A new optical method coupling light polarization and Vis–NIR spectroscopy to improve the measurement of soil carbon content

Visible and Near-infrared (Vis–NIR) spectroscopy is now a common analytical method to measure different physical and chemical properties of soils, including carbon content. However, prediction model accuracy is insufficient for Vis–NIR spectroscopy to replace routine laboratory analysis. One of the main issues of this technique is facing up to is light scattering due to soil particles. It causes departure in the assumed linear relationship between the absorbance spectrum and the concentration of the chemical components as stated by Beer–Lambert's law, which underpins the linear calibration models.Therefore, it is essential to improve the quality of the measured signal in order to optimize the calibration results. Optics can help to mitigate scattering effect on the signal.The aims of this study were to test the feasibility of a new optical setup, names PoLiS, coupling linearly polarized light with a Vis–VNIR (350–800nm) spectrometer to free the measured spectra from multiscattering effect. The measured signals were used to model the chemical absorbance of the soil samples using Dahm's equation in the frame of the Representative Layer Theory (RLT).The study was conducted using a set of 52 soil samples collected in France (in the French calcareous Prealps) to predict soil total organic carbon (TOC) content. The PoLiS absorbance signal tended to be more linearly related to the concentration of organic carbon, which is an important pre-requisite to perform linear multivariate modeling. In a second step, the PLS models achieved for TOC performed appreciably better than models based on classical reflectance spectra. The standard errors of cross validation decreased from 20.8 to 17.6gkg−1 and the coefficient of determination R2 improved from 0.82 to 0.87 on ground samples. To compare the added-value of the PoLiS method, we benchmarked the PoLiS prediction models against models built from empirically preprocessed spectra. Again, the PoLiS method showed better performances.This work confirmed that by optical means, it is possible to significantly improve the quality of a spectroscopic signal and as a consequence, improve also the quality of the linear model.