LAUSR

Abstract The Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (IMERG) produces high-resolution precipitation estimates and serves as an invaluable data source for data-scarce and ungauged basins. This study comprehensively evaluates two near-real time products (IMERG-E and IMERG-L) and a post-real time dataset (IMERG-F) of IMERG V06B from 2014 to 2019 on basin and grid scales in the Yellow River source region, a data-sparse basin in the northeastern Tibetan Plateau. Results show that the three IMERG products can well depict the spatial distribution of precipitation in the study area. Compared to IMERG-E and IMERG-L, the gauge-adjusted IMERG-F shows better agreement and event detection on both basin and grid scale, yet is limited in improving data accuracy. The near-real time products, IMERG-E and IMERG-L, underestimate precipitation across the basin (~−37.82% and ~−38.55%, respectively). At the same time, post-real time product exhibits negative errors in drier areas and positive errors in wetter regions that are roughly separated by the basin's 3900 m contour. The correlation, error metrics, and detecting ability of the three products upgrade from northwest to southeast as the climate becomes wet. Temporal properties exhibit dependence on meteorological conditions. Specifically, the performance in the wet season is superior to that in the dry season. For instance, the normalized root mean square error for IMERG-F was reduced by 55.10% from dry season to wet season. IMERG products' capability to detect precipitation is affected by temporal scale, intensity, and phase. The reliability of these satellite precipitation products increases when the time interval extends from daily to monthly. IMERG can better estimate moderate precipitation, especially in the range of 5–10 mm/d. The products generally perform worse for solid precipitation than for liquid one. However, interactions of precipitation phases and intensities lead to a decrease in accuracy for light precipitation (