Abstract: By using frequency matching method (FMM), an experiment is performed to correct the precipitation bias from Shenzhen Storm-Scale Ensemble Forecast system (SZ-SSEF). The principle of this method is to use ensemble member based on precipitation frequency to calibrate the simple ensemble mean frequency to correct smoothing effect caused by ensemble averaging. It is applied to experiments of extreme rainfall events in South China induced by southwest monsoon on June 19 and the strong Typhoon Hato on 23 August 2017. The results show that the method can effectively eliminate system biases in precipitation amount and area coverage. Large area of erroneous or spurious rain from model forecasts can be effectively removed, and ETS and BIAS scores for precipitation forecast are significantly improved. The results of MODE (Method for Object-based Diagnostic Evaluation) test indicated that the corrected precipitation is closer to the observations in the precipitation average area ratio, centroid distance separation and the intensity value ratio. This paper aims to provide references for the application and improvement of the SZ-SSEF.