Quantifying the retrieval uncertainties associated with systematic and random errors in multi‐satellite‐only precipitation estimates over the Chinese mainland

Abstract

AbstractCharacterizing the uncertainties in retrieval, which include both systematic and random errors, of satellite‐derived precipitation products (SPPs) is crucial for the development of bias correction techniques aimed at enhancing the accuracy of precipitation retrieval algorithms. This study aimed to comprehensively and quantitatively analyse the retrieval uncertainties in various SPPs across different seasons, climate regions, precipitation rates, and elevations. Three satellite‐only SPPs from the Tropical Rainfall Measuring Mission (TRMM) and the Global Precipitation Measurement (GPM) mission, namely the TRMM Multi‐satellite Precipitation Analysis (TMPA‐rt) and the Integrated Multi‐satellite Retrievals for GPM (IMERG‐early and IMERG‐late), were tested at the daily scale from April 2014 to March 2018, benchmarked by ~2400 weather stations in China. In comparison to TMPA‐rt, both IMERG‐early and IMERG‐late exhibited reduced systematic errors across the majority of western China, particularly over the Tibetan Plateau during the summer. The percentages of systematic errors increased gradually with the rise in rainfall intensities and became predominant in all three SPPs beyond rain rates of 50 mm·day−1. Moreover, the systematic error ratios of IMERG products generally exhibited an upward trend with increasing elevation, whereas this feature is not significant in TMPA‐rt. In contrast to TMPA‐rt, the two IMERG‐based SPPs demonstrated much lower magnitudes of random errors in all climate regions over the Chinese mainland, although they did not exhibit a significant reduction in systematic errors. However, these two IMERG products still suffer from certain uncertainties in the mountainous regions of western China during winter, primarily stemming from deficiencies in the IMERG retrieval system in the snow/ice‐cover areas. In summary, it appears timely for this study to intercompare and quantify the systematic and random errors in mainstream SPPs from TRMM to GPM, providing insights for future algorithm upgrades and data generation.