In order to make the HEC-HMS model universally applied in different hydrometeorological regions, the SCS-CN method in the HEC-HMS’s director runoff module was improved according to the small and medium-sized watersheds of Xianbeigou, Chengcun and Pinghe. The improvement measures include replacing the previous 5 days’ rainfall with the antecedent impaction precipitation P a to evaluate the previous soil wetness, adopting the linear interpolation to increase the classification of CN( Cx ) and selecting the appropriate initial loss rate λ according to the different meteorological and hydrometeorological region. The limitation of the inverse algorithm in the calibration of SCS model parameters is discussed. By calibrating the optimal Cx under the pre-affected rainfall in different watersheds and regressing the Cx-P a relationship, we explored the law of the change of actual wetness with the pre-affected rainfall. The results show that when the optimal initial loss rate is 0.05 in the humid region and 0.20 in the sub-humid region, a better simulation accuracy can be achieved. The improved method can effectively improve the application effect of HEC-HMS, and provide a reference for the rainfall and runoff simulation in the region with insufficient data.