The Waste Isolation Pilot Plant: A Potential Solution for the Disposal of Transuranic Waste (1996)

The purpose of this chapter is to review and comment on the state of the U.S. Department of Energy's (DOE's) understanding of hydrology in formations at the WIPP site above the Salado. Studies of these subsurface aquifers ideally should aim to define the flow of ground-water as well as possible. Virtually all feasible pathways by which radionuclides can move from the repository to the accessible environment depend on ground-water transport.

Site hydrology considerations are thus relevant to predicting the fate of any radionuclide release from a WIPP repository. Any such release will depend predominantly on three factors:

1. the volume of water (brine) that circulates through the waste-filled rooms and is released;¹
2. the concentration of radionuclides in this brine; and
3. the rate of transport of the radionuclide-contaminated brine through the permeable sediments above the Salado Formation (e.g., Culebra Dolomite, Dewey Lake Red Beds) to the accessible environment.

WIPP repository design measures that can be taken to reduce the volume of brine release have been discussed in Chapter 4. Detailed studies to determine radionuclide solubilities, and hence concentrations, in brine have been discussed in Chapter 5. This chapter considers the third factor, considerations of ground-water flow and radionuclide transport.

If it can be demonstrated reliably that the concentration of radionuclides in the water and their rate of movement through the repository to the accessible environment together do not cause the release limits specified in 40 CFR 191 to be exceeded, then the repository can be considered to be in compliance with these regulations. Such a demonstration would, in principle, involve a combination of the three factors enumerated above, with their relative importance dictated by the natural site characteristics and any engineered enhancements to the repository design.

In the interests of brevity in this overall report, the significant advances in the understanding of the hydrogeologic framework that have been made since the WIPP investigations were started (i.e., since the mid-1970s) are not discussed in detail in this chapter. Appendix A summarizes the regional hydrogeology of the WIPP site.

The latest formal analysis available to the committee and discussed here is the 1992 Performance Assessment (PA). This model assumed a two-dimensional, steady state, horizontal flow of ground-water in the Culebra Dolomite, with simple parameterizations of both chemical and physical retardation mechanisms to describe solute transport.

The discussion below outlines the main shortcomings of the 1992 PA models and analysis. Some of the criticisms reflect current limitations of the state of art of ground-water and transport study in general. Some criticisms can be resolved relatively easily, but other criticisms may require considerable effort to address.

The assumptions of the 1992 PA predicted a release of a mass of contaminants under disturbed conditions that, for a compliance demonstration, required a reliance on two features: long ground-water travel times in the Rustler Formation (primarily in the Culebra Dolomite) overlying the Salado Formation, and significant retardation of radionuclides. If this remains true in the PA calculation submitted as part of DOE's compliance certification submission in October 1996, then the committee believes that DOE will have to

improve the case presented in the 1992 PA substantially to establish the technical credibility of the assumptions and values used in the analysis.

Most of the critiques in the following discussion of non-Salado hydrology relate to assumptions or numerical methods in the 1992 PA that need to be documented better or evaluated further before a high degree of reliance can be placed in the overall model results. For example, the PA models allow chemical retardation or matrix diffusion to attenuate radionuclides after release to the Culebra. Although it seems likely to the committee that some chemical retardation and matrix diffusion will occur, the documented basis for the assumptions about retardation in the 1992 PA is very weak. It is possible, and in some cases probable, that the final PA models and more recent experimental results—which have not yet been reviewed by the committee, but which will be included in the final compliance submission—will resolve many of these concerns.

Considerable effort has been expended by DOE over the last four years to increase its understanding of the ground-water and transport regimes at WIPP, and informal reports of significant progress have been made. The considerations raised in the following discussions point out shortcomings in past (i.e., 1992) modeling efforts, but the committee has determined that they do not challenge its position that DOE should be able to complete the work required to show that WIPP can comply with 40 CFR 191, and that the repository can probably be shown to present a low risk of human exposure to radionuclides, even in the event of a severe (E1E2) intrusion. However, to the extent that compliance demonstrations may rely on ground-water flow and transport in the non-Salado formations, they will require examination of at least some of the hydrogeology and transport issues discussed below.

