PIXE elemental characterization of air masses using a multivariate statistical technique

Abstract

An example is given to show the possibility of using a multivariate statistical evaluation technique in order to extract more information from a multielemental PIXE data set. Four weeks of continuous sampling was carried out at a background air pollution monitoring station in Sweden. Samples were collected both in fine and coarse mode, with a cutoff at 2 μm. In the subsequent PIXE analysis of the samples, 12–16 elements were detected in the fine fraction and 9–12 elements in the coarse fraction. The fine fraction PIXE data was further analysed using the multivariate statistical programme package SIMCA, which combines a pattern recognition technique and principal component analysis. Based on 1000 mbar back trajectories for the sampling period, principal component class models were constructed for Easterly and North-Westerly air masses using 15 elements (S, K, Ca, Ti, V. Cr, Mn, Fe, Ni, Cu, Zn, Ga. As, Br and Pb). For these elements, mean concentration values and standard deviations for the two classes are given. A methodology is presented which excluded outliers and facilitated the calculation of classes with a restricted and definable data distribution representative of the elemental composition of the air masses originating from the two source regions.