Characterization of volatile organic compound emission sources in Fort Saskatchewan,Alberta using principal component analysis

Volatile organic compounds (VOCs) were measured at two sites in a highly industrialized zone in western Canada from September 2004 to March 2006. Principal component analysis (PCA) with varimax rotation was performed on 30 VOCs to identify the pollution sources. Aliphatics, aromatics, and halogenated aliphatics were studied. The two monitoring sites were 11 km apart, with site 1 closer to the city of Fort Saskatchewan and site 2 predominantly down wind from the industrial sources. PCA results provided the basis for interpreting the relationship between the ambient 24-h integrated VOC samples and the emission sources in the region. Challenges existed in interpreting the PCA results in such a highly industrialized region; however a unique feature to this study was the fact that the region was home to the only 1,2-dichloroethane emitting facility in Canada. Other specific industry related VOCs in the region were vinyl chloride, styrene and HCFC-22. Making use of these specific VOCs in the PCA allowed for easy identification of an industrial contribution. For factors that were not easily distinguishable, further PCA tests were conducted using carbon monoxide concentrations, wind direction data and seasonal splitting of the samples. The analysis found that five factors accounted for 82% of the variance at site 1 and five factors accounted for 81% of the variance at site 2. The factor accounting for the highest variability (∼40%) at the two sites was the most difficult to interpret, but showed contributions from both industry and vehicle related emissions. Specific industrial sources were identified using 1,2-dichloroethane as a chemical tracer or by corroborating wind speed with known industry VOC emissions. Both sites had two factors identified as specific industry sources and these factors totaled to over 20% of the variance. Long range transport of stable halogenated compounds accounted for greater than 10% of the variance, and seasonal effects accounted for 5% of the variance.