Return to the Main Document

I. Introduction

This is the final report of the Total System Performance Assessment (TSPA) Peer Review Panel (the Panel). Although the Panel has prepared three interim reports, this final report is intended to stand alone. Hence, some of the recommendations and observations contained in the interim reports are repeated here.

A. Nature of TSPA Peer Review Process

In the Energy and Water Appropriations Act for fiscal year 1997, Congress specified four components of a viability assessment for a proposed high-level radioactive waste repository at Yucca Mountain, Nevada. One of these was to complete:

…a total system performance assessment, based upon the design concept and the scientific data and analysis available by September 30, 1998, describing the probable behavior of the repository in the Yucca Mountain geological setting relative to the overall system performance standards.

The Total System Performance Assessment supporting the Viability Assessment (TSPA-VA)(
CRWMS M&O, 1998d) has now been completed by the Civilian Radioactive Waste Management System Management and Operating contractor (CRWMS M&O) for the U.S. Department of Energy (DOE) Yucca Mountain Site Characterization Office and publicly released. Note, however, that the U.S. Environmental Protection Agency (USEPA) has not yet issued proposed standards against which overall system performance will be measured. Regulations have been proposed by the U.S. Nuclear Regulatory Agency (USNRC), but they will be subject to revision once the USEPA standards are set.

The task of the Panel, according to the Peer Review Plan (Appendix B), was to:

…conduct a phased review over a two-year period to observe the development and completion of the TSPA-VA. The comments, concerns, conclusions, and recommendations of the final Peer Review Report will be provided to the M&O to support the development and conduct of the License Application TSPA (TSPA-LA). As such, the Panel members will consider not only the analytical approach of the TSPA-VA, but also its traceability and transparency.

The Panel members were to evaluate the analytical approaches used in preparing the TSPA-VA, with specific attention directed to a range of aspects, including:

  • physical events and processes considered in the analyses,
  • use of appropriate and relevant data,
  • assumptions made,
  • abstraction of process models into the total system models,
  • application of accepted analytical methods, and
  • treatment of uncertainties.
  • These aspects were to be evaluated within the context of their significance to the long-term performance of a repository at Yucca Mountain.

    During its two year review, the Panel has discussed and evaluated the adequacy of the total system framework used for the TSPA, the way the individual components of the system were modeled and analyzed, and the significance of the component models to the overall results.

    B. Content of Panel Reports

    As noted above, the Panel has issued three interim reports, written prior to the completion of the TSPA-VA, and has prepared this final report after reviewing the completed TSPA-VA. A brief review of the coverage in each of the reports is presented below.

    Interim Reports

    The three interim reports should be viewed as a series. In successive reports, the Panel did not repeat comments found in previous reports except where the Panel amplified, extended, or revised its previous comments.

    In its first report (
    Whipple et al., 1997a), submitted on June 20, 1997, the Panel provided an overview of the TSPA-VA approach and discussed its understanding of processes and events that would affect the long-range performance of a repository at Yucca Mountain and how they were being considered in the TSPA-VA.

    The second report (Whipple et al., 1997b), submitted on December 12, 1997, covered general topics that were not covered in depth in the first report and specific issues that the Panel selected because of their potential significance to the results of the TSPA-VA. In addition, the Panel discussed its view of the role of the TSPA-VA, the expectations that could reasonably be set for the TSPA-VA, and how results were being interpreted and limitations and uncertainties were being addressed by the TSPA staff. The report also described in more detail the Panel’s understanding of how the processes and events that could affect the performance of a repository at Yucca Mountain were being analyzed in the TSPA-VA.

    In the third report (Whipple et al., 1998c) submitted on June 25, 1998, the Panel discussed how the TSPA-VA project staff described the way the repository was expected to work based on what is called the "base case" analysis; the importance to the current analysis of focusing on early canister failures and other events that could lead to releases and doses within 10,000 years; the methodology being used for sensitivity analysis; and, as in the second report, the Panel’s understanding of how the processes and events that could affect the performance of a repository at Yucca Mountain were being analyzed in the TSPA-VA.

    The Panel provided a summary of its interim findings in each report. The three interim Panel reports are available online at http://www.ymp.gov/reference/va/tspa.htm.

    Final Report

    As required by the Peer Review Plan, the Panel had two major objectives in preparing this final report:

  • To describe the technical strengths and weaknesses of the TSPA-VA.
  • To provide suggestions for moving toward the TSPA-LA, if this step is deemed appropriate.
  • During earlier reviews when complete documentation was not yet available, the Panel supplemented its review of the draft TSPA-VA with formal and informal meetings and interactions with the project staff. In contrast, the Panel’s review in this final report is based primarily on documented work, namely, the completed TSPA-VA (
    CRWMS M&O, 1998d), the supporting Technical Basis Documents (CRWMS M&O, 1998c), and other project documents listed as references to this report.

    In Section II, the Panel presents its main findings. These are the issues that the Panel believes are important to the overall credibility and usefulness of the TSPA.

    In Section III, the Panel discusses the overall framework of the TSPA-VA, including the uncertainties inherent in modeling the probable behavior of a repository at Yucca Mountain, and draws a distinction between those contributors to uncertainty in performance that can be improved through refinement of models and collection and application of data versus those uncertainties that can be better managed by the use of bounding analyses or design changes. The section concludes with the Panel’s recommendations.

    In Section IV, the Panel describes its understanding of how the processes and events that could affect the long-range performance of a repository at Yucca Mountain were analyzed in the TSPA-VA and presents its findings and recommendations for each element. As in the interim reports, the discussion follows the major elements examined in the TSPA-VA analysis: (1) initial conditions of the site; (2) conditions as affected by the repository; (3) isolation as provided by the geologic setting in which the proposed repository is to be located, (4) isolation as provided by the waste form and the engineered barrier system; (5) release and transport of radionuclides from the repository; (6) their movement within the biosphere, interaction with exposed population groups, and estimates of the resulting doses; and (7) disruptive events and climate.

    II. Main Findings

    Although the TSPA-VA is a comprehensive and complex analysis about which few generalizations apply, the Panel has reached a few overall conclusions regarding those issues we consider to be most important to an overall understanding of its strengths and weaknesses. Our findings are often based on or refer to specific aspects of the TSPA-VA, and should be read in the context of the more detailed information provided in
    Sections III and IV.

    A. Reliability of the TSPA-VA Results

    Key Points

    Because of the inadequacy of the supporting evidence, the Panel could not confirm whether a number of the TSPA-VA component models are representative of the systems, components, and processes they were designed to simulate. In addition, several of the component models are likely to be conservative and others non-conservative. For these reasons, it is unlikely that the TSPA-VA, taken as a whole, describes the long-term probable behavior of the proposed repository.

    1. With the benefit of hindsight, the Peer Review Panel finds that a credible assessment of the future probable behavior of the repository is beyond current analytical capabilities, given the complexity of the system and the nature of the data that now exist or that could be obtained within reasonable time and cost. The TSPA-VA team has performed well, has developed numerous analytical innovations, and has produced technical reports of exemplary clarity. The failure of the TSPA-VA to capture the probable future behavior of the proposed repository system is due in large part to the difficulty of the problem, including the long time scales over which performance is to be described and the large and heterogeneous physical setting that is addressed by the analysis. This difficulty was compounded by a failure, in many elements of the analyses, to initiate and complete the necessary research, develop the appropriate models, and collect and apply the needed data and information.
    2. The TSPA-VA was a necessary and useful step in the evolving understanding of how a repository could be expected to perform at Yucca Mountain. It has produced valuable insights into the performance of various repository components, and has helped identify issues where additional data and analyses could lead to improved understanding of the repository’s performance. It is also useful in identifying aspects that are comparatively unimportant to performance, for which additional data and analysis are not likely to be beneficial.
    Until there are improvements in the specific subsystem models for key elements of the system, in their supporting databases, in the coupling between certain aspects of the modeling, and in the use of tests, overall conclusions based on the analyses should be viewed skeptically, and decisions based on the analyses should be made cautiously. For example, the comparison of alternative designs based on the estimated doses at a distance of 20 km from the repository and 10,000 or more years into the future requires a degree of resolution that the TSPA-VA may not be able to provide.

    The Panel recognizes that substantial amounts of field and experimental data have been developed in support of the TSPA-VA; to be credible, however, many elements of the analysis require additional data. These data needs are of two types: fundamental data that are essential to the development and implementation of the models, and data sets designed to challenge conceptual models and test the coupled models used in the TSPA-VA. While it is obviously not feasible to test the full TSPA-VA, it is feasible to test many of the individual component models experimentally. The Panel notes that many experiments are planned or in progress that should be useful in confirming, calibrating, or invalidating the models being applied to analyses of the anticipated conditions at Yucca Mountain. Such experiments cannot be replaced by sensitivity analyses because the sensitivity analyses used in the TSPA-VA often do not directly address the uncertainties associated with the experimental database or the selection of a conceptual model.

    The objective for the TSPA-VA was to assess the probable behavior of the repository. In contrast, the objective for the TSPA-LA will be to determine whether it can be shown with reasonable assurance that the repository complies with the applicable regulatory limits. These are significantly different objectives, and recognition of this distinction should be an important element of a path forward to the TSPA-LA. This issue is discussed in more detail in
    Section III.

    Elements of a Credible Analysis

    To be credible, the analysis would have needed to include:

  •  
  • Component subsystem models that capture important and relevant phenomena;
  •  
  • Adequate databases;
  •  
  • Proper coupling between the subsystem models; and
  •  
  • Tests of modeled behavior.

    Although the TSPA-VA offers many examples of partial, even substantial, success in each of these four areas, the Panel has also observed examples of important deficiencies in each.

  •  
  • Concerning subsystem models, the final dose estimates within the TSPA-VA rest in large part on potentially optimistic, or at least undemonstrated, assumptions about the behavior of certain barriers in the system (for example, performance of the cladding and the waste package).
  •  
  • Concerning databases, some of the important analyses are not supported by an adequate database, (for example, databases for the corrosion of spent fuel and the saturated zone analysis).
  •  
  • Concerning coupled processes (that is, thermohydrological, thermomechanical, and thermochemical effects) and the data and models that support them, the Panel believes that it may be beyond the capabilities of current analytical methodologies to analyze systems of such scale and complexity. For this reason, the effects of coupled processes can probably best be dealt with through a combination of bounding analyses and engineered features designed to minimize the effects of such processes.
  •  
  • Concerning tests of modeled behavior, the TSPA-VA does not contain the convincing direct measurements or confirmation of the modeled behavior of components or subsystems for which testing is feasible. This testing should be a part of the analyses of such a complicated system.

    Although the TSPA-VA contains many useful sensitivity analyses that illuminate the system’s behavior, these analyses taken as a whole do not provide sufficient insights to overcome the above deficiencies and uncertainties.

