Comparison of multi-source merged precipitation products using independent gauge observations

Authors: ZHANG, H - Tianjin University; WEI, L - Nanjing University; ZHU, J - Wuhan University; ZHOU, J - Tianjin University; LIU, S - Nanjing University Of Information Science And Technology (NUIST); SUN, X - Nanjing University Of Information Science And Technology (NUIST); KANG, X - Tianjin University; ZHOU, H - Tianjin University; GAO, M - Tianjin University; DUAN, Z - Lund University; Crow, Wade; DONG, J - Tianjin University

Submitted to: Atmospheric Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/9/2025
Publication Date: 8/10/2025
Citation: Zhang, H., Wei, L., Zhu, J., Zhou, J., Liu, S., Sun, X., Kang, X., Zhou, H., Gao, M., Duan, Z., Crow, W.T., Dong, J. 2025. Comparison of multi-source merged precipitation products using independent gauge observations. Atmospheric Research. 328. Article e108427. https://doi.org/10.1016/j.atmosres.2025.108427.
DOI: https://doi.org/10.1016/j.atmosres.2025.108427

Interpretive Summary: Accurate tracking of global rainfall is critical for large-scale agricultural drought monitoring. The current state-of-the-art for such tracking is the fusion of rainfall information obtained from a variety of sources (i.e., ground-based observations, models, and remote sensing). Such fusion relies on robust estimates of uncertainty in rainfall information acquired from each source - so that higher-quality sources of information can be prioritized over lower-quality sources. This paper evaluates a particularly promising statistical technique (i.e., statistical uncertainty analysis or SUA) for acquiring such estimates of uncertainty and illustrates, for the first time, that the application of SUA improves our ability to globally monitor rainfall - particularly in areas of the world that lack adequate ground-based information. The results of this analysis will help to refine the development of large-scale precipitation estimates that, in turn, will improve our ability to globally monitor the extent, severity, and impact of agricultural drought.

Technical Abstract: Data merging is frequently used to enhance large-scale precipitation estimates. However, traditional merging algorithms heavily rely on gauge data and are therefore subject to higher uncertainties over data-sparse regions. Statistical uncertainty analysis (SUA) algorithms can estimate optimal merging weights without reliance on gauge data and have recently been used to develop new precipitation merging frameworks. However, the assumptions underlying SUA may be violated in real-world applications, which could result in degraded precipitation estimates. The relative performances of SUA-merged versus traditional gauge-merged precipitation datasets is largely unknown due to the lack of truly independent validation datasets. Here, based on the analysis of 268 wholly independent rain gauges, we find that SUA-based merging can effectively suppress random precipitation errors and outperform state-of-the-art remote sensing and reanalysis precipitation datasets. Interestingly, relative to traditional gauge-merged datasets, SUA-merged precipitation is better correlated with independent gauge observations and possess a lower root-mean-square-error. These results directly confirm the robustness of SUA for reducing reliance on gauge observations in precipitation merging, which is particularly useful in data sparse regions. However, SUA-merged precipitation still contains higher uncertainties with respect to rain/no-rain classification. Therefore, algorithms for addressing false precipitation events should be prioritized in future SUA-based precipitation merging frameworks.