Abstract: Tropical monthly mean precipitation estimated by the latest GPCP version 2 datasets and PR products in large time and space scales were compared and the major causes for the differences between them were investigated. After quantitatively analyzing the relations between the difference and the rain intensity, the results were employed to estimate the error of GPCP climatology precipitation from 1979 to 1997. Results shown that both of the two datasets present similar distribution patterns of long-time mean tropical rainfall but with differences in amplitude and range. In general, the average difference over ocean (0.45 mm/d) was larger than that of over land (0.08 mm/d), and it was the overestimation of rainfall by SSM/I E over ocean and the absence of rain gauge station in some area over land that lead the differences. It was found that the magnitude of difference is highly related to the rain intensity. The precipitation of PR with maximum PDF (Probability Distribution Function) is linear with the given rain intensity of GPCP over oceans, but the relation is nonlinear over land. The departure between them is obviously larger in ocean than that over land when the rain intensity is weaker than 5 mm/d. Almost all the differences over ocean are positive, but negative difference can be found over land when the rain intensity is weaker than 6 mm/d. Using the regression function of the maximum PDF, an error field of the GPCP climatological precipitation was established before 1998, The error estimated by this study has the same distribution pattern with those estimated by GPCP itself , but the former carries some new information. It determined not only the absolute value of the error, just as GPCP did, but also the sign of the error. Additionally, it showed that GPCP has underestimated the error come introduced by passive microwave precipitation retrieval over ocean.