Environmental Cleanup at Navy Facilities: Risk-Based Methods (1999)

From the preceding chapter it should be apparent that a major concern with using risk-based methodologies for hazardous waste cleanup is the uncertainty associated with the risk assessment and risk management processes. Adherence to resource conservation, on the other hand, involves less uncertainty, because, ultimately, contamination is not left in place (although there may be episodic uncertainties regarding the effectiveness of source removal actions). This chapter explores the types of uncertainty encountered during risk assessment, including those associated with sources, pathways, and receptors. A case study illustrates the trade-offs of doing an uncertainty analysis versus using conservative cleanup goals. The chapter closes with an examination of the uncertainties associated with remedial options, including both technical and non-technical solutions.

Uncertainty in risk assessment can be dealt with in one of three ways: (1) using conservative cleanup goals, (2) conducting extensive long-term monitoring to reduce uncertainty, or (3) conducting a quantitative analysis of the uncertainty in the risk assessment. An analysis of existing risk-based approaches indicates that most do not adequately address uncertainty (ASTM, 1995, 1998; CA SWRCB, 1997). Decision-makers may recognize uncertainty in risk analysis, but there has been little agreement on how that uncertainty should be incorporated into the risk management process. Almost all risk-based approaches have traditionally relied on the use of conservative assumptions and cleanup goals to account for uncertainty. For example, an "upper-bound" point estimate of risk is often specified. This approach, unfortunately, does not convey the degree of

confidence or uncertainty in the risk estimate, and it provides no information about how likely that risk may be.

Long-term monitoring has also been historically conducted on a regular basis, although perhaps not with the stated intention of reducing uncertainty. Long-term monitoring is effective at reducing uncertainty because it helps to further refine the assumptions on which the risk assessment was based. It can also reveal the effectiveness of remedial actions and warn of potential increases in risk over time. However, long-term monitoring cannot reveal the relative importance of different sources of uncertainty, nor can it evaluate how different management strategies might alter those uncertainties.

Although not frequently conducted, a systematic, quantitative assessment of the full range of risk uncertainties and their implications for risk management has significant advantages over the previous two options. Formal uncertainty analyses of risk estimates can actually help to inform decision-makers and the public about the level of conservatism contained in the risk assessment. They can also give risk managers an opportunity to describe the knowledge and rationale used to develop the risk estimate. Such analyses are useful for identifying the uncertainties that create the greatest differences in risk estimates (as the case study presented in this chapter demonstrates). Research or data collection can then be directed toward reducing these major uncertainties. Recent trends indicate that uncertainty analyses may become more commonplace during risk assessment and management (Finkel, 1990; Morgan and Henrion, 1990; NRC, 1994a, 1996; Browner, 1995). The first half of this chapter discusses the major uncertainties to consider when conducting an uncertainty analysis during risk assessment.

Three conditions are necessary for a risk to exist: a source of contaminants, one or more pathways for contaminant migration, and receptors that are susceptible and exposed to the contaminants. There are significant uncertainties associated with each of these conditions that are briefly described in the following sections and listed in Table 4-1.

Pollutants originate in source areas, either primary or secondary. Primary source contaminants enter the natural environment as a result of specific activities. Examples include chemical storage and transmission (tanks, drums, and pipelines), facility operations (manufacturing, weapons training, civilian services), waste management operations (landfills, impoundments, or sludges), or other direct sources. Secondary sources refer to contaminants that have been transferred from primary sources to the adjacent natural environment and can serve as contaminant "reservoirs." Examples include residual nonaqueous phase liquids (NAPLs) in either the saturated or unsaturated zones and soils containing adsorbed or precipitated contaminants.

A principal uncertainty in carrying out a risk assessment stems from various

unknowns associated with the contaminant source. Often, there is a lack of information about (1) the location of the contamination source; (2) the chemical composition of the contamination; (3) the amount of contaminant released; (4) the time release history of the contaminant; and (5) the present mass and concentration distribution of the contamination.

Some of these uncertainties are amenable to significant reduction. Extensive, focused studies can usually reveal the composition of the contamination, the location of the source, and the approximate mass distribution of the contaminants. This ability to reduce uncertainty underscores the importance of site characterization in any risk-based methodology. Reducing the other uncertainties, though, can be much less straightforward. Typically, only bounds can be put on the quantity of contaminants present and the history of release, because records of facility waste disposal or materials usage are often poor, and leaks may have been undiscovered for long time periods.

There are also uncertainties in source characterization brought about by measurement errors during data collection. For example, delineating a primary or secondary source in three dimensions requires extensive sampling efforts, and a potential source of uncertainty lies in possible errors associated with the location, density, and handling of samples. Fortunately, these sources of uncertainty may be assessed (or even reduced through improved techniques). Thus, sampling and measurement errors may be considered independent of other types of uncertainties in source characterization.

Figure 4-1

The spatial distribution of a bromide plume in the subsurface 174 days after injection of the bromide. SOURCE: LeBlanc et al., (1991).

