以地面加密雨量计数据为基准，全面分析和评估全球降水测量（global precipitation measurement, GPM）卫星反演产品IMERG-F（integrated multi-satellite retrievals for GPM-final run）对中国西南地区复杂山地条件下的日及亚日尺度降水，以及多种雨强降水的监测能力。结果表明：在日及亚日尺度内，IMERG对于短时间内累计降水的临界成功率（critical success index, CSI）主要分布在0.2 ~ 0.6，相关系数则在0.25 ~ 0.5范围内波动，其中以日尺度降水探测精度最高，夏季月份表现优于冬季。对于不同强度降雨事件，IMERG在小雨量的探测上具有较高的命中率（probability of detection, POD），而虚警率（false alarm rate, FAR）相对较低。然而，在中到大雨量级时存在低估现象。海拔落差对IMERG产品的稳定性影响显著，但与海拔高度本身没有直接线性关系。相对于地形起伏较大的地区，IMERG-F在地形起伏较小地区探测弱降雨事件中的可靠性更高。IMERG产品用于中国西南地区时应考虑其季节性和地形特征的限制。

In order to comprehensively analyze and evaluate the monitoring capabilities of the integrated multi-satellite retrievals for global precipitation measurement (GPM) final run (IMERG-F) satellite retrieval product for daily and sub-daily scale precipitation, as well as various intensity rainfalls, under complex mountainous conditions in southwest China, ground-based dense rain gauge data was used to investigate these aspects. The results show that within the daily and sub-daily scales, the critical success index (CSI) of IMERG for short-term accumulated precipitation primarily ranges from 0.2 to 0.6, with the correlation coefficient fluctuating between 0.25 and 0.5. The daily scale precipitation detection accuracy is the highest, with better performance in summer months compared to winter. For different intensity rainfall events, IMERG exhibits a high probability of detection (POD) for light rainfall, while the false alarm rate (FAR) is relatively low. However, there is an underestimation phenomenon for moderate to heavy rainfall. The elevation difference significantly impacts the stability of IMERG products, but there is no direct linear relationship with the elevation itself. Compared to areas with significant topographic variations, IMERG-F demonstrates higher reliability in detecting weak rainfall events in areas with less topographic variation. It is concluded that the application of IMERG products in southwest China should consider the limitations imposed by seasonal and topographic characteristics.