Application of the WEPP model to determine sources of run-off

HYDROLOGICAL PROCESSES Hydrol. Process. (2014) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hyp.10168 Application of the WEPP model to determine sources of run-off and sediment in a forested watershed Bahram Saghafian,1* Amin Reza Meghdadi1 and Somayeh Sima2 1 Department of Technical and Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran 2 Department of Civil Engineering, Sharif University of Technology, P.O. Box 11155-9313, Tehran, Iran Abstract: This study investigates critical run-off and sediment production sources in a forested Kasilian watershed located in northern Iran. The Water Erosion Prediction Project (WEPP) watershed model was set up to simulate the run-off and sediment yields. WEPP was calibrated and validated against measured rainfall–run-off–sediment data. Results showed that simulated run-off and sediment yields of the watershed were in agreement with the measured data for the calibration and validation periods. While low and medium values of run-off and sediment yields were adequately simulated by the WEPP model, high run-off and sediment yield values were underestimated. Performance of the model was evaluated as very good and satisfactory during the calibration and validation stages, respectively. Total soil erosion and sediment load of the study watershed during the study period were determined to be 10 108 t yr 1 and 8735 t yr 1, respectively. The northern areas of the watershed with dry farming were identified as the critical erosion prone zones. To prioritize the subwatersheds based on their contribution to the run-off and sediment production at the watershed’s main outlet, unit response approach (URA) was applied. In this regard, subwatersheds close to the main outlet were found to have the highest contribution to sediment yield of the whole watershed. Results indicated that depending on the objective of land and water conservation practices, particularly, for controlling sediment yield at the main outlet, critical areas for implementing the best management practices may be identified through conjunctive application of WEPP and URA. Copyright © 2014 John Wiley & Sons, Ltd. KEY WORDS run-off; sediment yield; WEPP model; unit response approach; subwatershed prioritization; Kasilian watershed Received 27 February 2013; Accepted 30 January 2014 INTRODUCTION Sustainable land and water management under land use and climate changes is a key challenge worldwide (Gijsbers et al., 2001). Soil erosion in watersheds has become a major environmental concern impacting stream pollution and decreasing soil productivity (Singh et al., 2011). Surface run-off and soil erosion are typically low in forested watersheds because of surface litter cover. However, natural or human-induced disturbances can increase run-off and erosion by reducing surface cover and compacting soils (Elliot et al., 1999; Dun et al., 2009). To address soil erosion and water quality deterioration, the erosion prone hotspots throughout a watershed should be identified. This can help to plan the spatial distribution of the best management practices (BMPs). However, a prerequisite to any conservation strategy is the accurate estimation of run-off production and sediment transport *Correspondence to: Bahram Saghafian, Department of Technical and Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran E-mail: b.saghafi[email protected] Copyright © 2014 John Wiley & Sons, Ltd. along the whole watershed. This requires adequate knowledge of rainfall–run-off characteristics of the watershed, soil properties, prevailing land use, agricultural practices, and erosion controlling processes (Drohan et al., 2003; Kaleita et al., 2007). Coupled hydrological/erosion models are efficient tools to describe run-off and erosion processes, to reliably predict the quantity and the rate of run-off and sediment from land surface into hydrological networks, and to evaluate a variety of management scenarios without costly and lengthy field tests (Toy et al., 2002; Miller et al., 2007). A number of such models have been developed that may principally be classified into three categories: empirical models [e.g. Universal Soil Loss Equation (USLE), Revised USLE], physically based models [e.g. Water Erosion Prediction Project (WEPP)], and quasi-physically based models relying on mathematical process descriptions coupled with empirical relationships [e.g. Environmental Policy Integrated Climate (EPIC), Areal Non-point Source Watershed Environment Response Simulation (ANSWERS)] (Favis-Mortlock et al., 1996; Krysanova et al., 1998). Physically based models are generally superior to empirical models B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA because both spatial and temporal variability of natural processes may be considered in the simulation (Shen et al., 2009). Water Erosion Prediction Project is amongst the most promising physically based erosion models. Its hillslope module computes spatial and temporal distributions of soil loss and sediment deposition from overland flow on hillslopes, while its watershed module can also simulate soil loss and sediment deposition from concentrated flow in small channels (Flanagan and Nearing, 1995). Comparative studies performed on the efficiency of the WEPP model with several physically based and quasiphysically based models revealed that the WEPP outperformed EPIC and ANSWER (Bhuyan et al., 2002) and Soil and Water Assessment Tool (Shen et al., 2009) and Erosion 3D (Defersha and Melesse, 2012). Water Erosion Prediction Project has been successfully employed for run-off and erosion modelling in various geographical locations across USA (Huang et al., 1996; Tiwari et al., 2000; Laflen et al., 2004; Abaci and Papanicolaou, 2009), in Australia (Rosewell, 2001), and in Europe (Brazier et al., 2000; Pieri et al., 2007). However, little is known about the applicability of WEPP for countries with arid to semi-arid climate. Acceptable performance of the WEPP model was reported for simulating daily run-off and sediment yield in two catchments in India (Pandey et al., 2008, 2009; Singh et al., 2011). Moreover, Pandey et al. (2009) noted that WEPP can be efficiently used to identify the critical area in terms of sediment production by providing the output at any desired location within a watershed. On the contrary, several researchers discussed the unsatisfactory results of the WEPP model in arid regions. As an example, the efficiency of WEPP to predict soil erosion was assessed in a cultivated catchment in Tunisia (Raclot and Albergel, 2006). The results indicated that the model had poor performance in prediction of soil erosion, specifically at daily intervals. High discrepancies between the simulated and observed data