The Analysis workshop session included atmospheric dispersion modeling, receptor modeling, exposure modeling, emissions modeling, and other quantitative methods that might be used to estimate emission rates or ambient concentrations of air toxics. Differing atmospheric dispersion modeling approaches are often characterized into micro-scale, meso-scale, and regional scale modeling methods. Modeling methods originally developed by EPA to assess criteria pollutant impacts have largely been adopted to address impacts from air toxics.
Micro-scale modeling methods have been applied to estimate ambient concentrations at local levels (e.g., downwind of specific roadway projects) up to urban scales (e.g., less than 50 kilometers). However, for the purposes of modeling the impacts from transportation projects, downwind distances are often in the range from 10's of meters to a few hundred meters. Micro-scale models have also been developed to assess impacts adjacent to traffic intersections. Examples of micro-scale models include EPA's Industrial Source Complex model (and its successor, AERMOD), CALINE3, CALINE4, CAL3QHC, and box models. All of the above except box models are Gaussian models.
Meso-scale models use Lagrangian or "puff" modeling methods to simulate the transport, decay, and transformation of pollutants in discrete air parcels (puffs) over distances of 10's to 100's of kilometers. For the purposes of assessing transportation impacts, meso-scale models would be used to assess impacts on a domain up to 10's of kilometers (urban scale air quality) often with the incorporation of other air pollution sources. CALPUFF is an example of a meso-scale model.
Regional scale models are based on Eulerian or "grid-based" methods. These models are used to estimate ambient concentrations at urban to regional and even continental scales. These models are also used for assessing regional impacts of mobile source control measures compared with those from other emission sources. Examples include EPA's Community Multi-Scale Air Quality model, the Urban Airshed Model, and the Regulatory Modeling System for Aerosols and Deposition (REMSAD).
Receptor models are used to estimate source contributions to ambient concentrations. Hence, in addition to detailed knowledge of the chemistry of measured ambient concentrations, detailed knowledge of the chemical composition of sources is needed. Exposure modeling is performed using either modeled or measured ambient air concentrations to estimate a receptor's exposure over a specified time-frame. Historically, 70-year continuous exposures have been used to estimate chronic and carcinogenic health risks to residential receptors. Recently, more realistic assessments of the duration of exposure have been incorporated into assessments, including multiple exposure scenarios (e.g., standing near a roadway, commuter inside a car, and living near a roadway).
Mobile source emission rates are generally estimated using EPA's MOBILE6 model or the California ARB's EMFAC model for on-road sources. Nonroad engine emissions are often modeled using EPA's NONROAD model or ARB's OFFROAD model. EPA is currently developing a next generation model referred to as MOVES, which will handle both on-road and nonroad emissions. MOVES will use modal emission estimation methods to incorporate the effects of speed, grade, and other factors (i.e., those affecting engine load). MOBILE6 uses average emission factors for different roadway (facility) types. Among the foreseen capabilities of the MOVES model are the ability to estimate emissions at varying spatial scales (e.g., roadway segment to regional level) and incorporating modules to estimate uncertainty.
Workshop participants identified research needs in the three following areas: air quality modeling, including atmospheric chemistry of MSATs; development of data to support modal emissions modeling methods; and receptor modeling. Underlying each of these three areas, variability/uncertainty analysis and model reconciliation were considered to be important areas for additional work. Model reconciliation relates to making comparisons of modeled results with ambient measurements to evaluate model performance. For example, data from the PAMS monitoring stations could be useful for this task since 1- and 3-hour averaged concentration data are available for some pollutants (i.e., benzene). Also, model results should also be compared with each other as appropriate (e.g., differing air quality modeling methods). Finally, participants noted that there is a need for systematic documentation at all phases of mobile source air quality impacts assessments (e.g., emissions modeling, air quality modeling, health risk assessment).
In addition to the three research areas above, participants noted that appropriate inputs for exposure assessments should be evaluated and further developed. Specifically, these include the development of the various exposure scenarios applicable to the assessment of transportation project impacts.
In the atmospheric dispersion modeling area, one of the more important research needs identified by workshop participants was in the atmospheric chemistry of toxic air pollutants, where little is currently known about the atmospheric behavior of some of the MSATs (e.g., acrolein, 1,3-butadiene). Certain species (e.g. acetaldehyde, formaldehyde) may be formed from precursors. Other reactive toxic species may degrade following chemical or photo-chemical reactions in the atmosphere (1,3-butadiene). Research is needed on these issues specific to MSATs and on how best to incorporate these reactions into micro- to -meso-scale analyses.
