TL: SAFETY ASSESSMENT MODELLING - A REALISTIC OPINION [Radioactive Waste ] (GP) SO: Phil Richardson BSc CGeol EGeol FGS, Greenpeace International DT: March 1, 1991 Keywords: nuclear power iaea radioactive waste greenpeace gp reports / The purpose of the International Atomic Energy Agency/Nuclear Energy Agency (of the OECD)/Commission of the European Community (IAEA/NEA/CEC) Joint Collective Opinion is to show that the currently available methodologies for safety assessment of radio- active waste repositories demonstrate the reliability of the process. It is based upon the data presented at an IAEA/NEA/CEC Joint Symposium in Paris, in October 1989. If the actual data and opinions presented at that symposium are examined, such a situation can be seen to be far from fact. Whereas the claim is made that the models and scenarios used in safety assessment are well proven and soundly based, many of the participants expressed considerable doubts about the overall validity and applicability of the research thus far undertaken. The collective opinion is based on the summary of the conference and not on a detailed or unbiased examination of the data presented. This background paper gives an alternative view of the conclusions of the symposium, and hence criticises the bland and unrealistic tone of the opinion. It is useful to concentrate on a few specific areas where criticism is most relevant. 1. Scenario development This is concerned with defining the range of possible futures to be considered in a performance assessment. It was discussed at some length in Paris. It is claimed that scenario methodologies have been greatly improved over the last few years, that only a few new phenomena have been identified and that `human imagination and scientific judgement coupled with existing knowledge of natural systems' form the basis of the process. In fact, Schifferstein (GRS, Germany) actually suggested that long-term predictions were worthless, based on our present levels of knowledge of natural systems. Only results up to 10,000 years should be given, to avoid public ridicule, an understandable fear. The difficulties with even being able to erect scenarios for events far into the future were described by Johansson (National Institute for Radiation Protection, Sweden). The need to simplify models and scenarios just to be able to perform the calculations, was a very important factor in increasing uncertainties, he said, and there is wide scope for error at all stages of the process. He concluded that the ability to predict the future is very limited, and workers must be seen to base their predictions on sound science, and not science fiction. In discussion, Chapman (British Geological Survey) concluded that a major problem still exists in terms of moving from the definition of possible futures to the definition of specific calculations where the models and data are defined. This uncertainty in application was repeated by Peaudecerf (Bureau de recherches geologiques mineress, France). He showed that numerous scenarios for the future existed, but that they are very difficult to erect, with always the possibility that important factors have been missed out, and that many of the parameters necessary have not yet been determined. The laws are often poorly understood, he said, the results are often extremely costly to obtain and even then can be virtually incomprehensible. It has been claimed that an increasing effort has been made to develop a capability to model environmental effects such as glaciation. The symposium was told by Boulton (Dept Geology, Edinburgh University) that a glacial event is likely within the next 60,000 years, which could have severe consequences for a repository, due to increased pore-water pressures and altered site hydrogeological conditions. Estimates of likely effects on a repository have been included in the TIME4 model, used in the UK, and which is based on reconstruction of conditions during the last Ice Age. According to Boulton, however, it has not been possible to include the likely effects of global warming in the model. Ringrose (Dames and Moore Consultants, UK), in fact, considered that scenarios actually have `no useful contribution to make to a safety assessment', as they are generally merely science fiction. Stephens (AECL, Canada) described how up to 1000 possible futures were originally listed for a Canadian assessment, which have been amalgamated, screened and finally linked to produce just 3 alternative scenarios. However, due to uncertainties in data, scenarios involving gas generation were not even considered, and glacial effects were ignored as they are `unlikely to manifest themselves within the regulatory timescale.' In conclusion then, attempts are being made to produce lists of potential futures for repository sites, in terms of natural processes and events. It is clear that the modellers feel that this will provide `numbers' which they can feed into their computers, whereas those who have to quantify the natural processes and set constraints on probability feel far from confident. It is still far from clear in many people's minds how to set relative probabilities on future events over the timescales envisaged, bearing in mind our short historical database, and our lack of knowledge of many of the underlying natural cycles which govern the future. It is also still uncertain as to how to interpret the results which are obtained. An area described as receiving special attention recently is assessment of human intrusion. Reporting back on an international conference on the subject, Thegerstrom (NEA) stressed the limitations of the study, bearing in mind the difficulty in predicting the needs and activities of society thousands of years into the future. 