A general discussion of ground-water modeling efforts and common difficulties is provided in this section (see Box 6.1 for a summary of key features). Supplemental background discussions of these issues are included in Appendix F. Models that simulate ground-water flow and transport processes are an essential tool in the characterization, analysis, and performance assessment of WIPP. Model predictions also lie at the heart of the PA process for the WIPP site (see Chapter 2). Therefore, it must be demonstrated that a high level of confidence in the PA models is justified.

The accuracy and precision of predictions of transport of radionuclides through a deep regional aquifer system depend on the level of understanding of the processes governing flow and transport through that system, how well the relevant properties of the system are defined, and how accurately the boundary conditions of the system are known. If the assessment of the isolation capability of the site is expected to provide a reasonable level of confidence that the predictions bound the range of plausible values, then the predictions must be based on an appropriate and adequate level of understanding of the processes, properties, and boundaries of the aquifer system and of how these factors may change in the future. If the numerical PA models represent the wrong processes, ignore relevant processes, or incorporate erroneous or biased assumptions or coefficients, then the results may be misleading. These considerations of ground-water modeling efforts as applied to WIPP are discussed below.

In applying ground-water models to field problems, there are three primary sources of error, as described by Konikow and Bredehoeft (1993):

1. Conceptual errors: theoretical misconceptions about the basic processes that are incorporated in the model.
2. Numerical errors: the difference between a true and exact solution to the governing equations and the solution developed by the equation-solving algorithm.
3. Uncertainties and inadequacies in the input data: errors arising from the inability to describe actual aquifer properties, stresses, and boundaries. These include measurement errors, interpolation and extrapolation errors, and inaccurate predictions of future conditions.

In most model applications, conceptualization problems and uncertainty concerning the data are the most common sources of error. All three aspects of error are discussed below with reference to the PA models for WIPP.

If the conceptual models upon which the PA models are based are in error, then all of the complementary cumulative distribution functions (CCDFs; see Chapter 2 and Appendix B) and risk assessments derived from the PA also may be in error. Although the tool of PA is designed to account for uncertainties in the parameters, this assessment of uncertainty is necessarily within the context of the PA model used. Factors and processes that are not incorporated into this model are conceptual errors that are not treated by the PA sensitivity analysis.

The 1992 PA for WIPP did not assess adequately the uncertainties in the underlying conceptual model, which might be a more significant effect than the uncertainties in the parameters. For example, because the 1992 PA ground-water model was two dimensional, it cannot be used directly to evaluate the importance of a three-dimensional flow. Errors in a conceptual model are often difficult to detect, and sometimes field measurements can be reproduced by several alternative conceptual models. The alternative models may make little difference over the short duration of a field test but may make a large difference over the life of the repository and predictive horizon. It is therefore critical that the PA conceptual models be rigorously and independently tested.

There also must be consistency among the various conceptual models that describe different elements of the subsurface ground-water system. For example, hydrologic concepts indicate that ground-water in the Culebra Dolomite at and near the WIPP site flows generally from north to south. Chapman (1988), among others, notes that there are significant decreases in salinity and changes in water quality in the direction of flow. Siegel and Anderholm (1994, p. 2299), state that ''no plausible geochemical process has been identified that would cause this transformation in a hydrologically confined unit." Thus, geochemical interpretations appear to conflict with hydrologic

understanding.³ Such a discrepancy reflects a lack of comprehensive understanding of the subsurface system, which opens the possibility that some aspects of the overall conceptual model of the Culebra are flawed, and that predictions based on this model are likewise uncertain.

The 1992 PA model used the SECO ground-water modeling computer codes to simulate flow and transport. The SECO codes have been developed by recognized experts in numerical methods, and the transport code apparently represents a major advance in controlling and limiting numerical dispersion. However, because the SECO codes are not yet used widely, some "bugs" and other problems in the use of the codes may still remain undetected. Experience with other ground-water model codes indicates that code developers often cannot comprehensively test the code for all possible ranges and combinations of parameter values. Model applications by a variety of users and organizations to a range of problem types and hydrogeologic conditions often will result in the discovery of code deficiencies that were not detected by the model developers, regardless of how comprehensive their testing efforts.

