Accurate information on human exposure to hazardous air pollutants emitted by various sources is crucial to assessing their potential health risks. This chapter describes methods used to assess exposure to hazardous air pollutants. Section 112 of the Clean Air Act Amendments of 1990 applies to major sources that either singly or in combination emit defined quantities of one or more of the 189 hazardous air pollutants. The sources to which the act applies emit pollutants both continuously and episodically, and the pollutants can move from air to water, soil, or food.
In the terminology of the Environmental Protection Agency (EPA) and Title III of the 1990 Amendments, a major source of pollution is considered to be
any stationary source or group of stationary sources located within a contiguous area and under common control that emits or has the potential to emit considering controls, in the aggregate, 10 tons per year or more of any hazardous air pollutant or 25 tons per year of any combination of hazardous air pollutants. The [EPA] Administrator may establish a lesser quantity, or in the case of radionuclides different criteria, for a major source than that specified in the previous sentence, on the basis of the potency of the air pollutant, persistence, potential for bioaccumulation, other characteristics of the air pollutant, or other relevant factors.
A stationary source is "any building, structure, emission source, or installation which emits or may emit any air pollutant."
As part of determining the health threat of a pollution source to humans, EPA assesses how a pollutant moves from a source through the environment
until it makes contact with humans in its original form or after conversion to other substances. For most airborne substances, inhalation is assumed to be the primary route of entry into the body. There has recently been an extensive review of advances in assessing human exposure to airborne constituents (NRC, 1991a). That review attempted to define exposure carefully as a part of the overall continuum that leads to illness brought about by environmental contaminants. The definition of exposure as a part of this continuum has been incorporated into the 1992 revised guidelines for exposure assessment developed by EPA (1992a).
Human exposure to a contaminant is an event consisting of contact with a specific contaminant concentration at a boundary between a human and the environment (e.g., skin or lung) for a specified interval; total exposure is determined by the integrated product of concentration and time. The amount of a substance that is absorbed or deposited in the body of an exposed person in a given period is the administered dose. Calculating the dose from the exposure depends on a number of factors, including the mode of entry into the body. For substances that move into the body through an openingsuch as the mouth or nose via breathing, eating, or drinkingthe dose depends on the amount of the carrier medium that enters the body. For airborne substances, the potential dose is the product of breathing rate (volume of air inhaled per unit of time), exposure concentration, and fractional deposition of the substance throughout the respiratory tract. However, an inhalation exposure will not lead to a dose if none of the substance is absorbed through the lung or deposited on the surface of the lung or other sections of the respiratory tract.
A pollutant can also enter the body through the skin or other exposed tissues, such as the eyes. The substance is then directly absorbed from the carrier medium into the tissue, often at a rate that is different from the rate of absorption of the carrier. The pollutant uptake rate is the amount of the pollutant absorbed per unit of time, and the dose is the product of exposure concentration and uptake rate at that concentration. The NRC report on exposure assessment (NRC, 1991a) provides a scientific framework to identify routes of entry and degree of contact and indicates how exposure assessment integrates data on emitted pollutants with biological effects.
Exposure assessment involves numerous techniques to identify a pollutant, pollutant sources, environmental media of exposure, transport through each medium, chemical and physical transformations, routes of entry to the body, intensity and frequency of contact, and spatial and temporal concentration patterns of the pollutant. Mathematical models that can be used to describe the relationships among emissions, exposures, and doses are shown in Appendix C.
Exposure to a contaminant can be estimated in three ways. It can be evaluated directly by having a person wear a device that measures the concentration of a pollutant when it comes into contact with the body. Environmental monitoring is an indirect method of determining exposure, in which a chemical's concentration is measured in an environmental medium at a particular site, and the extent
to which a person is exposed to that medium is used to estimate exposure. Finally, exposure can be estimated from the chemical's actual dose to the body, if it manifests itself in some known way through a measurable internal indicator (biological marker), such as the concentration of the substance or its metabolite in a body tissue or excreted material (NRC, 1991a). This is a direct method of exposure estimation and, unlike the other two, accounts for the amount of contaminant absorbed by the body. Each of these methods provides an independent estimate of exposure; when it is possible to use more than one approach, comparison of results can be useful in validating exposure estimates.
