Click for next page ( 68


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 67
ASSESSING COMPLIANCE INTRODUCTION in the preceding chapter, we described our conclusion that the form of a Yucca Mountain standard should be based on limiting individual risk as measured by the average risk to individuals in a critical group. This group is defined as being composed of persons likely to be at highest risk from radionuclides released from the repository. Our judgment is that limiting individual risk in this way is also likely to provide adequate radiological protection for all relevant populations that might be exposed to radiation from radionuclides released from the proposed repository at Yucca Mountain (see Chapter 2~. The period over which this level of protection should be assessed should extend over the period of duration of hazard potential of the repository, that is, until the time at which the highest critical group risk is calculated to occur, within the limits imposed by the long-term stability of the geologic environment at Yucca Mountain, which is on the order of 1 o6 years. In this chapter, we discuss the analyses that must be undertaken to judge compliance with such a standard. Important questions to be answered are: Whether the scientific understanding of the relevant events and processes potentially leading to releases is sufficient to allow a quantitative estimate of future repository behaviors. Whether adequate analytical methods and numerical tools exist to incorporate this understanding into quantitative assessments of compliance. Whether the current scientific understanding and analytic methods are sufficient to evaluate performance with sufficient confidence to assess compliance over the long time periods required. 4. Whether the results of the analyses required to assess repository performance can be combined into an estimated 67

OCR for page 67
68 YUCCA MOUNTAIN STANDARDS risk for comparison with the standards in the licensing process. In particular, the estimated risk is defineti as the mean risk of members in the critical group. Risk is defined as the expected value of the probabilistic distribution of health effects experiences! by an individual member of the critical group. The main too! used to assess compliance is quantitative performance assessment, which relies upon mathematical modeling. We have evaluated the degree of confidence that can be placer! today in such assessments. We have also made a systematic analysis of the application of this methodology to the Yucca Mountain site. Based on these analyses, we conclude that: For those aspects of repository and waste behavior that depend on physical and geologic properties and processes, enough of the important aspects can be known within reasonable limits of uncertainty, and these properties and processes are sufficiently understood} and stable over the long time scales of interest to make calculations possible and meaningful. These properties and processes include the raclionuclide content of the waste (which changes over time clue to radioactive decay), the influx of water through the site and its effect on waste package integrity and other engineered barriers, the migration of wastes to ground} water after waste packages have lost their integrity, anti the subsequent dispersion and migration of wastes in ground water. While these factors cannot be calculated precisely, we believe that there is a substantial scientific basis for making such calculations, taking uncertainties and natural variabilities into account, to estimate, for example, the concentration of wastes in ground water at different locations and the times of gaseous releases. One critical gap in our understanding is with respect to future human behavior. Since there is no scientific basis for predicting human behavior, we recommend that policy decisions be made to specify default (or reference) scenarios .

OCR for page 67
ASSESSING COMPLIANCE 69 to be user} to incorporate assumed future human behavior into compliance assessment calculations. Available mathematical and numerical tools are neither perfect nor complete. Nevertheless, the currently available tools plus additional tools that we believe can be developed as part of the stanciarci-setting and compliance assessment efforts, or through other research, should be adequate for the analyses requires] to evaluate repository performance. So long as the geologic regime remains relatively stable, it should be possible to assess the maximum risks with reasonable assurance. The time scales of long term geologic processes at Yucca Mountain are on the order of 1 oh years. Other processes that operate on short time scales, such as seismic activity, can also be accommodates] in performance assessment if the maximum risks associateci with these processes depend] more on whether an event is likely to occur (at any time) than on the specific timing of the event. 4. Established procedures of risk analysis should enable the combination of the results of all repository system simulations into a single estimated risk to be compared with the standard. (Human intrusion is excluded from such a combination. See Chapter 4.) An element of judgment is contained in many of the conceptual assumptions to be macle, and those assumptions, methods, and the reference data will have to be specified. Similarly, reference exposure scenarios must be established clearly. This transparency in the use of assumptions is critical to evaluating the calculated risk. Because some readers might be unfamiliar with the technical aspects of a repository performance assessment, it is appropriate to provide an overview of the methodology, as we Rio in Part I of this chapter. We then consider the scientific basis for making an assessment of Yucca Mountain. We have found it useful to separate this evaluation into two parts, one dealing with the physical properties and geologic processes relevant to the behavior of the wastes and the other with those aspects of performance assessment that deal with assumptions about where ant] how people live, how they might be exposer] through the foot} en cl water they