Finally, variability plays a role in the uncertainties encountered during source characterization. As discussed in more detail later, aquifer soil and rock properties are spatially heterogeneous. Variability in aquifer structure gives rise to variability in the subsurface distribution of any contaminant spill. For example, Figure 4-1 illustrates the spatial distribution of a bromide contaminant plume at the U.S. Geological Survey's Cape Cod research site 174 days after injection. Extensive site characterization revealed a contaminant distribution that was highly irregular as a result of variability in aquifer properties. Because acquiring complete knowledge of the soil or rock properties is not possible, a complete description of the concentration distribution of contaminants in the subsurface can never be achieved. However, increasing the sampling of the subsurface (as demonstrated in Figure 4-1) can result in more effective delineation of contaminant plumes.

Pathway characterization with contaminant fate-and-transport models is an essential ingredient of risk assessment and management. The purpose of fate-and-transport modeling is to determine contaminant concentrations at human or ecological receptors as a function of time, given some measured or assumed source concentration. The receptor concentrations are used in exposure modeling to determine the health risk.

Development of quantitative fate-and-transport models requires understanding the physical, chemical, and biological processes that control the transport and fate of contaminants along pathways. These processes have been described extensively in the ground-water literature (e.g., NRC, 1990; Mercer and Faust, 1981). The physical processes of advection, diffusion, and dispersion are responsible for transporting the contaminants. Chemical processes (radioactive decay, sorption, dissolution/precipitation, complexation, and volatilization) and the microbial degradation of organic compounds redistribute mass among the solid, liquid, and gas phases.

Once the prevailing fate-and-transport processes are understood, a conceptual model of the system and its behavior is developed. This conceptual model leads to a mathematical model that can predict contaminant concentrations in space and time. Examples of fate-and-transport models used in risk assessment can be found in the ASTM RBCA standard guides (Appendixes 2 and 3) (ASTM, 1995, 1998) and other guidance documents (Moskowitz et al., 1996).

Table 4-2 summarizes the types of information required for modeling a contaminated site, all of which have some degree of uncertainty in their obtained values. Uncertainties in subsurface fate-and-transport modeling arise from the complexities of the process and the system being modeled, and because limited information is generally available about the site. Many of the uncertainties of fate-and-transport modeling can be reduced by collection of more data. How-

ever, the natural variabilities of aquifer parameters cannot be reduced, although their effects on model outputs can be assessed.

Aquifer heterogeneity and the spatial variability of parameters. The main feature of aquifers that gives rise to uncertainty in fate-and-transport modeling is the intrinsic heterogeneity of aquifer materials. Natural soils and fractured rock formations are highly variable because of fluctuations in the geologic and climatic processes that give rise to their formation. Hydraulic conductivity (or permeability), the most important aquifer property controlling water movement, has been found to vary over several orders of magnitude over small distances (LeBlanc et al., 1991; Rehfeldt et al., 1992).

Although the heterogeneity of subsurface materials affects source characterization by hampering a determination of the spatial extent of contamination, its effects on contaminant migration are more far-reaching. Not only do subsurface heterogeneities affect the physical movement of contaminants in the subsurface but they also affect the chemical form of those contaminants, the local sorptive capacity of the aquifer for the contaminants, and chemical and biological rates of contaminant degradation.

Variability in the hydraulic conductivity of subsurface materials directly affects the dispersive properties of an aquifer (i.e., how the aquifer material causes a contaminant to mix with the native ground water). When hydraulic conductivity varies over several orders of magnitude in a short distance, complex contaminant distributions can result, with portions of the contaminant plume traveling at differing velocities. Measuring hydraulic conductivity distributions in spatial detail and their effect on contaminant dispersion is, for all practical purposes, an impossible task. However, modeling methods can be used to consider the effect of hydraulic conductivity on contaminant transport (Gelhar, 1993).

In addition to physical heterogeneity observed in the subsurface, investigators have documented accompanying mineral or chemical heterogeneity. The distribution coefficient (K_d), which is used to quantify the extent of sorption to subsurface materials, has also been postulated to be correlated with hydraulic conductivity (Robin et al., 1991; Foster-Reid, 1994). The spatial variability of field-scale biotic and abiotic decay processes is also likely due to the variability in substrate and nutrient distribution in the subsurface. For example, variability of substrates, such as natural organic matter, benzene, and toluene, has been shown to impede determination of an apparent biodegradation rate constant for a given subsurface condition (Chapelle et al., 1996).

Geologic boundaries. Basic uncertainties exist regarding how to represent mathematically the geologic boundaries of a contaminated site. Necessary information includes whether the subsurface is porous or fractured, whether it is confined or unconfined, whether it can be represented as single or multiple aquifers, and the location of bedrock or other confining layers. Although this kind of

information is unknown at the beginning of a modeling project, it can be approximated with such routine methods as field reconnaissance and examination of any existing geologic maps. Therefore, this source of uncertainty can be reduced with a known level of confidence.

Hydrologic boundaries and recharge. Constant-head boundaries, such as streams, lakes, and the water table, exhibit temporal variability, a source of uncertainty that may be difficult to quantify and will require examination of historical hydrologic records. The temporal and spatial variability of recharge to an aquifer due to rainfall is also an important phenomenon that must be quantified because of its significant effect on the downward migration of contaminants. Although it is fairly easy to assess temporal variability of precipitation from rainfall records, spatial variability cannot generally be assessed.

Measurement errors. Field measurements used as direct inputs to fate-and-transport mathematical models, such as aquifer properties, hydraulic head, and solute concentration, are all subject to uncertainty because of errors associated with making such measurements. These sources of uncertainty can usually be quantified.

