Climate driven water balance stress in a reservoir using ANN corrected CORDEX-SEA projections

Abstract

Reservoir water balance assessment is essential for understanding how climate variability and climate change affect water availability and operational performance in monsoon regions. This study evaluates baseline (2006–2023) and projected (2024–2050) water balance stress in Wonogiri Reservoir, Indonesia, using bias-corrected Coordinated Regional Climate Downscaling Experiment for Southeast Asia (CORDEX-SEA) projections under representative concentration pathways (RCP4.5 and RCP8.5). The main contribution is an integrated reservoir-scale framework linking artificial neural network (ANN)-corrected precipitation and potential evapotranspiration with hydrological inflow simulation and differentiated forecasting of turbine and hollow jet valve (HJV) outflows to assess intra-annual storage stress under an operational constraint. An ANN was applied for bias correction and coupled with the F.J. Mock model for inflow simulation, while singular spectrum analysis (SSA) was used for outflow forecasting. Water balance was evaluated under probabilistic hydrologic conditions with a minimum intake elevation of 128 m. Baseline results indicate a monsoon-driven surplus–deficit regime with high inflow variability (0.04–221.93 m³∙s⁻¹) and strong dry-season vulnerability due to elevated evapotranspiration. The SSA shows higher predictability for turbine outflows (NSE = 0.76) than for HJV outflows (NSE = 0.58). Although dependable inflow increases in both scenarios, these gains do not translate into higher effective storage. Instead, additional inflow is largely discharged during wet periods, while dry-season storage declines because of unfavourable inflow timing and increased evaporation. Thus, climate change may intensify seasonal operational stress, requiring adaptation strategies focused on seasonal timing, evaporation losses, and release pathways rather than annual water availability alone.