OCR for page 67
70 YUCCA MOUNTAIN STANDARDS consume, and other factors that conic! affect exposures to radioactive wastes. We shall refer to this latter collection of factors that must be considered! as exposure scenarios. The reason for separating these two elements of performance assessment is that the nature of calculations in each is substantially different. We discuss these in Parts I] anti III. PART I: OVERVIEW OF PERFORMANCE ASSESSMENT Any standard to protect inclividuals ant] the public after the proposed repository is closecl wouIcl require assessments of performance at times so far in the future that a direct evaluation of compliance (for example by physical monitoring of system behavior) is out of the question. The only way to evaluate the risks of adverse health effects and to compare them with the standard} is to assess the estimated potential future behavior of the entire repository system anti its potential impact on humans. This procedure, involving modeling of processes and events that might Bali to releases and exposures, is called performance assessment. It involves computer calculations using quantitative models of physical, chemical, geologic, anti biological processes, taking uncertainties into account. Modeling repository performance is a challenging task because the rates of geochemical transformation and transport of the radionuclides are generally very slow and the times at which points distant from the repository become significantly affected by ra(iionuclide releases will be in the far future. Thus, to assess these effects requires projection of geochemical, hydrodynamic, and other processes over long time periods within rock masses whose properties are imperfectly known. Factors describing how humans can be exposed to radionuclides from the wastes are even more imperfectly known and these factors, inclucling the future state of technology and medicine, might be more changeable over time than are the physical processes. Reasonable Confidence One possible response to these difficulties is to conclude that they render any assessments of the ultimate fate of the waste materials too uncertain to be useful. However, we believe that such analyses do provide information for judging the quality of a disposal site. Even if the

OCR for page 67
ASSESSING COMPLIANCE 71 uncertainties involves! are large, some options for the disposition of the wastes can clearly be shown to result in worse consequences than other options would] produce. The results of compliance analysis should not, however, be interpreted as accurate predictions of the expecter} behavior of a geologic repository. No analysis of compliance will ever constitute an absolute proof; the objective instead! is a reasonable level of confidence in analyses that inclicates whether limits establisher! by the standard will be exceeded. Both the USNRC ant! EPA have explicitly recognized this objective. For example, EPA states in 40 CFR 191 that "unequivocal numeric proof of compliance is neither necessary nor likely to be obtained." In regulation 10 CFR 60, USNRC acknowledges that "it is not expected that complete assurance that "performance objectives] will be met can be presented." The USNRC requires instead} "reasonable assurance, making allowances for the time period, hazards, and uncertainties involved." EPA's required level of proof in 40 CFR 191 is "reasonable expectation." Time scale One commonly expressed concern regarding the performance assessment mo~ieling is that it requires simulating performance at such distant times in the future that no condolence can be placed in the results. Of course, the level of confidence for some predictions might decrease with time. This argument has been used to support the concept of a 10,000 year cutoff (DOE, 19921. We cio not, however, believe that there is a scientific basis for limiting the analysis in this way. One of the major reasons for selecting geologic disposal was to place the wastes in as stable an environment as many scientists consider possible. The deep subsurface fulfills this condition very well (NRC, 19571. In comparison with many other fields of science, earth scientists are accustomed to dealing with physical phenomena over long time scales. In this perspective even the longest times considered for repository performance models are not excessive. Furthermore, even changes in climate at the surface would probably have little effect on repository performance deep below the ground. We recommend calculation of the maximum risks of radiation releases whenever they occur as long as the geologic characteristics of the repository environment do not change

OCR for page 67
72 YUCCA MOUNTAIN STANDARDS significantly. The time scale for long-term geologic processes at Yucca Mountain is on the order of approximately one million years. After the geologic environment has changed, of course, the scientific basis for performance assessment is substantially eroded and little useful information can be cievelopect. Because there is a continuing increase in uncertainty about most of the parameters describing the repository system farther in the distant future, it might be expected that compliance of the repository in the near term could be assesses] with more confidence. This is not necessarily true. Many of the uncertainties in parameters describing the geologic system are clue not to temporal extrapolation but rather to difficulties in spatial interpolation of site characteristics. These spatial cliff~culties will be present at all times. Accordingly, even in the initial phase of the repository lifetime, a compliance decision must be baser! on a reasonable level of confidence in the predicteci behavior rather than any absolute proof. Under some circumstances, use of a shorter period for analysis court! in fact introduce additional uncertainties into the calculation. For example, uncertainties in waste canister lifetimes might have a more significant effect on assessing performance in the initial 10,000 years than in performance in the range of 100,000 years. Probabilistic Analysis of Risk To judge compliance against a risk-base(i standard of the type proposed, a risk analysis including treatment of all scenarios that might leac! to releases from the repository and to radiation exposures is, in principle, required. To include them in a stanciarci risk analysis, all these scenarios need to be quantified with respect to the probabilities of scenario occurrence ant} the probability distribution of their consequences to humans, such as health effects of radiation doses. In subsequent sections we specifically note that for some events or processes either the probability of occurrence or the estimates! consequences become very difficult to specify with confidence. Events caused by human activity are usually of this type. Incorporation of such events or processes into the formalized risk analysis sometimes is not justified on a scientific basis. Instead, how to deal with these events should be decided as a matter of policy.

