3.7.1 Safety assessment methodology
The IRT finds that the safety assessment methodology developed by SKB relies upon long national
experience and participation in several international working groups. The safety assessment is on par with
the international state-of-the-art and is coherent with the guidelines established by OECD/NEA through
long-lasting international cooperation.
The FEPs data base completes the previous SR-CAN 97 FEP catalogue with components not treated in SR-
CAN. Completeness was assessed by comparison/mapping with other national database and the
international NEA database. The fact that this FEP database is available on the web argues forcefully for
the transparency and traceability of the safety case. The links between FEPs and long-term processes are
clearly exposed.
The IRT considers that the methodology and safety demonstration takes the safety functions into account
in an effective way. The IRT recommends that SKB considers defining, before using the FEP chart,
intermediate-level safety functions that are not directly linked with the nature of the components and the
processes involved. Using these intermediate-level safety functions could be of some use with respect to
best available techniques (BAT), i.e., alternative ways of fulfilling a given safety function could be
explored.
The IRT considers that SKB‘s scenario selection methodology, which is based on safety functions, is on
par with the international state-of-the-art, i.e., it is described precisely and justified. The scenario analysis
is clearly explained and described and appears to be sound.
The IRT considers that the dose calculations based on the technical conclusions drawn from scenarios
analyses are sound. However, the IRT notes that the risk summation curve presents pulse-like features and
this is actually an artefact due to the small number of realisations leading to canister failure. Although this
does not have any consequence for the demonstration of risk compliance, the IRT is of the opinion that it is
a weakness with respect to the clarity of the presentation of the results. The IRT recommends that, to
remedy this, SKB refines the realisations of the probabilistic underlying models that control the points in
time of canister failure.
The IRT notes that SKB‘s safety assessment includes dose calculations to non-human biota, which is
consistent with regards to best practices.
SKB does not draw any strong conclusions from the assessment of future human actions. Nevertheless,
the IRT considers it beneficial to assess the sensitivity of the results to the pessimistic assumptions made
and to draw conclusions in the view of BAT that can reduce the risk, e.g., long-term memory, and the need
for its development.
The IRT considers SKB‘s use of stylized scenarios (hypothetical cases where different barriers are
assumed to be completely lost) as a good practice to build confidence in the safety case. These scenarios
do not represent any physical reality. If used cautiously, however, and with full understanding of their
limitations, they can give an illustrative view of the respective contributions of each of the barriers to
safety and provide a check for the robustness of the design.
3.7.2 Credibility of scenarios
The IRT is of the opinion that from a methodological point of view, SKB‘s scenario selection is on par
with international best practices.The IRT notes further that the general assumption of the repeatability of
future glacial cycles is common to performance assessments.
Regardless of the high confidence SKB has expressed, the IRT recommends that sensitivity of the
repository system to climate effects should be subject to further observation and assessment with respect to
the increasing state of knowledge about potential water pathways and rock properties between surface and
repository depth that will be developed during the coming project phases. The IRT recommends that future
safety assessments should address this issue and provide updated information on whether the originally
assumed favourable conditions are still valid. This would contribute to a higher degree of confidence in the
independence of the repository evolution from climate-related effects.
The IRT concludes that, with respect to the scenario of isostatic load leading to canister failure, the
conceptualization of the maximum hydrostatic load is correct, although the estimated values for the various
pressures that contribute to the isostatic load seem to be conservative.
The IRT also concludes that, with regard to the scenario of shear load leading to canister failure, fault
reactivation from pore pressure drawdown induced by the repository excavations will most probably not
lead to critical fault reactivations. However, SKB‘s assessment of the most critical earthquake triggering
shear failure scenario (seismic slip > 5 cm) is very complex, and based on a large number of assumptions.
Nevertheless, the IRT, although not staffed with a seismologist, supports the discussion on the probability
of future large earthquakes (as described in TR10-01, chapter 10.4.5) and the corresponding scenario
credibility.
3.7.3 Corrosion failure
The analysis of the corrosion failure scenario highlights the main factors of uncertainty in this scenario,
namely:
Uncertainty related to the DFN model
Uncertainty related to the sulphide concentration
Questions regarding the potential for that copper may corrode in pure water under exclusion of
oxygen and with production of hydrogen.
