Cluster analysis of electron microprobe analysis data of individual particles for source apportionment of air particulate matter

Resuspended soil particles were sampled in the vicinity of a zinc smelter and analyzed for 16 elements using automated electron microprobe X-ray analysis (EPMA) of the individual particles. The data sets containing the concentrations of the fine and coarse mode particles were interpreted using a combination of both hierarchical and nonhierarchical cluster analysis. The clusters were used as an initial training set in a nearest centroid sorting clustering over all receptor sites to gain insight in the spatial distribution of the clusters.Several aspects of the method are discussed, such as the data processing, the optimization of the cluster algorithm, the analytical error, the cluster compositions and the uncertainties on the cluster abundances.The composition of the samples was found to be mainly dominated by particles of crustal origin, associated with metals as Zn, Mn and Pb. Considerable differences in composition and abundance were observed between coarse and fine mode particle fractions and between different sampling sites.