Comparison of Anfis, Artificial neural network, and Gene expression programming to estimate the amount of Water hardness (Case study: Mazandaran Plain)

Rate of water hardness is an important factor in hydrogeology, particularly in groundwater quality researches. In recent decades, the artificial intelligence systems such as artificial neural networks have many applications in various sciences, including management of water resources. In this study, estimated rate of groundwater hardness in Mazandaran plain, using Gene expression programming have been studied and the results is compared with other intelligent methods such as artificial neural network and Anfis. For this purpose the hydrogen carbonate, chloride, sulfate, magnesium and calcium monthly time scale of the period (1994-2014) was selected as inputs and water hardness as output. Standard deviation of the correlation coefficient, root mean square error, and coefficient of Nash Sutcliff were used to assess various methods. The results showed that Gene expression programming model has the maximum correlation coefficient 0.960, minimum root mean square error 0.112, mean absolute error 0.171 coefficient of Nash Sutcliff 0.880 was in the verification phase. In overall, the results showed that the Gene expression programming model has high performance in estimating some maximum and intermediate values of groundwater hardness.