outlet low rainfall values and vice-versa but no definite correla-
and the transport capacity of the stream is not reached. tion is obtained from their relation.
Rainfall is the main driving force of sediment yield. Figure 14 shows the observed and simulated reservoir
Though there is not enough rainfall during the dry period, inflow during the period of 2000–2009. The observed data

123
 Sustain. Water Resour. Manag.

Fig. 11 Scattergrams of the

observed and simulated daily
sediment load during the
calibration and validation period
for the sub-watershed area
a Basantpur, b Baronda,
c Jondhra and d Bamnidih

of reservoir inflow was collected from the field visits to the between them. However, the model captures peak flows
Hirakud Research Station at Hirakud and Water Resources very well and the statistical values show a satisfactory
Department, Government of Odisha and supplemented correlation between the observed and simulated reservoir
with the data by Choudhury et al. 2012. It is found that the inflows. This study is helpful in determining the contribu-
simulated reservoir inflow is lower than the observed tion of each sub-watershed in the deposition of sediment in
inflow during the period which is due to the diversion of Hirakud reservoir.
water for agricultural uses and other human activities in the Different management techniques may be evaluated in
upper catchment of the reservoir (Mishra 2008; Dadhwala the watershed to minimize soil erosion and sediment yield.
et al. 2010). Irregular distribution of rainfall throughout the However, economic consideration is a major constraint to
basin also causes the difference between them. The con- implement management techniques. Therefore, identifica-
clusion drawn from the above observations is also reflected tion and prioritization of the critical erosion-prone areas in
in a Report on Mahanadi basin (2014). Initial retention and the whole watershed are very important. Based on the
soil water condition in the catchment may also affect the satisfactory results, SWAT was used to identify sub-wa-
observed values which ultimately cause the difference tersheds having statistically higher soil erosion. Following

123
Sustain. Water Resour. Manag.

Fig. 12 For Hirakud watershed

showing a correlation between
monthly distribution of rainfall,
runoff and sediment yield and
b 13 years average seasonal
evolution of the contribution of
different pathways to stream
water flow

the guidelines provided by Singh et al. (1992), the simu- of water yield are also susceptible to the highest value of
lated annual average soil erosion over the period of soil erosion because of their physiographic characteristics.
2000–2012 of Hirakud watershed was divided into six The total sediment yields of the whole watershed is
categories: slight (\5 t/ha/year), moderate (5–10 t/ha/ divided into three classes (\5, 5–20 and [20 t/ha/year) to
year), high (10–20 t/ha/year), very high (20–40 t/ha/year), facilitate the representation of their differences. It is
severe (40–80 t/ha/year) and very severe ([80 t/ha/year) as noticed that the sub-watersheds 2, 3, 6, 11, 13, 14, 20 and
shown in Fig. 15. As may be noticed, high rate of soil 22 have the highest rate of soil erosion as well as highest
erosion occurs mostly in the sub-watershed areas where rate of sediment yield due to the steep slopes and excessive
deforestation has taken place and soil is mostly coarse livestock. The highest value of annual sediment yield at the
loamy. At sub-watershed level, 48% of the watershed area sub-watershed level is 102.20 t/ha/year from the agricul-
contributes to half of the total soil erosion of the Hirakud tural lands whereas the average annual value of sediment
watershed and 34% area falls under the class of high soil yield for the whole watershed is about 17.61 t/ha/year. The
erosion ([10 t/ha/year) or above. The average soil loss in sub-watersheds having lower slopes and more forested
Hirakud watershed is 181.73 t/ha/year. Within the sub- areas have low value of sediment yield. This suggests that
watersheds, differences in surface runoff and soil erosion the prediction of sediment yield is highly sensitive to the
are observed due to the combined influence of spatial sizes of different sub-basins due to the sensitivity of
variation of land use and soil texture. The erosion rates are topographic factors. This is due to the fact that the MUSLE
seen to be the lowest at the outlet of watershed due to the equation is a function of the peak rate of runoff which, in
lowest slope and highest infiltration capacity of the soil. turn, is a function of time of concentration, a variable
The analysis of this result helps the soil and water con- depending on the channel length from the remotest point to
servation managers to take proper decisions for preventing the sub-basin outlet. The other reason is that the sediment
soil loss by taking biological or engineering measures. It is routing through streams and channels, which is a function
observed that the sub-watersheds having the highest value of channel length and its cross sections, is also affected by

