Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
FINDINGS The Limitations of Analysis Overview Engineers are unable to anticipate all of the potential problems that might arise in trying to site, build, and operate a repository. Nor can science prove that a repository will be absolutely "safe." This is so for two reasons. First, proof in the conventional sense cannot be available until we have experience with the behavior of an engineered repository system precisely what we are trying to predict ahead of time. And second, safety is in part a social judgment, not just a technical one. While technical analyses can greatly illuminate the judgment of whether a repository is safe, technical analysis alone cannot substitute for decisions about the degree of risk that is acceptable. These decisions belong to the citizenry of a democratic society. Both of these important limitations of technical analysis have been understated, a lapse that feeds the concern and magnifies the public's distrust of nuclear waste managment when these limitations are pointed out by the program's . . cntlcs. Uncertainty and Significant Risks A principal source of concern over Be U.S. program is the uncertainty in estimating the risks from a radioactive waste repository. Technical approaches are available to reduce or at least bound these uncertainties. Yet in focusing on ways to improve the analysis, public discussion has often overlooked a more important question: whether the uncertainty matters. This is, in principle, the domain of performance assessment, which draws together the different portions of the technical analysis so that one can see which parts of the waste confinement system may pose environmental hazards during or after the time when the repository receives waste. Performance assessment of a repository system is necessarily a task for computer modeling. The waste management system, which starts at the reactor and continues into the distant future of a sealed repository, includes many different parts and processes that are described through different kinds of data (with different levels of quality), and different kinds of analysis (with different levels of accuracy). It is a practical consequence of the complexity of high-level radioactive waste (HLW) disposal, together with the fact that no one has ever operated a repository, that performance assessment is, in the end, a matter of technical judgment. 13
14 The traditional approach in such cases, where an important social deci- sion hinges on uncertain scientific data and projections, is to inform the political decision through a consensus of the appropriate technical commu- nity. Such consensus is difficult to reach in this case, however, given the political controversy, conflicting value systems, and overlapping technical specialties involved in assessing repository performance. Indeed, the allowable residual risk associated with a permissible repository site is a political choice; EPA has taken the position that the implementation of their guidelines constitutes the exercise of this choice. Unfortunately, the number and magnitude of the uncertainties in the probabilistic approach may be expected to reintroduce political controversy. This was recognized by the High-Level Radioactive Waste Disposal Subcommittee of EPA's Science Advisory Board in their January 1984 report reviewing EPA Draft Standard 40 CFR 191. That subcommittee concluded there was insufficient basis for agreeing with the EPA staff that the proposed release criterion with its probabilistic corollary can be demonstrated to have been met with reasonable assurance, and that this could be argued definitively in a legal setting. The subcommittee strongly affirmed the validity of EPA's probabilistic approach, but warned that if EPA cannot have high confidence in the adequacy and workability of a quantitative, probabilistic standard, [it should] use qualitative criteria, such as recommended by [the US]NRC. Specifically, with regard to the first major topic of the Science Advisory Board's findings and recommendations, "Uncertainty and the Standard," the subcommittee recommended relaxing the nuclide release limits by a factor of 10, modifying the probabilistic release criteria so that analysis of repository performance shall demonstrate that there is less than a So-so chance of exceeding the Table 2 release limits, modified as is appropriate. Events whose median frequency is less than one in one-thousand in 10,000 years need not be considered, and, finally that use of a quantitative probabilistic condition on the modified Table 2 release limits be made dependent on EPA's ability to provide convincing evidence that such a condition is practical to meet and will not lead to serious impediments, legal or otherwise, to the licensing of high-level-waste geologic repositories. If such evidence cannot be provided, we recommend that EPA adopt qualitative criteria, such as those suggested by the [US]NRC.l Unfortunately none of these recommendations was adopted. The USNRC staff, in commenting on the EPA Draft Standard, strongly
15 questioned the workability of quantitative probabilistic requirements for the defined releases stating, in part numerical estimates of Me probabilities or frequencies of some future events may not be meaningful. The [US]NRC considers Mat identification arid evaluation of such events arid processes will require considerable judgment and therefore will not be amenable to quantification by stansiica~ alyses without the in- clusion of very broad Garages of uncertainty. These uncertainty ranges will make it difficult, if not impossible, to combine the probabilities of such events wide enough precision to make a meaningful contribution to a licensing proceeding.2 The problem is compounded when the adequacy of the performance as- sessment to determine if the allowable residual risk is achieved is judged by its political impact (i.e., the effect of reopening the discussion of what is allowable residual risk) as well as its technical accuracy. The difficulty of evaluating performance assessments is compounded by the fact that there is no actual experience in the disposal of HEW on which to base estimates of the risk. Some risk scenarios include low-probability/ high-consequence events. Others are based on explicit or implicit assump- tions that cannot be plausibly proved or disproved for example, the consequences of climatic changes that could increase rainfall and groundwater flows at a repository site. The data and methodologies for modeling of repository isolation performance are still under development. The actual performance of a repository is difficult to predict for many reasons. Geologists often disagree about the interpretation of data in analyzing the history of a site or geological structure. Long-term predictions are even more uncertain. Releases may occur thousands of years in the future, and they are likely to be diffuse and hard to detect. The potential for (and effects of) human exposure will be further shaped by unpredictable changes in demographics and technology. These uncertainties do not necessarily mean that the risks are significant, nor that the public should reject efforts to site the repository. Rather, they simply mean that there are certain irreducible uncertainties about future risk. An essential part of any successful management plan is how to operate with large residual uncertainties, and how to maintain full public accountability as information about the risks changes with experience. This is not an impossible task: public policy is made every day under these conditions, and private firms undertake all sorts of activities in the face of uncertainty. What is clear, however, is that a management plan that promises that every problem has been anticipated, or assumes that science will provide all the answers, is almost certainly doomed to fail. There have been many cases where attempts to understate uncertainty have damaged an agency's credibility and subverted its mission. For this reason, experienced regulatory agencies like EPA now pay careful attention to describing the uncertainties associated with their risk assessments.