The SECO codes have not yet been subject to extensive benchmarking or international comparisons with other widely accepted codes, such as have been performed for other codes under the recent International Hydrologic Code Intercomparison Project (HYDROCOIN) and the International Project to Study Validation of Geosphere Transport Models (INTRAVAL; Swedish Nuclear Power Inspectorate, 1987, 1990). The committee believes that it is inappropriate to rely exclusively on these relatively new codes before such benchmarking evaluations have been completed in an open, broad-based scientific forum to yield more confidence in the codes.

The PA models make predictions based on ranges of values of parameters. The 1992 PA assumed that most of the parameters in the model are "independent random variables even though it is known that some were dependent on others" (Sandia National Laboratories, 1992). It is further stated (Sandia National Laboratories, 1992, p. 1-21) that "the effects of neglecting correlations on the sensitivity/uncertainty analyses are generally unknown." It may be a reasonable first approximation to assume independence of parameters in sampling (i.e., to ignore interactions between variables and their codependence on other factors). However, such an assumption must be carefully tested, and refined if necessary, before a final PA is completed. Independent sampling can yield simulations based on implausible (or contradictory) combinations of parameters. If results of these simulations yield low-risk outcomes, then including the results of such simulations in the generation of CCDFs might skew or bias statistical distributions of the results.

The committee recommends that the assumed independence, ranges, and distributions for the various parameters that are sampled in PA analysis be examined closely for reasonableness and consistency with observed data.

As noted previously, the 1992 PA models assumed that ground-water flow through the Culebra Dolomite was two-dimensional, steady-state, and horizontal. Investigations of the larger three-dimensional regional system, of which the Culebra is actually a part, have begun relatively recently but have not yet been completed. These investigations include a new synthesis of all the regional hydrogeologic data and will lead to the development of a three-dimensional numerical flow model. This more comprehensive analysis of the ground-water flow system is very important and relevant to the larger issue of conceptual uncertainty. Among other benefits, this type of analysis may help to resolve existing discrepancies between the

geochemical data and the present conceptual model of the flow system. Until the results of these ongoing studies have been completed and their impact on assumptions about the Culebra has been evaluated, it is too early to determine whether or not many of the simplifying assumptions incorporated into the 1992 PA were reasonable.

Ground-water flow and solute transport are affected strongly by heterogeneity in aquifer properties. Fractures represent an extreme but prevalent type of heterogeneity, so their characterization is important for reliable model development. However, the properties, continuity, and extent of fractures in the Culebra are still highly uncertain. One cost-effective approach to characterize fractures is to complete detailed mapping of any surficial expressions of fracture traces and lineaments at both local and regional scales. This should have been completed in the early stages of site characterization, but there is no written documentation of such analyses having been done.

Fracture and lineament mapping have been applied successfully to several potential fractured-rock repository sites (e.g., for applications in Sweden, see Wikberg et al., 1991; Ahlbom et al., 1992; Swedish Nuclear Fuel and Waste Management Company, 1995). Surface and subsurface features that might represent fracture zones and lineaments were defined by using remote-sensing methods (such as aerial photography and satellite imagery), topographic data (including computer analysis of digital elevation data), and several geophysical surveying techniques (including geomagnetic, electrical, and seismic methods and some recently developed high-resolution, three-dimensional tomographic techniques).

These approaches have helped to clarify the possible existence and spacing of major through-going fracture zones. Although the lithology at WIPP is different from that at the Swedish sites, these approaches have worked in other sedimentary rock basins, and the committee believes that they should be applied in the characterization of the WIPP site. An approach to fracture definition that is potentially more definitive, but also significantly more expensive, would involve drilling of slanted or even horizontal boreholes. However, because this probably would require an extremely expensive and time consuming operation, the possible benefits should be carefully weighted against the likely costs before a decision to undertake such a study is made.

Significant accomplishments in site characterization have advanced the level of understanding of the regional ground-water flow system at the WIPP site during the past 20 years (see Appendix A for a summary discussion). However, the degree and comprehensiveness of the understanding of some aspects of the local and regional ground-water flow systems are lower than desirable relative to the importance of the repository and to the time and money spent studying the WIPP site and constructing the repository. The following points support this contention:

• There is a disparity (as discussed above) between the regional flow system as determined by geochemical and isotopic interpretations and that indicated by hydrologic analyses. This disparity reflects an inadequate level of understanding of the whole system, and it should be resolved.
• The water table in the region, which represents the top of the saturated zone, has not been defined. This important boundary should be defined, or a reasonable explanation of either why it is not important or why it is technically or economically infeasible to determine should be documented.
• The rates, locations, and mechanisms of natural recharge to and discharge from the Culebra Dolomite are poorly defined (see Brinster, 1991).
• Apparently, there have been no detailed hydrologic studies of the unsaturated zone at the site. Such studies would help to (1) estimate if, why, and how much recharge may exist in the vicinity of the site; (2) define the water-table position; (3) predetermine an appropriate response should there be a spill or accident on the land surface at the WIPP site; and (4) predict how much and where recharge might increase under a scenario of climate change that yields increased precipitation.
• An adequate explanation is lacking for observed changes in water levels in the Culebra, where trends of rising water levels have persisted for several years.

• Observed changes in water levels from assumed steady-state conditions were not incorporated into the 1992 PA analysis. However, if the causes of the observed water-level changes during the last several years are unknown, then how is it to be known that even greater changes in the flow field might not occur in the near future? Such changes might invalidate the PA assumptions and predictions.
• These water-level changes are affecting hydraulic gradients, flow directions, and flow velocities and thereby provide some evidence against the assumption that steady-state flow prevails in the aquifer, an assumption that is the basis of many WIPP hydrologic analyses. Small "violations" of this assumption probably would not affect the conclusions to a noticeable degree. However, if the trends observed over the past five or six years persist, recur, or become more widespread, flow systems might change enough to negate the PA analyses based on the assumed steady-state flow field.
• There has been too little accounting for the three-dimensional nature of ground-water flow and consequent leakage through confining layers. If vertical components of flow do materially affect the Culebra, then the ground-water models of the Culebra that were calibrated under the assumption that flow is horizontally two dimensional must include compensating errors to offset ignoring the vertical leakage. Such errors can induce a bias in long-term predictions of flow and transport.

Some necessary and feasible experiments and analyses (of the kind mentioned here), which would lead to a more comprehensive understanding of the ground-water flow system and to more reliable PA models, have either not been undertaken or are not completed yet.

Ideally, the regional system should be understood with sufficient accuracy and precision that professional hydrologists will have confidence in the analyses and predictions, will believe that additional data collection is not critical, and will be convinced that there are no major "surprises" left in the characterization of the system.

Although it is difficult to define precisely when a requisite level of confidence in the hydrogeologic analyses has been achieved, the committee does not believe it has been reached in the 1992 PA. Too many questions remain about the models, assumptions, and data used to analyze the cases in which a significant release of contaminants to the Culebra or other formations is projected to occur.

The potential weaknesses in the ground-water models will be important to a demonstration of compliance only if there are significant releases from WIPP to the stratigraphic units shallower than the Salado. The weaknesses in hydrogeologic modeling discussed in this chapter should be balanced against an examination of the assumptions from which the estimated releases are derived, and assumptions concerning pathways to cattle that drink from stockwells tapping the Culebra. For example, even if contaminants were transported through fast-pathway fractures in the Culebra and were not subject to any retardation processes, the very same conditions that would enable this to occur would assure that the water carrying the contaminants would be far too saline for cattle to consume. This would appear to eliminate the potential for future human exposure, which is an underlying basis for the overall concerns about radionuclide releases.

The non-potable saline character of the ground-water in the Culebra will greatly reduce the exposure hazard even though the radionuclides released to the accessible environment may exceed the limits required for compliance in 40 CFR 191. Should the 1996 PA analysis be able to demonstrate convincingly that the releases are sufficiently low to meet compliance standards, then the exposure hazard would be correspondingly lower.

Transport of radionuclides released into an active ground-water system can and may occur at a rate less than the rate of the ground-water flow, due to physical and chemical interactions with neighboring rock. This radionuclide retardation, where it can be shown to exist, can add important additional assurance with respect to the radionuclide containment capability of the repository. Hence, studies of the radionuclide retardation properties of water-conducting rock formations at WIPP (see Appendix A) are an important complement to the studies of ground-water flow.

Following this discussion of solute transport, the specific retardation mechanisms are treated in turn.