123
 Sustain. Water Resour. Manag.

Fig. 13 For Hirakud watershed

showing a monthly sediment
yield superimposed with
monthly sediment delivery ratio
and b changes of rainfall and
soil erosion over year

characteristics etc. It is found that the value of SDR is more

near the watershed outlet because there is a tendency to
transport all the eroded sediment load at the watershed
outlet. However, few sub-watersheds (sub-watershed 5 and
17) near the outlet have low SDR as they do not transport
significant sediment load to the watershed outlet due to low
generation of sediment load. During the simulation, it is
noticed that increasing the value of PUSLE decreases SDR,
as the proportion of sediment that can be transported to the
Fig. 14 Observed and simulated value of Hirakud reservoir inflow outlet decreases. For a particular PUSLE, the stream has
constant sediment transport capacity and generation of
the sub-basin size. Similar conclusions have also been sediment load decreases, the fraction of sediment trans-
reported by Setegn et al. (2010). ported to the outlet increases.
Identification of sub-watersheds with the highest SDR In this study, the authors have demonstrated the use of
will help in undertaking control measures within the empirical equations and physically based SWAT model
watershed to reduce sediment yield at the watershed outlet. (Table 5) for finding SDR in the Hirakud watershed. The
It is found that all the sub-watersheds do not yield high advantage of using SWAT model for predicting SDR is
sediment load due to the low surface runoff generation. that it takes into account the physiographic characteristics
Table 4 represents the classification of Hirakud watershed and detailed spatial data over the total watershed, while the
depending on their SDR values. The variation in SDR is other methods include lumping of temporal and spatial data
mainly due to the differences in land use–land cover, soil of sediment transport without considering the effect of
texture, size and slope of sub-watershed and reach changing of climate, LULC etc. (de Vente et al. 2007).

123
Sustain. Water Resour. Manag.

Table 5 Comparison between various sediment delivery ratio (SDR)

methods
SDR method Value of SDR

Boyce (1975) 0.16

Renfro (1975) 0.14
Vanoni (1975) 0.11
USDA-SCS (1979) 0.16
SWAT 0.10

they are easily detached and susceptible to transportation

by runoff.
Figure 16b shows the effect of three major land use
classes on soil erosion. It is evident that for agriculture
lands, 80.48% area falls under the soil erosion class of high
or above. For the deciduous and evergreen forest regions,
value is 63.08 and 55.06% ,respectively. It is clear that
agricultural land is the major contributing factor for soil
erosion within the catchment. Due to unsustainable agri-
cultural practices, the soil breaks up into finer particles
which are easily available for transport by surface runoff.
The rate of erosion is therefore the highest in agricultural
lands as the soil is not fully covered by protective layer of
plants. It also destroys the root system of plants that help to
hold the soil particles together.
Fig. 15 Soil erosion zones of different sub-watersheds in study area Figure 16c shows the areas of different soil erosion
classes under different combinations of land use and soil
Table 4 Classification of Hirakud watershed according to the value type. It is evident that the area under agricultural land
of SDR with coarse loamy classification is mostly sensitive to
SDR Value of SDR Sub-basins no. soil erosion. This figure helps us to make a conclusion
classification that soil erosion in different sub-watersheds occur in
different ranges due to their characteristic soil properties
Low \0.1 2, 9,10, 11, 13, 17,18, 19, 20, 21
and land use. This phenomenon has also been reported
Medium 0.1–0.15 1, 3, 4, 6, 8, 12,14, 15, 22, 23
by Kumar et al. (2015), who remark that erosion is quite
High [0.15 5, 7, 16
sensitive to changes in landuse and soil type within the
watershed.
Hence, it is seen that SWAT model may be used with a
great accuracy for adopting effective measures in the
watershed by predicting SDR at the sub-watershed scale to Conclusions
reduce soil erosion at the respective sub basin area.
Figure 16a shows the varying effects of different com- This study is a preliminary attempt of hydrological simu-
binations of soil type on soil erosion. It shows that 92.23% lation in Hirakud watershed to simulate daily discharge and
of total coarse loamy soil contributes high or above soil sediment load. The objective of this paper was to identify
erosion, whereas for fine, clayey and fine loamy these the most soil erosion-prone area in the whole Hirakud
values are 40.87, 36.87, and 4.27% respectively. It is watershed. The availability of limited data input degraded
concluded that coarse loamy soil contributes mostly the model simulation to some extent. The calibrated
towards high or above soil erosion in the whole watershed parameters of sediment load such as CH_COV,
as it contains higher percentage of sand and silt, relatively CH_EROD, SPCON and SPEXP cannot be measured
smaller amounts of clay and organic carbon. For very high physically in real field condition. So they are calibrated by
rainfall intensities coarse-textured soils cause soil erosion trial and error procedure or by reviewing the literatures
in spite of having high hydraulic conductivity. This is due published on it until the goal is reached. In spite of this,
to weak soil structure and poor aggregation. Therefore, graphical as well as statistical comparisons showed that

