The occurrence of many large and/or destructive global earthquakes over the past ten years has provided unprecedented seismic, geodetic, and tsunami recordings that reveal complex rupture processes advancing our understanding of earthquake physics. This thesis research has focused on seismological analysis of recent large earthquakes to extract observational insights that address two fundamental questions, “how do great earthquake rupture?”, and “what controls large earthquakes?”. We approach these two questions by providing an improved seismological understanding of large earthquake rupture processes, exploring the variation of kinematic source parameters, and placing the ruptures into the context of tectonic plate motions that drive the deformation.

Given the great diversity of earthquakes, various seismic tools have been explored to give a better robust characterization of large earthquake ruptures. It includes W-phase point source inversion, back projection of seismic array data to map the space-time distribution of high-frequency coherent seismic radiation, determination of broadband source spectra and radiated energy, waveform inversion for co-seismic finite-source slip distribution, and forward modeling of and joint inversion with tsunami and GPS data. By applying these methods, I have studied large events located in different areas, including 1) megathrusts (subduction zone plate boundaries) along the Japan trench, Middle American Trench, and globally; 2) the large transform fault boundary near Scotia-Sea-Antarctic plate boundary, and 3) intraplate events in subducted slabs near the Philippine trench, at intermediate depth (70-300 km) beneath Rat Island earthquake and in the mantle transition zone (300-700 km) beneath Sea of Okhotsk and Ogasawara Islands. The controlling parameters for earthquake-related hazards (e.g. tsunami and strong ground shaking) and earthquake physical mechanisms (e.g. brittle failure, thermal weakening process, stress transfer) have been investigated with an emphasis on the frequency-dependence seismic radiation.

In this thesis, I summarize the research that I have done at UC Santa Cruz involving my development of joint inversion approaches using hr-GPS, teleseismic body and surface waves, regional seismic, campaign GPS, InSAR and tsunami datasets, to investigate the kinematic rupture patterns of large earthquakes. In eight different studies of rupture models of the 2011 Tohoku earthquake, 2012 Indo-Australia earthquake, 2012 Costa Rica earthquake, 2013 Craig earthquake, 2010 Mentawai earthquake, 2013 Pakistan earthquake, 2010 Chile earthquake and 2014 Iquique earthquake, I adopted each available dataset progressively in my joint inversion algorithm, so that in my current approach I can model all of the types of datasets simultaneously. As noted in this thesis, the teleseismic datasets provide good temporal resolution of the rupture process, while geodetic datasets have good spatial resolution. Tsunami datasets have good spatial resolution of slip near the trench. The joint inversion combines the advantage of each dataset, yielding stable and high- resolution rupture models with detailed spatial and temporal information. Resolving a robust and detailed rupture model helps us to understand co-seismic rupture properties, such as depth dependent energy release patterns, super-shear rupture, and tsunami excitation. Comparing the inter-seismic locking pattern and post-seismic stress release pattern with the co-seismic rupture model helps to investigate the locking and releasing behavior of the fault plane through the earthquake cycle, the stress release level of large earthquakes and the relationship between the main shock ruptures, aftershocks and non-seismogenic deformation.

The Tonga subduction zone is among the most seismically active regions and has the highest plate convergence rate in the world. However, recorded thrust events confidently located on the plate boundary have not exceeded Mw 8.0, and the historic record suggests low seismic coupling along the arc. We analyze two major thrust fault earthquakes that occurred in central Tonga in 2006 and 2009. The 3 May 2006 Mw 8.0 event has a focal mechanism consistent with interplate thrusting, was located west of the trench, and caused a moderate regional tsunami. However, long-period seismic wave inversions and finite-fault modeling by joint inversion of teleseismic body waves and local GPS static offsets indicate a slip distribution centered ~65 km deep, about 30 km deeper than the plate boundary revealed by locations of aftershocks, demonstrating that this was an intraslab event. The aftershock locations were obtained using data from 7 temporary seismic stations deployed shortly after the mainshock, and most lie on the plate boundary, not on either nodal plane of the deeper mainshock. The fault plane is ambiguous and investigation of compound rupture involving co-seismic slip along the megathrust does not provide a better fit, although activation of megathrust faulting is responsible for the aftershocks. The 19 March 2009 Mw 7.6 compressional faulting event occurred below the trench; finite-fault and W-phase inversions indicate an intraslab, ~50 km deep centroid, with ambiguous fault plane. This event also triggered megathrust faulting. There continues to be a paucity of large megathrust earthquakes in Tonga.