OCR for page 67
ASSESSING COMPLIANCE 73 This approach implies a departure in part from common analytical techniques to assess risks and the introduction of more pragmatic procedures needed to provide an adequate decision basis. It is important, therefore, that the "rules" for the compliance assessment be established in advance of the licensing process; that is, that the scenarios that might be excluder! from the integrated risk analysis be identified. Human intrusion is an example of one scenario that we judge to be not amenable to incorporation in the risk assessment framework; this is discusser] further in Chapter 4. We believe that performance assessment using numerical models of physical anti chemical processes and quantitative estimates of probabilities is the key approach to assessing compliance. However, the confidence that can be placed in such analyses is also a key part of the compliance issues. To some extent, this degree of confidence can be quantified, for example, by performing rigorous uncertainty analyses that propagate uncertainties in parameter values through the analysis to produce estimates of uncertainties in estimated risks. Uncertainties due to modeling approaches can also be assesses! by comparing the results of assessments using various alternative models, or by comparing mocie! results with data collected! in experiments or in observations. In other cases, less rigorous but useful evidence of the adequacy of models or data can be obtained by, for example, comparisons with relevant natural analog systems. A final, important point to note is that performance assessments of the type summarized above are not likely to be performed only on a single occasion preparatory to licensing. Assessments will likely be performed iteratively during system clesign, construction and operation of a geologic repository, and finally at the time the repository is sealed, following decades of experience in which additional data on the performance of system components can be gatherer). QUANTITATIVE CALCULATION OF REPOSITORY PERFORMANCE In this section, we summarize general aspects of performance assessment modeling ant! sources of uncertainty in the modeling process before moving in subsequent sections to issues more specific to Yucca Mountain. The main thrust of performance assessment involves

OCR for page 67
74 YUCCA MOUNTAIN STANDARDS developing a quantitative understanding of system behavior, assembling a sufficient database of parameters describing the system, and producing simulations of possible future system behavior allowing as fully as possible for uncertainties in understanding or in databases. Figure 3.1 schematically illustrates the generic modeling process described in more detail below. Figure 3.l The Basic Steps in Performance Assessment Natural Observations Laboratory and Field Experiments develop system understanding (conceptual model) develop quantitative models (mathematical model) numerical implementation (numerical analysis) test models in relevant conditions accumulate input data on repository (in form of probability distributions) 1 run simulations derive estimates of consequences and probabilities derive risk estimate Elements of Performance Assessment Conceplual midge! The conceptual mode} reflects the scientists' understanding of how the important aspects of the system work. It answers questions such as:

OCR for page 67
ASSESSING COMPLIANCE 75 What are the limits of the system? What are the geometry and composition of the system? What are the significant physical processes? It is the conceptual mode] that dictates the selection of the mathematical formalisms that enable quantitative calculations to be performed. One special type of conceptual mocle! frequently employed in performance assessment is the scenario. In this context, a scenario means a description of how radionuclides might migrate from the repository and affect humans. For example, "the wastes are dissolved in ground] water, which is transported by natural processes to an agricultural area, where it is pumped out of the ground en cl used to irrigate crops and ingested] by humans" is a possible scenario for the Yucca Mountain repository. Quantitative performance assessment baser] on this scenario wouIc! then have to employ cletailec! conceptual models of release ant] transport processes specifying, among other things, how and where the ground water flows and exposure scenario mociels specifying where farmers live, what technologies they use ant! their patterns of consumption of food and water. The scenario thus constitutes a kind of master conceptual mocle! that guicles the selection of more detailed and specific conceptual moclels for each step of the process. The conceptual models are potentially the source of the most significant uncertainties regarding the outcome of the analysis. If the nature of the system has not been properly assessed, or the most important processes have not been included in the conceptual model, the mathematical mocle] based on the conceptual mode} will not properly simulate the behavior of the system regardless of how adequately the other elements of the analysis might be quantif~ecI. Inadequacies in conceptual models are a particularly worrisome aspect of the performance assessment process because a major error couIc! invalidate the entire exercise, yet be difficult or impossible to detect. Although, it is important to realize that this limitation is an aspect of all human problem-solving activities, it is particularly important for radioactive waste repository performance assessment computations because of their long-term considerations. The best way to guard against errors of this nature is to provide for multiple, rigorous, independent reviews of conceptual moclels that are clearly documentecI and widely disseminated.

