Based on observed meteorological data and measured soil loss data from field runoff plots in 1985-1992 at Ansai Comprehensive Experimental Station, the applicability of WEPP model to different slope lengths on hil-lgully region of the Loess Plateau was assessed. Results showed that simulated values of event runoff, annual runoff, and average annual runoff by WEPP model were not desirable for the slope lengths of 10 m, 20 m, 30 m and 40 m. How ever, the simulated values of event soil loss and annual soil loss by WEPP model were desirable for the slope leng ths of 10m, 20m, 30 m and 40 m, respect ively. The differences of simulated values between slope lengths and increasing trend with increase of slope length were close to observed data of the soil loss. The Nash-Sutcliffe model ef ficiencies of event runoff simulat ion and annual runoff simulation were 0.915 and 0.879, respectively; and the Nash-Sutcliffe model efficiencies of event soil loss simulat ion and annual soil loss simulat ion were 0.853 and 0.758, respect ively. The simulated accuracy of annual average runoff and soil loss by WEPP model ws satisfied. I30 had a great effect on runoff simulated values at individual storms; the error between simulated values and observed data probably increased when I30 was hig her than 0.92mm/ min. The soil loss simulated values at individual storm had a close correlat ion with PI30, and the error betw een simulated values and observed values increased when PI30 was greater than 129mm2/ min.PI30was an important factor affect ing simulated accuracy of WEPP model, and simulated accuracy decreased when the PI30 was higher than 150mm 2/ min;meanw hile, the effect of PI30 on runoff simulation was more greater than that on soil loss.