This Is AuburnElectronic Theses and Dissertations

Application and Evaluation of WEPP in a Forested Watershed with Perennial Streams




Srivastava, Anurag

Type of Degree



Civil Engineering


Forests generally have low erosion rates unless they are disturbed by harvesting operations or road construction. These common disturbances typically cause increase in the sediment yield which is reduced after vegetation and plant litter covers the surface. To better quantify and predict sediment loads adequate watershed scale erosion simulation tools are needed. The Water Erosion Prediction Project model (WEPP version 2008.907) is calibrated using 13 storm events in 2004 for runoff and sediment yield in an undisturbed forested watershed (4.41 km2) in East Texas having perennial streams. Although the WEPP model is not designed for application in watersheds having perennial streams, an attempt was made to validate the calibrated watershed model using observed post-harvest runoff and sediment yield from 19 storm events from 2005 to 2007. For calibration and validation, WHAT baseflow separator program is used to compare estimated runoff (subsurface and surface runoff) with corresponding simulated model runoff. Sediment rating curves developed from observed data were used to estimate sediment yield from runoff for comparison with model simulated sediment yield. WEPP performance in terms of calibration and validation of runoff and sediment yield prediction is evaluated with the correlation coefficient (R), coefficient of determination (R2), Nash-Sutcliffe efficiency (NSE), root mean square error (RMSE)-observation standard deviation ratio, and percent bias (PBIAS). A high value of correlation coefficient (R=0.98) is found between WEPP simulated runoff and WHAT baseflow separated runoff for the pre-harvest calibration. A high correlation coefficient (R = 0.98) indicates a positive relationship between WEPP simulated sediment and estimated sediment from runoff for pre-harvest conditions. NSE, RSR, and PBIAS statistical evaluation of model calibration is considered “very good” for runoff (NSE = 0.95, RSR = 0.22, and PBIAS = 11.67%) and ranges from “satisfactory” to “very good” for sediment yield (NSE = 0.90, RSR = 0.32, and PBIAS = 7%). In subsequent validation of post-harvest conditions, a road crossing, culvert, and forest roads were added to the pre-harvest model using field information available at the time of harvest. The resulting R between simulated runoff and WHAT baseflow separated runoff is 0.91. However, NSE, RSR, and PBIAS evaluation of WEPP performance is considered “unsatisfactory” for runoff (NSE = 0.46, RSR = 0.74, and PBIAS = 48.72%). Sediment validation results such as correlation coefficient (R = 0.91) indicated a positive linear relationship between simulated and estimated sediment from post-harvest runoff. Although the resulting R value for the linear regression is 0.91, NSE, RSR values are considered “unsatisfactory” (NSE = 0.11 and RSR = 0.94). The model failure to validate runoff and sediment for post-harvest conditions is due to major limitations of WEPP model not being intended for use in watersheds with perennial streams. Attempts to accurately simulate culvert, road crossing, and forest road surfaces in this study were not sufficient to overcome model limitations. In order to apply the WEPP model confidently in watersheds having perennial streams, model capabilities should be improved to include erosion and sedimentation occurring within the channel during baseflow conditions.