were attributed to the weakness of the model in considering processes related to seasonal effects that occur in Mediterranean conditions. A study conducted in a semi-arid Mediterranean catchment also showed that the WEPP hillslope model cannot provide acceptable estimates of the surface run-off and soil loss (Albaradeyia et al., 2011). Recently, two studies have investigated the efficiency of the WEPP model in two semi-arid watersheds in Iran. The first study was conducted in a watershed in northern Iran to predict run-off and sediment yield (Ahmadi et al., 2011). The results indicated that WEPP underestimates sediment volumes and overpredicts run-off volumes by almost 25%. However, the capability of the model in the prediction of the general trend, low and high run-off and Copyright © 2014 John Wiley & Sons, Ltd. sediment yield, was not discussed in this study. In another study, Landi et al. (2011) assessed WEPP modelling capability to estimate the average soil loss in a small watershed located in south-west of Iran. Their results were compared with the Modified Pacific Southwest Inter-Agency Committee model predictions, and a high correlation between the two models was reported. However, one shortcoming of these studies is the poor calibration of the model as a result of lack of field data. In addition, assessment of the model efficiency in daily simulations of run-off and sediment yields was not reported in detail. Unit response approach (URA) may be applied to prioritize subwatersheds based on their contribution to the run-off and sediment load of the entire watershed. The URA is helpful in the absence of measured data at subwatershed scale. URA was originally developed with the aim of ranking subwatersheds based on their contribution to the flood peak generation at the main outlet of a watershed. The advantage of the URA prioritizing procedure is inclusion of discharge (and sediment for that purpose) routing within the watershed stream network (Saghafian and Khosroshahi, 2005). Recent application of the URA has been reported by Saghafian et al. (2012) for spatial prioritization of run-off and sediment sources in Iran. URA can help managers to effectively perform area selection for land conservation and water quality control practices through identification of critical subwatersheds. The objectives of this study were to (1) evaluate the performance of the WEPP model in a forested watershed at a daily time step and (2) identify critical areas in terms of their contribution to the run-off and sediment yield at the main watershed outlet. To accomplish these objectives, the WEPP watershed model was calibrated and validated against observed run-off and sediment data at the watershed outlet. Subsequently, URA was applied to prioritize the contribution of subwatersheds based on the run-off and sediment yield at the watershed scale. STUDY AREA The Kasilian watershed is located north of Iran and lies between 53°1′–53°9′E longitude and 36° 4′–36°8′N latitude. The watershed covers an area of 69 km2 with an elevation ranging from 1120 to 3123 m (Figure 1). The watershed is steep with an average slope of 24%. It has a semi-rectangular shape (with a shape factor of 0.23), which signals a relatively slow hydrologic response. The climate of the watershed is moderately humid and designated as ‘moderate Caspian climate’. Mean annual rainfall of the study area is about 960 mm, varying from Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Figure 1. Location of the Kasilian watershed (left), hydrometric and meteorological stations (centre), and watershed digital elevation model (right) 58 mm in June to 108 mm in September. Minimum and maximum monthly temperatures are 15 to 37 °C, respectively. The mean daily relative humidity varies from a minimum of 24% in March to its maximum of 56% in August. Bright sunshine hours vary from 9 to 14 during the dry months and 5 to 9 during the rainy months. Dominant soil types of the watershed are sand and silt. Limestone soils also can be sparsely observed. More than 70% of the watershed consists of natural forests with dense to moderate biomass. Agriculture is the second dominant land use, covering 20% of the watershed. Remaining lands are occupied by pastures and residential areas. The length of the main channel is 15.4 km, extending from the mountainous area in the southern part of the watershed towards the main outlet in the north. Run-off and sediment concentration are regularly recorded at the Valikbon hydrometric station located at the watershed outlet. Mean annual run-off of the main river at the outlet is approximately 0.45 m3 s 1. METHODOLOGY WEPP and GeoWEPP models Water Erosion Prediction Project is a continuous, process-based model for simulating soil erosion along a hillslope or within a watershed (Flanagan and Nearing, 1995). It has been developed based on numerous physically based equations to calculate the watershed run-off and erosion on a daily basis in small cultivated or forested watersheds, where the sediment yield at the outlet is significantly influenced by hillslope and channel processes (Foster et al., 1987; Baigorria and Romero, Copyright © 2014 John Wiley & Sons, Ltd. 2007). The model can simulate hydrologic processes such as infiltration, surface run-off, and sediment yield at the hillslope and watershed scales. The watershed version of WEPP consists of nine components: weather generation, winter processes, irrigation, surface hydrology, soils, plant growth, residue decomposition, overland flow hydraulics, and erosion. Having known the intensity and duration of a certain rainfall, WEPP computes cumulative infiltration using a Green–Ampt Mein–Larson model (Chu, 1978). Run-off is calculated as a result of rainfall, infiltration, and deposition storage on each hillslope for the entire simulation period. Subsequently, simulation results from each hillslope are combined, and then, run-off and sediment routing are performed along the channels and impoundments. Further details about the model are presented in the technical manual of WEPP (Flanagan and Nearing, 1995). Since its initiation in 1985, various improvements in the WEPP model have been made (Flanagan et al., 2007). As a result of efforts to link the WEPP model with geographical information system (GIS), an ESRI ArcView extension known as GeoWEPP (Renschler and Harbor, 2002) was released in 2001. GeoWEPP provides spatial graphical display outputs of predicted erosion risk areas in a watershed. Integration of WEPP with GIS facilitates data management particularly for WEPP applications at the watershed scale (Renschler, 2003). GeoWEPP (v2008.9) watershed version was used in this study. Unit response approach Unit response approach was used to prioritize subwatersheds of Kasilian