Ensemble‐Based Runoff Reconstruction for Enhanced Hydrological Process Representation in the Jialing River Basin, China

ABSTRACT Accurate runoff reconstruction is crucial for managing water resources in basins with complex terrain and nonlinear hydrological responses. Traditional single‐model approaches often face systematic biases across different flow regimes. In contrast, ensemble‐based fusion provides a more robust solution. This study proposes a multi‐model framework for the Jialing River Basin, China. It integrates three distinct methods: Relationship fitting, an improved Manning's formula and the C/M signal method. These methods are combined using a Stacking ensemble strategy, with linear regression as the meta‐learner. The framework was validated using monthly observations from 2015 to 2020. The Stacking ensemble consistently outperformed individual models, achieving an NSE of 0.981 and a MAPE of 2.307%. Error diagnostics revealed minimal systematic bias (PBIAS = −0.283%), with only a slight underestimation of peak flows. Importantly, the framework improved the capture of runoff variability by reducing model‐specific errors and uncertainty propagation. These results suggest that ensemble fusion offers a more stable and accurate tool for hydrological modelling in complex river systems.