Comparison and Uncertainty Analysis on Source Identification Methods

Factor analysis/multivariate linear regression ((FA/MLR) and source profile/multivariate linear regression (SP/MLR) were used to investigate the sources of PAHs adsorbed on PM2.5 from Nanjing city. A total of 32 sets of data including the concentrations of 28 kinds of PAHs were utilized. The results from these two methods were compared. Uncertainty analysis was conducted on FA/MLR method using Monte Carlo method. The results showed that among the three and four pollution sources determined using FA/MLR and SP/MLR method respectively, only one pollution source was the same and the rest were all different. The uncertainty of FA/MLR method mainly came from the number of data used and the selection of tracer for each source. The number of pollution sources extracted was also different for the different number of data sets. In each calculation, the compounds with the largest contribution to factors extracted were not always the same. This research showed that large numbers of simulation were recommended to improve the reliability of the source identification results.