EPA's air-pollution regulatory programs have relied primarily on mathematical models to predict the dispersion of emissions to air and the potential for human exposure under different emission-control scenarios (see Appendix C for a description of EPA's Human Exposure Model). Source-emission estimates and meteorologic data were used to calculate the expected long-term ambient concentrations at various distances and directions from the source. Census data were used to estimate the number and location of people living near the source. A high-exposure scenario was estimated for a person (e.g., maximally exposed individual, MEI) assumed to be living near the source and constantly exposed for 70 years to the highest estimated air-pollutant concentration. EPA does not modify exposure estimates by including mobility of the population, shielding due to indoor locations, or additional exposures from indoor or other community sources. EPA also used a modeling approach to estimate the exposure of the local population to an average concentration of pollutant emitted from a source (EPA, 1985a).
1992 Exposure-Assessment Guidelines
EPA has recently promulgated a new set of exposure-assessment guidelines to replace the previous (1986) version (EPA, 1992a). The approach in the new guidelines is very different from that in the previous version and generally follows many of the concepts of exposure assessment presented in the 1991 NRC report (NRC, 1991a). The guidelines explicitly consider the need to estimate the distribution of exposures of individuals and populations and discuss the need to incorporate uncertainty analysis into exposure assessment. This approach is consistent with the most recent NRC recommendations on exposure analysis (NRC, 1993e).
The guidelines discuss the roles of both analytic measurement and mathematical modeling in estimating concentrations and durations of exposure. They do not recommend specific models, but suggest that models match the objectives of the particular exposure assessment being conducted and that they have the accuracy needed to achieve those objectives. They also call for detailed explication of the choices and assumptions that often must be made in the face of incomplete data and insufficient resources.
Exposure Calculation and the Maximally Exposed Individual
EPA has traditionally characterized exposure according to two criteria: exposure of the total population and exposure of a specified, usually highly or maximally exposed individual. The MEI's exposure is estimated as the plausible upper bound of the distribution of individual exposures. The reason for finding the MEI, as well as population exposure, is to assess whether any individual exposure might occur above a particular threshold that, as a policy matter, is considered to be important. Because the MEI's exposure level is intended to represent a potential upper bound, its calculation has involved a variety of conservative assumptions. Among the more conservative, and more contentious, were that the MEI lived for 70 years at the location deemed by the dispersion model to receive the heaviest annual average concentration, that the person stayed there 24 hours/day, and that there is no difference between outdoor and indoor concentrations. In practice, it is straightforward to estimate the exposure of an immobile MEI with the air-quality models described below. However, estimating exposure for a more typical person requires much more information as to his or her activities during the assessment period. Usually, these activities include spending a majority of time inside (where pollutant concentrations can be attenuated) and time spent in travel away from the residence. The 70-year, 24-hour/day and no-indoor-attenuation assumptions are, in effect, bounding estimates. Some people do live in a small community for a whole lifetime. Some people do spend virtually their whole life at home. And for some pollutants, there is little attenuation of pollutant concentrations indoors. Nonetheless, the occurrence of these conditions is rare, and it is even rarer that all these are found together.
In the most recent exposure guidelines, EPA no longer uses the term MEI, noting the difficulty in estimating it and the variety of its uses. The MEI has been replaced with two other estimators of the upper end of the individual exposure distribution, a "high-end exposure estimate" (HEEE) and the theoretical upper-bounding estimate (TUBE). The HEEE is not specifically defined ("the Agency has not set policy on this matter" [EPA, 1992a]); rather, the new exposure guidelines discuss some of the issues and procedures that should be considered as part of the choice of the methods and criteria. The HEEE is "a plausible estimate of exposure of the individual exposure of those persons at the upper end of an exposure distribution." High end is stated conceptually as "above the 90th percentile of the population distribution, but not higher than the individual in the population who has the highest exposure.'' As is implied by those statements, the new guidelines have adopted the use of individual exposure distributions, and the HEEE is a value in the upper tail of that distribution. The exact percentile for the HEEE that should be picked from the exposure distribution is not specified, but, according to EPA, should be chosen to be consistent with the population size in the particular application. The TUBE is a "bounding calculation that can easily be calculated and is designed to estimate exposure, dose, and
risk levels that are expected to exceed the levels experienced by all individuals in the actual distribution. The TUBE is calculated by assuming limits for all the variables used to calculate exposure and dose that, when combined, will result in mathematically highest exposure or dose. …" In addition, calculation of the TUBE includes using a limiting case for the exposure-dose and dose-response relationships in calculating risk.
