Inhalation Exposure and Risk from Mobile Source Air Toxics in Future Years

www.nature.com/jes

Inhalation exposure and risk from mobile source air toxics in future years

RICHARD COOKa, MADELEINE STRUMb, JAWAD S. TOUMAc,TEDPALMAb, JAMES THURMANd, DARRELL ENSLEYd AND ROY SMITHb aOfﬁce of Transportation and Air Quality, U.S. Environmental Protection Agency, Ann Arbor, MI, USA bOfﬁce of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC, USA cNational Oceanic and Atmospheric Administration/Atmospheric Sciences Modeling Division (In Partnership with the U.S. Environmental Protection Agency), Research Triangle Park, NC, USA dComputer Sciences Corporation, Research Triangle Park, NC, USA

Modeling ofinhalation exposure and risks resulting from exposure to mobile source air toxics can be used to evaluate impacts ofreductions from contr ol programs on overall risk, as well as changes in relative contributions of different source sectors to risk, changes in contributions of different pollutants to overall risk, and changes in geographic distributions of risk. Such analysis is useful in setting regulatory priorities, and informing the decision-making process. In this paper, we have conducted national-scale air quality, exposure, and risk modeling for the US in the years 2015, 2020, and 2030, using similar tools and methods as the 1999 National-Scale Air Toxics Assessment. Our results suggest that US Environmental Protection Agency emission control programs will substantially reduce average inhalation cancer risks and potential noncancer health risks from exposure to mobile source air toxics. However, cancer risk and noncancer hazard due to inhalation ofair toxics will continue to be a public health concern. Journal of Exposure Science and Environmental Epidemiology (2007) 17, 95–105. doi:10.1038/sj.jes.7500529; published online 27 September 2006

Keywords: exposure modeling, inhalation exposure, risk, air toxics, mobile sources.

Introduction matter and diesel exhaust organic gases (U.S. EPA, 2006a). Benzene and 1,3-butadiene are both known human carcino- Air toxics, which are also known in the United States’ Clean gens (U.S. EPA, 1998, 2002a). Air Act as ‘‘hazardous air pollutants’’ (HAPs), are those In recent years, the US Environmental Protection Agency pollutants known or expected to cause cancer or other serious (US EPA) as well as State and local agencies have released a health and environmental effects. For example, some of these number ofrisk characterizations addressing potential inhala- pollutants are known to have negative effects on human tion cancer risks from exposure to ambient sources of air respiratory, cardiovascular, neurological, immune, reproduc- toxics. Some studies estimate cancer risk assuming that tive, or other organ systems, and they may also have individual exposures are equivalent to average ambient developmental effects. They may pose particular hazards to concentrations in a given area (e.g. South Coast Air Quality more sensitive populations, such as children, the elderly, or Management District, 2000). Other studies attempt to people with pre-existing illnesses. estimate risk based on modeled exposures, such as US Mobile source air toxics are emitted by motor vehicles, EPA’s National-Scale Air Toxics Assessment, or NATA nonroad engines (such as lawn and garden equipment, (U.S. EPA, 2002b, 2006b). The analyses presented in this farming and construction equipment, aircraft, locomotives, paper build on the framework developed for NATA. and ships) and their fuels. Air toxics are also emitted by There are four basic steps in the NATA for 1999: stationary sources such as power plants, factories, oil reﬁneries, dry cleaners, gas stations, and small manufac- (1) Compiling a national emission inventory ofair toxics turers. Some mobile source air toxics ofparticular concern emissions from outdoor sources. The 1999 National include benzene, 1,3-butadiene, formaldehyde, acrolein, Emissions Inventory is the underlying basis for the naphthalene, polycyclic organic matter, and diesel particulate emissions information in the 1999 assessment. (2) Estimating ambient concentrations based on emissions as input to an air dispersion model (the Assessment System 1. Address all correspondence to: R. Cook, Ofﬁce of Transportation and for Population Exposure Nationwide, or ASPEN model; Air Quality, US Environmental Protection Agency, 2000 Traverwood U.S. EPA, 2000). Drive, Ann Arbor, MI 48105, USA. Tel.: 1 734 214 4827. þ (3) Estimating population exposures based on a screening- Fax: þ 1 734 214 4939. E-mail: [email protected] Received 12 April 2006; accepted 31 August 2006; published online level inhalation exposure model (Hazardous Air Pollu- 27 September 2006 tant Exposure Model, version 5, or HAPEM5) and the Cook et al. Inhalation exposure and risk