    In addition to the specific mechanisms that could cause the repository to fail to perform as projected, the overall complexity of the system and the resulting high uncertainties may lead to difficulties in licensing. The USNRC has considerable experience in defining "reasonable assurance" for the licensing of commercial nuclear power plants, but the short time scales and lack of geological complexity of such plants relative to those associated with the proposed repository could be interpreted to mean that the USNRC’s experience in defining reasonable assurance may not be applicable to a repository. In the case of the proposed repository, it is possible that "reasonable assurance" will require a degree of proof that is currently not available for the reference design and the site. For example, two aspects of repository performance with large uncertainties are the near-field geochemical environment and actinide transport by colloids. Depending on the degree of confidence that is necessary for licensing, the current analyses may not be adequate. As discussed in
    Section III, the Panel believes that, for many issues to be addressed in the TSPA-LA, use of simplified bounding analyses may be necessary to achieve the desired degree of confidence.

    B. Advances and Improvements in the TSPA-VA Analysis

    The TSPA-VA document contains many substantial advances and improvements over the previous TSPA reports issued in 1991, 1993, and 1995. It also provides a revised understanding of how a repository at Yucca Mountain would perform, in comparison to these earlier studies. Perhaps the most dramatic change has been in the estimate of the current infiltration rate, which has been revised upwards by one order of magnitude. This revision has led to many changes in the conceptual models of flow and transport in the mountain, in the repository, and in the saturated zone. In particular, the interpretation by the project staff of flow and transport has shifted from predominantly matrix flow-dominated regimes to fracture flow-dominated regimes that involve fast paths.

    A revised scenario on future climates, emphasizing higher rates of rainfall, was also a new addition to the TSPA-VA. The increased rate of infiltration assumed in the analysis leads to the projections of more rapid transport in the unsaturated zone and a larger volume of water coming into contact with waste canisters and waste, a change that, in turn, leads to the projections of earlier and more rapid releases of radionuclides from the waste packages. In response, design changes were made that were intended to strengthen the performance of the engineered barriers. In comparison to previous analyses, the degree to which waste isolation is dependent on the geologic setting has been reduced in the TSPA-VA; the required contribution to isolation from the engineered barriers has significantly increased.

    The revised interpretation of infiltration has led to a more realistic, although still far from complete, representation of the fracture-matrix interaction and to an improved characterization of the hydrologic properties of the mountain. Although much is still unknown and/or uncertain, there have been several advances in addressing the problem of modeling the performance of the site. A significant development in the current TSPA-VA is the incorporation, for the first time, of a model of seepage into the drift. This addition relates to the hydrology of the near-field environment, which is modeled at a scale much smaller than in the previous TSPA reports. Drift-scale hydrology critically affects the rates of canister failure, and thus represents a key consideration in the analysis of repository performance. The , currently under way, and other investigations in the engineered barrier system will be useful in testing the current models. The hydrology at this scale is also linked to processes at the mountain scale, at least during the thermal period, and to thermal-hydrologic, thermochemical, and thermomechanical processes. The coupling of these processes was not considered in the TSPA-VA.

    Substantial improvements have been made in the modeling of flow and transport at the various space and time scales, and in the linkage between these scales. The Panel notes, in particular, advances in the use of more sophisticated numerical models such as the dual permeability model, the finite element heat and mass model, and particle-tracking, and decreasing reliance on the limited equivalent-continuum model. The hybrid approach for assessing the effects of thermohydrologic processes at different scales and from different heat sources, in a computationally manageable scheme, as illustrated in Chapter 3 of the Technical Basis Document (
    CRWMS M&O, 1998c), represents a significant improvement, even though it is far from complete.

    The project team has incorporated a dramatic and needed improvement in numerical modeling in the area of transport in the saturated zone, where they have abandoned the previous finite-difference model in favor of a streamtube-based approach. Although the adoption of a streamtube approach based on an overall dilution factor is less desirable than a more detailed treatment of dispersion, it is appropriate, given the limitations in the data concerning the saturated zone. The new model eliminated numerical dispersion errors, inherent in the previous model, and may provide a more realistic prediction of dilution in the saturated zone. This model is not physically representative of the saturated zone transport for isolated waste package failures, however, although sensitivity analyses indicate that the model overestimates dilution for such cases by perhaps a factor of three. This factor is small in comparison to the other uncertainties in the assessment of the saturated zone.

    The project team and its supporting contractors have made advances that have not been included in the current version of the TSPA-VA. Many of these are improvements in basic scientific issues, and they lead to an improved understanding of fundamental processes and the projected repository performance. Of particular significance are the analyses of coupled thermohydrological, thermochemical, and thermomechanical effects described in publications by Wilder (1996), Hardin and Chesnut (1997), and Hardin (1998). These represent serious attempts to study the complex interaction problems resulting from the thermal pulse. Also worthy of note are the sensitivity analysis of saturated zone (Arnold and Kuzio, 1998), the introduction of geostatistics for assessing flow, transport and retardation in the unsaturated zone; and the analyses of the various tests currently under way, including the single heater test, the large block test, and the .

    C. Key Role of the Waste Package

    Corrosion Resistance

    Based on the TSPA-VA results, the analysts have judged that the rate of waste package degradation is a principal determinant of overall repository performance. Hence the integrity of the waste package as a barrier is crucial to the performances of the repository as projected within the TSPA-VA for each of the time periods considered: 10,000, 100,000 and 1,000,000 years (
    CRWMS M&O, 1998d, pages 0-16). The Panel agrees that this is a logical and credible conclusion. Therefore, it is important that the factors that control the life of the waste packages be carefully identified and analyzed.

    Corrosion is the most important and realistic threat to the deterioration of the waste packages. Although an outer canister of steel will readily corrode in hot, moist conditions, under the proper environmental conditions (temperature-relative humidity-water chemistry), C-22, a proposed waste canister material, can realistically be expected to remain passive for long periods of time. Since such conditions are anticipated to prevail within the proposed repository, it is likewise anticipated that C-22 will corrode at an extremely slow rate. At this stage of development, however, the analyses leading to this assumption are necessarily a simplification of a complex problem and far from rigorous.

    Need for Data

    A more rigorous treatment of the evaluation of the performance of the waste package material requires the determination of two important factors:

  •  
  • The realistic, extreme environments expected to come in contact with the C-22 metal surface; and
  •  
  • The critical temperature for crevice corrosion of C-22 in the presence of these environments.

    In the case of the TSPA-VA, estimates for both of these factors were based on expert elicitation. Although a large number of models that are reasonable and well conceived have been developed for evaluating various waste package processes, few of these have been validated and/or verified through the use of experimental data. In fact, experimental data are lacking throughout the treatment of the waste package and engineered barrier system (WP/EBS). These deficiencies will need to be resolved prior to the preparation of the TSPA-LA.

    Wet Waste Packages

    Because the spatial and temporal pattern of water seepage onto the waste packages is highly uncertain, it is prudent to design them with the anticipation that they will be wet during significant periods of time. The extent that waste packages are not wetted or are wetted to a lesser degree can be accepted as a source for increased confidence in the barrier properties and a hedge against uncertainty.

    The waste package surface will be dry when it is above a critical temperature (TWET) for the formation of a moisture film. This condition is expected to pertain after emplacement. It will persist until the waste packages cool below the temperature where moisture is stable. In the TSPA-VA, estimates for the critical temperature for moisture formation are based upon expert elicitation. Increased concentrations of ionic species in the water, capillary action from particulate matter on the surface, surface roughness, and the presence of crevices will increase the critical temperature. With respect to corrosion, the important properties of the moisture that forms on the waste package surfaces are chemical composition, the level of acidity (pH) and the oxidizing power (Eh). In the TSPA-VA, all of these important properties were based upon expert elicitation. Experimental data are required in this important area.

    When the waste package surface is wet, the important issue becomes whether the metal has sufficient corrosion resistance to remain passive (and therefore to corrode at extremely slow rates), or whether localized corrosion will occur. In the TSPA-VA, the treatment is based on the concept of a critical temperature (TCRIT) for localized corrosion. At temperatures in excess of this, localized corrosion can persist. Conversely, at lower temperatures, localized corrosion will not persist and the metal will remain passive. The difference between these two temperature limits (TWET and TCRIT) determines the critical temperature range in which localized corrosion can occur, i.e., at temperatures low enough for moisture to form and yet high enough for localized corrosion to occur. The time-temperature profile of a waste package will determine the time at which the critical temperature range is entered and the time duration that the package remains within this range. If the critical temperature for the metal is higher than the temperature for moisture formation, then no localized corrosion is expected to occur.

    The Panel concludes that this approach to localized corrosion is sound and consistent with the current knowledge of corrosion science and technology. For the TSPA-VA, the critical temperature for localized corrosion of C-22 was estimated to be 80°C and the critical temperature for moisture formation was estimated to be 100°C. An estimated probability function was used for estimating the onset of localized corrosion once the package reaches the critical temperature range. Estimates of corrosion rates as a function of temperature were used to determine the corrosion damage (penetration depth). While the approach is sound, once again it is important to recognize that all of the estimates of crucial parameters were based upon expert elicitation, not upon experimental data.

    Use of the TSPA to Evaluate Design Options

    Although the project staff has adequately and convincingly determined that waste package performance is critical, the TSPA is only one of several methods for the evaluation of base case design features and for the comparison of the base case with alternate designs. The Panel considers the TSPA-VA to be a useful tool for better understanding the performance and the effects of individual components on the expected repository performance; however, the TSPA treats a highly complex system and is a work in progress. The Panel concludes that the results of the TSPA-VA should be used cautiously, and that they should not be used as the primary criterion for design selection. This is particularly relevant to the evaluation of engineering components and structures. For example, the outcomes of the TSPA-VA clearly show that preventing water coming into contact with the waste packages is highly beneficial. However, the projected efficacy of an additional engineered water barrier, be it a drip shield, backfill barrier, or ceramic coating, is driven by the assumptions made within the TSPA. The Panel concludes that many of these assumptions do not have an adequate analytical and experimental justification. In a similar manner, the credibility of the conclusions is dependent upon the underlying engineering and science that supports the presumed performance of the enhanced engineering features.

    For purposes of evaluating alternative design features, the Panel recommends that a concentrated effort be undertaken to collect and collate the available experimental data germane to an analysis of waste package performance. These efforts should focus on the processes and components at the micro-level; they should not be aimed at the macro-level of the overall repository. The devil is in the details, and the details are lost at the overall repository response level. For example, the performance of the waste packages depends on the specifics of the methods used in their manufacture and fabrication. Alternative welding procedures are not amenable to evaluation within the TSPA; rather they need to be evaluated in the supporting documentation and engineering analyses. Development of an improved information base at the process level will increase the confidence in the TSPA results.

    D. Key Role of Infiltration and Seeps Analysis

    According to the TSPA-VA report, "limited water contact of waste packages" is one of the four basic attributes of the proposed repository. Infiltration and seepage into drifts are the main factors which control watercontact with the canisters.

    The base case in the TSPA-VA assumed that seepage into drifts is negligible during the thermal period, which was predicted to last for a few thousand years. Following the thermal period, seepage rates into the drifts were calculated by assuming a steady-state, in which the percolation rate near the drift was equated to the infiltration rate beneath the land surface, directly above the particular drift. The infiltration rate was estimated from the infiltration maps and the assumed climate. It was further assumed that the thermal period has no significant effect on the drift-scale hydrologic properties, which were taken in the analysis to be the same as present-day properties. The fraction of canisters that get wet and the corresponding seepage rates were determined from a random sampling of a response surface obtained from a detailed analysis of seepage in an individual drift. This probabilistic approach reflects the uncertainty in the capillary and flow properties of the fractures at the drift-scale.

