Three-dimensional finite-difference modeling of the San Andreas fault: Source parameterization and ground-motion levels

Abstract Olsen et al. (1995) recently simulated an Mw 7.75 earthquake on the San Andreas fault, predicting long-period (T 2.5 sec) ground velocities of 140 cm/sec in the Los Angeles basin, about 60 km from the fault. These motions are much larger than estimates derived from empirical relations or other numerical simulations. Standard area-magnitude relations predict that the 170 × 16 km fault used in the simulations would produce an Mw 7.5 earthquake, giving a moment of 2.0 × 1027 dyne-cm, which is 2.4 times smaller than the moment used by Olsen et al. (1995). Further, self-similar scaling predicts a rise time of 3 sec for an Mw 7.75 event and 2.2 sec for an Mw 7.5 event. The filtered impulse slip function used by Olsen et al. (1995) has an effective rise time of 1.6 sec, yielding a response that is about 2 times larger than expected for periods less than 5 sec. This combination of high seismic moment and short rise time, along with the use of a uniform slip distribution, leads to the extreme ground-motion levels predicted by Olsen et al. (1995). To quantify the sensitivity of the long-period ground-motion response to source parameterization, we have performed 3D finite-difference (FD) simulations using various combinations of seismic moment, source rise time, and slip heterogeneity. These calculations incorporate the same grid dimensions, fault size, and bandwidth employed by Olsen et al. (1995). With a moment of 2.0 ՠ1027 dyne-cm, a rise time of 2 sec, and a smoothly heterogeneous slip distribution, we simulate peak long-period ground velocities of 155 cm/sec in the near-fault region and 40 cm/sec in the Los Angeles basin. These values are much closer to (although still higher than) empirical predictions. A uniform slip distribution produces the largest peak motions, both in the near-fault region and in the Los Angeles basin, whereas a rough asperity slip distribution noticeably reduces the maximum near-fault ground velocities. Our results indicate that the accurate simulation of long-period ground motions requires a realistic source parameterization, including appropriate choices of seismic moment and rise time, as well as the use of spatial and temporal variations in slip distribution.