16 Perceptions of Risk Studies have linked the high public perceptions of the risk from nuclear power plants to certain qualities of that risk, in particular to perceptions that the risks are catastrophic, new, uncertain, and involuntary (i.e., beyond individual control). Radioactive waste poses risks with many of the same technical characteristics: the principal health risks (chiefly cancer and genetic defects) originate in the hazards of ionizing radiation. The risks from radioactive waste also have some of the same social characteristics as risks from nuclear reactors: a long time may pass before the hazards become apparent, dangers may be imposed involuntarily on populations, and there is a perceived pos- sibility of catastrophe. The last perception, in particular, is qualitatively incorrect for HLW, since radioactive waste materials have far lower energy levels in comparison to those of reactors, thereby limiting the risk associated with HLW to much lower levels in virtually all accident scenarios. Given the complexity of the potential risks from HLW, most people will transfer the judgment of the safety of geologic disposal to the experts. The key question is which experts they will listen to. The answer depends on who seems more trustworthy: citizens may have little experience with radioactive waste, but they have considerable experience in evaluating people. The perception of integrity and competence in risk managers depends not only on their personal attributes but also on the character of the policies they implement and the institutions they represent. The current decision process is structured in a way that does not promote trust in those who are implementing the waste management program. The current situation in Nevada, for example, demonstrates the importance of local input in the acceptance of risk. The political leadership of Nevada is fighting the proposed repository and portraying their State as a victim, reinforcing the perception on the part of the broader public that the program is beyond local control. The Department of Energy (DOE) should recognize that communications about the program will be ineffective so long as Nevadans believe they have no voice in the process. To the extent that DOE can share power, however, the increased perception of local control is likely to improve acceptance of a repository. The funding of a technical review group whose members are selected by the State government would be one positive step in this direction. In order to encourage rigorous technical analysis, it should be required that the findings of this review group include a statement of the technical evidence and reasoning behind the conclusions, as is done now by the State of New Mexico's Environmental Evaluation Group for the Waste Isolation Pilot Plant. Given the highly polarized reactions to radioactive waste disposal, it is reasonable to anticipate criticisms and challenges to the technical competence and integrity of the program and its participants. Critics of the program
17 point to the perceived incentives to find the proposed site and technology suitable, the motivation to meet schedules and budgets, and the resulting incentive to disregard or play down troubling findings. Claims to predict accurately events like earthquakes and climatic change are guaranteed to be challenged. These concerns have been addressed through a regulatory re- view process that is carefully designed to reveal errors, optimistic assump- tions, and omissions; but the perceived credibility of that process can be bolstered if state and local groups and individuals have an opportunity to participate, not only in the formal review process but also through informal working relationships with project staff. Those involved in HEW management must also avoid the trap of promising to reduce uncertainties to levels that are unattainable. Uncertainties are certain to persist. Whether the uncertainties in geologic disposal are too great to allow proceeding can only be judged in comparison to the projected risks and uncertainties for the alternatives, such as delayed implementation of disposal or surface storage of spent fuel. As a rule, the values determined from models should only be used for comparative purposes. Confidence in the disposal techniques must come from a combination of remoteness, engineering design, mathematical modeling, performance assessment, natural analogues, and the possibility of remedial action in the event of unforeseen events. There may be public desire or political pressure on implementing agencies to provide absolute guarantees, but a more realistic and attainable goal is to assure that the likelihood of unforeseen events is minimal, and that the consequences of such events are of limited magnitude. Technical program managers may ask whether it is better for the public to know too much or not enough. When unforeseen events occur, for example, the public can raise questions about the validity of the technical approach, as well as the competence of the risk analysis that was used to justify it. Conversely, when foreseen events occur, they lead to questions about why they were not prevented. The technical credibility of the project team suffers in either case, but it probably suffers more when the organization has understated the risk or uncertainty. Moral and Value Issues Overview The foregoing discussion suggests that, in the area of radioactive waste, ethical issues are as important as management and technical decisions. In- terested parties approach the issues with different views about the right way to proceed, often due to differences in moral and value perspectives. As a result, an exploration of ethical issues can illuminate the fundamental policy debates in this field by showing the technical issues in their political and
18 social context. Such an exploration also provides scientists with an oppor- tunity to explore their own ethical responsibilities as they provide society with technical advice on controversial subjects. During its 1988 study ses- sion, the Board examined recent work on ethical questions in radioactive waste management conducted by scholars from a variety of disciplines. These ethical concerns fall into two principal areas: (1) questions concerning the professional responsibility of scientists and engineers; and (2) questions concerning the appropriate uses of science in the decision-making process. Science and engineering are part of broader human activities, and as science enters the public arena, decisions can no longer be purely scientific; good science is not enough. Science has also become an important source of information and analysis for the public policy process, and scientists find themselves being called to account for, and to justify the results of, those decisions. Is this responsible, good, or desirable? How can the process be improved and the parties satisfied? Scientists have been sheltered from such questions in the past, but the increasing scale, sophistication, and per- vasiveness of technical information require a corresponding increase in the sophistication with which these value judgments are made. Three Issues of Equity To see how questions of equity apply to radioactive waste management, consider first a study by Roger E. Kasperson and Samuel Ratick.3 This project identified three sets of equity concerns, each of which raises questions of differential impact, public values, and moral accountability: · Labor. Who does the work and who pays for it? Congress