Box 6.2. (Continued)

In addition to the direct processes of advection, dispersion and diffusion, a range of coupled processes, including thermal, chemical and electrical osmosis, thermal diffusion, hyperfiltration and electrophoresis, can transport porewater and radionuclides in solution in response to gradients in temperature, pressure, solute concentration and electrical potential. Although such coupling is generally negligible for most practical applications, some of these processes, including chemical osmosis and hyperfiltration, can be significant in microporous media such as some argillaceous sediments.

Where there is significant groundwater flow, advection and mechanical dispersion are often far more effective transport processes than diffusion or the coupled processes mentioned above. This does not mean, however, that these processes can necessarily be neglected. In many media, groundwater flow is predominantly confined to discrete water-conducting features, such as fractures, sand lenses, thrust planes, alluvial or glacial channels and rock unconformities. Elsewhere in the rock body, groundwater may effectively be stagnant. In such cases, diffusion may, e.g., transport radionuclides from waterconducting features into stagnant porewater ‘‘matrix’’ regions and vice versa. This process of matrix diffusion (see Fig. a) is referred to as a retardation process, since it results in slower transport. In some media, matrix pores may be accessible to solutes by diffusion, but larger molecules, ions and colloids may be excluded due to size and/or charge effects (Fig. b). The retarding effect of matrix diffusion may thus apply to radionuclides associated with solutes, but not to those associated with colloids (although colloids may be retarded or immobilised in other ways, such as by filtration).

evaluations of, for example, the actual future release of radioactivity from the repository, the actual rate at which radioactivity from the repository enters the biosphere or radiation doses received by human populations living in the future. In fact, doses and risks calculated on the basis of stylised approaches and simplified models should be interpreted as illustrations based on agreed sets of assumptions for particular scenarios and well-defined, but not necessarily realistic, model assumptions, and not as actual measures of future health detriments and risks (ICRP, 2000).

6.4. Timescales for evaluation

As discussed earlier, the evolution of a repository and its environment is always subject to some uncertainties. These uncertainties tend to increase the further into the future that the assessment must consider and affect different components of the repository and its surroundings in different ways. Figure 6.8 illustrates how increasing uncertainty affects the extent to which the future behaviour of the different components of the disposal system considered in the Belgian SAPHIR 2 SA (ONDRAF, 2001a,b) could be reliably assessed. The most stable and predictable component of a geological disposal system is generally considered to be the host rock itself (‘‘Clay’’ in Fig. 6.8, which considers the Boom Clay in Belgium as a potential host rock), but even this will be subject to poorly predictable changes over long enough timescales.

The questions then arise:

Is it necessary to evaluate the performance and safety of a repository at distant times when even the evolution of the geological environment, which is typically chosen, at least in part, for its long-term stability, may be subject to poorly predictable changes?

If so, how can safety be evaluated when assumptions, such as geological stability that typically underlie the models used to evaluate dose and risk, cease to be valid?

Fig. 6.8. The extent to which the future behaviour of the components of the disposal system considered in the Belgian SAPHIR 2 SA could be reliably assessed – ‘‘robustness’’ in the terminology of this SA (from Fig. 4.2 of ONDRAF, 2001a). Note: Radioactive decay and the biosphere are included in the figure, even though they are not normally classed as components of the disposal system.

These questions were addressed at a recent NEA workshop (NEA, 2003, 2004b) which concluded that there are no ethical arguments that justify imposing a definite limit to the period addressed by SAs, in spite of the technical difficulties that this can present to those conducting such assessments.

‘‘It is an ethical principle that the level of protection for humans and the environment that is applicable today should also be afforded to humans and the environment in the future, and this implies that the safety implications of a repository need to be assessed for as long as the waste presents a hazard. In view of the way in which uncertainties generally increase with time, or simply for practical reasons, some cut-off time is inevitably applied to calculations of dose or risk. There is, however, generally no cut-off time for the period to be addressed in some way in SA, which is seen as a wider activity involving the development of a range of arguments for safety’’ (NEA, 2004b).

