SCEC Award Number 20080 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Optimizing and further developing simulations of sequences of earthquakes and aseismic slip
Investigator(s)
Name Organization
Nadia Lapusta California Institute of Technology
Other Participants Graduate students Kavya Sudhir and Valère Lambert, postdoc Luca Dal Zilio
SCEC Priorities 5a, 1e, 2e SCEC Groups FARM, SDOT, CS
Report Due Date 03/15/2021 Date Report Submitted 04/22/2021
Project Abstract
Physics-based predictive modeling of destructive large dynamic events relies on simulations of sequenc-es of earthquakes and aseismic slip (SEAS), because prior slip events, including aseismic slip, determine where earthquakes would nucleate and modify stress and other initial conditions before dynamic rupture. We have expanded our boundary-integral methodology (BIM) for SEAS simulations to include along-fault diffusion and an approximation for off-fault plasticity, and compared the accuracy and efficiency of different variable time-stepping schemes. We have also participated in the code comparison led by Drs. Erickson and Jiang and developed a benchmark for future code comparison on interaction between two seismogenic (velocity-weakening) fault segments separated by a velocity-strengthening barrier, a prob-lem of high relevance to rupture jumps between segments and UCERF. Our findings show that quasi-dynamic treatment of wave-mediated stress changes and using oversized cells significantly affects the long-term behavior of the model, including the probability of jumps. Moreover, the long-term simula-tions of long enough faults produce different earthquake sequences and different jump rates even when well resolved by all known metrics. We also find that matching frequency-magnitude statistics and static stress drops does not imply the same probability of ruptures jumping across the unfavorable barrier. In fact, the significant sensitivity of the rate of multi-segment ruptures to small changes in our numerical models implies that this hazard parameter may also be sensitive to physical perturbations on natural faults and motivates finding more reliable metrics for describing long-term fault behavior and assessing seismic hazard, tasks for which physics-based modeling is well-suited.
Intellectual Merit Our study aims to verify, optimize and develop codes that simulate earthquakes sequences and aseismic slip, including the problems of earthquake jumps between segments. Our goal is two-fold: (i) improve the boundary-integral approach for simulations on planar faults by incorporating a range of physical in-gredients, including representations of fluid flow, fault roughness, and off-fault inelasticity and (ii) use the simulations to understand the impact of different modeling assumptions on the modeling outcomes and connections between different observables. Our findings that (i) quasi-dynamic treatment of wave-mediated stress changes and using oversized cells significantly affects the long-term behavior of the model, including the probability of jumps and (ii) matching frequency-magnitude statistics and static stress drops does not imply the same probability of ruptures jumping across the unfavorable barrier have direct implications for large-scale simulations of earthquakes sequences and fault interactions and the resulting estimates of seismic hazard. These findings imply that care must be taken in evaluating the stability of the outcomes, such as rupture jump probabilities, to the assumptions of the model regarding inertial effects and resolution.
Broader Impacts The results of this project would (a) provide better understanding of appropriate ways of modeling the long-term behavior of faults; (b) provide better assessment of seismic hazard and evaluation of possible extreme events, based on physical models and integrations of laboratory, field and seismological stud-ies; and (c) contribute to the development of realistic scaling laws for large events. Two graduate stu-dents and an undergraduate research student have gained valuable research experience by participat-ing in the project and interacting with the SCEC community.
Exemplary Figure Figure 1. Different long-term earthquake sequences and different percentages of two-segment ruptures in three simulations with different resolution. a) Schematic for the model of two velocity-weakening segments separated by a velocity-strengthening barrier; we consider a depth-averaged representation of the model to enable fine numerical resolution. b) Resolution of the shear stress at the rupture front. For well-resolved simulations, the cohesive zone is resolved by several cells. (c-e) 4000 years of earthquake sequences for simulations using (c) oversized cells and (d-e) small enough cells to resolve the cohesive zone. Contours in blue and red represent earthquakes that do and do not rupture across both segments, respectively. The simulations with adequate discretization exhibit initially nearly identical slip histories that eventually substantially diverge. The resolution is excellent in both cases, with the quasi-static cohe-sive zone at the rupture tip resolved by 36 and 72 cells, respectively. The simulation using oversized cells that do not resolve the rupture front has different slip history and substantially different jump rate of 0.15. Adapted from Lambert and Lapusta, 2021. Minor differences in model assumptions can lead to substantial differences in the jump rate, as shown further in Lambert and Lapusta (2021).