Parameter estimates. The parameters of ground water flow and solute transport models are hydraulic conductivity, specific storativity, effective porosity, dispersivity, matrix diffusion, distribution coefficients, decay coefficients, and reaction rates. These parameters are inferred by fitting mathematical models to measured head and solute concentration data. If there are errors in the measured values of head and concentration, these errors will affect the calculated parameter values, which will give rise to uncertainty in the risk assessment.

Like the uncertainties in source characterization, many uncertainties in pathway characterization can be reduced with intensive data collection. However, for dissolved plume and vapor transport pathways, the most important uncertainty is the variability of aquifer properties. Although uncertainties due to variability cannot be reduced, they can be quantified by examining the statistical properties of model inputs and the way these statistics propagate to the model output.

Models that can represent variability in their output are termed stochastic, while deterministic models generate a single output. The use of stochastic modeling implies that the input parameters, and therefore the outputs, are not known with certainty. One of the most common methods of stochastic modeling is the Monte Carlo simulation, in which various realizations of equally plausible, spatially variable input parameters are numerically generated and then the appropriate transport model is applied. The mean output, as well as the variability about the mean, is calculated. In this way, confidence limits can be specified for both the inputs and outputs, and the inherent uncertainty in the model output due to aquifer heterogeneity can be quantified. Monte Carlo analysis has been used

successfully to represent uncertainty in fate-and-transport modeling derived from the variability in hydraulic conductivity (Maxwell et al., 1998) and biodegradation rates (McNab and Dooher, 1998).

Uncertainties in human and ecological receptors arise because basic biologic processes of most organisms and systems are not well understood, exposures are not easily established and are often complex, and sufficient data are not available to estimate health effects. In general, the uncertainties in receptor characterization are as significant as those in source and pathway characterization, and they are less amenable to assessment and reduction.

Human health was the initial focus of environmental regulations and has received considerably more study than ecological health. Assessing the impact of contamination on ecological health presents additional complexities because of the variety of potentially affected organisms and their inter-relations, and the impact of contamination on habitat and natural resources.

Receptor characterization during risk assessment requires the quantifying of two parameters, both of which are characterized by uncertainty: (1) probable exposure to chemical contamination and (2) the resulting outcome (how the receptor responds to a given exposure).

Exposure. Two principal types of uncertainty are characteristic of human exposure to toxicants: uncertainty concerning the bioavailability of toxicants in environmental media and uncertainty regarding actual doses. Bioavailability refers to the actual concentration of a toxicant that is accessible by humans, which depends greatly on the chemical form of the toxicant. Whether contaminants are bioavailable to humans depends on (1) the particular medium in which the contaminant occurs and (2) a variety of co-factors that will vary among individuals and in the same individual at different times. For example, absorption of ingested contaminants is strongly affected by the presence or absence of food and micro-nutrients in the gastrointestinal tract.

Determining the actual dose received by humans is challenging. Human exposure to chemicals occurs through intake of food, air, water, soils, and dust. If contaminants have traveled through the subsurface, variable aquifer parameters, such as hydraulic conductivity, will create variable exposure patterns for receptors. This variability can be eliminated only if toxicants are measured directly at the point and time of exposure.

Humans are often exposed to complex mixtures of chemicals. Uncertainty may derive from a lack of data about the effects of certain compounds, incom-

plete knowledge about chemical reactions occurring in the mixture, and not knowing how best to approximate exposure to mixtures (either by adding or multiplying the exposures from the individual contaminants).

Human behaviors vary greatly, resulting in substantial differences in exposure even in the same environment. For example, children eat, drink, and breathe more per unit of body weight than adults, and their normal behavior results in greater ingestion of soil and dust. Variations in cultural practices may also result in varying ingestion rates. Future research on identifying exposure biomarkers may improve our ability to assess human exposure to toxicants.

Outcome. The outcome of exposure to chemicals can be as difficult to assess as the actual exposure, since, for most chemicals, the health outcome is unknown. This uncertainty is troubling because of the enormous range of known toxic effects of chemicals. Outcome uncertainties can also be caused by variability in the sensitivity to chemical exposures. Among the most important factors contributing to variability are genetic make-up, age, state of health, and previous exposures. The unique genetic make-up of individuals results in inherent differences in the ability to absorb and metabolize toxicants and to repair cellular damage. Gender, body size, and relative fat content affect the response to chemical challenges. Age is possibly the factor that results in the greatest range of sensitivity to chemicals in otherwise healthy people. Children's immature physiologic systems may prevent them from efficiently detoxifying toxicants, while the elderly have a decreased ability to repair genetic and cell damage. The state of health at the time of exposure contributes to the variability in response. Cardiac or respiratory disease, liver disease, and pregnancy can influence susceptibility to toxicants. Finally, previous chemical exposure contributes to individual variability of response and can be difficult to assess.

Dose-Response Relationship. Uncertainties arise when relating exposure and outcome with a dose-response curve. These result from a lack of information about underlying principles of chemical toxicity and the need to make estimates based on available data. Use of a dose-response relationship requires extrapolation from high-dose animal experiments to the low-dose scenarios characteristic of human exposure. When adequate human data are available, extrapolation from short-term to long-term exposure scenarios is necessary.