The solute-transport model of the Culebra Dolomite is based on a theoretical understanding of transport and dispersion processes in ground-water. However, this underlying theory is itself somewhat weak because the scientific understanding of transport processes in heterogeneous media, particularly in fractured rock systems, is still in a stage of relatively rapid development and evolution. Much remains to be learned, and much ongoing research is devoted to developing a better understanding of the governing processes and parameters at large scales in complex field environments. Recent research indicates that many transport processes and parameters are highly scale dependent. This is one reason why there is often little basis for extrapolation of theory and lab tests to the field environment for predictive purposes.

The 1992 PA separately sampled and independently varied porosity, fracture spacing, and transmissivity. However, it is known that correlations exist among these parameters. Thus, it seems likely that the existing approach may yield unreasonable or unlikely combinations of these parameters in some of the individual simulations. The net impact on the overall PA is uncertain, but the committee believes that the parameter sampling procedure should be examined closely and refined if necessary.

The PA models represent leakage from a borehole into the Culebra as an initial concentration of solute in the transport model for a steady-state flow field. However, transient changes in the flow field, which are induced by the volume of water carrying the radionuclides into the Culebra, are not considered. While the leak is occurring, the velocity away from the well would be greater than the velocity under steady-state conditions, and some of the radionuclides would migrate faster and further than represented in the PA model. The committee recommends that the PA process carefully test and demonstrate that the steady-flow assumption and simplification, which ignore the initial fast advection, are reasonable and do not introduce significant errors.

The effective porosity (ε) is a critical transport parameter for which few measurements are available. If one assumes that ε = 0.001, which is representative of fracture flow, then travel times to the regulatory boundary would be on the order of hundreds of years. However, if ε = 0.16, which is representative of the rock matrix, travel times would be on the order of many thousands to tens of thousands of years. This could mean the difference between compliance and noncompliance of the site.

The 1992 PA assumed that porosity and transmissivity vary independently. This is a potential weakness in the analysis that needs to be tested, because physically reasonable arguments lead to correlations between the two variables. In this regard, a key issue for the PA flow model is whether the relation between transmissivity and porosity is in fact independent, or if the relation is direct or inverse. Either relation could be conceptualized (or perhaps rationalized), based on present knowledge of the Culebra Dolomite. However, any consistent relation would be contrary to the WIPP Project PA procedure of sampling independently from the populations (or distributions) of transmissivity and porosity. Thus, many simulations would be based on unlikely or contradictory combinations of parameter values. This could bias the PA model-generated statistics on which the CCDFs are based.

The PA sampling approach has yielded some individual outcomes to date that have exceeded the compliance standard. These particular simulations should be examined carefully to highlight and document the combinations of parameter values and boundary conditions that led to prediction of an unacceptable release of radionuclides to the environment. This evaluation would allow analysts and reviewers to focus on an assessment of the reasonableness of those specific and critical conditions, and would serve as a quality assurance check on the statistics generated by the PA analyses.

Some of the 1992 PA simulations of solute transport have allowed physical and/or chemical retardation, as would be derived from matrix diffusion and chemical sorption, respectively. To date, however, no reliable and representative field measurements of the

parameters controlling matrix diffusion and chemical retardation have been carried out at the WIPP site.

In the 1992 PA model of the Culebra Dolomite, physical retardation may occur in response to matrix diffusion in a dual-porosity system. The magnitude of the retardation is proportional to the effective surface area and volume of the blocks adjacent to the fractures through which advection is occurring. If the contaminant plume spreads through a larger volume of the aquifer, then matrix diffusion will be more effective in removing contaminants from the active flow field. If the transport model artificially spreads the contaminant plume into a larger volume of the aquifer than would actually occur, then the model will overestimate the retardation caused by matrix diffusion, even if the diffusion parameters themselves are estimated accurately.

The artificial spreading of contaminants can arise from errors in the numerical solution, from errors in the initial conditions, and from errors in the conceptual model. All are present to some degree in the WIPP PA model. The committee recommends that PA assumptions and analyses for cases including matrix diffusion be supported by additional model studies to demonstrate that the analyses do not overestimate the magnitude of matrix diffusion because of artificial spreading of the plume. This can be accomplished, for example, by applying a transient ground-water flow model that uses, for a relatively short time period, a much finer spatial grid and smaller time steps near the area close to a human intrusion (HI) borehole. These results can then be compared with those of the coarser PA model for equivalent travel distances and elapsed times.

