SCEC Award Number 18099 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title Advancing Simulations of Sequences of Earthquakes and Aseismic Slip (SEAS)
Investigator(s)
Name Organization
Brittany Erickson Portland State University Junle Jiang University of California, San Diego Yuri Fialko University of California, San Diego Michael Barall Invisible Software, Inc.
Other Participants Eric Dunham, Nadia Lapusta, Ruth Harris, Jean-Paul Ampuero, Eric Daub, Ahmed Elbanna, Yoshi Kaneko, Jeremy Kozdon, Yajing Liu, Shuo Ma, Paul Segall, Terry Tullis, Matt Wei, Ylona van Dinther, Bruce Shaw and others.
SCEC Priorities 1d, 1e, 3f SCEC Groups SDOT, FARM, CS
Report Due Date 03/15/2019 Date Report Submitted 05/05/2019
Project Abstract
 With SCEC support this past year, we have initiated and led the SEAS initiative, which consists of a series of community code verification exercises for SEAS models. Our main progress and achievements are to have:
- Gathered a group of researchers who are committed to recent benchmark exercises, or are interested in
our current activity and/or future participation (~30 PIs, ~20 students/postdocs).
- Established our online platform for benchmark comparisons (http://scecdata.usc.edu/cvws/seas/)
- Designed our first two benchmarks, BP1 and BP2, for 2D antiplane problems.
- Organized our first workshop in Apr. 2018 jointly with the dynamic rupture group for meet-and-greets
of modelers, introduction of numerical codes, and discussions on benchmark BP1 results.
- Organized a second SEAS-themed workshop in Nov. 2018 for sharing advancements in the field and
discussing results of benchmark BP2.
- In BP1 exercises, we found excellent agreements between 11 modeling groups, when computational
domain sizes, boundary conditions, and comparison schemes are appropriately chosen.
- In BP2 exercises, we found excellent agreements between most numerical codes, when important
physical length scales are well resolved. The simulated complexity in earthquake sequence in
insufficiently resolved models can be significantly different from the “true” fault behavior.
- Presented our group activities and benchmark results at 2018 SCEC/AGU Annual Meetings.
Intellectual Merit Developing robust predictive models of earthquake source processes is one of the main SCEC goals. Research groups within the earthquake science community are contributing to this goal through the development of computational methods for simulating Sequences of Earthquakes and Aseismic Slip (SEAS). ​In SEAS models, the goal is to capture the interplay of interseismic periods and the associated aseismic fault slip—that ultimately lead to earthquake nucleation—and earthquakes (dynamic rupture events) themselves, in an effort to understand which physical factors control the full range of observables such as aseismic deformation, nucleation locations of earthquakes, ground shaking during dynamic rupture, recurrence times and magnitudes of major earthquakes. One of the significant challenges in the SEAS modeling effort arises from the varying temporal and spatial scales that characterize earthquake source behavior. Computations are further complicated when material heterogeneities, bulk inelastic responses, fault nonplanarity, and their evolution with time and slip, are included. However, accounting for such complexity is widely recognized as crucial for understanding the real Earth and predicting seismic hazards.
Broader Impacts The SEAS initiative and code verification project has supported scientific training and communication within and across institutions, included the participation of undergraduate and graduate students, and postdocs. It has motivated community code development to include increased model complexity, which will improve seismic hazard estimates.
Exemplary Figure Fig. 2 Results from BP1. Left: plots of max shear stress during first event for all model results show qualitative similarities but discrepancies due, in part, to different domain sizes and boundary conditions. Right: time series of maximum slip rates over the first ~10 events with similar domain sizes show good agreement across multiple codes.