Of the uncertainties described above, the uncertainty of the toxic potential of unstudied or incompletely studied compounds is probably of greatest significance. Differences in chemical toxicity can vary by a billion-fold or more (Kamrin, 1990, p. 7). Among the remaining uncertainties, human variability as a function of age can greatly influence the effect of a toxicant. For reasons stated above, risk assessment models that consider only adult receptors may underestimate the impact of any particular waste site on children. This problem can be overcome by replacing single exposure models based on adult males with a series of models that

take into account children, pregnant women, and the elderly, with adjusted parameter values for food, water, and air intake and other significant physiological and behavioral parameters.

Assessing the uncertainty in human receptor characteristics can be accomplished with the same stochastic mechanisms used to model contaminant fate and transport. For example, Monte Carlo analysis can also be used to express the degree of variability of human exposure (Price et al., 1997; Maxwell et al., 1998). However, unless more animal studies, mechanistic toxicology studies, and well-defined epidemiological studies are conducted, reducing the uncertainties associated with the toxicity and dose-response curves of most chemicals is unlikely.

Most of the uncertainties that affect human health risk assessment have close analogues in ecological risk assessment. However, ecological risk assessment must address a vast array of organisms about which little is known, rather than a single species about which much is known.

Exposure. Like human receptors, uncertainty in the exposure of ecological receptors is related to the bioavailability of toxicants and to the actual doses received. Many factors influence the dose received by an organism. The mobility of organisms and the temporal and spatial variability of toxicants produce considerable uncertainty regarding the amount of a specific toxicant absorbed and its resulting internal concentration in target organs. For vertebrate animals exposed primarily through ingestion of contaminated food, rather than from direct dermal contact with environmental media, additional uncertainties relate to the chemical concentrations found in food, the ingestion rates of different foods, and the absorption of chemicals in the gut. Behavioral factors can cause variability in exposure as well, as animals may be either repelled by or attracted to contaminated media.

Outcome. As with humans, there is significant uncertainty about the way an ecological receptor will respond to toxicant exposures. These uncertainties derive from inherent limitations in the available toxicity testing methods, differences in individuals, and differences between species. Different ecotoxicological testing methods, which expose groups of organisms to known chemical concentrations, have inherent uncertainties related to extrapolating the results to the responses of intact organisms. Variability in individuals can also introduce uncertainties. The organisms tested in laboratories are typically genetically homogeneous and are raised under specific conditions. In addition to genetic differences, the age, developmental stage, nutritional status, parasite loads, reproductive condition, and previous exposure history can affect the response of an organism to chemical exposure.

The above uncertainties are essentially the same as the uncertainties in human health risk assessments. Two additional types of uncertainty are unique to ecological risk assessment, and are potentially more important. First, there is variability in the response of different taxonomic groups of organisms to a given contaminant release. The diversity of physiological processes and chemical sensitivities among green plants, fungi, arthropods, mollusks, annelids, vertebrates, and many other types of organisms is much greater than the differences between different human receptor groups. The second and more important uncertainty in ecological risk assessment concerns the ultimate effects of contaminant exposures on the integrity of ecosystems. Under normal conditions, the reproductive potential of most organisms is far greater than is needed for each generation to replace itself. This excess reproductive capacity permits populations to maintain themselves indefinitely in a variable environment. Variability outside the normal range, however, can cause irreversible collapse of ecosystems; the scale of disturbance that can be tolerated is generally unknown.

It might appear from the above discussion that successful management of ecological risks is an impossible task. However, over the past several decades, ecological risk assessors have developed a variety of strategies for identifying and dealing with these uncertainties, such as site-specific studies of impacted

Western gulls nesting on a landfill at the Alameda Naval Air Station.

Courtesy of the U.S. Navy.

ecological receptors. Although imperfect, these strategies have led to improvements both in the scientific foundation of ecological risk assessment and in the quality of ecosystems exposed to toxicants.

This chapter opened by identifying three options that could be taken to deal with uncertainty, assuming a risk-based approach is used: (1) use conservative cleanup goals, (2) perform more extensive long-term monitoring to reduce uncertainty, and (3) conduct an analysis of the uncertainty in the risk assessment. Depending on which option or combination of options is chosen, uncertainty in the risk assessment process will have a greater or lesser effect on risk management decisions.

The committee strongly supports the use of a properly designed and implemented long-term monitoring program to reduce uncertainty at contaminated sites. Long-term monitoring provides data that can be used to evaluate assumptions made during fate-and-transport modeling, and these data can be used to improve the estimates of various model parameters, thereby leading to a reduction in modeling uncertainty. Long-term monitoring is also the best way to demonstrate the effectiveness of risk management strategies that have been implemented, and to guarantee that risks do not increase over time. Because most cleanup scenarios already involve some long-term monitoring, enhancing the monitoring program to reduce uncertainty should not require substantial additional expenditures of time and money. It should be noted that although long-term monitoring is an effective strategy for reducing uncertainty, there will always be some ''residual" uncertainty about a site (unless the density of monitoring points is impractically large).

To complement long-term monitoring, the committee favors a formal analysis of uncertainty over the use of conservative cleanup goals. Uncertainty analyses allow the user to know the level of uncertainty in a risk estimate, and they reveal the relative significance of different sources of uncertainty (allowing uncertainties to be ranked). This has important implications for future site characterization efforts, for choosing the remedial option, and for design of long-term monitoring networks. As a hypothetical example of the way an uncertainty analysis might focus cleanup efforts, consider a site contaminated by a leaking underground storage tank (UST). If the UST contained highly degradable petroleum products for which the toxicological data, exposure pathways, and biodegradation pathways are well established, the most significant uncertainty may involve the quantity of contaminant released from the tank. At an identical site where MTBE is also present, the greatest uncertainty may shift to the toxicological effects of MTBE on human and ecological receptors. If the UST contains

primarily TCE, a highly mobile, relatively nondegradable substance, the variability in aquifer parameters may pose the greatest uncertainty.

