Multivariate Receptor Modeling Applied to PM2.5 and Air Toxics Monitoring Sites in the Midwest

Positive Matrix Factorization (PMF) receptor modeling was performed on four Midwestern sites: Northbrook, IL, Gary, IN, Dearborn, MI, and St. Louis, MO. The datasets included collocated samples of speciated fine particulate matter (PM2.5) and volatile organic compounds (VOCs, or air toxics) collected from 2003 to 2007. At each site, source factors were identified and quantified separately for PM2.5 and VOCs. The daily contributions of these two sets of sources were compared via pair-wise Spearman rank correlation. The Northbrook site was selected to model a combined PM2.5 and air toxics dataset.;PMF modeling of PM2.5 datasets produced results comparable to previous studies. This study did not support assumptions that organic carbon (OC) represents vehicle fuel emissions and elemental carbon (EC) signifies diesel fuel. PMF modeling of VOCs was limited by the number of compounds reported at each site and source factors were characterized by strong seasonal trends. Health risks from metals were attributed to metallurgic industry; toxic VOCs were apportioned to mobile sources and local industry.;Correlations between PM2.5 and VOC source factors were statistically significant but weak. The greatest relationship between factors related to mobile transport was between PM2.5-OC and VOC liquid fuel at Northbrook. Correlation of PM2.5 and VOC factors was weakest at the highly industrial Gary and Dearborn sites. PMF modeling of a combined Northbrook dataset served to confirm similarities between certain PM2.5 and VOC factors. This study did not demonstrate the usefulness of combined receptor modeling in a quantitative sense. However it did point to mobile-source related emissions as an important area to focus on for reduction of criteria pollutants and air toxics health risks.