Comparative hydro-climatic forecasting of reservoir storage and cross-scale bathymetric evaluation in Mingde and Shihmen reservoirs.

Accurate reservoir storage forecasting is critical for water security and risk management, yet most data-driven approaches emphasize short-term predictive skill without accounting for long-term changes in storage capacity caused by sedimentation. This study addresses this gap by examining whether monthly, hydro-climatic machine learning forecasts of effective water storage capacity can be meaningfully evaluated against observed, multi-year capacity changes derived from bathymetric surveys. Forecasting performance is evaluated at the rainfall-driven Mingde Reservoir and the highly regulated Shihmen Reservoir in Taiwan using linear regularized regression and a nonlinear tree-based model forced by precipitation, temperature, humidity, and evaporation, deliberately focusing on hydro-climatic forcing due to the temporal unavailability or static nature of monthly sediment transport, land-use, and operational release records. Model outputs are assessed using conventional performance metrics together with independent bathymetric benchmarking and interpretability analysis. Results show that model suitability depends strongly on reservoir behavior and operational regulation. At Mingde Reservoir, the linear model provides stable and accurate forecasts (Nash-Sutcliffe efficiency of 0.968), whereas the nonlinear model performs better at the highly regulated Shihmen Reservoir (Nash-Sutcliffe efficiency of 0.952). Interpretability analysis reveals that forecasts are dominated by antecedent storage, leading to systematic bias during extreme drought conditions. Although the models accurately reproduce storage variability when static capacity curves are assumed, bathymetric benchmarking demonstrates a failure to capture multi-year capacity loss caused by sedimentation, with relative errors reaching up to 292%. These findings identify a structural limitation of hydro-climatic forecasting frameworks and highlight the need to explicitly incorporate sediment-related processes to support long-term reservoir management.