STOCHASTIC GENERATION OF DAILY GROUNDWATER LEVELS BY ARTIFICIAL NEURAL NETWORKS

Abstract

Many studies have been carried out in order to produce measures capable of ensuring the sustainable use of the water resources. A better management of water resources systems can be achieved if the long term information of hydrologic variables is available, which is not always the case. Stochastic simulation of such hydrologic variables is an attractive alternative to extend the length of observed records. This paper applies an Artificial Neural Network (ANN) model for simulating daily groundwater levels for the city of Matsuyama, Japan. The stochastic generated series must keep not only the statistical properties but also the seasonal oscillations of the observed series. Ten years of observed daily data were used for calibrating the model parameters. The results show that the model preserves the major statistical properties.