2. Modelling The papers here were described as `an up-to-date overview of the current state-of-the-art in key modelling areas.' McCombie (Nagra, Switzerland), pointed out, however, that models are not currently available for many processes which are `thought' to be important. eg 2-phase flow colloidal flow transient flow flow in heterogeneous media (ie most host rocks) many aspects of near-field performance In many cases, model validation is very hard to carry out, and mostly consists of laboratory experiments and the use of analogues. As regards the future, he accepted that `some' models still need to be developed, and more validation will mean more extensive data gathering. He stressed the need to recognise the limitations of the models being used. Parker (BNFL, UK) repeated the statement that many key uncertainties have been recognised - including gas release, hydrogeology and the sorptive capacity of the `soil'. Indeed, `the more you look, the more you need to look' he said. Grambow (Hahn-Meitner Institut, Berlin, Germany), referring to the wasteform itself, pointed out the need for further research into the processes of degradation of glass etc. At present, he said, models can only predict processes which are already known, and duplicate experimental results already obtained. Carrera (Barcelona Univeristy, Spain) made the point that a certain degree of uncertainty is inherent in any model, and `no model can ever be proved right or wrong', it is merely a theory. This inherent uncertainty was also brought out by Cacas (Ecole nacionale superieure des mines de Paris, France). Models are already extremely complex, albeit with limited numerical accuracy. Major assumptions must be made about complex natural systems to enable the models to work at all, and there is, he said, a great need to design validation exercises for use in the field, so as not to just use simple experimental results. It is obvious that although the mathematical solutions are elegant in their complexity, knowledge in the areas of diffusion, thermal effects, sorption etc. is still lacking, and yet modellers are trying to come up with results for flow times that can be input directly into safety assessments. Neretnieks (Royal Institute of Technology, Sweden) pointed out that many of the uncertainties which still exist are concen- trated in specific subject areas, and their relative importances are still unclear. Geochemical models were reviewed by Sargent (AECL, Canada). When data and research needs were reviewed in 1983, areas of concern involved sorption, thermodynamic models etc. These are still valid today, with only limited site-specific data still available in many of these areas. Little or no consideration has thus far been given to bacterial action and its effects on the evolving chemical environment. This lack of adequate data was repeated by Carnahan (Lawrence Berkeley Laboratory, USA) who also pointed out that few of the models were actually linked to other models, and that the development of coupled models ought to be a high priority activity. Lack of adequate field validation was also pointed out as a weak area. According to de Marsily (Ecole nacionale superieure des mines de Paris, France), the problems inherent in predicting repository performance using models include incorrect modelling of the source term, use of the wrong model and input of incorrect parameters. He stressed the need to apply the models to real natural situations to check their performance. He also made the point that although models of highly permeable media exist in varying forms of complexity, very few have as yet been developed for low permeability rocks which are in fact most likely to enclose any repository. To claim that models exist in a sufficient form to allow confident predictions into the relative far future is extremely misleading. Unfortunately, many individual countries are claiming just this, and are using the models to justify repository siting studies. In the UK, for example, NIREX intend to justify a safety case based on an ongoing research programme described by even the IAEA as being in `a preliminary stage of development.' This is despite Laurens' (Electrowatt Consultants, UK) statement that the UK had expertise in sedimentary sites, but as both the potential sites are in crystalline rocks, the models have had to be improved to include fractured rocks. This work is still not complete, and even then will not take into account marine or coastal processes, gas migration, colloidal flow, the influence of organics in the repository or numerous engineering related aspects. Modelling of the biosphere was only directly dealt with by one paper. Smith (Intera, UK) compared and contrasted the number and scope of models available for geosphere and biosphere, concluding that considerably fewer exist for the biosphere, and that they are and must remain relatively simple, due to uncertainties in predicting long-term biosphere evolution. `Less difficult to justify' models may prove useful in future safety assessments. As Saltelli (CEC Ispra) pointed out, `most codes are still at the stage of research tools'. 