A coupled optimized hedging rule-based reservoir operation and hydrodynamic model framework for riverine flood risk management

Abstract

Long-term changes in reservoir inflow due to climate change and human interferences violate the assumptions of hydrologic stationarity, especially in the reservoir operation during high flood season for managing the downstream critical levee (DCL) sections from overtopping. Utilization of uncertain inflow forecast into a reservoir using the operating rule curve of certain forecast horizon reflects the challenges imposed by nonstationary conditions, downstream flood intensification with spatiotemporally distributed lateral flux and floodplain dynamics. Addressing these issues, this study develops four hierarchical frameworks considering single-stage hedging (1SH) and two-stage hedging (2SH) rules-based reservoir operation models optimized with Particle Swarm Optimization (PSO) and informed with rating curve uncertainty at DCL section. Further, these two frameworks are coupled with HEC-RAS-2D (H2D) hydrodynamic model to reduce the existing flood risk at DCL section. The efficiency of the advocated 1SH-PSO, 2SH-PSO, 1SH-PSOH2D and 2SH-PSOH2D are tested in the Rengali reservoir on the Brahmani River in eastern India. The inflow forecasts into the reservoir are simulated by the coupled SWAT-Pothole and Wavelet-based Bidirectional Long-Short-Term Memory (WBiLSTM) models forced with the bias-corrected GFS weather forecasts with up to 10 days’ lead-times. The results demonstrate that the best-performing 2SH-PSOH2D framework-based reservoir operation could reduce the average peak flow depth at the DCL station by 21 % from the baseline with an average reduction in levee failure risk by 22.28 % leading to effective management of high flood events. This advocated framework could be used in other reservoir systems worldwide in reducing the downstream flood hazards through enhanced reservoir operation.