Modelling of Sodium Adsorption Ratio of the Soil Using Adaptive Neuro Fuzzy Inference System

Abstract

Soil management for crop production is a major concern for sustainability agricultural. Sodium adsorption ratio (SAR) of the soil is needed to quantify the amount of amendments. The objective of this study was to evaluate the performance of Adaptive Neuro Fuzzy Inference System (ANFIS) for estimating the SAR of the soil. In this research, 153 observations of soil properties were collected from literature and actual laboratory analysis and SAR was calculated. Soil electrical conductivity (EC), soil pH, sand, silt and clay percentages were taken as inputs and the SAR in the soil was taken as output. Based on the membership functions, four ANFIS models were tested against the calculated sodium absorption ratio to assess the accuracy of each model. The tested membership functions were triangular-shaped membership function (trimf, ANFIS1), generalized bell-shaped membership function (gbellmf, ANFIS2), trapezoidal-shape membership function (trapmf, ANFIS3) and Gaussian curve membership function (gaussmf, ANFIS4). The results showed that ANFIS4 was the most accurate membership function where the training error was 0.10492.   Meanwhile, the training error for ANFIS1, ANFIS2 and ANFIS3 were 0.1945, 0.22751 and 1.4297, respectively. The comparison between results of ANFIS and observed SAR using testing data set shows that the coefficient of determination was 0.9907. Results indicate that ANFIS modeling is a promising alternative to the traditional approach and it significantly decreases calculation time in determining SAR of the soil.