Optimizing storage-based reservoir operation schemes for enhanced large-scale hydrological modeling: A comprehensive sensitivity analysis

Abstract

Accurate reservoir operation modeling is essential for hydrological simulations and climate impact assessments. This study evaluated three storage-based reservoir operation models across 289 global reservoirs using observed inflow, outflow, and storage data. Under default parameterizations, model performance varied, with median Nash-Sutcliffe Efficiency (NSE) values for outflow ranging from 0.30 to 0.38. A model incorporating target storage constraints consistently outperformed the others across most reservoir types and sizes. A Sobol-based sensitivity analysis, using the bounded NSE index (C2M) on outflows, identified distinct parameter influences. Two models exhibited balanced sensitivity to multiple parameters, whereas the third was primarily influenced by a single parameter related to peak flow management. Parameter optimization significantly improved outflow simulations, with median C2M increases of 29 %, 26 %, and 25 % across the three models. These improvements were particularly pronounced in regions with poor default performance, such as the eastern USA. Statistical analyses underscored the importance of calibration, as optimized models achieved C2M values above 0.6 for over 60 % of reservoirs. This study systematically evaluated the performance of storage-based reservoir operation models, parameter sensitivities, and the potential for improving reservoir representation in hydrological models. Enhanced reservoir simulations support the calibration and refinement of large-scale river models by identifying key parameters for reservoir modules.