2024年7月22-24日，甘肃省遭遇历史罕见特大暴雨，共计12个站（点）累积降水量超300 mm，最大达 351. 4 mm，综合强度为 1961 年以来西北地区最强。本文基于地面自动观测站（Automatic WeatherStation，AWS）降水实况观测数据，评估了中国区域融合降水分析系统（CMA Multi-source PrecipitationAnalysis，CMPA）、雷达估测降水（Radar Quantitative Precipitation Estimation，Radar-QPE）、风云 4B卫星估测降水（Fengyun 4B Quantitative Precipitation Estimation，FY4B-QPE）和欧洲中期天气预报中心的全球陆面再分析资料（European Centre for Medium Range Weather Forecasts Reanalysis v5，ERA5）四种降水产品在此次特大暴雨期间的监测能力。结果表明：（1）在空间分布上CMPA表现最佳，能够准确捕捉暴雨的核心区降水和极值，空间变异性最小，小时降水量平均误差（Mean Error，ME）仅为 0. 002 mm·h-1。Radar-QPE 能够识别暴雨区位置，但低估了核心区降水量，FY4B-QPE 对核心区降水有明显高估，而ERA5 则低估了核心区降水量，ME 分别为-0. 151、0. 192 和 0. 08 mm·h-1。（2）CMPA 在时间演变的捕捉上最为准确，误差最小，相关系数（Correlation Coefficient，CORR）高达0. 999。Radar-QPE在强降水时低估降水量，误差随降水强度增加显著增大，FY4B-QPE和 ERA5的误差在强降水期间显著增加，尤其是FY4B-QPE 在核心区的表现较差，CORR 分别为 0. 96、0. 24 和 0. 22。（3）CMPA 与 AWS 的日变化特征最为接近。Radar-QPE在降水峰值和分布上存在偏差。FY4B-QPE峰值位置偏东、偏北，且较降水时间提前。ERA5没有显著的经向峰值，表现为纬向偏北的负偏差。（4）CMPA与AWS在降水概率分布上高度一致，表现出最佳的时空一致性。Radar-QPE和 ERA5高估了首个降水峰值，而低估了 5. 0 mm·h-1以上区间的小时降水量。FY4B-QPE对弱降水低估、强降水高估。这些结果为不同降水产品在暴雨降水事件中的监测能力提供了详细的对比，为暴雨动态监测、预警和水文应用研究等方面提供参考。

From July 22nd to 24th，2024，Gansu Province was hit by an extremely rare torrential rain in history. A total of 12 stations accumulated rainfall exceeding 300 mm，with the maximum reaching 351. 4 mm. The over‐ all intensity was the strongest in the northwest region since 1961. Based on the Precipitation observation data from Automatic Weather Station（AWS），this study evaluated the CMA Multi-source Precipitation Analysis sys‐ tem in China（CMPA），Radar Quantitative Precipitation Estimation（Radar-QPE），Fengyun 4B Quantitative Precipitation Estimation（FY4B-QPE）and the European Centre for Medium Range Weather Forecasts Reanaly‐ sis v5（ERA5）monitoring capabilities of four precipitation products during this extremely heavy rainstorm. The results showed that：（1）CMPA had the best performance in spatial distribution，which could accurately capture the precipitation and extreme value in the core area of rainstorm，with the least spatial variability，ME was only 0. 002 mm·h-1. Radar-QPE could identify the location of the rainstorm area，but underestimate the precipitation in the core area，FY4B-QPE significantly overestimates the precipitation in the core area，while ERA5 underesti‐ mates the precipitation in the core area，ME was respectively -0. 151，0. 192 and 0. 08 mm·h-1.（2）CMPA was the most accurate in capturing time evolution with the smallest error，CORR was up to 0. 999. Radar-QPE under‐ estimated precipitation during heavy precipitation hours，and the error increased significantly with the increase of precipitation intensity，the errors of FY4B-QPE and ERA5 increased significantly during heavy precipitation hours，especially FY4B-QPE had worse behavior in the core area，CORR was respectively 0. 96，0. 24 and 0. 22.（3）The diurnal variation characteristics of CMPA and AWS were the closest. There were deviations in the peak value and distribution of precipitation in Radar-QPE. The peak position of FY4B-QPE was located to the east and north，and the precipitation time was advanced. There was no significant peak of ERA5 in the meridian direction，but showed a negative deviation in the zonal direction that was slightly northward.（4）CMPA and AWS were highly consistent in precipitation probability distribution，showing the best spatio-temporal consisten‐ cy. Radar-QPE and ERA5 overestimated the first precipitation peak and underestimated the precipitation above 5. 0 mm/h. FY4B-QPE underestimated weak precipitation and overestimates heavy precipitation. These results provided a detailed comparison of the monitoring capabilities of different precipitation products in rainstorm pre‐ cipitation events，and offered a reference for the dynamic monitoring，early warning and hydrological applica‐ tion of rainstorm event.