123
 Sustain. Water Resour. Manag.

Fig. 16 a Area of soil erosion

w. r. to different soil class,
b area of soil erosion w. r. to
different land use and c area of
soil erosion w. r. to different
combinations of soil and land
use type

there is a good agreement between the simulated and SWAT is unable to identify the reasons of occurring soil
observed data. However, SWAT overestimates runoff due erosion. For further studies, it is recommended to use sub-
to the inaccurate estimation of CN2 value during high flow daily data and to prioritize soil erosion at HRU level for the
events. In SWAT, CN2 value is dependent on soil moisture improvement of watershed management.
content of entire soil profile which is very difficult to Implementation of management practices across the
predict during high flood events. Therefore it is suggested total watershed is very time consuming, laborious and
that SWAT can be upgraded by modifying CN2 value costly. Hence, it is very essential to recognize the most
depending on moisture content of top-most layer soil only. erosion-prone area so that the measures may be undertaken
It then helps in predicting accurately runoff during high locally. This study helps us to identify critical erosion-
flow as well as low flow periods. SWAT can also be prone area in the whole Hirakud watershed with the iden-
improved by taking into consideration duration of rainfall tification of soil type and land use condition whose prop-
to simulate peak flow rate during flooding condition. erties are mostly responsible to cause erosion. The
From the analysis it is found that 34% area of total outcomes of this study indicates that several controlling
catchment falls under high or above soil erosion with measures like land contouring, terracing and planting of
combination of coarse loamy type of soil and agricultural certain trees may help in reducing the erosion from dif-
type of land use and land cover condition. However, ferent sub-basins. Sediment traps such as small check dams

