Modeling Qualitative Parameters of SAR, EC, and TDS in Groundwater using Optimized Artificial Neural Network Model (Case Study: Behbahan Plain)

Proper modeling of groundwater quality is an important tool for planning and decision making in water resources management. The present study was conducted to simulate the groundwater quality parameters of Behbahan Plain including SAR, EC, and TDS using ANN and ANN + PSO models and finally to compare their results with the measured data. Input information to the models gathered were for TDS quality parameter including electrical conductivity, absorption ratio of sodium, sulfate, calcium, magnesium and sodium, for SAR quality parameter including total dissolved salts, sodium, bicarbonate, and for EC quality parameter including sulfate, calcium, magnesium and ratio Sodium uptake from 2010 to 2017. The results indicated that the highest prediction accuracy of quality parameters of EC and TDS is related to the ANN + PSO model with the tangent sigmoid activation function and for the SAR parameter is related to the ANN + PSO model with the logarithm sigmoid activation function so that the MAE and RMSE statistics had the minimum and R2 had the maximum value for the model. In the test phase the values calculated were for EC parameter RMSE=14.61, MAE=9.27, NRMSE=0.41, EF=0.942, and R2=0.96 and for TDSparameter RMSE=22.21, MAE=18.32, NRMSE=0.398, EF=0.925, and R2=0.836 and for SARparameter RMSE=9.45, MAE=7.2, NRMSE=0.301, EF=9.27, and R2=0.974. In addition, the results of the mean comparison between measured and simulated data showed that the predicted values with models were not significantly different with the measured date.