Computational Screening of Environmental Proxies in Spectrometric Patterns from Humic Acids

Abstract

It is well known that humic acids (HAs) have a complex heterogeneous composition which depends not only on the precursor organic constituents but to large extent on the environmental factors in the soil system. In this study, an ‘omic’ approach was carried out taking advantage of the information from different data matrices obtained by analytical pyrolysis of HAs. It was intended to readily infer tentative environmental information avoiding complex, time-consuming spectral interpretations and supervised data processing. This approach is based on comparing the information extracted from data tables with compounds assemblages after HA degradation by (i) multivariate data treatments, mainly multidimensional scaling (MDS) and correspondence analysis, (ii) partial least squares (PLS) regression where the variable importance for projection (VIP) method was applied in order to identify the more informative chromatographic peaks and (iii) PLS-VIP method applied to ‘rebuilt pyrolysis-mass spectra’, i.e. to arrays consisting of one average mass spectra (2–60 min) of the total ion gas chromatogram. When the results obtained from approach (iii) were compared with those from approaches (i) and (ii), we realize the possibility for a tentative prediction of the relative amounts of almost all major families of degradation compounds and large number of individual compounds. Apart from this, a series of characteristics of HAs and whole soils (e.g. aromaticity, respiratory activity, extent of lignin alteration…) could be predicted with a variable success, not requiring previous interpretation of the mass spectra or integration of chromatographic peaks. The PLS-VIP plots in approach (iii) with the same format of a typical mass spectra lead to conclusions similar to MDS (used for variables ordination) or simple and multiple regressions, and illustrate the ion patterns relevant as regards to predicting soil and HA characteristics.