Earthquake Rupture Imaging And Multiscale Stress Drop Estimation

Investigation of earthquake source parameters is important for better understanding of the physics behind earthquakes, which is essential for mitigating the hazards of destructive earthquakes. This dissertation explores several different aspects of source parameter estimation. First, I image the rupture process of the 2004 M6.0 Parkfield earthquake in three dimensions using a waveform back-projection method. I identify a distinct secondary high-frequency phase that radiated from a point about 13 km northwest of the hypocenter. Together with existing slip model inversions of the Parkfield earthquake, these results constrain several physical properties of the rupture process. Second, I use an iterative least-squares approach on both local and global seismic data to isolate source spectra from attenuation and near-station site effects and apply an empirical Green's function correction for near-source attenuation. Brune-type stress drop estimates for small earthquakes in the Parkfield area and for moderate to large earthquakes worldwide range from 0.1 to 50 MPa, with robust lateral variations in average stress drop. In Parkfield some of these variations change temporally after the 2004 M6.0 Parkfield earthquake. On a global scale, the stress drop variations are usually confined to specific plate boundaries or tectonic regimes. The highest average stress drops occur for oceanic transform fault earthquakes and other strike-slip events. Intraplate events have a factor of two higher average stress drops than interplate events. Stress drops of both small earthquakes in Parkfield and larger global earthquakes show no dependence on moment. Our stress-drop estimates, combined with other published results, provide strong evidence that earthquakes exhibit self-similarity over most of the instrumentally observable magnitude range. Finally, using the spectral separation method to discriminate between natural seismicity and man-made explosions, I find that the best single waveform discriminant for data recorded in southern California is the RMS-misfit between quarry blast spectra and an w-2 source model.