Collaborators
- Brad Aagaard at United States Geophysical Survey
- Charles Williams at GNS Science Inc.
- Gerard Gorman at ICL Earth Science and Engineering
Short Projects
Incorporation of Poroelasticity into Crustal Dynamics
ProblemSimulation of earthquake rupture and the consequent post-seismic deformation are crucially important for risk management in the affected areas, as well as scientific understanding of fault triggering and plastic deformation mechanisms. It has been widely conjectured that subsurface fluids play an important role in both processes, but this has never been verified by either experiment or simulation.
ApproachUsing the PETSc Plex mesh and function space abstraction, we can assemble multiphysics systems for domains of any dimension, with any cell shape, and any finite element basis. We are now incorporating the faulting mechanism as a lower-dimensional manifold in the assembly process, in order to assemble and solve the fully coupled system.
ImpactThe ability to quantify the effects of subsurface fluids on fault behavior, through a poroelastic model, would have critical implications for seismic hazard models, which in turn by law determine insurance rates in high risk areas. Nearly 3B USD is underwritten each year in the United States alone, measured by direct premiums (see here).
PlanWe have already produced homogeneous multiphysics systems in PyLith, the crustal physics simulator sponsored by the NSF and USGS. We are now working to incorporate fault behavior in the multiphysics context.
- Multiphysics solvers for incompressible elasticity, temperature, and poroelasticity
- Adding in point source double-couples for faults
Motivation
We could do quick inversion with a NURB surface describing the fault surface, which gets sampled, but we still have material property variation using the FEM in PyLith.
Background
Project
- Fault system evolution by coupling with Underworld
Motivation
Background
Project
- Proof of convergence for fault preconditioner
- What can we say computationally about L from Scholtz's paper?
- Do Elastoplastic systems reduce to Viscous systems when averaged over long times? Can we prove that fault
networks arising from fracture become a certain viscous model?
Blaise might be interested. Can we run a fracture simulation which produces a fault network?