Constraints on stress and friction from dynamic rupture models of the 1994 Northridge, California, earthquake

We have simulated several scenarios of dynamic rupture propagation for the 1994 Northridge, California, earthquake, using a three-dimensional finite-difference method. The simulations use a rate- and slip-weakening friction law, starting from a range of initial conditions of stress and frictional parameters. A critical balance between initial conditions and friction parameters must be met in order to obtain a moment as well as a final slip distribution in agreement with kinematic slip inversion results. We find that the rupture process is strongly controlled by the average stress and connectivity of high-stress patches on the fault. In particular, a strong connectivity of the high-stress patches is required in order to promote the rupture propagation from the initial nucleation to the remaining part of the Fault. Moreover, we find that a small amount of rate-weakening is needed in order to obtain a level of inhomogeneity in the final slip, similar to that obtained in the kinematic inversion results. However, when the amount of rate-weakening is increased, the overall moment drops dramatically unless the average prestress is raised to unrealistic levels. A velocity-weakening parameter on the order of 10 cm per second is found to be adequate for an average prestress of about a hundred bars. The presence of the free surface and of the uppermost low-impedance layers in the model are found to have negligible influence on the rupture dynamics itself, because the top of the fault is at a depth of several kilometers. The 0.1-0.5 Hz radiated waves from the dynamic simulation provides a good fit to strong motion data at sites NWH and SSA. Underprediction of the recorded peak amplitude at JFP is likely due to omission of near-surface low velocity and 3-D basin effects in the simulations.