Several alternative DFN models are presented and assessed in SR-site. The selection of the correlation
between fracture size and transmissivity on one hand and the chosen realization of the model on the other
hand, have a strong impact on the number of advective positions, on the number of failed canister and on
the migration of released radionuclides. In order to capture this variability, SKB builds its assessment on a
small number of realizations of the DFN models. SKB considers that this statistical treatment is sufficient
to build confidence in the calculated mean number of failed canisters. Given the pessimistic assumptions
adopted by SKB, the order of magnitude used for the risk summation seems convincing to the IRT, and
hence also the fulfillment of the risk criterion.
However, the IRT still considers that the number of realizations is rather low with regard to the criticality
of the issue of the hydrogeological model. Moreover, SKB does not justify clearly enough that the
variability captured in these realizations corresponds to a rigorous confidence interval.
Sulphide concentration is a key factor for corrosion. Only the extreme value of the discrete sulphide
concentration leads to a canister failure for most of the DFN models. Thus this end tail of the distribution is
an important part of the distribution for the risk calculation, although its description is very rough, and is
represented by only one data point. A more rigorous statistical analysis of the confidence that can be put in
the sulphide distribution than the one SKB conducted in its sensitivity analysis would then be of interest.
However, the IRT recognizes that it is likely that the distribution used in SKB‘s calculations is a
pessimistic one.
As discussed in Section 3.3 on the process of corrosion of copper by water, SKB‘s calculations in TR-10-
66 are defensible as long as hydrogen evolution would come from the anoxic corrosion of copper with two
copper atoms oxidised for each hydrogen molecule (H
2) produced. If, however, other reactions or
stoichiometries are at play, SKB‘s calculations do not apply and this could lead to new questions about the
long-term durability of the copper canister. The IRT notes that the discussion of copper corrosion by
hydrogen evolution is important and must be resolved.
The IRT strongly recommends that, as soon as it can reduce uncertainties through further data acquisition,
SKB should reassess the central corrosion scenario and its risk contribution, which could vary in a non-
negligible range. Given the pessimistic assumptions taken by SKB in the safety case, this should lead to a
reduction of the calculated risk, and thus should not jeopardize risk compliance.
3.7.4 Shear failure
The IRT notes that the observation periods used in the determination of frequency-magnitude relationships
are relatively short (100-1000 years) in comparison to the duration of the safety assessment. Therefore
SKB systematically discussed the uncertainties of long term frequency relationships and included
paleoseismic frequency indicators as derived from glacially-induced faulting. Although the IRT supports
the discussion on the probability of future large earthquakes (as described in TR10-01, chapter 10.4.5) and
the corresponding scenario credibility, the IRT recommends additional studies on near-field conditions and
paleoseismic investigations to augment the understanding of the impacts of long-term seismicity at
Forsmark.
3.7.5 Evaluation of the Total System Performance Assessment Results
With the exception of relevant issues affecting overall repository performance discussed earlier in this
report, and those identified by SKB itself, the IRT finds SKB‘s approach to estimating repository
performance to be ―conservative‖.
Many of the properties of the geology and repository design are uncertain over the long assessment period.
SKB has assumed bounding values for many of these properties. This is in accord with standard
international practice to use bounding, or near-bounding properties when the actual distribution or
uncertainty range of geology and repository properties are unknown. This does not mean the IRT is certain
SKB has overestimated the buffer and canister failure rates or the dose rates to humans. Both the IRT and
SKB recognize that more work needs to be done to improve confidence of the repository system
performance.
The IRT notes that the understanding of the repository system based on the quantitative performance
assessment results and their uncertainties could change if a ―best estimate‖ approach were taken. The IRT
recognizes that SKB has addressed this in its responses to the IRT, and recommends that SKB provide an
expanded discussion of insights on how a ―best estimate‖ approach would change not only their
quantitative results, but more fundamentally, the understanding of the entire repository system.
In document
Revista ILLAPA nº 7, julio 2010.
(página 113-118)