Constraining seismic hazard in the Pacific Northwest through observation and direct modeling of earthquakes

Abstract

In this dissertation, we seek to expand knowledge of the processes controlling seismic hazard in the Pacific Northwest through observing and interpreting earthquake records, as well as through directly modeling earthquake source processes. In Chapter 2, we develop a catalog of earthquakes occurring in the vicinity of the Cascadia Subduction Zone using a four-year dataset of seismograms recorded with ocean bottom seismometers (OBS). We locate 271 earthquakes of M0.4-4.0 with epicenters in or directly adjacent to the subduction zone. The distribution of near-interface seismicity shows distinct along-strike variability, which we associate with changes in plate deformation and the amount of subducted sediments. Earthquakes off the coast of Vancouver Island and Northern Washington are typically sparse, coinciding with a relatively undeformed subducting plate and a thin layer of subducted sediments. Near-interface seismicity is relatively abundant off Southern Washington and Northern Oregon, where subduction bending has roughened the lower plate and where entrained sediments are still relatively thin. Seismicity abruptly decreases between Central and Southern Oregon. Despite significant deformation in the subducting plate, seismicity is only clustered in a few distinct swarms. We suggest this is due to a thick layer of largely unconsolidated sediments entering the subduction zone at this location, which encourages geodetically observed partial creeping; the earthquake swarms are likely related to seamounts on the subducting plate, which pierce through the thick sediment sequence to come in contact with the overriding plate. In Southern Oregon and Northern California, seismicity is abundant, corresponding with very little entrained sediments and a subducting slab significantly deformed by the complex plate interactions near the Mendocino Triple Junction. In Chapter 3, we explore the effect of topography on earthquake ground motion during finite fault ruptures by directly modeling M7 earthquakes on the Seattle Fault. Our study focuses on a ~60 X 60 km region centered on the Seattle Fault and simulates seismic wave propagation via a spectral element method (SEM) code. We accurately model ground motion up to 3 Hz using a realistic 3-D velocity model and a model mesh built with a 30 m topographic surface. We ultimately test nine different kinematic rupture scenarios, in which we vary the slip distribution and hypocenter location, to judge the sensitivity of topographic amplification to kinematic rupture parameters. We demonstrate that adding topography to a simulation does not significantly change the average strength of ground motion; however, amplification of shaking is common, with over a quarter of the model area experiencing short period (≤2 s) ground motion 25-35% greater than in a flat simulation. This amplification typically occurs at the top of topographic features, while the bottom of features experience either less amplification or de-amplification. S- and surface waves are more affected by amplification than P-waves. Amplification is sensitive to period, with ground motion experiencing the greatest amplification near a feature’s analytically predicted resonance frequency; shaking is also typically greatest at azimuths perpendicular to a feature’s primary axis of elongation. However, topographic response shows a strong sensitivity to kinematic rupture parameters, particularly at periods <1 s. Our results suggest that while topographic resonance does contribute to the ground motion modeled on a feature, other processes (e.g., localized focusing, improved free surface incidence, and scattering) also play a significant role in determining topographic response. In Chapter 4, we map depth to basement and calculate Qp-Qs relations in the Seattle and Tualatin Basins using converted seismic phases. We identify P-, P-to-S (Ps), S-to-P (Sp), and S-wave phase arrivals in records of earthquakes observed near these basins. The Ps and Sp phases are assumed to represent conversions off of the crystalline basement beneath the basins, as well as at interfaces within the basins. By looking at the relative arrival times between the parent and converted phases, we are able to calculate depth to basement in both the Seattle and Tualatin Basins using a simple velocity model. These depth values are generally similar to those predicted by other active and passive source geophysical methods. In addition, by taking the spectral ratio of the converted and parent phases, we determine the average Qp-Qs relationships within each basin. Through comparisons with linear Qp/Qs relationships, we find that the average Qp value in the Seattle Basin is near 71, while average Qs is near 60 for waves with frequencies between 2-25 Hz. In the Tualatin Basin, the Qp-Qs relationship suggests that average body wave attenuation is higher than in the Seattle Basin.