By using precipitation data collected every six hours by 2419 stations from April 1 to March 31 in 2013, precipitation forecast data of ECMWF and Japan high resolution model, and hourly rainfall data fusion by CMORPH (NOAA Climate Prediction Center Morphing Method) satellites and more than 30 thousand automatic stations based on contingency table and Taylor diagram statistical method, the precipitation forecasting capabilites of ECMWF and Japan high resolution model by every 6 h, 12 h, 24 h segment are comparatively analysed. The results show that: (1) Generally， ECMWF has an advantage over Japan high resolution model whose forecast dispersion is a bit larger. However, the forecast of ECMWF is relative steady, agreeing more with the observation. (2) The precipitation frequency predicted in weather forecasting by the two models is higher than the real case while the forecast of rainstorm frequency is lower than it. However as piecewise intervals increase, this situation is improved. (3) The 6 h fractional precipitation weather forecast score by ECMWF is lower than by the Japan 〖JP2〗model while the 12 h and 24 h fractional precipitation weather forecast scores of ECMWF are higher than those of Japan model, but its rainstorm forecast scores are all higher than those of the Japan model. (4) By changing forecast bias using threshold adjustment, the forecasting skill scores are improved to some extent. (5) Regarding the spatial distribution, the model can forecast precipitation more reasonably in coastal areas than in the northwestern area.〖JP〗