Comparing Regression and Artificial Intelligence Models for Estimating Soil Exchangeable Sodium Percentage from Sodium Absorption Ratio (Case Study: Miankangi Region Soils, Sistan)

Sodium absorption ratio (SAR) and exchangeable sodium percentage (ESP) are two indicators of sodic soils. Several approximate correlations between ESP and SAR for soils of different regions in the world have been reported. The purpose of this study is to find the relationship between ESP and SAR in Miankangi region, in Sistan plain, and assessing possibility of ESP calculation from SAR. Thus, 189 soil samples from the study area were collected and analyzed. Relationship between ESP and SAR was determined by using the logarithmic regression equation of ESP = 8.07 × ln(SAR1:1) 10.20 and linear equation of ESP = 0.78 SAR1:1 15.69 (SAR1:1 is SAR in 1:1 soil to water extract), which could explained 83% and 67% of ESP variations respectively. Then, performances of multi-layer perceptron (MLP) network and artificial neuro-fuzzy inference system (ANFIS) were studied. Results showed the capability and better outcomes of MLP and ANFIS in comparison to regression models (correlation coefficient and root mean square error values were 0.94 and 0.05, respectively). These results demonstrated the superiority of intelligent models in explanation of the relationship between ESP and SAR compared with linear and nonlinear regression relations.