OCR for page 67
76 Math etnatical mode' YUCCA MOUNTAIN STANDARDS By mathematical mocie} we mean the mathematical relationships that are used to describe the physical system quantitatively. The system of equations that is incorporated in the mathematical mocle! usually represents a simplification of the selected conceptual model. Mathematical simplification might be required because it is not possible to filch adequate descriptions of all the phenomena consiclerec] important, or because incorporation of all relevant equations would result in a mathematical system too cumbersome to solve, or because the data available do not justify the most complete description of the system that might be possible. Mathematical simplifications reduce the realism of the outcome of the moclel, but the degree to which the results are affecter! can be assessed by means of mathematical techniques, such as sensitivity analyses of numerical results. Numerical analysis Most mathematical models consist of sets of coupled differential equations. For the cases of interest to performance assessment, it is often difficult to solve such complex systems of equations analytically, or exactly, in which case approximate numerical methods are employed. Selection of appropriate numerical methods is important because more efficient numerical techniques can permit more complex (and thus, presumably, more realistic) physical models to be solved, ant! because inappropriate numerical schemes can introduce significant errors into results. However, numerical inaccuracies are rarely a major source of error in properly conducted modeling because well-established methods exist for assessing the accuracy of numerical schemes. Further, if one approach is found to introduce unacceptable error, it can either be replaced or modified to achieve the desired accuracy. Model parameters Physical ant] chemical models require the specification of the physical properties of the system to be modelecl. These properties are

OCR for page 67
ASSESSING COMPLIANCE 77 referred to as parameters. The parameters are represented by numerical functions or values in the mathematical models. Models of the type commonly used in performance assessment describe the behavior of the system as a function of both space and time. Spatially heterogeneous models of systems incorporate the spatial variations of the parameters throughout the physical domain that is being modeled. The neeci to provide numerical values for parameters is another source of uncertainty in mathematical modeling. It is a goal of geologic disposal of nuclear wastes to emplace them in an environment that is deep, remote, and clifficult to access. These same repository properties make it clifficult to obtain data on the spatial variations of physical parameters in the system. Furthermore, the very procedures necessary to collect the data, such as drilling exploratory holes to extract samples of rock might compromise the integrity of the geologic barriers. Boundary conditions Performance assessment models have both spatial and temporal boundaries, that is, times of the beginning and ending of simulations. In general, both mass and energy can flow across these boundaries. Thus, to perform model calculations it is necessary to specify the conditions at the spatial ant] temporal boundaries (the mode] calculates parameter values within the mocle] domain). Specification of the "boundary conditions" is subject to many of the same types of uncertainty that are involved in specifying parameter values, and they are usually ciealt with in a similar fashion. In general, spatial boundary conditions of regional scale subsurface flow models are considered to be constant over time. There is at least one important exception to this generalization. The upper boundary to the geologic environment around the repository is the atmosphere. The average of atmospheric conditions is the climate, and it is well known that climate can vary significantly over geologic periods of time. Although the typical nature of past climate changes is well known, it is obviously impossible to predict in detail either the nature or the timing of future climate change. This fact adds to the uncertainty of the model predictions. During the past 150,000 years, the climate has fluctuates} between glacial and interglacial status. Although the range of climatic conditions

OCR for page 67
94 YUCCA MOUNTAIN STANDARDS Studies have been made of the possibility that a seismic event could produce transient changes in the water table at Yucca Mountain sufficient to bring ground water through the repository to the surface (NRC, 19921. Results indicate a probable maximum transient rise on the order of 20 m or less. In summary, although the timing of seismic events is unpredictable, the consequences of these events are bounciable for the purpose of assessing repository performance. Volcanism A volcanic intrusion into the proposer! repository could be catastrophic, releasing a major part of the repository inventory directly into the biosphere. However, the overall risk might be very low, because it is also a very unlikely event. Like seismicity, volcanism is episodic. The two phenomena cocci also be linked, in that some seismic activity can be triggered during periods of volcanic activity. Unlike seismicity, volcanism in the Yucca Mountain region involves intermittent concentrated activity separated by long repose periods. Even so, like seismicity, estimates of future volcanic activity can be based on analysis of the geologic record, with the assumption that the same pattern of events will hold in the future. The risk from volcanism at Yucca Mountain is being examined using a probabilistic approach. According to Crowe et al. ~1994), current studies are designed to establish three components of an overall probability of magmatic disruption of a repository: 1. 2. 3. Future recurrence rate of volcanic events, such as volcanic centers or volcanic clusters; The probability that a future event will intersect a specified area, such as the repository or a controlled area beyond! the repository; The probability that an event occurring within the specified area will release radionuclides into the biosphere. The probability of occurrence of the second component depends upon the probability of the first component, and the overall probability of raclionuclide release due to volcanism in the Yucca Mountain region tiepentis on the combined probability of all three components. Emphasis

