Abstract: Climate models are useful tools in climate variability and climate change studies. However, the current state-of-the-art climate models generally show large biases in monsoon rainfall simulation. The sources of the model bias may result from either the atmospheric circulations or the physical parameterization schemes. The reanalysis datasets were produced by using the most advanced operational numerical models. Due to the assimilation of observational data, the atmospheric circulation in the reanalysis dataset is nearly“real”and thus the precipitation in the reanalysis data may be regarded as the output predicted by a“perfect”Atmospheric General Circulation Model (AGCM). In this“perfect” model, since the atmospheric circulation is predicted as the real world, any biases in the precipitation prediction should result from the model physics. In this study, the authors have compared the global monsoon precipitation derived from the NCEP1 reanalysis data (NCEP1 for short) against the observations derived from the GPCP data. The observational spatial patterns of climatology monsoon modes are reasonably reproduced in NCEP1, with a pattern correlation coefficient (PCC) higher than 0.8 and a root mean square error (RMSE) less than 2 mm/d. NCEP1 underestimates the accumulation of heavy and little rainfall, while it overestimates the accumulation of middle rainfall. Over the domains of eight sub-monsoon systems, the amounts of total summer precipitation are underestimated by NCEP1 in comparison to the GPCP data. Only the precipitation amount over the northwestern Pacific and South African monsoon regions is overestimated. The long-term trend and interannual variability of monsoon precipitation index (MPI) derived from NCEP1 are similar to those from the GPCP data, the skill in the Northern Hemisphere is better than that in the Southern Hemisphere. The authors also examine the variability of global monsoon rainfall by EOF analysis. The first EOF mode of Annual Range (AR) from NCEP1 is the same as that from the GPCP data, the corresponding principle component (PC) series all exhibit a significant decreasing trend. Examination on the statistical significance of AR trend at each grid point within the global monsoon domains based on MK (Mann-Kendall rank statistics) and T2N (trend-to-noise ratios) methods indicates that for the NCEP1, it agrees with the observations over most of global monsoon domains, but over the North African monsoon region it shows a decaying (increasing) trend in the north (south), which is contrary to the GPCP data.