A novel scaling factor correction for GRACE terrestrial water storage anomalies in the Yellow River Basin

Yellow River Basin, China Global climate change and human activities have intensified drought conditions, particularly in the Yellow River Basin (YRB), making accurate monitoring of Terrestrial Water Storage Anomalies (TWSA) critical. However, GRACE-derived TWSA is significantly affected by leakage errors. This study proposes a novel scaling factor, developed using the Root Mean Square and Nash-Sutcliffe Efficiency Coefficient (RMS_NSE), to correct the leakage errors from 2002 to 2024 by weighting different hydrological models. This approach combines the strengths of individual hydrological models and addresses the limitations of Single-Model Scaling Factors (SMSFs), which often suffer from regional differences and uncertainties, resulting in systematic biases. The reliability and robustness of the RMS_NSE-based scaling factor are validated through multi-perspective assessments. The results indicate that RMS_NSE outperforms SMSFs by improving trends and amplitudes at the spatio-temporal scales, avoiding erroneous or over-recovery issues, and surpassing Forward Modeling (FM) in capturing the spatial distribution of annual amplitude. The RMS_NSE-corrected terrestrial water storage changes (TWSC RMS_NSE ) demonstrate the improved consistency with water balance-derived (WB) TWSC WB , achieving higher correlation (average 0.484) and lower RMS (average 13.514 mm), compared to SMSFs and FM. Simulation results show that RMS_NSE outperformed FM and SMSFs, demonstrating higher correction accuracy and spatial robustness, especially under extreme conditions and when hydrological model reliability was limited. • The study proposes a novel RMS_NSE-based scaling factor for improving leakage error. • RMS_NSE outperforms SMSFs by improving trends, amplitudes, and avoiding errors. • RMS_NSE outperforms FM in capturing the spatial distribution of annual amplitude. • Simulation results show reduced extreme-case uncertainty and improved robustness.