A quantitative method for clustering size distributions of elements

A quantitative method was developed to group similarly shaped size distributions of particle-phase elements in order to ascertain sources of the elements. This method was developed and applied using data from two sites in Houston, TX; one site surrounded by refineries, chemical plants and vehicular and commercial shipping traffic, and the other site, 25 miles inland surrounded by residences, light industrial facilities and vehicular traffic. Twenty-four hour size-segregated (0.056 < D-p (particle diameter) < 1.8 mu m) particulate matter samples were collected during five days in August 2000. ICP-MS was used to quantify 32 elements with concentrations as low as a few picograms per cubic meter. Concentrations of particulate matter mass, sulfate and organic carbon at the two sites were often not significantly different from each other and had smooth unimodal size distributions indicating the regional nature of these species. Element concentrations varied widely across events and sites and often showed sharp peaks at particle diameters between 0.1 and 0.3 mu m and in the ultrafine mode (D-p < 0.1 mu m), which suggested that the sources of these elements were local, high-temperature processes. Elements were quantitatively grouped together in each event using Ward's Method to cluster normalized size distributions of all elements. Cluster analysis provided groups of elements with similar size distributions that were attributed to sources such as automobile catalysts, fluid catalytic cracking unit catalysts, fuel oil burning, a coal-fired power plant, and high-temperature metal working. The clustered elements were generally attributed to different sources at the two sites during each sampling day indicating the diversity of local sources that impact heavy metals concentrations in the region. (c) 2004 Elsevier Ltd. All rights reserved.