Abstract: To evaluate the accuracy of 3 typical precipitation products, including MSWEP V2.8,TRMM 3B42 V7,and China Meteorological Forcing Dataset(CMFD),with long time series and high temporal resolution in China′s mainland, we analyzed them from the perspectives of multi-scale, evolution function and spatial heterogeneities.Based on the hourly data of 2117 meteorological stations in recent 5 a, the 4 indexes of correlation coefficient(Corr),root mean square error(RMSE),median of relative error(REM),and probability of detection(POD)were selected to systematically evaluate their accuracy and spatial distribution characteristics at the hourly, daily, ten-day and monthly scales, and the evolution function models were established to describe the trends of these indices accompanied with the increase of temporal scale.The results reveals that the accuracies of the 3 precipitation products, generally, increase from the northwest to the southeast regions, and MSWEP exhibits a higher accuracy than CMFD and TRMM in POD and REM at all scales while they show inconsistency in Corr and RMSE at sub-daily and ten-day/monthly scales.As for the evolution function models, RMSE and REM can be described by a power function, and Corr and POD from hourly to monthly scale follow a modified logarithmic Gamma distribution function well.The research results can provide references for optimal identification, accuracy improvement, and deep fusion of precipitation production.