中国不同气候区多源遥感降水融合与性能综合评估

遥感降水产品种类繁多，单一遥感降水产品难以满足所有气候区的降水估算。多源遥感降水融合方法可弥补这一缺陷，进而提高降水估算精度。本研究选择四种全球典型遥感降水产品（CHIRPS v2.0、CMORPH v1.0、PERSIANN-CDR、TRMM 3B42V7），利用集成模型输出统计（EMOS）进行数据融合，并与均值（MME）对比分析，阐明遥感降水产品和EMOS，在不同气候区的基本统计性能（BIAS、RMSE、r和KGE）、等级性能（POD）和水文性能（NSE）。结果发现：① 四种遥感降水产品均能捕获多年降水的空间分布格局，但EMOS改进了遥感降水产品整体低估现象。相对于MME，EMOS显著地（P<0.05）减小了4个气候区的BIAS，减小全国降水BIAS和RMSE分别为36.9%和10.2%。全国尺度而言，EMOS提高r和KGE分别为18.2%和71.4%。② PERSIANN-CDR中寒旱区微量降雨的POD高于其他三种产品，但是对大雨的POD低于其他产品。EMOS对微量降雨POD的提高出现在干旱区和青藏高原，对中雨POD则在过渡区和湿润区，而对小雨则适用于全国各气候区。③ TRMM 3B42V7的径流模拟能力优于其他三种遥感降水产品，表现最差的为CHIRPS v2.0产品。相对MME而言，除过渡区外，EMOS均提高了其他气候区的水文模拟性能，且在青藏高原、干旱区和湿润区的NSE分别提高26.3%、8.5%和2.2%。研究结果不仅为遥感降水产品在不同气候区数据源的选择提供参考依据，而且为多源遥感降水数据的融合和应用开拓了新思路。

Improving and comprehensive evaluation precipitation estimation by merging multi-satellite precipitation products in different climate regions of China

Various remotely sensed precipitation products have been published, but it is difficult for a single remote sensing precipitation product to capture the temporal and spatial patterns of precipitation in different climate regions. In this study, four typical global remote sensing precipitation products (CHIRPS v2.0, CMORPH v1.0, PERSIANN-CDR, TRMM 3B42V7) were selected, and the ensemble model output statistics (EMOS) was used for data fusion. Compared with the mean (MME), performance of remote sensing precipitation products and EMOS were evaluated based on the basic statistical (BIAS, RMSE, r and KGE), categorical (POD) and hydrological (NSE) skills in different climate regions. Results show that: (1) All the four remote sensing precipitation products can capture the spatial distribution pattern of multi-annual precipitation, but EMOS improves the overall underestimation of remote sensing precipitation products. Compared with MME, EMOS significantly (P<0.05) reduced the BIAS of the four climate regions, and reduced the national precipitation BIAS and RMSE by 36.9% and 10.2%, respectively. On a national scale, EMOS increased r and KGE by 18.2% and 71.4%, respectively. (2) The POD of PERSIANN-CDR for tiny rain in cold and arid regions is higher than that of the other three remote sensing products, but the POD for heavy rain is lower than that of other remote sensing products. The improvement of POD for tiny rain by EMOS is mainly observed in arid regions and the Tibetan Plateau, for moderate rain in transitional and humid regions. However, for light rain, it is applicable to all the climatic regions. (3) The runoff simulation capability of TRMM 3B42V7 is better than that of the other three products, and the CHIRPS v2.0 product has the worst performance. Compared with MME, EMOS improved the hydrological simulation performance of the other climate regions except the transitional regions, and improved NSE by 26.3%, 8.5% and 2.2% in the Tibetan Plateau, arid zone and humid zone, respectively. This paper not only provides a reference for the selection of data sources for remote sensing precipitation products in different climatic regions, but also opens up new ideas for the fusion and application of multi-source remote sensing precipitation data.