Classification of forensic soil evidences by application of THM-PyGC/MS and multivariate analysis

The forensic classification of soil samples was carried out by thermally assisted hydrolysis and methylation (THM) of soil organic matters (SOM) using pyrolysis-gas chromatography/mass spectrometry (PyGC/MS). In this work, thirty-three THM derivatives were detected as SOM contained in <3 mg soil. The specific ions of the mass spectra were selected to separate and minimize the interference between SOM peaks. SOM data were normalized with the sum of peak areas to correct the amounts of SOM contained in the soil, and the chemometric approach based on principal component analysis (PCA), hierarchical cluster analysis (HCA) and linear discriminant analysis (LDA) was employed to evaluate and compare the soil classification. The first seven principal components (PCs) accounted for 94.8% of total cumulate variance and these PCs were statistically determined by multiple comparisons (Tamhaneʹs T2 and Dunnettʹs T3) for the post hoc test (p-value < 0.05) and were used to construct the LDA model. It was determined that multiple comparisons were a statistically good criterion for deciding on the number of PCs for the LDA model. It was also concluded that the discrimination model correctly classified 40 soil samples into six clusters with high accuracy. Furthermore, the eleven marker compounds were investigated according to the loadings of PCs and the normalized data. These results demonstrated that lignin, fatty acid and urea can be used as potentially useful compounds to characterize soil samples for forensic purposes.