If PA results indicate that repository compliance with EPA standards depends strongly on matrix diffusion, more reliable field evidence to support the assumed range of coefficients should be obtained. In this case, field tests should be completed at several different locations in the vicinity of the WIPP site because of the spatial variability in the properties of the system.

The PA models combine all of the complex chemical reaction processes, such as sorption or exchange, that might retard the movement of a particular chemical species relative to the movement of a nonreactive solute into a single simple parameter—the retardation factor R_f. This factor, in turn, is related to a distribution coefficient K_d. Although the concept of a retardation factor is computationally efficient, it inherently assumes that reaction processes are linear, instantaneous, reversible, and homogeneous.

The concept generally works well in chemical plants where chemical engineers apply it to fluid flow and solute transport through tanks filled with relatively homogeneous, reactive, porous media. However, in a complex regional aquifer system, the reactions rarely are subject to these ideal conditions. The theoretical basis for applying a retardation factor to a field environment is therefore weak, at best, and its application to large-scale hydrogeologic settings, without restrictions or qualifications, has been criticized in the literature (e.g., Reardon, 1981; Miller and Benson, 1983; Valocchi, 1984).

From the perspective of a compliance assessment, use of the simplifying retardation factor may either underestimate or overestimate the overall amount of retardation of a solute relative to the flow of water, depending on the particular chemical species and the types and rates of reactions that affect its fate and transport. To probe this issue, the nature of the expected isotherm for each solute of concern (and within the range of expected concentrations) should be determined experimentally.

Scientists from Sandia National Laboratories recognize the numerous difficulties and weaknesses in applying a linear sorption model for predicting radionuclide transport through the Culebra Dolomite, and many of these concerns have been documented (Novak, 1992). As an example of one of several serious concerns, Novak (1992) points out that available K_d data for uranium sorption in the Culebra are very sensitive to water composition and vary by more than three orders of magnitude for selected representative compositions of Culebra brine. Novak (1992, p. 55) states that "existing radionuclide K_d data are insufficient for predicting radionuclide migration in

the Culebra with a sound scientific basis within reasonable certainty." Ongoing experiments will help to resolve this uncertainty, but results and analyses are not yet available for review.

If performance assessment results indicate that compliance is contingent on the occurrence of chemical retardation, it will be necessary to provide more evidence that the PA approach is reasonable for these conditions. The committee believes that the use of sorbing tracer tests in the field would be highly beneficial for building confidence in the analysis and approach to chemical retardation, in addition to determining whether or not the conceptual model is applicable to the field environment. Because the specific solutes of interest here, such as plutonium and americium, cannot be used in tracer tests, reasonably appropriate surrogate tracers must be used, or field evidence of sorptive behavior must be ascertained from analysis of previous accidental releases elsewhere.

As hypothesized for flow parameters for the Culebra Dolomite, a correlation may also exist between K_d (or R_f) and either transmissivity or porosity. For example, relatively high transmissivity values might coincide with zones of low or zero clay content. Because clay probably yields a higher sorptive capacity per unit volume than dolomite, where transmissivity and flow velocity are the highest, there may be very little chemical retardation. This illustrates one type of predictive error that might be associated with applying one average effective R_f to the entire ground-water system. The greatest risk of release in that case is related to flow through transmissive fractures having a lower than average R_f.

It is unclear whether the PA approach to sampling values of R_f is appropriate if the chemical retardation in critical high-transmissivity channels is not represented accurately by the average R_f assumed for the entire aquifer system. Again, the only way to address this issue in a manner that would instill confidence in the predictive analyses is through a series of in-situ tracer tests or other field-based analyses.

This chapter and Appendix F have focused on potential sources of error in the preliminary PA modeling of the Culebra Dolomite. Therefore, the reader should not infer from the tone of this chapter that the WIPP ground-water models are necessarily inaccurate or unacceptable. Rather, the intent is to indicate where the preliminary analyses presented in the 1992 PA need to be revised, strengthened, or better documented. Because there have been four years of additional studies since the 1992 PA, it is possible that recent experimental results and the final PA models, which have not yet been reviewed by the committee, will have resolved many of these questions.

A key question is whether the uncertainties in the properties of the subsurface system above the Salado Formation need to be resolved (or minimized) further to demonstrate adequately the integrity of the WIPP site with respect to containment of radionuclides. Alternatively, is enough information about the subsurface system already available to predict that any potential release from the proposed repository will not cause radionuclides to migrate faster or further than is acceptable from a regulatory standpoint?