123
Sustain. Water Resour. Manag.

and sediment detention basins can also be constructed to Action research report, Odisha State Resource Centre, Forum for
tackle soil degradation into the watershed. Policy Dialogue on Water Conflicts in India
CWC (2001) Compendium on silting of reservoirs in India. Water
This paper attempts to demonstrate the effective use of Planning and Projects Wing, Environment Management Organ-
the SWAT model in a typical watershed of India which isation, Watershed and Reservoir Sedimentation Directorate,
may encourage other researchers to implement it in other CWC, Govt of India
watersheds for planning effective management techniques Dadhwala VK, Aggarwal SP, Mishra N (2010) Hydrological simu-
lation of Mahanadi river basin and impact of land use/land cover
for reducing soil erosion and sediment load transportation, change on surface runoff using a macro scale hydrological
ultimately helping to reduce sedimentation in the reservoirs model. ISPRS TC VII symposium—100 years ISPRS, Vienna,
existing or planned along the stream. Austria. IAPRS XXXVIII(Part 7B):165–170
de Roo APJ, Wesseling CG, Cremers NHDT, Offermans RJE,
Acknowledgements The author gratefully acknowledges the Ritsema CJ, Van Oostindie K (1994) LISEM: a new physically-
Department of Science and Technology, Government of India, for based hydrological and soil erosion model in a GIS-environment:
funding the research project. Tha–nks are also due to the Water theory and implementation. Var Stream Eros Sediment Transp
Resources Department, Government of Odisha, providing the sedi- 224:439–448
mentation reports of the Hirakud reservoir and to the Central Water de Vente J, Poesen J, Arabkhedri M, Verstraeten G (2007) The
Commission, Bhubaneswar, for providing relevant discharge and sediment delivery problem revisited. Prog Phys Geogr
sediment data. 31(2):155–178
Debele B, Srinivasan R, Parlange J (2009) Hourly analyses of
hydrological and water quality simulations using the ESWAT
model. Water Resour Manag 23(2):303–324
DHI (Danish Hydraulic Institute) (1999) SEAGIS: Soil Erosion
References Assessment using GIS. Documentation and user’s guide, version
1.48
Abbaspour KC, Yang J, Maximov I, Siber R, Bogner K, Mieleitner J, DLWC (Department of Land and Water Conservation) (1999)
Zobrist J, Srinivasan R (2007) Modeling hydrology and water Integrated water Quantity-Quality Model (IQQM) user manual.
quality in the pre-alpine/alpine Thur watershed using SWAT. Centre for Natural Resources, NSW Dept. of Land and Water
J Hydrol 333:413–430 Conservation
Allen RG, Jensen ME, Wright JL, Burman RD (1989) Operational Droogers P, Bastiaanssen WGM (2002) Irrigation performance using
estimates of reference evapotranspiration. Agron J hydrological and remote sensing modeling. J Irrig Drain Eng
81(4):650–662 128(1):11–18
Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area Dutta S (2016) Soil erosion, sediment yield and sedimentation of
hydrologic modeling and assessment Part 1: model development. reservoir: a review. Model Earth Syst Environ 2(123):1–18
J Am Water Resour Assoc 34(1):73–89 Dutta S, Sen D (2016) Sediment distribution and its impacts on
Babar S, Ramesh H (2014) Streamflow response to land use land Hirakud Reservoir (India) storage capacity. Lakes Reserv Res
cover change over the Nethravathi river basin, India. J Hydrol Manag 21:245–263
Eng 20(10):1–11 Easton ZM, Fuka DR, White ED, Collick AS, Ashagre BB,
Bagnold RA (1977) Bedload transport in natural rivers. Water Resour McCartney M, Awulachew SB, Ahmed AA, Steenhuis TS
Res 13:303–312 (2010) A multi basin SWAT model analysis of runoff and
Beasley DB, Huggins LF, Monke EJ (1980) A model for watershed sedimentation in the Blue Nile, Ethiopia. Hydrol Earth Syst Sci
planning. Tran ASAE 23:938–944 14:1827–1841
Bennett JP (1974) Concepts of mathematical modeling of sediment Elirehema YS (2001) Soil water erosion modeling in selected
yield. Water Resour Res 10:485–492 watersheds in Southern Spain. IFA 2001, ITC Enschede, The
Beskow S, Mello CR, Norton LD, Curi N, Viola MR, Avanzi JC Netherlands
(2009) Soil erosion prediction in the Grande river basin, Brazil, Elwell HA (1978) Soil loss estimation: compiled works of the
using distributed modeling. CATENA 79:49–59 Rhodesian multi-disciplinary team on soil loss estimation.
Bogen J, Bonsnes TE (2003) Erosion and sediment transport in high Institute of Agricultural Engineers, Salisbury
Arctic rivers, Svalbard. Polar Res 22(2):175–189 FAO (1995) Food and Agriculture Organization of the United
Boyce RC (1975) Sediment routing with sediment delivery ratios. Nations. Digital soil map of the world and derived soil
Present and Prospective Technology for Predicting Sediment properties, Rome
Yields and Sources, US Department of Agriculture Publication Feng X, Wang Y, Chen L, Fu B, Bai G (2010) Modeling soil erosion
ARS-S-40, pp 61–65 and its response to land-use change in hilly catchments Chinese
Cao W, Bowden WB, Davie T, Fenemor A (2009) Modeling impacts Loess Plateau. Geomorp 118:239–248
of land cover change on critical water resources in the Motueka Ferguson RI (1986) River loads underestimated by rating curves.
river catchment, New Zealand. Water Resour Manag Water Resour Res 22(1):74–76
23:137–151 Flacke W, Auerswald K, Neufang L (1990) Combining a modified
Chakrapani GJ, Subramanian V (1990) Factors controlling sediment universal soil loss equation with a digital terrain model for
discharge in the Mahanadi basin, India. J Hydrol 117:169–185 computing high resolution maps of soil loss resulting from rain
Chakrapani GJ, Subramanian V (1993) Rate of erosion in the wash. CATENA 17:383–397
Mahanadi river basin, India. J Hydrol 149:39–48 Foster GR, Yoder DC, Weesies GA, McCool DK, McGregor KC,
Chaplot V (2005) Impact of DEM mesh size and soil map scale on Binger RL (2002) RUSLE2 user’s guide. USDA-Agricultural
SWAT runoff, sediment, and NO3–N, loads predictions. J Hydrol Research Service, Washington, pp 1–76
312:207–222 Ganasri BP, Ramesh H (2015) Assessment of soil erosion by RUSLE
Choudhury P, Sandbhor J, Satapathy P (2012) Floods, fields and model using remote sensing and GIS—a case study of Nethra-
factories: towards resolving conflicts around the Hirakud dam. vathi basin. Geosci Front 7:953–961

