Volume 1, No. 1, June 2001, Pages 9-20 PDF(1.2 MB)
The Elemental Contents and Fractal Dimensions of PM2.5 in Taipei City
Jin-Lan Chen1, Chung-Te Lee1, Shyh-Yuh Chang1, Charles C. K. Chou2
1 Graduate Institute of Environmental Engineering, National Central University, Chungli, Taiwan, ROC
2 Center for Environmental, Safety and Health Technology, Industrial Technology Research Institute, Hsinchu, Taiwan ROC
Five sites (i.e. Sung-Shan, Wen-Lin, Gu-Ting, Wan-Hwa, and Yen-Ping) located in Taipei City were selected for simultaneously collecting PM2.5 using honeycomb denuders (Koutrakis et al., 1993). The collected filters were subsequently analyzed with a Computer Controlled Scanning Electron Microscope (CCSEM) (RJ Lee Group, Inc., PA, USA) to obtain aerosol elemental compositions. Multivariate statistics such as Cluster Analysis and Principal Component Analysis (PCA) were applied to the resolved elemental compositions for source apportionment. Additionally, aerosol morphology from selected aerosols was quantitatively analyzed using fractal geometry to reveal differences among source types. The measurement results display that predominant particles are either pure carbon or contain a high proportion of carbon. For particles associated with a high percentage of carbon, 63.7% of 9595 particles were found for the first season (fall), 71.3% of 5987 for the second season (winter), and an even higher 82.9% of 9000 for the third season (spring). Cluster analysis for the total amount of aerosols collected reveals 27 source types. Among the resolved source types, 73.3% of 24579 particles were from motor vehicles, 12.4% were from soil dusts, and the remainders were from industrial process, secondary reactions, and sea spraying. In contrast, only 9 source categories were resolved using PCA method, compared to the 27 identified from cluster analysis. Although the apportioned source types of aerosol types of aerosols identified using the cluster method can be lumped together for comparison with PCA, cluster analysis indicates a higher resolution in the source apportionment procedure. A discriminant function is formulated from the resolved source types of the first two seasons to verity the source apportionment using the cluster procedure. The discriminant function analysis demonstrated that 123 out of 1535 particles were misjudged (excluding high percentage carbon particles). The calculation of fractal dimension on the morphology of 15 selected particles illustrates that particles are differentiated from each other based on kinky periphery and compact surface.
Source apportionment; Single particles; Statistical methods; CCSEM; Fractal dimension of aerosols.