Representing small-scale heterogeneity in simulated radionuclide transport through brittle fault zones

Several new Discrete Fracture Network (DFN) modelling methodologies have recently been implemented in groundwater flow and transport models supporting upcoming Safety Assessment (SA) studies for the final disposal of spent nuclear fuel in Sweden and Finland. One improvement is to the representation of brittle fault zones (BFZs), zones of damaged rock with increased fracturing caused by brittle deformation around a fault core. Where earlier models described BFZs as simple planar features, they are now represented explicitly as finite-thickness swarms of fractures stochastically generated around each fault core. This increased realism is a major advance, as BFZs are a key factor in determining site-scale groundwater flow and radionuclide transport pathways.

However, the large number of swarm fractures in this representation carry a high computational cost, which can make it challenging to carry out all the transport calculations required in a typical SA study. To reduce this burden, codes such as ConnectFlow and MARFA provide a range of upscaling and downscaling methods. It is desirable that these methods preserve the small-scale heterogeneity structure of the explicit DFN model, as pathline results are sensitive to the local variability of velocity which this provides. Upscaling of DFN flow and transport properties to an Equivalent Continuous Porous Medium (ECPM) model is a common approach, but the resulting flow field can be smoothed out, affecting simulated radionuclide transport and retention. In a further simplification, ECPM properties can be projected back onto the BFZ midplane, producing a planar representation of the BFZ which retains some variability of flow properties. A downscaling algorithm allows the heterogeneity lost in each of these steps to be recovered by sampling from a library of DFN transport paths.

This study investigates these methods, applied to models of increasing levels of complexity and realism, as well as differing degrees of directionality and choking of flow. Groundwater flow and particle transport are simulated in ConnectFlow for each representation and the transport statistics of the resulting pathlines are compared. MARFA transport calculations are then used to investigate the degree to which the heterogeneity lost by upscaling can be recovered by downscaling.