Modelling of chemical fractionation patterns of metals in soils by two-way and three-way principal component analysis.

The 'pseudo-total' contents and the chemical fractionation pattern of eight toxic elements (As, Cd, Co, Cu, Cr, Ni, Pb and Zn) have been determined in 12 soil samples collected around a coal-fuelled power plant (Velilla del Río Carrión, Spain) by using, respectively, the US-EPA 3051 norm and the modified BCR (SM&T) chemical fractionation procedure. The 'pseudo-total' dataset has been analyzed by classical two-way principal component analysis (PCA) finding a PC accounting for the metal 'pollution' of the area. On the other hand, the three-dimensional (samples x metals x fractions) X array obtained after application of the modified BCR SM&T procedure, has been studied by matrix augmentation (MA-PCA) and three-way principal component analysis (3-PCA) using PARAFAC and TUCKER3 models. Whereas both MA-PCA and PARAFAC originated two-factor models biased towards the different chemical fractionation of the samples, the best TUCKER3 model [1, 2, 2] takes into account simultaneously both the 'pseudo-total' contents and the chemical fractionation of the soil samples. Therefore, the TUCKER3 originated a better representation of the global environmental impact caused by the power plant, and the plotting of the soil samples loadings, A1, in the physical space allowed to locate the most potentially hazardous areas.