estimated ambient concentrations (from the ASPEN time resulting from changes in emissions. For example, risk model) as input to the exposure model (Rosenbaum, estimates in US EPA’s recent 1999 National Scale Air Toxics 2005). Assessment assume individuals will be exposed to 1999 levels, (4) Characterizing 1999 potential public health risks due to on average, over a 70-year lifetime. Thus, they do not take inhalation ofair toxics from outdoor sources. This into account significant reductions in emissions that have includes cancer and noncancer effects, using available taken effect since 1999, future reductions anticipated from information on air toxics health effects, current US EPA mobile and stationary source control programs, facility risk assessment and risk characterization guidelines, and closures, industry initiatives and so on. They also do not estimated population exposures. consider potential increases in emissions due to population or economic growth. Such changes in emissions can have a large The models and data used for the NATA approach allow impact on lifetime risk (Cook et al., 2004a). wide geographic coverage and extensive pollutant coverage In this paper, we have conducted air quality, exposure and (nearly all the 188 HAPs listed in the Clean Air Act plus risk modeling for select mobile source air toxics for the years diesel particulate matter) but have inherent limitations. 2015, 2020, and 2030, using the same tools and methods as Among the significant limitations ofthe framework is that the 1999 National-Scale Air Toxics Assessment. Thus, our it cannot be used to identify ‘‘hot spots,’’ such as areas in results are comparable to the 1999 Assessment, other than in immediate proximity to major roads, where the air the few situations in which risk values were recomputed concentration, exposure, and/or risk might be significantly resulting from stationary source inventory errors, which were higher within a census tract or county. This limitation may determined to impact a tract or county-level risk estimate. result in underestimates ofexposure due to the design and Therefore, impacts of reductions from control programs on application ofASPEN. In addition, this type ofmodeling overall risk can be evaluated, as well as changes in relative assessment cannot address the kinds ofquestions an contributions of different source sectors to risk, changes in epidemiology study might allow, such as the relationship contributions ofdifferent pollutants to overall risk, and between asthma or cancer risk, and proximity ofresidences to changes in geographic distributions ofrisk. Such analysis is point sources, roadways, and other sources ofair toxics useful in setting regulatory priorities, and informing the emissions. The framework also does not account for risk decision-making process. The modeling includes the impacts from sources of air toxics originating indoors, such as stoves, ofmobile and stationary source control programs promul- out-gassing from building materials or evaporative benzene gated prior to 2005, as well as estimates ofthe effect of emissions from cars in attached garages. The ASPEN model economic growth on emissions. performs well for some pollutants, but has also been shown This paper addresses risks from the pollutants listed in to systematically underestimate concentrations relative to Table 1. These pollutants are all air toxic compounds listed in measured levels for some other pollutants such as certain section 112 ofthe Clean Air Act forwhich we had adequate highly reactive compounds and metals. data to estimate emissions. The paper does not address In NATA, quantitative estimates ofcancer risk are modeled air concentrations, exposure, or risk from diesel developed at the census tract level. Census tract median risk values are then used to calculate average individual risks and risk distributions at the county level and above. The Table 1 . Mobile source air toxics included in assessment. individual risk is the probability that an individual in a given population will contract cancer ifexposed to a given average 1,3-Butadiene Ethyl benzene concentration ofan airborne carcinogen over an assumed 2,2,4-Trimethylpentane Fluoranthene 70-year lifetime. It is calculated as the product of exposure Acenaphthene Fluorene and a cancer unit risk value (The Unit Risk Estimate is the Acenaphthylene Formaldehyde upper-bound excess lifetime cancer risk estimated to result Acetaldehyde N-Hexane Acrolein Indeno(1,2,3,c,d)-pyrene from continuous exposure to an agent at a concentration of Anthracene Manganese 3 1 mg/m in air). The unit risk estimates used in the 1999 Benzene Methyl tert-butyl ether (MTBE) NATA are upper confidence limits or maximum likelihood Benz(a)anthracene Naphthalene estimates, depending on the pollutant (U.S. EPA, 2006b). Benzo(a)pyrene Nickel Typically a unit risk estimate based on animal data will be an Benzo(b)fluoranthene Phenanthrene Benzo(g,h,i)perylene Propionaldehyde upper confidence limit, whereas a maximum likelihood Benzo(k)fluoranthene Pyrene estimate is used ifhuman data are adequate. Chromium Styrene The NATA estimates also assume that exposure levels Chrysene Toluene estimated for a given calendar year using HAPEM5 are Dibenzo(a,h)anthracene Xylenes representative ofaverage exposures over an entire lifetime, Compounds in bold were analyzed as polycyclic organic matter (POM) and do not account for expected changes in exposure over groups.

96 Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) Inhalation exposure and risk Cook et al. particulate matter and diesel exhaust organic gases. However, ASPEN to calculate air quality concentrations. In addition to it does include the contribution ofdiesel exhaust to emissions projecting stationary source emissions to future years for ofindividual mobile source air toxic compounds found in the some source categories, EMS-HAP spatially allocates emis- complex mixture. sions inventoried at the county level to the census tract level, and temporally allocates them to eight 3-h time periods throughout the day. Once the emissions were processed, they Methods were input into ASPEN to calculate air quality concentrations. In addition to the emissions, ASPEN uses meteoro- Inventory Development logical parameters and census tract internal point, or We projected the 1999 National Emission Inventory for centroid, locations for concentration calculations. ASPEN HAPs (U.S. EPA, 2004a) to numerous future years up to estimates do not account for day-of-week or seasonal 2030 using the following tools and data: variations in emissions. For all years modeled, meteorologi- cal conditions in 1999 and 2000 census tract data were used. the Emissions Modeling System for Hazardous Air Secondary formation through photochemical transformation Pollutants (EMS-HAP; U.S. EPA, 2004b) was modeled for acetaldehyde, acrolein, formaldehyde, and the National Mobile Inventory Model (NMIM; Cook propionaldehyde. et al., 2004b; Michaels et al., 2005) ASPEN only accounts for sources within a 50 km radius of emission reduction information resulting from national each source when calculating ambient concentrations. Thus, standards and economic growth data. the contribution to ambient levels ofair toxics fromsources Detailed description ofthe methodology used to develop further away than 50 km, as well as the contribution of inventories for 1999 through 2020 are described by Strum uninventoried sources, is addressed through the addition of et al. (2006). Emission reductions include impacts ofall US a ‘‘background’’ term (Battelle, 2003). Mobile source EPA control programs currently in place. For stationary pollutants, which include background components are 1,3- sources, this includes impacts ofMaximum Achievable butadiene, acetaldehyde, benzene, formaldehyde, and Control Technology (MACT) standards and rules requiring xylenes. Each ofthe three projection years used the same reductions in emissions in solid waste combustion under 1999-based background. Section 129 ofthe Clean Air Act. It should be noted that after 2020, stationary source emissions were assumed to Exposure and Risk Modeling remain constant. Reductions for highway vehicles reflect fuel The HAPEM5 exposure model used in this assessment is the programs such as Federal reformulated gasoline (RFG) most recent version in a series ofmodels that the EPA has standards, gasoline toxics emissions performance standards used to model population exposures and risks at the urban as required by US EPA’s 2001 mobile source air toxics rule, and national scale in a number ofassessments (U.S. EPA and low-sulfur gasoline and diesel requirements. Vehicle 1993, 1999, 2002b). HAPEM5 is designed to assess average programs include the national low-emission vehicle (NLEV) long-term inhalation exposures ofthe general population, or program, Tier 2 motor vehicle emissions standards and a specific subpopulation, over spatial scales ranging from gasoline sulfur control requirements; inspection and main- urban to national. HAPEM5 uses the general approach of tenance programs, on-board diagnostics, and heavy-duty tracking representatives ofspecified demographic groups as engine and vehicle standards. Reductions for nonroad they move among indoor and outdoor microenvironments equipment include impacts ofthe recent Clean Air Nonroad and among geographic locations. The estimated pollutant Diesel Rule promulgated by US EPA in 2004, as well as concentrations in each microenvironment visited are com- other controls on nonroad engines. US EPA rules, which bined into a time-weighted average concentration, which is have been proposed, but not yet issued as final rules, are not assigned to members ofthe demographic group. included in these projections. Nor are potential impacts of HAPEM5 uses four primary sources of information: requirements for increased renewable fuel use in the 2005 population data from the US Census Bureau for calendar Energy Policy Act. year 2000, population activity data, modeled air quality In addition to developing inventories for the modeled estimates, and microenvironmental data. Two kinds of pollutants, emission inventories for precursor compounds, activity data are used: activity pattern data and commuting which form these pollutants through photochemical pro- pattern data. The activity pattern data quantify the amount cesses, were also developed. oftime individuals spend in a variety ofmicroenvironments and come from EPA’s Consolidated Human Activity Air Quality Modeling Database (CHAD; Glen et al., 1997). The commuting data Prior to performing air quality modeling on the projected contained in the HAPEM5 default file were derived from a emissions, the emission inventories were processed in EMS- special 1990 US Census Bureau study that specifies the HAP, Version 3 to create the emissions input files used by number ofresidents ofeach tract that work in that tract and

Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) 97 Cook et al. Inhalation exposure and risk every other US Census tract. The air quality estimates come (UCLs) ofthe fitted dose–response curve, although where from ASPEN (after background has been added). The there are adequate epidemiological data, the unit risk may be microenvironmental data, developed from literature studies, based on a maximum likelihood (MLE), which represents the consists offactors that estimate air toxic concentrations in statistical best fit. The RfC is defined as an estimate, with specific microenvironments, based on penetration ofoutdoor uncertainty spanning perhaps an order ofmagnitude, ofan air into the microenvironment, proximity ofthe microenvir- inhalation exposure to the human population (including onment to the emission source, and emission sources within sensitive subgroups) that is likely to be without appreciable the microenvironment. These factors vary among pollutants. risks of deleterious effects during a lifetime. Average lifetime exposure for an individual in a census tract Also listed in Table 2 are the cancer weight ofevidence was calculated by building a lifetime exposure from classifications for carcinogens and target organ systems for appropriate age exposures for 70 years. noncancer calculations. The weight ofevidence classifications Once the exposure calculations were completed, cancer risk provided in this table were developed under U.S. EPA’s 1986 and noncancer hazard calculations were made for each of the risk assessment guidelines. Dose–response values were mobile source air toxic pollutants. Table 2 lists the pollutants selected using the same hierarchy followed in the 1999 included in this study, with their respective Unit Risk NATA. More details on the development ofthese unit risks Estimates (UREs) for cancer calculations and reference can be found on the website for the 1999 National Scale concentrations (RfCs) for noncancer calculations. These are Assessment and in Appendix H ofthe 2001 EPA draftreport the same values used in the 1999 National-Scale Air Toxics to the Science Advisory Board on the 1996 National-Scale Assessment. UREs represent the excess lifetime cancer risk Assessment (U.S. EPA, 2001). estimated to result from continuous lifetime exposure to an Cancer risk estimates (the product ofunit risk estimates agent at an average concentration of1 mg/m3 in air. These and exposure levels) for various pollutants were assumed to unit risks are typically statistical upper confidence limits be additive, since there was no evidence ofnonadditive

Ta bl e 2 . Dose response values used in risk modeling.

HAP Carcinogen class URE (per mg/m3) Source Organ systems RfC (mg/m3)Source

1,3-Butadiene A 3.0 Â 10À5 IRIS Reproductive 2.0 Â 10À3 2,2,4-Trimethylpentane N/A N/A N/A N/A Acetaldehyde B2 2.2 Â 10À6 IRIS Respiratory 9.0 Â 10À3 IRIS Acrolein N/A Respiratory 2.0 Â 10À5 IRIS Benzene A 7.8 Â 10À6 IRIS Immune 3.0 Â 10À2 IRIS Chromium III N/A N/A N/A N/A Chromium VI A 1.2 Â 10À2 IRIS Respiratory 1.0 Â 10À4 IRIS Ethyl Benzene N/A Developmental 1.0 IRIS Formaldehyde B 5.5 Â 10À9 CIIT Respiratory 9.8 Â 10À3 ATSDR Hexane N/A Respiratory, Neurological 2.0 Â 10À1 IRIS Manganese N/A Neurological 5.0 Â 10À5 IRIS MTBE N/A Liver, Kidney, Ocular 3.0 IRIS Naphthalene C 3.4 Â 10À5 CAL Respiratory 3.0 Â 10À3 IRIS Nickel A 1.6 Â 10À4 OAQPS Respiratory, Immune 6.5 Â 10À5 CAL Propionaldehyde N/A N/A N/A N/A POM1 B2 5.5 Â 10À5 OAQPS N/A POM2 B2 5.5 Â 10À5 OAQPS N/A POM3 B2 1.0 Â 10À1 OAQPS N/A POM4 B2 1.0x10À2 OAQPS N/A POM5 B2 1.0 Â 10À3 OAQPS N/A POM6 B2 1.0 Â 10À4 OAQPS N/A POM7 B2 1.0 Â 10À5 OAQPS N/A POM8 B2 2.0 Â 10À4 OAQPS N/A Styrene N/A Neurological 1.0 IRIS Toluene N/A Respiratory, Neurological 4.0 Â 10À1 IRIS Xylenes N/A Neurological 1.0 Â 10À1 IRIS

IRIS, EPA’s integrated risk information system; CIIT, CIIT centers for health research; CAL, California office of environmental health hazard assessment (OEHHA); OAQPS, US EPA office of air quality planning and standards; POM, polycyclic organic matter. Weight ofevidence classifications: A, known human carcinogen; B1, probable human carcinogen, based on incomplete human data; B2, Probable human carcinogen, based on adequate animal data; C, possible human carcinogen.

98 Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) Inhalation exposure and risk Cook et al. interactions for any of the pollutants. Most of the estimates that adverse effects will occur as they may not be proportional are based on the UCL, but the estimates for hexavalent to risk. A HI41 can be best described as indicating that a chromium, nickel, and benzene are based on the MLE. potential may exist for adverse health effects. To express chronic noncancer hazards, we used the RfC as part ofa calculation called the hazard quotient (HQ), which is the ratio between the concentration to which a person is Results and discussion exposed and the RfC. A value of the HQo1 indicates that the exposure is lower than the RfC and that no adverse Modeled Ambient Concentrations and Exposures health effects would be expected. A value of the HQ41 Table 3 presents the national annual average ofcensus tract indicates that the exposure is higher than the RfC. However, ambient concentrations (mg/m3) for background, stationary, because many RfCs incorporate protective assumptions in and mobile sources estimated using ASPEN, for 1999, 2015, the face of uncertainty, an HQ41 does not necessarily 2020, and 2030. Table 4 presents the national annual average suggest a likelihood of adverse effects. Furthermore, the HQ ofmean census tract exposure concentrations estimated using cannot be translated to a probability that adverse effects will HAPEM5. Overall, exposure concentrations tend to be less occur as it may not be proportional to risk. A HQ41can than modeled ambient concentrations because penetration best be described as indicating that a potential exists for rates to indoor microenvironments are typically less than one. adverse health effects. Following the approach used in the However, mobile sources make a larger contribution to 1999 National-Scale Assessment, combined noncancer ha- overall average population exposures than they do to zards were calculated using the hazard index (HI), defined as ambient levels. This is mostly because ofelevated exposures the sum ofhazard quotients forindividual air toxics experienced inside vehicles. compounds that affect the same organ or organ system. The HI is only an approximation ofthe combined effect, Estimated Cancer Risks because some ofthe substances may affectthe target organs in Table 5 presents national average inhalation cancer risks, different (i.e. nonadditive) ways. The assumption of dose based on these HAPEM5 modeled exposures, by carcinogen additivity is most appropriate for compounds that induce the class, in 1999, 2015, 2020, and 2030. The total cancer risk same effect by similar modes of action. As with the HQ, a from mobile source air toxics (including the stationary source value of the HI below 1 will likely not result in adverse effects contribution) was about 23 in a million in 1999. This over a lifetimeofexposure. However, a value ofthe HI 41 estimate does not reflect the small number ofcorrections does not necessarily suggest a likelihood ofadverse effects. made for stationary sources before release of the final 1999 Furthermore, the HI cannot be translated to a probability NATA. This compares to an overall nationwide average

