Multiphase flow and fault poromechanics : understanding earthquake triggering and seismic hazard

Abstract

In this Thesis, we investigate natural and engineered processes related to the assessment of the seismic hazard from the impact of anthropogenic operations on the stability of pre-existing geological faults. We do so by developing simulation tools that coupled multiphase flow and geomechanics, and apply them at the field scale using geologically realistic representations of the subsurface. In a first contribution at the scale of individual fractures, we study the impact of confining stress on the capillary pressure behavior during drainage through rough fractures, where we find that capillary pressure variations are sensitive to the degree of confining stress and the degree of spatial correlation of the fracture aperture.

By solving the elastic contact problem and simulating slow two-phase displacements through the fracture gap, we uncover the universality class of avalanche size in fluid displacement, and find that it is consistent with a process controlled by self-organized criticality. In a second contribution at the scale of hundreds of kilometers, we address the importance of long-term, large-scale crustal deformation on the spatiotemporal distribution of Slow Slip Events (SSEs) in the Guerrero Gap, putting forward an alternative explanation for SSE nucleation, interval time and arrest. We show, by means of finite element simulations with rate-state friction, that fault geometry and crustal deformation control the nucleation and arrest of SSEs, via normal stress changes along the subducting slab that act as a mechanism for SSE stabilization. In a third contribution, we develop a two-way coupled multiphase flow and geomechanics model that rigorously accounts for the fluid-solid interaction.

We do so by coupling two well-established open-source simulators, the open-source finite element mechanical simulator PyLith and the finite volume open source flow simulator MATLAB Reservoir Simulation Toolbox (MRST). We employ the fixed-stress split of the fully-coupled problem, which renders the sequential iterative scheme unconditionally stable. We validate our implementation using analytical solutions for single-phase flow for a range of model parameters, and find excellent agreement in all cases. We then apply our simulator to synthetic cases to illustrate the impact of CO₂ injection on earthquake triggering on a pre-existing fault, demonstrating that poroelastic effects can have a strong fault-weakening effect even through impermeable geologic strata. In the two final contributions in this thesis, we apply the coupled multiphase flow and geomechanics simulator described above to assess seismic hazard from fluid injection at the reservoir scale.

In our first application, we revisit the classical experiment in earthquake control from water injection at the Rangely oil field, Colorado. The coupled flow-geomechanics simulations on a geologically constrained structural model of the Rangely field, along with reservoir-pressure and seismological data, provide an unique opportunity to understand the mechanisms responsible for the observed seismicity. In particular, our analysis allows us to separate the contributions to fault destabilization from direct pore pressure diffusion and poroelastic effects and to elucidate the fundamental role of fluid flow along the fault. In our second field-scale application, we investigate the impact of industrial-scale CO₂ storage on the stability of, and potential leakage along, pre-existing faults in the Gulf of Mexico (GoM).

We do so by performing 3D numerical simulations of coupled flow and geomechanics using high-fidelity geological models of the Miocene section of the GoM, both at the field scale (10s of km) and at the regional scale (100s of km). We pay particular attention to the frictional and hydraulic properties of unlithified sedimentary faults, and incorporate a detailed, physics-based, probabilistic representation of clay and sand smearing to populate the flow properties of normal faults. We then investigate different scenarios of injection-well location in relation with faults' geometry and architecture, representing geologic settings corresponding to "open" and "closed" reservoirs.

The results of our flow-geomechanics simulations suggest that CO₂ injection results in small fault destabilization, and vanishingly small probability of leakage along faults--supporting the notion that large-scale (100s of Mt) CO₂ injection in the GoM is feasible, but that well location is key for the success of individual Carbon Capture and Storage (CCS) projects.