2005 NATA: Assessment Methods
This section contains brief descriptions of the methods used to implement various sections of the 2005 NATA. More detailed information may be found in the Technical Methods Document (TMD). The methods discussed are:
- Preparation of Emission Estimates
- Development of Background Concentrations
- Approach for Modeling POM
- Emissions Modeling with ASPEN, HEM, and CMAQ
- HAPEM5 Modeling
- Estimating Exposure Concentrations
- Risk and Hazard Characterization
- Assessment Uncertainty and Variability
The 2005 National Emissions Inventory is the underlying basis for the emissions used in the national-scale assessment. The NATA emissions inventory uses the NEI as the starting point but changes may be made in the inventory based on updated information and different data sources. Examples of these types of changes include:
- Adjustments to emission rates based on recent analyses conducted for other EPA programs and projects, such as the Risk and Technology Review and the development of emission standards for area source categories;
- Corrections and updates implemented as a result of the reviews and quality assurances of draft versions of NATA; and
- Updates to mobile source emissions for some chemicals based on the results of analyses using the Motor Vehicle Emissions Simulator (MOVES) model.
Other differences reflect the specific role and function of the resulting inventory within the context of the NATA risk assessment process and should be more accurately described as post-processing procedures rather than substantive changes. Examples of these types of changes include:
- Chemical name conversions, placing individual chemicals into groups, and similar transcription and phraseology conversions, e.g., for the purpose of cross-walking the identity of an emitted pollutant to a substance with a quantitative dose-response value;
- Adjustments to emission rates of metal compounds based on the toxic metal proportion of the compound's molecular weight, e.g., only the mass fraction of arsenic trioxide that consists of arsenic; and
- Speciation of some metal compounds into individual chemicals based on toxicity associated with a certain valence, particle size, or chemical reactivity.
The types of emissions sources in this inventory include major stationary sources (e.g., large waste incinerators and factories); area and other sources (e.g., dry cleaners, small manufacturers); and both onroad and nonroad mobile sources (e.g., cars, trucks, boats). In preparing the inventory for modeling, EPA used default, or simplifying assumptions where data were missing or of poor quality. For example, when data on stack height or facility location were not available or were flawed, they were replaced by default assumptions (e.g., a stack height for a facility might be set equal to stack heights at comparable facilities); the location of the facility might be placed at the center of a census tract; etc. A few changes in the 2005 NEI include the addition of 'airports' as point sources.
EPA compiled the 2005 National Emissions Inventory (NEI) using a variety of sources. These sources include:
- State and local toxic air pollutant inventories (developed by State and local air pollution control agencies)
- Existing databases related to EPA's air toxics regulatory program
- EPA's Toxic Release Inventory (TRI) database
- Estimates developed using mobile source methodologies (developed by EPA's Office of Transportation and Air Quality)
- Emission estimates generated from emission factors and activity data
- Revisions to source inventories made in response to the various Risk and Technology Review requests, where agencies and industries can provide extensive reviews of emissions inventories and screening-level analyses
- Emissions estimates generated as part of EPA's analyses supporting the development of standards to control emissions of air toxics from area sources
- The NATA review process
The models discussed below do not estimate outdoor concentrations of air pollutants attributable to long-range pollutant transport, unidentified emission sources, and natural emission sources. These "background" contributions can be significant for some air toxics and should be accounted for to accurately estimate ambient concentrations. For NATA, EPA uses background concentrations to represent the contributions to ambient concentrations of air toxics resulting from natural sources, emissions of persistent chemicals that occurred in previous years, and long-range transport from distant sources. These background concentrations are intended to represent levels of pollutants found in a particular year even if there had been no local anthropogenic emissions of those pollutants during that year.