Ta bl e 3 . Estimated US national average background, stationary, and mobile source contributions to modeled ambient concentrations (mgmÀ3)for each air toxic in 1999, 2015, 2020, and 2030.

HAP Background 1999 2015 2020 2030

Stationary Mobile Stationary Mobile Stationary Mobile Stationary Mobile

1,3-Butadiene 5.11 Â 10À2 2.24 Â 10À2 6.85 Â 10À2 2.27 Â 10À2 2.44 Â 10À2 2.28 Â 10À2 2.38 Â 10À2 2.28 Â 10À2 2.63 Â 10À2 2,2,4-Trimethylpentane 0 5.25 Â 10À2 7.71 Â 10À1 4.19 Â 10À2 2.86 Â 10À1 4.44 Â 10À2 2.59 Â 10À1 4.44 Â 10À2 2.75 Â 10À1 Acetaldehyde 5.17 Â 10À1 8.43 Â 10À2 8.28 Â 10À1 8.67 Â 10À2 3.62 Â 10À1 8.93 Â 10À2 3.29 Â 10À1 8.93 Â 10À2 3.50 Â 10À1 Acrolein 0 3.25 Â 10À2 8.10 Â 10À2 2.97 Â 10À2 3.41 Â 10À2 2.93 Â 10À2 3.37 Â 10À2 2.93 Â 10À2 3.74 Â 10À2 Benzene 3.94 Â 10À1 1.86 Â 10À1 7.81 Â 10À1 2.04 Â 10À1 3.16 Â 10À1 2.13 Â 10À1 2.95 Â 10À1 2.13 Â 10À1 3.18 Â 10À1 Chromium III 0 1.28 Â 10À3 1.58 Â 10À4 1.66 Â 10À3 2.11 Â 10À4 1.86 Â 10À3 2.30 Â 10À4 1.87 Â 10À3 2.73 Â 10À4 Chromium VI 0 3.06 Â 10À4 3.46 Â 10À5 4.08 Â 10À4 4.64 Â 10À5 4.61 Â 10À4 5.05 Â 10À5 4.61 Â 10À4 6.00 Â 10À5 Ethyl Benzene 0 1.04 Â 10À1 3.42 Â 10À1 1.24 Â 10À1 1.33 Â 10À1 1.37 Â 10À1 1.23 Â 10À1 1.37 Â 10À1 1.32 Â 10À1 Formaldehyde 7.62 Â 10À1 1.28 Â 10À1 7.10 Â 10À1 1.48 Â 10À1 3.13 Â 10À1 1.60 Â 10À1 3.03 Â 10À1 1.60 Â 10À1 3.31 Â 10À1 Hexane 0 5.14 Â 10À1 3.08 Â 10À1 5.91 Â 10À1 1.35 Â 10À1 6.38 Â 10À1 1.18 Â 10À1 6.38 Â 10À1 1.23 Â 10À1 MTBE 0 4.92 Â 10À2 6.90 Â 10À1 7.04 Â 10À2 1.26 Â 10À1 7.41 Â 10À2 1.05 Â 10À1 7.41 Â 10À2 1.05 Â 10À1 Manganese 0 7.56 Â 10À3 1.21 Â 10À4 6.14 Â 10À3 1.85 Â 10À4 6.80 Â 10À3 2.08 Â 10À4 6.80 Â 10À3 2.61 Â 10À4 Naphthalene 0 5.22 Â 10À2 1.88 Â 10À2 6.15 Â 10À2 1.24 Â 10À2 6.57 Â 10À2 1.26 Â 10À2 6.57 Â 10À2 1.46 Â 10À2 Nickel 0 2.19 Â 10À3 1.91 Â 10À4 2.50 Â 10À3 2.50 Â 10À4 2.74 Â 10À3 2.71 Â 10À4 2.74 Â 10À3 3.19 Â 10À4 POMa 0 2.13 Â 10À2 2.32 Â 10À3 2.25 Â 10À2 1.47 Â 10À3 2.34 Â 10À2 1.50 Â 10À3 2.34 Â 10À2 1.71 Â 10À3 Propionaldehyde 0 3.34 Â 10À2 2.07 Â 10À1 3.32 Â 10À2 9.20 Â 10À2 3.39 Â 10À2 8.21 Â 10À2 3.39 Â 10À2 8.62 Â 10À2 Styrene 0 3.94 Â 10À2 3.36 Â 10À2 4.90 Â 10À2 1.23 Â 10À2 5.53 Â 10À2 1.19 Â 10À2 5.53 Â 10À2 1.29 Â 10À2 Toluene 0 1.01 1.99 1.19 7.24 Â 10À1 1.31 6.63 Â 10À1 1.31 7.07 Â 10À1 Xylenes 1.70 Â 10À1 6.84 Â 10À1 1.37 8.62 Â 10À1 5.22 Â 10À1 9.52 Â 10À1 4.87 Â 10À1 9.52 Â 10À1 5.24 Â 10À1 aPOM, polyclic organic matter and is the sum ofall POM groups.

Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) 99 Cook et al. Inhalation exposure and risk

Ta bl e 4 . Estimated US national average stationary and mobile source exposure concentrations (mgmÀ3) for 1999, 2015, 2020, and 2030 by air toxic pollutant.

