Incorporation of parametric factors into multilinear receptor model studies of Atlanta aerosol

Abstract

In prior work with simulated data, ancillary variables including time resolved wind data were utilized in a multilinear model to successfully reduce rotational ambiguity and increase the number of resolved sources. In this study, time resolved wind and other data were incorporated into a model for the analysis of real measurement data. Twenty-four hour integrated PM2.5 (particulate matter ≤2.5μm in aerodynamic diameter) compositional data were measured in Atlanta, GA between August 1998 and August 2000 (662 samples). A two-stage model that utilized 22 elemental species, two wind variables, and three time variables was used for this analysis. The model identified nine sources: sulfate-rich secondary aerosol I (54%), gasoline exhaust (15%), diesel exhaust (11%), nitrate-rich secondary aerosol (9%), metal processing (3%), wood smoke (3%), airborne soil (2%), sulfate-rich secondary aerosol II (2%), and the mixture of a cement kiln with a carbon-rich source (0.9%). The results of this study indicate that utilizing time resolved wind measurements aids to separate diesel exhaust from gasoline vehicle exhaust. For most of the sources, well-defined directional profiles, seasonal trends, and weekend effects were obtained. © 2003 Elsevier Ltd. All rights reserved.