By using SAL quantitative verification method, a heavy rainfall prediction in Meiyu period, 2017 of EC-THIN,RIOF,NCEP and CMA models were verified from amplitude,location and structure aspects. The causes of precipitation forecast errors, and the stability of 72 h forecast of the models were analyzed. The results show that:(1)RIOF, EC-THIN and CMA forecast the trend of rainband is basically consistent with the observation. NCEP forecast the main rainband range is larger and the rainstorm area is east. (2)RIOF and EC-THIN are better in the structure prediction of rainy area, NCEP and CMA are better in amplitude precipitation, and the location prediction errors are all small, and the error of CMA is minimal. (3)EC-THIN and NCEP show better stability in structure amplitude and location prediction. CMA shows better stability in intensity prediction and a large adjustment in location prediction. RIOF shows better stability in structure prediction and a large variation in area prediction. (4)The primary cause of precipitation forecast error is the system forecast error. The intensity, position and movement of the system directly affect the precipitation. The forecast of vertical physicals also affect the precipitation period, enhancement and intensity.