Soil salinity mapping using Landsat 8 OLI data and regression modeling in the Great Hungarian Plain

Abstract

Salt's deposition in the subsoil is known as salinization. It is caused by natural processes such as mineral weathering or human-made activities such as irrigation with saline water. This environmental issue has grown more critical and is frequently occurring in the Hungarian Great Plain, adversely influencing agricultural productivity. This study aims to predict soil salinity in the Great Hungarian Plain, located in the east of Hungary, using Landsat 8 OLI data combined with four state-of-the-art regression models, i.e., Multiple Linear Regression, Partial Least Squares Regression, Ridge Regression, and Feedforward Artificial Neural Network. For this purpose, seventy-six soil samples were collected during a field survey conducted by the Research Institute for Soil Sciences and Agricultural Chemistry between the 15 of September and the 15 of October, 2016. We used the min–max accuracy, the root-mean-square error (RMSE), and the mean squared error (MSE) to evaluate and compare the four models' performance. The results showed that the ridge regression model performed the best in terms of prediction (MSEtraining = 0.006, MSEtest = 0.0007, RMSE = 0.081), with a min–max accuracy equal to 0.75. Hence, the application of regression modeling on spectral indices, principal component analysis, and land surface temperature derived from multispectral data is an efficient method for soil salinity assessment at local scales. The resulting map can provide an overview of salinity levels and evaluate the efficiency of land management strategies in irrigated areas. An increase in sampling density will be recommended to validate this approach on the regional scale.