Prediction the Groundwater Level of Bastam Plain (Iran) by Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS)

Abstract

Prediction of the groundwater level (GWL) fluctuations is very important in the water resource management. This study investigates the potential of two intelligence models namely, Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) in the forecasting of the groundwater level of Bastam Plain in Iran. For this purpose, 9 years data-sets including hydrological and hydrogeological parameters like rainfall recharge, irrigation returned flow and also pumping rates from water wells were used as input data to predict groundwater level. The results showed that ANN and ANFIS models can predict GWL accurately. Also, it was found that the ANFIS model (with root-mean-square-error (RMSE) 0.02 m and determination coefficient (R2) of 0.96) performed better than the ANN model with RMSE = 1.06 m and R2 = 0.83. Finally, three scenarios were considered to predict the groundwater level in the next 2 years as follows 1- The rainfall recharge and pumping rate of water wells will be constant, 2- The rainfall recharge will be constant, but the pumping rate of water wells will be reduced equal to the water deficit of the aquifer, 3- The pumping rate of water wells will be constant but the rainfall recharge will be reduced 30 %. The prediction with these scenarios showed that the groundwater level has the minimum reduction when the pumping rate of water wells is equal to the water deficit of the aquifer.