To be responsive to the concerns raised in the NRC (1991a) report, EPA changed its approach to the MEI. The TUBE is to be used only for bounding purposes and is to be superseded by the HEEE in detailed risk characterizations. Although the exposure guidelines are ambiguous in details about the determination of the HEEE, the HEEE is based on the estimation of the distribution of exposures that people might actually encounter. From the individual exposures, it is possible to develop population exposure (and risk) distributions and include uncertainty estimation, and personal-activity patterns. The details of these approaches are discussed in the applicable sections of this report (Chapters 10, 11, and 12).
The calculation of the exposure distribution for an individual requires knowledge of both the distribution of hazardous-pollutant concentrations and the distribution of times that the individual spends in places for which the concentrations are measured or modeled (time-activity patterns). For estimates of population exposure, the individual time-activity patterns are estimated for the population of the individuals that might be exposed.
The first step in exposure assessment is estimation of the quantity of toxic materials emitted by a given source. Emission characterization involves identifying the chemical components of emissions and determining the rates at which they are emitted. Although emission characterization is a necessary part of the exposure-assessment process, it is often conducted separately from exposure assessment to determine whether a given operation falls into one or another regulatory category.
Sources of Emissions
The emission rate often is considered to be proportional to the type and magnitude of industrial activity at a source. Emissions from a source might occur from process vents, handling equipment such as valves, pumps, etc., storage tanks, transfer, and wastewater collection and treatment. Process-vent emissions are released to the atmosphere from the use, consumption, reaction, and production of chemicals. Fugitive emissions are produced when chemicals "escape" from handling equipment, such as pumps and valves. Storage-tank emissions are released from the locations where chemical feedstocks or products are
stored. These emissions depend on the chemical properties of the product stored (e.g., the vapor pressure), the atmospheric conditions (e.g., temperature), the type of tank (e.g., fixed or floating roof), and the type of seal and venting used. Transfer emissions are produced as material is received from or loaded into storage tanks, tank trucks, rail cars, and marine vessels (e.g., barges and ships). When material is added to a storage tank, for example, it can displace contaminated air into the atmosphere. Wastewater collection and treatment emissions can be released into a plant's wastewater system when chemicals are processed and released from the wastewater treatment plant. In continuous processes, a malfunction (upset), startup, or shutdown of the process can result in a much greater emission than normal.
Emission Estimation Methods
EPA (1991c) has provided a detailed procedure for estimating the emissions from facilities that use hazardous chemicals. In estimating emissions, information is generally needed on the magnitude of use of given chemicals, the chemical characteristics of the chemicals, and the efficiency with which the emissions are controlled.
The EPA protocols (1991c) provide a tiered approach to emission estimations ranging from relatively simple emission factors to material balances and direct measurements. These approaches have varied accuracy in estimation and a wide range of costs.
An emission factor is a multiplication factor that allows determination of the average emissions likely to come from a facility on the basis of its level of activity (EPA, 1985b). Emission factors are calculated on the basis of average measured emissions at several facilities in a given industry (Compilation of Air Pollutant Emission Factors, commonly known as AP-42 [EPA, 1985b]).
A material balance is performed by assuming that the sum of the mass of chemical inputs minus the sum of the outputs, after all chemical changes and accumulation within the process or equipment have been accounted for, is the emission. In general, material balances produce information about emissions that depends on relatively small differences between the large numbers that characterize inputs (raw materials) and outputs (finished products, byproducts, and other wastes).