These issues are addressed separately for the undisturbed case and the disturbed (i.e., HI) cases described in Chapter 2.

The committee finds that it is reasonably certain that the WIPP repository will provide an adequate level of long-term isolation of TRU waste under a broad range of natural conditions and stresses for tens of thousands of years into the future. Provided that seals and plugs (as discussed in Chapter 4) are effective, there is no evidence to indicate that any brine that might be contained in the repository after it is closed is likely to leak through the Salado Formation and be released into the regional ground-water flow system in shallower formations. Nor would the committee expect that ground-water flowing above or below the Salado might dissolve the salt or in any other way come in contact with the wastes that would be stored in the repository.

For consideration of the undisturbed case, the present state of understanding of the ground-water flow system above the Salado is adequate. The committee finds that the overall permeability and creep properties of the Salado are such that, with effective repository seals and without some form of human intrusion, the flow of radionuclide-contaminated brine from the Salado is very unlikely. Further studies of the hydrogeology of non-Salado units are therefore not necessary to support compliance for the undisturbed case.

Provided that plugs and seals in the Salado Formation are effective (see Chapter 4), there is high confidence in the conclusion that the WIPP repository will isolate TRU waste adequately from the environment for many thousands of years if it remains undisturbed by future societies. For this undisturbed case, continued studies of non-Salado hydrogeology are not necessary.

In the WIPP PA (see Chapter 2), the HI scenario assumes that future drilling will facilitate upward leakage of brine from the repository through a borehole, thereby permitting radionuclides to be released into an overlying geologic formation through which lateral migration offers a possible pathway for release of radionuclides to the accessible environment. If releases exceed regulatory limits, then a predicted rate of transport in the regional ground-water system above the Salado Formation becomes a critical factor in estimating cumulative releases.

The analysis of the disturbed case hinges on predicting flow and transport through rock formations above the Salado (see Appendix A). The issue of concern to the committee is whether the 1992 PA was based on an adequate conceptual understanding of the regional ground-water flow system near WIPP and whether the relevant parameters had been quantified adequately to support either "realistic" or "bounding" models in the PA.

For example, the 1992 PA assumed that the most likely pathway for a lateral release of radionuclides is through the Culebra Dolomite. However, a release to the shallower Dewey Lake Red Beds also is possible and does not appear to have been considered or evaluated adequately. Releases of saline ground-water from leaking deep boreholes directly to the water table are known to have occurred historically in other areas (for example, see Van der Leeden et al., 1975). The Dewey Lake is a plausible exposure pathway because it is known to contain some potable water. Among other concerns, climate changes can increase the recharge rate and raise water levels. In the committee's opinion, releases to the Dewey Lake cannot be discounted summarily; if a borehole to the Salado or to the Castile Formation were to connect these formations to brine at a pressure near lithostatic, then the hydraulic gradient (driving Darcy flow) would be sufficient to enable leakage into both the Culebra and the Dewey Lake if a pathway to either formation were to exist.

Also, the amount and rate of brine release appear to be very sensitive to the assumed permeability of the material that is filling the leaky borehole. In light of the importance of that parameter, values for borehole permeability that are used must be realistic and technically defensible.

If the final analyses of the disturbed-case scenarios yield projected releases of contaminants to formations above the Salado that are less than the release limits of 40 CFR 191 (i.e., for which compliance results even if the releases moved across the regulatory boundary in less than 10,000 years), then no further hydrogeologic characterization work is needed. This means that compliance could be demonstrated even if one assumes that no retardation occurs and that ground-water flow and transport occur through fractures.

For the disturbed case, if the mass of contaminants and brine released is as great as assumed in the 1992 PA, then the ability of the site to isolate waste satisfactorily under such "disturbed" conditions has not yet been demonstrated adequately. However, such a demonstration appears to be a plausible outcome if the following problems, issues, or limitations of present models and data are resolved through further studies.

