Evaluation of the CMB and PMF models using organic molecular markers in fine particulate matter collected during the Pittsburgh Air Quality Study

This analysis investigated different possible strategies for source apportionment of airborne fine particulate matter (PM2.5) using data collected as part of the Pittsburgh Air Quality Study (PAQS). More specifically, we apportioned the organic fraction of the winter and summer season PM2.5 using two source-receptor models - the EPA Chemical Mass Balance 8.2 (CMB) and EPA Positive Matrix Factorization 1.1 (PMF) models - and tested several case scenarios with each model by varying either the chemical species or source profiles used as model input. Moreover, we added the constraint of selecting only individual molecular marker species with concentrations above their minimum quantitative limits. Model results suggest that the molecular marker and source profile selection can strongly affect the model, as reflected in the source contribution estimates determined by both CMB and PMF. Biomass burning and mobile emissions sources were identified by both models as being major source contributors in Pittsburgh. A third source was consistent with a meat cooking profile but was more likely a combination of cooking and secondary organic aerosol. As expected, the relative proportion of each source's contribution depended on both the season and on whether the CMB or PMF model was applied. Selecting fewer species in CMB resulted in less mass being apportioned, and an unrealistically large wood burning contribution estimate. Swapping a wildfire profile for one of the two wood burning profiles also resulted in less mass being apportioned in the winter. The results suggest that CMB can distinguish between fireplace burning and wildfire contributions when appropriate species are included. The gasoline/diesel split also varied by up to an order of magnitude, depending on which model was applied and which species were fit. Published by Elsevier Ltd.