This dissertation focuses on the slip history of two well-observed devastating earthquakes: the 2008 Mw 7.9 Wenchuan, China event and the 2011 Mw 9.1 Tohoku-Oki, Japan event. We adopt a non-linear finite inversion code, developed by Ji et al. [2003], to image the kinematic rupture processes of those two earthquakes by jointly modeling seismic waveforms and geodetic measurements.

Our results indicate that the Wenchuan earthquake simultaneously ruptured two sub-parallel faults with about 100 km on the low-angle Pengguan Fault and about 270 km on the high-angle Beichuan Fault. The rupture initiated at the intersection between those two faults and propagated only on the lower potion of the Pengguan Fault in the first 16 s. It then triggered the rupture on the Beichuan Fault about 40 km northeast of the hypocenter, where the Xiaoyudong Fault intersects those two faults on the surface. Although the Xiaoyudong Fault did not experience significant slip during the Wenchuan earthquake, it might act as a local tectonic boundary as inferred from the aftershocks and their focal mechanisms. For the Tohoku-Oki earthquake, our preferred model suggests that it has broken a big near-trench asperity with a peak slip up to 60 m. It re-ruptured the slip region of its March 9th M7 foreshock with a much larger slip, rise time, and slip velocity. One possible scenario model for this kind of foreshock-mainshock sequence is that the foreshock had only ruptured a relatively weaker patch inside a strongly coupled asperity, and the giant Tohoku-Oki event broke the strong asperity and released the deficit slip in the foreshock slip region.

In addition to resolving the slip distributions for realistic earthquakes, we used the earthquake source inversion validation BlindTest 1 to investigate uncertainties associated with earthquake inversions based on strong motion data. Our results suggest that the fault slip and rise time distributions of the BlindTest 1 can be well constrained, but inversions using the variance reduction function of velocity waveforms as the objective function have low sensitivity to the total seismic moment and peak slip. Therefore, it is crucial for source studies to develop new inversion schemes that can properly honor the frequency and time dependent energy distribution of seismic radiation.