Community-Driven Code Comparisons for Three-Dimensional Dynamic Modeling of Sequences of Earthquakes and Aseismic Slip (SEAS)

Junle Jiang, Brittany A. Erickson, Valere R. Lambert, Jean-Paul Ampuero, Ryosuke Ando, Sylvain D. Barbot, Camilla Cattania, Luca Dal Zilio, Benchun Duan, Eric M. Dunham, Alice-Agnes Gabriel, Nadia Lapusta, Duo Li, Meng Li, Dunyu Liu, Yajing Liu, So Ozawa, Casper Pranger, & Ylona van Dinther

In Preparation September 30, 2021, SCEC Contribution #11680

Crustal faulting and earthquakes are inherently multi-scale dynamic processes. Numerical modeling of sequences of earthquakes and aseismic slip (SEAS) provides a self-consistent framework to connect, interpret, and predict geophysical observations across spatial and temporal scales. With the increasing complexity and growing applications of SEAS models, numerical code verification is imperative and essential to producing reliable simulation results to better understand earthquakes and assess seismic hazard. Here, we pursue community-driven efforts to verify three-dimensional (3D) SEAS models, building on our earlier success in two-dimensional (2D) problems (Erickson et al., 2020). We design two 3D benchmarks to test the capabilities of different numerical codes to resolve mathematically well-defined earthquake faulting problems, balancing the computational cost and desired complexities of 3D model setup. These benchmarks consider quasi-dynamic problems involving a 2D planar vertical strike-slip fault obeying rate- and state-dependent friction laws, embedded in a 3D homogeneous, linear elastic whole space (BP4) or half-space (BP5), where sequences of earthquakes and slow slip arise due to tectonic loading. We use simulations from 10 modeling groups to assess the agreement among model observables, including local and overall fault behavior and earthquake parameters, and explore how computational factors such as domain size, grid spacing, and boundary conditions contribute to model discrepancies. We found excellent quantitative agreement among simulated outputs for sufficiently large model domains and fine grid spacing. Agreement among simulated earthquake sequences requires finer discretization than that needed for individual ruptures, including additional considerations when resolving free surface effects. Some simulated properties of individual ruptures, such as rupture style, duration, stress drop and magnitude, are comparable among even marginally resolved simulations. However, details of longer-term fault behavior, such as earthquake recurrence intervals and nucleation phase, are more sensitive to numerical resolution and domain-size-dependent loading. These results lend confidence to the accuracy of participating numerical codes and reveal the relative sensitivities of different observables to computational and physical factors. Both aspects will be important for our efforts to test, validate, and integrate SEAS models with geological and geophysical observations.

Key Words
earthquake source processes, aseismic slip, numerical modeling, code verification, community project

Citation
Jiang, J., Erickson, B. A., Lambert, V. R., Ampuero, J., Ando, R., Barbot, S. D., Cattania, C., Dal Zilio, L., Duan, B., Dunham, E. M., Gabriel, A., Lapusta, N., Li, D., Li, M., Liu, D., Liu, Y., Ozawa, S., Pranger, C., & van Dinther, Y. (2021). Community-Driven Code Comparisons for Three-Dimensional Dynamic Modeling of Sequences of Earthquakes and Aseismic Slip (SEAS). Journal of Geophysical Research - Solid Earth, (in preparation).


Related Projects & Working Groups
SEAS, FARM, SDOT