Abstract: Based on the bias characteristics of the ECMWF (European Centre for Medium-range Weather Forecasts) ensemble statistic fusing prediction products in the Haihe River basin, the frequency matching method was used to correct the precipitation bias. The performance before and after the correction was examined. Using the results from four months（May-August, 2016）experiment, we demonstrate t that the positive biases in the precipitation level and range from the original products can  be significantly improved using the improved fusion products. The average intensity of the modified precipitation forecasts is closer to observations. The longer the valid forecast time, the greater the precipitation level, and the bigger the prediction bias , the better the improvement effect. To some extent, the scores of TS (threat score) and Bias of the improved fusion products are improved, especially for the light rain, rainstorm and the heavy rain cases. It eliminates the large false areas and reduces  the false-alarm  rates  significantly.  However, the missing- rates are slightly increased.