Evaluation of Intelligent Models in Water Quality Simulation of the Babolrood and Sefidrood Rivers, Iran

The amount of Total Dissolved Solids (TDS) is an important factor in determining the water quality of rivers. The purposes of this paper was to investigate the efficiency of four intelligent models including, SVM-IWO, SVM-PSO, SVM-ABC, LS-SVM and Bayesian Neural Network, in predicting water quality of Babolrood and Sefidrood rivers, north  of Iran. To achieve this aim, unpublished measured monthly data including; Ca, Mg, HCO3, Na, SO4, EC, pH, and TDS from Quran-Talar station of Babolrood river and Prorijabad station of Sefidrood river were analyzed in the periods of 1966-2015 and 2002-2015, respectively. Evaluation of the four above-mentioned models, based on the correlation coefficient (R2), Root Mean Square Error (RMSE) and the Nash-Sutcliff coefficient (NSE) showed that the SVM-ABC model in both Babolrood and Sefidrood rivers with the highest R2s equal to 0.985 and 0.989 and the least amounts of RMSE equal to 10.493mg/l and 5.289 mg/l and the highest Nash-Sutcliffe equal to 0.983 and 0.992, respectively, has better and faster performance compared to others models in assessing water quality.