There is currently no standard protocol for the transportation community on which modeling methods to apply in specific situations. For example, a protocol could specify which models to apply to assess air quality impacts at different spatial scales. The areas of most interest are methods to apply in micro- (e.g., local roadway projects) to meso-scale (e.g., transportation corridor) studies. The protocol should also provide methods for quantifying uncertainty in the modeled concentrations from these studies. Given the direct association between emissions modeling, air dispersion modeling, and exposure modeling (described below), a protocol covering all of these areas would be beneficial.
Another common theme was the need to incorporate modal emissions modeling methods into standard practices to assess the effects of speed, acceleration, and road gradient. EPA has proposed using vehicle specific power (VSP) to characterize modal emission rates for the MOVES model (Koupal et al., 2003). The proposed light-duty VSP bin distributions in MOVES will be based on the MOBILE6 framework for facility-specific driving patterns. Another set of facility-based cycles will be produced for heavy-duty vehicles (both sets would include the 12 highway performance monitoring system roadway types). A general lack of toxic air pollutant emissions data to support modal emission estimation methods was noted (e.g., differing toxic air pollutant compositions for idling versus in-use heavy-duty diesel vehicles). As part of this, toxic air pollutant emission dependencies on driving cycles should be identified and characterized.
Also related to improved emissions modeling methods, toxic air pollutant emission profiles for high-emitters need to be incorporated into the new models. Until these new methods are available, there is a need to update the MOBILE6/MOBTOX emission factors derived from the COMPLEX model because they are outdated and inaccurate. In addition, there is a need to validate the air toxic emission factor estimates produced by these models.
Participants agreed on the need to data-mine sources of toxic air pollutant emissions data to develop these new methods or refine existing ones. The group pointed out that care must be taken while evaluating these data, as emission measurements are often collected using sampling techniques that may not be comparable with each other. It was noted that data-mining efforts were also likely to help identify toxic air pollutant surrogates that are needed due to the large expense associated with the measurement of multiple chemical compounds.
Workshop participants were divided on the issue of re-entrained road dust and its potential contribution to toxic air pollutant emissions. Some felt that the issue was highly uncertain and in need of study, while others felt that re-entrained road dust is typically deposited near the source and was therefore less important. The ARB is currently incorporating re-entrained road dust emissions into its detailed study of toxic air pollutant health risks in the Wilmington, California Air Quality Study (Sax, 2003). It should be noted that, in multi-media exposure assessment, both direct inhalation exposure as well as indirect exposure pathways are considered (e.g., deposition to soils or vegetation that can then be ingested, etc.).
Also, relative to emissions, accurate VMT data to support new generation emission models (i.e., based on modal methods) are needed. Essentially, the group recommended gathering and evaluating link-based VMT and characteristics (e.g., hourly speed, grade, altitude).
Analogous to issues discussed under the monitoring issues area, participants noted the need to identify source-specific marker compounds for use in source apportionment/receptor modeling studies. Unique marker compounds are used to apportion the contributions of toxic air pollutants to sources, as well as to differentiate the contributions of various mobile source sectors (e.g., light-duty gasoline and heavy-duty diesel) that are similar in emissions composition. It was also suggested that research be performed to determine if the marker compounds change under varying motor vehicle operating conditions.
Relative to receptor modeling, questions arose as to the accuracy of tools such as Positive Matrix Factorization that only use ambient data to estimate source contributions. There is a need to validate these results using source data (e.g., chemical mass balance or emissions and dispersion modeling). Reviews of existing data are also needed to refine PM and VOC speciation profiles, especially related to source apportionment efforts. It should be noted that EPA is currently funding work to update the SPECIATE database with many new profiles from data obtained from the California ARB, the Texas Commission on Environmental Quality, and research organizations.
Priority program areas for research in emissions and air quality analysis to support the transportation community's need for air toxics modeling tools are described below. These four priority research areas cover emissions and air quality (atmospheric dispersion) modeling; developing methods on how to perform and interpret on-road vehicle emissions uncertainty analyses; developing improved inputs for emissions and receptor modeling of toxic air pollutants; and performing research on the atmospheric chemistry of MSATs.
