runoff and sediment yield modeling from a small agricultural watershed in india using the wepp model

Journal of Hydrology (2008) 348, 305–319

ava i lab le a t www.sc iencedi rec t . com

journal homepage: www.elsevier .com/ locate / jhydrol

Runoff and sediment yield modeling from a smallagricultural watershed in India using the WEPP model

Ashish Pandey a,*, V.M. Chowdary b, B.C. Mal c, M. Billib d

a Department of Water Resources Development and Management, IIT, Roorkee 247 667, Indiab CSEAS, Kyoto University, Japanc Department of Agricultural and Food Engineering, IIT, Kharagpur 721 302, Indiad Institute of Water Resources Management, Hydrology and Agricultural Hydraulic Engineering, University of Hannover,Germany

Received 30 July 2006; received in revised form 1 October 2007; accepted 2 October 2007

00do

*

fa

erbm(M

KEYWORDSGIS;Hydrological modeling;Remote sensing;Watershed;WEPP model

22-1694/$ - see front mattei:10.1016/j.jhydrol.2007.10

Corresponding author. Telx: +91 1332 271073.E-mail addresses: ashish

net.in (A. Pandey), [email protected] (B.. Billib).

r ª 200.010

.: +91 13

_nerist@ry_isro@yC. Mal),

Summary The WEPP (Water Erosion Prediction Project) watershed model was calibratedand validated for a small hilly watershed (Karso) of India. Sensitivity analysis of the modelwas carried out for the input parameters. The analysis shows that the sediment yield ishighly sensitive to interrill erodibility and effective hydraulic conductivity, whereas, run-off is sensitive to effective hydraulic conductivity only. Initially, the model was calibratedusing data from the 1996 monsoon season and subsequently its performance wasevaluated by estimating the daily runoff and sediment yield using the monsoon seasondata of different years. Coefficient of determination (R2) (0.86–0.91), Nash–Sutcliffesimulation model efficiency (0.85–0.95), and percent deviation values (7.90–15.15) indi-cate accurate simulation of runoff from the watershed. Performance of the WEPP modelfor simulation of sediment yield was also evaluated. High value of coefficient of determi-nation (R2) (0.81–0.95), Nash–Sutcliffe simulation model efficiency (0.78–0.92) and per-cent deviation values (4.43–19.30) for sediment yield indicate that the WEPP model canbe successfully used in the upper Damodar Valley, India.ª 2007 Elsevier B.V. All rights reserved.

7 Elsevier B.V. All rights reserved

32 285872;

yahoo.co.in, [email protected] (V.M. Chowdary),[email protected]

Introduction

Reliable prediction of the quantity and rate of runoff andsediment from land surface into streams and rivers is diffi-cult, expensive and time consuming. In India, an estimated175 Mha of land constituting about 53% of the total geo-graphical area suffers from deleterious effect of soil erosionand other forms of land degradation (Reddy, 1999). Active

.

mailto:[email protected]

mailto:ashisfwt@iitr.




306 A. Pandey et al.

erosion caused by water and wind alone accounts for150 Mha of land, whereas 25 Mha has been degraded dueto ravine/gullies, shifting cultivation, salinity/alkalinity,and water logging (Reddy, 1999). At the same time, avail-ability of accurate runoff and sediment yield data is scar-cely available at few selected places. Hence, thisnecessitates the simulation of processes like runoff andtransport of sediment as well as pollutants from watershedsthrough hydrological modeling. Estimation of runoff andsediment yield is necessary for developing watershed man-agement plans involving soil and water conservation mea-sures. Thus, research in hydrological modeling and relatedwatershed planning issues form a strong component of theenvironmental activities. During the last three decades,researchers have developed hydrological models of empiri-cal or conceptual nature for prediction of hydrological vari-ables. Hydrological models like SWAT (soil and waterassessment tool) (Arnold et al., 1993), AGNPS (agriculturalnon-point source pollution) (Young et al., 1989), ANSWERS(areal non-point source watershed environment responsesimulation) (Beasley et al., 1980) and WEPP (Water ErosionPrediction Project) (Laflen et al., 1991) are being exten-sively used for sustainable development of watersheds.Thus, hydrological and water quality models provide the ba-sis for improved understanding of hydrological processesand also for assessing the impact of human activities onenvironment and agricultural production. A major limitationin hydrology is the lack of availability of adequate data toquantitatively describe a hydrologic process accurately. Ra-pid parameterization of hydrologic models can be derivedusing remote sensing (RS) and geographic information sys-tems (GIS) as remotely sensed data provides valuable andup-to-date spatial information on natural resources andphysical terrain parameters. Numerous studies describedthe use of RS and GIS in hydrologic modeling (Hession andShanholtz, 1988; Tim et al., 1992; Maidment, 1993a; Srini-vasan and Engel, 1994; Bhaskar et al., 1992; Sekhar andRao, 2002; Chowdary et al., 2004; Pandey et al., 2005,2007). In all these studies, the potential benefits of RSand GIS in hydrologic and water quality modeling have beenclearly demonstrated.

The upper Damodar Valley is facing serious problems ofland degradation due to soil erosion and about 66% of thisregion is affected by different types of erosion and 35% ofthe agricultural land is under moderate to severe sheet ero-sion (Misra, 1999). Hence, in order to preserve natural re-sources and the useful life of the reservoirs, there is aneed to identify the critical areas in this region that contrib-ute higher runoff and sediment. The WEPP (Water ErosionPrediction Project) (Laflen et al., 1991) model was usedfor prediction of runoff and sediment yield for the casestudy area. The WEPP model is capable of (a) identifyingzones of sediment deposition and detachment within per-manent channels or ephemeral gullies, (b) accounting theeffects of backwater on sediment detachment, transportand deposition within channels, and (c) representing spatialand temporal variability in erosion and deposition processesas a result of agricultural management practices (Ascough IIet al. 1995a). It is intended for use on small agriculturalwatersheds (less than 260 ha) in which the sediment yieldat the outlet is significantly influenced by hillslope andchannel processes (Ascough II et al. 1995b). Model applica-

tion is constrained by the following limitations: (1) no par-tial area response; (2) no headcutting; (3) no banksloughing; and (4) no perennial streams (Ascough II et al.1995c). The WEPP model has been widely applied to predictrunoff and sediment yield at field and watershed scales(Chaves and Nearing, 1991; Tiscareno-Lopez et al., 1994;Risse et al., 1994; Favis-Mortlock et al., 1996; Zhanget al., 1996; Ghidey and Alberts, 1996; Baffaut et al.,1997; Flanagan and Nearing, 2000 and Renschler and Har-bor, 2002). Contrary to earlier studies, Gronsten and Lun-dekvam (2006) reported that the yearly and daily surfacerunoff and soil loss simulated by the WEPP Hillslope modelv. 2002.7 from two different soil erosion plot sites in south-eastern Norway did not yield satisfactory results as equa-tions used for predicting erodibility parameters in theWEPP model are not applicable to Norwegian soil types.Nearing et al. (1990) performed a sensitivity analysis ofthe WEPP hillslope model and identified precipitation, rillerodibility, rill residue cover, and rill hydraulic friction fac-tors as dominant factors while saturated hydraulic conduc-tivity and interrill erodibility were found to be moderatelysensitive parameters.

