Comparison of numerical model, neural intelligent and GeoStatistical in estimating groundwater table

Study and analysis of groundwater due to the complexity are difficult and time consuming. Modeling as an efficient method with the lowest cost provides studding of groundwater managers. The purpose of this study was comparison of the numerical model, neural intelligent and GeoStatistical in groundwater table changes modeling. The information of Hamedan – Bahar aquifer as one of the most important water sources in Hamedan province, was studied. In this study, MODFLOW numerical code in GMS software, artificial neural network (ANN) and neural – fuzzy (CANFIS) method in NeuroSolution software, wavelet- neural method in MATLAB software and GeoStatistical method in ArcGIS software were used. The results showed that the accuracy of methods in estimation of the groundwater table with the lowest Normal Root Mean Square Error (NRMSE) include Wavelet-ANN, CANFIS, GeoStatistical, ANN and numerical model, respectively. The NRMSE value in Wavelet-ANN method as optimization method was 0.11 % and in numerical model was 2.2 %. Also the correlation coefficients were 0.998 and 0.904, respectively. So, application of neural combination models, specially, wavelet theory in estimated the groundwater table is most suitable than GeoStatistical and numerical model. Moreover, in the neural intelligent models were applied latitude, longitude and altitude as available variables in input models.