Towards High-Frequency Ground Motion Prediction Using a Dynamic/Pseudo-Dynamic Approach

The past several years have seen substantial progress in physically based approaches to earthquake source modeling for the prediction of strong ground motion (e.g., Mai et al., 2001; Guatteri et al., 2002, 2003). We explore the use of this approach for predictions of wave generation from the source at higher frequencies, where important engineering needs exist. To do this we apply two physical constraints. We assume that wave generation is controlled by the product of maximum slip velocity and effective pulse width. The Kostrov-like slip velocity function (t ^-1/2) has a too steep spectral decay (f ^-1/2) to reproduce the high-frequency content observed in seismograms. Previous dynamic rupture simulation (< 2 Hz) in Guatteri et al. (2003) shows that maximum slip velocity of each subfault correlates well with the static stress drop. Moreover, the rise time in the heterogeneous slip distribution varies inversely with the time after rupture initiation, as long as the slip-weakening constitutive law is adopted. Using these characteristics of the dynamic faulting, we construct high-frequency source description that satisfies the above constraints. Once the high frequency slip velocity functions are obtained, ground motions are simulated by convolution of the slip velocity and Green_fs functions using the discrete wave-number method (e.g., Bouchon, 1981) in the low-frequency range, and by the ray-theory calculation using isochrone integration (Spudich and Frazer, 1984) in the high-frequency range. We examine the variability of the ground motions and response spectra, particularly in the near-fault region within 5 km of the fault. Even assuming the same slip distribution, variations in hypocentral location result in strong variations in the intensity of strong ground motions, indicating that the influence from the hypocenter location cannot be neglected for ground motion prediction in empirical attenuation relations.