QUASI-DYNAMIC VERSUS FULLY-DYNAMIC SIMULATIONS OF SLIP ACCUMULATION on FAULTS WITH HETEROGENEOUS FRICTION PROPERTIES

A number of factors which might be inferred from geological and geodetic observations are thought to influence fault seismic behavior: they include lithology which might control mechanical properties of fault zone, pore pressure and faults geometry. In theory the imfluence of these factors might be estimated from theoretical fault models and computer simulations. This is computatioannly challenging because this kind of study requires a proper account of the effect of these factors on rupture dynamics, at the scale of individual seismic event, as well as long sequence of rupture to capture the stochastic behavior of faults systems. In such studies it is computationally advantageous to not incorporate full inertial effects during simulated fast slip. That is why so-called quasi-dynamic methods have become increasingly popular, which approximately account for inertial effects (and hence seismic radiation) during simulated earthquakes through a radiation damping term. Such methods allow to continue simulations through the seismic phase, without having to pay significant additional memory and computational costs associated with modeling true wave-mediated effects. However, the resulting seismic events tend to have much slower slip velocity and rupture speeds and may modify significantly the resulting seismic events and hence the long-term fault behavior. In this study, we compare the results of quasi-dynamic and fully dynamic simulations, with wave effects during simulated earthquakes. We consider the long-term fault behavior in two problems: (i) interaction of two velocity-weakening regions separated by a small velocity-strengthening patch and (ii) segments with additional pronounced rate-weakening during seismic slip. We find that, in the absence of additional seismic weakening, the two methods generally result in the same qualitative behavior, with similar slip patterns, although there are quantitative differences. However, in simulations with additional rate weakening, the two methods produce qualitatively different long-term results, with different slip patterns and significantly different levels of shear stress on the fault. Our eventual goal is to determine the range of applicability for the quasi-dynamic approaches.