以位于不同水文气象分区的屯溪流域和绥德流域为研究对象，选取 TIGGE(THORPEX?Interactive?Garnd Global?Ensemble) 数 据 集 中 NCEP（ National?Centers?for?Environmental?Prediction） 、 ECMWF（ European?Centre?for Medium-range?Weather?Forecasts）、CMA（China?Meteorological?Administration）3 种预报产品的 2010—2015 年控制 预 报 数 据 ， 基 于 分 位 数 映 射 法 中的 QUANT（ non-parametric ?quantile ?mapping ?using ?empirical ?quantiles） 法 和 RQUANT（non-parametric?quantile?mapping?using?robust?empirical?quantiles）法进行预报降雨修正，并采用多分类预报 检验、连续型预报检验和概率型预报检验等方法，对不同水文气象分区、不同预报产品和不同修正方法进行比较 与适用性分析；同时，以屯溪流域实测降雨为例，通过增加噪声项对降雨重采样，基于新安江模型分析降雨不确定 性对水文模拟结果的影响。结果表明：在研究流域，所选的预报产品对无雨和小雨期的预报精度都较高，但随着 降雨量的增加，各产品的预报能力均出现较为明显的下降。多分类和连续型检验表明绥德流域的降雨预报效果 更佳，NCEP 和 ECMWF 在研究流域的整体预报精度较高，CMA 的整体预报精度在研究流域略低于其他产品。各 产品在修正后大部分检验指标预报精度提高，其中：ECMWF 在绥德流域修正后预报精度最高，对两种修正方法都 有很好的适用性；在屯溪流域，NCEP 和 ECMWF 在不同修正方法后各指标预报精度各有高低，CMA 在修正后仅 在大雨量级的 TS 评分预报精度高于其他产品。降雨的不确定性会对水文模拟产生消极影响，并导致参数的不确 定性和水文模拟精度的下降。

Rainfall data is one of the main factors that can affect the accuracy of flood simulation in small and medium rivers, and rainfall forecasting is the most direct means of extending the forecast period. At present, there are many precipitation forecast product evaluations and post-processing methods, but the research on the applicability evaluation, accuracy improvement, and forecast of each product is still insufficient in different hydrometeorological subregions. Through the detailed evaluation after applying different correction methods, the numerical forecast products and suitable rainfall correction methods are sought for small and medium basins in different hydrometeorological regions. This research aimed to provide a reliable basis for improving the flood forecasting accuracy and forecast period of medium and small rivers.This research took Tunxi and Suide basins and selected the 2010 to 2015 control forecast data of three products (NCEP, ECMWF, and CMA). The QUANT method and the RQUANT method were used for forecasting rainfall correction, and the multi-classification forecast test, continuous forecast test, and probabilistic forecast test methods were used to compare and analyze the applicability of different forecast products and different correction methods. Taking the measured rainfall in the Tunxi basin as an example, the rainfall was resampled by adding a noise term, and the influence of rainfall uncertainty on the hydrological simulation results was analyzed based on the Xin'anjiang model.The results showed that the selected forecast products had high forecast accuracy for no-rain and light-rain periods. With the increase in rainfall, the forecast ability of each product decreased significantly. The test showed that the rainfall forecast effect in the Suide basin was better, and the overall forecasting accuracy of NCEP and ECMWF was well in the study basin. The forecast accuracy of CMA is slightly worse than that of other products. The prediction accuracy of most of the inspection indicators for each product was improved after the correction. Among them, ECMWF had the highest prediction accuracy in the Suide basin and had good applicability to both correction methods. The prediction accuracy of NCEP and ECMWF after different correct methods was different in the Tunxi basin. CMA's corrected TS score forecast accuracy was superior to other products only in the heavy rains weight level. The uncertainty of rainfall had a negative impact on the hydrological simulation, which led to the uncertainty of the parameters and a decrease in the accuracy of the hydrological simulation.The research provided a reference for selecting suitable numerical forecast products and correction methods for different hydrometeorological regions. Semi-arid areas like Suide basin should use ECMWF revised forecast products, and in humid watersheds like Tunxi basin, NCEP and ECMWF products and correction methods should be selected according to the needs of application indicators, which can reduce rainfall forecast errors in certain procedures and improve the accuracy of flood simulations.