A Numerical Study of Channeling in Heterogeneous vs. Calibrated Homogeneous Discrete Fracture Network Realizations

It has been observed that groundwater flow through fractures is channelized due to variations in fracture aperture, fracture roughness and different fracture intersections. Nevertheless, a common practice in DFN modelling is to model each fracture as hydraulically homogeneous, which is an obvious simplification of reality. In our study, we have evaluated whether a flow-calibrated DFN model, where each fracture is assumed hydraulically homogeneous, can reproduce the particle tracking behavior of a more realistic DFN model where each fracture is hydraulically heterogeneous. We used a concept of a synthetic reality where a model with heterogeneous hydraulic properties was used to produce simulated field data measurements to provide a benchmark for particle tracking analyses. Multiple synthetic reality models were used to reproduce a borehole pumping test mimicking a Posiva Flow Log (PFL) tool. Results from the pumping tests served as a calibration target for a model with the same fracture geometry, but homogeneous fracture properties. PEST ® was used as a tool for the calibration of homogeneous models so that they match the distribution of inflows obtained from the synthetic reality model. By using PEST®, it was possible to successfully calibrate a homogeneous models so that the distribution of inflows matched results from the synthetic reality model. A particle tracking analysis was then conducted on both heterogeneous and calibrated homogeneous models. Particle pathways were statistically analyzed in terms of particle travel distance, travel time and F-factor, which is a key parameter governing the transport of radionuclides within fractured rock. Results show that homogeneous calibrated models can produce non-conservative results in terms of the F-factors. In our study, we have demonstrated that DFN models successfully calibrated to boreholes inflow data could give a poor prediction in terms of particle tracking analysis.