1. The PA approach must assess the sensitivity of the predictions to uncertainties in the underlying conceptual

1. models. Alternative conceptual models are feasible, and those that are plausible and that could lead to significant releases should be evaluated. Completion of ongoing three-dimensional analyses of the regional ground-water flow system is critical to this effort and should be given a high priority.
2. The non-Salado site characterization efforts and PA models have been overly focused on a two-dimensional flow analysis of the Culebra Dolomite. The minimal levels of consideration and evaluation of other potential hydrogeologic release pathways (such as the Dewey Lake Red Beds) and of the interconnection between the Culebra and adjacent formations have not been justified adequately.
3. Fractures are a dominant control on the transmissivity of the aquifer system at the WIPP site and represent the highest-velocity channels for migration of contaminants. A better definition and understanding of the nature, density, spacing, length, and interconnectedness of fractures and fracture networks are critically needed. As one step to help achieve this understanding, the committee recommends that modern topographic and geophysical methods of analysis be applied in the site characterization efforts.
4. In general, the types of chemical and geochemical reactions, rates of reactions, and their dependence on spatially heterogeneous mineralogy and changing aqueous geochemical parameters (e.g., salinity, pH, and Eh) are not well characterized for the HI scenario at the WIPP site. Representing all of these sources of variation in the PA model as a single parameter (i.e., an effective retardation factor), as was done in the 1992 PA, has a very weak scientific basis. If the PA assumes that chemical retardation occurs in the Culebra, then supporting evidence from the field environment should be available before any predictions based on that assumption are accepted.
5. Similarly, if matrix diffusion is to be relied upon in the PA models to demonstrate compliance, then additional controlled field experiments must be completed to define rates of diffusion more accurately and to assess spatial variability in the process and parameters. Furthermore, additional testing and documentation are needed to eliminate concern that the net retardation caused by matrix diffusion is being calculated inaccurately in the numerical models.
6. For the disturbed case, if it can be shown that the mass of contaminants and brine released would be significantly less than previously estimated (as might be derived, for example, due to changes in the estimated severity of human intrusion, refinements in estimates of the actinide source term, or in the effects of borehole closure), then further studies of the Culebra would not be necessary. That is, if the total amount of solute entering the non-Salado formations is determined to be sufficiently small to assure compliance, then studies of ground-water transport are not needed for this purpose.

Insight into the difficulties of hydrogeologic site characterization issues is offered based on examples from and comparisons with international siting efforts. In efforts to site geologic repositories in other countries, a good quantitative understanding of the regional ground-water flow regime and the influence of the geological environment of the repository on radionuclide transport are often the most critical factors needed to determine whether or not a potential repository site can provide adequate long-term isolation of the waste. This is usually true for both the undisturbed and the disturbed cases.

Several countries have opted to site repositories in rock masses of low permeability, such as granites, in regions of low hydraulic gradient, so that flow rates will be very small. Determination of the rock-mass permeability in these cases is often complicated by the presence, on the field scale, of fractures that are not usually present or tested in laboratory specimens. Fractures also complicate larger-scale tests conducted directly in the field, making it difficult to extrapolate to the regional scale. Fractures tend to follow a lognormal distribution in which well-developed, extensive, and often highly conductive fractures are spaced widely apart, with smaller, less extensive fractures becoming progressively more abundant. On the smaller scales, these are not always interconnected and hence may be nonconductive. Definition of rock-mass permeability of fractured systems is difficult, and factors that influence radionuclide transport processes are at least as complex.

Predicting regional ground-water flow and radionuclide transport behavior with the reliability and detail required for repository performance assessment is a complex and formidable challenge. It has been a main focus of international research related to repository siting and design for more than 25 years. Although substantial advances have been made—with project scientists and consultants making notable contributions—significant problems remain. Each potential repository site has unique features. To the extent that long ground-water travel times and significant physical and chemical retardation are necessary to ensure adequate isolation of radionuclides from the accessible environment, characterization of the site is correspondingly necessary to address these problems.

Most repository sites worldwide are located in rock that is part of an active regional ground-water regime. In these cases, except for engineered barriers in the immediate vicinity of the waste, the repository is physically close to circulating ground-water. By contrast, locating WIPP centrally within the 400-m-thick, essentially impermeable, Salado bedded salt deposit, some 200 m below the permeable, saturated sedimentary Rustler Formation, which consists of fractured limestones and dolomites, is a major design advantage of this repository. The restrictions to ground-water transport provided by the Salado Formation provide a valuable "buffer zone" between the repository and the non-Salado permeable formations above it, to which the site hydrology considerations of this chapter apply.