Multi-source precipitation fusion for hydrological models: Correction and metrics importance analysis

The Ganjiang River Basin, a major humid subtropical tributary of the Yangtze River located in southeastern China. This study develops a lightweight and interpretable precipitation merging framework based on Absolute Distance Inverse Weighting (ADIW) to integrate eight mainstream precipitation datasets (CHIRPS, CMORPH, ERA5, GSMaP, IMERG, PERSIANN, SM2RAIN, TRMM) from 2008 to 2020. The performance of the merged product, corrected with four methods (Linear Regression-LR, Linear Scaling, Quantile Mapping, Quantile–Quantile), was rigorously evaluated through hydrological simulation using the HYPE and VIC models. Results show that: (1) The ADIW merging framework effectively synthesizes multiple datasets, yielding a product with superior correlation and rainfall detection skill compared to any individual input, despite a slight underestimation. (2) While all four bias-correction methods effectively mitigated systematic errors, Linear Regression (LR) proved the most robust and consistent in enhancing both precipitation accuracy and subsequent runoff simulations. (3) Consequently, the combined ADIW+LR approach delivered the optimal hydrological performance, achieving the highest Nash–Sutcliffe and Kling–Gupta efficiency values in both models. (4) Diagnostic analysis identified ERA5 and GSMaP as the most influential datasets contributing to the merged product, and established that relative bias (RB) and mean absolute error (MAE) are the key metrics controlling hydrological reliability. • ADIW method enables lightweight merging of multiple precipitation data sources. • Four bias-correction methods significantly improve hydrological modeling performance. • Dynamic ADIW with moving window enhances merged precipitation for runoff simulation. • RB and MAE identified as most critical precipitation metrics for hydrological modeling. • Merged-LR outperforms benchmark MSWEP in both HYPE and VIC hydrological models.