has determined that DOE will be responsible for the work and that the beneficiaries of nuclear power will pay for it through a surcharge on their electric rates. · Legacy. What do we owe to future generations? Moral intuition tells us that our descendants deserve a world that we have tried to make better.4 Posterity matters to us, independent of economic trade-offs; policy should therefore take that interest into account. The EPA regulation requiring evidence that radioactive waste releases will be limited for 10,000 years and more is an illustration of such a concern for the distant future. · Locus. Who benefits, and who is exposed to risk? A repository is the final resting place for the waste from nuclear power plants that provide benefits spread over the whole nation for a short time; but it also concentrates risks and burdens along transportation routes and, for a much longer time, at the disposal site. A radioactive waste repository poses additional complications: it will be the first facility of its kind; the risks it poses are uncertain and, to the extent they exist at all, are likely to emerge over very long time spans; public fears are unusually high; and the history of federal action has raised
19 concerns about whether the interests of local populations will be treated equitably. These ethical questions, when applied to radioactive waste management, demonstrate that once science enters the policymaking arena, good science is no longer enough, because technical decisions are no longer simply sci- entific. When the questions are no longer scientific, scientists alone cannot be expected to answer them. Sheldon Reaven suggests that the Nuclear Waste Policy Act POPPA) creates a "scientific trap," in which citizens are encouraged to expect certainty from flawless science, and in which scientists and engineers are encouraged to believe or pretend that they can supply it.5 Sheila Jasanoff makes the same point: the political need for accountabil- ity in the United States pressures regulators to seek a "scientifically correct" answer, even when there is none.6 The attempt is doomed to scientific and political failure. It is therefore critical to recognize the boundaries of scientific understanding as it can be applied to a societal problem. Five Issues of Policy These ethical considerations have been applied to the current HEW situa- tion by an interdisciplinary team led by E. William Colglazier.7 For each of five key policy issues, the study discusses the "fairness" and appropriateness of the procedures for making decisions, the distribution of costs and benefits, and the type of evidence that is considered sufficient and admissible. The study places special emphasis on the role of scientific evidence because of the large scientific uncertainties and the continuing controversy, even among experts, on what is known and not known. The study's observations include the following: · The need for the repository. The core policy dispute concerns the choice between permanent disposal in a geologic repository and long-term monitored storage in an engineered facility (including at-reactor storage) at or near the surface. The controversy has been over the distribution of costs and benefits to current and future generations and to various stakeholder groups: Pro-nuclear groups feel that the federal government promoted nuclear power and therefore has a special responsibility (spelled out in contractual obligations) to accept spent fuel in a timely manner for permanent disposal. Many environmental groups, on the other hand, view radioactive waste as a special threat to people and the environment; they also favor permanent disposal in order to fulfill this generation's responsibility, and view interim storage as an unfair "legacy" to future generations. Some proponents of interim storage, however, argue that this gen- eration should not make decisions that would be costly to correct in the
20 future; new technological developments may occur over the next century that could change our view of how to handle nuclear waste. In short, all stakeholder groups agree that this generation should ful- fill its responsibility to future generations, but they disagree on how to turn this value principle into policy. · Siting. In making politically difficult siting decisions, political lead- ers have two basic options: make the choice internally and impose it on a weak constituency; or set up and follow a selection process perceived as objective, scientifically credible, and procedurally fair. When NWPA was passed in l9X2, the latter course appeared necessary for both technical and political reasons. However, critics soon claimed that DOE was being political rather than objective in its decisions, citing as evidence DOE's choice of first-round sites and its decision to defer the second round of site selection. This perception led to a stalemate: DOE lacked credibility, and credibility is essential to implement the siting approach set forth in the NWPA. This stalemate was broken by Congress win the 1987 NWPA amendments, which designated Yucca Mountain, Nevada (one of DOE's first-round choices), as the initial site to be characterized and, if acceptable, to be licensed. · Intergovernmental sharing of power. Procedural values were also important in NWPA, which established rules for sharing power among the affected governmental entities. However, the states feel that federal agencies, and especially DOE, have generally chosen to try to meet milestones rather than slow down the process to live up to the spirit of "consultation and cooperation." DOE, for its part, feels that it has a mandate to move forward expeditiously; it has tried to accommodate the states, which (in DOE's view) seek delays to throw obstacles in the way of efficient implementation. Nevada, in particular, interprets the 1987 NWPA amendments as unfair on procedural (as well as distributional and evidential) grounds. · Safer. The fundamental safety issue is the determination of a fair evidential process and standard of proof for showing that the repository is acceptably safe for the thousands of years over which the waste will remain dangerously radioactive. The United States has adopted a set of licensing criteria (e.g., groundwater flow time, package lifetime, waste release limits, and so on) that require considerable certainty. As is often the case with frontier science, however, knowing more may actually increase rather than decrease the uncertainties, at least in the near term. The evidential uncertainties in assessing repository safety may point to a more flexible and evolutionary approach (see below); but this conflicts with the concerns to keep to a fixed schedule, so as to limit costs, discharge obligations to future generations, and meet contractual commitments to utilities holding spent fuel. · Impacts. The debate over the distributional impacts of the repository program include such issues as who should pay for the program, how the impacts can be fairly calculated, and what is fair compensation for negative
