Abstract: Based on the surface observed precipitation data in Sichuan Province, aiming at a sustained heavy rainfall process in Sichuan in August 2020, four merged precipitation products (FAST_5KM, FRT_5KM, RT_1KM, NRT_1KM) developed by National Meteorological Information Center were compared and assessed by using various assessment indexes. The results show that, for the accumulative precipitation of the process, the four merged products can reflect the precipitation process well, the 1KM products are more detailed to describe the rainfall area, and have a high consistency with the actual, the deviation of NRT_1KM is the smallest and FAST_5KM is the biggest. The precipitation cells of the four merged products are slightly different from the actual, and the precipitation cells of 1KM products are closer to the actual. The trend of hourly precipitation of the four merged products is consistent with the actual, but the peak value is lower than the actual hourly precipitation peak value, and the hourly precipitation of 1KM products is closer to the actual. For 24h accumulative precipitation, the four merged precipitation products can represent the daily movement of rainfall area, and the rainfall area, trend and shape of rain band are more consistent with the actual, while the rainfall density and area of 1KM products are closer to the actual. For the accuracy of clear-rainy forecast, the rainless stations of the four merged products are less than the actual, the 1KM products are the closest to the actual, and the accuracy is basically maintained at 95%. The hourly maximum rainfall density of merged precipitation products has a good corresponding relationship with the actual, the 1KM products are better than 5KM products, FRT_5KM is better than FAST_5KM, the NRT_1KM is the best. In general, the 1KM products are better than 5KM products, 1KM near real-time products are better than real-time products, and 5KM three-source products are better than two-source products.