Proposed Programmatic Initiatives
Programmatic Initiative P7: Development of a Protocol on Emissions, Atmospheric Dispersion, and Exposure Modeling for Transportation Projects.
A protocol that provides the appropriate methods to use to estimate emissions, perform dispersion modeling, and conduct exposure assessments is needed. The focus of this protocol should be on project-level analyses (e.g., roadway improvement projects, intersections), although it should also include methods to assess impacts at meso-scales (e.g., transportation corridors). The protocol should address which air dispersion models to use at varying spatial scales (i.e., what is the spatial point-of-departure from micro-scale methods). An exploration of the use of GIS tools to assess exposure to project-level impacts would also be useful (e.g., GIS data sources and methods to characterize the population adjacent to roadway projects).
Due to the inter-dependence of exposure assessment with the selection of the appropriate emissions and dispersion models, they should all be addressed in a single document. Among other issues in emissions modeling, the protocol should cover alternatives to the use of emission factors from MOBILE6/EMFAC that may be based on trip averages or facility averages in micro-scale applications (e.g., intersections or off-ramps where emissions from stops and starts are much different than represented by the underlying average emission factor). As part of this effort an assessment will be needed as to how well EPA's upcoming MOVES model will satisfy the transportation community's needs, particularly for micro-scale impacts analysis.
The protocol should also include a plan for phased implementation. This should include descriptions of how existing models can be used for micro-scale analysis, and how standard adjustments can be applied to allow modeling results for relatively inert gases like CO to be used to provide ballpark estimates of air toxic concentrations. If new or revised models are needed, the protocol should allow for phased implementation, with descriptions of needed computational and data resources provided.
Estimated Cost: $350,000
Estimated Duration: 12 - 18 months
Programmatic Initiative P8: Conduct Research on Appropriate Methods to Estimate Variability and Uncertainty of On-Road Emission Estimates. Identification of key sources of uncertainty can be used to target resources to reduce uncertainty.
There is a need to develop systematic estimates of uncertainty in mobile source emission inventories. However, the focus of this work needs to be determined. For example, should these efforts focus on establishing uncertainty estimates for new generation models (e.g., modal emissions models) or for existing emissions models (e.g., MOBILE6, EMFAC)? For MOVES, FHWA should partner with EPA to develop uncertainty estimation methods during the development of MOVES inputs (e.g., distributions of each parameter used in estimating emissions). Hence, the first step under this program area is to develop a needs assessment for uncertainty analysis. It is also expected that an expert workshop would be convened in order to incorporate expert input into the process.
Estimated Cost: $250,000
Estimated Duration: 24 - 30 months
Programmatic Initiative P9: Develop Improved Inputs for Emissions and Receptor Modeling
Relative to receptor modeling, research should be conducted and data mining should take place to identify the organic marker compounds that can be used to characterize mobile source contributions (especially the contributions of on-road diesel and light-duty gasoline vehicles). Using these marker compounds, different source profiles may be needed based on different operating modes (e.g., idling versus in-use heavy-duty diesel). Another objective of this program area is to define where additional testing is needed to fill the observed data gaps.
Another set of inputs that requires additional study is the toxic air pollutant composition of re-entrained road dust and brake/tire wear. There is likely not a lot of information in the literature in this area, so testing programs will need to be developed.
Estimated Cost: $750,000
Estimated Duration: 24 months
Programmatic Initiative P10: Research the Atmospheric Chemistry of MSATs
In this program area, a review of the atmospheric chemistry of MSATs will be performed. Sources of information include the scientific literature, work supporting EPA's National Air Toxics Assessment project, and ongoing work at EPA to incorporate toxic air pollutant chemical mechanisms into air quality models (e.g., Community Multi-Scale Air Quality). The objective would be to develop recommendations for incorporating formation/degradation/condensation algorithms or other adjustments (deposition velocities for particulate toxic air pollutants) into standard micro-scale and meso-scale models that are important tools in assessing transportation project impacts. One product of this research would be to provide the information needed to develop a better line source model for the purpose of evaluating transportation projects for near roadway dispersion of air toxics and PM. This model should be able to simulate secondary formation of air toxics and PM. The results of this work could be incorporated into the protocol described under Programmatic Initiative A1 above.
Estimated Cost: $250,000
Estimated Duration: 12 months