    In the base case, the project staff conservatively ignored the effect of the matrix, which would be to delay the onset of seepage and to reduce the seepage rates. The assumption in the seeps analysis that all percolation flows through fractures led to a conservative estimate of seepage. The analysis predicted a percolation flux threshold (in the range of 2-3 mm/yr) such that for percolation rates below this threshold, seepage does not occur. Above this threshold, the seepage rate was found to be a non-linear increasing function of the percolation rate. However, the results depend sensitively on the permeability and capillary structure of the fracture continuum in the immediate neighborhood of the drifts. As a result, the base case projections are subject to the large uncertainties in the knowledge of the heterogeneity, spatial correlation and anisotropy of these properties. Figure 3-13 of the TSPA-VA, for example, shows that for a wide range of percolation flux values, the variance in the seepage fraction is almost equal to its mean value. It is not surprising therefore, that seepage into drifts was found to be one of the most sensitive parameters in the final dose resultsin the TSPA-VA report.

    Of interest is the predicted sensitivity of the seepage fraction on the percolation flux for values above the threshold and below 10 mm/yr, as shown in Figure 4-3 in Volume 3 of the TSPA-VA (
    CRWMS M&O, 1998d). On the basis of data presented in this figure, it follows that predictions of the fraction of wet canisters under present-day infiltration, which falls in this range, will be subject to relatively large uncertainties. This underscores the need for an accurate estimate of the magnitude of the percolation flux. Inthe long term, the effect of percolation rate, per se, on seepage cannot be estimated with a large degree of accuracy, given the uncertainties in long-term climate predictions.

    The Panel considers the analysis of seepage into drifts novel and informative. Given that it was only recently performed, however, it is understandable that the resulting analysis represents only a first-order approximation and that further improvements will be necessary before the accompanying estimates can be adopted with confidence. The following issues are of particular concern to the Panel.

    The analysis relies on a conventional but questionable van Genuchten formalism applied to a fracture continuum. This approach ignored the unstable nature of gravity-driven infiltration in real fractures, the possibility of hysteretic (and chaotic) behavior during episodic flow, as documented in recent experiments in related systems (Faybishenko et al., 1998), the discrete nature of the fracture network, and a detailed characterization of the capillary barrier condition at the drift surface. Thus, it is questionable that the representation assumed for purposes of developing the model actually reflects the true physics of seepage in a fractured system. Furthermore, the analysts ignored the possibility of drift collapse as a result of thermomechanical or seismic events, except for the analysis of waste package damage from rockfalls. Damage to the drifts would alter seepage rates and locations in two ways: As a result of the different boundary condition at the drift ceiling and as a result of the presence of rockfall on the canisters. These two combined effects will alter the predictions on seepage patterns andrates and the contact of waste packages with water.

    The hypothesis that negligible seepage occurs during the thermal period is based on certain assumptions of spatial homogeneity and symmetry. However, the possibility cannot be excluded that episodic seepage events will occur during the thermal period, in which canisters get wet as a result of instabilities at the overlying heat pipes or of focused flow, driven by heterogeneities in flow properties, canister heat output, and canister location (edge vs. center). Episodic seepage during the heating period was recently observed in the large block test (Hardin, 1998). If this were to occur in the proposed repository, it would lead to premature canister wetting in some locations. The absence of such a possibility needs to be convincingly and unambiguously demonstrated.

    Because of the steady-state assumptions made, seep locations and rates are estimated to be time-independent, under conditions of constant climate. Specifically, the project staff has assumed that water will come into contact with (drip onto) some patches some of the time, but water will not come into contact with other patches for periods as long as 1,000,000 years. Although a case was made in the TSPA-VA to support this assumption, the associated understanding of the features of the mountain, including the location and size of fractures, is not adequate. In addition, thermomechanical and thermochemical effects on the permeability and capillary structure of the fracture network will alter seepage patterns as a function of time, not only during the period of thethermal pulse, but also in a longer time horizon (recall that thermomechanical effects will last as long as the mountain is at a temperature higher than the ambient). This raises the possibility that seep locations and rates will shift with time. The consequences of this possibility should be investigated. Conversely, if precipitation caps develop (Hardin, 1998), they may act to reduce the amount of seepage in drifts over which such a cap forms over a long time period. This effect was not considered in the TSPA-VA.

    Finally, the Panel notes that waste packages at the edges of, or in isolated zones within, the repository may differ considerably from the standpoint of their exposure to water. Another uncertainty, in addition to the volume and distribution of water flowing, is whether water flows as droplets, films, or streams. The effects of rock particulate, debris, backfill, and corrosion products on water distribution and time-of-wetness were not well defined in the TSPA-VA.

    For these reasons, it is unclear to the Panel that the base case approach of the TSPA-VA correctly captures the behavior of seepage into drifts in the proposed repository and for the unprecedented periods of time considered in the TSPA-VA. Better characterization of the hydrologic properties near the drifts, improved modeling, consideration of coupled effects, and additional experimentation at the drift scale would add confidence to the approach taken. We note that efforts in these directions are currently under way.

    E. Potentially Non-Conservative Aspects of the Analysis

    Cladding

    The outcome of the TSPA-VA analysis depends to a considerable extent on the performance of the fuel cladding (
    CRWMS M&O, 1998c, Chapter 4, see page 4-12 and Figure 3-54), combined with an extended waste package lifetime. Despite the acknowledged corrosion resistance of Zircaloy cladding, this is a remarkably optimistic view of the long-term performance this cladding. Zircaloy cladding is typically in the range of 600 to 900 microns thick (less than a millimeter) and, during its life in a reactor, has experienced high temperatures and neutron fluxes. Important changes in mechanical properties can also occur due to thermally induced chemical reactions (oxidation or hydride formation). Another concern is embrittlement.

    To substantiate these comments, the Panel notes the following:

    1. The TSPA-VA (page 3-101) cited the work of Rothman (1984) in the discussion of oxidation rates for Zircaloy as part of the basis for the credit taken for extended cladding lifetime. This same paper notes,
    1. As noted in the TSPA-VA (page 3-102), Zircaloy may be susceptible to corrosion under certain chemical conditions. The authors of the TSPA-VA noted, "However, the chemistry within the waste package, hence the long term performance of Zircaloy cladding, is not well understood and has considerable uncertainty." Additionally, stress corrosion cracking is sensitive to chemical conditions (see extensive literature survey in Sidky, 1998). These chemical processes were explicitly not considered in the TSPA-VA (Siegmann, et al., 1996, page 16).
    2. The Panel notes, as it did in its third report, that additional mechanisms of failure remain to be investigated experimentally: (1) pitting and crevice corrosion; (2) hydride-induced embrittlement and cracking; and (3) "unzipping" of cladding due to secondary phase formation, particularly uranyl oxy-hydroxides which form immediately as alteration products of UO2 under moist, oxidizing conditions.
    3. Although the Panel strongly urges that the project team initiate and complete the necessary experimental programs, we note that time is limited. Quoting from Siegmann et al. (1996):
    To quote from the TSPA-VA, the "base case cladding model does not have a very wide uncertainty range, so the parameter does not show up in section 4.3 as a top rank-regression parameter" (TSPA-VA, page 5-25). In the Panel’s view, this is an instance in which the TSPA-VA analysts have failed to identify the critical importance of a parameter because of optimistic assumptions in the analysis both in terms of performance and the uncertainty in that performance.

    Soil Buildup in Biosphere Analysis

    As part of its biosphere analysis, the project staff has estimated the doses that would result from the use of contaminated groundwater for drinking and irrigation. Inhalation of dust containing radionuclides that would be found in irrigated soils has also been evaluated. Such an analysis has many components, including the estimation of the quantity and type of foods consumed, the factors that indicate how radionuclides in water are transferred to soil, from soil to plants, and, for forage crops, from plants to animals.

    The quantities of radionuclides that are assumed to transfer from soil to plant roots depend on their concentrations in the soil, and these concentrations can increase with ongoing irrigation with contaminated water. Eventually, a steady-state condition is reached in which the annual radionuclide additions to the soil through irrigation with groundwater will be equal to the annual losses. Soil losses can occur by wind or water erosion, and radionuclides can be removed from surface soils by transferring to crops that are subsequently harvested, or by weathering into the soil below the crop root zone.

    The analyses of the biosphere dose conversion factors were conducted using the GENII-S model (
    Leigh, 1993). This model permits the user to specify the length of time that irrigation water is deposited on the soil prior to the intake period for which a dose is estimated. In the TSPA-VA model runs, this time was assumed to be one year. Taking into account the fact that irrigation in some locations may continue for a period of hundreds or thousands of years, this one-year assumption could lead to estimates of radionuclide concentrations in the soil that will significantly underestimate the radionuclide uptake by root crops.

    The degree to which the failure to consider soil buildup leads to an underestimation of the dose rate depends on the specific radionuclides of concern. For technetium and iodine, the default assumption in GENII-S is that these radionuclides are rapidly washed through the soil column. This assumption appears to be inconsistent with measured iodine concentrations in surface soil near release sites for iodine (Kantelo et al, 1982, Straume et al., 1996, and Straume et al., 1997). Data from these studies indicate that iodine tends to remain in near-surface soils for extended periods. For radionuclides such as neptunium and plutonium, which are readily adsorbed by the soil, the degree by which the dose is underestimated could be significant.

    F. Potentially Conservative Aspects of the Analysis

    Transport through Penetrations in Waste Packages

    The Panel concluded that the TSPA-VA treatment of the movement of water into a damaged waste package and the transport of radionuclides from such a package were highly conservative. This is due in part to the preliminary state of the analysis of the likely evolution of penetrations through a damaged waste package. As the TSPA staff moves ahead, there is a need for an improved description of the progression of corrosion damage to waste packages, the size and shape of the assumed penetrations, the distribution of penetrations on an individual waste package, and the distribution of penetrations across the inventory of waste packages. There is also a need for a more realistic conceptual description and treatment of the evolution of corrosion damage. The results of these efforts will be coupled to and have a significant effect on several other process models and abstracted models. The important parameters are (1) the time sequence of waste form exposure to the repository environment; (2) the transport of water and other species into the waste packages; and (3) the release and transport of radionuclides from the waste packages.

    Should the corrosion-resistant metals fail by localized corrosion, the likely shapes of the penetrations will be small pits, tight cracks, or narrow channels. The size, shape, and distribution of penetrations in thick layers of corrosion-resistant metals were not analyzed as part of the TSPA-VA; the Panel recommends that this topic be examined in anticipation of the potential LA phase. Waste package penetrations from general corrosion are likely to be broad patches ranging from a few centimeters to tens of centimeters in diameter. The spatial distribution of penetrations at the top or bottom and along the sides of the waste packages is poorly understood and defined. The issue of the likelihood of penetrations remaining open or becoming blocked by corrosion products or deposits, has been introduced, but the understanding is not well developed. Once a waste package has been penetrated, water and air from outside will have access to the waste package internals and the spent fuel. The subsequent processes that ensue among internal metal structures, cladding, and spent fuel were not addressed in any depth within the TSPA-VA.

