Fengzhou Tan

Topic

Exploring Complex Earthquake Sequences through Innovative Automatic Detection and Location Methods

School of Earth and Ocean Sciences

Date & location

• Tuesday, August 13, 2024
• 1:30 P.M.
• Clearihue Building, Room B017

Examining Committee

Supervisory Committee

• Dr. Edwin Nissen, School of Earth and Ocean Sciences, University of Victoria (Co-Supervisor)
• Dr. Honn Kao, School of Earth and Ocean Sciences, UVic (Co-Supervisor)
• Dr. Andrew Schaeffer, School of Earth and Ocean Sciences, UVic (Member)
• Dr. Kwang Moo Yi, Department of Civil Engineering, UVic (Outside Member)

External Examiner

• Dr. Greg Beroza, Department of Geophysics, Stanford University

Chair of Oral Examination

• Dr. Brenda Matthews, Department of Physics and Astronomy, UVic

Abstract

Earthquakes present significant risks to both human lives and infrastructure. Most large earthquakes occur in sequences, with thousands of smaller events after, and sometimes also before, the main event, known as aftershocks and foreshocks, respectively. Precise, comprehensive catalogs for these events are essential for advancing our understanding of the seismogenic processes, regional fault maps, and active tectonics, which form the foundation for effective earthquake study and hazard mitigation efforts. Existing automatic workflows for earthquake monitoring usually lack completeness and accuracy during busy earthquake sequences. Developing automatic methods capable of producing consistently high-quality catalogs in near real-time for major earthquake sequences is crucial. In this dissertation, I develop innovative earthquake sequence observation techniques and apply them to three recent large earthquakes around the world.
In the first project, I improve the Seismicity-Scanning based on Navigated Automatic Phase-picking (S-SNAP) workflow to enable it to delineate the spatiotemporal distribution of dense foreshock and aftershock sequences in real time. Applied to the 2019 M 7.1 Ridgecrest, California earthquake sequence, the S-SNAP catalog usually contains 1.4–2.2 times as many events as the TriNet catalogue, a customized real-time earthquake information system for southern California. In addition, S-SNAP is more likely to solve phase association ambiguities correctly and provide a catalogue with consistent quality through time. Our new catalog details the spatiotemporal evolution of the sequence, including early foreshocks fours days before the mainshock, a subsequent acceleration in foreshock activities, a seismicity gap before the main shock around its epicentre, seismicity on discrete structures within a broad fault zone, and triggered earthquakes outside the main fault zone.

In the second project, I propose a novel approach that utilizes three‐dimensional image segmentation—a computer vision technique—to detect and locate seismic sources, and develop this into a complete workflow, Source Untangler Guided by Artificial intelligence image Recognition (SUGAR). In synthetic and real data tests, SUGAR can handle complex, energetic earthquake sequences in near real time better than skillful analysts and other artificial intelligence (AI) and non‐AI based algorithms. I apply SUGAR to the 2016 M 7.8 Kaikōura, New Zealand earthquake sequence and obtain five times more events than the analyst‐based GeoNet catalog, providing the most complete catalog of the immediate aftershock sequence to date, independent from all existing catalogs. The improved aftershock distribution shows continuous clusters of seismicity under the highly segmented surface ruptures and reveals a connection between well‐studied onshore faults and a hitherto poorly characterized offshore thrust fault, suggesting the possibility of an unusual simultaneous rupture of the onshore and offshore faults. The diffuse nature of the seismicity indicates widespread secondary faulting extending well beyond the normal kilometric‐width damage zones associated with the major faults.

In the third project, I apply SUGAR to the 2023 M 7.8 and M 7.6 Turkey earthquake doublet sequence with a neural network retrained using the regional station coverage. The SUGAR catalog has 2–4 times as many events as the Turkey Disaster and Emergency Management Authority (AFAD) catalog. The result resolves the detailed spatiotemporal distribution of the seismicity, showing the branch initiation of the first mainshock, the separation between the two mainshock ruptures, bifurcated aftershock clusters at the southern end, and the shallow dipping structure at the western end. More complex secondary structures, a single foreshock, and delayed aftershocks are found in and around the rupture zones, shedding lights on the dynamics of the mainshocks.

These innovative methods have yielded more comprehensive and accurate catalogs for complex earthquake sequences, providing essential data for studying earthquake statistics and physics. Broader impacts can be achieved through additional case studies, applications to other seismic sources and data types, interdisciplinary research and close collaboration with hazard management departments and other stakeholders.