Though the preceding example is highly simplistic, uncertainty analyses are complex and must be crafted on a site-by-site basis. It is possible that Navy facilities will need to call on the expertise of ground-water modelers and risk assessment professionals. Forming interdisciplinary partnerships to carry out risk assessment calculations may be a future direction that should be promoted if risk assessment calculations are to become more accurate. The decision to conduct an uncertainty analysis must be made early in the cleanup process prior to the collection and assessment of data. Ideally, the interested public should be involved in this decision-making so that the uncertainty analysis is not perceived as a way to avoid conservative cleanup goals.

Little information is available concerning the cost of conducting an uncertainty analysis. Because it relies on extensive site characterization and the expertise of professional risk assessors, it is likely to be an expensive undertaking. However, the monetary benefits to be gained from an uncertainty analysis can be substantial if such an analysis results in less overestimation of the risk than is typical when using conservative exposure parameters. This is clearly shown in the unique case study by Maxwell et al., (1998) found in Box 4-1. First, the case study demonstrates how an uncertainty analysis resulted in a substantially less stringent cleanup goal than would have been specified by a conservative cleanup goal. The study suggests that quantifying uncertainties may save money in the long run, even though the cost of modeling and data collection to quantify the uncertainties may be high up front. Second, the case study provides a clear example of how variability in the hydraulic conductivity of natural materials and variability in human receptors can be quantitatively incorporated into risk assessment calculations. The same method could be expanded to include uncertainties in other parameters. Finally, the example demonstrates that risk should not be thought of as a single value but rather as a range of values.

The uncertainties inherent in conducting risk assessment comprise a major source of the overall uncertainties in the cleanup of hazardous waste sites. Beyond uncertainties in the risk assessment calculation, though, there are more qualitative uncertainties associated with the risk management process. Decisions on appropriate treatment technologies and whether to use engineering and institutional controls in situations where cleanup is not possible are marked with significant uncertainty. Much of the uncertainty arises from factors that have already been discussed; for example, the effectiveness of some treatment technologies is uncertain because of subsurface heterogeneities. These factors, as well as the qualitative uncertainties unique to risk management, are discussed below.

One of the difficulties in managing risks from ground water and soil contamination is the uncertainty of treatment technologies and engineering controls. In making risk management decisions for contaminated ground water and soil, policy maker and risk managers have generally not had enough information to judge the capabilities of technological solutions for the diversity of conditions at waste sites. The historic premise of CERCLA is that ground water should be restored to its beneficial uses (generally to drinking water standards), and many state ground water cleanup programs require the restoration of ground water to pristine conditions (NRC, 1994b). Until recently, insufficient attention was paid to the feasibility of restoration using current technologies. Even now, evaluating the performance of treatment and containment technologies is generally limited to short-term considerations since many of these systems have not been operating for long periods of time (more than 10 years).

Determining the feasibility of technological solutions is subject to significant uncertainty for many of the reasons already discussed in this chapter. Difficulties in characterizing the subsurface affect not only source and pathway characterization but also complicate the design of effective cleanup and containment systems. Many technologies require the flushing of fluids, such as water, air, and steam, through the subsurface. Uncertainty about the physical heterogeneity of the subsurface complicates predictions of the flow paths of these treatment fluids (NRC, 1994b, 1997). In addition, cleanup and containment systems often require direct contact between treatment fluids and contaminants. The difficulty in locating contaminant mass, along with inaccessibility of contaminants entrapped in low-permeability zones or in micropores of geologic materials, will limit contact between the treatment fluids and the contaminants, contributing to the uncertainty in system performance (NRC, 1994b, 1997).

Considerable attention has recently been given to the limitations of both treatment technologies and engineering controls. In a 1994 study, the NRC reviewed the performance of conventional ground water cleanup systems, known as pump-and-treat systems, in achieving cleanup standards (NRC, 1994b; MacDonald and Kavanaugh, 1994). Whether these systems would achieve cleanup standards was found to be highly uncertain. After examining 77 active pump-and-treat systems in detail, it was determined that only 8 of the systems had achieved ground water cleanup goals. Whether the remaining systems would achieve cleanup goals was unlikely at 34 of the 77 sites. The NRC found that it was common for contaminant concentrations to rebound above cleanup goals after these goals had apparently been achieved and the pump-and-treat systems shut down. Although the limitations of pump-and-treat systems for complete remediation are now widely

This Soil Vapor Extraction Treatment System is being used for removing petroleum vapors from the underlying vadose zone above the ground water aquifer.

Courtesy of the U.S. Navy.

recognized, these systems are still the most common treatment technology at large contaminated sites (used at 93 percent of CERCLA sites where ground water restoration is under way, according to the most recent available EPA data [NRC, 1997]).