21 1 impacts. NWPA determined that the costs should be paid by the beneficia- ries of nuclear-generated electricity through fees, initially, of one mill per kilowatt-hour. An evidential dispute concerns the potential "stigma effect," including lost jobs and lost tax revenues, due to nuclear waste; the social science methodologies for assessing this effect are still controversial. An- other issue concerns the use of incentives and compensation: in the 1987 NWPA amendments, Congress authorized special payments for the host state, provided it forgoes its right to object. This runs the risk of being perceived by opponents as a bribe, offered in exchange for taking otherwise unacceptable risks. Congress also sought a procedural solution to these distri- butional impacts through creation of the Office of Special Negotiator, hoping that the negotiator might find an acceptable arrangement with the host state. Consideration of these policy debates regarding the disposal of radioactive waste leads to three important conclusions: · no interested party has an exclusive claim to be rational or to articulate the public interest; · what is considered fair or unfair is subjective and can change over time; · and with regard to repository safety, the issue is acceptability rather than certainty acceptability being what is acceptable to society, given the evidential uncertainties, perceptions of risk, and contentious stakeholder debates. These conclusions highlight the advantage of an empirical approach one that examines fairness in process, outcomes, and evidence; one that reflects an understanding of the values as well as the interests of the stakeholders. Such an approach may lead to policies that have a greater chance of surviving over time because they are more widely perceived as fair. Modeling and Its Validity Overview Models based on geological principles play a central role in the design and licensing of a waste repository. Because this is where science enters into the design and evaluation process, the Board discusses the appropriate use of models at some length, including the following topics: the purposes for which models are used; the relationship among modeling, treatment of uncertainties, and regulation; and supplements to the use of models in the current program. The role of models in the design and licensing of the repository should properly be understood to be different from the use of models in designing airplanes or licensing nuclear reactors. There are major sources of uncertainty
22 in quantitative geophysical modeling even geohydrology, the best developed, can provide only approximate answers. Geoscientists will need more time to learn how to do more reliable predictive modeling of near-term events, and some events may prove to be chaoticthat is, impossible to predict in detail. In particular, there is a critical need for (1) better communication between modelers and geological experts, in order to improve model prediction; and (2) a more open, quality-reinforcing process such as could be obtained through a peer-reviewed research program at universities and elsewhere. This would do much to improve technical and public confidence in models. DOE could support such an effort by allocating R&D funds, possibly through or in cooperation with the National Science Foundation, for model improvements. In the meantime, however, models can be useful in identifying and evaluating significant contributors to risk and uncertainty. Models are not well suited to describe the risk and uncertainties to lay audiences, however. Natural analogues, if they can be found, are far more useful for this purpose (see below). Problems of repository performance assessment, according to the scheme shown in Figure 1, belong in Region 2 or at the border between Regions 4 and 2. However, there is a general tendency to assume that we can address them using a Region 3 approach: that is, start with a deterministic model that incorporates all "relevant" contributors to overall behavior, and then attempt to collect enough data to move the problem from Region 2 into Region 3. In reality, however, this approach leads to increasingly complex models and increasingly expensive site evaluations, without a concomitant improvement in either understanding or design. Anthony M. Starfield and More Data Figure 1. Types of modeling problems. Region 1 Region 3 Region 4 Region 2 More Understanding
23 P. A. Cundall have suggested that we sometimes demand answers that the model is incapable of providing because of complexity or input demands. The design of the model should be driven by the questions that the model is supposed to answer, rather than by the details of the system that is being modeled. Under the present HEW program, geophysical models are being asked to provide answers to questions that they were not designed to tackle.8 Models and Modeling Problems Figure 1 illustrates a general classification of the types of modeling problems taken from C. S. Holling.9 In Region 1 there are good data but little under- standing; this is where statistics is the appropriate analytic tool. In Region 3 there are both data and understanding; this is where models can be built, validated, and used with conviction. The use of finite-element models in structural design is a good example of Region 3 models. Regions 2 and 4 contain problems that are data-limited in the sense that the relevant data are unavailable or cannot be placed in a rigorous theoretical framework. In Region 2 the understanding of basic mechanisms is good; it is the detailed information that is unobtainable. In Region 4 there is not even a sound understanding of the basic mechanisms and interactions. Appropriate Uses for Geophysical Models In the Board's judgment, a scientifically sound objective of geophysical modeling is learning, over time, how to achieve the long-term isolation of radioactive waste. That is a profoundly different objective from predicting the detailed structure and behavior of a site before, or even after, it is probed in detail. Yet, in the face of public concerns about safety, it is the latter use to which models have been put. The Board believes that this is scientifically unsound. This conclusion is based on review of the modeling approach used by DOE and the regulatory agencies in order to implement the NWPA. In order to support the regulatory and political argument that a site will be safe, it is necessary to make detailed, expensive, and extended extrapolations. These are informed speculations based on existing knowledge. In many instances the guesses are likely to be correct. The geotechnical models used to assure that the foundations of a building or bridge will be secure in the event of earthquakes provide an example of a well-founded predictive use of geophysical modeling. But to predict accurately the response of a complex mass of rock and groundwater as it reacts over thousands of years to the insertion of highly radioactive materials is not possible. This point is important to the public concerns that have surrounded the U.S. radioactive waste program. Use of complex computer models is neces-
