Determining Multivariate Short-Term Forecasts of Groundwater Levels andReservoir Inflows by Artificial Neural Networks

摘要

A feedforward Artificial Neural Network (ANN) model is applied for themultivariate forecast of daily groundwater levels and reservoir inflows up to five daysahead. The ANN forecasting model relates previous groundwater level, previousreservoir inflow and daily precipitation forecasts of five days ahead in order toestimate daily groundwater levels and reservoir inflows for these five days. The short-term precipitation values are assumed to be deterministic since meteorological short-range forecasts are generally available. The methodology is applied for the watersupply system of Matsuyama City, in Japan. Scarcity of water is a periodical problemin this city and thus accurate forecasts of groundwater levels and reservoir inflows arevery important to improve the water resources management in the region. The goodaccuracy obtained by the ANN forecasting model indicates that it is very reliable forshort-term daily estimations. As a result, this multivariate model may generateconsistent data for the application of optimization techniques to the sustainablemanagement of Matsuyama City's water supply system.