Tiwari et al. (2000) evaluated the prediction of soil lossfrom natural runoff plots at 20 different locations in theUnited States using the WEPP model and compared the re-sults with measured data and with the predictions madeby USLE and RUSLE. They concluded that the model perfor-mance is close to the traditional empirical methods withoutcalibration of any parameter. Laflen et al. (2004) reportedthat WEPP performs very well as compared to USLE andRUSLE based models in different conditions. Bhuyanaet al. (2002) compared the soil loss predictions using WEPP,EPIC and ANSWERS model and concluded that all three mod-els performed reasonably well and the predicted soil losseswere within the range of measured values. For managinglarge quantities of data for WEPP applications at the wa-tershed scale, integration of WEPP with GIS is desirable be-cause it can facilitate and possibly improve the applicationof the model. An initial application of the WEPP model witha raster-based GIS was conducted by Savabi et al. (1995) in asmall watershed in Indiana. Cochrane and Flanagan (1999)developed an interface between WEPP (Watershed version),and Arc View GIS for small basins (0.59–29 ha), comparingthe results obtained from the manual application of WEPPwith those obtained using the interface, and studying theeffect of DEM resolution on the results from the GIS WEPPinterface. There was no significant difference between themanual and the automated applications, and results ob-tained from different classes of resolution were also not sta-tistically different. Further development in techniques toautomate the application of the WEPP model has resultedin GeoWEPP (Renschler, 2003).

The WEPP model was extensively used worldwide by sev-eral researchers viz., Spain (Soto and Diaz-Fierros, 1998),UK (Brazier et al., 2000), Australia (Yu et al., 2000; Yuand Rosewell, 2001), Norway (Gronsten and Lundekvam(2006)) and Brazil (Bacchi et al., 2003). Though, severalstudies were carried out using the WEPP model, furtherrefinement and additional testing of the model is still re-quired for wide range of conditions and agricultural water-sheds. From the literature, it is evident that very limitedinformation on application of WEPP model using RS and

Runoff and sediment yield modeling from a small agricultural watershed in India using the WEPP model 307

GIS is available for Indian watersheds. In India, topographi-cal conditions, soil conditions, rainfall pattern and cultiva-tion practices are different from those in other parts ofthe world. Therefore, it is necessary to evaluate the physi-cal based models such as WEPP for an Indian watershed.Hence, the present study was carried out with the specificobjective of calibration and evaluation of the WEPP modelfor estimation of runoff and sediment yield under Indianconditions by selecting a case study area located in theupper Damodar Valley, Hazaribagh, Jharkhand State, India.

Study area description

The Karso watershed is a part of the Damodar Barakarcatchment and is situated between 85�23 0 and 85�28 0E lon-gitude and 24�12 0 and 24�18 0N latitude (Fig. 1). DamodarValley Corporation (DVC) is the first River Valley Project lo-cated in the eastern region of India to tackle soil and waterconservation problems in an integrated manner. The topog-raphy of the study area is hilly and undulating with an eleva-tion ranging from 390 to 650 m above MSL and extends overa total area of 2793 ha. The general slope of the watershedarea varies from 0% to 8% and maximum slope of some hillyparts of the watershed is up to 22%. The watershed receivesan average annual rainfall of 1300 mm and more than 80% ofthe rainfall occurs during the monsoon season (June–Sep-tember). The minimum and maximum monthly temperaturevaries from 3 �C to 42 �C. The daily mean relative humidityvaries from a minimum of 40% in April to a maximum of 85%in the month of July. The overall climate of the area can beclassified as sub-humid tropical. Two crop seasons viz. Kha-rif (Monsoon season) extending from June to September andRabi (Non-monsoon season) extending from October to Jan-uary are mainly followed. Major crops grown in the area arepaddy, maize, sorghum, soybean and peanut in Kharif sea-son and mustard, potato and onion in Rabi season. The an-nual crop rotations are mainly of paddy–mustard-fallowand corn–potato–onion.

Description of Water Erosion PredictionProject (WEPP) watershed model

The WEPP watershed model is a process-based, continuoussimulation erosion prediction model built as an extension ofthe WEPP hillslope model (Flanagan and Nearing, 1995) thatcan be used to estimate watershed runoff and sediment yield(Ascough et al., 1997). The WEPP model is based on the fun-damentals of infiltration theory, hydrology, soil physics, plantscience, hydraulics and erosion mechanics (Nearing et al.,1989). It consists of nine components: climate generation,winter process, irrigation, hydrology, soils, plant growth, res-idue decomposition, hydraulics of overland flow, erosion anddeposition. The surface hydrology component of WEPP com-putes the surface runoff and peak discharge using the kine-matic wave equation. The WEPP erosion model computessoil loss along a slope and sediment yield at the end of a hills-lope. Interrill and rill erosion processes are considered, and ituses a steady-state sediment continuity equation as a basisfor the erosion computations. The detailed mathematicalrepresentations of the channel hydrological processes arepresented in technical manual of WEPP model (Flanagan

and Nearing, 1995). The application of WEPP to a watershedrequires that hillslopes be delineated and channels identified(Baffaut et al., 1997). Each hillslope (represented as a rectan-gle inWEPP) consists of a representative length (L), width (W)and slope profile. Hillslopes drain into the top, left side, orright side of a channel, eventually leading to the watershedoutlet. In theWEPPmodel the smallest possiblewatershed in-cludes one hillslope and one channel. Runoff, detachmentand deposition are first calculated on each hillslope withthe hillslope component of WEPP for the entire simulationperiod. Then the model combines simulation results fromeach hillslope and performs runoff and sediment routingthrough the channels and impoundments. It is intended foruse on small agricultural watersheds in which the sedimentyield at the outlet is significantly influenced by hillslope andchannel processes. An advantage of WEPP over other existingmodels such as thepopularUniversal Soil Loss Equation (USLE)(Wischmeier and Smith, 1978) is that the soil loss and deposi-tion of sediment is estimated spatially along a profile. In otherwords, soil loss and deposition on a complete continuous hills-lope profile can be calculated, which is important in wa-tershed modeling because it enables enhanced prediction ofsediment yield to channels and to the watershed outlet.

Methodology

Meteorological data

Historical daily rainfall data for nine years (1992–2000) werecollected fromthe raingauge station located in thewatershedandanalyzed todetermine the various statistical parameters.Data on intensity and duration of rainfall for the years 1994,1998 and 1999 could not be collected, as the recording typerain gauge was inoperative. Other meteorological data suchas maximum and minimum air temperature and relativehumidity were collected from a meteorological observatoryat Hazaribagh, located 30 km away from the outlet. Due tonon-availability of location specific data, some of the weath-er data such as solar radiation and wind velocity were col-lected from the meteorological observatory, Ranchi located90 km away from the study area as the climate and the topo-graphical conditions of these meteorological observatoriesare similar to that of the study area.

Hydrological data

Soil Conservation Department of Damodar Valley Corpora-tion, Hazaribagh and Indo-German Bilateral Project (IGBP)on ‘Watershed Management’, New Delhi, India monitorhydrological data in some of the watersheds of DamodarValley, including the Karso watershed. Watershed daily sur-face runoff and daily sediment were collected for the mon-soon season of the years 1992–2000. A set of instrumentsconsisting of continuous recording rain gauge, water levelstage recorder and silt sampler (bottle type) were used torecord rainfall, stream flow (seasonal) and sediment flowdata, respectively. The sediment yield data were measuredby manual sampling using point integrating Punjab bottleand USDH-48 bottle type samplers. The sediment concentra-tion was obtained by filtration and evaporation (oven dry-ing) methods.

Figure 1 Location map of the study area.


Satellite data

Cloud free digital data of IRS-1C LISS-III (Linear Imaging SelfScanner) path-105 and row-55 of imagery of 23.5 m spatialresolution pertaining to 28th October, 1996 in four spectralbands (band 1: 0.45–0.52 lm; band 2: 0.52–0.59 lm; band3: 0.62–0.68 lm and band 4: 0.77–0.86 lm) was used.