Emissions can be estimated with calculation methods presented in EPA (1988d) publications, such as Protocols for Generating Unit-Specific Emission Estimates for Equipment Leaks of VOC and VHAP (used for fugitive emissions). This emission-estimation method allows the development of site-specific emission factors based on testing a statistical number of sources at a facility. These site-specific emission factors can be used to develop emission estimates in the future.
Ideally, emissions from a source can be calculated on the basis of measured
concentrations of the pollutant in the source and the emission rate of the source. This approach can be very expensive and is not often used. The emission rates, characteristics of the source facility (stack height, plume temperature, etc.), and local topography (flat or complex terrain) are used to estimate the ambient concentrations of the hazardous pollutants to which people can be exposed.
The concentration of a given pollutant can be measured in each microenvironment. A microenvironment is a three-dimensional space with defined boundaries of which contaminant concentration is approximately spatially uniform during some specific period (Sexton and Ryan, 1988). There have been substantial improvements in analytic methods to measure concentrations, as described in a 1991 NRC report (NRC, 1991a). Modern methods in computerization of instruments, data recording, and data processing also permit much greater capability to obtain detailed information on the temporal and spatial variability of contaminants over a range of microenvironments. Other substantial improvements have enhanced the utility of personal monitors, which are worn by subjects directly and record the concentration or collect time-integrated samples of specific pollutants with which the wearers come into contact for specific intervals. For example, assessment of exposure to radiation has long made use of inexpensive, accurate, integrating dosimeters that were first developed when research on radioactive materials and the use of radioactivity were expanding rapidly. There are often substantial variations in the spatial distribution of radiation within a microenvironment, so individual dosimeters have been thought to provide the best estimates of individual exposure. Individual monitoring and extensive microenvironmental measurements are not generally practical for assessing exposures of the general population, but because of cost and the unwillingness of individuals to participate in exposure assessments, new instruments, including passive dosimeters for airborne chemicals, are likely to permit a similar strategy. These methods have been used in the TEAM studies (Wallace, 1987) to examine the total exposure of individuals to a number of volatile organic compounds in several locations around the country. This approach to exposure assessment has been applied in other research studies. One important finding of the TEAM studies (and others) is that substantially greater exposures to many contaminants occur indoors, both because of the higher concentrations and because most people spend considerably more time inside.
Although field measurement studies are generally expensive and require careful planning, organization, and quality-assurance programs, measurement programs can provide the large amounts of high-quality data needed to characterize environmental systems, to estimate exposure, and to develop, test, and evaluate models for evaluating exposure. Documented reliable models can then be used in place of more expensive, direct measurements. Reliable measure-
ments are generally needed to provide knowledge of emissions of chemicals that give rise to human exposures. However, measurements provide only information on the current status of the system. To allow for a broader range of meteorologic conditions, estimate the effects of changes in plant operating capabilities and procedures, or estimate the effects of an accident or upset condition, models are needed to estimate emissions and the transport of emitted materials in the atmosphere.
Modeling Used In Exposure Assessment
Mathematical models used in exposure assessment can be classified in two broad categories: models that predict exposure (in units of concentration multiplied by time) and models that predict concentration (in units of mass per volume). Exposure models can be used to estimate population exposures from small numbers of representative measurements. Although concentration (or air-quality) models are not truly exposure models, they can be combined with information on human time-activity patterns to estimate exposures.
Air-quality models are also used to predict the fate, such as deposition or chemical transformation, of atmospheric pollutants to which people can be exposed indirectly (e.g., through deposition of pollutants from air onto surface water followed by bioaccumulation in fish). Such models are central to risk assessment (see Figure 3-1). They constitute the only method of determining the total impact of diverse emissions on air quality and are key tools in assessing the impact of specific sources on future air-pollutant concentrations and deposition.