    Regarding the base case conditions, the TSPA-VA staff assumed that the spent fuel and cladding would be instantly covered by a water film at the time a waste package was penetrated. Transport of moisture and air into the packages and the transport of products from the packages through such penetrations were judged not to provide any significant retardation to radionuclide releases. The Panel does not accept this view; we believe that it would have been more realistic to have assumed that the resulting penetrations will likely retard radionuclide releases from the waste packages. Although the task will be difficult, the Panel recommends that steps be taken to develop better methods for analyzing the movement of radionuclides into and from the waste packages.

    Retention of Radionuclides in Alteration Products of Spent Fuel

    As discussed in the first interim report of the Panel (
    Whipple et al., 1997), the alteration and corrosion rate of UO2 is relatively rapid under moist, oxidizing conditions. As modeled by the TSPA-VA (CRWMS M&O, 1998C, Chapter 6, Section 4.6.2.2.2), the UO2 in the spent fuel will be completely converted to secondary uranyl phases within 100 to 1,000 years after waste package and fuel cladding failure and exposure to water. In the present analysis, the corrosion of UO2 results in releases of radionuclides which are then available for transport in water, and only the solubility limits for individual radionuclides place an upper limit on the radionuclide concentrations in water. However, under such conditions, an assemblage of secondary uranyl oxyhydroxides, silicates, and carbonates will form depending on groundwater compositions. It is expected (Burns et al., 1997) and has been shown experimentally (Buck et al., 1998) that certain radionuclides, such as 237Np, will be incorporated into the structures of these secondary phases. Thus, the formation of these secondary alteration phases may remove certain radionuclides from solution by co-precipitation or sorption and retard their release from the near-field environment.

    At present, the TSPA-VA does not take credit for this type of radionuclide retardation; in this sense, the analysis is conservative. For some radionuclides (237Np and 79Se) some degree of co-precipitation is expected, and for other radionuclides (99Tc and 129I) this type of process is unlikely. For those radionuclides for which this is a likely retardation process, a well-defined experimental program (discussed in section IV.G of this report) may provide a substantive basis for increased retardation of key radionuclides, e.g., 237Np. The inclusion of this type of analysis in the TSPA would, however, increase the general level of complexity of the analysis of spent fuel corrosion and create new data needs which will require further experimental work.

    Potential Sorption of Technetium and Iodine

    The TSPA-VA analysis of the performance during the first 10,000 years after repository closure indicates that the calculated doses are due to 99Tc and 129I. Ultimately, the doses from neptunium and plutonium are estimated to be larger, but these radionuclides are not expected to reach the accessible environment in significant concentrations until much later, because their flow through the unsaturated and saturated zones is assumed to be retarded by chemical sorption. No retardation credit is taken for technetium and iodine (or for three other radionuclides), based on the lack of observed sorption in batch measurements of Kd values. The decision is described in the TSPA-VA as conservative.

    However, the data cited above regarding the potential accumulation of iodine in surface soils (Kantelo et al., 1982,
    Straume et al., 1996, Straume et al., 1997) indicate that iodine deposited on the ground from the Chernobyl accident is retarded in the upper soil layer to about the same extent as is plutonium and cesium. Given these data indicating that iodine does not move readily through soil, it could also be the case that iodine could be retarded in transport through the unsaturated and saturated zones. The Panel has not conducted a literature review on this issue.It seems likely that measurements taken of areas near the Chernobyl site should also provide relevant data on the retention or lack of retention of technetium in soil. Regarding the retardation of iodine, additional data sets are likely to be available from environmental measurements taken at the Hanford site, where radioactive iodine was released during spent fuel reprocessing. Although it appears that some fraction of the deposited radionuclides may be transported to groundwater, as was the case with cesium at the Hanford tank farm, field data suggest that radioiodine does not move unretarded through soil.

    Due to the difference between surface soils and the properties of the rock in the unsaturated zone and saturated zone flow paths, the retardation observed at the surface may not occur during underground transport. It is the Panel’s view that this question should be explored. It is also the Panel’s view that the TSPA-VA analysts over-emphasized laboratory Kd measurements and did not appropriately consider opportunities to observe the mobility of radionuclides in the environment.

    G. Potentially Important but Omitted Processes

    The Panel has identified several processes that can have major detrimental effects on repository performance, but which are currently not extensively analyzed. These processes require further analysis and an assessment of their importance.

    Expansion of Steel Corrosion Products

    For waste package performance, the detrimental effect of the expansion of steel corrosion products on the inner barrier and canister internals has not been addressed. When steel corrodes, the iron oxide that forms occupies a volume two to three times larger than the corroded metal. When these corrosion products form in tight spaces, e.g. the crevice between the outer steel barrier and the inner C-22 barrier of the base case design, high stresses are produced that will lead to plastic deformation of the remaining metal structures. Such "pack-out" failures have been observed for structural steel beams where atmospheric corrosion caused the expansion of steel corrosion products. "Denting" failures of nickel alloy tubes in steam generators in nuclear power plants were caused by iron oxide growth due to corrosion of the steel support plates. This process represents a serious threat to the integrity of the inner C-22 barrier of the waste package, once the outer steel barrier has been penetrated and water gains access to the crevices between the barriers.

    Hydrogen Embrittlement of Zirconium Cladding

    Damage due to hydrogen is a major threat to the integrity of zirconium cladding. When cladding is embrittled by hydrogen, it loses its mechanical strength and ductility and fails by through-wall cracks. One possible source of hydrogen in the proposed repository is the corrosion of dissimilar metals in contact with zirconium. Hydrogen formed by the corrosion of steels and stainless steels has damaged and led to failures of zirconium components in industrial applications. If the internal barrier of the waste package is penetrated, water can contact the package internals. The resulting corrosion and hydrogen production represent a significant threat to the integrity of the cladding. This degradation process was not addressed adequately in the TSPA-VA. As a result, the extent of the credit taken for cladding in the analysis is questioned.

    Stress Corrosion Cracking

    Stress corrosion cracking of the C-22 barrier is a realistic threat to waste package performance. The possibility of such a threat was not adequately addressed in the TSPA-VA. A proper evaluation will require more experimental data in realistic, repository environments. The Panel supports the recommendation to use double, U-bend specimens in stress corrosion cracking tests for realistic simulation of repository conditions.

    H. Data Needs

    The Panel is recognizes that substantial amounts of field and experimental data were developed in support of the TSPA-VA; however, the future success of the project depends even more critically on the acquisition of additional data, particularly as more sophisticated models are incorporated into the analysis.

    Additional data needs are of two types:

  • Fundamental data that are essential to the development and implementation of the models, and
  • Data sets designed to challenge conceptual models and test the success of coupled models used in the TSPA-VA.
  • Fundamental Data

    A substantial part of the knowledge base, presently rests on expert elicitations (Table 3-2, page 3-2 of the TSPA-VA) for data on: flow in the unsaturated zone, the near-field environment, waste package degradation, waste form degradation, radionuclide mobilization and flow/ transport in the saturated zone. Additional site characterization in the unsaturated and saturated zones, as well as experimental programs in waste package and waste form degradation are required.

    Solubility limit distributions for the key radionuclides (Table 3-15, page 3-99 of the TSPA-VA) have only a limited experimental basis. Indeed, most are based on expert elicitations or a reanalysis of previous experiments. If there is any area amenable to experimental study, it should be the determination of concentration limits in relevant solution compositions. Project scientists are well aware of the need for data, particularly for actinides (Near-Field/Altered-Zone Models Report, page 5-15). The NEA data base for uranium (
    Grenthe et al., 1992) has thermodynamic data for fewer than five uranyl oxyhydroxides and silicates that may be important during the corrosion of the UO2 in spent fuel. In 1993, J. Fuger reviewed the database for actinides, and concluded that not only are there inconsistencies in the database for uranium, but that for the transuranium elements many fundamental data are lacking. The present situation is not much improved.

    An experimental program should be developed to advance the spent fuel corrosion model beyond its present empirical representation by a response surface. "Currently, detailed knowledge is not available for the atomic (mechanistic) steps and the sequence of chemical/electrochemical reaction steps to describe the dissolution process over the range of spent fuel inventory, potential water chemistries, and temperatures" (CRWMS M&O, 1998c, Chapter 6, page 6-60). Since over ninety percent of the radioactive waste intended to be disposed in the repository is spent fuel, a considerable knowledge of spent fuel corrosion is likely to be required.

    There are similar lacks of data for Zircaloy cladding corrosion, secondary phase formation, colloid formation and transport, Kd, and saturated zone characterisitics.

    Testing Models

    The project staff should provide and regulators should require, where possible, demonstrations that the TSPA "works". This can be accomplished by designing experiments and field tests that are driven by the TSPA-VA analysis and that challenge the conceptual models used in the analysis. This is a standard approach in any scientific and engineering study, particularly one as complex as the TSPA.

    An example of such a laboratory study was recently reported by
    Werme and Spahiu, (1998). These authors illustrate the difficulty of modeling actinide concentrations in well controlled experiments. They conclude, "There is a large body of data on the solubilities of pure actinide phases; however, it appears that the information available is insufficient to explain the experimental results." These conclusions speak directly to the uncertainty in the modeled results in the TSPA-VA. Such experiments cannot be replaced by sensitivity analyses, because the sensitivity analyses used in the TSPA-VA do not directly address the uncertainties associated with the experimental data base or the selection of a conceptual model.

    Concerning field tests, many of the questions raised about the effects of the thermal pulse can be (at least partly) answered from the results of the , which is one part of the in situ thermal testing program (DOE, 1995) for Yucca Mountain. This test is an important experiment that will provide the first large scale underground investigation of the thermomechanical and thermochemical processes which control the thermal behavior of the rock mass surrounding the proposed repository. The Panel believes that the will constitute a major step forward in the process of understanding the complex behavior of the proposed repository under the impact of the thermal field. Underground testing in fractured tuff on this scale has never before been performed. It is anticipated that the results will provide data that will lead to a reduction of uncertainties.

    I. Insights from the TSPA-VA

    The Panel observes that the analysis indicates that the performance of the proposed repository depends primarily on the functions and efficiencies of certain major elements of the system. The TSPA-VA (Section 2.2.1) describes the four key attributes that contribute to the safety of the repository as:

  • Limited water contacting waste packages;
  • Long waste package lifetime;
  • Slow release of radionuclides from waste package; and
  • Reduction in the concentration of radionuclides during transport from the waste package.
  • These four attributesof the repository system control the radionuclide concentrations that may ultimately reach the accessible environment. These system elements, in turn, can be grouped into two spatial and functional groups:

  • Near-field: delay in the release and mobilization of radionuclides; and
  • Far-field: transport of radionuclides, with associated delay and dilution.
  • In the TSPA-VA analyses of the performance out to the time of peak doses, typically at several hundred thousand years, almost all of the protection was found to be provided by the engineered waste package, the cladding, and the dilution that occurs in the saturated zone. From this long-term perspective, the early thermal period during which liquid water does not contact the waste packages and the required times for transport of the radionuclides through the unsaturated or saturated zones were found not to be important to overall performance. This has led to criticism that the project is relying principally on engineered features for protection, and that the natural features of the site contribute little to safety.

