Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4445116 | Atmospheric Environment | 2006 | 22 Pages |
Four gas components (CO, SO2, HNO3 and NOy) and PM2.5 (particulate matter ⩽2.5 μm in aerodynamic diameter) composition data including eight individual carbon fractions collected at four sites in Georgia and Alabama were analyzed with the positive matrix factorization (PMF) method. Multiple linear regression (MLR) was applied to regress the total PM mass against the estimated source contributions. The regression coefficients were used to scale the factor profiles. Nine factors were resolved at two urban sites (Atlanta, GA (JST) and Birmingham, AL (BHM)) and one rural site (Centerville, AL (CTR)). Eight factors were resolved at the other rural site (Yorkville, GA (YRK)). Six factors we refer to as soil, coal combustion/other, diesel emission, secondary sulfate, secondary nitrate, and wood smoke are common among the four sites. Two industry-related factors are similar at the two sites in the same state, but differ between states. Contrary to previous results using only PM2.5 data with non-speciated EC and OC data, diesel and gasoline emission factors were resolved at the two urban sites instead of only one single motor vehicle factor; diesel and gasoline factors were also resolved at the CTR site and a diesel factor was found at YRK instead of no motor vehicle factors at the two rural sites. The inclusion of gas components also improved the identification of the coal combustion/other factor among the four sites. This study shows that inclusion of gas phase data and temperature-resolved fractional carbon data can enhance the resolving power of source apportionment studies, especially for the factors we refer to as gas, diesel, and coal combustion/other.