To learn more about the performance of GRAPES_Meso 3km (Global/Regional Assimilation and Prediction System) model for precipitation prediction in the first flood season of South China and to provide reference for improvement and operational application, the hourly precipitation data of 86 stations in Guangdong Province were selected. According to the distance to the coastline, the stations were divided into three sub-regions: the eastern coastal region, the western coastal region and the inland region. By using the Threat score (TS), the Equitable Threat Score(ETS), the forecast bias score and other indicators, the quantitative evaluation was carried out from 18 to 6 may 2020 The forecast effect of precipitation in the first flood season of South China on June 18. On the whole, the TS and ETS scores of light rain, moderate rain, heavy rain and rainstorm in the eastern coastal area and inland area are higher than those in the western coastal area, but the rainstorm forecasting skills of the three sub regions are lower; from the 3H cumulative precipitation score, the forecasting scores of the three sub regions have obvious diurnal variation characteristics, but the performance of the eastern coastal area and the western coastal area is different from that of the inland area In the eastern and western coastal areas, the night forecast skill is relatively low (the bias is high), and the daytime forecast skill is relatively high (the bias is low), while the daily variation in inland areas is relatively high (the bias is low), and the daytime forecast skill is relatively low (the bias is high); the reason for the relatively low forecast skill in the western coastal areas is due to the verification period There is a weak wind shear in Inner Guangdong, and most of the western coastal areas are just controlled by the high temperature area on the south side of the shear line. However, the daily average temperature in this area simulated by the model is lower than the actual situation, and the relative humidity is higher, which leads to more air forecasts of precipitation in the western coastal areas, and reduces the skills of precipitation forecast.