DRIFT spectroscopy and permutation importance algorithm in quantitative analysis of organic matter in soil model systems

Abstract

In order to obtain useful MIR spectrochemical data of soil organic matter for the development of remote sensing methods, synthetically prepared soils with artificial precisely defined organic matter fractions, DRIFT spectroscopy and permutation importance algorithm were used in this paper. In terms of imitation of soil organic matter, sample model systems were prepared with precisely defined added values of added organic components. After MSC and SNV spectral treatments using PCA and LDA techniques and DRIFT spectra, the soil was classified according to the percentage of organic matter. Using the KDE+permutation importance algorithm, three significant MIR spectral regions were obtained for percentage grouping: 600- 1000cm-1 (skeletal vibrations of organic matter); 1750-2250cm-1 (Total reflectance+quartz overtones) and 3250-3950cm1 (Hydroxyl groups). In terms of the potential for quantitative analysis, the calculated wavelength ranges match well with the classical spectrochemical theoretical basis of analytical methodologies. Also, extracted useful spectrochemical data can be potentially used in the development of new remote-satellite detection methods (ASTER satellite in SWIR and MIR range).