OCR for page 67
ASSESSING COMPLIANCE . , 95 is being given to estimating the combination of the first ant! second! components to determine the combiner] probability that a future event will intersect a specifier! area. This analysis is baser! on extrapolations into the future of volcanic activity from the historic record, and on assumptions about the spatial distribution of future volcanic eruptions in the Yucca Mountain region. Crowe suggests that a probability of ~ 0~~/yr, which is a 1 in 10,000 possibility of a disruption over 10,000 years or ~ in 1,000 possibility in 100,000 years ~ or less might be sufficiently low to constitute a negligible risk. If the combined probability of the first two components can be shown to be below this level, then it might not be necessary to consider the third component. Efforts are underway to refine the intrusion distribution models by incorporating geologic structure constraints. It is noteci, for example, that the volcanic eruptions in Crater Flat appear to be aligned in the northeast direction of the extensional faulting (across the Yucca Mountain site). If this constraint is confirmed and included in the distribution, the probability of a future event intersecting the repository site might fall below ~ 0~8 per year. While acknowledging the complexity of estimating the release of radionuclicles to the biosphere, it seems possible, given the knowledge of material ejected} from various types of volcanic eruptions ant! study of the circler cones in the region, to clevelop reasonable estimates of the health consequences from radionuclides releaser] by a volcanic eruption through a repository at Yucca Mountain. Thus, it is believer] that the radiological health risk from volcanism can ant] should! be subject to the overall health risk standard] to be requires] for a repository at Yucca Mountain. PART III: EXPOSURE SCENARIOS IN PERFORMANCE ASSESSMENT As noted above, we believe that it is feasible to calculate, to within reasonable limits of certainty, potential, defined as possible but not necessarily probable concentrations of radionuclides in ground water and air at different locations and times in the future. To proceed from the calculation of ra~iionucli~ie concentrations to calculations of risks that would result from a repository, many additional factors or assumptions about the nature of the human society at or near the repository site must be

OCR for page 67
96 YUCCA MOUNTAIN STANDERS consiclered. These factors must be included in an exposure scenario that specifies the pathways by which persons are exposer] to radionuclicles releaser} from the repository. As we note in Chapter 4 with regard to the feasibility of making projections of future human intrusion into a repository, based on our review of the literature we believe that no scientific basis exists to make projections of the nature of future human societies to within reasonable limits of certainty. Therefore, unlike our conclusion about the earth science ant! geologic engineering factors described in Part lI of this chapter, we believe that it is not possible to predict on the basis of scientific analyses the societal factors that must be specified in a far-future exposure scenario. There are an unlimited! number of possible human futures, some of which would involve risks from a repository and others that would not. Although the nature of future societies cannot be preclicted, it is possible, at least conceptually, to consider several characteristics of future society that would indicate whether a repository is likely to pose a risk to people. A repository wouicl be unlikely to pose significant risks to future societies: if the area near the repository were not occupied, if future societies do not use ground water from the contaminated region, or if future societies routinely monitor ground-water quality ant] either treat or avoid use of contaminated sources. Conversely, exposures would result if water wells were drilled into the contaminateci areas and the water consumed by people or used to irrigate crops. As far as we are able to determine, there is no sound basis for quantifying the likelihood of future scenarios in which exposures do or do not occur; about all that can be said is that both are possible. It is our view, however, that once exposure scenarios have been adopted, performance assessment calculations can be carried out for the specified scenarios with a degree of uncertainty comparable to the uncertainty associated with geologic processes and engineered] systems. The more difficult task is the specification of reasonable scenarios for evaluation. Any particular scenario about the future of human society near Yucca Mountain that might be adopted for purposes of calculation is likely to be arbitrary, and should not be interpreter} as reflecting conditions that eventually will occur. Although we recognize the burden on regulators to avoid regulations that are arbitrary, we know of no scientific method for identifying these scenarios.