As yet, no innovative ground water cleanup technology that can overcome all of the major difficulties in subsurface cleanup has been identified. In a recent review of cleanup technologies, the NRC concluded, "The current state of remediation technology development is relatively rudimentary" (NRC, 1997). Technologies are available for treating mobile and reactive contaminants (such as petroleum hydrocarbons and, to a lesser extent, chlorinated solvents) in permeable, relatively homogeneous geologic settings. However, the treatment of recalcitrant contaminants (such as metals and polychlorinated biphenyls) in complex geologic settings is subject to significant technological limitations.

Reviews of the performance of containment systems have also produced mixed results. Although shown to have limited effectiveness as a treatment technology, pump-and-treat systems that establish hydraulic barriers to isolate contaminant sources can be highly effective in preventing the spread of contaminant plumes. Of the 77 sites reviewed by the NRC (1994b), containment of the contaminant plume was achieved at 40 sites, although cleanup goals were achieved at

only 8 sites. Of the 40 sites, 31 were categorized as hard-to-clean sites because of unfavorable hydrogeologic conditions and the type of contamination. Thus, there is less uncertainty associated with the use of pump-and-treat systems when the objective is containment rather than achievement of cleanup goals.

Physical containment systems, such as engineered barriers, have also been investigated for their effectiveness. A review of 34 sites containing vertical barrier walls or caps found that only 4 systems had detectable leaks (Rajaram, et al., 1997). Of 27 horizontal containment systems using liners made of various materials, 10 were found to be ineffective (Bass et al., 1985). A clear understanding of the performance and long-term reliability of containment systems will be critical to the use of risk-based approaches to remediation, which rely less on source removal and more on engineering and institutional controls. In fact, the development of innovative containment systems has been partially attributed to the increased use of risk-based approaches (International Containment Technology Conference, 1997).

The technologies available for treating major classes of contaminants (WASTECH, 1994; NRC, 1997; WASTECH, 1998) and for containing sources of contamination (Rumer and Mitchell, 1995) are extensively discussed in the literature. Whether the technologies can remove contaminants to achieve cleanup goals or contain contaminant sources may be uncertain at the outset of remediation and may need to be determined by trial and error after remediation commences. Risk managers may need to adjust risk management goals based on the capabilities of currently available remediation technologies. This feature is discussed in Chapter 5 as an important criterion for a risk-based approach.

Several approaches can be taken to manage the uncertainties in the use of treatment technologies and engineering controls.

Use an iterative approach to remediation in which the cleanup or containment system is adjusted in response to data from a monitoring system. With an iterative approach, the accuracy of initial performance estimates can be determined and the system can be adjusted accordingly. Although this approach is presently available, in the committee's experience there is little evidence that it is used on a routine basis or written into decision documents.

At complex sites make greater use of expert panels to review cleanup plans and the feasibility of achieving cleanup goals. A panel of independent reviewers could provide guidance on the site conceptual model, design of the cleanup system, probability of achieving cleanup standards, and the need for a containment system when achievement of cleanup goals is not feasible. This level of oversight will ensure that the remedial options are more durable and scientifically defensible. It will also result in reduced operations and maintenance costs and sustained risk management results.

Develop a data base containing information on the performance of cleanup and containment systems at Navy sites that can be used to reduce uncertainties in future applications of cleanup technologies. This data base would contain information on the specifications of each contaminated site, design parameters for the cleanup or containment system(s) employed, and performance and cost data. The EPA has data bases of RODs and other technologies, but they contain limited information on system performance and cost.

Institutional controls are an essential element of the risk-based approach to remediation. When the elimination of pathways or removal of receptors is used to reduce risk, rather than removal of the source itself, there must be some mechanism to ensure an adequate margin of safety for the life of the hazard. These mechanisms often are controversial, because the contamination problem passes from the responsible party to neighbors or future users of the property. For many reasons, institutional controls generally require long-term monitoring and strict enforcement to be effective.

Institutional controls are defined as non-engineering measures that restrict the use or access to a site or facility to prevent or reduce exposure to hazardous substances (EPA, 1998). The most common scenario involves restrictions on use that significant reduce public access to contaminated land (e.g., fences, deed restrictions, and restrictive zoning). Another common institutional control is to limit access to ground water by shutting down drinking water wells and restricting new well drilling.

Although risk-based cleanups are often described with zoning terminology (e.g., residential, commercial, industrial), a zoning ordinance is not an effective institutional control, as it is rarely considered sufficiently permanent to limit exposures to hazardous material. Risk assessors, therefore, do not consider zoning; rather, they focus directly on exposure pathways and potential receptors. Institutional controls should block exposure pathways, including inhalation, ingestion, and dermal contact with contaminated media.

At both active bases and closed or former military bases, the lack of enforcement of institutional controls is a major source of uncertainty. Initial decisions about appropriate institutional controls can be difficult because of the conflicting goals of the military, nearby residents, and potential land owners. The military, in an effort to lower cleanup costs, tends to support less stringent cleanup goals that will restrict land use. In contrast, nearby residents and future land owners do not want to be restricted from construction on or access to land, and generally advocate the highest possible cleanup level.

Security fencing is one type of access control used at Navy facilities.

Courtesy of the U.S. Navy.

At closed and former military bases, regulators have used various legal instruments to impose long-term controls, including deed restrictions, deed notices, and negative easements. These mechanisms are generally unreliable because they are not supported by long-term oversight. Often institutional controls are adopted with no review by local planning bodies. Even when local communities are consulted (through Restoration Advisory Boards, Local Reuse Authorities, or local governments) there is no guarantee that restrictions will last.

