Combined Use of Principal Component Analysis (PCA) and Chemical Mass Balance (CMB) for Source Identification and Source Apportionment in Air Pollution Modeling Studies

Chemical mass balance (CMB) and principal component analysis (PCA) are used together for source identification and source apportionment in this air pollution modeling study. Source profile sets, each of which contains five source profiles based on ten pollutant species, were generated using a computer program. Another algorithm was implemented to produce ten random data sets, which was composed of 100 simulated measurement results for all of ten pollutant species. Ten source profile sets were selected. Five of them contained sources of dissimilar characteristics, whereas the other five were chosen from those of similar emission profiles. Ten simulated data sets for each source profile set were used in the analyses. PCA was applied to all simulated data sets; a number of principal factors were extracted and interpreted. The identified sources for each data set were used in fitting with CMB analyses, and source contributions were estimated. The performance of PCA–CMB combination was evaluated in the aspect of percent variance explained, percent apportionment, R ², and χ ². PCA was able to explain 89.6% to 100% of the variance within the data sets used. Two to five sources were extracted depending on the characteristics of source profile sets used. CMB was found to be successful in the aspect of percent apportionment since 95.4% to 100% of mass concentrations were apportioned. The values of R ² and χ ² were found out to range from 0.981 to 1.000 and from 0.000 to 29.947, respectively. Evaluating overall results from the analyses, PCA–CMB combination produced satisfactory results in the aspect of source identification and source apportionment.