    From the perspective of the time frames associated with peak doses, it is the case that the major share of protection comes from the engineered barriers, along with whatever dilution occurs in the saturated zone. But for an initial 10,000 year assessment, the relative contributions from the initial thermal period and from the travel times through natural barriers are more significant, in a relative sense, in delaying the arrival of radionuclides at the accessible environment. The thermal period is projected to prevent the onset of liquid water reaching the waste packages for around 2,000 years, and the travel times for transport in the unsaturated zone and saturated zone are estimated to be around 1,000 years for nonretarding radionuclides, and longer for the others.

    In the Panel’s view, the confidence that the public can have in the TSPA results will, to a large degree, depend on how the analyses of the major attributes of the repository system are conducted and presented. These four attributes can be presented in a framework that includes the supporting models and their underlying physical and chemical principles, conformance with available laboratory and field data, experiences with similar models in comparable systems, and sensitivity analyses based on alternative plausible models. If such a framework can be effectively developed, the strategy of "defense-in-depth" will have been applied successfully to the design and analysis of the proposed repository.

    III. THE TSPA-VA METHODOLOGY

    In this section, the Panel discusses the overall framework of the TSPA-VA, including the uncertainties inherent in modeling the probable behavior of a repository at Yucca Mountain. A distinction is drawn between those contributors to uncertainty in performance that can be improved through the refinement of models and the collection and application of data, versus those uncertainties that can be more effectively managed through the use of bounding analyses or design changes. The section concludes with the Panel’s recommendations.

    A. Expectations for the TSPA-VA

    Differing Objectives for the TSPA-VA and the TSPA-LA

    As part of its review, the Panel considered the relation between the TSPA-VA and the TSPA-LA. The Panel’s Peer Review Plan addresses this issue: "The Peer Review Panel members will conduct a phased review … to observe the development and completion of the TSPA-VA. The comments, concerns, conclusions, and recommendations of the final Peer Review Report will be provided to the M&O to support the development and conduct of the License Application TSPA (TSPA-LA)."

    The Congressional mandate that a viability assessment be conducted was stipulated in the Energy and Water Development Appropriations Act of 1997. Specifically, the Congress directed DOE, in the course of this assessment, to analyze "the probable behavior of the reference design for the engineered repository components in the expected natural conditions at the Yucca Mountain site." If the project moves on to the license application phase, the goal will be different. The USNRC, for example, in 10 CFR Part 60 and in its draft regulations in 10 CFR Part 63 for the proposed Yucca Mountain facility, has stipulated that the goal is to provide "reasonable assurance" that a facility of this type will comply with its regulations. While it is not possible for the Panel to know at this time how this requirement will be interpreted, it is our opinion that the Commission may likely require a higher standard of proof than is associated with the Congressional concept of "probable behavior" at the VA stage. That is to say, the USNRC may require that the results be developed and defended in a manner so as to demonstrate a higher level of confidence. At the same time, however, it is important to keep in mind that neither the USEPA nor the USNRC expects DOE to be able to perform an assessment that provides proof positive that the proposed repository will perform in a given manner.

    In addition to the degree to which the TSPA-VA and TSPA-LA may require different goals for the analyses, the Panel noted in its third report that the draft TSPA-VA analysis was based on models that may be better suited for estimating peak doses at some later time than for assessing performance within the first 10,000 years. At the LA stage it will be necessary to estimate the performance during the time frame specified by the regulations. Based on draft guidance from the USNRC and on recommendations to the M&O from the DOE, it is anticipated that this time frame may be 10,000 years. Given these fundamental differences in the objectives for the degrees of certainty and the time frames of the TSPA-VA and the TSPA-LA, it follows that the two analyses may differ significantly.

    In addition to meeting the needs of the VA and LA, a version of the analysis will be required in support of the preparation of the associated Environmental Impact Statement (EIS). Performance assessments will also be necessary for evaluating various alternative designs. The degree to which the EIS and the design review analyses will address "probable behavior," versus a determination of "reasonable assurance" of performance, is a consideration for the TSPA staff. It is beyond the scope of this peer review.

    The main point of the Panel in noting the differing objectives for the TSPA-VA and the LA is to make it clear that there are several options available to the TSPA staff to address deficiencies in the VA. Certain of these deficiencies are noted in
    Section IV of this report. Several approaches that the TSPA staff may want to consider as the analysis moves forward to the possible LA phase are addressed in Section III.D.

    Inherent Uncertainties in the Assessment

    In developing its standards for the geologic disposal of radioactive wastes, the USEPA recognized that large uncertainties would be associated with performance assessments over long time scales. For this reason, in 40 CFR Part 191.13(b) of its standards for spent nuclear fuel and high-level and transuranic radioactive wastes (which now apply to the Waste Isolation Pilot Plant, but not to the proposed repository at Yucca Mountain), the Agency included the following statement regarding the degree of confidence that one must have that the containment requirements are met:

    Performance assessments need not provide complete assurance that the requirements of Sec. 191.13(a) will be met. Because of the long time period involved and the nature of the events and processes of interest, there will inevitably be substantial uncertainties in projecting disposal system performance. Proof of the future performance of a disposal system is not to be had in the ordinary sense of the word in situations that deal with much shorter time frames. Instead, what is required is a reasonable expectation, on the basis of the record before the implementing agency, that compliance with Sec. 191.13 (a) will be achieved.

    As the USEPA noted, factors such as the complexity of the total system, the variability in the natural setting, and the need to estimate the performance of the repository over extended periods of time result in large inherent uncertainties. These uncertainties, in turn, have an impact on the overall credibility of the TSPA-VA. Many uncertainties are evident in the characterization of the site, in the development of conceptual models, in the selection of values for various input parameters, and in the determination of boundary conditions. Other uncertainties developed as an outgrowth of the presence of the novel and not yet field-tested aspects of the analyses of the many processes that may occur in a geological setting. The TSPA-VA staff acknowledged these uncertainties and devoted a large fraction of their report to a discussion of the associated probability density functions and sensitivity analyses.

    For these and other reasons, it is important that the TSPA staff continue to recognize the limitations that these uncertainties place on their analyses, especially as they move into the LA phase. The Panel has been mindful of the need to acknowledge, in the course of our review, what is feasible and what is not.

    B. Methodology

    Overall Framework of the Analysis

    The Panel concluded that the overall performance assessment framework and approaches used in the TSPA-VA are sound and follow accepted methods for risk analysis. However, some of the technical complexities associated with the analysis are unusual if not unique.

    Conventional Aspects

    The project staff has followed conventional methods in the TSPA-VA analysis in that it:

    1. Begins with the characterization of the inventory of radioactive materials to be placed in the repository.
    2. Characterizes the site.
    3. Conducts analyses of the processes and events that could cause the engineered barriers to fail and radionuclides to be released. Such barriers include the waste form, the cladding, and the waste package. The outcomes of these analyses provide estimates of the rate of release of radionuclides through each of these barriers and form the basis for subsequent calculations of radionuclide transport and exposures.
    4. Accounts for radioactive decay, both during the time prior to the release of radionuclides from the waste packages and during their transport through the environment.
    5. Models the transport of the released radionuclides through the unsaturated zone, with consideration given to mechanisms that would lead to retardation of certain radionuclides.
    6. Models the transport of the radioactive material within the saturated zone. This analysis includes retardation and dilution, and leads to a projection of the concentration of key radionuclides in groundwater at offsite locations where people could be exposed.
    7. Considers a full range of pathways in the exposure assessment, including the direct consumption of groundwater, ingestion of foods contaminated through the use of groundwater for irrigation, inhalation of contaminated dust, and direct radiation from the ground surface.
    8. Evaluates several exposure scenarios, in which the principal differences are with the assumed fractions of locally produced foods consumed by members of the exposed group.
    Novel Features

    The TSPA-VA includes several features and analyses that are novel. These include the:

    1. Unprecedented time periods over which the analyses were performed. The TSPA-VA extends from up to 10,000 to up to 1,000,000 years into the future, with unknown changes occurring over those times (e.g., climate, locations of people and their sources of food and water). These time periods are also long compared to those available for testing the corrosion rates of materials, thus making the extrapolation of materials performance uncertain.
    2. Length (20 kilometers) of the subsurface pathway over which transport was modeled.
    3. Heterogeneity of the site and of the paths through which radionuclides could be transported and ultimately come into contact with people. The movement of radionuclides occurs as a result both of water flow through fractures and its interactions with the rock matrix. The site cannot be characterized at a sufficiently detailed scale to define precisely the flow paths or material interactions.
    4. Complexity of the processes that interact to affect hydrologic and chemical conditions within the proposed repository and adjoining areas. The coupled interactions among heat, moisture, and the chemical environment, and the responses of the proposed repository to the associated mechanical stresses are complicated and cannot be modeled with precision. Material performance will depend on the thermal, chemical, and hydrological environments as they evolve over time, yet material performance can also alter these conditions, e.g., corrosion byproducts from steel may affect water flow, colloid formation, and water chemistry. The site conditions measured during the site characterization phase of the project may be significantly altered by the heat output of the waste.
    5. Sensitivity of the performance of the proposed repository to features that cannot be precisely assessed. For example, if the infiltration rate of water through the mountain is higher than anticipated (or if climate changes lead to wetter conditions), there will be an increase in the portion of unsaturated zone flow through fast pathways, in the number of waste packages in contact with seeps, and in the amount of water contacting waste packages. Such enhanced flow will result in increased releases of solubility-limited radionuclides. At the same time, the lifetimes for the waste packages and the time required for radionuclides released from the waste packages to reach the saturated zone would both decrease.
    These novel features limit the confidence that can be placed in the outputs of the TSPA-VA.

    Use of Model Abstractions

    Because of the inherent complexity of the proposed facility and site and the number of simulations that need to be analyzed, the detailed process-level models are abstracted, i.e., replaced with simplified models or with lookup tables generated by the underlying process models. The purpose of this approach is to reduce the computational requirements associated with the analyses. The Panel agrees that this approach is sound in principle. For example, the use of model abstractions avoids the need to conductnumerous runs of the complex biosphere analyses, incorporating a host of individual assumptions regarding the multiple possible exposure pathways and food consumption rates for each model realization. Instead, the results of a series of biosphere analyses can be used to generate distributions for the doses that would result from a specified groundwater concentration of each radionuclide under consideration. These distributions, referred to as biosphere dose conversion factors, are used in the integrated TSPA calculations to estimate doses. This approach is computationally efficient and technically appropriate. In addition, the use of abstractions may prove helpful in future licensing reviews by making it easier to separate the evaluation of the component models from the consideration of how the results of these models are integrated into the overall analysis.

    For several of the mel components, however, it is not easy to review or evaluate the degree to which the model abstractions are equivalent to the underlying more complex analyses. For example, the TSPA-VA staff has stated that the modeling of the behavior of the near-field geochemical environment is based on a mixture of "abstracted models with some process level components" (
    CRWMS M&O, 1998c, Chapter 4, page 4-38). The document indicates that in at least one area (composition of gas in or around drifts), there is no process-based model. The Panel is unable, based on these comments, to determine what the basis is for the abstracted models. This matter needs to be clarified.

    To help resolve these questions, the Panel recommends that the staff pay careful attention to its own definition of the model abstraction process. In each case, the abstraction should be a simplification of a more fundamental process-based model, and it should provide results consistent with the process-based models over the same range of parameter and input values as can be treated by the more complex process-based model.

