Enhancing Hydrological Model Calibration for Flood Prediction in Dam-Regulated Basins with Satellite-Derived Reservoir Dynamics

The construction and human operation of reservoirs have made terrestrial hydrological processes increasingly complex, posing challenges to large-scale flood modeling. While many hydrological models have incorporated reservoir operation schemes to improve discharge estimation, the influence of reservoir representation on model parameterization has not been sufficiently evaluated—an issue that fundamentally affects the spatial reliability of distributed modelling. Additionally, the limited availability of reservoir regulation data impedes dam-inclusive flood simulation. To overcome these limitations, this study proposes a synergistic modeling framework for data-scarce dammed basins. It integrates a fully satellite-based reservoir operation scheme into a distributed hydrological model and incorporates reservoir processes into model parameter calibration. The framework’s feasibility was tested using the DRIVE flood model (coupled version named DRIVE-Dam) through a case study in the Nandu River Basin, southern China. Two calibration strategies, with and without dam operations (CWD vs. CWOD), were compared. Results show that reservoir dynamics were effectively reconstructed by combining satellite altimetry with FABDEM topography, successfully supporting the development of the reservoir scheme. Multi-station comparisons across the basin indicate that, while CWD slightly improved streamflow estimation (NSE and KGE > 0.75, similar to CWOD), it enhanced cross-basin peak discharge and flood event duration capture with reduced bias, boosting flood detection probability from 0.54 to 0.60 and reducing false alarms from 0.28 to 0.15. The improvements stem from refined parameterization enabled by a physically complete model structure. In contrast, CWOD leads to subdued flood impulses and prolonged recession due to spurious parameters distorting baseflow and runoff responses, highlighting that neglecting reservoir processes can result in unrealistic parameter estimations and compromised model reliability across space. The proposed methodology provides a technical reference for flood forecasting in dammed watersheds. The findings reveal the enhancement effect of reservoir representation on conventional model parameterization and emphasize the great potential of satellite observations for improving hydrological modeling in data-limited regions.