Abstract: Verification and integration of numerical weather prediction (NWP) products is an important step in the application and development of numerical prediction. Based on the NCEP, ECMWF, JMA and KMA NWP models from TIGGE center, the basin of Chitan Reservoir in Fujian province was selected as the research object, the multimodal precipitation forecast products were evaluated from the aspects of precipitation classification forecast, precipitation level and process forecast comprehensively. Meanwhile, three methods of regression integration, TS integration and Nash coefficient integration were used to integrate multimodal precipitation forecast products, and the influence of different integration methods on the final precipitation forecast effect was discussed. The results show that the four products had good prediction effect for both no rain and light rain events. In the precipitation forecast of different magnitudes, the JMA model was the best for the precipitation forecast of the magnitude less than 25 mm, but the prediction of ECMWF for precipitation over 25 mm was better. For the daily precipitation forecast, the NCEP model had a poor performance, while the ECMWF model was the most accurate. For the precipitation process forecast, the prediction performance of KMA was obviously worse than the other three models. The integrated forecast had a good effect on reducing the prediction error and improving the precipitation process forecast, among which the TS integrated method weighted by different magnitudes had the best forecast effect, and had better improved the prediction effect of high level precipitation events.