    Uncertainty and Sensitivity Analyses

    The TSPA-VA summarizes the results of extensive sensitivity analyses, conducted for different time periods (typically 10,000, 100,000, and 1,000,000 years), and provides estimates of the effects on isolated subsystems of changes in various performance parameters or site conditions. For example, estimates of the distribution of travel times for radionuclides through the unsaturated zone were based on three different climate regimes. This approach is informative and can provide helpful insights into the likely performance of the proposed repository system.

    The degree to which such analyses reliably indicate which aspects of the system are more or less important is limited by the inconsistent degree of realism versus conservatism in the various analyses incorporated into the TSPA-VA. Because a mixture of both conservative bounding and more-or-less realistic analyses were used, the interpretation of the outcome of the sensitivity analyses is not straightforward. The Panel knows of no methodologically sound approach to quantify sensitivities for a given analysis that uses such an approach. This stems, in part, from the fact that the degree to which the actual performance of some aspect of the repository system differs from an estimate of that performance based on a bounding analysis is not known (if the actual performance were known, a bounding analysis would not be needed).

    The Panel’s point in noting that the TSPA-VA will inevitably be an uneven mixture of bounding analyses and more realistic assessments is to caution against overconfidence in the validity of the results of the sensitivity analyses. Because the TSPA-VA incorporated many assumptions of varying validity, the results of these analyses need to be interpreted with judgment and their conditional status recognized. Even with these limitations, however, the sensitivity analyses can be valuable if it can be shown that certain aspects of the repository system have little effect on the performance of the repository. However, the Panel notes that the TSPA-VA staff did not, at this stage, seek to use the sensitivity analyses to demonstrate that certain aspects and/or issues are unimportant and therefore need not be further considered.. These judgments and/or decisions may be more appropriately made during the possible TSPA-LA phase.

    In seeking to understand the performance of the proposed repository, the TSPA-VA staff has, in general, modeled the various issues separately. In some cases, they have performed limited coupled-process analyses. They have also conducted a variety of sensitivity studies and bounding-type evaluations, typically to examine how the estimated overall performance of the proposed repository depends on a single parameter or aspect, while holding constant the other interacting parameters or aspects. One example was the evaluation of thermomechanical effects in the unsaturated zone, which were not incorporated in the base case analysis.

    While it may be possible to analyze some components and systems in a realistic manner, the analysis of others may of necessity, because of data or modeling limitations, have to be based on bounding and therefore conservative assumptions.

    This can lead to several problems:

    1. Sensitivity analyses may indicate incorrectly that a particular feature of the site or design is unimportant to performance. This could be the case, for example, where the given feature has been analyzed by an overly conservative bounding analysis. When a parameter point value is used, a sensitivity analysis typically cannot identify whether that parameter is important or unimportant to performance.
    2. Similarly, a sensitivity analysis may be performed over too narrow a range of values to reflect the actual sensitivities. It is important that a full range and distribution of the uncertainties in the values for each key input parameter be considered, and appropriately justified. An analysis that may be unrealistically optimistic can mask the actual sensitivities in the performance of that system and/or component. For example, the cladding analysis produces an estimate of the cumulative cladding failures over time. The lower bound of the failure rate curve (that is, the "worst case") indicates that the cladding is so robust that repository performance is comparatively insensitive to this aspect of performance.
    3. It may be difficult to assess the relative importance of components and systems.
    In its first interim report, the Panel discussed the importance of viewing sensitivity analyses from multiple perspectives and over differing periods of time. At that time, the Panel noted that, while an aspect of performance may not seem important when viewed from an overall perspective, it may be important on the basis of considerations of subsystem performance measures. The TSPA team has been responsive to this recommendation. In its third interim report, the Panel recommended "….that the sensitivity analysis results not be used to identify key analytical uncertainties as the program progresses toward the TSPA-LA. Instead, the Panel recommends that the TSPA sensitivity analyses be viewed as an input to the collective judgment of the TSPA and other project staff. In addition, where sensitivity analyses produce results that are inconsistent with the intuitive judgments of the project staff or advisors, the underlying models and parameters should be examined to ensure that uncertainties in performance are appropriately represented." We continue to endorse this recommendation.

    Use of Expert Elicitations

    Several of the more important components of the TSPA-VA were based on information derived through expert elicitations. The results of these elicitations have been used in analyses of the probabilistic volcanic hazard, probabilistic seismic hazard analysis, waste-package degradation, and radionuclide mobilization, saturated-zone-flow issues, radionuclide solubility limits, and near field/altered zone coupled effects. The expert elicitations followed a defined protocol and were extensively documented.

    The value of a properly executed expert elicitation under these circumstances is that it provided the TSPA-VA staff with the full, and fully documented, range of interpretations of the data or models currently considered valid or respectable. Such a process can also, if properly applied, direct the thinking of the experts toward the specific questions being faced, including where the data or models need to be applied and how. Through the process of being forced to interact on the subjects at hand, the experts can often resolve conflicting interpretations and provide a more unified view than the TSPA-VA staff could reach on its own.

    Nonetheless, the Panel is concerned that expert elicitation could have been misused by the TSPA-VA staff through its application as a comparatively rapid and inexpensive way to synthetically generate "data" as inputs to the TSPA-VA, in place of actual laboratory or field measurements. Unfortunately, in several instances, noted in
    Section IV, this has occurred. However, there were several positive results from the use of expert panels that clearly make evident their value to the project. One major contribution, in a general sense, was in challenging the TSPA-VA staff’s basic approach to, and conceptual modeling of, a particular issue. For example, the streamtube analysis used to assess the transport of radionuclides in the saturated zone was developed in response to criticisms of the previous model by the expert panel. A second benefit was that often the experts were able to identify relevant studies or data with which the TSPA-VA staff members were not familiar.

    On balance, the contributions from the expert panels to the TSPA-VA were positive. The panels improved the analysis of many component parts of the TSPA-VA. Although there are some issues for which expert opinion (rather than laboratory data) served as the primary basis for estimating the projected performance of the proposed repository, the process through which the experts worked generally provided adequate feedback to enable them to confirm whether their opinions were appropriately used in the assessment. This process should not be seen by the project staff as completed; expert opinions should not be used when the required data can be obtained from experiments or field studies, in a reasonable amount of time.

    Use of Expected Values

    The results of the TSPA-VA analysis are presented in terms of the expected values of the doses, calculated as a function of time, based on multiple runs (or realizations) of the system of models. In these model runs, the ranges of parameter values were sampled according to their estimated underlying probability distributions. These calculations are probabilistic in the sense that each run involved estimation of the fraction of waste packages that experience seeps; the fraction that have failed through corrosion at any given time was estimated probabilistically. In addition to the expected value, curves reflecting the fifth and ninety-fifth percentile distributions were provided in some cases. This approach follows the normal and accepted methodology for such analyses.

    However, the results need to be understood in the sense that the expected (or mean) value of a calculation is not necessarily informative about the nature of the underlying distribution. For example, a scenario that leads to a dose rate of 1 mrem per year has the same mean value as an event with a 1% probability per year of occurring, and which results in a dose rate of 100 mrem per year if the event occurs. In this latter case, the expected value for the scenario is 1 mrem per year, but 1 mrem per year need not be a likely or even possible result of the event. While these two cases have equal expected values, they may not be viewed as equivalent risks.

    The situation where the expected annual dose does not describe the results of specific scenarios occurs in the TSPA-VA at "early" time periods. At early times, a single juvenile waste package is assumed to occur at 1,000 years, presumably due to poor fabrication. Isolated waste package failures are also estimated to occur due to corrosion within the time period between several thousand years to ten thousand years after repository closure. As described in
    Section IV, the analysis assumes that stainless steel cladding (slightly over 1% of the total spent fuel) provides no barrier to the release of radionuclides. In addition, a smaller amount of spent fuel is assumed to have cladding that is not completely intact at the time of disposal. In total, 1.25% of the spent fuel is assumed to lack a cladding barrier when the waste package fails.

    It is unlikely that the stainless steel clad fuel will be uniformly distributed throughout the waste packages. For this reason, the juvenile failure of one waste package will most likely involve no immediate release of radionuclides, assuming that the cladding is intact and acts as an effective barrier. But for 1.25% of the time, the full contents of the waste package are available for release and transport after waste-package failure. The analysis treats the source term for the juvenile failure as equal to 1.25% of the inventory of the average package of spent fuel, and the annual dose due to a juvenile failure is calculated on this basis. It may not be clear to the TSPA-VA reader that the estimated annual dose is an expected value resulting from 98.75% of scenarios that produce no radiation exposure and 1.25% that produce exposures that are higher than the mean (higher by a factor of 80, in this example).

    C. Complexities of the System and of Its Components

    Modeling of Coupled Processes

    One critical aspect of the TSPA-VA is the degree to which an appropriate conceptual model and approach were applied to three key groups of phenomena that involve complex coupled processes or data-intensive analyses of a large heterogeneous site, as exemplified by Yucca Mountain. These are:

  •  
  • Phenomena in the unsaturated zone above and near the repository during the first several thousand years that govern whether the various waste packages will be wetted, will be in a humid environment, or will remain essentially dry;
  •  
  • Waste-degradation phenomena inside the waste packages, after canister degradation has begun, that govern how much, how quickly, and in what forms the radionuclides may move from outside the waste packages into the local environment; and
  •  
  • The behavior of the radionuclides in the unsaturated and saturated zone environments.

    It is difficult to model adequately each of these sets of phenomena over the anticipated 10,000-year regulatory period. For the first two groups, this difficulty is due to the fact that each is influenced by complex coupled interactions. Phenomena in the unsaturated zone involve a combination of thermal, hydrological, mechanical, and chemical processes and effects; in the waste-degradation process, they involve a combination of physical, chemical, and mechanical processes and effects. For the third group of phenomena, the difficulty stems from the heterogeneity of the site and the large distances over which the transport of radionuclides is to be assessed. It is not feasible to know the structure of the flow paths in sufficient depth to model these phenomena in detail.

    In neither of the first two cases were sufficiently detailed coupled models developed in the TSPA-VA to permit an integrated analysis to be performed. The development of a fully-coupled model based on first principles for the reference design may be beyond current analytical capabilities. Making such an analysis especially difficult is the lack of an adequate theoretical basis for such models, compounded by the lack of sufficient site-specific data to support them.

    While the Panel does not think that a fully-coupled, theoretically-defensible, first-principles analysis of coupled processes is possible, it believes that a considerable amount of data exists that could have been incorporated into the modeling approaches used in the TSPA-VA. The ongoing thermal testing program is designed to investigate precisely these issues, namely, to provide data that will lead to a better definition of the effects of coupled processes. The Panel addresses these issues in
    Section IV.

    Limitations of the Component Models

    The Panel’s detailed comments on the TSPA-VA component models are presented in
    Section IV. Although many of the processes addressed in the TSPA-VA are extremely difficult to analyze and would perhaps be better addressed through bounding analyses, the Panel has reviewed the application and analyses produced by these component models against the Congressional charge to assess the probable behavior of the repository system. As discussed in the Panel’s Second Interim Report (Whipple et al., 1997), the capability to analyze such complex systems over long time periods is uneven at best. In general, significant errors in performance assessment may occur due to the selection of the wrong deterministic model for specific phenomena, an incorrect analytical solution for the model, or an incomplete description of the system to be modeled.

