HYBRID MODELING FOR FUTURE INFLOW PREDICTION OF HUAI LUANG RESERVOIR UNDER CLIMATE CHANGE

Abstract

This study presents a hybrid modeling approach integrating the hydrological model HEC-HMS with machine learning techniques to predict the future inflow of the Huai Luang Reservoir under climate change scenarios. Rainfall projections from three CMIP6 global climate models (CanESM, CESM2, and GFDL-ESM4) under SSP245 and SSP585 scenarios for the period 2023–2044 were used as key inputs. Historical inflow data from 2001 to 2022 were employed for model training and validation. The calibration phase (2011–2015) achieved a coefficient of determination (R²) of 0.62 and an RMSE of 0.70, while validation (2016–2020) resulted in an R² of 0.56 and an RMSE of 0.70, demonstrating moderate predictive performance. The hybrid modeling approach reveals a declining trend in annual inflow, with projections ranging from 54.08 million cubic meters (GFDL-ESM4 under SSP245) to 172.71 million cubic meters (CanESM under SSP245), while the highest average inflow projection reaches 120.48 million cubic meters (CanESM under SSP585). These findings highlight the potential hydrological impacts of climate change and underscore the necessity of adaptive reservoir management strategies to ensure sustainable water resource availability in the Huai Luang watershed.