Three methods were used to estimate background concentrations for NATA. The ambient method uses available monitoring data, the emissions method uses NEI emissions data, and the uniform method assumes a uniform nationwide concentration for the pollutant. The ambient-based method is preferred because the background estimates are based on measured air toxics concentrations throughout the United States. Background concentrations were estimated for 14 HAPS using this method. However, reliable ambient measurements are not always available for every pollutant of interest. Background estimates based on the available ambient data for these pollutants could either have too few sites to extrapolate from or poor quality measurements on which to base background estimates. Therefore, an emissions-based method was developed to handle those pollutants with inadequate ambient measurements. Background concentrations were estimated for 16 HAPs using this method. In addition, a few pollutants were assigned uniform spatial concentrations based on their long lifetimes and well-characterized concentrations. These pollutants are carbon tetrachloride, methyl chloride, methyl bromide, and methyl chloroform. All are routinely measured at remote sites and have well-mixed concentrations in the Northern Hemisphere. More complete discussions of this may be found in the Section 2.5 of the TMD.
Not all POM reported to EPA's national emission inventory (NEI) are speciated into individual compounds. As a result, we apply some simplifying assumptions in order to model and assess the risk from the individual pollutants that comprise POM. This involves establishing different POM "groups" and modeling them as separate pollutants. In establishing these groups, we considered the need to provide the most detailed information about risks from the individual pollutants within POM while taking into account the inconsistencies in how much speciation is reported in the inventory. A more refined method to establish POM groups was used for the 1999, 2002, and 2005 NATAs than was used for the 1996 NATA. The newer methods reflect improvements in the speciation of POM within the inventory. Specifically, for the 1996 NATA we had two overlapping POM groups, and for the 1999, 2002, and 2005 NATAs, we have eight distinct groups. See the Section 5.4.2 of the TMD document for more detailed discussion.
For the 2005 NATA, three computer models were used:
- Human Exposure Model-3 (HEM-3), AERMOD Version;
- Assessment System for Population Exposure Nationwide (ASPEN); and
- Community Multiscale Air Quality (CMAQ).
The HEM - AERMOD was used to model emissions from point, on-road and non-road mobile sources. ASPEN was used for the non-point sources, and CMAQ was used to estimate the secondary formation concentrations of 4 air toxics. The 2005 NATA emission estimates were used as input to EPA's air dispersion models. The Emissions Modeling System for Hazardous Pollutants (EMS-HAP) was used to process the emission inputs for modeling.
The non-point sources modeled by ASPEN are associated with a census tract and are not situated at specific locations. Therefore, ASPEN treats each source as a "pseudo-point" source located at the centroid of the census tract where it is located (the sources resident tract). ASPEN estimates ambient concentrations for a pre-set receptor grid, and then interpolates ambient concentrations from the grid receptors to each census-tract centroid outside of the source's resident tract (the sources non-resident tract). To estimate the average concentration for a resident tract, ASPEN represents the area source for a tract as multiple pseudo-point sources geographically dispersed throughout the tract, rather than as a single source. Ambient concentrations in the resident census tract are estimated with spatial averaging of the ambient concentrations at all grid receptors that fall within the bounds of the tract. When these resident tract and non-resident-tract concentrations are calculated for all sources, the concentrations are summed for each tract.
The ASPEN model takes into account important determinants of pollutant concentrations, such as:
- The rates of emission releases
- The location of these releases
- The height from which the pollutants are released
- The wind speeds and directions from the meteorological stations nearest to the release
- The breakdown of the pollutants in the atmosphere after being released (i.e., reactive decay)
- The deposition or settling of pollutants out of the atmosphere, and
- The atmospheric transformation of one pollutant into another (i.e., secondary formation)
More information about the ASPEN model is available at the SCRAM website.
Stationary source (point source) emissions were modeled using the EPA's Human Exposure Model - AERMOD (HEM-AERMOD). This model is used primarily for performing risk assessments for major point sources (usually producers or large users of specified chemicals) of air toxics. Like ASPEN, HEM-AERMOD only addresses the inhalation pathway of exposure, and is designed to predict risks associated with emitted chemicals in the ambient air (i.e., in the vicinity of an emitting facility but beyond the facility's property boundary). The HEM-AERMOD used for NATA 2005 contains: 1. The AERMOD, an atmospheric dispersion model, with included meteorological data, and 2. U.S. Bureau of Census population data at the census block level. Census blocks are the smallest spatial area modeled in NATA. Census blocks contain approximately 40 people, but the number of people may vary in size depending on where the blocks are located.