Recent SAs have recognised the limitations of the models used to calculate dose and risk and impose ‘‘cut-off’’ times or use presentational techniques, so that calculations are not reported without suitable qualification when the underlying model assumptions may no longer hold. In the Swiss Project Opalinus Clay SA, e.g., the timescale beyond which significant geological changes cannot be ruled out is judged to be one million years. Beyond this time, doses calculated on the basis of an assumption of geological stability may not be meaningful. This is reflected in the dose curves presented in the safety report by using dark background shading for the time interval between 106 and 107 years, as illustrated in Fig. 6.9. Most calculations in this assessment are truncated at 107 years.

When considering times when calculated doses and risks may no longer be meaningful, some SAs, including the Belgian SAPHIR 2, the Swedish SR 97 and the Swiss

Fig. 6.10. Radiotoxicity of spent fuel as a function of time after discharge from the reactor for Swedish BWR fuel with a burnup of 38 MWd/tU. Radiotoxicity pertains to ingestion via food (from Fig. 1–2 of SKB, 1999).

The construction and use of safety cases have recently been discussed in an NEA brochure (NEA, 2004a). An initial safety case can be established early in the course of a repository project. A safety case will, however, generally become more comprehensive and rigorous as a programme progresses and as the depth of understanding and the technical information available increase. It is generally presented in the form of a structured set of documents that are tailored in their style of presentation to the intended audience (i.e., usually the regulators and the government) and is typically required at major decision-points in repository planning and implementation, including decisions that require the granting of licences.

Input to the safety case comes not only from the findings of SA, but also more directly from site selection and characterisation and design studies; in particular, evidence that the system has been well chosen, has a range of positive attributes that intrinsically favour safety and that no obviously better system exists. The favourable characteristics of the geosphere that can be cited in a safety case were discussed at a recent NEA workshop, the first in the AMIGO series (Approaches and Methods for Integrating Geologic Information in the Safety Cases, NEA, 2004c), which took the Swiss Project Opalinus Clay (Nagra, 2002a–c) as a case study. Examples of such characteristics from this study are given in Box 6.3.

Most national regulations give safety criteria in terms of dose and/or risk. The results of a SA may well indicate that such criteria are satisfied. The credibility of these results, however, depends on the reliability of the analyses and the adequate treatment of uncertainties (i.e., the adequacy of the range of scenarios, alternative conceptualisations and parameter variations considered). The reliability of the analyses, in turn, depends on the quality of the models, computer codes and databases used to analyse assessment cases and the adequacy of the QA procedures for performing SA calculations.

Factors also relevant to a safety case thus include, e.g., whether sensitivity analyses have been carried out to ensure that scenarios and calculational cases address key uncertainties affecting the performance of the disposal system, whether calculational

Box 6.3. Favourable characteristics of the geosphere that can be cited

in a safety case – examples from the Swiss Project Opalinus Clay (from NEA, 2004c).

Long-term geological stability, implying, e.g., a low-rate of uplift and erosion and insensitivity of the geochemical and hydrogeological environment to geological and climatic changes;

Favourable physical, chemical and structural properties, including thickness of the host formation, low rates of groundwater movement, a geochemical environment that is beneficial in terms of radionuclide retention and protection of the engineered barrier system, and rock mechanical properties that support the feasibility of construction (although not strictly part of the safety case, engineering feasibility is relevant in that the system described in the safety case must be one that can be realised in practice);

Sufficient lateral extent, which gives flexibility in the location and layout of the repository;

Absence of, low likelihood of, or insensitivity to detrimental phenomena and perturbations, including climatic and geological events and processes, perturbations caused by the repository itself (gases, chemical alterations), and future human intrusion;

Explorability, or the ability to characterise the rock at any stage of the project to a degree that is adequate to support a decision to proceed (or not) to the next stage (e.g., site characterisation from the surface can provide sufficient evidence to support the decision to proceed with further characterisation from underground tunnels); and

