SCEC Award Number 20033 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Extending Earthquake Sequence Simulations to Plane Strain and Multiple Faults
Name Organization
Eric Dunham Stanford University
Other Participants Martin Almquist (postdoc, Stanford)
Yongfei Wang (postdoc, USC/SCEC)
SCEC Priorities 2e, 3d, 1c SCEC Groups CS, FARM, SDOT
Report Due Date 03/15/2021 Date Report Submitted 03/25/2021
Project Abstract
Models of sequences of earthquakes and aseismic slip (SEAS) are used throughout SCEC and the broader earthquake science community to understand controls on rupture behavior, whether slip is seismic or aseismic, and to integrate constraints from the lab and field into a comprehensive modeling framework. The new generation of SEAS models encompass dynamic ruptures as the quasi-dynamic approximation is replaced with full inertial dynamics. Our group has contributed to advances in SEAS modeling by exploring the coupling between the elastic upper crust and viscoelastic lower crust and upper mantle, effects of shear heating, and interactions with evolving pore fluid pressure within the fault zone. However, this work was limited to planar faults in 2D antiplane shear. In this project, we introduced a new, open-source SEAS code for the 2D plane strain problem with an arbitrary number of possibly nonplanar fault segments. Key to this effort was the development of a new, high-order accurate finite difference discretization of the elastic operator on curvilinear, multiblock meshes. This operator was verified with rigorous convergence tests for fully anisotropic elasticity problems in complex geometries. We then introduced rate-and-state friction on fault interfaces and adaptive time stepping, giving us earthquake sequence simulation capabilities. The new code was verified by solving the recent SEAS TAG benchmark problem BP3 for dipping faults (both normal and reverse). Planned future work includes investigation of earthquake sequences on nonplanar and branching faults and dipping faults, as well as inclusion of fault-zone fluid transport and viscoelastic material response.
Intellectual Merit Earthquake models integrate lab and field constraints on source processes and connect with seismic, geodetic, and other data from actual events to provide insight into how and why earthquakes occur. This project introduces new, open-source modeling software for earthquake rupture propagation and earthquake sequence simulations in complex geometries involving multiple, nonplanar fault segments. This software will be used to determine controls on rupture path selection in complex fault networks, with self-consistent stress conditions prior to ruptures determined self-consistently with the past history of seismic and aseismic slip. Additional processes thought to be important for earthquake behavior, including viscous flow of rocks at depth and fluid transport and pore pressure dynamics within the fault zone will also be added.
Broader Impacts This project trained one postdoctoral fellow, an applied mathematician specializing in numerical methods and scientific computing, building connections between the Earth sciences and applied mathematics. The modeling software has been released to the community using an open-source license and can be obtained from (for wave propagation only, with documentation) and (for earthquake sequences, no documentation yet).
Exemplary Figure Figure 6. (left) Injection into a velocity-strengthening fault drives an aseismic slip front outward. Simulation is symmetric about the injection site at z=0. (center) Pressurization weakens the center of the fault, causing slip, which then transfers stress outward to drive the slip front. Suctions from inelastic pore dilation strengthen the fault, which slows, but does not prevent, slip front advance. (right) Slip front migration rate depends on injection rate and initial stress (reported here as the ratio of shear to effective normal stress). (bottom) Snapshots of shear stress change, pressure change, slip, and slip velocity at 144 d from the start of injection, from the same simulation shown at top. From Yang & Dunham (2021).