HAP Background 1999 2015 2020 2030

Stationary Mobile Stationary Mobile Stationary Mobile Stationary Mobile

1,3-Butadiene 3.90 Â 10À2 1.88 Â 10À2 8.04 Â 10À2 1.91 Â 10À2 2.77 Â 10À2 1.92 Â 10À2 2.68 Â 10À2 1.92 Â 10À2 2.94 Â 10À2 2,2,4-Trimethylpentane 0 4.47 Â 10À2 8.69 Â 10À1 3.61 Â 10À2 3.16 Â 10À1 3.83 Â 10À2 2.86 Â 10À1 3.83 Â 10À2 3.03 Â 10À1 Acetaldehyde 4.00 Â 10À1 7.14 Â 10À2 9.01 Â 10À1 7.40 Â 10À2 3.84 Â 10À1 7.62 Â 10À2 3.47 Â 10À1 7.62 Â 10À2 3.67 Â 10À1 Acrolein 0 2.76 Â 10À2 8.68 Â 10À2 2.53 Â 10À2 3.40 Â 10À2 2.50 Â 10À2 3.33 Â 10À2 2.50 Â 10À2 3.68 Â 10À2 Benzene 3.01 Â 10À1 1.61 Â 10À1 8.91 Â 10À1 1.78 Â 10À1 3.54 Â 10À1 1.86 Â 10À1 3.29 Â 10À1 1.86 Â 10À1 3.53 Â 10À1 Chromium III 0 5.14 Â 10À4 8.77 Â 10À5 6.68 Â 10À4 1.21 Â 10À4 7.51 Â 10À4 1.33 Â 10À4 7.51 Â 10À4 1.60 Â 10À4 Chromium VI 0 1.26 Â 10À4 1.95 Â 10À5 1.687 Â 10À4 2.70 Â 10À5 1.90 Â 10À4 2.96 Â 10À5 1.90 Â 10À4 3.57 Â 10À5 Ethyl Benzene 0 9.07 Â 10À1 3.79 Â 10À1 1.08 Â 10À1 1.44 Â 10À1 1.19 Â 10À1 1.32 Â 10À1 1.19 Â 10À1 1.41 Â 10À1 Formaldehyde 5.91 Â 10À1 1.09 Â 10À1 7.93 Â 10À1 1.27 Â 10À1 3.25 Â 10À1 1.38 Â 10À1 3.12 Â 10À1 1.38 Â 10À1 3.40 Â 10À1 Hexane 0 4.46 Â 10À1 3.50 Â 10À1 5.12 Â 10À1 1.50 Â 10À1 5.53 Â 10À1 1.30 Â 10À1 5.53 Â 10À1 1.36 Â 10À1 MTBE 0 6.33 Â 10À2 7.82 Â 10À1 5.95 Â 10À2 1.40 Â 10À1 6.27 Â 10À2 1.14 Â 10À1 6.27 Â 10À2 1.14 Â 10À1 Manganese 0 1.98 Â 10À3 7.93 Â 10À5 2.48 Â 10À3 1.23 Â 10À4 2.74 Â 10À3 1.39 Â 10À4 2.74 Â 10À3 1.74 Â 10À4 Naphthalene 0 4.44 Â 10À2 2.13 Â 10À2 5.25 Â 10À2 1.34 Â 10À2 5.61 Â 10À2 1.36 Â 10À2 5.61 Â 10À2 1.57 Â 10À2 Nickel 0 9.05 Â 10À4 1.02 Â 10À4 1.03 Â 10À3 1.35 Â 10À4 1.12 Â 10À3 1.48 Â 10À4 1.12 Â 10À3 1.77 Â 10À4 POMa 0 1.31 Â 10À2 1.85 Â 10À3 1.40 Â 10À2 1.12 Â 10À3 1.46 Â 10À2 1.14 Â 10À3 1.46 Â 10À2 1.31 Â 10À3 Propionaldehyde 0 2.81 Â 10À2 2.22 Â 10À2 2.80 Â 10À2 9.68 Â 10À2 2.86 Â 10À2 8.58 Â 10À2 2.86 Â 10À2 8.98 Â 10À2 Styrene 0 3.31 Â 10À2 3.87 Â 10À2 4.09 Â 10À2 1.36 Â 10À2 4.61 Â 10À2 1.30 Â 10À2 4.61 Â 10À2 1.42 Â 10À2 Toluene 0 8.69 Â 10À1 2.26 1.03 8.06 Â 10À1 1.14 7.35 Â 10À1 1.14 7.83 Â 10À1 Xylenes 1.28 Â 10À1 5.96 Â 10À1 1.51 7.52 Â 10À1 5.61 Â 10À1 8.31 Â 10À1 5.20 Â 10À1 8.31 Â 10À1 5.59 Â 10À1 aPOM, polycyclic organic matter.

Ta bl e 5 . Estimated US national average inhalation cancer risks per one million individuals for stationary and mobile sources for individual air toxics, each carcinogenic class and total risk (all air toxics).

HAP Carcinogen Class Back-ground 1999 2015 2020 2030

Stationary Mobile Stationary Mobile Stationary Mobile Stationary Mobile

1,3-Butadiene A 1.15 5.63 Â 10À1 2.41 5.73 Â 10À1 8.30 Â 10À1 5.76 Â 10À1 8.03 Â 10À1 5.76 Â 10À1 8.82 Â 10À1 Benzene A 2.32 1.25 6.95 1.38 2.76 1.45 2.57 1.45 2.75 Chromium VI A N/A 1.51 2.34 Â 10À1 2.01 3.24 Â 10À1 2.28 3.56 Â 10À1 2.28 4.29 Â 10À1 Nickel A N/A 1.45 Â 10À1 1.64 Â 10À2 1.64 Â 10À1 2.17 Â 10À2 1.80 Â 10À1 2.37 Â 10À2 1.80 Â 10À1 2.83 Â 10À2 Class A A 3.47 3.47 9.61 4.14 3.94 4.48 3.75 4.48 4.09

Formaldehyde B1 3.22 Â 10À3 6.00 Â 10À4 4.36 Â 10À3 7.01 Â 10À4 1.79 Â 10À3 7.57 Â 10À4 1.73 Â 10À3 7.57 Â 10À4 1.87 Â 10À3 Class B1 B1 3.22 Â 10À3 6.00 Â 10À4 4.36 Â 10À3 7.01 Â 10À4 1.79 Â 10À3 7.57 Â 10À4 1.73 Â 10À3 7.57 Â 10À4 1.87 Â 10À3