123
 Sustain. Water Resour. Manag.

Geza M, McCray JE (2008) Effects of soil data resolution on SWAT Labs Clark (2001) Idrisi 32, Release 2, Tutorial manual. Idrisi
model stream flow and water quality predictions. J Environ Production, USA
Manag 88(3):393–406 Li XH, Yang J, Zhao CY, Wang B (2012) Runoff and sediment from
Green WH, Ampt GA (1911) Studies on soil physics, Part I, The flow orchard terraces in southeastern China. Land Degrad Dev
of air and water through soils. J Agric Sci 4(1):1–24 25:184–192
Griensven AV, Popescu I, Abdelhamid MR, Ndomba PM, Beevers L, Licciardello F, Rossi C, Srinivasan R, Zimbone SM, Barbagallo S
Betrie GD (2013) Comparison of sediment transport computa- (2011) Hydrologic evaluation of a Mediterranean watershed
tions using hydrodynamic versus hydrologic models in the using the SWAT model with multiple PET estimation methods.
Simiyu River in Tanzania. Phys Chem Earth 61–62:12–21 Trans ASABE 54(5):1615–1625
Hanley N, Faichney R, Munro A, Shortle JS (1998) Economic and Littleboy M, Silburn DM, Freebairn DM, Woodruff DR, Hammer GL,
environmental modeling for pollution control in an estuary. Leslie JK (1992) Impact of soil erosion on production in
J Environ Manag 52:211–225 cropping systems. I. Development and validation of a computer
Hargreaves GH, Samani ZA (1985) Reference crop evapotranspira- simulation model. Aust J Soil Res 30:757–774
tion from temperature. Appl Eng Agric Sci 4(1):1–24 Mahmood K (1987) Reservoir sedimentation: impact, extent, mitiga-
Hunink JE, Niadas IA, Antonaropoulos P, Droogers P, Vente JD tion. World Bank Technical Report No. 71, Washington, DC
(2013) Targeting of intervention areas to reduce reservoir Manaswi CM, Thawait AK (2014) Application of soil and water
sedimentation in the Tana catchment (Kenya) using SWAT. assessment tool for runoff modeling of Karam river basin in
Hydrol Sci J 58(3):600–614 Madhya Pradesh. Int J Sci Eng Technol 3(5):529–533
Jha MK, Gassman PW, Arnold JG (2007) Water quality modeling for Merritt WS, Letcher RA, Jakeman AJ (2003) A review of erosion
the Raccoon river watershed using SWAT. Trans ASABE transport models. Environ Model Softw 18:761–799
50(2):479–493 Mishra N (2008) Macroscale hydrological modeling and impact of
Jha MK, Wolter CF, Schilling KE, Gassman PW (2010) Assessment landcover change on streamflows of the Mahanadi river basin.
of total maximum daily load implementation strategies for A Master thesis submitted to Andhra University, Indian Institute
nitrate impairment of the Raccoon river, Iowa. J Environ Qual of Remote Sensing (National Remote Sensing Agency) Dept. of
39:1317–1327 Space, Govt. of India
Johanson RC, Imhoff JC, Davis HH (1980) User manual for the Hydrologic Mishra A, Froebrich J, Gassman PW (2007) Evaluation of the model
Simulation Program-Fortran (HSPF). U.S. Environmental Protection for assessing sediment control structures in a small watershed in
Agency, Office of Research and Development, Environmental India. Trans ASABE 50(2):469–477
Research Laboratory, Athens, EPA-600/9-80-105, p 678 Misra RK, Rose CW (1990) Manual for use of program GUEST. Div.
Johnson YW (1943) Distribution graphs of suspended-matter con- of Australian Environmental Studies Rep. Griffith University,
centration. Trans ASCE 108:941–964 Brisbane
Kannan N, White SM, Worrall F, Whelan MJ (2007) Sensitivity Morgan RPC, Quinton JN, Smith RE, Govers G, Poesen JWA,
analysis and identification of the best evapotranspiration and Auerswald K, Chisci G, Torri D, Styczen ME (1998) The
runoff options for hydrological modeling in SWAT-2000. European Soil Erosion Model (EUROSEM): a process-based
J Hydrol 332:456–466 approach for predicting sediment transport from fields and small
Kaur R, Singh O, Srinivasan R, Das SN, Mishra K (2004) catchments. Earth Surf Process Landf 23:527–544
Comparison of a subjective and a physical approach for Moriasi DN, Arnold JG, Van Liew MW, Bingner RL, Harmel RD,
identification of priority areas for soil and water management Veith TL (2007) Model evaluation guidelines for systematic
in a watershed—a case study of Nagwan watershed in Hazarib- quantification of accuracy in watershed simulation. Trans
agh District of Jharkhand, India. Environ Model Assess ASABE 50(3):885–900
9(1):115–127 Morris GL, Fan J (1998) Reservoir sedimentation handbook. McGraw
Kim NW, Lee J (2008) Temporally weighted average curve number Hill, New York
method for daily runoff simulation. Hydrol Process Mukherjee S, Veer V, Tyagi SK, Sharma V (2007) Sedimentation
22(25):4936–4948 study of Hirakud reservoir through remote sensing techniques.
Kirkby MJ, Jones RJA, Irvine B, Gobin A, Govers G, Cerdan O, Van J Spatial Hydrol 7(1):122–130
Rompaey AJJ, Le Bissonnais Y, Daroussin J, King D, Mon- Mukhtar MA, Dunger V, Merkel B (2014) Assessing the impacts of
tanarella L, Grimm M, Vieillefont V, Puigdefabregas J, Boer M, climate change on hydrology of the upper reach of the spree
Kosmas C, Yassoglou N, Tsara M, Mantel S, Van Lynden G J and river: Germany. Water Resour Manag 28:2731–2749
Huting J (2004) Pan-European Soil Erosion Risk Assessment: The Musgrave GW (1947) The quantitative evaluation of factors in water
PESERA map, version 1 October 2003 Explanation of Special erosion: a first approximation. J Soil Water Conserv 2:133–138
Publication Ispra 2004 No. 73 (S.P.I. 04.73), European Soil Bureau Nash JE, Sutcliffe JV (1970) River flow forecasting through
Research Rep. No. 16, EUR 21176 and map in ISO B1 format, conceptual models: part I. A discussion of principles. J Hydrol
European Comm. Joint Research Centre, Ispra/Institute for 10(3):282–290
Environment and sustainability. Available from: Office for Ndomba P, Mtalo F, Killingtveit A (2008) SWAT model application
Official Publ. of the European Communities, Luxembourg in a data scarce tropical complex catchment in Tanzania. Phys
Knisel WG (1980) CREAMS: a field scale model for chemicals, Chem Earth 33(8–13):626–632
runoff and erosion from agricultural management systems. Nearing MA, Foster GR, Lane LJ, Finkner SC (1989) A process-
Conservation Research Rep. No. 26, U.S. Dept. of Agriculture based soil erosion model for USDA-water erosion prediction
Kumar S, Mishra A, Raghuwanshi NS (2014) Identification of critical project technology. Trans ASAE 32(5):1587–1593
erosion watersheds for control management in data scarce Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2011) Soil and Water
condition using the SWAT model. J Hydrol Eng 20(6):1–8 Assessment Tool theoretical documentation version 2005. Texas
Kumar S, Raghuwanshi NS, Mishra A (2015) Identification and A&M University, College Station
management of critical erosion watersheds for improving O’Loughlin EM (1986) Prediction of surface saturation zones in
reservoir life using hydrological modeling. Sustain Water Resour natural catchments by topographic analysis. Water Resour Res
Manag 1:57–70 22:794–804

