Intercomparison of GPM hourly precipitation products: Assessing the strengths in capturing precipitation events and their properties

Abstract

Spatially and temporally enhanced satellite precipitation products (SPPs) from the GPM mission offer a viable alternative to ground-based measurements. Although the SPPs have been extensively evaluated at specific time scales, such as hourly and daily, the temporally continuity feature of precipitation events is often overlooked. This study assessed and intercompared the performance of six GPM hourly SSPs, i.e. Early, Late and Final from IMERG, and NRT, MVK, NRT_GC (hereafter GC) from GSMaP, focusing on capturing three key precipitation properties (i.e. depth, duration and intensity) along with peak magnitude and timing, in order to obtain a comprehensive understanding of their capability and potential for precipitation monitoring and related hydrometeorological applications. The minimum inter-event time (MIT) approach, with a range of threshold values (i.e. 1, 2, 6, 10 and 24 h), was adopted to delineate precipitation events for the SSPs and the reference data (site observations and CLDAS datasets). IMERG and GSMaP products revealed their respective strengths in detection capacity of precipitation events, with IMERG achieving lower FAR and GSMaP showing higher POD. Despite comparable capabilities in estimating the depth of precipitation events by all the SPPs, IMERG products tended to be more effective in characterizing precipitation intensity and event peak while GSMaP better identified event duration. Notably, near/post real-time products outperformed gauge-corrected versions in certain aspects. Early was more capable of capturing event-based precipitation properties compared to the other IMERG products, especially for shorter MITs, suggesting the potential of near-real-time products in precipitation monitoring and early warning of associated hazards. In addition, MVK excelled in characterizing event duration, intensity, and peak magnitude within the GSMaP system, revealing the possible effectiveness of gauge-free satellite retrieval algorithms. Subregion analysis using the K-means clustering algorithm demonstrated the integrated influence of multiple factors on the performance of the SPPs, also with region-dependent sensitivity to a specific factor. The findings obtained from this study could provide new insights into the practical application and possible further enhancement of the GPM SPPs.