24 sary to apply well-known geophysical principles in order to estimate or to set bounds on the behavior of a site, so that its likely suitability for a waste repository can be evaluated. But it is impossible to stretch the almost always incomplete understanding of a site into an accurate quantitative pro- jection of whether a repository will be safe if constructed and operated there. Even after a detailed and costly examination of the site itself, only an informed judgment can be reached, and even then there will be uncertainties. As modelers have become more aware of the processes they are attempt- ing to model, they are also recognizing that the geological environment is more complex than originally thought and that quantitative prediction is correspondingly more difficult and uncertain. Many computer simulation models of geological environments are based on deterministic models that have been used successfully in branches of mechanics such as aerospace engineering, where the basic phenomena are much better defined. Such models are of limited value for the ill-defined, data-limited, long-term situations such as the repository isolation problem. It is illusory to expect accurate quantitative estimates of radionuclide releases from them. Sources of Uncertainty in Geophysical Models Performance assessmentsestimates of the repository's ability to isolate HEW are based on current computer simulations and parameters derived from laboratory and field measurements. As a consequence, they will have large uncertainties associated with the predicted performance. These uncertainties could pose serious obstacles in demonstrating compliance with licensing requirements. Discussions at BROOM's 1988 study session identified four principal causes of uncertainty: 1. Structural uncertainty. Do the equations adequately represent the operative physical processes? Do we in fact understand the system well enough to model it mathematically? Modeling will be most successful in solving Region 3 problems (see Figure 1), where we have a great deal of data and a good understanding of how the system works. 2. Parametric uncertainty. Have we chosen the right values for the variables (e.g., permeability) in the equations? Have we in fact chosen the right variables to represent the behavior of the system? Are our measurement techniques valid? Will they produce enough, and good enough, data? 3. Uncertainties in initial arid boundary conditions. Have we inter- polated adequately from a few spatially isolated point measurements to a broad three-dimensional domain (e.g., groundwater, heat, in situ stress)? 4. Uncertainties in forcing functions. How well can we characterize past and future events that might play a part in the fate of the repository (e.g., climate, tectonics, human intrusion)?
25 Urgent attention should be given to examining these and other causes of uncertainty, but even with continuing research along the present lines, im- provement will come slowly. It may even turn out to be appropriate to delay permanent closure of a waste repository until adequate assurances concern- ing its long-term behavior can be obtained through geophysical studies. Judgments of whether enough is known to proceed with placement of waste in a repository are needed throughout the life of the project. But to repeat the Board's earlier point: these judgments should be based on a comparison of the available alternatives, rather than just a simplistic debate over whether, given current uncertainties, a repository site is "safe." Even when the detailed behavior of an underground repository is still under study, it may well be safer to put waste there, in a way that permits retrieval if necessary, rather than leaving it at reactors or in storage at, or near, the surface of the earth. Modeling Limitations An Example The inherent diff~culdes of modeling are illustrated by the case of Groundwater flow, which is used as an example precisely because it the best developed in terms of modeling. Groundwater flow has been extensively modeled for a broad range of engineering problems, and it consequently has a richer base from which to draw than do many other aspects of repository isolation. Groundwater flow is also generally accepted as the primary mechanism by which radionuclides could move from the repository to the biosphere, so it has been emphasized in modeling studies of repository isolation. Several experts, however, have commented on the difficulty of applying classical hydrology models to the problem of radioactive waste isolation. Groundwater hydrologists are becoming increasingly aware that inadequate and insufficient data limit the reliability of traditional deterministic [distributed- parameter] Groundwater models. The data may be inadequate because aquifer heterogeneities occur on a scale smaller than can be defined on the basis of available data, time-dependent variables are monitored too infrequently, and measurement errors exist.10 To carry out these [repository flow] calculations, hydrogeologists are apply- ing geostatistical models and stochastic simulation methods originally developed to assess piezometric response in near-surface unconsolidated aquifers over limited spatial distances and short time frames with relatively abundant data. . . . These techniques may not be as valuable when applied to the assessment of radionuclide transport in deep rock formations, over large distances and long time frames, under conditions of sparse data availability.... [The authors] have repeatedly drawn attention to the potential problems associated with the geostatistical methods (Bayesian and otherwise) when data networks
26 are sparse and sample sizes small. Id our opinion, this is the potential Achil- les heel for Me application of geostatistics at nuclear repository sites. With regard to repository isolation modeling, increased study has thus far resulted in the identification of greater complexity. Progress is being made toward including some of this complexity in the models, at least in terms of groundwater studies; but other geotechnical aspects of repository isolation (such as constitutive properties of rock joints, excavation and repository scale deformation behavior, and regional in situ stress) are far less developed. It will take years of additional research to represent them adequately in the models. As a result, the prospects are poor, especially in the short term, for models that can produce reliable quantitative measures of isolation performance. Appropriate Objectives for Modeling Repository performance assessments are unlikely to prove beyond doubt that risks are below established limits. Nor do the regulations require it- EPA requires only a "reasonable assurance." The problem is that in a case without clear precedents, it is unclear what is "reasonable." The Board's point is that unsound use of technical information is not a proper substitute for the political reasoning that, in a democratic society, must in the end win consent for taking reasonable steps to advance public health and safety. In light of the limitations of technical knowledge, the Board concludes that it makes sense to conduct the assessments through an iterative process, in which the assessment provides direction to those characterizing a repository site and developing the repository engineering features. As further information is developed about the candidate site, it is also used in the performance assessment. Many of the uncertainties associated with a candidate repository site will be technically interesting but irrelevant to overall repository performance. Conversely, the issues that are analytically tractable are not necessarily the most important. A key task for performance modeling is to separate the significant uncertainties and risks from the trivial. Similarly, when there are technical disputes over characteristics and processes that affect calcula- tions of waste transport, sensitivity analysis with alternative models and parameters can indicate where further analysis is required and where enough is known to move on to other concerns. Using Models to Reduce Uncertainty Models do have an indispensable role in developing understanding of such problems, provided that the models are developed and used within the proper limitations. In other words, modeling can be used to improve models.
