By Stuart Batterman and Yu-Ling Huang (Click here to view the entire P&R issue)
This article explores how to determine whether disproportionate impacts have occurred or are likely to occur from facilities handling or emitting hazardous substances. It emphasizes the use of risk assessment and other technical assessment techniques in these determinations. A number of suggestions and issues are highlighted to ensure that equity concerns are adequately addressed in environmental analyses.
Environmental impacts and health risks from hazardous chemicals and industrial facilities in communities are evaluated using three general approaches: monitoring, modeling and proxy techniques.
Monitoring. Environmental monitoring of air, water and soil is often used to indicate the presence and extent of contamination and to identify some environmental impacts. For example, a local air quality monitoring site can quantify levels of particles (e.g., lead) and gases (e.g., sulfur dioxide gas, carbon monoxide). More sophisticated monitoring approaches to assess exposure and dose are available but rarely used—for example, indoor air sampling, personal monitoring and sampling of breath, blood, urine, hair, etc. (called “biological monitoring”), from which chemicals exposure may be inferred. Monitoring can produce exposure estimates free of the many assumptions necessary using models and other techniques discussed below.
However, monitoring is only relevant for facilities or contaminants already present in communities: a proposed facility cannot be evaluated in this manner. There are other important limitations. Serious questions can be raised regarding the extent and a~curacy of available monitoring data, the ability to identify specific sources contributing to the pollution (called “source apportionment”), the inability to estimate long-term impacts from the “snapshot” of monitoring information usually available, and the need to interpret monitoring information in order to gauge its health significance.
Modeling. Environmental and health impacts may be predicted by mathematical modeling processes that govern the distribution of contaminants in the environment, and by modeling the human uptake and dose-response relationships, based on toxicological studies. Well developed models exist for contaminants in air, surface water, ground water and soil. To estimate impacts at a specific location (called “receptors”), these models take input data regarding pollution sources (e.g. type and quantity of emissions) anc combine them with site-specific data (e.g., the local meteorology, topography and hydrology).
Three model methodologies are mos’ relevant to evaluate disproportionate impact. First, exposure assessments evaluate the nature and severity of chemical exposures to specific groups. Exposure assessments are routinely performed, for example, in workplace settings by industrial hygienists. Second, environmental impact assessments evaluate the probable and possible effects of construction projects, chemical use and other actions. Such assessments are routinely conducted by state, federal and international organizations. Third, risk assessments combine exposure assessment information with chemical toxicity and demographic information to estimate the potential health threats from the manufacture, use and disposal of chemicals. Risk assessments have been extensively employed in the last decade, especially at toxic waste sites, due to EPA requirements.
Economically disadvantaged and some racial populations tend to have higher incidence of cancer and other diseases. The exposure, uptake and effects of toxic chemicals in an individual—and in a population—depend on many factors. For example, the absorption of lead is higher in poorly nourished individuals. Also, fish-eating populations may have greater exposure to certain contaminants. In most cases, however, insufficient information exists to separate environmental risks by race, income and other factors. Instead, risk analyses use “default” or average assumptions designed to represent typical or sometimes conservative parameters for dietary and other factors.
These exposure, environmental impact and risk analyses, called technical analyses in this article, are very flexible. They may be used to estimate historical impacts and predict future impacts. Individual pollution sources as well as multiple sources can be evaluated. The analyses are imperfect, of course, subject to limitations of available knowledge and data. The cumulative uncertainty grows as the analysis grows in complexity. Consider, for example, the dispersal of a contaminant from a smokestack to ambient air, the deposition of pollutants from that air to soil and vegetation, the incorporation of those contaminants in forage, their ingestion by cattle and dairy cows, consumption of contaminated beef or milk, and breast-feeding of fat-soluble contaminants to infants. In some cases, these steps are incompletely understood, and many processes are likely to be site-specific. Further, consider that food preferences may vary by race, and nutritional status and possibly genetics may alter the uptake and effects of the same contaminant in different individuals. The number of “exposure pathways” and other “mode parameters,” such as the populations being considered, must be held to some reasonable bound to make analysis manageable, communicable to interested parties and feasible, given time and cost constraints.