Generation of input parameters for WEPP model

Generation of land use/cover from remote sensing dataThe IRS-1C LISS-III satellite image of the year 1996 was clas-sified using supervised classification (after several groundtruth verifications) with maximum likelihood classificationalgorithm in ERDAS IMAGINE software with an overall


accuracy of 88.71% and kappa coefficient value of 0.867(Pandey et al., 2007). The major land use classes of thestudy area are paddy, upland crop (Soybean), forest andwasteland/fallow land (Pandey et al., 2007).

Spatial database generation under GIS environmentDrainage network and elevation contours at 10 m intervalwere digitized using Survey of India toposheets at1:50,000 scales using ARC INFO GIS software. Subse-quently, these digitized contours were used to generatea Digital Elevation Model (DEM) with a grid cell resolutionof 200 m (Kienzle, 1996; Renschler, 1997). Further, slope,aspect and slope shape factor grids were also generatedfor the study area using the DEM. The study watershedwas then delineated into seven sub-watersheds using theDEM and drainage network. The soil map of the study areaat a scale 1:10560 of the watershed was collected from theSoil Conservation Department of DVC, Hazaribagh (India)and digitized using ARC-INFO GIS. Sub-watershed wise in-put soil parameters used in the WEPP model are presentedin Table 1.

Hydrological data processing

Daily rainfall charts (15 min) for the years 1992, 1993, 1995,1996, 1997 and 2000 were collected for the study watershedand analyzed for rainfall amount, intensity and duration togenerate the break point data sets for each interval in theday. Minimum and maximum temperatures and wind veloc-ity and direction at 8 and 18 h of the day, daily values ofradiation and dew point temperatures were used for thepreparation of the WEPP input climate file. The observeddischarge (m3 s�1) at the outlet of the watershed was con-verted to daily runoff depth (mm) using the drainage areaof the watershed. The daily observed sediment yield wasconverted from grams per liter (gm l�1) to tons per hectare(t ha�1) using the watershed area and runoff. The averagedensity of runoff water was found to be 1.5 g cm�3. Theseobserved runoff and sediment yields were used to comparethe simulated values of runoff and sediment yield usingWEPP model.

Model input parameters

In the WEPP model hillslopes are defined as a set of gridcells in the DEM that drain to the left, right, or top of eachindividual channel. If the channel is a secondary channel,meaning that the junction of two other channels creates

Table 1 Sub-watershed wise input soil parameters used in WEPP

Soil parameters Parameters for different sub-w

1 2 3

Albedo 0.49 0.46 0.Initial saturation level (%) 70 70 70Sand (%) 18.70 21.80 18Clay (%) 11.40 8.90 11Organic (%) 0.517 0.638 0.CEC (meq/100 g) 10 10 10

it, then there will be one hillslope to the left and anotherone to the right of the channel, but no hillslope drainingto the top of the channel. In the present study, methodologyfor delineation of hillslopes was adopted from Amore et al.(2004). The hillslope method consists of identifying thechannel network, defining hillslopes draining into eachchannel segment, and creating a representative slope pro-file for each hillslope.

The channel network was digitized from the Survey ofIndia topographic map at a scale of 1:50000. The channelnetwork was further updated using IRS 1C LISS III data hav-ing 23 m resolution. Initially the DEM with a cell resolutionof 200 m was constructed using the topogrid function ofGIS by inputting different thematic layers such as contourcoverage, drainage network and spot heights. The resul-tant DEM was used to generate a flow direction grid whichidentifies the flow direction out of each grid cell. Subse-quently, hillslopes inside the basins were delineated basedon the flow direction grid which was used to detect wa-tershed divides within the basin. However very small hills-lope basins with irregular shapes were dissolved intoneighboring hillslope basins. Sub-division of hillslopes werecarried out by overlaying different thematic layers such asslope coverage, soil coverage and land use coverage, sothat each hillslope is characterized by topography, soil,and land use. Parameters of the watershed such as over-land and channel slope, channel length and hillslope lengthwere extracted from different thematic layers (i.e. con-tour, slope and drainage map). The number of channelsidentified for each sub-watershed is presented in Table 2and Fig. 2. Manning’s roughness coefficient values werealso selected based on the condition of the channels ofsub-watersheds from the literature (Table 2). Other WEPPchannel input parameters such as actual width, shape anddepth were collected either through communication withthe field workers during field visits or by actual measure-ment in the study area, as these cannot be extracted orderived from the DEM.

The climate (.cli) file for the WEPP model was generatedusing BPCDG (Breakpoint Climate Data Generator) that gen-erates breakpoint climate data using observed standardweather data sets. The Hillslope slope file (.slp) was builtwith the interface slope file builder. Sub-watershed wisesoil characteristics, such as percent of sand, silt, clay, or-ganic matter, rock fragment fraction, and cation exchangecapacity were obtained from the measured data in the studyarea. The soil file (.sol) was built through soil file builderwithin the WEPP interface.

model

atersheds

4 5 6 7

49 0.38 0.50 0.29 0.4570 70 70 70

.60 20.60 24.40 7.50 11.30

.40 4.80 6.30 20.80 15.60517 1.140 0.450 1.790 0.741

5 5 15 10

Table 2 Number of hillslopes and channels in the study sub-watersheds

Sub-watershed number 1 2 3 4 5 6 7

Number of hillslopes 28 69 31 41 79 70 39Number of channels 17 39 19 24 46 41 25Manning’s roughness coefficient 0.04 0.03 0.027 0.04 0.03 0.03 0.03

Figure 2 Delineated hillslopes and channels in sub-watersheds of the study area.

y = 0.27x + 2.98R2 = 0.41

0

20

40

60

80

0 20 40 60 80 100Observed Runoff (mm)

Sim

ulat

ed R

unof

f (m

m)

1:1 Line Regression Line100

Figure 3 Pre calibrated observed and simulated runoff forthe year 1996.


Model evaluation

The model evaluation procedure included calibration, sensi-tivity analysis and validation. In general, the estimation ofparameters of the WEPP model is often a stumbling blockto new users of the model who are faced with the task ofpreparing an input file for the first time. To overcome thisproblem, more critical parameters can be identified by sen-sitivity analysis and can be fine tuned or calibrated to im-prove the agreement between the simulated and observeddata. Sensitivity analysis provides a method for examiningthe response of a model in a way that eliminates the influ-ence of error related to natural variation of the model inputparameters (McCuen and Snyder, 1983). Thus, sensitivityanalysis of the WEPP model is necessary for assessing therationality of the model, to provide insight into the overallphysical system, which the simulation model represents,and to help identify research needs (Nearing et al., 1990).Initially, the WEPP model was evaluated with the model in-put values calculated by the equations presented in thetechnical manual of the WEPP model (Flanagan and Nearing,1995). Model input values were estimated at the sub-wa-tershed level and averages of these values were used inthe model simulation. Pre calibrated model results for run-off and sediment yield are presented in Figs. 3 and 3a. Thelow values of coefficient of determination (0.41 and 0.23,respectively for the runoff and sediment yield) and Nash–

Sutcliffe model efficiency (0.19 and 0.22, respectively forthe runoff and sediment yield) indicates the pre calibratedmodel performance is poor. Thus, in the present study inputparameter values required by the model were obtained fromdirect field and laboratory measurements, literature (Misra,1999), remote sensing, GIS or through the calibration pro-cess of the models, where selected model parameters wereadjusted within an expected range and the discrepanciesbetween the measured and model predictions could be min-imized (Donigian and Rao, 1990). Subsequently, a base datafile was established to perform the sensitivity analysis, andbase output variables were determined. Each input variablewas varied within the prescribed range keeping the others

y = 0.24x + 0.07R2 = 0.23

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2

Observed Sediment (t ha-1)

