Abstract: Using ECMWF ensemble forecasts, a heavy rainfall event in the Yangtze River delta was analyzed in this paper. The"good"and "bad"clusters of ECMWF ensemble members were also compared. The results show that the ensemble products, such as ensemble mean and quantile products have better forecast skills for the heavy rainfall event compared to the deterministic forecasts. The precipitation forecast of high quantile products is a good indication to heavy rainfall, especially for a long forecast length. The continuous forecasts of heavy rainfall by high quantile products, and the increasing trend of precipitation forecasts by ensemble mean and lower quantile products indicate that heavy rainfall is very likely to happen. The forecast differences of weather systems and meteorological elements, including the trough in 500 hPa, vortex and wind shear in 850 hPa, southwesterly and easterly winds, between the"good"and"bad"clusters of ensemble members have key impacts on the forecast skills.