Abstract: Numerical weather prediction errors come from the initial conditions and model errors. Ensemble forecasting technique is an effective way to diminish the errors. Short-range ensemble forecasting experiments are made for three precipitation cases during the 1999 Meiyu period in the East China area. The MM5 model is used as the experimental model configuration. Eight ensemble members are created by choosing four kinds of cumulus parameterization schemes and two kinds of PBL parameterization schemes. The four kinds of cumulus parameterization schemes are Anthes-Kuo, Grell, Kain-Fritsch and Betts-Miller schemes. The two kinds of PBL parameterization schemes are MRF and Eta schemes.The resul ts indicate that different ensemble members have dif ferent forecasting result s . For the precipi tation fo recast ing results , the inf luence of cumulus parameterization scheme is larger than the influence of the PBL parameterization scheme .For the bias sco re , most ensemble members have a “wet” bias .The bias score is larger for large precipitation than that for small precipitation .The ef fects of ensemble averaging increase the bias score for small precipitation and reduce the bias sco re for large precipitation .Fo r different cases, the member w ho has the best precipitation forecasting results is not the same one .Af ter ensemble averaging , stable precipi tation fo recasting results can be got ten .Also the objective and quantitative precipitation probability fo recasts can be obtained f rom the ensemble forecasting .