Sim

ulat

ed S

edim

et (

t ha

-1)

1:1 Line Regression Line

Figure 3a Pre calibrated observed and simulated sedimentyield for the year 1996.


constant. The output values were then analyzed to deter-mine their variation with respect to the base values. As eachvariable was varied about the base value, the mean outputwas compared with the mean of the base value prediction asa measure of the sensitivity. The input parameters thatshowed negligible variation were not calibrated and weretaken as model default values. Thus, the calibration processfocused mainly on input parameters such as interrill erod-ibility and effective hydraulic conductivity that control run-off and sediment production. The model was calibrated forthe study watershed conditions using 1996 monsoon seasondata. The 1996 dataset such as rainfall, temperature, rela-tive humidity, wind velocity and land use/cover were cho-sen for calibration of the model because the weatherconditions in 1996 were less extreme than in other years.After each parameter adjustment and simulation run, thesimulated and observed hydrographs were visually com-pared to examine the improvement in the match. Afterproper calibration, the model was evaluated for the estima-tion of daily runoff and sediment yield using the monsoonseason data of the years 1992, 1993, 1995, 1997 and 2000.Due to non-availability of data, the model could not be val-idated for the years 1994, 1998 and 1999.

Criteria for model evaluation

Hydrological models are used most frequently to simulate orpredict flow either on a continuous basis or for a particularevent. The graphical representation of the result could eas-ily be interpreted if the calibration is done for only one wa-tershed at one stream gauging location. In the presentstudy, the model performance was evaluated on the basisof test criteria recommended by the ASCE Task Committee(1993) and graphical performances criteria suggested byHaan et al. (1982). The deviation of runoff values, Dv givenby the following equation is another criterion for goodness-of-fit.

Dv ð%Þ ¼V � V 0

V� 100 ð1Þ

where V is the measured daily runoff volume; V 0 is the mod-el computed daily runoff. The smaller the value of Dv, thebetter the model results. Dv would equal to zero for a per-fect model. The use of Dv provides an immediate comple-ment to a visual inspection of the continuous hydrograph.The prediction performance of the model was decidedbased on the criterion suggested by Bingner et al. (1989).

Another goodness-of-fit criterion recommended by ASCETask Committee (1993) is Nash–Sutcliffe coefficient orcoefficient of simulation efficiency (ENS) (Nash and Sutcliffe,1970) given by:

ENS ¼ 1�Pn

i¼1ðQi � Q 0iÞ2

Pni¼1ðQi � QÞ2

ð2Þ

where Qi is the measured daily discharge, Q 0i is the com-puted discharge and Q is the average measured discharge.The ENS values can vary from 0 to 1, with 1 indicating a per-fect fit. However, a shortcoming of the Nash Sutcliffe statis-tics is that it does not perform well in the period of lowflow. If the daily measured flow approaches the average va-lue, the denominator of the equation becomes to zero andENS approaches minus infinity with only minor model misspredictions. This statistics works well when the coefficientof variation for the observed data set is large.

Results and discussion

Calibration and sensitivity analysis

The calibrated parameters and input variables used for sen-sitivity analysis of the WEPP model are presented in Table 3.A major limitation in the present study is that the WEPPmodel was calibrated with the measured data at the wa-tershed level due to non-availability of measured runoffand sediment data at the sub-watershed level. The resultsof the sensitivity analysis for the runoff and sediment yieldare presented in Table 3. The results of sensitivity analysisreveal that the runoff is sensitive to effective hydraulic con-ductivity value alone whereas, sediment yield is sensitive tothe interrill erodibility, effective hydraulic conductivity, rillerodibility and critical hydraulic shear stress as could beseen by either increasing or decreasing these parametersby 10%, 20%, 25% and 50%, respectively. The lower sensitive-ness of interrill erodibility, rill erodibility and criticalhydraulic shear stress to runoff production may be attrib-uted to the fact that these parameters are more dominantonly in erosion process. Thus, runoff in WEPP is more sensi-tive to changes in the physical environment than the erosionparameters, but in both cases management and soil attri-butes are integrated in the estimates. However, other fac-tors such as topography, soils and land cover may alsoinfluence the results to some extent. Thus, from the sensi-tivity analysis it was possible to determine the importantvariables that needed to be precisely estimated to makeaccurate prediction of watershed yields. Similar resultswere reported by Nearing et al. (1990), Baffaut et al.(1997) and Bhuyana et al. (2002). Therefore, more precisecalibration was carried out for the parameters such aseffective hydraulic conductivity for runoff and both effec-tive hydraulic conductivity and interill erodibility for sedi-ment yield. In the present study, although rill/interrillplot studies are not available for verification purposes, thecalibrated values are consistent with values reported in lit-erature (Bhuyana et al., 2002). Although the present studyapplies WEPP to a large watershed, these applications arejustifiable if the physical modeling processes are appropri-ately represented. For example, Amore et al. (2004) usedWEPP to model three large Sicilian basins of size 115, 185

Table 3 Sensitivity analysis of calibrated parameters of WEPP model for whole watershed

Soil parameters Value Percent deviation of sediment yield from measured

+10 +20 +25 +50 �10 �20 �25 �50Interrill erodibility (kg s m�4) 8.17 · 105 1.24 5.8 11.55 17.90 �2.77 �2.90 �3.05 �3.15Effective hydraulic conductivity (mm h�1) 4.52 �1.25 �6.15 �6.26 �8.30 1.14 6.43 8.40 11.15Rill erodibility (s m�1) 0.009 1.8 1.85 2.0 3.17 �0.48 �1.59 �1.67 �3.62Critical hydraulic shear stress (Pa) 1.28 0.26 5.06 6.11 7.66 �0.49 �0.76 �2.81 �7.15+: percent increase in the parameter, �: percent decrease in the parameter.

y = 0.79x + 1.71R2 = 0.95

0

20

40

60

80

100

0 20 40 60 80 100

Observed Runoff (mm)

Sim

ulat

ed R

unof

f (m

m)


Figure 4a Calibrated observed and simulated runoff for theyear 1996.

0.0

0.3

0.6

0.9

1.2

1.5

25-J

un

2-Ju

l

9-Ju

l

16-J

ul

23-J

ul

30-J

ul

6-A

ug

13-A

ug

20-A

ug

27-A

ug

3-S

ep

10-S

ep

17-S

ep

24-S

ep

Time (days)

Sed

imen

t (t h

a-1)

0

50

100

150

200

250

300

Rai

nfal

l (m

m)

Rainfall Observed Sediment Simulated Sediment

Figure 5 Calibrated observed and simulated sediment yieldfor the year 1996.

y = 1.09x - 0.001.2

ha-1

)



and 570 sq. km, respectively. Amore et al., 2004) reportedthat the WEPP is not sensitive to the size of the hillslopearea within the considered range of values. Further, theyalso suggested that a finer sub-division may better approxi-mate the experimental condition (plot or field area) butmay not necessarily be needed for a better estimate ofthe erosion model.

