Propagation of precipitation errors to streamflow: A large-sample evaluation of IMERG V06 and V07 in southern China

Study region 312 small- and medium-sized watersheds across humid southern China. Study focus Using the Xinanjiang (XAJ) model together with an effective automatic calibration algorithm, this study systematically evaluates the streamflow simulation performance of the Integrated MultisatellitE Retrievals for Global Precipitation Measurement (IMERG) Version 06 (V06) and Version 07 (V07) Final Run products across 312 small- and medium-sized watersheds (SMWs). Hydrological impacts of two types of IMERG precipitation error characteristics are quantified through parallel experiments and sensitivity analyses. New hydrological insights for the region Simulation skill varied markedly among products and regions. The calibrated IMERG V07 product achieved the highest median Nash–Sutcliffe efficiency (NSE) during calibration, but its validation performance deteriorated because of the anomalous overestimation associated with the Global Precipitation Climatology Centre (GPCC) after 2019. After removal of this temporal anomaly, the calibrated IMERG V07 and V06 products achieved median NSE values of 0.61 and 0.58, respectively, over the full period. Recalibrating the XAJ model accommodated the modest underestimation of the uncalibrated IMERG V07 product in eastern watersheds, whereas severe underestimation in western high-elevation watersheds still resulted in substantial streamflow-volume underestimation and degraded simulation skill. This study provides baseline streamflow simulation performance for IMERG precipitation in humid southern China. More broadly, it demonstrates that precipitation-error magnitude and temporal anomalies can constrain streamflow simulation under model recalibration.