Proxy. The third approach to determine pollution impacts is based on measures of industrial activity or location. Most environmental equity studies have used this approach, with proximity to a pollution source or the total tonnage of emissions in the county being the major criteria identifying potentially affected populations. While convenient, proximity and tonnage may not be a good indicator of impacts, especially for elevated air pollution sources such as smokestacks that disperse pollutants over large distances. In addition, these measures do not account for the toxicity of contaminants, which varies over an enormous range. On the other hand, proximity may be a reasonable surrogate for noise, visual impacts (aesthetics) and possibly stress.
Summary. The development and analysis of monitoring, modeling and proxy measures represent technical exercises that by themselves cannot be expected to provide all information needed to address and resolve an issue. However, completing and understanding them offers several advantages: The analyses should provide an explicit presentation of assumptions, parameters and data. Environmental impact and risk assessment studies are designed to be comprehensive, although important processes (pathways) accounting for the major portion of the risk or pollution burden may be emphasized at the expense of minor processes. They are scientific in the sense that “objective” techniques are used, i.e., informed and rational persons should reach similar numerical estimates of impacts (with similar assumptions). Finally, because the technical analyses should be explicit, comprehensive and objective, they form a potentially effective communication vehicle aiding discussion and hopefully resolution among the parties involved.
Determining Disproportionate Impact
The investigation and possible determination of a disproportionate impact is a straightforward extension to a monitoring, modeling or proxy analysis. In essence, demographic information, giving population and racial distributions, is simply coupled to the exposure, impact or risk estimate.
Demographic information. Demographic information is collected and distributed by the Bureau of the Census, and 1990 information is available on CD ROM. Major updates are performed every ten years. Census information includes population, racial composition, housing, income, occupation and zip code. Information at the “block” level gives the most detailed and highest spatial resolution. Each block typically contains 250 to 550 housing units. The next level of aggregation, “blockgroup,” aggregates a number of blocks and may constitute the most relevant scale for determining disproportionate impacts for chemical exposures, although most studies have aggregated these data at much larger scales, e.g., by zip code and county. The most appropriate Census information to use should match the spatial scale of the technical analysis. For example, while air pollution concentrations resulting from an incinerator may be slightly elevated over a radius of several miles, localized pollution hotspots that are much smaller—perhaps 0.1 to 0.25 miles across—may occur. This area more closely matches blockgroup data; thus, a more accurate reflection of the minority population affected will be obtained.
Impact measures. The next step in assessing disproportionate impact is to select an appropriate measure of chemical exposure or risk. At least two different measures should be evaluated. These should examine pollution hot-spots that indicate both maximum individual risk and more broadly dispersed pollution patterns and demographics that show population risk. Either is sufficient to show disproportionate impact.
Individual risk—hotspot analysis. The first approach follows EPA guidance for conventional risk assessments aimed at evaluating worst-case scenarios. It involves examination of the racial composition in pollution hotspots or the risks to the “most exposed individual.” For an air pollution impact, for example, the population in the pollution hotspot downwind of the source should be examined. If the ground water is polluted by fuel or chemical spills and local (residential) wells are used, then the composition of the population near and downstream of the spill should be examined. If risk occurs due to the accumulation of pollutants in fish, for example, then the specific populations eating contaminated fish should be examined. In all cases, disproportionate impact is demonstrated by a high minority fraction in the population currently or potentially affected by the pollution in the hotspot area. The impact may be disproportionate if this fraction exceeds that in the larger region, state or county.
Population risk. Disproportionate impact may also be shown if a minority population bears most of the aggregate risk experienced by the affected population, rather than just the population in the hotspot. Many environmental contaminants present risks at low levels, and while the calculated risk will be reduced, exposure of large populations to a low level of contaminants still increases the expected incidence of harm among the population. Incidence refers to the fraction or number of individuals affected, e.g., the number of excess cancer cases. For example, if the pollution exposure is half as much in one area as another, but the population density is double, then the incidence of disease or death will be the same, all else being equal. The analysis of the pattern of disease in a population has tremendous relevance to public health, as all potentially affected individuals are considered, not just those in an identified hotspot. Because a population analysis employs exposure or risk information and demographic statistics over a broader area, it is more complex than the hotspot analysis.
An example will clarify distinctions between the two approaches. The Flint woodwaste facility described in Kary Moss’ case study in the May-June P&R was located near an elementary school and minority neighborhood. But the pollution hotspot from this facility occurred several miles away in a largely non-minority, but sparsely populated, area. The precise location, in fact, fell within a county recreation area. However, a portion of urban Flint with a population approximately ten times that in the hotspot area was exposed to a pollution level roughly one-third that of the hotspot. Because this was primarily a minority population, disproportionate impact is demonstrated, as the expected incidence of risk will be approximately three times higher among minorities than in the non-minority population.