Predictability, meaning that the range of possible geological evolution scenarios is sufficiently limited over the timescale for which the geological environment plays a role in the safety case (perhaps, e.g., a million years).

results are in agreement with simplified calculations and understandable from a scientific perspective, whether models and databases are consistent with wide-ranging field, natural analogue, URL and laboratory evidence and with fundamental scientific principles, whether models and databases can be shown to err on the side of conservatism3, whether they have been subjected to peer review and whether the computer codes used have been developed and applied in the framework of a QA procedure and have been adequately verified and tested.

As mentioned earlier, at sufficiently distant times in the future, and particularly when the stability of the geological environment cannot be relied upon with confidence, the models used to calculate dose and risk may not be appropriate and arguments based on radioactive decay and the resulting decrease in the radiological hazard presented by the waste may receive more prominence in a safety case. In particular, the radiotoxicity of the waste at distant times in the future may be compared with that of naturally occurring mineral deposits and rocks (cf. Fig. 6.10). As discussed in NEA (2004b), although an evaluation of dose or risk may still be required by regulations, a less rigorous assessment of these indicators may well be acceptable on account of the decreased hazard potential.

3 This aspect of SAs is often difficult for many people to understand: an important premise of a SA is to show that, despite assuming the worse case scenario for each process or mechanism (e.g., ignoring geosphere retardation or irreversible sorption), a repository can be shown to satisfy the safety criteria set out by the regulatory authorities. Expressed in another way, the repository is deliberately ‘‘over-engineered’’ to provide large margins of safety (i.e., belt and braces).

also discussed at the recent AMIGO workshop. Examples from the case study, the Swiss Project Opalinus Clay, are given in Box 6.4.

A safety case may indicate what enhancements to the system itself or the models and databases used in SA may be considered eventually to make a case that is adequate for licensing purposes. In the Swiss Project Opalinus Clay SA, phenomena were identified that were omitted from the evaluation of assessment cases on the grounds that it was conservative to do so, but for which there were considered to be good prospects for improved scientific understanding, models and data, so that they might be included at a later stage of the repository programme. In the terminology of that assessment, these are ‘‘reserve FEPs’’, and their existence is said to constitute an additional argument for safety. Particularly when a SA is carried out in the early stages of a programme, there may well be a number of uncertainties with the potential, at least, to call safety into question, as well as open issues regarding, e.g., specific design options that will eventually be selected. It is often, however, still possible to make a safety case that is adequate to support the decision at hand, as long as these uncertainties and unresolved issues are acknowledged and a strategy is set out to address them (see Chapter 8).

A safety case needs to be presented in a style that is understandable and useful to its intended audience, taking account of their interests, concerns and level of technical knowledge. The audience may include the regulator, political decision-makers and the

Box 6.4. Key safety-relevant properties of the Opalinus Clay that are supported by multiple lines of evidence (from NEA, 2004c).

Low rate of uplift and erosion, consistent evidence for which comes from:

basin modelling (burial history) of the area of the proposed repository in northern Switzerland, which takes into account stratigraphic evidence, apatite fission track analysis, organic matter maturity studies and investigations of diagenetic cements and their fluid inclusions;

geomorphological studies of the elevation and age of river terraces in northern Switzerland, from which the rates of linear erosion since the time of deposition can be evaluated, as well as an evaluation of erosion rate from the assumption that the pre-glacial landscape was a peneplain whose elevation corresponds to present day hill and mountain peaks; and

geodetic studies using precision levelling, which is available over a period of almost 100 years, relative to a point where the underlying crystalline basement is exposed.

Low hydraulic conductivity and groundwater flow in the bulk rock, evidence for which comes from:

in situ and laboratory hydraulic testing;

tests for consistency with the porosity/conductivity relationship for clay formations investigated world-wide;

the existence of hydraulic overpressures, which are interpreted as relics of burial history or as a result of the compressive stress field, but can only be understood if the hydraulic conductivity is even smaller that those derived from hydraulic tests; and

concentration profiles of numerous elements and isotopes in porewater which suggest a diffusiondominated system.

