abstract
- © 2019 John Wiley & Sons, Ltd.Fine particulate matter is of great concern in urban areas because it is one of the most heavily emitted pollutants by numerous emission sources. Several receptor models have been developed to estimate contributions of emission sources to the observed concentrations of air pollutants. Most of these tools require source profiles and uncertainties of the measured concentrations of the pollutants in the atmosphere. In addition, these receptor models have been widely applied using a trace element approach. However, a quick overview of the source contributions and key chemical species is desirable before selecting and applying a robust receptor model. Here, partial least squares (PLS) was applied as a preliminary source apportionment step for particulate matter with aerodynamic diameter ¿ 2.5 ¿m (PM2.5) attribution based on molecular organic markers. Given that PLS is an unweighted uncertainty model and that the uncertainties for organic species are not easy to quantify, Monte Carlo sampling (MC-S) was performed as a complement to PLS to assess the variation of source contributions on the total PM2.5. These methods were performed using 43 samples collected during the spring and fall season monitoring campaigns. In general, source contributions obtained from the MC-S results were not statistically different from the PLS original results, and uncertainties in ambient measurements did not have a significant influence on the source contributions. Finally, the source contributions were compared with those obtained by positive matrix factorization and chemical mass balance showing insignificant differences. The PLS and MC-S fit well as preliminary and complementary tools for source apportionment based on organic molecular markers.