Both the ASPEN and HEM-AERMOD models utilize the 2000 Census data. Each source must be specifically located by latitude and longitude, and its release parameters must be described. These include stack height, exit velocity, emission rate, etc. Based on the inputs for source parameters and the meteorological data, the model estimates the magnitude and distribution of ambient air concentrations in the vicinity of each source (at the census block or tract levels depending on which model was used). These models generally are used to estimate ambient concentrations within a radial distance of 50 kilometers (30.8 miles) from the source.
A third model, the Community Multiscale Air Quality Model (CMAQ), was used to model the secondary formations of formaldehyde, acetaldehyde, acrolein, and the decay of 1,3-butadiene into acrolein. "Secondary formation" occurs when an emitted substance chemically transforms in the atmosphere to become another (i.e., secondary substance). CMAQ is a multi-pollutant, multi-scale air quality model that simulates the atmospheric and land processes affecting the transport, transformation, and deposition of air pollutants and their precursors on large and small scales. It is intended to holistically consider major pollutant issues, such as photochemical oxidants, particulate matter, acidic deposition, and nutrient deposition. CMAQ calculates ambient concentrations within model grid cells, with detailed treatment of atmospheric chemistry and physics. For NATA, these grid cell concentrations are interpolated to census-tract centroids. The secondary air toxic concentrations from CMAQ are presented in the results under the "secondary formation" heading. This separation shows the contribution of atmospheric transformation to the total cancer risk or noncancer hazard results compared to the other sources modeled.
The concentrations estimated by these models are ambient air concentrations in micrograms per cubic meter. Ambient concentration estimates are actually surrogates for exposure, as important exposure variables (e.g., duration, human activity patterns, residential occupancy period, etc.), are not explicitly addressed by these models. The ambient air concentrations were then used as inputs to the exposure model, HAPEM5, which is described below. More information about these models and the process for generating ambient concentration data may be found in Section 3 of the TMD.
The outputs from the models (ASPEN, HEM-AERMOD, and CMAQ) were used as input to the screening-level inhalation exposure model, HAPEM5. Estimating exposure is a key step in determining potential health risk. People move around from one location to another (e.g., outside to inside; or, commuting to work, etc.) Exposure isn't the same as concentration at a static site. People also breathe at different rates depending on their activity levels. For these reasons, the average concentration of a pollutant that people breathe (i.e., exposure concentration), may be significantly higher or lower than the concentration at a fixed location.
HAPEM uses a general approach of tracking representative individuals of specified demographic groups as they move among indoor and outdoor microenvironments and between geographic locations. Personal activity and commuting data specific to individual demographic groups are used to determine the census tracts containing residential and work locations and the microenvironments within each tract. Empirically based factors reflecting the relationship between exposure concentrations within each microenvironment and the outdoor (ambient) air concentrations at that location are selected by the model through a stochastic sampling process to estimate exposure concentrations. Through a series of calculation routines, the model makes use of census data, human activity pattern data, ambient air quality levels, climate data, and indoor/outdoor concentration relationships to estimate an expected range of "apparent" inhalation exposure concentrations for groups of individuals.
To simulate long-term exposures,, the estimated pollutant concentrations in each microenvironment visited are combined into an average concentration weighted by daily activity patterns, and this time-weighted average is assigned to members of a demographic group. From this set of model estimates, a single population-weighted, median exposure concentration, which represents the best estimate of exposure for a "typical" person for a given census tract, can be estimated. In this case, "typical" does not refer to a specific individual in the population or even the average over a group of individuals. Rather, this person is a hypothetical individual residing at the centroid of a census tract and engaging in a range of activities (both indoor and outdoor) that are representative of those in which individuals in that census tract might engage. For more information about the model, see the HAPEM5 Users Guide. For more information about the method, see Section 4 of the TMD.