based on their contribution to the run-off and sediment generation at the main outlet of Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA the watershed. It requires a calibrated rainfall–run-off model for performing discharge and sediment routing through the watershed. Within the URA, the two following indices are introduced to quantify the contribution of each subwatershed: the run-off index (YIk) and sediment index (SIk). Q QO:all K (1) YI K ¼ O:all QO:all SI K ¼ SO:all SO:all SO:all where K (2) Slope. A digital elevation model of the study area with a grid cell resolution of 50 m was produced based on the 1 : 25 000 topographic map of the watershed (Figure 1). Maps of slope, aspect, and slope shape factor were then extracted using GIS. The watershed was delineated into 305 hillslopes, with an average area of 22.28 ha, and 112 channels. Plant/Management. The plant/management input file contains all information related to plant variables (rangeland plant communities and cropland annual and perennial crops), tillage sequences and tillage implement parameters, plant and residue management, initial conditions, contouring, subsurface drainage, and crop rotations. Based on the land use map (Figure 2a), values of the crop variables and forest specifications were selected from the WEPP default database for each land type (type 1: forest perennial, 2: forest 5 year perennial, 3: agriculture alfalfa; soybean no till, 4: residual areas, and 5: pasture). The amount of ground cover is also determined based on the growth and mortality parameters. YIk: run-off index of the Kth subwatershed SIk: sediment index of the Kth subwatershed Qo, all: outlet discharge with all subwatersheds present in the base simulation (m3 s 1) Qo, all K: outlet discharge with the Kth subwatershed removed (m3 s 1) So, all: sediment load with all subwatershed units present Soils. Soil types of the Kasilian watershed were in the base simulation (t) obtained from a digital soil map of 1 : 25000 and used So, all K: sediment load with the Kth subwatershed in the simulation (Figure 2b). Supplementary physical and removed (t). chemical properties of the watershed soils collected along The first step in applying URA is to simulate the 12 well-distributed locations within the watershed at four watershed’s base state in which all subwatersheds contribute depths were also received from local offices. Summary of to the discharge and sediment yield at the main outlet. The land use and soil contents of subwatersheds is presented result of the base state is then used for comparison with other in Table I. states resulting from individual removal of subwatersheds. Afterwards, the effect of each subwatershed is individually Calibration and validation of WEPP quantified by removing it during the simulation. Then, its The model evaluation procedure consists of calibration, contribution to run-off and sediment at the main outlet is sensitivity analysis, and validation. Similar to any other estimated by Equations (1) and (2). hydrological model, WEPP should be accurately calibrated against field data in order to reduce uncertainty in WEPP model set-up model simulations (Engel et al., 2007). The split sample WEPP input data. The WEPP model requires four calibration approach was applied over the available data input files: topography, climate, soil, and plant/manage- between 2000 and 2001. The data was partitioned into ment file. These data were obtained from the available two parts: the first year for calibration and the second year local databases, meteorological stations, and the model for validation. default values as described in the following sections. Because soil input variables have been recognized as the foremost sensitive inputs of the model, the calibration Climate. The CLIGEN model was set up to generate process was performed on soil input parameters such as the climate file including daily precipitation, temperature, rill erodibility, interrill erodibility, effective hydraulic solar radiation, and wind speed. CLIGEN is an auxiliary conductivity, and effective hydraulic shear stress. The stochastic weather generation model developed to provide values of these parameters were chosen within the daily or single-storm climate input required to run WEPP prescribed range (Flanagan and Livingston, 1995) and (Nicks et al., 1995). The meteorological variables of the were adjusted through several simulations until a study area including maximum and minimum air minimum value of the root-mean-square error (RMSE) temperature, relative humidity, precipitation, solar radia- was obtained. tion, and wind speed were obtained from the Sangdeh and Sensitivity analysis, as a crucial part of the simulation, Valikben weather stations (Figure 1) and subsequently examines the response of a model output over a range of transformed into the CLIGEN format. input variables that determines how a relative perturbation Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Figure 2. Land use and soil type maps of the Kasilian watershed Table I. Land use and soil contents of subwatersheds Land use (%) Subwatershed 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 High density forest Medium density forest Dry farming 87.2 64 62 95.9 97.1 96.8 100 69.7 1.3 4.6 2.3 11.6 4.1 0.3 5.1 2.2 9.7 1.5 61.5 4.4 71.8 90.3 Soil content (%) Residential area 2.9 3.2 30.3 87.1 84.6 38.2 31.3 9.5 11.6 15 46.2 66.8 38.5 95.6 28.2 0.2 of the parameter is propagated into the relative perturbation of the prediction (Hantush and Kalin, 2005). Through using the relative sensitivity relationship (McCuen, 1973), the effect of changes in input variables can be assessed. Several researchers have employed a relative sensitivity concept to perform sensitivity analysis of WEPP (e.g. Nearing et al., 1990; Brunner et al., 2004; Singh et al., 2011). Relative sensitivity analysis was also adopted in this study. Weather and soil variables are recognized as the two important inputs for many hydrological models (Nearing Copyright © 2014 John Wiley & Sons, Ltd. 0.4 5.9 0.4 Rangeland Clay Silt Sand Organic matter 7.9 28.6 24.2 32.1 27.9 39.2 28.3 28.9 29.1 30.4 32.2 20.5 31 30 27 28 14.5 36 36.4 28 25.1 20 13 35 37.3 28 23.8 41 43.3 41.9 43 26 56.4 26 42 24 30 20.8 38.7 20.1 19.9 14.2 28 27.5 15.1 14.2 15 35 19.9 19 14 3.9 3.2 2 3.8 4.1 4.5 3.2 4 6.1 5.9 7 6.2 5.3 7.2 3.1 2.8 et al., 1990; Baffaut et al., 1997). Because weather data are commonly recorded by precisely calibrated automatic weather stations, manual error in their measurement is minimal. Thus, sensitivity analysis of the model was carried out only on soil input variables. The value of each variable was altered within a range of ±50% of its calibrated value, while keeping other parameters constant. Subsequently, sensitivity ratios were determined by comparing the corresponding simulated run-off and sediment yield. When the calibrated values of the model parameters were determined, the model was validated. Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA Recorded data Run-off and sediment concentration data corresponding to each rainfall event were obtained from the hydrometric records of the Valikbon station. Then, run-off depths (in mm) were calculated from the observed run-off data (m3 s 1) divided by the area of the watershed. The observed sediment yield data (in kg ha 1) were also estimated using the recorded sediment concentration data, watershed area, and run-off volume. The daily run-off depth and the sediment yield data at the watershed outlet were used for calibration and validation of the model. Model performance criteria Numerous model evaluation criteria have been proposed for assessing the performance of watershed models. However, because performance measures are model and project specific, no universal measure exists (Moriasi et al., 2007). Amongst many, Pearson’s correlation coefficient (r) and coefficient of determination (R2), RMSE (Thomann, 1982), Nash–Sutcliffe efficiency (NSE) (Nash and Sutcliffe, 1970), and percent bias (PBIAS) (Gupta et al., 1999) are the most widely used measures. Moriasi et al. (2007) investigated the suitability of several performance measures and recommended three quantitative statistics including NSE, PBIAS, and ratio of the RMSE to the standard deviation of measured data (RSR). They also proposed numerical threshold values for these measures and defined corresponding performance ratings (Table II). We used several statistical measures including R2, NSE, RSR, and PBIAS for quantitative evaluation of the WEPP model. The formulations of the NSE, RSR, and PBIAS for run-off discharge are as follows:
2 ∑ni¼1 Qi ^ Qi NSE ¼ 1 (3)
2 ∑ni¼1 Qi Q qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi n
^ 2 ∑ Q Q i RMSE i i¼1 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (4) RSR ¼
2 STDEVobs ∑ni¼1 Qi Q
 ∑ni¼1 Qi ^ Qi PBIAS ¼  100 (5) ∑ni¼1 Qi where Qi and^ Q i are the measured and simulated daily run-off in (m3 s 1), respectively. Q is the average measured discharge in (m3 s 1) and n is the number of observations. RESULTS AND DISCUSSION Sensitivity analysis Calibrated parameters and input variables used in the sensitivity analysis of the model are presented in Table III. All parameters considered for the sensitivity analysis were soil-related parameters. We found that run-off is mainly sensitive to the effective hydraulic conductivity (Ke), while sediment yield is sensitive to interrill erodibility (Ki), rill erodibility (Kr), and critical hydraulic shear stress (τc). Similar results have been reported for sensitivity analysis of the WEPP model (e.g. Brunner et al., 2004; Pandey et al., 2008; Singh et al., 2011). High sensitivity of the model to interrill erodibility indicated that interrill erosion was the dominant process in sediment production in the Kasilian watershed. Therefore, accurate estimations of Ke and Ki parameters are needed for predictions of watershed yield. Lack of field data on Ke and Ki prohibited us from direct verification of their values. However, the calibrated values of these parameters were within the reported ranges in the literature (e.g. Singh et al., 2011). Calibration of the WEPP model Observed and simulated daily run-off and sediment yield data of the Kasilian watershed, for a total of 54 events in 2000, are compared in Figures 3 and 4, respectively. It is seen that the overall trend of the simulated values closely matches the trend of the measured values for both run-off and sediment yields, although very low run-off and sediment yields are mostly underpredicted by the model. Furthermore, while high run-off values are reasonably predicted, peak values of sediment yield are overestimated. Nevertheless, the cumulative curve of simulated sediment yield is well consistent with that of the observed data (Figure 4). This Table II. Performance ratings for recommended statistics to assess the watershed models at a monthly time step (Moriasi et al. 2007) PBIAS Performance rating RSR NSE Stream flow Sediment Very good Good Satisfactory Unsatisfactory 0.00 ≤ RSR ≤ 0.50 0.50 < RSR ≤ 0.60 0.60 < RSR ≤ 0.70 RSR > 0.70 0.75 < NSE ≤ 1.00 0.65 < NSE ≤ 0.75 0.50 < NSE ≤ 0.65 NSE ≤ 0.5 PBIAS < ±10 ±10 ≤ PBIAS < ±15 ±15 ≤ PBIAS < ±25 PBIAS ≥ ±25 PBIAS < ±15 ±15 ≤ PBIAS < ±30 ±30 ≤ PBIAS < ±55 PBIAS ≥ ±55 Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Table III. Sensitive soil-related parameters of the WEPP model Relative change in output variables with respect to variation in input parameters (%) Output variable Sensitive parameters Run-off Sediment yield Ke Ke Ki Kr τc (mm h 1) (mmh 1) (kg s m 4) (s m 1) (N m 1) Calibrated value 10 20 25 50 10.75 10.75 1.9105 0.011 6.4 1.22 0.98 0.93 1.3 0.13 1.89 2.25 3.1 1.9 2.01 3.2 4 5.15 2.1 3.11 6.8 7.2 8.23 3.05 4.23 10 20 25 50 1.98 1.2 1.77 0.58 0.61 4.43 4.32 2.6 1.3 0.82 7.11 6.26 3.05 1.72 1.75 11.25 8.22 3.8 4.41 3.68 Figure 3. Observed and simulated daily run-off during the calibration period Figure 4. Observed and simulated daily sediment yield as well as cumulative sediment yield during the calibration period Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA may be attributed to the fact that uncertainties in daily modelling of soil erosion are high and rooted in the complex interactions between rainfall and watershed characteristics. Typically, as the model time scale increases, its performance improves by reducing the randomness of model parameters (Engel et al., 2007). Several researchers confirm that model simulations are poorer for shorter time scales than for longer time scales (e.g. daily vs monthly or yearly) (Santhi et al., 2001; Van Liew et al., 2007). Figures 5 and 6 also show the scatter plots of the simulated run-off and sediment yield against the observed data. The data points are properly scattered around the 45° lines with small bias values of 0.79 (mm) and 2.57 (kg ha 1) for run-off and sediment yields, respectively. In general, the model underestimates high run-off values, whereas it slightly overestimates the sediment yield. Moreover, for the case of the Kasilian watershed, performance of the WEPP model in estimating run-off and sediment yield was found well. To quantify the performance of the model during the calibration and validation periods, several goodness-of-fit statistics for daily run-off and sediment were calculated and are presented in Table IV. Mean and standard deviation of the observed and simulated run-off and sediment are close. Discrepancies of 17 and 7% were identified between the predicted and observed values of maximum run-off and sediment. For the minimum