Using image analysis to estimate the soil organic carbon content

Abstract

Soil test for organic carbon content using the Walkley-Black standard method is a laborious task that takes almost 5 working days to report the results. In order to improve lab analysis efficiency, a new analytical method using digital images is proposed. Using multivariate image analysis (MIA) soil organic carbon (SOC) contents can be determined in <3 days. The MIA method for SOC was developed using 177 soil samples collected from 3 regions of Brazil (North, West Central, and Northeast). Digital images of soil samples were correlated with the organic carbon contents determined by the Walkley-Black standard method using multivariate regression algorithms. In the end, MIA model employing a machine learn approach using least squares support vector machine (LS-SVM), presented an excellent correlation, r2 > 0.93, between image data and SOC contents measured by the standard analytical method. The proposed MIA method is eco-friendly, cheap, fast, and a clean alternative that can be employed by soil testing laboratories for measuring SOC contents.