Resolving simulated sequences of earthquakes and fault interactions

Abstract:

Physics-based numerical modeling of earthquake source processes aims to combine available real-world data and physical principles to improve our understanding of fault behavior. The ultimate aspiration is to develop models that have predictive power for quantities of interest for seismic hazard, such as the probability of an earthquake rupture jumping from one fault segment to another. However, the outcomes of numerical simulations can depend on choices in numerical procedures and physical approximations, and assessing the predictive power of numerical models remains a topic of ongoing research.

Here, we investigate the sensitivity of numerical simulations of sequences of earthquakes and aseismic slip (SEAS) to choices in numerical discretization and approximations for inertial (wave-mediated) effects, using a simplified 2-D model of a crustal fault with two co-planar segments separated by a relatively strong creeping barrier. Our simulations demonstrate that the frequency of two-segment ruptures and the complexity of the resulting earthquakes sequences significantly depend on the modeling assumptions. In part, simulations with different numerical discretization (including insufficient discretization) and different treatment of inertial effects that result in similar complexity of earthquake sequences can have very different probabilities of multi-segment ruptures.

Even among the simulations that are well-resolved by standard considerations, some properties of the simulated events are similar and some are not, when the fault is long enough with respect to the nucleation size. Some simulated properties, such as average slip, average static stress drop, and the history of average stress on the fault, are similar among adequately resolved simulations, suggesting that they may be reliably estimated from well-formulated numerical models. However, other properties, such as the frequency of two-segment ruptures, are highly sensitive to the numerical discretization even among well-resolved simulations and cannot be reliably inferred from such numerical studies. Our results emphasize the general need to examine the potential dependence of simulation outcomes on modeling assumptions, including resolution, particularly when evaluating their predictive value for seismic hazard assessment.