values, as previously discussed, the model cannot properly represent extreme low run-off and sediment values. NSE, reflecting the overall fit of a hydrograph (Sevat and Dezetter, 1991), also turned out to be 0.76 and 0.82 for run-off and sediment yield, respectively. According to the performance ratings described by Moriasi et al. (2007), model prediction in the calibration stage can be rated as very good (Table II). Calculated values for other performance statistics such as R and RMSE also support the fact that overall prediction of daily surface run-off and sediment by the WEPP model during the calibration period is satisfactory. Thus, the model can reproduce the watershed hydrologic response. Validation of the model Figure 5. Measured and simulated daily run-off values for model calibration Figure 6. Measured and simulated daily sediment yield for model calibration Copyright © 2014 John Wiley & Sons, Ltd. The calibrated WEPP model was used to simulate daily run-off and sediment yield during the year of 2001 with 45 rainfall events (Figures 7 and 8). Figure 7 shows that the temporal variation of the run-off is consistent with the seasonal pattern of the rainfall over the watershed. This is because of the fact that in a steep forested watershed mainly covered by clay loam, both infiltration rate and time of concentration are low. Thus, appreciable rainfall events may easily produce considerable run-off volume at the watershed outlet. The simulated run-off and sediment patterns also suggest that the general variations of the hydrograph and sediment yield can be reasonably predicted by the model. Low values of the simulated run-off match relatively well with the corresponding measured run-offs. Nevertheless, the model underpredicts a few peak run-off values. Looking at the predicted sediment yield, the model fails to capture very high peaks. On the contrary, very low sediment yields occurring in the summer months are well predicted. Moreover, medium sediment yield values can be reasonably simulated by the model. Underprediction of high run-off and sediment by the WEPP model has been also reported in several studies Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Table IV. Goodness-of-fit statistics for measured and simulated daily run-off and sediment yield during calibration and validation periods Calibration Run-off (mm) Validation Sediment yield (kg ha 1) Run-off (mm) Sediment yield (kg ha 1) Statistics Observed Simulated Observed Simulated Observed Simulated Observed Simulated Mean Stddev Minimum Maximum R RMSE NSE PBIAS RSR 8.20 9.13 0.79 44.85 7.62 8.69 0.14 37.35 28.49 30.21 6.39 152.22 28.32 35.60 0.10 162.28 6.21 9.02 0.89 37.10 4.83 3.73 1.56 16.49 20.94 38.14 0.12 183.46 20.16 40.70 0.12 224.08 0.87 4.47 0.76 7.07 0.49 0.92 14.05 0.82 0.60 0.47 0.89 6.05 0.54 22.30 0.67 0.89 18.14 0.77 3.73 0.48 Figure 7. Observed and simulated daily run-off during the validation period (e.g. Abaci and Papanicolaou, 2009; Dun et al., 2009; Singh et al., 2011). There may be several reasons for underprediction of larger run-offs. First, this can arise from the difference between the temporal variations of high rainfall during the calibration and validation years. In the case of Kasilian, high rainfall typically occurs during spring. However, in 2001, the rainfall distribution has been considerably varied in comparison with its longterm distribution and that of 2000 (Figure 9). Because the model parameters have been calibrated based on the runoff and sediment data of 2000, this may cause the underestimation of unusual peak run-offs occurring in fall. Second, to properly model hydrologic processes in a forested watershed, it is crucial to adequately simulate Copyright © 2014 John Wiley & Sons, Ltd. lateral flow processes as a dominant variable (Covert et al., 2005). However, previous studies on the application of WEPP in forest areas indicted that WEPP underestimates subsurface lateral flows (Elliot et al., 1995; Dun et al., 2009). This drawback, in turn, leads to underprediction of run-off and sediment yields for forested watersheds. Third, underprediction of peak runoff may arise from insufficient calibration of subsurface parameters. To improve the simulation accuracy of WEPP, particularly in a forested watershed, an elaborate calibration of the subsurface flow parameters is required (Singh et al., 2011). Cumulative sediment yield was compared with that of the observed in Figure 10. Overall, the variation of the Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA Figure 8. Observed and simulated daily sediment yield during the validation period Figure 9. Monthly distribution of rainfall in the Kasilian watershed during the study period cumulative predicted sediment yield follows the cumulative observed sediment yield trend, having a negative marginal shift. Total predicted sediment yield (823 kg ha 1) of all events is slightly less than the total observed sediment yield (855 kg ha 1). To analyse the discrepancies between the measured and modelled predictions, values of percent error (PE) were plotted against rainfall intensity for run-off and sediment values (Figures 11 and 12). For run-off, nearly all values corresponding to the rainfall events of higher than 20-mm Copyright © 2014 John Wiley & Sons, Ltd. depth are predicted with the PE of higher than 50%. However, large errors in the estimation of the sediment yields mostly occur at low rainfall intensities. Consequently, maximum discrepancies for run-off prediction are not necessarily concurrent with those of the sediment yield. Additionally, almost three quarters of the predicted run-off values were underestimated, while for sediment yield, the number of overestimations and underestimations is equal. Table IV further presents calculated values of some performance measures during the validation period. Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Figure 10. Observed and simulated cumulative sediment yield during the validation period Figure 11. Percent error in daily run-off simulation during 2001(errors and rainfall heights are shown by diamonds and bars, respectively) Subsequently, the efficiency of the model was judged based on the threshold values of these measures reported by Moriasi et al. (2007) (Table II). Through comparing NSE, PBIAS, and RSR values with their recommended values for performance rating, WEPP predictions can be assessed as satisfactory for run-off and sediment yield. Copyright © 2014 John Wiley & Sons, Ltd. Sediment load and soil loss Mean annual run-off and sediment load of subwatersheds at their own outlets are presented in Table V. The maximum run-off generation belongs to subwatershed 4, followed by subwatersheds 1, 2, and 11, which are amongst the largest subwatersheds. The least amount of run-off is produced at the outlet of subwatershed 14, the smallest of Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA Figure 12. Percent error in daily sediment yield simulation during 2001(errors and rainfall heights are shown by diamonds and bars, respectively) Table V. Annual run-off and sediment loads of Kasilian subwatersheds based on the WEPP model outputs and URA indices in 2001 Subwatershed 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Area (km2) Slope (%) Run-off (km3) Sediment load (t) YI (%) SI (%) 8.0 8.4 3.2 8.1 4.9 3.1 1.5 2.5 1.5 1.8 10.3 5.3 3.3 0.8 3.8 3.3 41 58 27 24 18 10 18 16 18 18 8 6 28 20 9 19 3.9 3.4 0.9 4.1 2.1 1.4 0.9 1.1 1.2 1.3 3.4 3.3 0.6 0.3 3.2 1.9 1068 1212 358 825 452 464 234 112 46 120 997 1009 606 431 1035 28 13.6 9.7 6.5 10.3 5.0 6.9 2.1 2.5 2.9 1.8 11.7 7.9 4.8 2.6 6.8 5.0 8.1 4.1 4.4 4.3 3.1 2.7 1.0 1.2 3.6 5.5 14.4 15.9 8.7 5.9 13.2 4.0 all subwatersheds. For sediment load, area, slope, and soil texture of the subwatersheds are the prominent factors. The steep large subwatersheds 1 and 2 are prone to produce the highest loads. On the other hand, subwatersheds 16 and 9 produce minimum sediment loads. Moreover, both run-off and sediment load positively correlate with the area of subwatersheds, having an R2 of 0.76 and 0.61, respectively. Because the large portion of the watershed’s soil consists of clay, to assess the effect of soil texture on the run-off and sediment load of each subwatershed, Copyright © 2014 John Wiley & Sons, Ltd. regression analysis was performed between the percent of the clay loam, as the prevailing soil texture, and the run-off volumes and sediment loads of subwatersheds. While there is no meaningful correlation between the runoff and clay loam content of the soil (R2=0.26), sediment loads reasonably correlate with the clay loam percent content of the soils (Figure 13). This implies that the amount of fine-texture soils in the watershed adversely affects the sediment load because the clay binds soil particles and produces more resistance to erosion. Run-off generation can be typically intensified in fine-texture soils Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Figure 13. Correlation between the sediment load and the clay loam content of the soils in the Kasilian watershed because of low infiltration. However, the regression analysis indicates that the effect of area and slope is more dominant in producing run-off compared with that of the soil texture. Additionally, run-off and sediment load of subwatersheds are correlated with each other (R2 = 0.61). Therefore, those areas producing larger volumes of run-off generally generate higher sediment load. This can be attributed to the increased flow carrying capacity. However, at low run-off volumes, sediment load is not necessarily minimal. For instance, subwatersheds 8 and 9 have similar run-off volumes, whereas sediment load produced at subwatershed 8 is 2.5 times that of subwatershed 9. The soil loss map through the entire watershed was obtained from the results of the validated model (Figure 14). The majority of the watershed area has an erosion rate of less than 0.25 t ha 1 yr 1. The average erosion rate of the Kasilian watershed for 2001 turned out to be 1.2 t ha 1 yr 1. According to Pandey et al. (2009), there is a need to impose soil conservation measures on the subwatersheds where the average sediment production rate exceeds 2.5 t ha 1 yr 1. In Kasilian, more than 20% of the watershed has a soil erosion rate of over 4 t ha 1 yr 1. As illustrated in Figure 14, the three critical soil erosion zones within the watershed include (1) areas under dry farming, (2) rangeland in the southern part of the watershed, and (3) scattered spots in the middle reaches of the watersheds. The dominant erosion prone areas are dry farming lands at the north part of the watershed (Figure 2a). This arises from the fact that dry farming is Copyright © 2014 John Wiley & Sons, Ltd. usually associated with significant cultivation activities, often in the spring when erosion may intensify. The high erosion areas in the southern part of the watershed are also related to steep lands having the sandy soil with little organic matter. Sandy soils lacking cohesiveness may easily erode while subject to considerable rainfall in these regions. Moreover, high soil loss in the middle reaches of the watershed along the stream network is because of the stream erosion. Stream erosion is intensified during the storm events and widens the bed into the already narrow floodplain and hillslopes. Distribution of high soil loss patches adjacent to the river network may be an evidence of the stream erosion in the middle of the watershed. From the extracted soil erosion map, it can be deduced that managing inappropriate conventional farming and adopting proper tillage should be applied to effectively control erosion. Determining critical subwatersheds The URA was applied to prioritize subwatersheds based on their contribution to the run-off and sediment yield at the main outlet. The mean annual run-off and sediment yield of the whole watershed simulated by the WEPP model were used to calculate the YI and SI indices (refer to the last two columns in Table V). Relative share of the subwatersheds can be compared in Figure 15. Subwatersheds 1, 11, 4, and 2 contribute most to the runoff production at the watershed main outlet, whereas the minimum contribution belongs to subwatershed 10, which is a small subwatershed. Subwatershed 12 has the highest contribution considering the sediment load at Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA Figure 14. Soil erosion map of the Kasilian watershed in 2001 the main outlet, while the lowest share belongs to subwatershed 7. Overall, in the Kasilian watershed, ranking of the subwatersheds based on the YI index is almost similar to the WEPP model outputs. This means that subwatersheds having the highest run-off at their own outlets are the critical subwatersheds in run-off contribution at the main outlet of the whole watershed. The R2 value of 0.84 between the YI values and subwatershed area also reveals that larger subwatersheds are expected to have the most determining effect on the run-off production at the watershed outlet. As a result, through controlling runoff volumes at these subwatersheds, it is possible to control the run-off volume exiting the whole watershed. For sediment yield, on the contrary, URA results indicated that subwatersheds located close to the watershed outlet have the highest contribution to the sediment yield of the entire watershed. Copyright © 2014 John Wiley & Sons, Ltd. These are mostly subwatersheds containing dry farming. URA can also be used to assess the effect of the subwatersheds with mixed land use on the overall