Comparison of the abilities of WEPP and the USLE-M to predict event soil loss on steep loessal slopes in China

WEPP was calibrated to minimize the mean squared error (MSE) between measured runoff and predicted runoff for the events where WEPP predicted runoff and soil loss to occur. The number of events where WEPP predicted runoff and soil loss to occur varied from 25% to 41% of events that actually produced soil loss. In the comparison with WEPP, the USLE-M erodibility factor was calibrated by dividing the total observed soil loss where WEPP predicted runoff and soil loss to occur by the total of the USLE-M erosivity factor for the same set of events. Consequently, the total amount of soil lost estimated for these events by both models were the same. This enabled the ability of the models to predict event soil loses to be examined using the Nash-Sutcliffe model efficiency index. In every case, the efficiency index values for the USLE-M were higher than for WEPP irrespective of whether the values of USLE-M erosivity index were determined using observed or WEPP predicted runoff. As a model of rainfall erosion, focusing on a process based approach has not resulted in WEPP performing as well as or better than the USLE-M on the steep bare fallow plots of loessal soil at the Ansai Research Station in China.