    In the case of the analyses of the behavior of some components, the refinement of the relevant models and the acquisition of additional data would permit significant improvements to be made. In other instances, the problem being analyzed may be essentially intractable given current analytical capabilities, or intractable within the time constraints under which the TSPA staff is operating. The distinction between those analyses that may be intractable and those that can be improved is important, because the approaches to deal with the two situations, as the project moves into the anticipated TSPA-LA phase, are different. In Section III.D, approaches for dealing with the analyses that appear to be intractable are discussed. Improvements that can be made through updating the component models and the acquisition or use of additional data are discussed in Section IV.

    D. Managing Complexities and Component Model Limitations

    On the basis of its review, the Panel has concluded that there are two types of processes that should be analyzed as part of the possible upcoming TSPA-LA, particularly in terms of meeting the anticipated "reasonable assurance" requirements of the USNRC. These are (1) those for which analytical models are available, and (2) those that may be essentially intractable given current analytical capabilities, or intractable within the time constraints under which the TSPA staff is operating. Although both of these types of processes are complex and extremely difficult to analyze, each has distinct characteristics from the standpoint of the approaches that can be used to analyze them. These approaches include:

  • updating the component models;
  • expanding the quality and quantity of data available as input into these analyses;
  • use of bounding analyses, that is, intentionally conservative assumptions, parameters, and models; and
  • design changes.
  • Also to be considered is the incorporation of the "defense-in-depth" concept into the design of the overall repository system. Effective use of this concept, in concert with the approaches enumerated above, can enhance the confidence of the designers, the analysts, and the regulators that there is reasonable assurance that the proposed repository design will meet the regulatory requirements.

    As would be expected, the four identified approaches are closely related and, in fact, they are intertwined. Furthermore, the applicability of a given approach to a specific type of process will depend on the nature of the process. In the case of processes for which analytical models are available, significant improvements can be made through updating the component models and the acquisition and use of additional data. In the case of processes that may be essentially intractable, the only available option may be to treat them through the use of bounding analyses and/or design changes.

    In contrast to the goal in the preparation of the TSPA-VA, the objective for the TSPA-LA should be to provide sufficient documentation so that it can be more readily defended as being either realistic or conservative. With these thoughts in mind, the Panel offers the following comments on the use and application of each of the identified approaches to these two types of processes.

    Updating the Component Models

    This approach primarily involves the improvement of the experimental and/or theoretical basis for the models. While, due to the physical size of the proposed repository and the temporal scale of the analyses, the site and design cannot be tested in a realistic manner, component models can be improved through refinements in the underlying subsystem models. At the same time, as the Panel noted in its Second Interim Report (
    Whipple et al., 1997), the capabilities of the project staff to analyze such complex systems over long time periods is uneven at best. In general, the performance assessment of complex systems may have significant errors if the wrong deterministic models are selected, if an inadequate or incomplete description of the system to be modeled is used, or if an incorrect analytical solution is applied.

    Acquisition of Additional Data

    Much can be accomplished through the development and acquisition of new data. This can be achieved through the conduct of well designed experiments, through observations and studies of natural analogues, and through analyses of other relevant systems in the field. All such steps can enhance the ability of the analysts to test the validity of existing component models and to increase their confidence in the estimates these models provide. Data that can be used to challenge the basic model formulations can be particularly useful. For example, data on 36Cl collected during the site investigation imply faster travel of water in the unsaturated zone than was estimated using analytical models. Similarly, the modeling of transport has been revised in response to measurements of the groundwater transport of plutonium-bearing colloids at the Nevada Test Site. Additional discussion on this subject is provided in Section IV.

    Bounding Analyses

    Applications of bounding analyses generally produce results that are conservative. For this reason, the outcomes of such analyses are generally assumed to be highly credible by regulatory agencies. In addition, such analyses are commonly less data-intensive than those conducted on amore realistic basis. As a result, bounding analyses are particularly useful in cases where the existing analytical models have significant deficiencies that would be difficult and time consuming to correct. A good example of processes that fall into this category are those that are highly complex and extensively coupled. The application of bounding analyses would appear to be especially appropriate as the project staff approaches the preparation of the anticipated TSPA-LA. The chosen applications must, however, be defensible, and care should be taken to ensure that the performance of the systems to which the analyses are applied have only a minor effect on the results of the overall assessment. Otherwise, the use of bounding analyses may result in unacceptably conservative projections of the performance of the overall repository system.

    That this can be the case was demonstrated in several instances in the preparation of the TSPA-VA. One example was the case in TSPA-95 (CRWMS M&O, 1995) where the analysts assumed that a waste package failed completely at the time of the first pinhole leak. For purposes of the TSPA-VA, a less conservative analysis was used to model this process. At the same time, it should be recognized that there were some issues for which the TSPA-VA was intentionally conservative. One example was the use of a bounding analysis to assess the risks from potential criticalities.

    There are other cautions that should be observed in the application of bounding analyses. For a complex, non-linear system, it is not always readily apparent how conditions that bound performance should be defined. This makes it difficult to judge whether, and the degree to which, the generated results are conservative. Because of the difficulties inherent in developing fully-coupled models for analyzing the flow and transport in the unsaturated zone, it may prove advantageous to begin with a simpler set of models, and then to evaluate the more complex issues through either sensitivity studies or bounding evaluations. If these efforts demonstrate that certain aspects of the complex coupled phenomena can be ignored or treated one-dimensionally, the overall analysis will be vastly simplified. More effort, however, needs to be directed to defending this approach and ensuring that coupled effects, that are potentially detrimental to repository performance, are addressed in this manner.

    Design Changes

    As noted above, the incorporation of changes in the design of the repository system is suggested primarily as a means for addressing processes whose analyses are essentially intractable. Although this is not likely to be applicable to the processes affecting the transport of radionuclides in the saturated zone or biosphere, it does appear to be relevant to the analyses of the complexities associated with the coupled processes that occur as a result of the increased heat loading within the proposed repository. This is clearly a design issue. It is important to note, however, that the trade-offs between hot designs that delay water contact with the waste packages, versus cooler designs that may induce smaller changes in the natural system, are complex. The degree to which cooler designs are likely to be easier to analyze should be considered in the trade-off analysis.

    Although the subject of changes in design is outside the scope of its charge, the Panel recognizes that there are other examples that might be considered. These include the possible use of backfill to avoid damage to the waste packages and other protective barriers by rockfalls. Again, however, this would likely result in higher temperatures inside the waste packages which, in turn, leads to a trade-off between protecting the waste packages or limiting the peak cladding temperature. Furthermore, such a design change is likely to lead to additional uncertainties in the chemical environment that will be experienced by the waste packages.

    Concerns regarding the use of the results of the TSPA as a guide for making changes in the design of the proposed repository were discussed in Section II C.

    Defense-in-Depth

    As noted above, incorporation of the "defense-in-depth" concept into the design of the overall repository system can also provide increased confidence in the performance of the proposed repository system. In fact, the viability of Yucca Mountain as a nuclear waste repository finally must rest on the evaluation of safety (expressed as some measure of radiation exposure to individuals or to a critical population). The TSPA is the primary tool for this evaluation; however, its inevitable complexity may obscure or even confound the safety analysis. For this reason, it is likely that, at the anticipated LA phase, the TSPA staff will be required to establish a fundamental safety case for the proposed repository. Developing such a case basically involves describing why and how the staff believes the repository could perform safely.

    Accomplishing this goal is intimately involved with the defense-in-depth safety philosophy long used to provide an acceptable level of safety for other types of nuclear operations, that is to say, the analysts should take advantage of the use of redundancies in various barriers and systems to protect the public from being exposed at unacceptable levels. In nuclear power plant regulation, the objective in applying this philosophy is to assure that the system will perform safely even if one or more individual barriers has failed. Theanalogous defense-in-depth requirements for a repository are not yet clear. While the assessment of defense-in-depth takes place outside the TSPA and, therefore, outside the Panel’s charge, we believe that the TSPA methodology can be a useful tool for assessing defense-in-depth. Just as in the case of the analyses of the performance of nuclear power plants, we believe that the TSPA methodology can provide a means of estimating how well a system of barriers within the proposed repository would perform, even when one or more of the barriers within the system is assumed to have failed (USNRC, 1998; CRWMS M&O, 1998e).

    Conclusions and Recommendations

    While incremental improvements in the analyses of coupled processes can be made, the Panel has concluded that a detailed, technical defense of these analyses cannot be demonstrated at the present time. As a consequence, the Panel recommends that in the TSPA-LA phase, the processes be treated by the use of either bounding analyses or design changes, supported by the incorporation of the defense-in-depth philosophy into the overall design of the proposed repository system. In the case of thermal-hydrologic-mechanical-chemical coupled processes, the Panel has the following suggestion on how such a bounding analysis could be performed. The suggestion is based on the observation that the output of an ideal analysis of these processes would describe the duration of the thermal period, and the pattern and flow rates of water after the thermal period has ended. Although the Panel does not think that it is possible to analyze these coupled processes in detail, we have concluded that it may be possible to determine reasonable bounds for the following factors. The time curve over which the repository will heat up and then cool. This will enable the analysts to estimate when the waste packages will experience increasing humidity levels, subsequently followed by the flow of liquid water;

  •  
  • The quantity of water that will flow on or into waste packages. This is likely to be bounded by the estimated infiltration rate at the repository horizon for each particular climate regime considered. The infiltration associated with the long-term average climate appears to the Panel to provide a reasonable bounding value.
  •  
  • Where the water will go and how many waste packages will experience liquid drips. The uncertainties in the SEEPS model may be such that the TSPA staff may consider the bounding case in which all the waste packages are assumed to be wet.

    Other effects such as those on the chemistry of the water entering the drifts are likely to be of less importance, because the chemical conditions at the waste package surface are likely to be determined more by the local versus the far-field environment

    E. Overall Conclusions about the TSPA-VA Methodology

    The Panel believes that the basic framework or architecture of the TSPA-VA is sound, as is the use of abstractions of component models for purposes of computational efficiency. Where the Panel has concerns, it is more often due to the specific methods applied and the details of the component models, rather than with how the models are linked. In some instances, as noted in
    Sections II and IV, the Panel is concerned that inappropriately optimistic analyses may have been made.

    In summary, the challenging features of the present TSPA-VA are that: (1) the already complex models are coupled; (2) the models are being extrapolated into temporal and spatial scales that are well beyond experimental data bases or human experience; and (3) there is little testing of the component submodels. Compounding the problem, there can be no test of the fully-coupled and extrapolated models used in the TSPA.

    As summarized above, the TSPA-VA staff used a mixture of analyses, some of which were intended to be realistic, and others of which were intended to be conservative. For these reasons, and the limitations of the component models cited in Section IV, the Panel concludes that the TSPA-VA cannot be viewed as an accurate projection of the "probable behavior" of the repository.

    In the course of its review, the Panel has noted the inherent difficulty of several aspects of the performance assessment. Our purpose in doing so is to distinguish between those cases where refinements in the modeling and the acquisition of additional data will permit significant improvements to be made in the analysis, and those cases that may be essentially intractable within the time constraints under which the TSPA staff is operating. Our comments are not meant to excuse the Department of Energy from meeting its obligation of demonstrating with the required degree of confidence that the repository will meet or exceed the specified performance targets, should a license application be submitted the USNRC. Instead, they are to suggest that the approach to resolving deficiencies in the TSPA-VA, and the work toward preparation of the TSPA-LA, should be based on a clear understanding of the nature and cause of each deficiency.

