Corrigendum to “A source apportionment of U.S. fine particulate matter air pollution”[Atmos. Environ. 45/24 (2011) 3924–3936

Abstract

Subsequent to this manuscript going to press, we discovered that approximately one-third of the nitrogen dioxide (NO2) data downloaded from the Health Effects Institute (HEI)/Atmospheric and Environmental Research (AER) database for the “nearest neighbor” site (16,641 of 46,478 observations) were actually from sites outside the metropolitan statistical area (MSA) where the Chemical Speciation Network (CSN) site being considered was located. In this Corrigendum, we evaluate what aspects of the reported results were affected, and re-estimate the source impacts and profiles after addressing this NO2 data issue. To evaluate the importance of this NO2 data issue to the initial factor analysis, we repeated the factor analysis: 1) for the entire dataset considering only trace element data (i.e., without the NO2 variable) (n 1⁄4 46,478 observations); and, 2) for the subset of the data having NO2 data fromwithin the same MSA as the CSN site (n 1⁄4 29,837 observations). In the first case, all source-related factors except for Traffic were reproduced with negligible change, producing correlations to the original factors as follows: Crustal/Soil r 1⁄4 1.00, Metals-Related r 1⁄4 0.99, Salt Aerosols r1⁄4 1.00, Residual Oil r1⁄4 0.98, Steel Industry r1⁄4 1.00, Coal-Burning r1⁄4 0.99, and Biomass burning r1⁄4 0.98. However, the Traffic factor without the NO2 variable had a distinctly lower correlation with the original Traffic factor (i.e., r 1⁄4 0.86). As previously noted in the original paper, “Without the inclusion of NO2 to the analysis, it was not possible for the factor analysis to separate a distinct traffic component”. Clearly, the Traffic component was most notably affected by the elimination of NO2 variable because it was the component with the largest NO2 loading in the original analysis. In our follow up factor analysis (for the subset with within-MSA NO2 data, n 1⁄4 29,837 observations), we found that the Traffic component was again resolved, just as in the original analysis (r 1⁄4 0.95 with the original scores for these 29,837 data observations), indicating that the original Traffic component was not an artifact of the non-MSA NO2 data, and that it was a valid representation of the Traffic component, with the exception of the observations for out-of-MSA NO2 values.