Source Apportionment of Fine and Coarse Particulate Matter and Daily Mortality in Two U. S. Cities – A Comparison of Different Methods

ISEE-243 Aim: The objective of this study was to identify sources of fine and coarse-mode particles (diameter less than 2.5 μm or between 2.5 and 10 μm, respectively) in two U. S. cities by different source apportionment methods and to estimate their associations with daily mortality. Methods: Fine and coarse particulate matter were collected every other day between 1979 and 1986 on dichotomous filters, in centrally located measurement stations in Boston, MA, and St Louis, MO. Their elemental composition was determined by X-ray fluorescence. We used two techniques, Principal Component Analysis (PCA) followed by a varimax rotation, and Positive Matrix Factorization (PMF), to identify and quantify sources. Results: In both cities, both methods identified factors indicating airborne soil, traffic, secondary sulphates, and coal combustion. In Boston, both methods identified factors indicating oil combustion, sea salt, fine crustal material, and another traffic factor. In St Louis, both methods identified several factors indicating industrial activities. A factor indicating wood combustion could be identified in both cities only by PMF. The correlations between corresponding factors identified by both methods range from 0.45 to 0.97, with most correlations being greater than 0.8. The R2 of the regression of the measured PM10 concentration on the factor scores was in both cities 0.87 for the PCA model and 0.88 for the PMF model. Daily counts of deaths were extracted from National Center for Health Statistics records. Associations of mortality with each source factor were estimated in Poisson regression models, adjusting for time trends and weather. They are reported as the percent increase in daily deaths associated with a 10 μg/m3 increase in PM10 mass concentration, from the respective source. In Boston, statistically significant associations were found for the oil combustion (3.2%, 95% confidence interval (CI): 0.5–6.0%) and secondary sulphate / coal combustion (1.5%, 95% CI: 0.3–2.6%) factors (PCA), as well as for the sulphate (2.3% (95% CI: 1.2–3.5%) and wood smoke (9.5% (95% CI: 6.1–13.1%) factors (PMF). In St Louis, statistically significant associations were found for the traffic (2.7%, 95% CI: 0.6–4.9%) factor (PCA) and for the lead smelter (6.5%, 95% CI: 1.2–12.1%) and traffic (6.4%, 95% CI: 2.4–10.6%) factors (PMF). Here the wood smoke factor (PMF) was positive and of borderline significance (10.6%, 95% CI: −0.5–23.1%). Conclusion: We conclude that both methods are able to identify similar factor patterns and factor scores. Furthermore, PMF seems to be able to resolve more factors and correlated factors. Besides combustion-related particles from traffic and secondary sulphates, wood smoke particles also seem to be a risk factor for mortality. These results are consistent with the key findings of Laden et al (2000), despite analyzing an extended dataset including also coarse-mode particles and using different source apportionment techniques. Disclaimer: This abstract does not necessarily reflect EPA policy.