    For cases in which it is feasible to improve either the component models or their underlying data, the Panel recommends that efforts be made to implement such improvements wherever such changes would affect the overall assessment. Where conservative bounding analyses do not result in unduly pessimistic estimates of the total system performance, the Panel recognizes that it may not be cost-effective to spend additional time and effort refining the assessments and making them more realistic. For those issues for which, by virtue of their complexity, it is not feasible to produce more realistic models supported by data, the Panel recommends that a combination of bounding analyses and design changes be applied.

    Our purpose in distinguishing between these situations is to acknowledge that there are some aspects of the analysis for which additional data collection and modeling will produce only small reductions in uncertainty. In such cases, we recommend that the TSPA staff demonstrate, where possible, either in the TSPA-VA reference design or in a revised design, that the cited uncertainties have only limited consequences with respect to the overall repository performance.

     

     

    IV. COMPONENT MODELS OF TSPA-VA

    A. The Unsaturated Zone under Initial Conditions

    Three issues are addressed in Chapter 2 of the Technical Basis Document (
    CRWMS M&O, 1998c) on unsaturated zone flow: (1) estimating the infiltration rate and infiltration maps; (2) characterizing the hydrologic properties of the site; and (3) estimating seepage in the drifts under postulated (ambient) conditions. In general, substantial improvements over previous TSPA efforts are reported in all these areas. However, considerable uncertainties still remain.

    Infiltration Rate

    The infiltration rate, its variation in time, and to a lesser degree its variation in space (infiltration maps), are key variables to repository performance. Infiltration affects practically all other TSPA components and plays a crucial role on the rates of release of radionuclides and their transport in the unsaturated zone (UZ) and the saturated zone (SZ). It is also a key parameter in the characterization of the hydrologic and transport properties of the site.

    Estimating the magnitude and the temporal and spatial variability of the infiltration rate is a difficult task. The issue is complicated by the fact that direct measurements of the infiltration rate are not available or easily obtainable (although there is anecdotal reference in the UZ expert elicitation report of a measurement of 50 mm/yr in the Exploratory Studies Facility). Therefore, infiltration rates, infiltration spatial maps and their variation in time are inferred in the TSPA-VA from indirect measurements or from mathematical models. Compared to previous reports, the TSPA-VA team has implemented a substantial change in its estimate of the current infiltration rate, which has been revised upwards by one order of magnitude. This change is a reflection of field evidence accumulated in recent years. The dramatic increase is consistent with the recent discovery of fast flow paths in the mountain, as evidenced by chlorine-36 findings. The revision has far-reaching effects, and has led to many qualitative changes in the conceptual models of flow and transport in the mountain. In particular, it has shifted the project's interpretation of flow and transport from predominantly matrix-dominated regimes to predominantly fracture-dominated (fast path) regimes.

    In parallel with the revised estimate of a higher infiltration rate, the project also adopted a new approach for representing variation in precipitation in the base case by considering three substantially different climates of variable duration, dry (present), wet (long-term-average) and super-pluvial. The approach towards wetter conditions in the mountain, both present and future, is conservative and represents a drastic improvement over previous TSPA models, in which a constant (and much smaller) infiltration rate was used. At the same time, the climatic conditions for the base case and the postulated scenarios for switching between various climates are simply hypothetical and have large associated uncertainties (see also the related discussion of climate in subsection
    IV.K.).

    Currently, the estimation of the spatial and temporal variability of infiltration is based on the use of water balance models at the mountain surface, coupled with hypotheses about future climates. Infiltration maps are computed from the models but are not directly measured. These models contain a number of simplifying assumptions on processes, such as precipitation, evapotranspiration, run-on and run-off, and on parameters, such as soil depth, flow dimensionality, etc. In the absence of experimental data to confirm these approximations, it is unclear how realistic they and the resulting infiltration maps are. The accuracy of the current infiltration maps was subject to some criticism by the UZ expert elicitation panel. The validity of future projections is also questionable, given that they are based on present-day values for the various model parameters (including vegetation, cloudiness, etc.).

    Hydrologic Properties

    An accurate characterization of the hydrologic and transport properties of the site is essential for obtaining reliable predictions of the repository performance. Mountain-scale hydrology will influence the response to the thermal pulse, seepage into drifts, and the transport rates of released radionuclides. For example, it will provide the velocity fields to be used for flow and transport in the UZ during the life of the repository. The large disparity in the length scales involved, the intrinsic heterogeneity of the site, and the fractured nature of the mountain, however, present some difficult problems for the detailed characterization of the site. Ongoing efforts to address these problems through research are impressive and commendable, and although much is still unknown and/or uncertain, multiple advances have been made (
    Bodvarsson and Bandurraga, 1996).

    Currently, the characterization of Yucca Mountain is based on a combination of experimental data from boreholes, laboratory measurements, pneumatic tests, air injection tests, field measurements and inverse computer modeling to estimate parameter values. The revised interpretation of infiltration has led to a more realistic and improved characterization of the hydrologic properties. Key findings include: the need for a substantial reduction in the fracture-matrix interaction in order to match saturation data under the revised infiltration rate; the observation of a strong anisotropy (ratio of order 10) between vertical and horizontal fracture permeabilities above the repository; the estimation of the capillary properties of the fracture continuum; and the identification of perched water zones. Using inverse modeling and matching of data to simulations, a set of parameter values has been obtained. They form the basis for the computation in the TSPA-VA of the key hydrologic aspects of the site, such as the flow fields, the saturation profiles, and the transport pathways.

    Considering the complexity of the problem, the project has made many important strides. At the same time, significant uncertainties remain. The origin of these can be variously traced to the unprecedented scope of the problem, the novelty of some of the analytical problems involved, the inadequate representation of some physical processes, and/or the non-uniqueness inherent to inverse modeling. These uncertainties will propagate through many components of the report, closely linked to hydrology and transport, with significant ramifications on the final results. The Panel has singled out the following two areas, related to UZ flow in a fractured system and to inverse modeling, where additional advances will help increase the level of confidence on the TSPA estimates.

    UZ flow in a fractured system

    Determining UZ flow in a fractured system, such as at Yucca Mountain, involves the elucidation and representation of the process at different scales and geometries, such as a single-fracture, the interface between a single fracture and its matrix, the network of fractures, the matrix continuum and the interface between fracture continuum and matrix continuum. By necessity, however, the TSPA-VA approach is based on modeling the dual continuum of fractures and matrix, which is the simplest approximation that can capture the main macroscale (computational grid-scale) aspects of flow and transport in this complex geological setting. This approach is essentially a scale-up (an abstraction) of flow phenomena that occur over the multitude of smaller scales included in the respective fracture and matrix continua. Relevant issues that arise in this context include the representation of the flow and capillary properties at various scales and of the fracture-continuum/matrix-continuum interaction.

    With respect to upscaled flow and capillary properties, the approach taken in the TSPA-VA is to use the van Genuchten model derived for UZ flow in homogeneous soils. Although convenient, use of this model is not justifiable in the present context. Ignored in this approximation are a multitude of processes: the unstable nature of gravity-driven infiltration in real fractures; the possibility of hysteretic (and chaotic) behavior during episodic flow, as documented in recent experiments in related systems (Faybishenko et al., 1998); the effect of subgrid-scale heterogeneities, including correlated structures, anisotropy in fracture permeability, and saturation gradients; the effect of the connectivity of the fracture and matrix continua; and the effect of abrupt changes in properties on transport fluxes expected along stratigraphic discontinuities. The last item has already been shown to be sensitive to the particular flux-weighting scheme used in the simulations. Also ignored are the differences between wetting and drying cycles, which are expected to develop during the heating period. A similarly questionable approach is used in the modeling of heat pipes in thermal hydrology, where recent findings have shown a complex flow behavior (Hardin and Chestnut, 1997).

    With respect to the representation of the fracture-continuum/matrix-continuum coupling, the increase in the estimated infiltration rate has forced the introduction in the dual continuum (DKM) model of an adjustable fracture-matrix interaction factor. In this way, a non-trivial fraction of the infiltration is forced to partition in the fracture continuum. Using the inverse modeling calibration procedure, reducing this interaction factor by as much as four orders of magnitude, has enabled the TSPA team to accommodate changes in the revised infiltration rate, without producing unphysical changes in other hydrological properties.

    The introduction of a reduction factor is reasonable and appropriate in order to account for a variety of processes, which are not included currently in the description of physics at the various scales, such as the scale of a single fracture and the scale of a numerical grid block, as mentioned above. However, in the current approach of the TSPA, this reduction factor is simply an adjustable parameter, devoid of convincing physical meaning and often taking values as small as 0.0001. This is not satisfactory and reflects a lack of understanding of the actual physics of the process and, more generally, the lack of progress in the scale-up of two-phase flow in the fractured system, as also noted above.

    The difficulties in the above two issues are compounded by the lack of convincing field data to support the representations taken, inasmuch as reliable flow data have only been gathered from core studies. As a result, the Panel is skeptical of the validity of the base case set of hydrologic parameters and particularly of the van Genuchten-type capillary and flow properties of the fracture network and of the fracture-matrix reduction factor. These are all key variables in the partition of flow between fractures and matrix. Given the significance to other TSPA components (seepage fluxes into drifts, thermohydrology, and UZ transport), the Panel believes that efforts should be made to reduce the existing uncertainties, using analytical studies and field tests. Although acknowledging that the upscaling of UZ flow in a fractured system is a non-trivial task, the Panel believes that such a step is also necessary in order to conclusively and unambiguously determine the relevant hydrologic response of the site in developing the TSPA-VA.

    Inverse modeling

    The extraction of the base case set of parameters in the TSPA-VA is based on matching the predictions of the DKM model to various data, including borehole data on saturations, by an inverse modeling procedure. The approach is novel, compared to previous work, and commendable. The complexity of the problem, however, requires that a large number of parameters (more than 150) be estimated from a small number of data sets (less than 300). In addition, various assumptions are introduced for computational or other reasons. For example, the assumption is made of strictly vertical flow and vertical variation in properties. This 1-D philosophy is also adopted in other components of the TSPA (such as the simplified description of thermohydrology). Although consistent with a strong anisotropy in some of the layers, this assumption may fail across stratigraphic discontinuities which favor lateral flow (as in the perched water zone). For certain sets of data, the inverse modeling involves matching of point values with the grid-block values of the upscaled variables of the DKM as discussed above. Given that the upscaling procedure is yet to be validated, such matching may not be meaningful. Several of these data have considerable uncertainties. Finally, in order to match the observed perched water zones with the results of the computer model, it was necessary to reduce the fracture permeabilities below the perched water zones by orders of magnitude. Whether this is representative of the true conditions remains to be demonstrated by field tests.

    Coupled with the inherent non-uniqueness of the inverse modeling procedure and the questions on the realism of the modeling cited above, these approximations cast doubt on the validity of the obtained base case set of parameters. The Panel believes that at this stage there is still a considerable need fo