123
Sustain. Water Resour. Manag.

Overton D (1966) Muskingum flood routing of upland streamflow. Sardar B, Singh AK, Raghuwanshi NS, Chatterjee C (2013)
J Hydrol 4:185–200 Hydrological modeling to identify and manage critical erosion
Pandey A, Chowdary VM, Mal BC (2007) Identification of critical prone areas for improving reservoir life-a case study of Barakar
erosion prone areas in the small agricultural watershed using river basin. J Hydrol Eng 19(1):196–204
USLE, GIS and remote sensing. Water Resour Manag Schiefer E, Petticrew EL, Immell R, Hassan MA, Sonderegger DL
21:729–746 (2013) Land use and climate change impacts on lake sedimen-
Pandey BK, Gosain AK, Paul G, Khare D (2016) Climate change tation rates in western Canada. Anthropocene 3:61–71
impact assessment on hydrology of a small watershed using Schmidt J (1991) A mathematical model to simulate rainfall erosion.
semi-distributed model. Appl Water Sci, pp 1–13 Erosion, transport and deposition processes—theories and mod-
Panigrahi B, Paul JC (2015) Managing drainage congestion to els. CATENA 19:101–109
increase crop production and productivity in Hirakud command, Sean A, Woznicki A, Nejadhashemi P (2013) Spatial and temporal
India. J Agric Eng Biotechnol 3(1):32–40 variabilities of sediment delivery ratio. Water Resour Manag
Parajuli PB, Nelson NO, Frees LD, Mankin KR (2009) Comparison of 27:2483–2499
AnnAGNPS and SWAT model simulation results in USDA- Setegn SG, Srinivasan R, Dargahi B, Melesse AM (2009) Spatial
CEAP agricultural watersheds in south-central Kansas. Hydrol delineation of soil erosion vulnerability in the Lake Tana basin,
Process 23:748–763 Ethiopia. Hydrol Process 23:3738–3750
Prabhanjan A, Rao EP, Eldho TI (2015) Application of SWAT model Setegn SG, Dargahi B, Srinivasan R, Melesse AM (2010) Modeling
and geospatial techniques for sediment-yield modeling in of sediment yield from Anjeni-gauged watershed, Ethiopia using
ungauged watersheds. J Hydrol 20:1–6 SWAT model. J Am Water Resour Assoc 46(3):514–526
Priestley CHB, Taylor RJ (1972) On the assessment of surface heat Shao Y, Lunetta RS, Macpherson AJ, Luo J, Chen G (2013)
flux and evaporation using large-scale parameters. Mon Weather Assessing sediment yield for selected watersheds in the Lauren-
Rev 100(2):81–92 tian Great Lakes basin under future agricultural scenarios.
Prosser IP, Young WJ, Rustomji P, Hughes AO, Moran CJ (2001c) A Environ Manag 51:59–69
model of river sediment budgets as an element of river health Shrestha NK, Leta OT, De Fraine B, Griensven AV, Bauwens W
assessment. In: Proceedings of the international congress on (2013) Open MI-based integrated sediment transport modeling
modeling and simulation (MODSIM’2001), pp 861–866 of the river Zenne, Belgium. Environ Model Softw 47:1–14
PSIAC (Pacific Southwest Inter Agency Committee) (1968) Rep. of Singh G, Ram B, Pratap N, Bhushan LS, Abrol IP (1992) Soil erosion
the water management subcommittee on factors affecting rate in India. J Soil Water Conserv 47:97–99
sediment yield in the Pacific Southwest area and selection and Singh J, Knapp HV, Arnold JG, Demissie M (2007) Hydrological
evaluation of measures for reduction of erosion and sediment modeling of the Iroquois river watershed using HSPF and
yield. Sacramento, California: Pacific Southwest Region Plan- SWAT. In: AWRA spring specialty conference on agricultural
ning Office, U.S. Dept. of the Interior, ASCE, vol 98, p l0 hydrology and water quality, Kansas City, pp 343–360