27 The following quotations from those concerned with such problems illustrate this point: . . . much time can be saved in the early stages of hypothesis formulation by Me explarai~on of these hypotheses Trough mathematical models. Similarly, mathematical models cart be used to investigate phenomena from the view- point of existing theories, by the integration of disparate theories into a single working hypothesis, for example. Such models may quickly reveal inadequacies in the current theory and indicate gaps where new theory is required. The updating properties of Me Bayesian approach . . . are well suited to Me iterative approach we espouse for die model~ng/data ga~er~ng sequence at a site. We feel that the first modeling efforts should precede or accompany initial site investigations.13 A good example of this general approach is the "regionalized sensitivity analysis" approach, by which G. M. Hornberger and his collaborators have been able to identify the "critical uncertainties" in applying a particular model to several data-sparse ecological problems and, thereby, to define programs of investigations to reduce those uncertainties.~4 In summary, models should be qualitatively sensible, robust to sensitivity analysis, and independent of minor effects or processes, and they should include acceptable levels of uncertainty. However, models cannot prove that the repository is safe, nor can they resolve public concerns about the repository. Supplements to Modeling Natural Analogues. Because models cannot be conclusive with regard to the safety of a repository site, it is important to think carefully about natural analogues. These are natural "test cases," geological settings in which naturally occurring radioactive materials have been subjected to environmental forces for millions of years. These natural experiments demonstrate the action of transport processes that are similar to those that will govern the release of man-made radionuclides from a repository in a similar setting. The natural analogue approach depends, of course, on whether the natu- ral case is in fact an analogue for a repository situation. Where there is scientific agreement that the analogy applies, however, the approach is powerful because it allows us to predict processes with confidence over many millennia. And natural analogues can serve two additional roles: (1) they can provide a check on performance assessment methodology, and (2) they may be more meaningful than sophisticated numerical predictions to the lay public. The alternative management strategy described in the following section would make substantial use of natural analogues, such as undisturbed natural de-
28 posits of radioactive elements and groundwater systems, in order to illumi nate the behavior of the geologic environment. Professional Judgment. A second approach is to use the professional judg- ment of technical experts as an input to modeling in areas where there is uncertainty as to parameters, structures, or even future events. Such judgments, which may differ from those of DOE program managers and their staffs, should be incorporated early in the process. A model created by this process can redirect the DOE program substantially. It is important to bear in mind that all uses of technical information entail judgments of what is important and what is less so. If the technical community is to learn from the successes and failures of the DOE program, it is essential that these technical judgments be documented. Setting out the reasoning of DOE staff and of independent outside experts contributes to learning and builds credibility in the process even when the experts disagree with DOE staff and among themselves. Implications for Program Management The Board has concluded that geological models, and indeed scientific knowledge generally, are being inappropriately applied in the U.S. radioac- tive waste repository program. That misapplication prompts this Board to outline an alternative management strategy. The next section describes an alternative management approach that employs natural analogues and professional judgment in a program design that uses science appropriately in the search for a safe disposal system. Putting such an approach into effect, however, would require major changes in the way Congress, the regulatory agencies, and DOE conduct their business. Such changes will be difficult to achieve, but the Board has reluctantly concluded that nothing else will put to rest the problems that plague the national program today. Strategic Planning Overview There is no scientific reason to think that an acceptable HEW repository cannot be built and licensed. For historic and institutional reasons, however, DOE managers often feel compelled to "get it right the first time." This management strategy runs the risk of encountering "show-stopping" problems that may delay licensing and will certainly cause further deterioration of public and scientific trust. The alternative would be a more flexible, experimental strategy that em- bodies the following principles:
29 · respond with conservative design changes as site attributes are discovered; · use modeling to identify areas where more information is needed; and · allow for remediation if things do not turn out as planned. Implicit in this approach is the need to revise both technical design and regulatory criteria as more information is discovered. This is difficult to achieve in a governmental structure that disperses authority among legisla- tive and executive agencies and separates regulation from implementation. When presented with intense controversy, such an institutional arrangement breeds disgust among governmental units and the public. In that setting, partial remedies further entangle the procedural morass. More practically, however, DOE can enhance the credibility of the program and reduce the likelihood of late-stage surprises by (1) encouraging effective communication within its complex management structure; and (2) providing incentives for field personnel to identify and solve problems. DOE and the USNRC can also enhance credibility by encouraging periodic external reviews of the repository design, construction, and licensing requirements and associated processes. Policy Context The present U.S. approach to HLW disposal is increasingly vulnerable to being derailed by minor surprises. This vulnerability does not arise from a lack of talent or effort among the federal agencies and private contractors working on the program. Nor does the design or construction of the repository represent an unusually difficult technical undertaking. Instead, the program is at risk because it is following the wrong approach to implementation. The current predetermined process, in which every step is mandated in detail as in the more than 6,000-page "Site Characterization Plan,"is is in- appropriate. The current policy calls for a sequential process in which EPA and the USNRC first establish the criteria for safe disposal, and then DOE describes in detail what steps will be taken to move through site characterization, licensing, and operation of the facility. The result of this approach is that any late change, by any of the participating agencies, is taken as an admis- · ~ slon ot error. And late changes are bound to happen. One worker was killed and five injured in an HLW repository under construction in West Germany when a support ring failed unexpectedly. At the Waste Isolation Pilot Plant (WIPP) in New Mexico, the discovery of pockets of pressurized brine in formations below the repository level led to public outcries and a continual National Research Council review of the suitability of the site. The United States seems to be the only country that has taken the ap- proach of writing detailed regulations before all of the data are in. Almost