Modeling Airborne Concentrations
Mathematical air-quality models used in air-pollution analysis are in two classes: empirical and analytic. The former type of model statistically relates observed air quality to the accompanying emission patterns, with chemistry and meteorology included only implicitly. Although they hold promise for use in some aspects of air-pollutant risk assessment, these models are not commonly used by EPA in its risk-assessment practice and will be discussed later. EPA and others more commonly use the form of analytical models, in which analytic or numerical expressions describe the complex transport processes and chemical reactions that affect air-pollutant concentrations. Pollutant concentrations are determined as explicit functions of meteorologic and topographic characteristics, chemical transformation, surface deposition, and source characteristics. In exposure assessments of air pollutants, the most widely used set of models has been the class called Gaussian-plume models. Gaussian-plume models are derived from atmospheric diffusion theory assuming stationary, homogeneous turbulence or, alternatively, by solution of the atmospheric-diffusion equation assuming simplified forms of the effective diffusivity (Seinfeld, 1986). Within the limits of the simplifications involved in their derivation, they can describe the individual processes that affect pollutant concentrations, such as diffusion, bulk transport by the wind, and deposition. These models are a type of a much broader family of models called dispersion or atmospheric-transport models. See Appendix C for more information.
Modeling Multimedia Exposure to Air Pollutants
In some cases, exposure to toxic pollutants emitted into the atmosphere occurs by pathways other than, or in addition to, inhalation. An example is deposition of metals like mercury in surface waters followed by the bioaccumulation of methyl mercury in fish and then ingestion of contaminated fish. Another is exposure of an infant ingesting the breast milk of a mother exposed to a toxic pollutant, such as polychlorinated biphenyls; this can be an important route for lipophilic compounds (NRC, 1993e), and EPA has investigated it in some exposure assessments. Recent studies (Travis and Hattemer-Frey, 1988; Bacci et al., 1990; Trapp et al., 1990) have also found significant bioaccumulation of chemicals from the atmosphere in plant tissues, particularly of nonionic organic compounds. These studies have found that the degree of bioaccumulation depends on solubility, and models for the uptake have been developed (Stevens, 1991). Such "indirect" pathways can concentrate pollutants and thus result in significant increases in exposure.
Multimedia exposure and indirect exposure have been considered more frequently in hazardous-waste site (e.g., Superfund) cleanup than in the management of exposure to industrial air pollutants. One example of multiple-path
exposure to a source of primary air pollutants conducted by EPA is found in Cleverly et al. (1992). Multiple air pollutants, including heavy metals and organic chemicals, were followed after emission from a municipal-waste combustor. Atmospheric transport and deposition were modeled with a Gaussian-plume model modified to include wet and dry deposition. Other models were used to assess pollutant concentrations in nearby bodies of water; bioaccumulation; consumption of animal tissue, plants, and water; soil ingestion; and total potential dose.
Alternative Transport and Fate Models
The 1992 EPA guidelines for exposure assessment offer an approach to selection and use of models to estimate transport and fate, as well as exposure, so a variety of models can be used. For rapid screening analyses, Gaussian-plume models are adequate for limited distances around the source. However, for a more complete characterization of the distribution of concentrations downwind of a source, more refined modeling approaches may be needed.
In recent years, stochastic modeling of atmospheric dispersion has increased in popularity because of its relatively simple concept, its applicability to more complicated problems, and the improvements in computer capability and costs that make such models practical. Stochastic models can easily incorporate real physical phenomena, such as buoyancy, droplet evaporation, variations in the dispersity of released particles, and dry deposition. Stochastic modeling is typically implemented as a numerical Monte Carlo model in which the movement of a large number of air parcels is tracked in a Lagrangian reference frame. The concentration profile is then obtained from the air-parcel positions.
Boughton et al. (1987) described a Monte Carlo simulation of atmospheric dispersion based on treating either parcel displacement or parcel velocity as a continuous-time Markov process (a one-step-memory random process like Brownian movement). They simplified the problem by restricting themselves to crosswind-integrated point sources and assumed that dispersion in the mean wind direction is negligible. Thus, they reduced the problem to a one-dimensional model. Liljegren (1989) extended the model to incorporate both horizontal and vertical dispersion perpendicular to the mean wind direction. He found good agreement between the results of the three-dimensional stochastic model with concentration data found in the literature. Recent measurements of the dispersion of ground-released smokes and obscurants have shown excellent agreement of his stochastic model both with the average concentration values, including the profile across the plume, and with the time-varying concentrations observed (pers. comm., W. E. Dunn, U. of Illinois, 1988). It appears from those results that stochastic models offer considerable improvement over conventional Gaussian-plume models. Thus, there will soon be a substantially improved ability to predict average and time-varying ground-level concentrations.
