Poster #028, Computational Science (CS)

High-performance computing and multi-physics earthquake modeling towards next generation earthquake simulations

Alice-Agnes Gabriel, Ahmed E. Elbanna, & Yehuda Ben-Zion
Poster Image: 

Poster Presentation

2021 SCEC Annual Meeting, Poster #028, SCEC Contribution #11621 VIEW PDF
Earth modeling is highly complex – and when the earth quakes, there are additional dynamic complexities. While computational seismology has been a pioneering field for high-performance computing, earthquake source processes are very ill-constrained and highly non-linear. Earthquake science is increasingly data-rich, but the lack of quantitative data on timescales capturing multiple large earthquake cycles is a fundamental impediment for progress in the field. Physics-based simulations provide the only path for overcoming the lack of data and elucidating multi-scale dynamics and spatio-temporal patterns that extend the knowledge beyond sporadic case studies and regional statistical laws.
...
/>
We present recent developments and future directions for data-driven seismological, geodetic, tectonic and experimental analyses that can be integrated synergistically with multi-physics modeling. Physic-based modeling can shed light on the dynamics, severity and cascading hazards of earthquake behaviour and enables an unparalleled degree of realism exploiting high-performance computing. We demonstrate the potential of Solid Earth community software for performing data-integrated large-scale scenarios of recent powerful multi-fault earthquake cascades; simulating 3D partially and fully-coupled Earth and ocean models of tsunami generated during earthquakes; evolution of complex localization patterns in fault zones and coupling between changes of elastic moduli in failing regions and subsequent rupture properties. The degree of realism achieved in these simulations is enabled by modern software, e.g. allowing for multi-petascale computational efficiency and high-order accuracy in time and space. The inclusion of probabilistic and Bayesian frameworks, geometric transformations and diffuse interface approaches to avoid manual meshing are future directions for exploiting the expected exascale computing infrastructure.

SHOW MORE