Multi-Block Data Modeling for Characterization of Soil Contamination: A Case Study

Multi-block (heavy metals, pesticides, physico-chemical parameters) data set pertaining to the soils of alluvium region in Indo-Gangetic plains was analyzed using principal component analysis (PCA) and multiple factor analysis (MFA) methods to delineate the contaminated sites and to identify the possible contamination sources in the study region. In normal PCA, the first three factors were dominated mainly by heavy metals, pesticides and physico-chemical variables, respectively, thus identifying samples/sites contaminated with these. The MFA results, due to its unique weighting scheme of variables of different blocks extracted, to more realistic information about the spatial distribution of samples and relationships among the variables. MFA minimized the influence of variables of one single block on the first few components, allowing variables of all blocks equally to share the common MFA space. This resulted in delineating the sites/regions contaminated with variables (Al, Co, Cu, Mn, Ni, Pb, V, Na, SO₄, aldrin, lindane, HCB, HCH, DDT, and endosulfan) of all the blocks, rather than by particular block variables as in case of normal PCA, where, the variables of single block dominate the first factors, suppressing other block variables. MFA which can be considered as a method for standardization of the multi-block variables was successfully applied to the three block data set of soils.