OCR for page 67
ASSESSING COMPLIANCE Selection of Exposure Scenarios for Performance Assessment Calculations 97 Any approach to assessing compliance with the standard! must make assumptions about the nature of the human activities and lifestyles that provide pathways for exposure. For example, people could drink water containing radionuclicies, irrigate crops with the water, eat these crops, and bathe in the water. Quantification of the doses receiver} from the various pathways requires detailed data on these pathways. For the example above, the average amount of water-ingested per day (not including other beverages constituted with uncontaminated water) should be known, as should the type of crops grown, the amount eaten, ant! the frequency of bathing. The set of circumstances that affects the dose received, such as where people live, what they eat and drink, and other lifestyle characteristics including the state of agricultural technology, are part of what we refer to as the exposure scenario. Unfortunately, many human behavior factors important to assessing repository performance vary over periods that are short in comparison with those that should be consiclered for a repository. The past several centuries (or even decades) have seen radical changes in human technology and behavior, many or most of which were not reasonably pre(lictable. For example, within the past one hundred years, our society has evolved from one in which drilling and pumping technology did not exist for production of water from the depths of grounc! water at Yucca Mountain to a level of technology where such production is feasible. Within this same time period, we have seen U.S. (demographic patterns shift from a time where a majority of U.S. residents were engages! in farming and grew their own food to the present day in which only a few percent of the work force is employee] in farming, ant! in which most people's diet includes foot] producer! outside their local area. Given this potential for rapic! change, it is unknowable what patterns of human activity might exist 10,000 or 100,000 years from now. Incleed, the perioc! during which repository performance might be relevant, on the order of a million years, is sufficiently long that any number of different societies might reside near the repository site. Several glacial periods probably will have occurred, making estimates of human society even more difficult. Given the unknowable nature of the state of future human societies, it is tempting to seek to avoid the use of such assumptions

OCR for page 67
98 YUCCA MOUNTAIN STANDARDS in performance assessment calculations. In our view, however, it is not possible for a reasonable stanciar~i for the protection of human health to avoid use of some specified assumptions about future populations, patterns, and lifestyles around a proposed repository site. Even regulatory standards stated in terms of geologic and engineering factors are not inciependent of assumptions about future exposure scenarios. For example, the containment requirements of 40 CFR 191 were apparently developed based on consideration of a global release scenario in which average doses to large populations were considered. The problem is how to pick an exposure scenario to be user! for compliance assessment purposes. Given the lack of a scientific basis for doing so, we believe that it is appropriate for the regulator to make this policy decision. One specific recommendation we make is to avoid placing the burden of postulating and defencling assumptions about exposure scenarios on the applicant for a license. The regulator appears to be better situates! than the applicant to carry the responsibility because of the perception that any future scenario cievelopeci by the applicant could have been chosen to give the desired outcome. On the other hand, the results of calculations from a scenario specified by the regulator in an open process designed to consider the views of all the interested! parties might be seen as a fair test of the suitability of a site and design. In addition, we recommend against an approach under which a large number of future scenarios are specified for compliance assessment, since such an approach could be seen as putting both the regulator and the applicant in the indefensible position of claiming to have considered a sufficient number of scenarios and that all reasonable future situations are represented] in the analysis. The purpose of making exposure scenario assumptions is not to iclentiiTy possible fixtures, but to provide a framework for the analysis and evaluation of repository performance for the protection of public health.2 2 Another argument for using a large number of scenarios is that iterative analysis of repository performance will lead to the most cost-effective repository design. This might be true, but we believe that the regulator must in the end assess compliance with a single level of protection as defined in the standard. Therefore, one (or at most a few) exposure scenarios must be specified for compliance assessment purposes.

OCR for page 67
ASSESSING COMPLIANCE 99 Specification of the exposure scenario assumptions to be user} in performance assessment at Yucca Mountain will greatly influence whether the site and ciesign can comply or not. The selection of exposure scenarios is perhaps the most challenging and contentious aspect of risk and compliance assessment. For example, EPA guidlines for exposure assessment reflect a philosophical disagreement over the question of when and how to depart from the theoretical upper bounc] estimate of exposure and to employ probabilistic techniques (Federal Register 57 fMay 29, 19921: 22888-229381. These questions, which are at the interface between science anti policy judgment, are also acIdresseci in Science and Judgment in Risk Assessment (NRC, 19941. For these reasons, we strongly recommend that the decision be made through a public rulemaking process. This process will provide a more complete analysis of the advantages and iisadvantages of alternative scenarios than we have been able to perform, and do so with the benefit of full public participation.3 As with other aspects of defining the standards and demonstrating compliance that involve scientific knowlecige but must ultimately rest on policy judgments, we considered what to suggest to EPA as a useful starting point for rulemaking on exposure scenarios. Reflecting the disagreement inherent in the literature, we have not reached complete consensus on this question. We do agree, however, that the exposure scenario user} to test compliance should not be based on an inclividual dei ined by unreasonable assumptions regarding habits and sensitivities affecting risk. It is essential that the exposure scenario that is ultimately selected be consistent with the critical-group concept that we advanced in Chapter 2. The purpose of using a critical group is to avoid using the standard to protect a person with unusual habits or sensitivities. The critical-group approach does this by using the average risk in the group for testing compliance. To ensure that this average risk nevertheless affords a high level of protection to most persons, the group must contain the persons at highest risk within the group and must be homogeneous in risk. An exposure scenario selected for 3 This rulemaking need not be done before the promulgation of an individual-risk standard that we recommended in Chapter 2. Indeed, we would not want the selection of that standard to be colored by foreknowledge of the assumptions incorporated in the exposure scenario.