sediment production at the watershed main outlet. The results of URA showed that as dry farming in subwatersheds expands, the value of the SI index increases. This implies the increased contribution of the subwatersheds with dry farming land cover to the sediment production at the main outlet of the whole watershed. However, an accurate interpretation of the conjunctive effect of the mixed land uses on the sediment yield of the whole watershed is difficult because of the interaction between vegetation cover, soil type, area, and slope. The area and slope of subwatersheds were identified as the key parameters affecting the sediment yield of each subwatershed. However, for determining sediment hotspots based on the sediment yield contribution at the Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Figure 15. Comparison of run-off and sediment indices for the Kasilian subwatersheds in 2001 watershed outlet, area, land cover, and distance from the outlet become important. Consequently, depending on the land and water conservation objectives, different parts of the watershed should be considered for effective implementation of BMPs. • CONCLUSIONS • This study reported the application of the WEPP model to simulate daily run-off and sediment yield in an Iranian forested watershed. After calibration and validation, the model was used to determine the critical areas in terms of sediment yield and soil loss. Furthermore, to prioritize the subwatersheds based on their contribution to the run-off and sediment yield at the watershed main outlet, URA was employed. The following conclusions were drawn from this study: • • Sensitivity analysis showed that effective hydraulic conductivity and interrill erodibility are the most sensitive parameters in sediment yield simulation, while run-off is mainly sensitive to effective hydraulic conductivity. • Based on the calculated values of several performance statistics, the model prediction at the calibration and validation stages was rated as very good and satisfactory, respectively. • Simulated values of run-off showed that the maximum run-off generation at subwatershed scale belongs to subwatershed 4 followed by subwatersheds 1, 2, and 11, which are amongst the largest subwatersheds. This Copyright © 2014 John Wiley & Sons, Ltd. • is while lower run-off volume is produced by subwatershed 14, being the smallest subwatershed. Area and slope of the subwatersheds were distinguished to be the most influential factors on the sediment yield of each subwatershed. Thus, large subwatersheds with steep slopes are known to produce the highest sediment loads. Erosion risk areas are located in the northern part of the watershed in which dry farming is the dominant land use. In the Kasilian watershed, the critical subwatersheds in terms of run-off contribution at the main outlet were the four largest subwatersheds. However, subwatersheds close to the main outlet were found to have the highest contribution to the sediment yield of the whole watershed. To improve the simulation capability of WEPP, it is recommended that run-off and sediment yield data are recorded at hillslope/subwatershed scales and are used for calibration. Successful application of the WEPP model in conjunction with URA, as presented in this case, can provide insights into the site selection for implementation of BMPs throughout a watershed whether with the objective of land conservation and sediment load minimization at dam sites or pollutant control at the watershed main outlet. ACKNOWLEDGEMENT The authors appreciate Ms.Allison Fonga for her review and assistance in improving the language of the paper. Hydrol. Process. (2014) B. SAGHAFIAN, A. R. MEGHDADI AND S. SIMA REFERENCES Abaci O, Papanicolaou ANT. 2009. Long-term effects of management practices on water-driven soil erosion in an intense agricultural subwatershed: monitoring and modeling. Hydrological Process 23: 2818–2837. Ahmadi H, Taheri S, Feiznia S, Azarnivand H. 2011. Runoff and sediment yield modeling using WEPP in a semi-arid environment (Case study: Orazan Watershed). Desert 16: 5–12. Albaradeyia I, Azzedine H, Shahrour I. 2011. WEPP and ANN models for simulating soil loss and runoff in a semi-arid Mediterranean region. Environmental Monitoring Assessment 180: 537–556. Baffaut C, Nearing MA, Ascough JC, Liu BY. 1997. The WEPP watershed model: II. Sensitivity analysis and discretization on small watersheds. Transactions of the ASABE 40(4): 935–943. Baigorria GA, Romero CC. 2007. Assessment of erosion hotspots in a watershed: Integrating the WEPP model and GIS in a case study in the Peruvian Andes. Environmental Modelling& Software 22: 1175–1183. Bhuyan SJ, Kalita PK, Janssen KA, Barnes PL. 2002. Soil loss predictions with three erosion simulation models. Environmental Modelling & Software 17(2): 135–144. Brazier RE, Beven KJ, Freer J, Rowan JS. 2000. Equifinality and uncertainty in physically based soil erosion models: application of the GLUE methodology to WEPP the Water Erosion Prediction Project, for sites in the UK and USA. Earth Surface Processes Landforms 25: 825–845. Brunner AC, Park SJ, Ruecker GR, Dikau R, Vlek PLG. 2004. Catenary soil development influencing erosion susceptibility along a hillslope in Uganda. CATENA 58: 1–22. Chu ST. 1978. Infiltration during an unsteady rain. Water Resources Research 14: 461–466. Covert SA, Robichaud PR, Elliot WJ, Link TE. 2005. Evaluation of runoff prediction from WEPP-based erosion models for harvested and burned forest watersheds. Transactions of the ASAE 48: 1091–1100. Defersha MB, Melesse AM. 2012. Field-scale investigation of the effect of land use on sediment yield and runoff using runoff plot data and models in the Mara River basin, Kenya. CATENA 89: 54–64. Drohan PJ, Ciolkosz EJ, Petersen GW. 2003. Soil survey mapping unit accuracy in forested field plots in Northern Pennsylvania. Soil Science Society of America Journal 67: 208–214. Dun S, Wu JQ, Elliot WJ, Robichaud PR, Flanagan DC, Frankenberger JR, Brown RE, Xu AD. 2009. Adapting the Water Erosion Prediction Project (WEPP) model for forest applications. Journal of Hydrology 336(1-4): 45–54. Elliot WJ, Robichaud PR, Luce CH. 1995. Applying the WEPP Erosion Model to Timber Harvest Areas. In: Proc. ASCE Watershed Manage Conference, August 14–16, 1995. San Antonio, TX, pp. 83–92. Elliot WJ, Page-Dumroese D, Robichaud PR. 1999. The effects of forest management on erosion and soil productivity. Proceedings of the Symposium on Soil Quality and Erosion Interaction, Keystone, CO, July 7, 1996. Ankeney, IA: Soil and Water Conservation Society. 