Investigating adaptive hedging policies for reservoir operation under climate change impacts

Abstract

Concerns about whether the reservoirs against climate change will be able to fulfill their missions in the future have revealed the necessity of adapting their operations to changing circumstances. On the other hand, the fact that ensembles of hydrological climate-impact projections contain uncertainties originating from the general circulation models (GCMs) and the representative concentration pathway (RCP) scenarios is also an unavoidable challenge. This study, carried out to address these two issues together, corroborates a two-dimensional hedging model as an adaptive measure for climate change impacts and undertakes an uncertainty investigation over the Gordes reservoir in western Turkey. First, the statistically downscaled and bias-corrected meteorological data from twelve GCMs under two RCPs were transformed into inflow projections employing a hydrological modeling procedure for the period 2021-2050. Then, a parameterization-simulation-optimization framework imposed on adapting to all scenario-GCM variations was developed by integrating the hedging rule with the differential evolution algorithm. The results demonstrated that adaptive hedging policies mitigated potentially derived vulnerabilities from standard operating policies, ranging from 68% to 97%, and kept the sustainability index above 0.75 against climate change-induced water deficits. Finally, the decomposition of uncertainty contribu-tions in reservoir operation optimization along the GCM-RCP-rule curve chain was made by the analysis of variances. It is obvious from the findings that the GCM was the foremost source of uncertainty in both predicted releases and key performance indices, while the uncertainty from the RCP-GCM interaction is of secondary importance. But a noteworthy detection is that the contributions of GCMs to the total variance for vulnerability and sustainability indices could be reduced by about 20% with a refined ensemble consisting of the outputs from four GCMs that better represent the local predictors. The interpretation of the uncertainty analysis approach, which was previously performed in the literature for runoff quantiles, at the reservoir scale as well as the reservoir operation optimization model, which has been verified to be viable, might be advantageous for water resources management.