Forecasting the underground water level with artificial neural networks model (Case study: Shabestar plain)

Groundwater has always been considered as one of the major sources of drinking and agricultural water supply, especially in arid and semi-arid areas. Ground water systems simulations because of their inherent complexity are not readily possible. The goal of this research is prediction of groundwater level fluctuation in Shabestar Plain, West of East Azerbaijan province, using artificial neural networks (ANN). The 9 years data sets (1380-1388) from 15 piezometers that uniformly disperse on the whole plain is applied for ANN model training. The four inputs ANN model were monthly data set from temperature, rainfall, Daryanchay discharge, groundwater level in each piezometer with one month delay (t0-1). The groundwater level was the only output of the ANN model. The results of this research showed that the ANN model with TRAINLM training function and with TANSIG activation function are capable to monthly prediction of groundwater level in 3 years period with high accuracy R2=99.63, RMSE=1.43 in training step and R2=99.16, RMSE= 1.167 in validation step in study area.