De minimis impact. The methods suggested above for determining disproportionate impact should require that the severity of the potential or the existing impacts be above de minimis levels. That is, impacts due to pollution in the hotspot or elsewhere should not be minimal or insignificant, but should result in a meaningful likelihood of harm. This determination can raise numerous questions that should be dealt with in a risk assessment—for example, the amount of toxic chemicals released, their toxicity, the exposed population, etc. The determination of disproportionate impacts can be credible despite large uncertainties in estimating the existing or potential harm.
Most technical analyses require the use of many quantitative parameters. These numbers may be uncertain and controversial. Take incinerators, for example: emissions of greatest health significance may include various gases (carbon monoxide, nitrogen oxides), metals (arsenic, cadmium, lead) and organic compounds (benzo(a)pyrene, dioxin). Emission rate measurements of certain gases are easy, inexpensive and required by regulators, but measurements of many other gases and particles are difficult, imprecise, expensive, infrequent and sometimes unavailable. Moreover, emission rates are a function of the fuel and waste burned and operating conditions, factors which can vary significantly. Finally, for a new facility, emission rates must be predicted by analogy, modeling or engineering estimation, and the applicability and reliability of the predicted rates for the site-specific conditions are not always clear.
Thus, the emission rates of many pollutants for both existing and proposed facilities are uncertain. In some cases, the uncertainty of the emission rate is relatively- small, e.g., factor of two. In others, the uncertainty may be a factor of 10, 100 or more. Predictions of the severity of environmental and health impact are proportional to the emission rate; thus, the severity of possible impacts may vary over a very large range. However, the location of the hotspot will not vary. Thus, disproportionate impact exists if a high fraction of minorities live, work or go to school in a high-impact location where there is a reasonable likelihood that impacts will be above de minimis levels.
There is little guidance or precedent that indicates what is a reasonable likelihood of impact. In a somewhat similar application, EPA guidance suggests a 95% confidence level for risk assessments designed to protect public health around Superfund sites. This confidence level is usually factored in the dose-response calculation, that is, the exposure-to-risk extrapolation. The guidance recognizes that information regarding the toxicity of chemicals, based largely on animal studies, may not be directly applicable to humans, and that people vary considerably in their responses to chemicals.
However, no standard approach exists for adjusting most parameters in technical assessments, e.g., emission data. In instances where a facility already exists, the one or two emission tests often available may not permit confidence levels to be established. While a range of emission estimates is sometimes presented, procedures to estimate confidence levels in predicted emission levels have not been widely employed. Such estimates are not trivial to perform. In the Flint facility, for example, the amount of lead emitted depends on the fraction of demolition debris burned, the lead content in this debris, the effectiveness of hand-sorting to remove undesirable material, the amount of lead in vapor and particulate phases after combustion, and the efficiency of the air pollution control system.
Finally, even less guidance exists regarding de minimis levels of harm where population risk is considered. No new issue arises if the likely exposure or risk is above de minimis levels. This will generally not be the case for the larger population, since such a showing would clearly indicate unacceptable impacts without consideration of environmental equity. Consider, however, whether disproportionate impact occurs if individual risks fall below commonly used de minimis levels (say, one-in-ten-thousand to one-in-a-million chance of cancer), but a large population is exposed (say, hundreds of thousands or millions); thus, excess morbidity or mortality is expected in a minority group. Clearly, the number of people affected influences policy judgments, but exactly how many people must be affected to constitute disproportionate impact? It should be further noted that population risks should reflect expected risks, yet most risk analyses are protective in nature, using conservative assumptions. In some cases, the information necessary to calculate expected risks is not available.
Conclusion
The use of technical analyses like risk assessment in environmental equity studies to evaluate disproportionate impacts is feasible and has been demonstrated. It represents a logical extension of environmental assessment methodologies. While some work to develop appropriate guidelines and information is necessary, these methodologies can provide flexible, objective, comprehensive and transparent techniques that can be used to investigate potential harm to minorities from pollution sources.
Stuart Batterman is Associate Professor, Yu-Ling Huang a doctoral student in the Departmnet of Environmental and Industrial Health at the University of Michigan.
