We made intercomparison about characteristics of daily precipitation among GPCP1DD, TRMM3B42 (version5) and Radar-AMeDAS in order to enhance our climate model ability. Radar-AMeDAS is produced by Japan Meteorological Agency, whose horizontal resolution is up to 5km (2.5km is currently used on operating analysis). This radar-based data is corrected by high-density gauge network named AMeDAS (Automated Meteorological Data Acquisition System). This covers only Japan domain, but it is very useful to evaluate characteristics of GPCP and TRMM. We compared wet-day frequency, mean daily precipitation and probability of strong daily precipitation, which is derived as scale parameter of gamma distribution function fitted to daily precipitation intensity distribution. Evaluation was made for each season (DJF, MAM, JJA, SON). Spatial pattern of mean daily precipitation is similar between GPCP1DD and TRMM, but TRMM has slightly larger on land area. As for wet-day frequency, TRMM is higher than GPCP1DD especially on Asian-Monsoon area and Intertropical Convergence Zone(ITCZ). By comparing GPCP1DD and Radar-AMeDAS, we found that GPCP1DD underestimates wet-day frequency due to poor ability for catching weak precipitation. Frequency of strong precipitation in TRMM is generally lower than GPCP1DD, and GPCP1DD seems to better agree with Radar-AMeDAS. These results suggest that model evaluation should take into account of uncertainty of observational data.