Event-Based and Error-Component Evaluation of GPM IMERG and GSMaP_Gauge for Monitoring Hourly Extreme Precipitation over North China

Hourly satellite precipitation products are often evaluated using aggregate statistics that can conceal compensating errors in precipitation occurrence and event duration, accumulation, and intensity. Using all matched observations from 522 meteorological stations in North China between 1 October 2019 and 26 September 2022, this study evaluated GPM IMERG Final Run V07 and GSMaP_Gauge V8. The framework combined precipitation-amount and occurrence metrics, event characteristics defined using local 90th percentiles of wet-hour precipitation, and additive and multiplicative error decomposition. IMERG showed a persistent dry bias, with a mean annualized total error of −348.2 mm yr−1. GSMaP had a higher probability of detection than IMERG (77.6% versus 38.6%) but also a higher false alarm ratio (67.5% versus 29.7%); its mean total error of +99.7 mm yr−1 masked substantial compensation among the error components. For extreme precipitation events, IMERG underestimated annualized event frequency by 33.9% but reproduced mean duration comparatively closely (−11.0%). GSMaP overestimated event frequency by 31.6% and duration by 198.6%. Its near-unbiased mean event total (+1.1%) resulted from prolonged duration despite underestimation of mean and peak intensity by 67.3% and 67.4%, respectively. Within the approximately three-year evaluation period examined here, these results show that product selection should be application-specific and that event-scale correction should address precipitation occurrence, duration, accumulation, and intensity jointly.