Based on the hourly gauge precipitation data from the China Meteorological Administration, we used eight statistical metrics to evaluate the accuracy of the Final data from three IMERG (Integrated Multi-satellitE Retrievals for GPM) versions (i.e., Versions 3, 4 and 5) over Mainland China across multiple scales. We quantified the improvement of the latest Version 5 relative to previous versions and analyzed the problems in the current IMERG algorithm. Our result shows that: The IMERG data can well capture regional precipitation characteristics over Mainland China, but in northwest China where ground stations are sparse the error is larger and the accuracy is lower, making it difficult to estimate actual precipitation. The Versions 5 outperforms the Versions 3 and 4, with a higher correlation coefficient of 0.75 and a lower root mean squared error of 7.03 mm/d. Though partly corrected for the underestimate problem in northwest China, the Version 5 still performs poorly in winter and does not handle the overestimate problem. This latest version generally surfers from overestimate problems, and the ability to capturing and monitoring heavy rainfall events is less satisfying, and therefore cautions should be taken for the cases of heavy rainfall events. The correction algorithm is still imperfect, in particular for the historical data. Meanwhile, the algorithm may upraise satellite precipitation values in excess as a result of the correction for underestimate problems, leading to high false alarm rates and overestimate problems in the cases of heavy rainfall events. This has an impact on the quality of IMERG data that retrospect to TRMM (Tropical Rainfall Measuring Mission) times and also the follow-on data.%以中国气象局逐小时地面降水数据集为参考基准, 采用8种统计评价指标综合评估对比了美国NASA研发的全球降水计划 (GPM) 多卫星降水联合反演IMERG (Integrated Multi-satellitE Retrievals for GPM) 卫星降水产品的三个不同版本的Final数据, 分析了三套卫星降水在中国大陆地区多时空尺度下的反演精度, 探讨了IMERG最新版本V5数据的改进情况及反演中仍然存在的问题.结果表明:IMERG数据能够准确地捕捉到中国大陆地区的降水区域特征, 但是在中国西北部地面站点稀疏地区误差较大, 精度较低, 难以精确估测该地区的实际降水值.最新版本V5数据精度整体上优于先前的V3和V4数据, V5与地面观测数据的相关系数为0.75, 均方根误差为7.03 mm/d, 较V3、V4有明显提高, 改善了V3、V4在中国西北部出现的降水低估问题;但是V5在冬季表现较差且没有解决前期版本存在的高估问题, 整体上相对实际降水仍处于高估状态;同时V5在对高雨强事件的捕捉监测能力方面还存在一定的不足, 因此建议在强降雨事件监测中需谨慎使用卫星降水IMERG数据集.目前V5系统中的校正算法还存在部分缺陷:为消除全球降水系统性低估问题, 目前的校正算法整体性抬升了卫星降水值, 从而导致卫星降水反演在中国地区高雨强事件下出现高误报以及高估问题, 进而影响到IMERG数据回推以及后续再生数据的精度.
展开▼
