Runoff Prediction under Different Precipitation Scenarios Based on SWAT Model and Stochastic Simulation of Precipitation

To predict runoff under different annual precipitation and reflect the impact of annual precipitation and its inner-annual distribution on the runoff process, the soil and water assessment tool (SWAT) model is combined with the stochastic simulation of precipitation based on the torrential rain analysis. Taking the watershed above Wangkuai Reservoir in China as an example, the SWAT model is constructed. and the stochastic simulation models of precipitation under three annual precipitation states are established. Then, based on the torrential rain analysis, five precipitation scenarios with annual precipitation of 300, 600, and 900 mm are assumed, and the daily precipitation process of each scenario is generated as the input of the SWAT model. The results are as follows: the SWAT model has a very good performance for monthly runoff simulation; the maximum monthly runoffs of the five precipitation scenarios are 5.99, 7.09, 9.14, 17.48, and 23.71 m3/s, respectively, and the annual runoffs are 2.24, 3.02, 3.30, 4.75, and 5.08 m3/s, respectively. When the annual precipitation is about 600 mm, the influence of precipitation inner-annual distribution on the monthly runoff process is mainly reflected in July to October. When the annual precipitation is about 900 mm, the influence is mainly reflected in August. This study provides a new idea for runoff prediction and provides reference for the study of the rainfall–runoff uncertainty relationship. Moreover, the improved precipitation stochastic simulation and its combination with the SWAT model can be applied to the study of other basins for further discovery.