Uncertainty-aware reservoir operation projections using multi-model weighting and adaptive hedging rules

Abstract

While reservoirs designed based on historical records are increasingly vulnerable to inflow variability driven by climate change, conventional rules such as the standard operating policy (SOP) often amplify projection uncertainties by failing to anticipate future shortages. To tackle these challenges, this study adopts an integrative, uncertainty-aware reservoir operation framework that jointly addresses upstream inflow projection uncertainty and downstream operational uncertainty, representing a research gap not yet explored. Specifically, a dynamic weighting scheme, Uncertainty Optimizing Multi-Model Ensemble (UO-MME), was used for credible inflow projections, which were then incorporated into a two-dimensional hedging model (HDG-2d) to optimize reservoir releases. The methodology was applied to the Tahtali Reservoir (western Türkiye) considering a modeling chain comprising five GCMs, two emission scenarios, two downscaling methods, and seven hydrological models. The climate inputs needed by hydrological models and UO-MME include bias-corrected CORDEX outputs and statistically downscaled CMIP5 data covering the projection span 2021–2099. The inflow projections from different variants were used to operate both SOP and HDG-2d models. Compared with SOP operated under unweighted data, where projection uncertainties tend to accumulate, HDG-2d emerged as the main uncertainty-reducing mechanism for reservoir outputs, while UO-MME played a supportive role. Overall, the combined strategy achieves ensemble consistency in most variants, with dimensionless vulnerability remaining below 0.25, reduces its related uncertainty by 23% through mitigating single-period shortages, and further narrows projected release and sustainability index uncertainties by 43% and 36%, respectively. Combining hydrological model weighting with climate-informed hedging is shown to provide a more robust basis for reservoir planning.