Two factor analysis (FA)-based receptor modeling methods were applied to a polycyclic aromatic hydrocarbon (PAH) dataset from extracts of 75 PM(10) air particulate samples collected concurrently at 4 sampling sites proximate to the urban-industrial area in Hamilton, Ontario, Canada. The total PAH concentrations of 48 target compounds ranged from 0.23 to 172 ng m(-3). Principal component analysis (PCA) and positive matrix factorization (PMF) analysis were followed by multilinear regression analyses to identify and quantify PAH source contributions, together with spatial and temporal trends. The correlations between predicted and observed total PAH levels were excellent in both models (R(2) > 0.98). The PCA afforded large negative contributions in a number of samples, so further analysis was abandoned. The PMF analysis showed 3 factors which were identified as gasoline emissions, diesel emissions and coke oven emissions. Contributions of gasoline emissions and diesel emissions factors were surprisingly similar at all 4 sites indicative of a background of vehicle emissions across the city. The PMF coke oven emission factor showed the greatest variability in total loadings, consistent with the large PAH emissions from the steel industries and the large influence of wind direction on PAH concentrations. The highest coke oven contributions were observed at sites closest to the industrial area on days when these sites were downwind of the industries. The PMF coke oven impact factor showed good correlations with two commonly used PAH diagnostic ratios when the ratios were combined into a single ratio. This integrated approach allowed us to categorize >90% of the samples based on the wind direction of the impacting source.
Read full abstract