Discrete fracture network (DFN) models provide a natural analysis framework for rock conditions where flow is predominately through a series of connected discrete features. Mechanistic models to predict the structural patterns of networks are generally intractable due to inherent uncertainties (e.g. deformation history) and as such fracture characterisation typically involves empirical descriptions of fracture statistics for location, intensity, orientation, size, aperture etc. from analyses of field data. These DFN models are used to make probabilistic predictions of likely flow or solute transport conditions for a range of applications in underground resource and construction projects. However, there are many instances when the volumes in which predictions are most valuable are close to data sources. For example, in the disposal of hazardous materials such as radioactive waste, accurate predictions of flow-rates and network connectivity around disposal areas are required for long-term safety evaluation. The problem at hand is thus: how can probabilistic predictions be conditioned on local-scale measurements? This presentation demonstrates conditioning of a DFN model based on the current structural and hydraulic characterisation of the Demonstration Area at the ONKALO underground research facility. The conditioned realisations honour (to a required level of similarity) the locations, orientations and trace lengths of fractures mapped on the surfaces of the nearby ONKALO tunnels and pilot drillholes. Other data used as constraints include measurements from hydraulic injection tests performed in pilot drillholes and inflows to the subsequently reamed experimental deposition holes. Numerical simulations using this suite of conditioned DFN models provides a series of prediction-outcome exercises detailing the reliability of the DFN model to make local-scale predictions of measured geometric and hydraulic properties of the fracture system; and provides an understanding of the reduction in uncertainty in model predictions for conditioned DFN models honouring different aspects of this data.

Modelling physical processes such as flow and transport within a fracture network can be challenging, both conceptually and numerically. A typical approach is to upscale the properties of the network onto a regular grid of elements, which is then used to simulate the required physical processes. However, this method can be inaccurate if not carefully applied. For example, when most of the flow passes through a few choke points in the network, upscaling may introduce extra connectivity that is not present in the original network. The ConnectFlow software package has an alternative method for representing fracture networks, whereby the fractures are explicitly modelled as a network of intersecting two dimensional planes. The algorithms within ConnectFlow are very efficient, allowing millions of separate fractures to be simulated, each discretised using hundreds of finite elements. Here we present recent updates to this functionality: 1) to allow the advection-diffusion equation to be solved for multiple solute species (which is fully coupled to the pressure solution via the buoyancy term in Darcy’s equation); 2) to model the diffusion of solutes into the pore space of the surrounding rock (i.e. matrix diffusion); and 3) to carry out chemical reactions between solutes and minerals (which coat the fractures and/or the rock pores) using an interface to the IPhreeqc library. The ConnectFlow algorithms have also been parallelised to improve the tractability of this new functionality. This implementation represents a significant step forward in capability that allows groundwater flow, transport and hydrogeochemical reactions to be properly represented in the context of structurally constrained fractured bedrock. This has a wide range of potential applications, particularly for future safety assessments of nuclear waste facilities. Example calculations are presented for the Onkalo disposal facility in Olkiluoto, Finland, and the proposed Swedish repository for long-lived waste, SFL.