30 all other countries have established limitations on the allowable levels of radiation dose to individuals or populations resulting from repository estab- lishment but have taken a "wait and see" approach on design, while collecting data that may be of use in setting design. The United States, on the other hand, seems to have felt that detailed regulations can be, in fact must be, written without regard to any particular geological setting or other circumstance. As a direct consequence, the U.S. HEW program is bound by requirements that may be impossible to meet, even though overall dose limits can be achieved. Alternative Management Strategies The preceding sections have shown that there are a number of unresolved issues in the U.S. radioactive waste disposal program, as well as (and in part because of) high levels of uncertainty and public unease about the performance of the repository. The Board's consideration of these subjects indicates that the proper response to distrust is greater openness in the process, and that the proper response to uncertainty is greater knowledge and flexibility, as well as redundancy of barriers to nuclide transport. The U.S. program will continue to face controversy until it adopts a management strategy based on these principles. The current approach to the design, construction, and licensing of the Nevada site is derived from the philosophy and procedures used for licens- ing nuclear power plants. The characteristics of the repository and its geological setting are carefully determined and specified as a basis for a complex set of calculations that describes the behavior of the system. This model is used to generate predictions of the migration of radioactive elements into the biosphere and analyzes the consequences of various events ("scenarios") that might affect the site over the next 10,000 years, in order to demonstrate that the repository site meets regulatory requirements (i.e., is "safely. Based on the model and geologic studies of the site, the construction of the repository is specified in detail and then carried out under an aggressive quality assurance program, which is designed to withstand regulatory review and legal challenge. Within these requirements it is the geological setting that ensures isolation, not the engineered characteristics of the system; closure aims for complete entombment and discourages subsequent remediation. For all the reasons discussed above, a management process based on the regulation of nuclear power stations (a Region 3 type problem: see Figure 1) is inappropriate to the development of a waste repository. A well-documented alternative to this approach is being followed, to various degrees, by countries such as Canada and Sweden. The exploration and construction of a geological test facility and a low-level waste repository, respectively, follow a flexible path, allowing each step in the characterization
31 and design to draw on the information and understanding developed during the prior steps, and from prior experience with similar underground construction projects. During and subsequent to the closing of the repository, the emphasis will be on monitoring and on the ability to repair, in order to minimize the possibility that unplanned or unexpected events will compromise the integ- rity of the disposal system. Engineered modifications can be incorporated (e.g., in the waste containers or in the material used to backfill the repository) if the computer models suggest unacceptable or irreducible uncertainties in the performance of the unmodified containment system. The Canadian experience at their Underground Research Laboratory provides a good example. All of the major rock structures and groundwater conditions were defined from surface and borehole observations before shaft construction began. Detailed geological structure can never be totally determined from surface information, however, and the final details of the facility design were modified to take account of information gathered during shaft construction. What are the risks and benefits of the two approaches? The U.S. approach facilitates rigorous oversight and technical auditing. Its goals and standards are clear, and, if carried out according to specifications, this approach is robust in the face of administrative or legal challenge. It is designed to create a sense of confidence in the planning and operation of the repository, and it facilitates precise answers to specific technical questions. However, such an approach is not consistent with normal geologic or mining practice. It assumes that the properties of the geologic medium can be determined and specified in advance to a degree analogous to that required for man-made components, such as reinforcing rods, structural concrete, or pipes. In reality, geologic exploration and mine construction never proceed in this way. Most underground construction projects are more qualitative, using a "design (and improve the design) as you go" principle. New sections of drill core often reveal surprises that must be incorporated into the geologists' concept of the site, integrated with past experience, and used to modify the exploration plan or mine design. In a project where adherence to predeter- mined specifications is paramount, the inherent variability of the geologic environment will result in endless changes in the specifications, with resultant delays, frustration for field personnel, high overhead costs, and loss of public confidence in both the suitability of the site and the competence of the professionals working on the project. The second approach has more in common with research than with con- ventional engineering practice. This approach continually integrates new data into the expert judgments of geologists and engineers. It makes heavy use of natural analogues, such as undisturbed natural deposits of radioactive elements and groundwater systems, in order to illuminate the behavior of the geologic environment. It can immediately take advantage of favorable surprises and compensate for unfavorable ones. That this approach works
