Estimation of Water Quality Parameters in the Sepidrood River by ANFIS, GEP and LS-SVM Models

Abstract

Rivers are the most important water supply resource for the drinkable, agricultural and industrial demands. Therefore, estimation of water quality parameters in rivers is an essential and necessary task. This research applies the Adaptive Neuro-Fuzzy Inference System (ANFIS), the Least Squares-Support Vector Machines (LS-SVM) and the Gene Expression Programming (GEP) for estimation of Total Dissolved Solids (TDS), Electrical Conductivity (EC) and Total Hardness (TH) in the Sepidrood River and a 40 year period. The applied performance criteria are the correlation coefficient (R), the Nash-Sutcliffe model Efficiency coefficient (NSE), the Normalized Mean Squared Error (NMSE) and the Mean Absolute Error (MAE). These methods have high ability for estimation of water quality parameters. The best method is LS-SVM method for estimation of TDS (RTrain=0.95 RTest=0.96). The best method is GEP method for estimation of EC (RTrain=0.94 RTest=0.95). The best method is ANFIS method for estimation of TH (RTrain=0.92 RTest=0.94). This research shows that intelligence methods can estimate unmeasured concentration of qualitative parameters by concentration of other qualitative parameters.