Abstract:In semi-arid and semi-humid regions, due to the spatial topography and rainfall heterogeneity, the runoff generation mechanisms are mixed and variable. As a result, the simulation accuracy of Grid-Xin"anjiang model (Grid-XAJ) is low. In order to improve the forecasting accuracy of Grid-XAJ in semi-arid and semi-humid areas, based on Grid-XAJ, the spatio-temporal dynamic combination of the saturation-excess and infiltration-excess runoff was considered to further research in this paper. The dominant runoff generation mechanism of the basin grid in the process of runoff generation was dynamically identified by determining the relationship between rainfall and soil moisture. Constructing the Grid-Xin’anjiang and saturation-infiltration double excess distributed model(Grid-XAJ-SIDE). The semi-arid Suide watershed was selected for comparing the simulation results of the Grid-XAJ-SIDE, the Grid-XAJ and the Grid-GA models. The results show that the qualified rate of runoff simulation of the three models in the Suide watershed are all above 80.0%. Compared with the Grid-XAJ (53.0%) and Grid-GA (40.0%) models, the Grid-XAJ-SIDE (66.7%) model has a higher flood peak simulation qualified rate. The Grid-XAJ-SIDE model can capture the control effects of rainfall and soil water content on runoff generation process, and achieve a clear description of runoff generation processes in the basin.