Qiu L, Zheng F, Yin R (2012) SWAT-based runoff and sediment Singh HV, Kalin L, Srivastava P (2011) Effect of soil data resolution
simulation in a small watershed, the loessial hilly-gullied region on identification of critical source areas of sediment. J Hydrol
of China: capabilities and challenges. Int J Sediment Res Eng 16(3):253–262
27(2):226–234 Smith RE, Goodrich DC, Quinton JN (1995) Dynamic, distributed
Rathore DS, Choudhary A, Agarwal PK (2006) Assessment of simulation of watershed erosion: the KINEROS2 and EURO-
sedimentation in Hirakud reservoir using digital remote sensing SEM models. J Soil Water Conserv 50:517–520
technique. J Indian Soc Remote Sens 34(4):377–383 Song S, Schmalz B, Fohrer N (2015) Simulation, quantification and
Renard KG, Laursen EM (1975) Dynamic behavior model of comparison of in-channel and floodplain sediment processes in a
ephemeral streams. J Hydrol Div 101:511–526 lowland area—a case study of the Upper Stor catchment in
Renard KG, Foster GR, Weesies GA, McCool DK, Yoder DC (1996) northern Germany. Ecol Ind 57:118–127
RUSLE user’s guide. Soil Water Conserv Soc, p 173 Srinivasan R, Zhang X, Arnold JG (2010) SWAT ungauged:
Rendon-Herrero O (1978) Unit sediment graph. Water Resour Res hydrological budget and crop yield predictions in the upper
14:889–901 Mississippi river basin. Trans ASBE 53(5):1533–1546
Renfro GW (1975) Use of erosion equations and sediment delivery Takken I, Beuselinck L, Nachtergaele J, Govers G, Poesen J, Degraer
ratios for predicting sediment yield. Present and Prospective G (1999) Spatial evaluation of a physically-based distributed
Technology for Predicting Sediment Yields and Sources, US erosion model (LISEM). CATENA 37:431–447
Department of Agriculture Publication ARS-S-40, pp 33–45 Talebizadeh M, Morid S, Ayyoubzadeh SA, Ghasemzadeh M (2009)
Report on Mahanadi basin, Version 2.0 (2014). Government of India, Uncertainty analysis in sediment load modeling using ANN and
Ministry of Water Resources SWAT model. Water Resour Manag 24:1747–1761
Rose CV, Williams JR, Sander GC, Barry DA (1983) A mathematical Tang Z, Engel BA, Pijanowski BC, Lim KJ (2005) Forecasting land
model of soil erosion and deposition process. I. Theory for a use change and its environmental impact at a watershed scale.
plane element. Soil Sci Soc Am J 47:991–995 J Environ Manag 76:35–45
Rose NL, Morley D, Appleby PG, Battarbee RW, Alliksaar T, Tripathi MP, Panda RK, Raghuwanshi NS (2003) Identification and
Guilizzoni P, Jeppesen E, Korhola A, Punning J (2011) Sediment prioritization of critical sub-watersheds for soil conservation
accumulation rates in European lakes since AD 1850: trends, management using SWAT model. Biosyst Eng 85(3):365–379
reference conditions and exceedance. J Paleolimnol 45:447–468 Turner A (2013) The Indian monsoon in a changing climate. http://
Saghravani SR, Mustapha SA, Ibrahim S, Randjbaran E (2009) www.rmets.org/weather-and-climate/climate/indian-monsoon-
Comparison of daily and monthly results of three evapotranspi- changing-climate
ration models in tropical zone: a case study. Am J Environ Sci USDA-SCS (1979) United States Department of Agriculture-Soil Conser-
5(6):698–705 vation Service. National Engineering Handbook, Sec 4. Hydrology
Sahu GC, Das GN (2012) Hirakud: its background and performance. USEPA (U.S. Environmental Protection Agency) (1994)
Indian National Committee on Irrigation and Drainage (INCID), SWRRBWQ: window’s interface users guide. U.S. Environ-
Constituted by MOWR, Government of India, New Delhi, p 474 mental Protection Agency, Washington, DC

