Modeling of groundwater resources is considered as a tool in water resource management. The interaction of groundwater and surface water resources is important in water balance. Artificial Neural Networks are powerful tool for modeling of hydrological and groundwater parameters with considering all the parameters affecting on them. In this study, using correlation analysis methods in the time and frequency domain under the influence of aquifer parameters affecting groundwater level divided and changes in groundwater level with time delay with respect to any of the parameters measured and with the average level of groundwater in the region and neighboring regions artificial neural networks are considered as inputs. Output models are the groundwater level in all region pizometers separately .ANN models using two methods, the Levenberg–Marquart and Bayesian regularization were trained and was evaluated using the criterion of mean squared error (MSE). After the end of the modeling process, In order to investigate spatial variation of groundwater level in this study interpolation method by using Cross Validation method were compared. Results indicate the ability of feedforward artificial neural networks with Levenberg-Marquardt training algorithm with average parameter error (MSE) equal 0.87 square meters in time modeling and ordinary Kriging method with average parameter error (RMSE) equal 11.62 meters in estimating spatial variations in groundwater level for study area. Keyworld: Groundwater, Modeling Temporal-Spatial, Artificial Neural Networks, Correlation Analysis, Frequency Analysis, Estimates Spatial, Geostatistics, Kriging.