Using mesh morphing and reduced-order modeling to quantify the influence of fault geometry on earthquake dynamic rupture

Gabrielle Hobson, Dave A. May, & Alice-Agnes Gabriel

Submitted September 7, 2025, SCEC Contribution #14759, 2025 SCEC Annual Meeting Poster #TBD

Natural faults have complex, non-planar geometries with rough surfaces that are challenging to constrain observationally. Fault geometries often remain uncertain at depth, even if surface rupture expressions are observed, and assumed fault geometries in models should incorporate uncertainty. Earthquake dynamic rupture simulations rely on meshes that represent fault geometries; however, quantifying model sensitivity to variations in fault geometry is challenging due to the effort required to create a single mesh, combined with the need to explore a high-dimensional parameter space with computationally expensive models.

We present a mesh morphing approach that enables sensitivity analysis with respect to varying fault geometries via an efficient and flexible framework. Mesh morphing takes a single reference mesh and deforms, or “morphs”, the mesh to represent a new fault geometry, without repeating the meshing process. Mesh morphing preserves mesh connectivity, and therefore we can construct data-driven, non-intrusive reduced-order models (ROMs) from dynamic rupture simulation output. The ROMs enable robust sensitivity analysis and uncertainty quantification with respect to varying geometries.

We demonstrate mesh morphing for the SCEC TPV 13-3D benchmark exercise, showing that morphed meshes maintain high quality as the fault dip angle varies across a 40° range. Simulations run using morphed meshes are accurate and agree closely with simulations run using exactly-generated meshes. Additionally, we construct ROMs that accurately predict (i) velocity time series at off-fault receivers and (ii) surface displacement at a single simulation time step, in the latter case with a speedup of 1e8 relative to the SeisSol full-order simulations. Using the ROMs combined with mesh morphing, we efficiently explore a parameter space that includes geometric variability and quantify the sensitivity of surface displacement and velocity time series to variable fault dip angle. We observe that subtle changes in fault geometry can influence rupture speed, slip distributions, and surface displacement, emphasizing the important role of geometric variability in earthquake modeling. The combined mesh morphing and ROM framework is an efficient, flexible tool that enables the incorporation of realistic geometric variability in dynamic rupture modeling. Extensions of the mesh morphing methodology to complex fault geometries will be discussed.

Key Words
Fault geometry, Dynamic rupture, Mesh morphing, Reduced-order modeling

Citation
Hobson, G., May, D. A., & Gabriel, A. (2025, 09). Using mesh morphing and reduced-order modeling to quantify the influence of fault geometry on earthquake dynamic rupture. Poster Presentation at 2025 SCEC Annual Meeting.

Related Projects & Working Groups
Fault and Rupture Mechanics (FARM)