Modeling of qualitative parameters (Electrical conductivity and total dissolved solids) of Karun river at Mollasani, Ahvaz and Farsiat stations using data mining methods

Background and Objective: In the present study, EC and TDS quality parameters of Karun River were modeled using data-mining algorithms including LSSVM, ANFIS, and ANN, at Mollasani, Ahvaz and Farsiat hydrometric stations. Material and Methods: Eight different inputs including the combination of Cl-1, Ca+2, Na+1, Mg+2, K+1, CO32-, HCO3, and SO42- with discharge flow (Q) were selected as non-random and random calibration inputs for these algorithms. Then, in order to guarantee the accuracy of the results, the simulation was performed by random calibration and the results of the two methods were compared. In the next step, the EC and TDS parameters were modeled based on the four parameters of Na+1, Cl-1, Ca+2, and Q and a lag time of zero to three months. Results: Modeling results indicated that Na+1, Cl-1, and Ca+2 have the highest influence on modeling of EC and TDS parameters. The LSSVM algorithm was the most accurate in modeling EC and TDS parameters. Among the studied stations, the highest precision for EC and TDS modeling belongs to Ahvaz and Mollasani station, which has 16% and 36% higher coefficient of determination. LSSVM has highest accuracy in modeling the oscillation and peak EC and TDS parameters in during times. Conclusion: The methods and models applied in the present study especially the LSSVM algorithm, can be a useful decision-making tool for predicting and qualitative management of rivers, including rivers in the Karun catchment area. The results of modeling the quality parameters of the rivers were reliable and usable by using both non-random and random calibration methods. However, the accuracy of the random calibration method was slightly higher.