Self-sealing capacity, evidence for which comes from:

in situ experiments of artificially induced fractures at the Mont Terri URL; and

the absence of mineral veins and alterations, suggesting that there was not significant water flow through natural discontinuities in the past.

public, as well as technical specialists advising external groups and organisations, or the personnel of the implementing organisation itself. Multiple levels of documentation may thus be required, ranging from detailed technical reports designed to record all key assumptions and data in a traceable manner to more accessible forms such as brochures and video presentations. As pointed out, e.g., in (NEA, 2004a) and Chapter 9, where the audience is primarily the general public, highlighting less quantitative evidence for safety, including evidence from natural analogues, may be more accessible, more convincing and of more interest than, say, the results of complex mathematical models.

To present the Japanese H12 SA in a way that makes the SA process clear and the implications of the results meaningful both to workers within the SA field and to a wider technical audience, a report complementing the main SA reports (JNC, 2000a–d) was produced that examines the aims, procedure and results of the assessment from a wider perspective (Neall and Smith, 2004). Similar reports have been produced for earlier SAs in Japan and Switzerland (e.g., Neall, 1994). In these reports, the reasonableness of the assessment results is argued, in part, by making comparisons with results from SAs conducted by other national programmes for systems that are in some way similar. As part of this comparison, Fig. 6.12 shows the calculated annual individual dose as functions of time for the Reference Cases of H12 and eight other HLW and SF assessments conducted internationally. Doses are compared with the range of natural radiation exposure in Japan (approximately 900–1200 mSv a 1) and to the range of regulatory guidelines in various countries (100–300 mSv a 1). The nuclides that contribute most to dose at different times are also indicated.

There are many detailed technical reasons that can be identified for the differences in the results, including differences in the waste forms, inventories, disposal concepts and the models and data used for analysis. Perhaps the most striking point about the figure is, however, how little the calculated dose maxima (although not their times of occurrence) vary between most of the assessments, given these often very significant differences. At least part of the reason for this may be related to the high reserves of safety that are, in reality, built into all of the repository concepts that are assessed. The models used for SA, on the other hand, are usually highly conservative, and disregard or simplify the treatment of many potentially favourable features and processes. It may be that, if it appears that an analysis will give results near to or exceeding regulatory guidelines, then effort is spent in developing and testing more realistic models and databases that reduce the level of conservatism, thereby reducing the calculated doses or risks to levels that are well below the relevant guidelines.

A final point to note is that the lowest dose maximum in Fig. 6.12 comes from the oldest of the SAs in the figure, the Swiss Projekt Gewa¨hr (Nagra, 1985). This assessment was conducted at a time when the host rock under consideration, the crystalline basement of northern Switzerland, was less well understood than it was for the more recent Kristallin-I SA (Nagra, 1994), which addressed the same host rock and a similar repository concept. This illustrates that, although the availability of information and understanding always increases over the course of a repository programme, the level of confidence in the performance of a system may go down as well as up. It may well be that early conceptual models disregard some important phenomena that are subsequently discovered in the course of more detailed study. This is the reason why repository programmes take a cautious, step-wise approach to planning and implementation, and flexibility, as well as high reserves of safety, are built into the strategies for siting and design.

Fig. 6.12. Calculated annual individual dose as a function of time for the Reference Cases of H12 (JNC, 2000a–c) and eight other HLW and SF assessments conducted internationally: SKB-91 (SKB, 1992), SITE-94 (SKI, 1996), TVO92 (Vieno et al., 1992), TILA-99 (Vieno and Nordman, 1999), AECL EIS (AECL, 1994), Project Gewa¨hr 1985 (Nagra, 1985), Kristallin-I (Nagra, 1994), H3 (PNC, 1992).