16 p. Engel B, Storm D, White M, Arnold J, Arabi M. 2007. A hydrologic/water quality model application protocol. Journal of American Water Resource Association 43(5): 1223–1236. Favis-Mortlock DT, Quinton N, Dickinson WT. 1996. The GCTE validation of soil erosion models for global change studies. Journal of Soil and Water Conservation 51(5): 397–402. Flanagan DC, Livingston S J. 1995. USDA-Water Erosion Prediction Project-WEPP User Summary. NSREL Report No.11. USDA-ARS National Soil Erosion Res. Lab: West Lafayette; IN, USA. Flanagan DC, Nearing MA. 1995. USDA-Water Erosion Prediction Project: Hillslope Profile and Watershed Model Documentation. NSERL Report No. 10. USDA-ARS National Soil Erosion Res. Lab: West Lafayette; IN, USA. Flanagan DC, Gilley JE, Franti TG. 2007. Water erosion prediction project (WEPP): Development history, model capabilities, and future enhancements. Transaction of the ASABE 50(5): 1603–1612. Foster GR, Lane LJ, Schertz DL, Nordin JO, Wingate GD. 1987. User Requirements: USDA-Water Erosion Prediction Project (WEPP), NSERL Report 1. USDA-ARS National Soil Erosion Research Laboratory: West Lafayette; 43 pp. Copyright © 2014 John Wiley & Sons, Ltd. Gijsbers G, Janssen W, Odame HH, Meijerink G (Eds.). 2001. Planning Agricultural Research: A Source book. CABI Publishing: New York; 285. Gupta HV, Sorooshian S, Yapo PO. 1999. Status of automatic calibration for hydrologic models: Comparison with multi-level expert calibration. Journal of Hydrologic Engineering 4(2): 135–143. Hantush MM, Kalin L. 2005. Uncertainty and sensitivity analysis of runoff and sediment yield in a small agricultural watershed with KINROS2. Hydrological Science Journal 50(6): 1151–1171. Huang CH, Bradford JM, Laflen JM. 1996. Evaluation of the detachment transport coupling concept in the WEPP rill erosion equation. Soil Science Society of American Journal 60: 734–739. Kaleita AL, Hirschi MC, Tian LF. 2007. Field-scale surface soil moisture patterns and their relationship to topographic indices. Transactions of the ASABE 50(2): 557–564. Krysanova V, Becker A, Klocking B. 1998. The linkage between hydrological processes and sediment transport at river basin scale, Modelling Soil Erosion, Sediment Transport and Closely Related Hydrological Processes, Proceedings of a symposium held at Vienna, July 1998. IAHS Publication, No. 249, 1998. Laflen JM, Flanagan DC, Engel BA. 2004. Soil Erosion and Sediment Yield Prediction Accuracy Using WEPP. Journal of American Water Resource Association 40(2): 289–297. Landi A, Barzegar AR, Sayadi J, Khademalrasoul A. 2011. Assessment of Soil Loss Using WEPP Model and Geographical Information System. Journal of Spatial Hydrology 11(1). Spring 2011. McCuen RH. 1973. The role of sensitivity analysis in hydrologic modeling. Journal of Hydrology 18: 37–53. Miller SN, Semmens DJ, Goodrich DC, Hernandez M, Miller RC, Kepner WG, Guertin DP. 2007. The automated geospatial watershed assessment tool. Environmental Modelling and Software 22(3): 365–377. Moriasi DN, Arnold JG, Van Liew MW, Bingner RL, Harmel RD, Veith TL. 2007. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Transaction of the ASABE 50(3): 885–900. Nash JE, Sutcliffe JV. 1970. River flow forecasting through conceptual models: Part 1. A discussion of principles. Journal of Hydrology 10(3): 282–290. Nearing MA, Deer-Ascough L, Laflen JM.1990. Sensitivity analysis of the WEPP hillslope profile erosion model. Transaction of the ASABE 33(3): 839–849. Nicks AD, Lane LJ, Gander GA. 1995. Weather generator. Chapter 2 in USDA–Water Erosion Prediction Project. Hillslope Profile and Watershed Model Documentation. D.C. Flanagan M A, Nearing Eds. NSERL Report No. 10. West Lafayette, IND, USDA–ARS National Soil Erosion Research Laboratory. Pandey A, Chowdary VM, Mal BC, Billib M. 2008. Runoff and sediment yield modelling from a small agricultural watershed in India using the WEPP model. Journal of Hydrology 348: 305–319. Pandey A, Chowdary VM, Mal BC, Billib M. 2009.Application of the WEPP model for prioritization and evaluation of best management practices in an Indian watershed. Hydrological Process 23: 2997–3005. Pieri L, Bittelli M, Wu JQ, Dun S, Flanagan DC, Pisa PR, Ventura F, Salvatorelli F. 2007. Using the water erosion prediction project (WEPP) model to simulate field-observed runoff and erosion in the Apennines mountain range, Italy. Journal of Hydrology 336: 84–97. Raclot D, Albergel J. 2006. Runoff and water erosion modelling using WEPP on a Mediterranean cultivated catchment. Physics Chemistry of the Earth 31: 1038–1047. Renschler CS. 2003. Designing geo-spatial interfaces to scale process models: The GeoWEPP approach. Hydrological Processes 17(5): 1005–1017. Renschler CS, Harbor J. 2002. Soil erosion assessment tools from point to regional scales-the role of geomorphologists inland management research and implementation. Geomorphology 47: 189–209. Rosewell CJ. 2001. Evaluation of WEPP for runoff and soil loss prediction in Gunnedah, NSW, Australia. Australian Journal of Soil Research 9: 230–243. Saghafian B, Khosroshahi M. 2005. Unit response approach for priority determination of flood source areas. ASCE Journal of Hydrologic Engineering 25: 270–277. Saghafian B, Sima S, Sadeghi C, Jeirani F. 2012. Application of unit response approach for spatial prioritization of runoff and sediment sources. Agricultural Water Management 109: 36–45. Hydrol. Process. (2014) USING WEPP FOR RUN-OFF AND SEDIMENT SIMULATION IN A FORESTED WATERSHED Santhi C, Arnold JG, Williams JR, Dugas WA, Srinivasan R, Hauck LM. 2001. Validation of the SWAT model on a large river basin with point and nonpoint sources. Journal of American Water Resource Association 37(5): 1169–1188. Sevat E, Dezetter A. 1991. Selection of calibration objective functions in the context of rainfall-runoff modeling in a Sudanese savannah area. Hydrological Science Journal 36(4): 307–330. Shen ZY, Gong YW, Li YH, Hong Q, Xu L, Liu RM. 2009. A comparison of WEPP and SWAT for modeling soil erosion of the Zhangjiachong Watershed in the Three Gorges Reservoir Area. Agricultural Water Management 96(10): 1435–1442. Singh RK, Panda RK, Satapathy KK, Ngachan SV. 2011. Simulation of runoff and sediment yield from a hilly watershed in the eastern Copyright © 2014 John Wiley & Sons, Ltd. Himalaya, India using the WEPP model. Journal of Hydrology 405: 261–276. Thomann RV. 1982. Verification of water quality models. Journal of Environmental Engineering 108(5): 923–940. Tiwari AK, Risse LM, Nearing MA. 2000. Evaluation of WEPP and its Comparison with USLE and RUSLE. Transaction of the ASABE 43(5): 1129–1135. Toy TJ, Foster GR, Renard KG. 2002. Soil Erosion: Processes, Prediction, Measurement and Control. Wiley: New York. Van Liew MW, Veith TL, Bosch DD, Arnold JG. 2007. Suitability of SWAT for the conservation effects assessment project: A comparison on USDA-ARS experimental watersheds. Journal of Hydrologic Engineering 12(2): 173–189. Hydrol. Process. (2014)