The observed and simulated daily runoff and sedimentvalues of Karso watershed for the calibration period(June–September, 1996) were compared graphically asshown in Figs. 4, 4a, 5 and 5a. The observed and simulateddaily runoff and sediment yield for the calibration periodalong with 1:1 line are shown in Figs. 4a and 5a, respec-tively. It is observed from Fig. 4a that the simulated runoffvalues are distributed uniformly about the 1:1 line for lowervalues of observed runoff. For higher values of observedrunoff, the simulated values are slightly below the 1:1 line,indicating that the model under-predicts the high values ofrunoff, whereas for well distributed daily rainfall events itpredicts the runoff quite satisfactorily. Similarly, daily pre-dicted sediment yield values were plotted against the mea-sured values and their distribution about the 1:1 line isshown in Fig. 5a. The simulated sediment yields are foundto be distributed uniformly along the 1:1 line for both thelower and higher values of the observed sediment yields.Higher values of the coefficient of determination indicatea close relationship between the measured and simulatedrunoff and sediment yield (Table 4). It is seen from Table4 that during calibration the overall deviations betweenthe observed and simulated values for runoff and sedimentyield are 6.45% and 8.60%, respectively. There is slight un-der-prediction in the peak runoff which results in slightlylower standard deviation and mean for the simulated run-off. The maximum sediment yield simulated by the model

0

20

40

60

80

100

23-J

un

30-J

un

7-Ju

l

14-J

ul

21-J

ul

28-J

ul

4-A

ug

11-A

ug

18-A

ug

25-A

ug

1-Se

p

8-Se

p

15-S

ep

22-S

ep

Time (days)

Run

off

(mm

)

0

50

100

150

200

250

300

Rai

nfal

l (m

m)

Rainfall Observed Runoff Simulated Runoff

Figure 4 Calibrated observed and simulated runoff for theyear 1996.

R2 = 0.90

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1 1.2


Sim

ulat

ed S

edim

ent

(t

Figure 5a Calibrated observed and simulated sediment yieldfor the year 1996.

is found to be slightly lower than the observed maximumsediment yield at the beginning of the season (July, 13)(Fig. 5). However, student’s t-test shows that (t-calcu-

0

10

20

30

40

50

25-J

un

2-Ju

l

9-Ju

l

16-J

ul

23-J

ul

30-J

ul

6-A

ug

13-A

ug

20-A

ug

27-A

ug

3-Se

p

10-S

ep

Time (days)

Run

off

(mm

)

0

50

100

150

200

250

Rai

nfal

l (m

m)


Figure 6 Observed and simulated runoff for the year 1992.

y = 0.76x + 0.97R2 = 0.89

0

10

20

30

40

50

0 10 20 30 40 50

Observed Runoff (mm)Si

mul

ated

Run

off

(mm

)


Figure 6a Comparison between the observed and simulatedrunoff for the year 1992.


lated < t-critical) the means of observed and simulated run-off and sediment yield are not significantly different at the95% confidence level. High values of Nash–Sutcliffe modelefficiency also indicate that the model can be well adoptedfor simulation of runoff and sediment. Thus, the resultsindicate that the overall prediction of daily surface runoffand sediment by the WEPP model during the calibration per-iod was satisfactory and therefore, accepted for furtheranalysis.

Assessment of runoff and sediment yield from thestudy area

The calibrated WEPP model was then used to estimate dailyrunoff and sediment yield from the selected study area forthe years 1992, 1993, 1995, 1997 and 2000 using the dataof monsoon season (June to September). The measured dai-ly runoff values and sediment yield values for all the yearswere compared with the simulated values to evaluate themodel validation performance. The daily values of the ob-served runoff for the study years were compared graphicallywith the simulated runoff and are shown in Figs.6,6a,7,7a,8,8a,9,9a,10,10a. The peak values of the simu-lated runoff match consistently well with the peak valuesof measured runoff throughout the season in all the years.However, the model slightly under-predicts a few peak val-ues of runoff and over-predicts the small values of runoff.The scattergrams for comparison of the simulated runoffwith the measured runoff for all the years along with 1:1line are presented in Figs. 6a, 7a, 8a, 9a and 10a. It is seenfrom the figures that the points are almost evenly distrib-uted about the 1:1 line, except for few events located be-low the line. Similarly, sediment yield from the study areawas estimated using the data of monsoon season for theyears 1992, 1993, 1995, 1997 and 2000. The observed andsimulated daily sediment yields of Karso watershed for allthe years were compared graphically as shown in Figs.11,11a,12,12a,13,13a,14,14a,15,15a. The scattergram ofthe observed and simulated daily sediment yields for thestudy period along with the 1:1 line is shown in Figs. 9a,11a, 12a, 13a and 14a. The points obtained by plotting the

Table 4 Statistical analysis of observed and simulated daily runo

Statistical parameters Runoff (mm)

Observed

Mean 11.53Standard deviation 12.67Maximum 57.29Total 242.03Events 21t-calculated at 95% level 0.96t-critical(two tailed) 2.08Deviation (%) 6.45R2 0.95ENS 0.92

t-cal is the Student’s t-calculated for equal means at 95% confidenceR2 is the coefficient of determination.ENSis the Nash–Sutcliffe simulation efficiency.

simulated and observed values of sediment yield are evenlydistributed about the 1:1 line indicating a close agreementbetween them. Regression analysis was also performed be-tween the observed and simulated yield values.

The descriptive statistics for both the measured and sim-ulated daily runoff and sediment for all the years are shownin Tables 5 and 6, respectively. From Tables 5 and 6, it maybe observed that the values of deviation varied from 19.3%to 15.15% indicating almost a close agreement between theobserved and simulated daily runoff and sediment. The un-der-prediction or over-prediction limits for the model simu-lation are within ±20% from the measured values and are

ff and sediment yield, 1996

Sediment yield, (t ha�1)

Simulated Observed Simulated

10.78 0.12 0.1310.24 0.19 0.2146.35 0.84 1.01

226.43 2.48 2.6921 21 21

�0.662.08�8.600.900.85

level.

0

10

20

30

40

5027

-Jun

4-Ju

l

11-J

ul

18-J

ul

25-J

ul

1-A

ug

8-A

ug

15-A

ug

22-A

ug

29-A

ug

5-Se

p

12-S

ep

19-S

ep

26-S

ep

3-O

ct

10-O

ct

Time (days)

Run

off

(mm

)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 0.90x - 0.27R2 = 0.89

0

10

20

30

40

50


Sim

ulat

ed R

unof

f (m

m)


Figure 7a Comparison between the observed and simulatedrunoff for the 1993.

0

10

20

30

40

50

23-J

un

30-J

un

7-Ju

l

14-J

ul

21-J

ul

28-J

ul

4-A

ug

11-A

ug

18-A

ug

25-A

ug

1-Se

p

8-Se

p

15-S

ep

22-S

ep

Time (days)

Run

off

(mm

)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 0.85x + 0.18R2 = 0.91

0

10

20

30

40

50


Sim

ulat

ed R

unof

f (m

m)



0

10

20

30

40

50

26-J

un

3-Ju

l

10-J

ul

17-J

ul

24-J

ul

31-J

ul

7-A

ug

14-A

ug

21-A

ug

28-A

ug

4-Se

p

Time (days)

Run

off

(mm

)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 0.89x - 0.10R2 = 0.90

0

10

20

30

40

50

0 10 20 30 40 50

Observed Runoff (mm)Si

mul

ated

Run

off

(mm

)



0

10

20

30

40

50

15-J

un

22-J

un

29-J

un

6-Ju

l

13-J

ul

20-J

ul

27-J

ul

3-A

ug

10-A

ug

17-A

ug

24-A

ug

31-A

ug

7-S

ep

14-S

ep

21-S

ep

Time (days)

Run

off

(mm

)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 0.86x + 0.64R2 = 0.86

0

10

20

30

40

50

0 10 20 30 40 50

Observed Runoff (mm)

Sim

ulat

ed R

unof

f (m

m)




considered as the acceptable levels of accuracy for the sim-ulations as reported by Bingner et al. (1989). Student’s t-test was also performed to test the similarity between the

means of the observed and simulated runoff and sediment.Student’s t-test shows that (t-calculated < t-critical) themeans of observed and simulated runoff and sediment are

0.0

0.2

0.4

0.6

0.8

1.026

-Jun

3-Ju

l

10-J

ul

17-J

ul

24-J

ul

31-J

ul

7-A

ug

14-A

ug

21-A

ug

28-A

ug

Time (days)

Sedi

men

t (t

ha-1

)

0

50

100

150

200

250

Rai

nfal

l (m

m)


Figure 11 Observed and simulated sediment yield for theyear 1992.

y = 0.8088x + 0.0616R2 = 0.81

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1


Sim

ulat

ed S

edim

ent

(t h

a-1)


Figure 11a Comparison between the observed and simulatedsediment yield for the year 1992.

