Abstract: The comparison between the precipitation forecast of the ECMWF model and the observed precipitation at Xiangyang Station in Hubei province during the flood season from 2016 to 2020 is conducted,and the results show that the forecast of moderate rain and above in the first and second days are smaller,while the forecast in the third day is larger.The deviation of the heavy rainfall center in the three forecast periods is irregular.In order to better apply ECMWF products and improve the accuracy of precipitation forecast in flood season,the revised values of different precipitation magnitude are studied from the perspective of probability matching.And the precipitation forecast of ECMWF in flood season in 2021 was tested daily.The results show that the probability matching correction method can effectively improve the rainfall prediction performance of the model,and a good correction effect on the moderate rain and above has been verified,especially on the rainstorm forecast of the first day.The average TS score of 228 stations increases by 6% from 11.1% to 17.1%,and the missing rate decreases by 13.5% from 85.0% to 71.5%.The probability matching correction method is favorable of quantitative precipitation forecast,due to the addition of background information on the probability distribution of local rainfall.