Abstract: Based on the physical factors output by SWC-WARMS and hourly precipitation of automatic stations in Sichuan province from June to August 2020, the random forest model was used to forecast hourly precipitation for the rainfall process, and TS score was conducted to understand the local availability of this method. The results show that: (1) Partition of the province is conducive to the improvement of hourly heavy precipitation. (2) The overall trend of hourly heavy precipitation forecast obtained by the random forest model is consistent with that of the original mode, which has a certain response to the hourly heavy rainfall not predicted by the model, and the precipitation magnitude of the regions without actual heavy rainfall has decreased after correction, but there are a lot of uneven distribution of precipitation and many false predictions.