Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4439279 | Atmospheric Environment | 2012 | 11 Pages |
Coarse particles exposures are expected to be highly heterogeneous in an urban area. However, little data are available to explore the extent of heterogeneity of coarse particles, especially on a local scale. An extensive sampling program for the coarse particles was conducted using University of North Carolina (UNC) passive aerosol samplers. The samplers were deployed for 4–5 week periods during four seasons, fall, winter, spring, and summer at 25 different sites across Syracuse, a small city located in central New York. The substrates from the UNC passive samplers were analyzed with computer-controlled scanning electron microscopy (CCSEM) providing size, shape, and elemental composition in the form of fluoresced X-ray spectra. Adaptive resonance theory (ART-2a) based neural network algorithm was applied with processed X-ray data to identify homogenous particles classes of 25,437 coarse particles from all four seasons. Thirty-four particle classes were identified with similar chemical characteristics. The mass fractions of particles in each identified class were then used to assess the homogeneity of composition and concentration across the measurement domain for each season. Road/soil dust, carbonaceous dust, biological materials, and deicing road salt were identified as the major sources of the urban coarse particles. Spatial and seasonal variations in both composition and concentration were observed and a noticeable heterogeneity between adjacent sites is indicated by the coefficient of divergence and correlation coefficient analysis.
► UNC Passive Sampler deployed at 25 locations in 4 seasons in Syracuse, NY. ► Coarse particles characterized using computer-controlled scanning electron microscopy. ► Particle classes identified using an adaptive theory resonance neural network. ► Spatial and temporal patterns of coarse particle mass and composition were determined. ► Heterogeneity of particle data was assessed using coefficient of divergence analysis.