Relative Congener Scaling Of Polychlorinated Dibenzo-P-Dioxins And Dibenzofurans To Estimate Building Fire Contributions In Air, Surface Wipes, And Dust Samples
Notice - This site contains archived material(s)
Archive disclaimer
Archive
disclaimer
Archived files are provided for reference
purposes only. These files are no longer maintained by the Agency and may be outdated. For
current EPA information, go to www.epa.gov. It is EPA's policy to
support reasonable accommodation to persons with disabilities, pursuant to the Rehabilitation
Act of 1973, 29 U.S.C. 791. If you need assistance with accessing archived files, contact
EPA's Reasonable Accommodations
or submit a request using the Contact Us form.
Abstract
EPA collected ambient air samples in lower Manhattan for about nine months following the September 11, 2001 (9/11) World Trade Center (WTC) attacks. Measurements were made of a host of airborne contaminants including volatile organic compounds (VOCs), polycyclic aromatic hydrocarbons (PAHs), asbestos, lead, and other contaminants of concern. The present study focuses on the broad class of dioxins and furans with specific emphasis on seventeen congeners that exhibit mammalian toxicity. Samples taken shortly after 9/11 through the end of October, 2001, contained some of the highest levels of dioxins ever measured in ambient air.
Concentrations ranged as high as 200 pg toxic equivalents (TEQ)/m3 greatly exceeding typical urban background levels of 0.1 pg TEQ/m3. A subset of 29 samples all taken before October 31, 2001, were treated as a basis set known to be heavily impacted by the WTC building fire source. A second basis set was created using data from Los Angeles and Oakland, CA as published by the California Air Resources Board (CARB) and treated as the archetypical background pattern for dioxins. The CARB data had a congener profile appearing similar to background air samples from different locations in America and around the world, and in different matrices, such as background soils. Such disparate data would normally be interpreted with qualitative pattern recognition based on congener bar graphs or other forms of factor or cluster analysis that group similar samples together graphically.
The procedure developed here employs aspects of those statistical methods to develop a single continuous output variable per sample. Specifically, a form of variance structure-based cluster analysis is used to group congeners within samples to reduce co-linearity in the basis sets, new variables are created based on these groups, and multivariate regression is applied to the reduced variable set to determine a predictive equation. This predictive equation is designed to discriminate between background samples and samples influenced by the WTC building fire; it is assigned a nominal value of zero (0.00) for a background congener profile and one (1.00) for the profile characterized by the WTC air profile. Although this empirical method is calibrated with relatively small sets of airborne samples, it is shown to be generalizable to other WTC, fire source, and background air samples as well as other sample matrices including soils, window films and other dust wipes, and bulk dusts. Specifically, when applied to other WTC-related and fire-related data sets, it yields values near 1.00, and when applied to other background data sets, it yields values closer to 0.00.
Concentrations ranged as high as 200 pg toxic equivalents (TEQ)/m3 greatly exceeding typical urban background levels of 0.1 pg TEQ/m3. A subset of 29 samples all taken before October 31, 2001, were treated as a basis set known to be heavily impacted by the WTC building fire source. A second basis set was created using data from Los Angeles and Oakland, CA as published by the California Air Resources Board (CARB) and treated as the archetypical background pattern for dioxins. The CARB data had a congener profile appearing similar to background air samples from different locations in America and around the world, and in different matrices, such as background soils. Such disparate data would normally be interpreted with qualitative pattern recognition based on congener bar graphs or other forms of factor or cluster analysis that group similar samples together graphically.
The procedure developed here employs aspects of those statistical methods to develop a single continuous output variable per sample. Specifically, a form of variance structure-based cluster analysis is used to group congeners within samples to reduce co-linearity in the basis sets, new variables are created based on these groups, and multivariate regression is applied to the reduced variable set to determine a predictive equation. This predictive equation is designed to discriminate between background samples and samples influenced by the WTC building fire; it is assigned a nominal value of zero (0.00) for a background congener profile and one (1.00) for the profile characterized by the WTC air profile. Although this empirical method is calibrated with relatively small sets of airborne samples, it is shown to be generalizable to other WTC, fire source, and background air samples as well as other sample matrices including soils, window films and other dust wipes, and bulk dusts. Specifically, when applied to other WTC-related and fire-related data sets, it yields values near 1.00, and when applied to other background data sets, it yields values closer to 0.00.
Citation
U.S. EPA. Relative Congener Scaling Of Polychlorinated Dibenzo-P-Dioxins And Dibenzofurans To Estimate Building Fire Contributions In Air, Surface Wipes, And Dust Samples. Joachim D. Pleil and Matthew N. Lorber (ed.), Submitted to: ENVIRONMENTAL SCIENCE AND TECHNOLOGY. John Wiley & Sons, Ltd., Indianapolis, IN, 41(21): 7286 -7293, (2007).