123
 Sustain. Water Resour. Manag.

Van Griensven A, Meixner T (2006) Methods to quantify and identify Tex.: USDA-ARS Blackland Research Center, Texas Agricul-
the sources of uncertainty for river basin water quality models. tural Experiment Station, and Grassland, Soil and Water
Water Sci Technol 53(1):51–59 Research Laboratory
Vanoni VA (1975) Sediment deposition engineering. ASCE Manuals Wischmeier WH, Smith DD (1965) Predicting rainfall-erosion losses
and Reports on Engineering Practices, p 54 from cropland east of the Rocky Mountains: a guide for selecting
Verstraeten G, Poesen J (1999) The nature of small-scale flooding, practices for soil and water conservation. USDA Agricultural
muddy floods and retention pond sedimentation in central Handbook No. 282
Belgium. Geomorphology 29:275–292 Wischmeier WH, Smith DD (1978) Predicting rainfall erosion losses:
Vertessey RA, Watson FGR, Rahman JM, Seaton SP, Chiew FH, a guide to conservation planning. USDA Agriculture Handbook
Scanlon PJ, Marston FM, Lymbuner L, Jeanelle S, Verbunt M No. 537, US Govt. Printing Office, Washington, DC
(2001) New software to aid water quality management in the Woolhiser DA, Smith RE, Goodrich DC (1990) KINEROS, a
catchments and waterways of the south-east Queensland region. kinematic runoff and erosion model: documentation and user
In: Proceedings of the third Australian stream management manual U.S. Dept. of Agriculture, Agricultural Research
conference, vol 2, Brisbane, pp 611–616 Service, ARS-77, p 130
Viney NR, Sivapalan M (1999) A conceptual model of sediment Wu Y, Chen J (2012) Modeling of soil erosion and sediment transport
transport: application to the Avon river basin in Western in the East rver basin in southern China. Sci Total Environ
Australia. Hydrol Process 13:727–743 441:159–168
Walling DE (1983) The sediment delivery problem. J Hydrol Xu ZX, Pang JP, Liu CM, Li JY (2009) Assessment of runoff and
65(1–3):209–237 sediment yield in the Miyun reservoir catchment by using
Water Resources Department, Government of Odisha (2007) Report SWAT model. Hydrol Process 23:3619–3630
of the high level technical committee to study various aspects of Young RA, Onstad CA, Bosch DD, Andersen WP (1989) AGNPS: a
water usage for Hirakud reservoir nonpoint-source pollution model for evaluating agricultural
Williams JR (1969) Flood routing with variable travel time or watersheds. J Soil Water Conserv 44:168–173
variable storage coefficients. Trans ASAE 12(1):100–103 Zhang X, Srinivasan R, Hao F (2007) Predicting hydrologic response
Williams JR (1978) A sediment graph model based on an instanta- to climate change in the Luohe river basin using the SWAT
neous unit sediment graph. Water Resour Res 14:659–664 model. Trans ASABE 50(3):901–910
Williams JR, Berndt HD (1972) Sediment yield computed with Zhang X, Srinivasan R, Debele B, Hao F (2008) Runoff simulation of
universal equation. J Hydraul Div 98(12):2087–2098 the headwaters of the Yellow river using the SWAT model with
Williams JR, Jones CA, Dyke PT (1984) A modeling approach three snowmelt algorithms. J Am Water Resour Assoc
determining the relationship between erosion and soil produc- 44(1):48–61
tivity. Trans ASAE 27:129–144
Winchell M, Srinivasan R, Di Luzio M and Arnold JG (2007)
ArcSWAT Interface for SWAT 2005: user’s guide. Temple,

123