32 well is evidenced by the enormous number of underground construction projects in diverse geologic settings that have been completed successfully around the world. These projects were not designed to contain radioactive waste for thousands of years, but many of them faced technical problems of comparable magnitude, such as crossing active faults, sealing out massive groundwater flows, or stabilizing highly fractured and structurally weak rock masses. The second approach, with its reliance on continuous adaptation, would be much more difficult to document, audit, and defend before a licensing authority or court of law than is the more prescriptive approach. Some aspects of quality assurance can work well, such as document and sample control, the use of standard procedures and tools, and personnel qualifications. Other quality assurance techniques are likely to be contentious and may be impossible to implement in the same way they are implemented in nuclear power plants, including design control, instructions, procedures, drawings, inspections, and control of nonconforming items. An alternative is to use an aggressive and independent peer review system to appraise the decisions made and the competence of the technical personnel and managers responsible. The legal system is able to accept expert opinion as a basis for action or assessments of action, but one cannot predict whether a repository could ever be licensed in the face of the batteries of opposing "experts" who would inevitably be called on to assess a flexibly designed and constructed repository for HEW disposal. The debate will hinge in part on a clear understanding of the alternatives against which a proposed "solution" will be judged. By contrast, the EPA standards and USNBC regulations define requirements that, if met, form the basis for the presumption that the facility is "safe." Given the unhappy history of radioactive waste disposal in the United States, however, one very real and likely alternative is that nothing at all will be done. In judging disposal options, therefore, one should also adopt inaction or some other likely scenario as a default option, so that comparisons can be made and progress consistently assessed over time. The combination of a conservative engineering approach and designed-in maximum flexibil- ity, to allow unanticipated problems to be corrected, should reassure both technical experts and concerned nonexperts. The barrier is not logical but institutional, and the prescriptive approach in the U.S. program is dictated by a governmental structure that separates regulation from implementation. Within the present program, for example, "quality assurance" has be- come the bete noire of frustrated field personnel, who are trying to work within a system that is hostile to surprises in a world that is full of them. Because almost any geologic phenomenon has more than one possible cause, flexibility (including the recognition that uncertainty is inevitable and must be accommodated) is more likely to lead to the design and construction of a
33 safe repository system than are rigid, predetermined protocols. In employ- ing and evaluating such an adaptive approach to construction, emphasis focuses on those decisions that have irreversible or noncorrectable consequences on disposal, rather than on the myriad small adjustments that do not affect the basic flexibility and robustness of a repository. The Elements of a More Flexible System In a program governed by this alternative approach, change would not be seen as an admission of error; the system would be receptive and responsive to a continuing stream of information from site characterization. The main actors would reduce their reliance on detailed preplanning during initial site characterization, making it possible to debug the preliminary design during rather than before characterization. But the necessary conditions of the system are flexibility and resiliencyflexibility to respond rapidly to ongoing findings in the geology, geohydrology, and geochemistry (within broad constraints); and resiliency to continuously adjust the performance assessment to reflect new information, especially where such information indicates possible precursors of substantial increases in risk. These qualities could be devel- oped through the following steps: Iterative performance assessment. The basic approach outlined here would start with a simplified performance assessment, based on known data and methods of interpretation. Given the inherent uncertainties and techni- cal difficulties of the process, the present system may well expend large efforts on small risks, and vice versa. An iterative approach, on the other hand, could allow characterization efforts to give priority to major uncertainties and risks, while there is still time and money left to do something about them. As in probabilistic risk assessment, analysis focuses on efforts to reduce the important risks and uncertainties. In this case, that means acquiring information on the design features and licensing criteria that are most likely to determine whether the site is suitable or should be abandoned. · Fixing problems vs. anticipati1lg problems. The underlying concept of the present, anticipatory U.S. management strategy is "Get it right the first time." One result is a 6,300-page "Site Characterization Plan" for Yucca Mountain. For the reasons described above, however, a process based on getting all of the needed measurements and analysis on the first pass, with acceptably high quality, is not likely to succeed. The geological environment will always produce surprises, like the pockets of pressurized brine at WIPP. No matter what technical approach is initially adopted, the design can be improved by matching it with specific features of the site. Experiments are now being conducted at WIPP with backfill material and other engineered barriers that were not part of the original design. These ~ , _ , _ ~
34 are being Died as ways to make the disposal system as a whole robust in the face of newly discovered uncertainties in the geology. · Deft ne the problem broadly. As characterization proceeds, especially if it is done without the guidance of iterative performance assessment, DOE may eventually find it difficult or impossible to meet some of the criteria set by the USNRC and/or EPA. This will not mean that Yucca Mountain is unsuitable for a repositorythe problem could be with the detailed criteria. This is no reason to arbitrarily abandon the release limits it is the more detailed requirements that may need to be reconsidered, since they ultimately affect the release limits and the imputed dose. However, one should not take EPA's release standards or the USNRC's detailed licensing requirements as immutable constraints. They are roadmarkers to, and surrogates for, dose limits. Although the EPA standards and the USNRC regulations recognize and accept a certain level of uncertainty, the discussion to date of the application of these standards and regulations does not warrant confidence in the acceptance of uncertainty in licensing procedures. Some process is needed in order to determine whether DOE's inability to meet a particular requirement is due to a disqualifying deficiency in the site or to an unreasonable regulatory demand, one that is unlikely to be met at any site and is unnecessary to protect public health. And to the extent that regulatory criteria can be corrected earlier instead of later in the process, they are more likely to be perceived as technical adjustments rather than as a diminution of public safety. Given the history of U.S. efforts to dispose of radioactive waste, current plans for the program have little chance of progressing without major modification in the 20 years or more that will be required to get a repository into operation.