For the 2002 and 2005 NATAs, exposure level concentrations for each source (i.e., stationary, area, mobile, background, and secondary formation (2005 NATA only)) were not estimated by direct modeling in HAPEM. Instead the ambient level concentrations for each source (i.e., the outputs from the dispersion models), were multiplied by a ratio (or tract-level exposure factor) of the estimated HAP-specific exposure concentrations to HAP-specific ambient concentrations (HAPEM output / ASPEN output for the same HAP, source, and census tract) that were developed from the 1999 HAPEM5 modeling. When census block level ambient concentrations (stationary sources were initially modeled at the census block level) were being converted to tract-level exposure concentrations using the tract-level exposure factors, the exposure factors were applied to each census block equally. As described below, census block-level concentrations were aggregated up to the census tract-level which is the level presented for all HAPs in the results section of the 2005 NATA. See the following section for a brief description of the aggregation method. These exposure factors from the 1999 NATA assessment were used to save modeling time and because the demographic distributions within the census data were the same (i.e., all assessments since 1999 used the 2000 Census data as the source of census tract demographic data).
The risk and hazard quotients for each toxic air pollutant modeled is estimated from the exposure concentrations (not ambient concentrations) by combining the exposure concentration estimates with available unit risk estimates and inhalation reference concentrations, respectively. More details about the dose-responses used are found in the document Health Effects Information (PDF) (12pp, 220k) and in Section 5 of the TMD.
As described above, ASPEN or HEM modeling produced census tract-level or census block-level ambient concentration estimates, respectively, for each HAP. To get all concentrations at the same spatial level, tract-level concentrations were divided evenly across all census blocks in the tract, thereby creating a data set of census block-level ambient concentrations for all census blocks in the country. These sets of data while not presented in the results section of the NATA 2005 website (these block-level files are very large), were used as the starting point for estimating cancer and noncancer effects, and for aggregating up to larger spatial scales (i.e., tract, county, state, and US levels).
- Block level ambient concentrations were aggregated up to the tract level by taking the average of the block level concentrations and applying it to the tract.
- Block level ambient concentrations were aggregated up to county, state, or US level by using the population-weighted method described below.
The ambient block-level concentrations were also used to estimate exposure concentrations using the exposure factors (HAPEM/ASPEN ratios) described above. Since the exposure factors from the 1999 NATA were at the tract-level, each census block was assigned the tract-level factor and census block-level exposure concentrations were estimated. As was done with the ambient-level concentrations above, the block-level exposure concentrations were used to estimate cancer and noncancer effects and to aggregate these concentrations up to larger spatial scales. To aggregate tract-level concentrations up to county, State, or US level concentrations, the tract-level concentrations were population-weighted. For example, census tract exposure concentration estimates were multiplied by the population of each tract, summed, and divided by their county populations to get a final population-weighted county level concentration. The formula for this is:
concentrationc = county-level population weighted concentration
concentrationt = tract concentration
populationt = tract population
populationcty = county population
The same method would be applied when aggregating up to a state level (or higher), except that the populations used would be for the state (or higher) level.
The 2005 NATA results section presents ambient and exposure concentrations, cancer and noncancer risk at the tract, county, and state levels. The noncancer results from previous NATA assessments have shown that the respiratory and neurologic endpoints were the drivers of noncancer hazards. In the NATA 2005, all noncancer results (i.e., other endpoint results), are presented as well but the summary focus, as in previous assessments, is on the neurologic and respiratory endpoints. A more complete discussion may be found in Section 6 of the TMD.
A discussion of the components of variability and uncertainty will help understand why it's important to use the NATA results to answer only those questions for which the assessment is suitable. Only a brief overview is provided here, but a more complete discussion of variability and uncertainty is found in Section 7 of the TMD.
Variability: Emissions, air concentrations, exposures and risks are not the same throughout the U.S., and are not the same for everyone. Some geographic areas have higher concentrations than others, and there are some periods of time when the concentration is higher at a given location than at other times. Some individuals have an exposure and/or risk below the national average, while others have an exposure and/or risk above the national average. For these reasons, it is necessary to have some idea of how the ambient air concentration, exposure, and risk from HAPs varies throughout the U.S.
Uncertainty: The EPA seeks to protect health with reasonable confidence. But scientific estimates of air concentrations, exposures and risks always involve assumptions that simplify the real situation but make the assessment possible given available information and resources. These assumptions introduce uncertainties into the results, since there is never complete confidence that the assumptions are entirely correct. It is necessary to understand the size of these uncertainties, the level of confidence that can be placed on any statement made about the assessment, and how this confidence affects the ability to make reasoned decisions.