Spatiotemporal Predicting of Groundwater Level Using Artificial Intelligence Models and Geostatistics Model (Case study: Duzduzan plain)

No permanent surface water resources in many parts of the country resulted in overdraft of limited underground water resources. Duzduzan plain is one of the UromiaLake sub basins. In this area, indiscriminate harvesting of groundwater resources has caused an average decline of 76 centimeters per year. The purpose of this research is Groundwater level spatiotemporal predicting using Artificial intelligence models and Geostatistics model. To predict the groundwater level in the duzduzan plain, initially the piezometera in the plain were classified. The groundwater level in each piezometers category were introduced as output for each of AI models and input of these models include a evaporation and a precipitation and grounwater level of the considered piezometers with one time delay (t0-1), respectively. Ann's model and Sugeno fuzzy (SF) model applied to predict groundwater level. The resulted values of Groundwater level were evaluated by statistical measures, includes root mean square error and correlation coefficient. The obtained results showed ANNs model has better performance. Then the result of ANNs model, including two year monthly groundwater level prediction data in selected piezometers, were used as inputs of geostatistics model (Kriging and Co Kriging) for predating spatially ground water level in the study area. Obtained results showed Co Kriging model has better performance.