Modeling of soil exchangeable sodium percentage using easily obtained indices and artificial intelligence-based models

Abstract

Salinization and alkalization of land resources are the major obstacles to their optimal usage in many arid and semi-arid regions of the world, including Iran, since potential evapotranspiration is more noteworthy than precipitation in these areas. The amount of water that enters the soil is low and this results in salt accumulation in soils, which makes the soil infertile. Moreover, existence of salts, for example, sodium, in soils causes dispersion of soil particles and soil degradation, and intensifies soil erosion too. Monitoring exchangeable sodium percentage (ESP) variability in soils is both time-consuming and costly. However, in order to estimate the amounts of amendments and land management, it is necessary to know ESP variation and values in sodic or saline and sodic soils. Thus, introducing a method, which utilizes easily obtained indices to estimate ESP indirectly is more optimized and economical. Input and output data, i.e., EC e (dS m-1), clay (%), pH and ESP (%) were collected and measured from 100 soil samples in light of a stratified random sampling from Mashhad Plain, Khorasan-e-Razavi Province, Northeast Iran. This study aims to propose some models to estimate ESP by easily obtained properties of soil. In this regard, the efficiency of artificial intelligence-based (AI) models (i.e., Artificial Neural Network, ANN, and Adaptive Neuro-Fuzzy Inference System, ANFIS) was investigated and compared. Accuracy results showed that owing to highest R 2 and the lowest mean square error (MSE), ANFIS model predictions were superior to the MLP model for indirect estimation of soil exchangeable sodium percentage.

Prediction of Soil Salinity/Sodicity and Salt-Affected Soil Classes from Soluble Salt Ions Using Machine Learning Algorithms

“…By adding the soil pH, EC e , TOC, clay, silt, and sand to the matrix of predictor variables, only the performances of PLS and SVM regressions to predict soil ESP showed a significant improvement (Table 5) compared to those in Table 1. These results partly contrast with those of Keshavarzi et al [56] who obtained R 2 /MSE values of 0.84/5.36 and 0.90/5.09 for the AI-based models Multi-Layer Perceptron and Adaptive Neuro-Fuzzy Inference System, respectively, for predicting ESP from EC e , pH, and clay. Although the RF classification model obtained a slight increase in effectiveness (Table 5), should be noted the redundancy caused by the soil EC e and pH as explanatory variables and as classifiers of the explained categories at the same time; however, their further inclusion might be pertinent if more easily obtained parameters are used, such as EC and pH measured in soil-water suspensions.…”

Prediction of TBM penetration rate from brittleness indexes using multiple regression analysis

“…probabilistic and soft computing techniques, such as artificial neural networks, regression trees, adaptive neuro-fuzzy inference systems, fuzzy inference systems, a hybrid ANN and GA, hybrid ANN and imperialist competitive algorithm, and hybrid ANN and particle swarm optimization technique Heddam 2016a, b;Keshavarzi et al 2016). Simple regression is the simplest form of regression analysis involving one independent variable and one dependent variable.…”

Determination of Soil Agricultural Aptitude for Sugar Cane Production in Vertisols with Machine Learning

“…The parameters predicted with more accuracy were Ca/Mg (1.0 using KNN), B (86% using RF) and OM (86% using LR) and the parameter with least accuracy was P with 45%, using LR. Other authors have aimed to predict different parameters, with accuracy ranging from 86% to 97%, such as the following: exchangeable sodium percentage with different models (ANN (89%) and Adaptive Neuro Fuzzy Inference System (92%)) [40], OM with different models (Kennard-Stone (KS), Successive Projections Algorithm (SPA), Competitive adaptive weight weighting algorithm (CARS) and their combination).…”