3. Confidence Building An important part of efforts to convince the public of the case for deep disposal is concerned with ensuring that everybody involved in the decision making process has confidence in the information used and results obtained. Not least amongst the techniques employed to build this confidence is model validation, ensuring that the models adequately represent real system behaviour. Tsang (Lawrence Berkeley Lab., USA) stressed the need for a parallel development of field testing techniques, as no validation could be usable if these were shown to be insufficiently accurate. It is important to mention here that in Luxembourg last September, a conference was held to review the progress made in the CEC's 5 Year Research Programme on radioactive waste. Despite initial statements from Rometsch (Nagra, Switzerland) claiming that all the problems had been solved, it was obvious that much research and development work still remains to be done, a fact evidenced by the continuance of the Programme for at least another 5 years. An important conclusion from Luxembourg was that much of the so- called `proven technology' for site investigation and other related activities has either yet to be developed on an industrial scale or has yet to be shown to work successfully. International co-operation exercises are often claimed to have demonstrated the validity of models. Most however merely show that they can handle data from other projects, with varying degrees of success. Grundfelt (Kemakta Consultants, Sweden) dealt with HYDROCOIN, which although completed in 1987, is still only at the reporting stage. It appears that much of the work of stage 2 of the exercise actually involved getting the model output to match the experimental data supplied, which proved almost impossible in the case of a low-permeability medium. It is clear that claims to have validated codes using HYDROCOIN are not correct, and a significant conclusion expressed in the pape r isthat `the simpler the system being studied is, the better the agreement between model predictions and observations'. The follow-up exercise to HYDROCOIN is INTRAVAL, which includes natural analogue studies and more complex test-cases for geosphere models. The 14 test-cases were described by Andersson (Swedish Nuclear Power Inspectorate) and it is obvious that this is a long term project which is still only in its early stages. [The codes being assessed are, however, claimed by the relevant national bodies to be validated simply because the exercise is in progress]. Andersson stressed the fact that the project is only `an important step in the development of validation strategies'. To claim more is very misleading. CHEMVAL is more specifically an exercise dealing with models of chemical conditions in the near- field and geosphere, and will be completed by early 1990. However, as Broyd (W S Atkins Consultants, UK) pointed out, although initial findings are encouraging, they cannot be applied to the field situation yet. Numerous generalisations, simplifications and downright omissions have been made, with the roles of colloids, solid solution series, rection kinetics, sorption and organic ligands still open to interpretation. Great claims have been made following the completion of the CEC PAGIS project. This was a very useful survey of available methodology and modelling capabilities. However, it was of course media-specific, not site-specific in nature. Although the results show no significant doses up to 10,000 years post-closure, it should be realised that the `normal' scenario used to predict those doses assumed that no water ingress was possible. Another major assumption was that the mathematical codes used in the exercise were all correct, perfectly validated and totally verified. Much remains to be done, but `given the uncertainties inherent in natural system studies, complete validation of computational methods can rarely, if ever, be achieved.' CONCLUSIONS 1. Scenario development Our ability to model the future is severely limited, given the major uncertainties inherent in modelling the natural world. It is unclear as to how to apply even the limited knowledge that we do have. Efforts to model the effects of major climate change have only just begun. Some workers are not even sure that the application of `science fiction' has any role in safety assessment. Models do not yet exist for many areas of potential importance. For most models to work at all, major assumptions and generalisations need to be made. It is not even clear yet as to how significant uncertainties in model input parameters are. Very few integrated models have yet been developed, such is the complexity of the issues involved. Models of low-permeability, heterogeneous and fractured media are still in their infancy. Models of gas migration, colloidal flow and the influence of organics on near-field chemistry are still under development. Biosphere models are even further behind. 3. Confidence building International validation exercises have thus far only managed to handle relatively simple data. Numerous generalisations, simplifications and downright omissions have had to be made, with the roles of colloids, solid solution series, rection kinetics, sorption and organic ligands still open to interpretation. ---------- =======#=======