Acetaldehyde B2 8.74 Â 10À1 1.57 Â 10À1 1.98 1.63 Â 10À1 8.45 Â 10À1 1.68 Â 10À1 7.64 Â 10À1 1.68 Â 10À1 8.08 Â 10À1 POMa B2 N/A 1.19 1.21 Â 10À1 1.31 7.29 Â 10À2 1.36 7.44 Â 10À2 1.36 8.51 Â 10À2 Class B2 B2 8.74 Â 10À1 1.35 2.10 1.47 9.18 Â 10À1 1.53 8.38 Â 10À1 1.53 8.93 Â 10À1

Naphthalene C N/A 1.51 7.24 Â 10À1 1.79 4.57 Â 10À1 1.91 4.63 Â 10À1 1.91 6.35 Â 10À1 Class C C N/A 1.51 7.24 Â 10À1 1.794.57 Â 10À1 1.91 4.63 Â 10À1 1.91 6.35 Â 10À1

Total Risk All 4.35 6.33 12.4 7.39 5.31 7.92 5.05 7.92 5.52

aPOM, polycyclic organic matter.

population cancer risk from all air toxics in the 1999 risk from mobile sources is from gasoline vehicles and engines National-Scale Assessment of42 in a million (U.S. EPA, (see Figure 2). Other signiﬁcant contributors to cancer risk 2006b). In all projection years, benzene emissions are by far from mobile source air toxics include 1,3-butadiene, acet- the largest contributor to cancer risk from mobile sources (see aldehyde, and naphthalene. Figure 1). Furthermore, about 90% ofthe mobile source risk Despite signiﬁcant predicted reductions in risk from mobile from all air toxics included in this assessment is due to source air toxics, average inhalation cancer risk estimates for gasoline vehicles and engines, and about 95% ofthe benzene these pollutants, accounting for both mobile and stationary

100 Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) Inhalation exposure and risk Cook et al.

Formaldehyde 2.5E-05 POM Nickel 0% All MSATS 1% 0% Benzene Other 0% 2.0E-05 Chromium VI 7% Naphthalene 1.5E-05 9%

1.0E-05

5.0E-06 Acetaldehyde Benzene

15% 52% Average Nationwide Cancer Risk

0.0E+00 1999 2015 2020 2030 Year Figure 3. Average nationwide (50-State) cancer risk from emissions of mobile source air toxics from both mobile and stationary sources 1,3-Butadiene across census tracts, 1999–2030. 16% Figure 1. Contributions to inhalation cancer risk from air toxics emitted by mobile sources, 2020. Other 1E-06 <= Risk < 1E-05 1E-05 <= Risk < 1E-04 400000000 Total Mobile Source Contribution to Benzene Risk (Not Including 350000000 Contribution to Background) = 2.6×10-6 300000000 remaining nonroad 4% 250000000 200000000 nonroad gasoline 22% 150000000 100000000 50000000 0 highway diesel 1999 2015 2020 2030 2% Figure 4. US population at various cancer risk benchmarks due to exposure to mobile source air toxics, 1999–2030 (does not include Alaska and Hawaii). highway gasoline 72% thousand (1 Â 10À5) increases from 214 million in 1999 to 240 million in 2030. Figure 2. Contributions ofmobile source categories to mobile source Figure 5 depicts the geographic distributions ofannual benzene risk in 2020. median county cancer risks (the median ofaverage census tract risks in that county) in 2020 for all mobile source air toxics across the US. Not surprisingly, the highest risks are source contributions, remain well above 10 in 1,000,000 in found in major population centers of the country where 2030 (Figure 3). In addition, average risk from exposure to mobile source activity is the greatest. As benzene accounts for benzene remains above 5 in 1,000,000. It should also be such a large proportion ofrisk from mobile source air toxics, noted that because ofpopulation growth projected to occur relatively high risks are also seen in areas ofthe country in the US, the number ofAmericans above cancer risk where fuel benzene levels are higher, based on fuel survey ‘‘benchmark’’ levels will increase. Figure 4 depicts the US data (U.S. EPA, 2006a). These areas include the Paciﬁc population relative to cancer risk levels for mobile source air Northwest, parts ofAlaska, and the Upper Great Lakes toxics in 1999, 2015, 2020, and 2030. These statistics do not Region, since higher fuel benzene levels lead to higher include populations in Alaska and Hawaii. From this ﬁgure benzene emissions and higher exposures. it can be seen that the vast majority ofthe population experiences risks between one in a million (1 Â 10À6)andone Estimated Noncancer Hazards in ten thousand (1 Â 10À4). However, the number ofpeople The respiratory system is the only target organ system where predicted to experience risks above one in a hundred the estimated average HI exceeds one. Although the average

Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) 101 Cook et al. Inhalation exposure and risk

Risk (N in a million)

0.0 - 5.5

5.6 - 10.3

10.4 - 17.4

17.5 - 27.9

28.0 - 47.3

47.4 - 96.3

Figure 5. 2020 annual county median cancer risk due to exposure to mobile source air toxics.

7 source contribution to concentrations and exposure is largely background stationary mobile attributable to the contribution from mobile source 1,3- 6 butadiene, which is transformed to acrolein in the atmo-

5 sphere. Moreover, projected growth in the US population will increase the number ofAmericans with a respiratory HI 4 for mobile source air toxics above one, from 250 million in 1999–273 million in 2030 (Figure 7). Figure 8 depicts the 3 geographic distribution ofannual median county respiratory Hazard Index 2 hazard indices in 2020. The high HI in Idaho are the result of high inventory estimates for wildfires and reflect a known 1 error in the Idaho inventory for this source. It should be noted that the hazard indices were adjusted to account for 0 1999 2015 2020 2030 this issue in the final version ofthe 1999 NATA. Year

Figure 6. Average respiratory HI for US population (aggregate of Limitations and Uncertainties hazard quotients for individual pollutants). Limitations and uncertainties in the data and tools used to perform the modeling presented in this paper must be respiratory HI for mobile source air toxics decreases by considered when interpreting results, and are discussed almost 50% between 1999 and 2030 (Figure 6), it is still over below. Additional information on limitations and uncertain- 3 in 2030, indicating a potential for adverse health effects. In ties can be found elsewhere (U.S. EPA, 2006b, 2001). addition, about 95% ofthis noncancer hazard is attributable Among the limitations and uncertainties associated with the to acrolein in all projection years. Table 6 presents the inventory (Strum et al., 2006; U.S. EPA, 2006a) are: national respiratory HI for chronic noncancer effects across census tracts, 1999–2030. About 25% ofprimary acrolein Inventory estimates are based on limited toxic emissions emissions are from mobile sources, and about 70% of data for some sources. ambient concentrations ofacrolein (and about 75% of There are uncertainties associated with development of exposure) are attributable to mobile sources. The mobile county-level and census tract-level emissions estimates

102 Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) Inhalation exposure and risk Cook et al.

