Understanding Earthquake Source Processes in Real-Time with Seismogeodetic Techniques for Applications in Early Warning

Abstract

Early estimation of strong ground motions is essential for earthquake and tsunami early warning systems. Early warning systems typically rely on seismic recordings, which work well for small-to-moderate events (Mw < 7.5). However, for large magnitude events (Mw ≥ 7.5), seismic systems can saturate, leading to underprediction of magnitude, and they struggle to record the full range of ground displacements. Geodetic observations provide an unfiltered record of ground motion, which can be utilized to improve estimates of magnitude, mechanism, and ground motion for large magnitude events. This dissertation focuses on understanding earthquake source processes with seismogeodetic techniques for applications in early warning. It is made up of a relevant literature review and three original chapters, where the first original chapter “GNSS Velocities in Earthquake Early Warning” consists of three individual projects. First, I explore the applicability of Global Navigation Satellite System (GNSS) velocities in existing earthquake and tsunami early warning systems and post-event response tools. I develop a new distance metric from geodetic ground velocities for use in the geodetic magnitude scaling law of the US earthquake early warning system, ShakeAlert. I demonstrate that ground motions are better characterized with the geodetically derived distance metric and can account for complicated rupture characteristics, thus reducing the urgency for modeling faulting extent in real-time operations. Next, for earthquake rapid response product integration, I create a workflow for including geodetically derived velocities as additional observations into the USGS ShakeMaps module. I showcase the broad applications of the additional data constraints for several major earthquake events globally and find that these additional data constraints generate a better characterization of realized ground motion for the surrounding regions of these events. For tsunami early warning systems, I rapidly estimate focal mechanisms using long period W-phase energy derived from GNSS velocity waveforms. I show the earthquake’s centroid moment tensor, often indicative of if an event with be tsunamigenic or not, and three-dimensional hypocenter can be accurately determined within minutes after rupture when utilizing real-time GNSS velocities. This significantly reduces the timeline for existing tsunami early warning systems. Secondly, I explore machine learning prediction of post-processed quality GNSS displacement time series from real-time velocity time series. In seismology, ground displacement is the ideal metric for characterizing earthquake sources, but seismic observations suffer from filtering and integration issues. GNSS displacements are attainable but contain substantial noise in real time. GNSS velocities are accurate real-time observations, motivating a machine-learning approach to predict post-processed quality GNSS displacement time series from real-time velocity data. I developed a functional model that demonstrates the viability of predicting displacements in real-time and the potential for further noise reduction of those time series. Thirdly, I present a case study of recent moderate sized events around the Shumagin Gap of the Aleutian Trench to assess deformation in a seismically active region with available geodetic and seismic data. I computed co- and post-seismic finite fault slip models for five moderate (Mw > 7) events and explored locking models to evaluate stress release and remaining stress in this region. This showcases the real-world implications of utilizing seismogeodetic observations for recently seismically active regions that are capable of generating devastating earthquakes and tsunamis and provides further details on the stress state of the region for targeted disaster mitigation efforts.