Figure 1. Top: BP3-QD considers a planar, dipping fault embedded in a homogeneous, linear elastic half-space with a free surface where motion is plane-strain. The fault is governed by rate-and-state friction down dip to a distance Wf and creeps at an imposed constant rate Vp down to the infinite dip distance. The left and right sides of the fault are labeled with “(-)” and “(+)”, respectively. For the detailed benchmark description, please see https://strike.scec.org/cvws/seas. Bottom: Time series from two different codes (boundary element and finite difference) showing discrepancies attributable to thrust versus normal faulting assumption.

Conveners: Brittany Erickson and Junle Jiang
Date/Time:  Friday, October 30, 2020 (09:00-12:30 Pacific Time or UTC-7)
SCEC Award and Report: 20120

Our SCEC group advances ​computational methods for simulating Sequences of Earthquakes and Aseismic Slip (SEAS) by conducting a suite of code verification exercises. In SEAS models, the objective is to capture the interplay of aseismic fault slip—that ultimately leads to earthquake nucleation—and dynamic earthquake ruptures themselves. We want to understand which physical factors control the full range of observables such as aseismic deformation, earthquake initiation locations, ground shaking during dynamic rupture, recurrence times and magnitudes of earthquakes.

The online workshop on October 30, 2020 brought together 54 scientists from the U.S., England, Netherlands, Japan, Canada, Germany, Brazil, New Zealand, and Switzerland. Almost half the participants were graduate students and postdocs. The purpose of the workshop was to review results from the latest benchmark problems (BP3-QD and BP5-QD) and plan future research targets and SEAS activities that would continue to advance earthquake models with more robust physical features.

Workshop Presentations

Figure 2. Cumulative slip profiles from a BP3-QD simulation with a 30-degree dipping fault and (top) thrust and (bottom) normal faulting assumption from a boundary-element based code.

Goals and Introductions. Brittany Erickson (University of Oregon) and Junle Jiang (University of Oklahoma) kicked off the workshop sharing SEAS group activities over the last year, including presentations at SCEC2019 and the AGU fall meeting, as well as the first SEAS group publication (Erickson, Jiang, et al. 2020). Then all participants were invited to introduce themselves and/or their group’s research using 1-2 slides. These “lightning intros” helped everyone to meet old and new participants, to learn about SEAS-related science activities within the community, and to understand people’s motivations for attending the workshop and participating in SEAS.

Benchmark Results—BP3-QD (a planar, dipping fault embedded in a homogeneous, linear elastic half-space with a free surface where motion is plane-strain, Figure 1, top). The fault is governed by rate-and-state friction down dip to a distance Wf and creeps at an imposed constant rate Vp down to the infinite dip distance. The simulations include the nucleation, propagation, and arrest of quasi-dynamic earthquakes, and aseismic slip in the post- and inter-seismic periods. We asked modelers for results from three dipping angles, at 30, 60 and 90 degrees. Results were analyzed from three different modeling groups, including those from boundary-element based codes, and a finite difference code. We realized the problem did not state whether to assume normal or thrust motion—an oversight that took time to find the cause of major discrepancies (Figure 1, bottom). Once sorted, we were able to make better informed comparisons and found good agreements across codes, with discrepancies attributable to computational domain size and cell size. We were also able to make sense of results from thrust versus normal faulting assumptions by plotting normal stress changes, as well as profiles of slip contours (Figure 2). A thrust assumption (top) is associated with increased normal stress change, which decreases the critical nucleation length, allowing events to nucleate sooner and deeper than the normal faulting case (below). In a thrust scenario, therefore, earthquake cycles occur with a decreased recurrence interval and less slip with each rupture. The vertical fault case (not shown) is characterized by repeated (uni-modal) events, whereas faults that dip at 30 degrees (Figure 2) show bi-modal event sequences. We concluded discussion of this benchmark by creating a “to-do” list including the task of updating the benchmark description to include a specification of thrust versus normal faulting, as well as plans to do additional comparisons of time-series of off-fault surface stations, and (for codes with a volume discretization) to further explore dependency on computational domain size and cell size.