0.0

0.2

0.4

0.6

0.8

1.0

27-J

un

4-Ju

l

11-J

ul

18-J

ul

25-J

ul

1-A

ug

8-A

ug

15-A

ug

22-A

ug

29-A

ug

5-S

ep

12-S

ep

19-S

ep

26-S

ep

3-O

ct

10-O

ct

Time (days)

Sed

imen

t (t h

a-1)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 1.05x + 0.02R2 = 0.95

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0


Sim

ulat

ed S

edim

ent

(t h

a-1)



0.0

0.2

0.4

0.6

0.8

1.0

27-J

un

4-Ju

l

11-J

ul

18-J

ul

25-J

ul

1-A

ug

8-A

ug

15-A

ug

22-A

ug

29-A

ug

5-S

ep

12-S

ep

19-S

ep

26-S

ep

Time (days)

Sed

imen

t (t h

a-1)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 1.09x - 0.002R2 = 0.89

0.0

0.2

0.4

0.6

0.8

1.0

0.00 0.20 0.40 0.60 0.80 1.00


Sim

ulat

ed S

edim

ent

(t h

a-1)



0.0

0.2

0.4

0.6

0.8

1.0

26-J

un

3-Ju

l

10-J

ul

17-J

ul

24-J

ul

31-J

ul

7-A

ug

14-A

ug

21-A

ug

28-A

ug

4-S

ep

11-S

ep

Time (days)

Sed

imen

t (t h

a-1)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 1.01x + 0.004R2 = 0.85

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1


Sim

ulat

ed S

edim

ent

(t h

a-1)


Figure 14a Comparison between he observed and simulatedsediment yield for the year 1997.


0.0

0.2

0.4

0.6

0.8

1.015

-Jun

22-J

un

29-J

un

6-Ju

l

13-J

ul

20-J

ul

27-J

ul

3-A

ug

10-A

ug

17-A

ug

24-A

ug

31-A

ug

7-S

ep

14-S

ep

21-S

ep

Time (days)

Sed

imen

t (t h

a-1)

0

50

100

150

200

250

Rai

nfal

l (m

m)



y = 1.09x + 0.01R2 = 0.95

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0


Sim

ula

ted

Sed

imen

t (t

ha-1

)



Table

5Statistica

lan

alysisofobservedan

dsimulateddaily

runoff

Statistical

parameters

Runoff

(mm)

Year19

92Year19

93Year

1995

Year

1997

Year

2000

Observed

Simulated

Observed

Simulated

Observed

Simulated

Observed

Simulated

Observed

Simulated

Mean

8.76

7.67

5.46

4.64

6.50

5.69

7.28

6.39

10.41

9.58

Standard

deviation

7.89

6.41

6.85

6.54

6.18

5.48

6.41

6.04

7.16

6.62

Max

imum

26.14

23.59

27.57

24.96

25.06

24.24

27.21

28.54

23.63

23.63

Total

148.94

130.34

103.83

88.10

123.52

108.05

145.57

127.87

156.10

143.77

Count

17

1719

1919

1920

2015

15t-ca

lculated

at95

%leve

l1.58

1.55

1.90

1.93

1.21

t-critical

(twotailed)

2.11

2.09

2.09

2.08

2.13

Deviation(%)

12.49

15.15

12.53

12.16

7.90

R2

0.89

0.89

0.91

0.90

0.86

ENS

0.85

0.95

0.89

0.88

0.85


not significantly different at 95% confidence level for all theyears. From Tables 5 and 6 it may be seen that the highcoefficient of determination indicates a positive relation-ship between the measured and simulated runoff and sedi-ment yield in all the years. Further, reasonably highvalues of Nash–Sutcliffe model efficiency shows satisfac-tory performance of the model. The peak values of the pre-dicted sediment yield match consistently well with that ofthe measured values for the entire season. However, themodel predicted values are both higher and lower thanthe observed values at different times during the validationperiod. For intermediate rainfall events, the model gener-ally over-predicts the sediment yield whereas, for otherevents it generally predicts the sediment yield quite well.The over-prediction of sediment yield for some of the rain-fall events might be due to the existing conventional conser-vation practices not precisely accounted in the model. Itmay be observed from the figure that most of the predictedvalues are either equal to or slightly greater than the ob-served values. Total predicted sediment yield (2.48 tha�1) of all the events is slightly more than the total ob-served sediment yield (2.10 t ha�1). At low sediment yield,some of the predicted values are higher, except for the low-est values of sediment yield at the beginning of the season.For high rainfall events, model predicted values are veryclose to the observed sediment yield with marginal overprediction in a few cases. Relatively simplistic sedimentrouting equations in the model and assumption of statichillslope and channel dimensions during the simulationmay also be the contributing factors for the deviations.The result is in agreement with the ideas presented by Near-

Table

6Statisticalan

alysisofobservedan

dsimulateddaily

sedim

entyield

Statisticalparameters

Sedim

entyield

(tha�

1)

Year199

2Year

1993

Year

1995

Year

1997

Year

2000

Observed

Simulated

Observed

Simulated

Observed

Simulated

Observed

Simulated

Observed

Simulated

Mean

0.16

0.19

0.14

0.17

0.09

0.10

0.11

0.12

0.13

0.15

Standard

deviation

0.16

0.15

0.19

0.21

0.10

0.12

0.11

0.12

0.16

0.18

Max

imum

0.56

0.52

0.69

0.74

0.35

0.43

0.44

0.43

0.59

0.65

Total

2.25

2.68

2.10

2.48

1.22

1.30

2.22

2.32

1.83

2.09

Count

1414

1515

1313

2020

1414

t-ca

lculated

at95

%leve

l�1.62

�2.09

�0.55

�0.48

�1.55

t-critical

(twotailed)

2.14

2.13

2.16

2.08

2.14

Deviation(%)

�19

.30

�18

.36

�6.59

�4.43

�13

.96

R2

0.81

0.95

0.89

0.85

0.95

ENS

0.78

0.92

0.84

0.82

0.91


ing (1998) regarding the cause of over-prediction of smallsoil losses and under-prediction of large soil losses. This isconsistent with the previous studies using USLE (Risseet al., 1993), RUSLE (Rapp, 1994) and WEPP hillslope models(Zhang et al., 1996; Nearing and Nicks, 1997; Liu et al.(1997)) that have shown over-estimation on plots with rela-tively low erosion rates and under-estimation on plots withhigh erosion rates. The phenomenon of erosion modelsover-estimating small events and under-estimating largeevents is inherent to all erosion models as reported by Near-ing (1998), Ghidey et al. (1995) and Kramer and Alberts(1995).