OCR for page 67
100 YUCCA MOUNTAIN STANDARDS compliance assessment should procluce a critical group with these characteristics. Aciditionally, we note that the {CRP (1985a) recommends that the critical group be defined using present knowledge4 and cautious, but reasonable, assumptions. Although this guidance was originally intended for the regulation of dose limits, we believe that it is generally appropriate in applying the critical-group concept to risk, as we have recommencied. EPA should rely on this guiciance when choosing the assumptions for the exposure scenario to be used for performance assessment. Finally, we have considered the design of an exposure scenario that EPA might propose when it initiates the rulemaking process. We have considered} two illustrative approaches for this purpose. We describe the two approaches in Appendixes C and D, and summarize their important characteristics below. A substantial majority of the committee considers that the approach outliner! in Appendix C is more clearly consistent with the foregoing criteria for selecting an exposure scenario than is the alternative in Appendix D, and therefore believes that: EPA should propose an approach along the lines of Appenclix C. Of course, other methods might also meet these criteria, ant! some of the methods might be less complex than the method illustrated in Appendix C. Although the following discussion highlights differences between the two approaches, we wish to stress that the approaches are similar in many ways. The approach in Appendix C makes use of information that can be collected on the factors that influence human behavior in the present. Assumptions about factors such as the source of foot! would! be based on the source of food for today's population near the repository site. The Appendix C approach bases the exposure scenario on a population distribution rierived from observer! statistical associations between environmental parameters and the population distribution of actual population groups. For example, such parameters couic} inclucle depth to 4 We understand "present knowledge" to mean any knowledge that is available today, and so should be read as an injunction against making assumptions about knowledge that might exist in the future. For example, assuming that future societies will have found a cure or prevention for cancer would not be present- day knowledge.

OCR for page 67
ASSESSING COMPLIANCE 101 water, soil type and depth, land slope, and growing season. This approach uses statistical techniques to compute a critical group for each of a large number of simulations of the contaminated ground-water plume and then averages over these calculations to identify the average critical group for compliance purposes. Important characteristics of this approach include the following. First, it extends the probabilistic methods that have been applied to simulations of physical processes (such as transport of ground-water contaminants) to analysis of the factors affecting exposure. Second, although mathematically complex, the model is based on currently observable data and does not require assumptions regarding specific values of parameters, only ranges within which the parameters might fall. Third, the degree to which conservatism is incorporated is determined not only by the analyst in selecting the ranges of parameters that describe farming lifestyles but also by the regulator when the standard is set. Fourth, it requires that the probability that persons occupy specific parcels of land for farming be determined statistically by the relevant characteristics of the land, ground water, and technology that influence farming, avoiding the potential that the standard could be influenced by a situation in which the maximum dose occurred at a place that was uninhabitable or otherwise unsuitable for farming. The approach in Appendix D specifies a priori one or more subsistence farmers as the critical group and makes assumptions designed to define the farmer at maximum risk to be included in the critical group. The subsistence farmer would be a person with eating habits and with response to doses of radiation that are normal for present-day humans. All food eaten over the lifetime of the subsistence farmer would be grown with water drawn from an underground aquifer contaminated with radioactivity from the repository. The water would be withdrawn at a location outside the footprint of the repository and near that maximum potential concentration of the most critical radioactive contaminant in the ground water so that the scenario describes the maximum dose and risk. All of the farmer's drinking water would come from that same source. For compliance assessment purposes, it is assumed that the homogeneity criterion (see the definition of critical group in Chapter 2) applies and that the risk to the average member of the critical group is about one-third that of the subsistence farmer.