Ta bl e 6 . Estimated US national respiratory HI for chronic noncancer effects across census tracts.

Respiratory system average HI

Year Background Major Area & other Onroad Nonroad Total (including background)

1999 1.04EÀ01 1.49EÀ01 1.28 3.52 1.02 6.07 2015 1.04EÀ01 1.65EÀ01 1.16 9.88EÀ01 7.99EÀ01 3.22 2020 1.04EÀ01 1.85EÀ01 1.13 9.08EÀ01 8.38EÀ01 3.17 2030 1.04EÀ01 1.85EÀ01 1.13 9.79EÀ01 9.48EÀ01 3.35

0.1 <= HI < 1 1<= HI < 10 HI >= 10 near road impacts suggests that modeled exposures to 400 some individuals are up to 50% higher than those 350 predicted by HAPEM5 (U.S. EPA, 2006a). The results do not include impacts from sources in Canada 300

Millions or Mexico. 250 Exposure modeling does not fully reflect variation among 200 individuals or noninhalation exposure pathways. 150 Indoor sources are not accounted for. For some 100 compounds such as formaldehyde, these indoor sources 50 Population can contribute significantly to the total exposure for an 0 individual, even ifonly inhalation exposures are consid- 1999 2015 2020 2030 ered. Year The assessment does not fully reflect variation in back- Figure 7. US population at various noncancer hazard benchmarks for ground ambient air concentrations. Instead, it uses average respiratory effects, due to exposure to mobile source air toxics, 1999– 2030. values over broad geographic regions. Exposure levels estimated for a given calendar year using HAPEM5 are assumed to remain constant over an entire from broader geographic data (i.e., state, regional or lifetime in estimating risk, and do not account for expected national). changes in exposure over time resulting from changes in When developing projection year inventories, there are emissions. uncertainties associated with estimating growth in activity, The assessment does not include default adjustments for and impacts ofcontrol programs on emissions. early life exposures recently recommended in the Supple- Since the time this analysis was performed, US EPA has mental Guidance for Assessing Susceptibility from found that current technology vehicles emit higher Early-Life Exposure to Carcinogens (U.S. EPA, 2005). hydrocarbon and gaseous air toxic emissions at cold Incorporation ofsuch adjustments would lead to higher temperatures during vehicle starts than estimated by estimates oflifetime risk. NMIM. Cancer and noncancer risk values used are uncertain. Sources ofuncertainty include extrapolating from animals There are a number ofadditional significant uncertainties to humans and extrapolating from high doses to low doses. associated with the air quality, exposure, and risk modeling. Among these uncertainties are: As part ofthe 1999 NATA, EPA compared ASPEN- Parameters used to characterize photochemical processes, modeled concentrations with available, but geographically long-range transport, terrain effects, deposition rates, limited, ambient air quality monitoring data for 1999 (U.S. human activity patterns, relationships between ambient EPA, 2006b). For each monitor-pollutant combination, levels in different microenvironments, and dose–response EPA compared the annual average concentration estimated may not appropriately reflect real world conditions. by the ASPEN model at the exact geographical coordinates As mentioned in the introduction, results are most ofthe monitor location with the annual average monitored accurate for large geographic areas and cannot be used value to get a point-to-point comparison between the model to identify ‘‘hot spots,’’ such as the near road microenvir- and monitor concentrations. The agreement between model onment. More refined local-scale modeling is needed for and monitor values for benzene was very good, with a this purpose (Kinnee et al., 2004; Pratt et al., 2004; Cohen median model to monitor ratio of0.95, and 74% ofsites et al., 2005; Touma et al., 2006). Limited modeling results within a factor of 2. Agreement for acetaldehyde was almost using a new version ofHAPEM which better accounts for as good as benzene, but data suggest that ASPEN could be

Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) 103 Cook et al. Inhalation exposure and risk

Hazard Index

0.000 - 0.883

0.884 - 2.175 2.176 - 5.099 5.100 - 10.182

10.183 - 16.707

16.708 - 31.635

Figure 8. 2020 annual county median noncancer HI for respiratory effects, due to exposure to mobile source air toxics. underpredicting for other mobile source air toxics, including the contribution from both mobile and stationary sources, is formaldehyde, chromium, manganese and nickel, by 30– 18 in a million. Furthermore, the average respiratory HI is 70%. predicted still to be over 3, indicating a potential for adverse In addition to the limitations and uncertainties associated health effects. In fact, in 2030 there will be more people with modeling the 1999 base year, there are additional ones exposed to the highest levels ofrisk, because ofanticipated in the projection year modeling. For instance, the modeling is growth in the U.S. population. The number ofAmericans not accounting for impacts of demographic shifts that are above the 10 in a million cancer risk level from exposure to likely to occur in the future. A key limitation is using 1999 MSATs is projected to increase from 214 million in 1999 to ‘‘background’’ levels to account for mid-range to long-range 240 million in 2030, and the number ofAmericans with a transport. However, since background is related to emissions respiratory HI for mobile source air toxics above one, from far away from receptors, these levels should decrease as those 250 million in 1999 to 273 million in 2030. These emissions decrease. Results ofa sensitivity analysis for conclusions, however, are dependent on the accuracy ofthe benzene where background estimates were scaled by the model predictions, and there are signiﬁcant limitations and change in the inventory for a future year relative to 1999 uncertainties in the data and tools used to perform the indicated that using a scaled background reduced benzene modeling presented in this paper. concentrations by about 15% across the US in 2015, 2020, and 2030 (U.S. EPA, 2006a). Acknowledgements

Conclusions We acknowledge the contributions ofChad Bailey, David Brzezinski, Harvey Michaels, and Kathryn Sargeant, US Results ofthe analyses described in this paper suggest that EPA, Office of Transportation and Air Quality, and Anne US EPA emission control programs will substantially reduce Pope, Laurel Driver, Richard Mason, and Stephen Shedd of average inhalation cancer risks and potential noncancer US EPA, Office of Air Quality Planning and Standards, to health risks from exposure to mobile source air toxics, the work presented in this paper. We would also like thank between 1999 and 2030. However, cancer risk will continue Dr. Chon Shoaf, US EPA, Office of Research and to be a public health concern. The predicted national average Development, and Dr. Kenneth Mitchell, US EPA, Region cancer risk from mobile source air toxics in 2030, including 4, for comments on a draft of this paper.

104 Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) Inhalation exposure and risk Cook et al.