Near-source path effects imprint the wave field emanating from a seismic source and, if not well resolved, can obscure the details of source characteristics determined from observations of the seismic waves at regional and teleseismic distances (>200 km). These effects are particularly strong for crustal sources such as shallow earthquakes and underground nuclear explosions. First, I explore 2D effects of random seismic P-wave velocity heterogeneity resulting from volumetric heterogeneity in the upper mantle and variability of the Moho on the amplitude decay of the regional phase Pn. Results indicate that the pattern of amplitude decay due to geometric spreading for a simple Earth model is more complex than that for an Earth model containing strong heterogeneity in the mantle lid. Next, I implement the representation theorem in a method which collects displacement and strain components output from a 3D finite difference program capable of including realistic surface topography and geologic structure in a 3D velocity model, and calculates teleseismic 3D Green functions (3DGFs) to specified receiver locations. Green functions produced from a 3D source model match Green functions produced from a 1D source model for theoretical source-receiver geometries. This new method is then applied to the problem of constraining the source depth and location of the three nuclear tests conducted by North Korea, by using a realistic topography model for the mountainous test region to calculate 3DGFs for several possible locations of each event. Amplitude ratios of P and pP from 3DGFs are correlated to those in observed stacked traces. Results show a sensitivity of this method to source depth and location across the test site region with source depths slightly greater than published estimates, but relative locations consistent with other studies. Finally, I determine a rupture model of the 2008 Wenchuan earthquake using 3DGFs calculated in a velocity model containing the dramatic topographic contrast in the fault area, and compare results with a rupture model produced using 1DGFs. Rupture models derived from 1D and 3D synthetic source regions are very similar to each other, and both are consistent with other studies.

Observations of shallow focus megathrust earthquakes are difficult to obtain due to limited instrumentation in the near-source underwater region, and teleseismic observations have limited resolution that is strongly influenced by poorly resolved shallow seismic velocity structures near the toe of the accretionary prism (Lay et al., 2012). Slip under deep water can establish strong water reverberations that persist into seismic coda that is not usually inverted for the source. In this study, P wave coda is examined for many recent major and great earthquakes with independently determined slip distributions and known tsunami excitation to evaluate the prospect for rapidly constraining the up-dip rupture extent of large megathrust earthquakes. For narrowband filtered (7-15 s) data at an 80°-120° distance range, final event average rms Pcoda/rms P (RMS C/P) ratios above 0.610 are found exclusively for events that are believed to have ruptured to shallow depths, while ratios < 0.610 are found for all but one of ruptures that did not extend to the trench. Average RMS C/P ratios computed at all azimuths for each event are larger at azimuths in the direction of the trench for all cases, and events with deep slip beneath continental margins have pronounced reduction of reverberative coda amplitudes in the landward directions. This procedure is successful in separating ruptures that have shallow slip from those that do not for events larger than MW 7.5. It is likely to be very effective for smaller events as well, as their entire rupture area will span a smaller depth range than for many of these events. Given the overall success of the classification using all azimuths, which is the simplest for fast implementation, an optimal azimuthal windowing measure is not necessary. Our specific procedure is of somewhat limited value for assessing likelihood of enhanced tsunami excitation relative to the seismic moment magnitude of the event for very nearby coastlines; however, it could still be valuable for tsunamis with coastal arrival times that are longer than the average 15-18-minute lag time required for measuring numerous relative coda levels in the 80°-120° distance range.

