Abstract: Water shortages and nonpoint pollution caused by soil erosion have reached crises point in many regions. Hydrological model, soil erosion model and nonpoint pollution model have been used as the main technique tools which evaluate the risk of water resources, soil erosion and water pollution. Runoff calculation is an important part in the hydrological models and is the basic element in the calculation of water contaminants including sediment, nutrients and pesticides etc. Soil Conservation Service Curve Number (SCS-CN) method has been widely used in many countries and regions. Runoff curve number (CN) is an integrated parameter in the SCS-CN method, which reflects the effect of soil water content, soil type and land use etc. on runoff depth. But SCS-CN method may result in bad prediction accuracy in China because the values of curve number provided by USDA Soil Conservation Service can not be directly used in China due to the differences in landuse，hydrologic soil groups and slope gradient etc. Thus the purpose of this study was to compare the differences in determining CN value by using mean method, arithmetic-mean method, median method, S probability distribution method and asymptotic method and to compare the accuracy of runoff predicted by different CN value. Data of rainfall - runoff from three runoff plots at Shixia, Miyun, Beijing, with a period from 1993 to 2006, were used. Runoff depth was calculated by using SCS-CN method with different CN determining methods. The index of Nash model efficiency, correlation coefficient and acceptance rate were used to evaluate the accuracy of runoff depth predicted by different CN determining methods. The results show that the runoff depth predicted by five methods was significantly different from the measured ones. The asymptotic method had the best Nash model efficiency. But arithmetic-mean method provided with the best correlation coefficient and acceptance rate. Compared to asymptotic method, arithmetic-mean method is simple, thus we recommend that arithmetic-mean method be used to determine the CN value. The antecedent moisture condition had important effects on the results of runoff prediction. The acceptance rate under the AMCII was the best whereas that under the AMCIII was the worst. The results also indicate that it is necessary for further research on the initial abstraction ratio in Beijing area.

Key words: SCS CN method, SCS curve number, runoff