OCR for page 67
102 YUCCA MOUNTAIN STANDARDS The important features ofthe subsistence-farmer mode! include the following. First, it has been used extensively in radioactive waste management programs in the United States ant! other countries, so a body of experience with it exists on which to draw. Second, it is straightforward en c! relatively simple to understand and calculate. Third, while it incorporates a series of assumptions about the lifestyle of the hypothetical farmer, any degree of conservatism can be built into the mode! by choices among alternative assumptions, which can be based on current conditions in the Amorgosa Valley; these assumptions need not be constrained by the characteristics of the current population of the region. Fourth, it makes the most conservative assumption that wherever en c} whenever the maximum concentration of raclionuclides occurs in a ground water plume accessible from the surface, a farmer will be there to access it. These approaches have many elements in common. Most important, both rely on probabilistic methods of estimating the distribution of raciionuclides in the environment. Both also incorporate knowledge of the natural geologic features of the environment that influence the potential for exposure and both are intended to incorporate cautious, but reasonable' assumptions about lifestyles of the affected populations that the EPA might propose in a rulemaking. For example, both assume eating habits and response to radiation doses that are normal for present-day humans. Despite these similarities between the approaches, two major issues that differentiate them have emerged from our consideration. These issues are summarized below: Assumptions about the location and lifestyle of persons who might be exposed to radionuclides released from the repository are crucially important because they affect the identification of the person at highest risk that must be contained in the critical group. The two approaches differ in their treatment of these assumptions. For example, the approach in Appendix D specifies a priori that a person will be present at the time and place of highest nuclide concentrations in grounc} water and will have such habits as to be exposed to the highest concentration of radiation in the environment. This person is assumed to define the upper limit of risk in the critical group. Appendix C treats the distribution of potential farmers probabilistically based on

OCR for page 67
ASSESSING COMPLIANCE 103 current technical understanding of farming in the region. Because the person at highest risk might not be the same uncler the two approaches, the critical group selected for compliance assessment could be different. The second difference involves the method of calculating the average risk of the members of the critical group. Appendix C uses detailed statistical analysis to clef~ne the critical group. Specifically, it identifies a "critical subgroup" for each of a large number of Monte CarIo realizations of the contamination plume. The critical group risk is determiner} by averaging over the average risks to each of these subgroups. In contrast, the Appendix D approach approximates the average critical group risk at about one- thirc! of the risk faced by the person at highest risk, since the requirement that the critical group be homogeneous in risk implies that the overall range of risks in the critical group be limited to about a factor of ten. If the distribution of risk among members of the critical group is not relatively uniform, these approaches could produce different averages. As notes! earlier, we agree that unrealistic assumptions are inappropriate. Our divergence of view is on the extent to which the alternative sets of assumptions embodies} in Appendixes C ant} D are cautious, but reasonable. The approach of Appendix C has the advantages of explicitly accounting for how the physical characteristics of the site might influence population distribution ant! of identifying the makeup of the critical group probabilistically. Most of the committee regard these as desirable features of exposure scenarios that are intended to be consistent with the critical-group concept. We emphasize, however, that specification of exposure-scenario assumptions is a matter for policy decision. Exclusion Zone The original standard, 40 CFR 191, contained a provision for an exclusion zone in the immediate vicinity of the repository. The purpose was to provide a boundary for calculating releases.

OCR for page 67
104 YUCCA MOUNTAIN STANDARDS In light of our conclusion in Chapter 4 that there is no scientific basis for assuming that institutional controls can be maintained for more than a few centuries, we also conclude that there is no scientific basis for assuming that human activity can be prevented from occurring in an exclusion zone or that defining such a zone will provide protection to future generations from exposures in the vicinity of the repository. The question remains whether an exclusion zone serves a useful purpose for compliance assessment. In our analysis, we have assumed that some human activities, such as drilling into or through the repository, should be treated as special cases of human intrusion (see Chapter 4~. If, as we recommencI, human intrusion is treater] separately from the performance of an undisturbeci repository, it is reasonable in our view to clefine a region in which human activities are to be regarded as intrusion and to exclude that region from calculation of the undisturbed repository performance. For example, if we assume that all drilling for water wells is vertical, the area directly above the repository plan (or footprint) would be consiclered an exclusion zone for purpose of calculating compliance with that part of the standard that applies to undisturbed performance. Drilling in that zone would be a case of human intrusion. Beyond the repository footprint, however, there seems to be no practical purpose for defining a larger exclusion zone for the form of the standard we recommenct. Without either a release limit or a time limit for the standard for undisturbed performance, an arbitrary boundary serves no purpose. In the approach we recommend, an objective of performance assessment calculations is to determine the time in the future when risks from exposure to radionuclides released from the repository are greatest ant] to base the regulatory judgment about compliance on a comparison of the risks at that time to the standard. Furthermore, neither of the alternatives for treating the critical group requires an exclusion zone larger than the repository footprint.