MODELING FIELD DIFFUSION EXPERIMENTS IN CLAY ROCK: INFLUENCE OF NUMERICAL REPRESENTATION OF BOREHOLE AND ROCK INTERFACE

Argillaceous rocks are considered in Switzerland and in other countries as potential host rocks for the deep geologic disposal of radioactive waste. Opalinus Clay, a formation of Jurassic age, is at present the most favored candidate in Switzerland. It has a low hydraulic conductivity, no or only few active fractures, and a large retention capacity for sorbing solutes, which make this sediment well suited as an additional barrier for the spreading of contaminants. Transport through Opalinus Clay is typically dominated by diffusion. Accordingly, the diffusion of tracers through this material is intensively studied. In the underground research facility at Mont Terri in Switzerland, field experiments are performed that aim at investigating the diffusion and retention of sorbing tracers at intermediate scales and under relevant in-situ conditions. The field experiments use a borehole, from which tracers diffuse into the surrounding rock. The tracer cocktail in the borehole is continuously circulated, which allows to monitor the tracer decrease over time. When modeling these experiments, care has to be taken to correctly represent the inlet system. In this paper, it is shown how the numerical representation of the borehole and inlet system and the spatial discretization affect the simulation results of mobile and sorbing tracers. For mobile tracers, it is generally sufficient to use an effective diffusion coefficient for the circulated fluid about 30 times larger than that in the rock to mimic the continuous mixing. In contrast, for sorbing tracers even a 7000 times larger diffusion coefficient may, at early times, not lead to homogeneous borehole concentrations. This is because the equilibrium sorption on the rock quickly and drastically reduces the tracer concentrations at the interface. As a consequence, the simulated flux into the rock becomes too small, and the calculated average decrease of the borehole concentration is much too slow even for larger times. The slow decrease can be similar to that simulated for a much less sorbing solute, which of course would critically affect parameter estimation when fitting the model to observed data. Increasing the borehole diffusion coefficient to very large values to obtain a complete mixing can lead to a too fast simulated concentration decrease, if the spatial discretization is insufficient. Thus, a careful checking of the numerical results is required for the strongly sorbing tracers.