Flow, Transport and Deformation in 3D Fractured Media : Sensitivity to Model Conceptualization

Predictive modeling of fluid flow and solute transport in fractured rock is central to many subsurface applications, including nuclear waste isolation, geothermal energy, carbon storage, and hydrocarbon recovery. Such predictions depend strongly on the conceptualization of fractured media and on the assumptions used in model formulations. This thesis investigates how conceptual modeling choices influence the analysis of fluid flow, mass transport, and mechanical deformation behaviors in three-dimensional fracture networks by examining four key sources of uncertainty: multiscale anisotropy, geomechanical model simplification, mechanical boundary conditions, and ensemble structure and variability for modeling flow and transport in fractured rocks. Results demonstrate firstly that fracture-scale and network-scale anisotropy features interact in a non-linear manner. Alignment between the two scales enhances preferential flow, while misalignment suppresses connectivity and delays transport, producing travel-time differences of up to two to three orders of magnitude even for networks with similar equivalent permeability. Secondly, simplified geomechanics models (omitting fracture nonelastic deformation, stress variability, and fracture interaction) fail to capture complex stress-induced deformation of fractures under high stress ratio conditions, leading to underestimation of flow capacity and faster transport pathways, although they may remain robust for low stress ratios. The findings highlight that the validity of geomechanical simplification is strongly stress-dependent. Thirdly, mechanical boundary-condition effects are negligible for networks dominated by smaller fractures but become significant in networks containing larger fractures, especially under strongly anisotropic stress conditions. The higher connectivity of larger fractures allows boundary-induced stresses to propagate deeper into the system, resulting in stronger deformation, increased aperture, and enhanced transport. Fourthly, we investigate the ensemble structure and variability of flow and transport in fracture networks with varying connectivity. A suite of statistical metrics is introduced to quantify deviations of realizations from the ensemble average. We show that near the percolation threshold, strong variability among realizations limits the reliability of ensemble-averaged predictions. This variability diminishes away from the percolation threshold. This thesis shows that flow and transport predictions in fractured rock systems are highly sensitive to model conceptualization and assumptions, highlighting the need for robust and uncertainty-aware modeling frameworks to guide model selection and evaluation for subsurface engineering applications.