However, deviation (moderate as well as low) observedin the peaks may be attributed to the errors associated withmanual measurement and calculation in estimating runofffrom the measured cross sectional area and runoff velocityat the watershed outlet. Generally it is observed that duringthe initial phase of commencement of monsoon rains, theobserved runoff was less than the simulated runoff. Thismay be due to the fact that major portion of the rainfallis stored in the bunded paddy fields (Pandey et al., 2005).In general, high intensity rainfall occurs during the monthsof August–September and in most of the cases the soil is al-ready saturated, resulting in high runoff. This may be one ofthe reasons for under-prediction of runoff, as the modelcannot take into account the variation in the saturation le-vel of the soil. On the contrary, the model may over-predictthe runoff if high intensity rainfall occurs following few daysof dry spells which leads to unsaturated soil condition. Nev-ertheless, the model predicted runoff is found to be close tothe observed values for well-distributed daily rainfallevents. It implies that the model was accurately calibratedand hillslope wise parameters were evaluated correctly.Thus, the results confirm that the WEPP model could accu-rately predict the runoff and sediment from the Karso wa-tershed with marginal deviation as discussed earlier forthe study years. Though in the present study, the WEPPmodel was validated with measured values at the outlet ofwatershed, the simulated runoff and sediment yield dataat the hillslope level are expected to provide some directionabout the status of watershed. Thus, physically based modelsuch as WEPP can be used not only to identify the criticalareas but also to explore the capabilities of the model toidentify best management practices.

Conclusions

The present study was carried out to evaluate the physicallybased WEPP model for estimation of runoff and sedimentyield from Karso watershed (India). The coefficient of deter-mination (R2), Nash–Sutcliffe efficiency, and percent devi-ation were used for performance evaluation, and theseranged from 0.86 to 0.91, from 0.85 to 0.95, from 7.90 to15.15, respectively, indicating satisfactory model perfor-mance. Similarly, the high values of R2 ranging from 0.81to 0.95, Nash–Sutcliffe efficiency from 0.78 to 0.92, andpercent deviation values from 4.43 to 19.30 indicated satis-factory simulation of sediment yield. The sediment yieldcomputation was quite sensitive to interrill erodibility andeffective hydraulic conductivity parameters, for the changein each of the parameters by ±50% in a run exhibited thesediment yield to vary from 1.24% to 17.90% and from


1.25% to 11.15%, respectively. On the other hand, the runoffvaried from 1.1% to 10% with ±50% change in effectivehydraulic conductivity. In pre-calibration, the WEPP modeldid not predict erodibility parameters satisfactorily in bothsurface runoff and sediment yield simulations, largely dueto the empiricism involved in model equations. However,the present calibrated model results could be of use in ero-sion-based watershed prioritization and evaluation of crop-ping management practices in the study area.

Acknowledgements

The authors gratefully acknowledge the help provided byHead, Regional Remote Sensing Service Center, Kharagpurand other scientists of the centre in carrying out this study.The authors are also thankful to the Soil Conservation Dept.,DVC Hazaribagh, India for providing valuable data for valida-tion of the simulation results. The financial support pro-vided by the DAAD (Deutscher Akademischer AustauschDienst), to conduct a part of the research work in the Uni-versity of Hannover, Germany is gratefully acknowledged.The authors would also like to thank the anonymous refer-ees for contributing insightful remarks and useful sugges-tions, which led to a substantially improved manuscript.

References

Amore, E., Modica, C., Nearing, M.A., Santoro, V.C., 2004. Scaleeffect in USLE and WEPP application for soil erosion computationfrom three Sicilian basins. J. Hydrol. 293 (1–4), 100–114.

Arnold, J.G., Allen, P.M., Bernhardt, G., 1993. A comprehensivesurface-groundwater flow model. J. Hydrol. 142, 47–69.

ASCE Task Committee, 1993. Criteria for evaluation of watershedmodels. J. Irrig. Drain. Eng. 119 (3), 429–442.

Ascough II, J.C., Baffut, C., Nearing, M.A., Liu, B.Y., 1997. TheWEPP watershed model: I. hydrology and erosion. Trans. ASAE 40(4), 921–933.

Ascough II, J.C., Baffaut, C., Nearing, M.A., Flanagan D.C.,1995a,b,c. Watershed Model Channel Hydrology and ErosionProcesses. In: Flanagan, D.C., Nearing, M.A., (Eds.), USDA –Water Erosion Prediction Project Hillslope Profile and WatershedModel Documentation NSERL Report No. 10 July 1995, NationalSoil Erosion Research Laboratory, USDA-ARS, W. Lafayette, IN(Chapter 13).

Bacchi, O.O.S., Reichardt, K., Sparovek, G., 2003. Sediment spatialdistribution evaluated by three methods and its relation to somesoil properties. Soil Tillage Res. 69, 117–125.

Baffaut, C., Nearing, M.A., Ascough II, J.C., Liu, B.Y., 1997. TheWEPP watershed model: II. Sensitivity analysis and discretizationon small watersheds. Trans. ASAE 40 (4), 935–943.

Beasley, D.B., Huggins, L.F., Monke, E.J., 1980. ANSWERS: A modelfor watershed planning. Trans. ASAE 23 (4), 938–944.

Bhaskar, N.R., James, W.P., Devulapalli, R.S., 1992. Hydrologicparameter estimation using geographic information system. J.Water Resour. Plann. Manag. 118 (5), 492–512.

Bhuyana, S.J., Kalita, P.K., Janssenc, K.A., Barnesa, P.L., 2002. Soilloss predictions with three erosion simulation models. Environ.Model. Software 17 (2), 135–144.

Bingner, R.L., Murphee, C.E., Mutchler, C.K., 1989. Comparison ofsediment yield models on various watershed in Mississippi.Trans. ASAE 32 (2), 529–534.

Brazier, R.E., Beven, K.J., Freer, J., Rowan, J.S., 2000. Equifinalityand uncertainty in physically based soil erosion models: appli-

cation of the glue methodology to WEPP—the water erosionprediction project for sites in the UK and USA. Earth Surf.Process. Landforms 25, 825–845.

Chaves, H.M.L., Nearing, M.A., 1991. Uncertainty analysis of theWEPP soil erosion model. Trans. ASAE 34 (6), 2437–2444.

Chowdary, V.M., Yatindranath, Y., Kar, S., Adiga, S., 2004. Modelingof non-point source pollution in watershed using remote sensingand GIS. J. Indian Soc. Remote Sensing 32 (1), 59–73.

Cochrane, T.A., Flanagan, D.C., 1999. Assessing water erosion insmall watersheds using WEPP with GIS and digital elevationmodels. J. Soil Water Cons. 54 (4), 678–685.

Donigian Jr., A.S., Rao, P.S.C., 1990. Selections, application, andvalidation of environmental models. In: Decoursey, D.G., (Ed.),Proceedings of International Symposium on Water QualityModeling of Agricultural Nonpoint Sources. Part 2. June 19–23,1988. Logan, UT. USDA-ARS Report No. ARS-81. pp. 577–604.

Favis-Mortlock, D.T., Quinton, J.N., Dickinson, W.T., 1996. TheGCTE validation of soil erosion models for global change studies.J. Soil Water Cons. 51 (5), 397–402.

Flanagan, D.C., Nearing, M.A., 1995. USDA Water Erosion PredictionProject: Hillslope profile and watershed model documentation.NSERL Report No. 10. USDA-ARS National Soil Erosion ResearchLaboratory, West Lafayette, IN, USA, 47097–1196.

Flanagan, D.C., Nearing, M.A., 2000. Sediment particle sorting onhillslope profiles in theWEPPmodel. Trans. ASAE 43 (3), 573–583.

Ghidey, F., Alberts, E.E., 1996. Comparison of measured andpredicted runoff and soil loss for Midwest claypan Soil. Trans.ASAE. 39 (4), 1395–1402.

Ghidey, F., Alberts, E.E., Kramer, L.A., 1995. In: Comparison ofRunoff and Soil Loss Predictions from the WEPP Hillslope Modelto Measured Values for Eight Cropping and Management Treat-ments. ASAE Paper No. 95-2383, ASAE, St. Joseph, MI.

