Short slip duration in dynamic rupture in the presence of heterogeneous fault properties

Recent studies of strong motion data consistently show that the risetime (duration of slip at particular locations on the fault) is significantly shorter than the overall rupture duration. The physical explanation for this observation and its implications have become central issues in earthquake source studies. Two classes of mechanisms have been proposed to explain short risetimes. One explanation is that velocity-weakening frictional behavior on the fault surface causes the fault to self-heal. This possibility is suggested by rate-dependent friction observed in laboratory experiments and by some two-dimensional dynamic numerical simulations of earthquake rupture. It has recently been demonstrated, however, that the velocity dependence of friction observed in the laboratory is too weak to cause faults to self-heal. An alternative explanation for short risetimes is that spatially heterogeneous fault strength (e.g., barriers) limit the slip duration. In this paper we investigate this second explanation for short risetimes by constructing a three-dimensional dynamic rupture model for the 1984 Morgan Hill, California earthquake (M(W)=6.2) using a kinematic model previously obtained from waveform inversion of strong motion data. We assume velocity-independent friction and a critical stress fracture criterion and derive a dynamic model specified by the spatial distribution of dynamic stress drop and strength excess that reproduces the slip and rupture time of the kinematic model. The slip velocity time functions calculated from this dynamic model are then used in a subsequent inversion to fit the strong motion data. By alternating between dynamic and kinematic modeling, we obtain a dynamic model that provides an acceptable fit to the recorded waveforms. In this dynamic model the risetime is short over most of the fault, which is attributable entirely to the short scale-length slip/stress drop heterogeneity required by the strong motion data. A self-healing mechanism, such as strongly velocity-dependent friction, is not required to explain the short risetimes observed in this earthquake.