Integrating error decomposition and feature-based diagnosis for multi-source precipitation estimates in the Yangtze River Basin

Study region Yangtze River Basin, China. Study focus This study presents an integrated diagnostic framework that combines four-component error decomposition with feature-based analyses of error amount, frequency, and intensity. The component-feature framework is applied to seven representative multi-source precipitation estimates (MPEs) spanning four major product categories. Using a gauge-based reference, spatial and temporal variations in the error components are systematically examined across the plateau, complex-terrain, and plain subregions of the basin. New hydrological insights for the region Among the satellite-based products, CMORPH and GSMaP exhibit broadly consistent error characteristics, featuring the smallest total biases and the strongest detection skill within the component-feature framework. IMERG exhibits a nearly balanced distribution of positive and negative hit biases (39% vs. 37%), which helps offset systematic errors. In contrast, PERSIANN shows the weakest detection skill and the largest intensity biases among the satellite products, consistent with the limitations of infrared-based retrievals. Regional differences are observed for specific products and error components, mainly reflected by spatial variations in false-alarm frequency for MSWEP and ERA5, with pronounced contrasts between high- and low-elevation regions. The monthly variations of these error characteristics also exhibit regional contrasts. Additionally, MSWEP exhibits intermediate error characteristics while maintaining the most stable monthly proportions among the four error components. SM2RAIN is characterized by persistently high false-alarm frequencies (>30%) together with pronounced hit-negative bias intensity across the basin, indicating a distinctive error structure among the evaluated MPEs. Overall, this component-feature framework provides a structured basis for understanding precipitation uncertainty and may support more informed use of MPEs.