Earthquakes occur suddenly and often are followed with many serious consequences. Interpreting what causes earthquakes and how they respond to outside perturbations is a crucial goal for many geophysical scientists. Our approach to understanding the earthquake nucleation process depends on the amount and quality of data available. Given that the basic physical processes of earthquakes are now fairly well understood and that high-quality geophysical data are being collected, it should be possible to improve our general understanding of how earthquakes are triggered. Seismology is a key geophysical tool that can help us study the Earth’s interior and gain insights into processes that are difficult to observe directly. It also allows us to measure the internal disturbances in the crust where earthquakes nucleate. In addition, hydrological measurements, such as liquefaction, stream discharge, temperature, and turbidity, fluctuations in well water levels, and the eruption of mud volcanoes, have been observed to respond to earthquakes. By analyzing changes in water levels, we can estimate the hydraulic properties of fault damage zones, understand how fluid interacts with faults and quantify poroelastic responses to static stress changes or the dynamic stress associated with seismic waves. Besides, geological measurements also provide a comprehensive view of the history and properties of faults, adding constraints to geophysical data and improving our understanding on a tectonic scale. We use these geophysical tools to study the earthquake-triggering processes, which are challenging to observe directly due to the lack of in situ measurements at subsurface. This dissertation encompasses two seemingly distinct areas: fault maturity and earthquake triggering. These areas of study can ultimately be instrumentally and scientifically connected in understanding faults through various types of observation techniques for dual-purpose research goals as suggested by the title of this dissertation. Chapter 1 explores the relationship between the structure of faults and the seismological aspects of an earthquake rupturing. Earthquakes occur on faults, which are often complex structures. Variations in fault zone maturity have intermittently been invoked to explain variations in some seismological observations for large earthquakes. However, the lack of a unified geological definition of fault maturity makes quantitative assessment of its importance difficult. Here we empirically compare geological and geometrical aspects of strike-slip fault zones and surface ruptures of major events with seismological attributes of the events. We consider factors such as the total offset on the fault that has accumulated over geological time and the number of segments in maps of earthquake rupture and investigate how they correlate with aspects of the resulting earthquake such as the number of aftershocks or the rupture speed. Several of these factors co-evolve as a fault accumulates slip and matures, thus the trends can be interpreted as indicative of the type of earthquakes observed on mature versus immature faults. We find that less mature faults tend to generate more aftershocks and have lower rupture velocity. Energy estimated from seismograms is relatively low for very immature faults, increases with fault evolution and then decreases as maturity further increases. These relationships help elucidate seismic hazard for fault systems of different maturity and delineate the important fault zone factors that have bearing on earthquake rupture and aftershock generation process. In Chapters 2 and 3, this dissertation investigates various types of earthquake triggering, the process through which stress changes associated with an earthquake can either induce or inhibit seismic activity in the surrounding region or trigger other earthquakes over considerable distances. Chapter 2 studies the hydraulic parameters of a fault zone which help understand the triggering mechanisms of a fluid-driven active fault, including either pore pressure changes or poroelastic stresses perturbations to the surrounding rock. Chapter 3 explores dynamic triggering, a rapid form of triggering over large distances explained by the passage of transient seismic waves, which may immediately induce Coulomb-type failure or initiate a secondary mechanism that leads to delayed triggering. Chapter 2 focuses on determining the hydraulic properties (e.g., diffusivity, permeability) which are critical factors determining how fast the fluid can flow and are necessary and useful for studying induced seismicity. Pore pressure diffusion along faults influences induced seismicity and rupture mechanics. However, in situ hydraulic diffusivity measurements along faults are rare and generally lower than inferred from seismicity migration. Here we use the tidal response of deep geothermal boreholes to measure fault diffusivity and permeability. Initial interpretations of the observation with a homogeneous confined aquifer model result in diffusivities of 10-3-10-1 m2/s. However, this model mixed signals from both the conduit and the host rock. We develop a model for tidal response with a fault passing through the aquifer based on the fault-guided fracture network and solve for hydraulic properties in both the fault and the host rock. The resulting fault permeability is 2×10-14-7×10-14 m2 (90% CI) and fault diffusivity is 0.08-0.33 m2/s (90% CI), which is 2 orders of magnitude higher than the host rock diffusivity in some wells, thus highlighting the role of faults as fluid conduits. In Chapter 3, we work on quantifying dynamic triggering which presents a unique opportunity to understand earthquake interactions and associated hazard implications. The extent and timing of dynamic triggering at given specific stress changes still remain inadequately predicted due to limited studies and datasets. In particular, the requirement for complete, well-characterized catalogs to detect triggering systematically seriously limits the types of studies possible. To address this, we utilized 7-year continuous waveform data from 239 stations in southern California and used PhaseNet for phase picking to identify local earthquakes and measure triggering without constructing any earthquake catalog. Our analysis reveals a similar power-law relationship between triggering intensity and peak stress changes as prior works. Spatial mapping of triggering intensity can be affected by the Ridgecrest earthquake and shows variations across the region. We further observe a slow decay rate of dynamic triggering and conclude that low-frequency waves (0.04-0.1 Hz) may be more effective in dynamic triggering than high-frequency waves (1-3 Hz) which is consistent with a rate-state assisted aseismic creep or hydrological triggering mechanism.