Comparison of PM2.5 source apportionment using positive matrix factorization and molecular marker-based chemical mass balance.

A comprehensive comparison of positive matrix factorization (PMF) and molecular marker-based chemical mass balance (CMB-MM) modeling on PM2.5 source contributions was conducted for particulate matter measurements taken at Jefferson Street, Atlanta, Georgia (JST). The datasets used in each type of receptor modeling were different: CMB-MM used data of primarily organic tracers plus a couple elements measured from 51 24-h PM2.5 samples collected in July 2001 and January 2002. While for PMF, with elements, ions, five gaseous components, and eight temperature-resolved carbon fractions as the input data, both source profiles and contributions were resolved from a total of 932 daily PM2.5 samples covering a 3-year period between January 2000 and December 2002. The model results for the overlapping periods (July 2001 and January 2002) were extracted for comparison. Seven primary sources and three secondary sources were resolved by CMB-MM, while a total of nine primary and secondary factors were resolved by PMF. On average, 107% and 85% of PM2.5 mass were explained by CMB-MM and PMF, respectively, with secondary aerosols handled differently in the above two methods. Four similar sources were resolved by both methods, with good correlation for road dust, but fair for gasoline exhaust and wood combustion. The CMB-MM diesel exhaust has very poor correlation with the PMF resolved diesel exhaust. However, the CMB-MM combined mobile source has improved correlation with the PMF result as compared with the diesel exhaust source. If only the winter data were included, the CMB-MM combined mobile source shows enhanced correlation with the PMF combined source, as compared with the single source of diesel exhaust or gasoline exhaust.