Technical Assistance to MARAMA on EPA Cumulative Exposure Project (CEP): Review of Air Toxics Modeling, January 1999, P.T. Roberts, H.H. Main, S.B. Hurwitt, Sonoma Technology, Inc.
This report prepared for MARAMA by Sonoma Technology researchers is a review of toxics modeling conducted by EPA under the Cumulative Exposure Project, comparing model-estimated concentrations of various toxic air pollutants with ambient concentrations in Philadelphia, New Brunswick (NJ), and Baltimore. Ambient concentrations were for 1996, while the model estimates were for 1990 In addition, the ambient data is from PAMS monitoring sites, which only operate during the summer, while the model estimates are annual averages. Morning data was used for the comparisons, because a comparison of summer morning data to 24-hour, year round canister data for 8 HAPs at Philadelphia showed good agreement.
Sonoma identified three overall issues in reviewing EPA's modeling:
It is inappropriate to rely on model results at the census-tract level. The model is more reliable over larger areas.
The expected model performance and uncertainty of model estimates will vary widely for individual hazardous air pollutants (HAPs). HAPs vary in their emission, chemical reactivity, transport and dispersion, and background characteristics, and all these factors influence model performance.
EPA modeling guidance for dispersion models indicates that short-term concentrations should be within a factor of two of monitored values, and annual average model performance is expected to be better than short-term estimates. EPA's own model performance evaluation performed as part of the CEP compared available ambient data for 19 HAPs with model results and found that 68% of the comparisons were within a factor of three. Sonoma raises the question of whether this is acceptable performance.
Sonoma's comparison of monitored and modeled concentrations in Philadelphia, New Brunswick, and Baltimore led them to several conclusions, including those listed below. Of the HAPs investigated in this study, benzene, formaldehyde, and carbon tetrachloride are the most important with respect to cancer risk based on ambient data. Measured concentrations of these HAPs were well above cancer a risk benchmark of 1 in 1 million lifetime exposure risk.
Measured concentrations of species highly associated with motor vehicle emissions (e.g., benzene, toluene, and xylenes) were typically within a factor of two or three of the model's estimates. (The ambient concentration was usually lower than predicted by the model. This is consistent with recent studies showing that benzene has declined in the 1990s due to the use of reformulated gasoline.)
PAMS carbonyle compound concentrations were typically about twice those modeled. The model chemistry may be inadequate to reflect the secondary formation of carbonyles (e.g., formaldehyde).
Measured and modeled concentrations of methylchloride and carbon tetrachloride concentrations were similar, and were typically at about the level the model assumed as a background concentration. (Data only for Philadelphia.)
Several differences between modeled and monitored concentrations may indicate problems with emissions factors or activity data used to estimate emissions:
Ethylene dibromide was monitored at 20 times the level predicted by the model. The model may be missing key sources. (Data only for Philadelphia.)
Measured ethylene dichloride was at least an order of magnitude lower than modeled. Emissions estimates may be overstated. (Data only for Philadelphia.)
Ambient variability in concentrations is often much greater than reflected in the model.
Some pollutants are measured at very low concentrations, near the detection limits of monitoring equipment.
It should be noted that the modeling methods used at the time this report was prepared are not necessarily the same as modeling methods currently used by EPA.