Source apportionment of atmospheric PAHs and their toxicity using PMF: Impact of gas/particle partitioning

24-h PM2.5 samples were simultaneously collected at six sites in a subtropical city of South China during November–December, 2009. Particle-phase concentrations of polycyclic aromatic hydrocarbons (PAHs) and organic tracers such as hopanes for vehicular emissions (VE), levoglucosan for biomass burning (BB) and picene for coal combustion (CC) were determined. Meanwhile, their gas-phase concentrations were calculated from gas/particle (G/P) partitioning theory using the particle-phase concentrations. The 4 ring PAHs (fluoranthene to chrysene) had lower particle-phase fractions (10%–79%) than other species. Estimated BaPeq and lifetime cancer risk for particle-only (P-only) vs gas + particle (G + P) data sets showed similar values, indicating PAHs with 5–7 rings dominated the carcinogenicity of PAHs. Positive Matrix Factorization (PMF) was applied on both P-only and G + P data sets to estimate the source contributions to PAHs and their toxicity. Three common sources were identified: VE, BB and CC, with CC as the most significant source for both particulate (58%) and total (G + P, 40%) PAHs. While CC exhibited consistent contributions to BaPeq for P-only (66%) vs G + P (62%) solutions, VE and BB contributions were under- and overestimated by 68% and 47%, respectively by the P-only solution, as compared to the G + P solution. The results provide an insight on the impact of G/P partitioning on the source apportionment of PAHs and their toxicity. •PAHs and organic tracers were measured in 136 PM2.5 samples.•Gas-phase concentrations of these SVOCs were predicted.•PMF was applied to both P-only and G + P data sets.•Sources contributions to PAHs and their toxicity were estimated.•P-only solution significantly underestimated vehicular emission contributions to BaPeq.