Disclaimer Touma J.S., Isakov V., Ching J., and Seigneur C. Air Quality Modeling of This paper has been reviewed in accordance with the US EPA Hazardous Air Pollutants: Current Status and Future Directions. JAirWaste Manage Assoc 2006: 56: 547–558. peer and administrative review policies and approved for U.S. Environmental Protection Agency. Motor Vehicle-Related Air Toxics Study, presentation and publication. Mention oftrade names or Office of Mobile Sources, Ann Arbor, MI, 1993 Report No. EPA commercial products does not constitute endorsement or 420-R-93-005, 1993 (available at http://www.epa.gov/otaq/regs/toxics/tox_ archive.htm). recommendation for their use. U.S. Evironmental Protection Agency. Carcinogenic Effects of Benzene: An Update, National Center for Environmental Assessment, Washington, DC, 1998 EPA600-P-97-001F. (available at http://www.epa.gov/ncepihom/ Catalog/EPA600P97001F.html). References U.S. Environmental Protection Agency. Analysis ofthe Impacts ofControl Programs on Motor Vehicle Toxics Emissions and Exposure in Urban Areas Battelle. Estimated background concentrations for the National-Scale Air Toxics and Nationwide. Prepared for U. S. EPA, Office of Transportation and Air Assessment. Prepared for U.S. EPA, Office of Air Quality Planning and Quality, by Sierra Research, Inc., and Radian International Corporation/ Standards. 2003. Contract No. 68-D-02-061, Work Assignment 1-03, 2003 Eastern Research Group, 1999 Report No. EPA 420 –R-99-029/030. (available at http://www.epa.gov/ttn/atw/nata1999). (available at http://www.epa.gov/otaq/regs/toxics/tox_archive.htm). Cohen J., Cook R., Bailey C.R., and Carr E. Relationship between motor vehicle U.S. Environmental Protection Agency. User’s Guide for the Assessment System emissions ofhazardous pollutants, roadway proximity, and ambient concen- for Population Exposure Nationwide (ASPEN, Version 1.1) Model, Office of trations in Portland, Oregon. EnvironModeling&Software2005: 20: 7–12. Air Quality Planning and Standards, Research Triangle Park, NC, 2000 Cook R., Glover E., Michaels H., and Brzezinski D. Modeling ofmobile source Report No. EPA-454/R-00-017. (Available at http://www.epa.gov/scram001/ air toxics using EPA’s National Mobile Inventory Model. Proceedings, 2004 userg/other/aspenug.pdf). Emission Inventory Conference, Clearwater Beach, Fl,, 2004b: http:// U.S. Environmental Protection Agency. National-Scale Air Toxics Assessment www.epa.gov/ttn/chief/conference/ei13/poster/cook.pdf. for 1996: Draft for EPA Science Advisory Board Review, 2001 Report Cook R., Jones B., and Cleland J. A cohort based approach for characterizing No. EPA-453/R-01-003. (available at http://www.epa.gov/ttn/atw/sab/ lifetime inhalation cancer risk from time-varying exposure to air toxics from natareport.pdf). ambient sources. Environmental Progress 2004a: 23(2): 120–125. U.S. Environmental Protection Agency. Health Assessment of1,3-Butadiene, Glen G., Lakkadi Y., Tippett J.A., and del Valle-Torres M. Development of Office of Research and Development, National Center for Environmental NERL/CHAD: The National Exposure Research Laboratory Consolidated Assessment, Washington, DC, 2002a Report No. EPA600-P-98-001F Human Activity Database. Prepared by ManTech Environmental Technology, (Available at http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid¼ 54499). Inc. EPA Contract No. 68-D5-0049, 1997. U.S. Environmental Protection Agency. National-Scale Air Toxics Assessment for Kinnee E.J., Touma J.S., Mason R., Thurman J., Beidler A., Bailey C., and Cook 1996, 2002b (Available at http://www.epa.gov/ttn/atw/nata). R. Allocation ofOnroad Mobile Emissions to Road Segments forAir Toxics U.S. Environmental Protection Agency. 1999 Final National Emissions Inventory Modeling in Harris County, Texas. Transport Res Part D 2004: 9: 139–150. Data and Documentation, 2004a (available at http://www.epa.gov/ttn/chief/ Michaels H., Brzezinski D., and Cook R. EPA’s National Mobile Inventory net/1999inventory.html). Model (NMIM), A Consolidated Emissions Modeling System for MOBILE6 U.S. Environmental Protection Agency. User’s Guide for the Emissions and NONROAD, U.S. Environmental Protection Agency, Office of Modeling System for Hazardous Air Pollutants (EMS-HAP, Version 3.0), Transportation and Air Quality, Assessment and Standards Division, Ann Office of Air Quality Planning and Standards, Research Triangle Park, NC, Arbor, MI, March 2005; Report No. EPA-420-R-05-003. 2004b Report No. EPA-454/B-00-007. (available at http://www.epa.gov/ Pratt G.C., Wu C.Y., and Bock D., et al. Comparing air dispersion model scram001/tt22.htm#aspen. predictions with measured concentrations ofVOCs in urban communities. Env U.S. Environmental Protection Agency. Supplemental Guidance for Assessing Sci Technol 2004: 38: 1949–1959. Susceptibility from Early-Life Exposure to Carcinogens, 2005 Report Rosenbaum A. The HAPEM5 User’s Guide: Hazardous Air Pollutant Exposure No. EPA/630/R-03/003F. (available at http://cfpub.epa.gov/ncea/cfm/ Model, Version 5. Prepared by ICF, Inc. for U.S. EPA, 2005 (available at recordisplay.cfm?deid¼ 116283). http://www.epa.gov/ttn/fera/hapem5/hapem5_guide.pdf). U.S. Environmental Protection Agency. Draft Regulatory Impact Analysis: South Coast Air Quality Management District. Multiple Air Toxics Exposure Control ofHazardous Air Pollutants from Mobile Sources, Office of Study in the South Coast Air Basin – MATES-II, 2000 (available at http:// Transportation and Air Quality, Ann Arbor, MI, 2006a Report No. www.aqmd.gov/matesiidf/matestoc.htm). EPA420-D-06-004. (available at http://www.epa.gov/otaq/toxics.htm). Strum M., Cook R., Thurman J., Ensley D., Pope A., Palma T., Mason R., U.S. Environmental Protection Agency. National-Scale Air Toxics Michaels H., and Shedd S. Projection ofhazardous air pollutant emissions to Assessment for 1999, 2006b (available at http://www.epa.gov/ttn/atw/ future years. Sci Total Env 2006: 366: 590–601. nata1999).

Journal of Exposure Science and Environmental Epidemiology (2007) 17(1) 105