Technical Report NTB 00-09

Grimsel Test SiteInvestigation Phase IV (1994 – 1996):The Nagra-JNC in situ study of safety relevant radionuclide retardation in fractured crystalline rockI: Radionuclide Migration Experiment – Overview 1990-1996

The Nagra-JNC Radionuclide Migration Experiment (MI) is one of several investigations into the behaviour of radionuclides in the geosphere carried out at the Grimsel Test Site. The aims of MI were the development of methodologies for site characterisation, the study of the hydrology and geochemistry of fractured rock, the focussing of laboratory, field and modelling studies on the detailed characterisation of a single water-conducting feature, and the testing of models of radionuclide transport. The present report summarises the results of MI, updating an earlier overview report, provides a guide to detailed technical reports describing specific aspects of MI, and discusses the contribution of MI to building confidence in radionuclide transport models, including the applicability of these models to the performance assessment of radioactive waste repositories.

The central component of MI is a series of radionuclide transport tests performed in a single, approximately planar shear zone, in which well-defined dipole flow fields could be generated by injecting water at one location and extracting it at another. Tests were performed in flow fields of various sizes, strengths and locations and with a range of sorbing and non-sorbing radionuclides. In the course of carrying out the tests, a number of improvements were made to the test equipment, enhancing its reliability and minimising experimental artefacts that might affect tracer breakthrough curves. Accuracy of the measured breakthrough curves is particularly important since the effects of various retardation processes are apparent in the form of the curves only if they are measured for time periods that are several orders of magnitude longer, and for concentrations that are orders of magnitude lower, than the breakthrough peak. 

In order to support the development and parameterisation of transport models, a detailed geological, geochemical and hydrogeological characterisation of the MI shear zone was carried out. The geological characterisation included an evaluation of the small- to micro-scale structure of the shear zone and the distribution of both flow and rock matrix porosity, which provide direct input to the development of transport models, as does the hydrogeological characterisation. Furthermore, natural decay-series isotope profiles perpendicular to the MI shear zone provide unambiguous evidence of connected, diffusion-accessible porosity extending for some centimetres into the rock matrix. 

A series of laboratory and field experiments was performed to support the selection of sorbing tracers for the radionuclide transport tests, to determine the sorption properties of these tracers in laboratory conditions and to predict the in situ sorption properties of these tracers in the MI shear zone. The laboratory experiments centred on rock-water interaction tests and batchsorption measurements on material from the shear zone (crushed mylonite). They included the evaluation of sorption kinetics, reversibility and concentration-dependence ("non-linearity"). A field hydrogeochemical equilibration experiment was also carried out and yielded further information on sorption properties. Three sorbing radionuclides were identified for transport tests: weakly-sorbing sodium and moderately-sorbing strontium, both of which sorb linearly, and more strongly-sorbing caesium, which displays non-linear sorption. A mechanistic sorption model was used to predict the in situ sorption behaviour of caesium, in order to take account of competition with potassium, which is present at lower concentrations in natural Grimsel groundwater than in the laboratory experiments. 

Three different models were developed, and applied to MI breakthrough curves, by the Paul Scherrer Institute (PSI) and by the Federal Institute of Technology (ETH) in Switzerland, and by JNC in Japan. All of the models assume the dominant mechanisms for radionuclide transport to be advection and dispersion in the MI shear zone, with retardation due to matrix diffusion into stagnant pore water (largely within the fault gouge) and, for sorbing tracers, sorption on mineral surfaces. The conceptual differences between the models were relatively minor. The JNC model, unlike the PSI and ETH models, took into account transmissivity heterogeneity in the shear zone, as derived from pumping tests. The ETH model, unlike the other two, incorporated transverse, as well as longitudinal dispersion. The models also differed in the manner in which the dipole flow field was discretised in order to solve the governing equations for transport. 

All of the models were tested in terms of their ability to fit break-through curves and the consistency of the fitted parameters with what is known about the system. Furthermore, in order to maximise the contribution of MI to confidence building, emphasis was, where possible, placed on predictive model testing. Model parameters were derived from independent observations (e.g. sorption parameters inferred from laboratory experiments) and by inverse modelling (fitting) of the breakthrough curve for a non-sorbing tracer. Predictions were then made of the break-through curves for sorbing tracers (in advance of each test). 

The PSI model was applied to a particularly wide range of tracer tests, in both inverse and predictive modelling exercises, including those with a shorter dipole flow field, where the different transport mechanisms are weighted differently. The success of these exercises for the majority of tests not only supports concepts underlying the model, indicating that no relevant processes have been overlooked, but also indicates that, for tracers that sorb rapidly and exhibit reversible cation exchange, the results of laboratory experiments can be extrapolated reasonably well to in situ conditions. Adequate care must, however, be taken in selecting and preparing rock samples, so as to ensure that they properly reflect the geological character of the site. There are indications that sorption kinetics may influence the breakthrough curve of caesium, which has slower sorption kinetics, in the case of the shorter dipole flow field. Such effects are not, however, relevant to the much longer spatial and temporal scales of performance assessment. 

There are limitations to the degree to which the success of modelling breakthrough curves may be said to "validate" the models, in the sense of building confidence in their applicability in performance assessment. No information is provided, for example, on processes that, though irrelevant on the spatial and temporal scales of the transport tests, may be important over the scales relevant to performance assessment. Thus, the success of the predictions of the PSI, JNC and ETH models does not necessarily mean that the differences between the models are insignificant if applied in performance assessment and these differences may require further consideration. 

Overall, MI has demonstrated that the methodology adopted for the characterisation of waterconducting features, the simplification of this characterisation for modelling purposes, the adaptation of laboratory data (particularly sorption data) to field conditions, the selection of transport-relevant processes and the numerical solution of the governing equations, is applicable to the modelling of solute transport through a fractured crystalline rock. Apart from the technical achievements (and limitations) noted above, the participation in MI has resulted in the development of a culture of rigorous model testing and, arguably more importantly, of predictive model testing in the organisations involved. Furthermore, experience of working in a multi-disciplinary team on a long-term, challenging project has been considered to be of value to all participants, establishing successful communications between geoscientists, laboratory and field experimenters and modellers.