عنوان مقاله [English]
Changes in groundwater levels are important factors and effectiveness in study and management of groundwater resources use in agricultural, domestic and industrial. Lack of harvesting and reading points will limit access to this information due to financial and time constraints. Thus, application of interpolation and approximation water level methods are based on information from neighboring regions and have a special place in the study of groundwater resources. In this study, conventional interpolation method, the kriging and artificial intelligence systems evolve neural - fuzzy adaptive were evaluated land Dashtenaz Sari city. The research shows that three minimum curvature method, kriging and neural-fuzzy inference system adaptive values RMSE, 56.32, 55.12 and 53.97 cm, and the mean absolute error of 43, 42.54 and 42.94 percent and as well as the values of R2 0.56, 0.54 and 0.55, are the same accuracy in the interpolation of groundwater levels in the study area. However, due to large fluctuations in water levels in this area varies between 12 and 315 cm, None of these methods are of high accuracy and acceptable. In addition, a Neural-fuzzy inference system adaptive, Withdrawal of membership function trapmf the input layer and 5 for the membership function for each input and output layers with constant learning algorithm for hybrid and conventional interpolation and kriging parameters, the default get the best results by software surfur.