Abstarct: Bastam plain with an area of about 406 km2 is located 8 km northeast of Shahrood. Over exploitation of groundwater, mainly for agricultural uses, has significantly lowered water table. The main objective of this research is to simulate groundwater behavior using nonlinear ANN (artificial Neural Network) and ANFIS (Adaptive Neuro-Fuzzy Inference System). The ANN has been performed using a feedforward multi-laxyer perceptron network, applying transfer function of Sigmoid, Gauss, Tangent hyperbolic and Secant hyperbolic. The model was trained using a Back propagation (BP) method. Groundwater discharge, irrigation return flow and effective precipitation were the modeling input data. The results showed that both models can predict water table elevation with a high accuracy; For ANFIS modelingthe Sugeno system with trapezoidal shape membership function and minimum mass method with regression coefficient of 0.99 and 0.91, for training and verification levels respectively, the best answers were results.In modeling with ANN, the best result achieved when transfer function was selected as tangent hyperbolic with RMSE=0.36 and R2=0.99 for training stage and RMSE=1.06 and R2=0.83 in verification stage. The modeling results showed more accuracy of ANFIS compared to ANN. The ANFIS was selected for predicting elevation of water table at the subsequent two years, considering the following strategies, A) constant groundwater discharge and precipitation. B) constant precipitation and The production rate as much as shortage in aquifer storage. C) constant discharge and 30% decline in precipitation (draught period). On the basis of results, when the production rate is decreased equivalent to aquifer storage shortage, total drawdown in whole plain decrease from 1.8 to 0.7 m, and in the draught scenario, the drawdown increase from 1.8 to 2m.