Abstract:Flood forecasting for mountain rivers realizes accurate simulation and early warning of the rainfall- runoff process in complex terrain by constructing high- precision hydrological models and optimizing key parameters. This paper expounds the methods of element integration and spatial resolution configuration in constructing hydrological models, introduces the technical points of calibrating parameters such as soil curve number, hydraulic conductivity, and channel resistance coefficient based on sensitivity analysis, multi- objective evolutionary algorithms, and hybrid optimization strategies. Through typical cases, it verifies the effects of the optimized model, such as increased flood peak prediction accuracy (NSE increased from 0.65 to 0.84), significantly reduced RMSE, and extended forecast lead time. The results show that automatic parameter optimization can greatly improve the reliability of mountain flood forecasting, providing scientific support for flood control scheduling and risk management.