Figure 3. This benchmark is a 3D problem with a planar fault embedded vertically in a homogeneous, linear elastic half-space. The fault is governed by rate-and-state friction in the colored region, surrounded by regions in gray with an imposed rate Vp. A favorable nucleation zone (dark green square) is located at one end of the VW patch.

Figure 4. Results from BP5-QD. Top: The first coseismic event rupture front across several modeling groups. Bottom: Time series of horizontal surface slip rate showing good initial agreement.

Benchmark Results—BP5-QD (a 2D planar fault embedded in a 3D, homogeneous half space, with quasi-dynamic events, Figure 3). The objective of BP5-QD is to understand resolution issues and verify models in 3D, following what we learned from our last 3D benchmark BP4. The model output requested from each modeling group included time series from a set of on-fault and off-fault stations, data for plotting coseismic rupture contours of the first event, cumulative slip and stress evolution along several depths and strike locations, and an earthquake catalog recording earthquake magnitude, recurrence times and durations of events. An overview of the participating modeling groups and codes was given and the comparison strategy for assessing model agreements described. For two different cell sizes (1 km and 500 m), Jiang presented (a) long-term time series at various fault locations, which showed good agreements across modeling groups, and (b) plots of the rupture contours showing minor spatial offsets due to different treatments of the surface node locations (Figure 3). Figures of along-depth and along-strike cumulative slip demonstrate the high similarity of earthquake rupture patterns in different models. Some differences in the coseismic behavior are persistent from the first or later events in the simulations, suggesting some other factors are at play (Figure 4). The group agreed that more simulations are needed for more informed comparisons, in particular to improve agreements in coseismic behavior. We were encouraged by evidence of convergence of many models with decreasing cell size and increased domain size, and by the fact that many earthquake characteristics agreed well despite variability in rupture direction and nucleation processes. We conclude that we have achieved good benchmark verification results for BP5-QD, having addressed some issues encountered for BP4.

Future Directions for SEAS. Bruce Shaw (Columbia University) gave a talk entitled “Earthquake complexity, simulators, and SEAS” aimed at bridging connections between current earthquake simulators efforts and SEAS modeling. The presentation provided good discussion points for future SEAS activities. Jiang and Erickson concluded the workshop with a group discussion, motivated by a participant survey sent out prior to the workshop. The group strategized on criteria for success for the SEAS group, how to constrain SEAS models, how to continue testing and validating our models when increased complexities are present, how to recruit additional effort and support new modelers, what resources and/or tools would facilitate better coordination, and what the SEAS research targets for the near and long-term for SCEC and benchmarks should be over the next 1-2 years.

References

• Erickson, B.A. et al. (2019), The Community Code Verification Exercise for Simulating Sequences of Earthquakes and Aseismic Slip (SEAS): Initial Benchmarks and Future Directions, poster 174 presented at the 2019 SCEC Annual Meeting, Palm Springs, CA. SCEC Contribution 9618
• Erickson, B.A. et al. (2019), The Community Code Verification Exercise for Simulating Sequences of Earthquakes and Aseismic Slip (SEAS): Initial Benchmarks and Future Directions, poster T13D-0309 presented at the 2019 AGU annual meeting, San Francisco, CA.
• Erickson, B. A., Jiang, J., Barall, M., Lapusta, N., Dunham, E., Harris, R., Abrahams, L. S., Allison, K. L., Ampuero, J.-P., Barbot, S., Cattania, C., Elbanna, A., Fialko, Y., Idini, B., Kozdon, J. E., Lambert, V., Liu, Y., Luo, Y., Ma, X., Mckay, M. B., Segall, P., Shi, P., van den Ende, M., Wei, M. The Community Code Verification Exercise for Simulating Sequences of Earthquakes and Aseismic Slip (SEAS) (2020) , Seismological Research Letters, 91 (2A): 874-890, doi: 10.1785/0220190248. SCEC Contribution 9066

Presentation slides may be downloaded by clicking the links below. PLEASE NOTE: Files are the author’s property. They may contain unpublished or preliminary information and should only be used while viewing the talk. Only the presentations for which SCEC has received permission to post publicly are included below.