Author(s): Thannob Aribarg; Serre Supharatid
Keywords: Climate change; Discrete wavelet transform; Nonlinear autoregressive exogenous neural network; RCPs; Reservoir inflow
Abstract: Accurate and reliable long-term forecasting of reservoir inflow is necessary for efficient water resources’ planning and management. In this study, a hybrid model using discrete wavelet transform (DWT) and the nonlinear autoregressive exogenous (NARX) neural network is developed for the simulation of the monthly inflow into Bhumibol and Sirikit reservoirs in Thailand under present and future climate scenarios. For this purpose, we have compiled an ensemble of nineteen downscaled climate data from NASA earth exchange global daily downscaled projections (NEX-GDDP). Two climate scenario projections (RCP 4.5 and RCP 8.5) are used to evaluate the climate change impacts for the future period up to 2099. Results indicate that climate change has a clear impact on both reservoirs inflow and show an increase in annual inflow into both reservoirs except in dry seasons. In the wet season (May-October), the inflow of Bhumibol and Sirikit reservoirs will increase by 6.61% and 17.41%, respectively, in the far future period (2079 - 2099) under RCP 8.5. Findings from this study imply how to adapt for the optimize water resource management in the future.