Exposure occurs when someone is in contact with a substance for some period. To estimate exposures, it is necessary to estimate the time spent in various activities that provide the opportunity for exposure. Figure 3-2 shows one such analysis. Various methods are available (NRC, 1991a), including recording of activities in a time-use diary (which might be automated to facilitate the recording of locations at specific times of the day and might use questionnaires to help reconstruct kinds and duration of activities). Some participants are careful in recording their activities; others might not provide accurate accounts, because of oversight or carelessness. The framing and wording of questionnaires can substantially affect the results of a survey and thus bias the resulting estimates of time spent in various activities and locations. Further work in the measurement and modeling of time and activity is needed; research recommendations were presented in an earlier report (NRC, 1991a).
The 1992 guidelines call for the development of distributions, instead of point estimates, for exposure parameters. It is the exposure-prediction models that combine microenvironmental concentration estimates with information on time-activity patterns of people to estimate individual exposures or the distribution of individual exposures in a typical population. Activity patterns and microenvironmental concentrations can both the measured or modeled. Microenvironmental concentrations and activity patterns can vary from person to person, and from period to period. Three types of models have been developed to esti-
mate population exposures: simulation models, such as the simulation of human air pollution exposure (SHAPE) model (Ott, 1981, 1984; Ott et al., 1988) and National Ambient Air Quality Standards (NAAQS) Exposure Model (NEM) (Johnson and Paul, 1981, 1983, 1984), the convolution model of Duan (1981, 1987), and the variance-components model of Duan (1988) and Switzer (1988) (see Appendix C for additional information). The development of total-exposure models is one of the advances in modeling.
Several of the models for predicting exposures assume some correlation between measured contaminant concentrations in a microenvironment and the time spent by the exposed person in that space. Studies by Duan et al. (1985) suggested, on the basis of data from the Washington, D.C., carbon monoxide (CO) study (Akland et al., 1985), that there is no correlation between CO concentrations and time. However, there will be problems in existing models if occupancy times and concentrations of other contaminants correlate, as they might for irritating toxicants, such as formaldehyde.
Current exposure models use a variety of crude assumptions about the constancy of concentrations in microenvironments, the human activity patterns that determine the amount of time people spend in each microenvironment, and how representative the sampled population is to the total population that might be exposed to a contaminant.
Long-Term Exposure Modeling
Modeling very-long-term exposures, as is required for cancer risk assessment, presents several major difficulties. The current practice is to measure or model the concentration of a contaminant at one time and determine lifetime exposure by multiplying that concentration by a fixed number of years, e.g., the lifetime of an exposed person. However, the nature of exposure sources (e.g., changes in industrial processes) and activity patterns can change substantially over a lifetime. New sources or uses of sources can be introduced into the environment (e.g., the spreading use of wood-burning stoves), and old sources can be eliminated or modified (e.g., by the use of catalytic converters in motor vehicles). Typically, large facilities have a design life of 30 years, so considerable change can be anticipated in sources over the 70 years of a typical lifetime-exposure calculation.
Time-activity patterns of people can also vary substantially over very long periods. In the United States, people generally change their place of residence frequently, although some live in the same place over a lifetime. Population mobility can have a large impact on exposure assessments of agents, such as radon, that require reasonable estimates of long-term and highly variable exposure concentrations.
A person's activity pattern changes from childhood through young adulthood to middle and old age. Some efforts have addressed age-related differenc-
es in exposure that arise because of age. However, that aspect of variability in exposure over long periods has generally not received much attention in exposure modeling.
Short-Term Exposure Modeling
The typical steady-state airborne-concentration models are not able to provide estimates below 1-hour averages and have difficulty in modeling concentrations that vary widely over time and that can lead to short-term high exposures. If an exposure model is to estimate the effects of peak exposures on sensitive populations, the concentration model must provide reliable estimates for the time scales needed. There have been some important developments in stochastic models that could provide such estimates, but these developments have not yet been incorporated into the procedures for estimating exposure.