Predicting the electrical conductivity of groundwater using the adaptive neuro-fuzzy inference system method (ANFIS) Case Study: Azarshahr, Ajabshir and Maragheh Plains

The aim of this study was to investigate the electrical conductivity of groundwater due to physical and chemical parameters of water using ANFIS-FCM method in Azarshahr, Ajabshir and Maragheh study areas of Urmia Lake catchment area. To achieve this goal, 82 water samples were taken from wells and springs in the plains and the data were chemically analyzed in the laboratory. Descriptive statistical data and correlation matrix of the studied parameters were obtained using SPSS software. By forming the correlation matrix, it was found that the four salinity parameters, soluble oxygen (DO), total soluble solids (TDS) and pH, have the highest correlation with electrical conductivity (EC) compared to other existing parameters. Therefore, the inputs of the model included the four mentioned parameters and the output was selected according to the purpose of the research, electrical conductivity.  After standardization, the data entered the MATLAB environment and using ANFIS-FCM method, the electrical conductivity of groundwater was predicted. In this method, 80% of the data (66 samples) were randomly selected for the training data set and 20% of the data (16 samples) were randomly selected for the test data set. For ANFIS-FCM training data set, R2, RMSE and VAF values were 0.9999, 0.0032399 and 0.99993, respectively, and also for ANFIS-FCM test data set, R2, RMSE and VAF values were 0.9998, respectively, 0.0029949 and 0.99972 were obtained. Using the results of this model, it was found that the estimated electrical conductivity in the studied areas had a very good accuracy and high correlation with the measured values. As a result, the ANFIS-FCM intelligent method is an effective, efficient and accurate way to estimate the physical and chemical parameters of water.