Enforcement of legal institutional controls at active military bases is even more difficult, since there is no property deed, and local governments have no zoning authority. The military has a poor record for keeping track of toxic hazards because it is not subject to local building codes. In addition, the repeated turnover of management, including base commanders, undermines the enforcement of controls. In California and Florida, regulators and the military have developed a plan to rely on a variety of documents, including Federal Facilities Agreements and base master plans, to reinforce institutional controls, but it is too soon to measure their success.

The effectiveness of access controls—responses designed to keep receptors away from hazards—varies with ownership. At active bases, the Navy can effectively enforce such controls, since access to most facilities is restricted for na-

tional security purposes. (Whether enforcement actually occurs, however, is dependent on facility size, the land use in surrounding areas, and other factors.) On former military property now in the hands of other federal agencies (the Department of the Interior, for example), access controls are more difficult to maintain. The mission of land management agencies is to encourage, not limit, public use of public lands. Finally, at property transferred to private and other non-federal entities, access controls are also difficult to sustain. Most new uses (economic development, recreation, transportation, and education) require greater public access to the property. Local reuse authorities, developers, and owners have virtually no interest in fencing or patrolling potentially hazardous property.

Even when an institutional control has been agreed to by all relevant parties, there may be unforeseen future changes in land use that negate the effectiveness of the control. At active military bases, if the military must change its use for a section of the base or, more likely, if the installation is later designated for closure, remedies based on institutional controls could backfire.

Institutional controls are most successful in areas that have physical attributes that are consistent with the controls, or in combination with containment systems. Examples of such controls include:

• building restrictions on a capped landfill in an area already designated, because of its unique habitat, permanent open space;
• building restrictions on runways constructed to withstand a nuclear attack in an area that wants a large airport; or
• restrictions on well drilling into aquifers already classified as nonpotable because of high levels of saltwater intrusion.

Such physical foundations for institutional controls, however, do not eliminate the need for continued monitoring of their performance.

In the absence of favorable physical conditions, enforcement of institutional controls is more difficult. The most important requirement is that institutional controls be reliable in the long run, when the present-day stakeholders are gone. The following recommendations for improving enforceability are made.

Strengthen the legal basis of deed restrictions and easements by modifying real estate law and by writing them explicitly into cleanup Records of Decision or other enforceable agreements. Today, such restrictions are subject to challenge, particularly after multiple transfers of ownership.

Encourage local jurisdictions to sustain land use restrictions through zoning, subdivision maps, building permits, and other activities. Currently, institutional controls may be imposed as part of a cleanup without notifying the local planning body. Furthermore, since local planners have no jurisdiction over

federal property until transfer takes place, local reuse authorities need to negotiate institutional controls so their rules and designations match those assumed by the military.

For current and former federal properties, establish a national entity that regularly audits the effectiveness of institutional controls. This could be done by the EPA's Federal Facilities Enforcement Office. The EPA should develop criteria for determining whether institutional controls are functioning properly and under what conditions controls can be relaxed. (The five-year review under CERCLA is designed to make this determination, but there appears to be no standard process for translating such reviews into the arena of institutional controls.)

Encourage public pressure as an oversight tool by systematically providing more information to the public. (such as the Toxic Release Inventory and the 800-number hotlines used for utility trenching). A comprehensive national registry, or a network of standardized state registries, of contamination-based institutional controls would be particularly effective. Increasing public awareness about the use of institutional controls should help ensure their protectiveness.

Uncertainties that arise in the long-term effectiveness of institutional controls should make responsible parties and regulators cautious when proposing them. Interim controls remain necessary under any system, and at sites where cleanup is technically impracticable, they may be unavoidable. In other cases, decisions to substitute institutional controls for cleanup to unrestricted use must be carefully weighed. The following two case studies (Boxes 4-2 and 4-3) illustrate the success and failure of institutional controls.

Abernathy, A. 1998. EPA Office of Solid Waste and Emergency Response. Personal communication.

American Society for Testing and Materials (ASTM). 1995. Standard Guide for Risk-Based Corrective Action Applied at Petroleum Release Sites (E 1739–95). Annual Book of ASTM Standards. West Conshohocken, Pa.

ASTM. 1998. Standard Provisional Guide for Risk-Based Corrective Action (PS 104–98). Annual Book of ASTM Standards. West Conshohocken, Pa.

Applegate, J. S., and S. Dycus. 1998. Institutional Controls or Emperor's Clothes? Long-Term Stewardship of the Nuclear Weapons Conflict. Environmental Law Reporter 28(11) 10631–10652.

Bass, J. M., W. J. Lyman, and J. P. Tratnyek. 1985. Assessment of Synthetic Membrane Successes and Failures at Waste Storage and Disposal Sites. Project Summary, EPA/600-S2-85/100. U.S. Environmental Protection Agency, Cincinnati, Ohio.

Browner, C. M. 1995. Policy for Risk Characterization at the U.S. Environmental Protection Agency. Washington, D.C.

California State Water Resources Control Board. 1997. Resolution #01–21–97. Sacramento, Calif.

Chapelle, F. H., P. M. Bradley, D. R. Lovley, and D. A. Vroblesky. 1996. Measuring Rates of Biodegradation in a Contaminated Aquifer Using Field and Laboratory Methods. Groundwater 34(4):691–698.

