Soil analysis using diffuse reflectance infrared Fourier transform spectroscopy and partial least square model

Abstract

Nowadays, the increasing population urged us to have soil quality determination techniques that can be conducted on-site, inexpensive, and much more rapid resulting in cost savings and faster decisions. Traditionally, the method that uses capillary Gas Chromatography/ Flame Ionization Detector (GC/FID) is time-consuming and expensive. This study evaluates the performance of the current practice, Diffuse Reflectance Infrared Fourier Transform (DRIFT) spectroscopy, coupled with multivariate modelling methods, to analyse soil samples. Partial least square regression (PLSR) and partial least square-discriminant analysis (PLS-DA) methods were implemented using soil spectra obtained to predict soil properties consisting of sand and clay plus classify the soil type. Using the PLS-DA method, the classification of soil types shows a promising result with the area under the receiver operating characteristic curve (AUC) of 0.95, meaning that the model is a very good classifier. This study demonstrates the potential of using the PLS-DA method in DRIFT spectroscopy to be a rapid and accurate method in classifying the soil type. However, moderate accuracy in PLSR analysis needs further investigation and exploration in sample size and methodology.