Gronsten, H.A., Lundekvam, H., 2006. Prediction of surface runoffand soil loss in southeastern Norway using the WEPP HillslopeModel. Soil Till. Res. 85 (1–2), 186–199.

Haan, C.T., Johnson, H.P., Brakensiek, D.L., 1982. Hydrologicalmodeling of small watersheds. Pub. Am. Soc. Agri. Eng., ASAE,St. Joseph, Michigan.

Hession, C.W., Shanholtz, V.O., 1988. A Geographic InformationSystem for targeting non-point source agricultural pollution. J.Soil Water Cons 43 (3), 264–266.

Kienzle, S.W., 1996. Using DTMs and GIS to define input variablesfor hydrological and geomorphological analysis. In: Kovar, K.,Nachtnebet, H., (Eds.), Applications of Geographic InformationSystems in Hyrology and Water Resources Managemnt, IAHS-AISHPublication No. 235. pp. 183–190.

Kramer, L.A., Alberts, E.E., 1995. In: Validation of WEPP 95.1 DailyErosion Simulation. ASAE Paper No. 95-2384, ASAE, St. Joseph,MI.

Laflen, J.M., Lane, J.L., Foster, G.R., 1991. WEPP – A NewGeneration of Erosion Prediction Technology. J. Soil Water Cons46 (1), 34–38.

Laflen, J.M., Flanagan, D.C., Engel, B.A., 2004. Soil Erosion andSediment Yield Prediction Accuracy Using WEPP. J. Am. W. Res.Asso. 40 (2), 289–297.

Liu, B.Y., Nearing, M.A., Baffaut, C., Ascough II, J.C., 1997. TheWEPP watershed model: III. Comparisons to measured data fromsmall watersheds. Trans. ASAE 40 (4), 945–951.

Maidment, D.R., 1993a. GIS and Hydrologic Modeling. In: Goodchild,M.F., Parks, B.O., Steyaert, L.T. (Eds.), Environmental Modelingwith GIS. Oxford University Press, NY.

McCuen, M.A., Snyder, W.M., 1983. Hydrologic Modeling: StatisticalMethods and Applications. Prentice-Hall, Englewood Cliffs, NJ.

Misra, K., 1999. Watershed management activities in DamodarValley Corporation at a glance, Soil Conservation Department.Damodar Valley Corporation, Hazaribagh, India.

Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting throughconceptual models Part 1-A discussion of principals. J. Hydrol.10 (3), 282–290.


Nearing, M.A., 1998. Why soil erosion models over-predicts smallsoil losses and under-predicted large soil losses. Catena 32 (1),15–22.

Nearing, M.A., Nicks, A.D., 1997. Evaluation of WEPP: Hillslope andsmall watershedsIn NATO-ASI book Global Change: Modeling SoilErosion by Water. Springer-Verlag, Oxford, England.

Nearing, M.A., Foster, G.R., Lane, L.J., Finkner, S.C., 1989. Aprocess-based soil erosion model for USDA – water erosionprediction project technology. Trans. ASAE 32 (5), 1587–1593.

Nearing, M.A., Deer-Ascough, L., Laflen, J.M., 1990. Sensitivityanalysis of the WEPP Hillslope profile erosion model. Trans. ASAE33 (3), 839–849.

Pandey, A., Chowdary, V.M., Mal, B.C., Dabral, P.P., 2005.Estimation of surface water potential of agricultural watershedusing geographic information system. Asian J. Geoinformatics 5(4), 29–36.

Pandey, A., Chowdary, V.M., Mal, B.C., 2007. Identification ofcritical erosion prone areas in the small agricultural watershedusing USLE, GIS and Remote Sensing. Water Resour Manage. 21(4), 729–746.

Rapp, J.F., 1994. Error assessment of the revised universal soillossequation using natural runoff plot data. M.S. Thesis, Universityof Arizona, Tucson, Arizona.

Reddy, M.S., 1999. Theme paper on ‘‘Water: Vision 2050’’. IndianWater Resources Soc. Roorkee, pp. 51–53.

Renschler, C.S., 2003. Designing geo-spatial interfaces to scaleprocess models: The GeoWEPP approach. Hydrol. Process. 17(5), 1005–1017.

Renschler, C.S., Harbor, J., 2002. Soil erosion assessment toolsfrom point to regional scales-the role of geomorphologists inland management research and implementation. Geomorphol-ogy 47, 189–209.

Renschler, C., Diekkruger, B., Mannaerts, C., 1997. Regionalizationin surface runoff and soil erosion risk evaluationRegionalizationof Hydrology, vol. 254. IAHS Publishers, UK, pp. 233–241.

Risse, L.M., Nearing, M.A., Nicks, A.D., Laflen, J.M., 1993. Assess-ment of error in the Universal Soil Loss Equation. Soil. Sci. Soc.Am. J. 57 (3), 825–833.

Risse, L.M., Nearing, M.A., Savabi, M.R., 1994. Determining Green-Ampt effective hydraulic conductivity from rainfall-runoff datafor the WEPP Model. Trans. ASAE 37 (2), 411–418.

Savabi, M.R., Flanagan, D.C., Hebel, B., Engel, B.A., 1995.Application of WEEP and GIS-GRASS to a small watershed inIndiana. J. Soil Water Cons. 50 (5), 477–483.

Sekhar, K.R., Rao, B.V., 2002. Evaluation of sediment yield by usingremote sensing and GIS: a case study from the Phulang Vaguwatershed, Nizamaba District (A.P). Int. J. Remote Sensing 23(20), 4449–4509.

Soto, B., Diaz-Fierros, F., 1998. Runoff and soil erosion from areasof burnt scrub: comparison of experimental results with thosepredicted by the WEPP model. CATENA 31, 257–270.

Srinivasan, R., Engel, B.A., 1994. A spatial decision support systemfor assessing agricultural nonpoint source pollution. WaterResour. Bull., AWRA 30 (3), 441–452.

Tim, U.S., Milner, M., Majure, J., 1992. Geographic informationsystems/simulation model linkage: Processes, problems andopportunities. ASAE paper No. 92-3610, ASAE, St. Joseph,Michigan.

Tiscareno-Lopez, M., Lopes, V.L., Stone, J.J., Lane, L.J., 1994.Sensitivity Analysis of the WEPP Watershed Model for Range LandApplications II. Channel Processes. Trans. of ASAE 37 (1), 151–158.

Tiwari, A.K., Risse, L.M., Nearing, M.A., 2000. Evaluation of WEPPand its Comparison with USLE and RUSLE. Trans. of ASAE 43 (5),1129–1135.

Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosionlossesUSDA Agricultural Research Services, Handbook 537. USDA,Washington, DC, pp. 57.

Young, R.A., Onstad, C.A., Bosch, D.D., Anderson, W.P., 1989.AGNPS: A Non-point Source Pollution model for evaluatingagricultural watersheds. J. Soil Water Cons. 44 (2), 168–173.

Yu, B., Rosewell, C.J., 2001. Evaluation of WEPP for runoff and soilloss prediction at Gunnedah, NSWAustralia. Aust. J. Soil Res. 39,1131–1145.

Yu, B., Ciesiolka, C.A.A., Rose, C.W., Coughlan, K.J., 2000. Avalidation test of WEPP to predict runoff and soil loss from apineapple farm on a sandy soil in subtropical QueenslandAustralia. Aust. J. Soil Res. 38, 537–554.

Zhang, X.C., Nearing, M.A., Risse, L.M., McGregor, K.C., 1996.Evaluation of runoff and soil loss predictions using natural runoffplot data. Trans. ASAE 39 (3), 855–863.

runoff and sediment yield modeling from a small agricultural watershed in india using the wepp model

Documents