Environmental Law Institute. 1995. Institutional Controls in Use. Washington, D.C.

Environmental Protection Agency. 1998. Institutional Controls: A Reference Manual (Working Group Draft).

Finkel, A. M. 1990. Confronting Uncertainty in Risk Management A Guide for Decision-Makers. Center for Risk Management, Resources for the Future. Washington, D.C.

Foster-Reid, G. H. 1994. Variability of hydraulic conductivity and sorption in a heterogeneous aquifer. S. M. Thesis, MIT, Cambridge, Mass.

Gelhar, L. W. 1993. Stochastic Subsurface Hydrology, Prentice Hall, Englewood Cliffs, N.J.

International Containment Technology Conference. 1997. U.S. Environmental Protection Agency, St. Petersburg, Fla., February 9–12.

Kamrin, M. A. 1990. Toxicology Primer. Lewis Publishers, Inc., Chelsea, Mich.

LeBlanc, D. R., S. P. Garabedian, K. M. Hess, L. W. Gelhar, R. D. Quadri , K. G. Stollenwerk, and W. W. Wood. 1991. Large-Scale Natural Gradient Tracer Test in Sand and Gravel, Cape Cod, Massachusetts. 1. Experimental Design and Observed Tracer Movement. Water Resour. Res. 27(5):895–910.

Mackay, D. M., W. P. Ball, and M. G. Durant. 1986. Variability of Aquifer Sorption Properties in a Field Experiment on Groundwater Transport of Organic Solutes: Methods and Preliminary Results. J. Contam. Hydrol. 1:119–132.

MacDonald, J. A., and M. C. Kavanaugh. 1994. Restoring Contaminated Groundwater: An Achievable Goal? Environ. Sci. Technol. 28(8):362A-368A.

Maxwell, R. M., S. D. Pelmuder, A. F. B. Tompson, and W. E. Kastenberg. 1998. On the Development of a New Methodology for Groundwater-driven Health Risk Assessment. Water Resour. Res. 34(4):833–847.

McNab, W. W., and B. P. Dooher. 1998. Uncertainty Analyses of Fuel Hydrocarbon Biodegradation Signatures in Ground Water by Probabilistic Modeling. Ground Water 36(4):691–698.

Mercer, J. W., and C. R. Faust. 1981. Ground-Water Modeling. National Water Well Association . Worthington, Ohio.

Morgan, M. G., and M. Henrion. 1990. Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge University Press. Cambridge, U.K.

Moskowitz, P. D., R. Pardi, V. M. Fthenakis, S. Holtzman, L. C. Sun, and B. Irla. 1996. An Evaluation of Three Representative Multimedia Models Used to Support Cleanup Decision-Making at Hazardous, Mixed, and Radioactive Waste Sites. Risk Anal. 16(2):279–287.

National Research Council (NRC). 1983. Risk Assessment in the Federal Government: Managing the Process. National Academy Press. Washington, D. C.

NRC. 1990. Ground Water Models, Scientific and Regulatory Applications. National Academy Press. Washington, D.C.

NRC. 1993. Issues in Risk Assessment. National Academy Press. Washington, D.C.

NRC. 1994a. Science and Judgment in Risk Assessment. National Academy Press. Washington, D.C.

NRC. 1994b. Alternatives for Ground Water Cleanup. Washington, D.C.: National Academy Press.

NRC. 1996. Understanding Risk. Informing Decisions in a Democratic Society. National Academy Press. Washington, D.C.

NRC. 1997. Innovations in Ground Water and Soil Cleanup: From Concept to Commercialization. Washington, D.C.: National Academy Press.

Price, P. S., R. E. Keenan, J. C. Swartout, C. A. Gillis, H. Carlson-Lynch, and M. L. Dourson. 1997. An Approach for Modeling Noncancer Dose Responses with an Emphasis on Uncertainty. Risk Anal. 17(4):427–437.

Rajaram, V., P. V. Dean, S. A. McLellan, A. Mills, P. L. Chandler, G. W. Snyder, and D. L. Namy. 1997. Performance of Engineered Barriers. In Proceedings of International Containment Technology Conference. St. Petersburg, Fla., February 9–12.

Rehfeldt, K. R., J. M. Boggs, and L. W. Gelhar. 1992. Field Study of Dispersion in a Heterogeneous Aquifer, 3. Geostatistical Analysis of Hydraulic Conductivity. Water Resour. Res. 28(12):3309–3324.

Robin, M. J. L., E. A. Sudicky, R. W. Gillham, and R. G. Kanachanoski. 1991. Spatial Variability of Strontium Distribution Coefficients and Their Hydraulic Conductivity in the Canadian Forces Base Borden Aquifer . Water Resour. Res. 27(1):2619–2632.

Rumer, R. R., and J. K. Mitchell. 1995. Assessment of Barrier Containment Technologies: A Comprehensive Treatment for Environmental Remediation Applications. Prepared under the auspices of the USDOE, USEPA, and DuPont Co. NTIS #PB96-180583.

WASTECH. 1994. Innovative Site Remediation Technology: Phase I (Process Descriptions and Limitation). American Academy of Environmental Engineers, Annapolis, Md.

WASTECH. 1998. Innovative Site Remediation Technology: Phase II (Design and Application). American Academy of Environmental Engineers, Annapolis, Md.