An improved method that incorporates the estimated runoff for peak discharge prediction on the Chinese Loess Plateau

An accurate prediction of peak discharge in watersheds is critical not only for water resource management, but also for understanding the complex relationships of hydrological processes. In this study, a modified peak discharge formula based on the Chemicals, Runoff, and Erosion from Agricultural Management Systems (CREAMS) model was developed by introducing rainfall intensity and soil moisture factors. The reliability of the proposed method was tested with data from 1464 storm events in 41 watersheds and was applied to 256 storm events in five remaining typical watersheds using the optimized parameters. The results indicate that the proposed method is highly accurate in terms of model efficiency, as determined by Nash–Sutcliffe efficiencies (NSEs) of 88.60%, 74.04%, and 90.12% during the calibration, validation, and application cases, respectively. Furthermore, it performed better than the original and modified CREAMS methods. Subsequently, using the parameters derived from the initial 41 watersheds and the runoff estimated using the modified Soil Conservation Service curve number (SCS–CN) method, the proposed method was used to predict the peak discharge from the last five typical watersheds. Large NSE (63.88–80.83%) and low root mean square error (RMSE) values (0.31–35.93 m3s-1) were obtained for the five watersheds. Overall, the proposed peak discharge model, combined with the modified SCS-CN method, may accurately predict event-based peak discharge and runoff for general applications under various hydrological and geomorphic conditions in the Loess Plateau region.