Validating ground-motion simulations on rough faults in complex 3D media

We utilize a two-step process to validate 0-4 Hz ground motion simulations using the 1989 Loma Prieta earthquake. In the first step we run multiple realizations using the Graves and Pitarka hybrid method as implemented on the SCEC Broadband Simulation Platform and compare these with near-fault (R < 40 km) recorded motions. A total of 648 rupture scenarios are examined and from these results we select the best fitting ruptures, which are then used in the second step of the validation. This second step consists of running 3D simulations using a finely sampled seismic velocity mesh. The base 3D structure is obtained from the USGS SF Bay Area velocity model, which is also modified to include small-scale 3D stochastic perturbations. The 3D results are compared with the recorded motions and show very good agreement over the frequency band 0-4 Hz for distances out to 40 km. Most importantly, the newly added features to the simulation process reduce the coherency of the radiated higher frequency (f > 1 Hz) ground motions, and homogenize radiation-pattern effects in this same bandwidth, bringing the simulations into closer agreement with the very near-fault records.