Based on the new generation numerical prediction model GRAPES (Global/ Regional Assimilation and Prediction Enhanced System), with three kinds of physi cs option parameterization schemes(Kessler schemes、NCEP3－class simple ice sc hemes and simple ice schemes (developed by Liu Qinjun, CAMS) ) and two kinds of cu mulus parameterization schemes (KainFritsch (new Eta) and BettsMillerJanji c), six combinational schemes are formed. 48hour predicting experiments are made for the precipitation cases in the rainy season in Guizhou. Comparisons among the experiments and analyses of the observations indicate that GRAPES model can be used for predicting Guizhou precipitation. The predicting effect of cumulus para meterization schemes is better than that of physics option parameterization sche mes. The predicting Guizhou precipitation ability of combinational schemes with simple ice scheme physics option parameterization, BettsMillerJanjic cumulus parameterization is superior to the others. These results are useful for improvin g the GRAPES model for predicting Guizhou precipitation.