Skip to main content
Loading Events

« All Events

  • This event has passed.

Spencer Wilbur Thesis Defense

May 18 @ 12:00 pm - 1:00 pm MDT

Spencer Wilbur

Machine Learning Reveals Aftershock Locations for Three Idaho Earthquake Sequences

I explore spatial and temporal aftershock patterns related to three instrumentally recorded earthquakes within Idaho. Bordering the eastern extent of the Snake River Plain, within the Intermountain Seismic Belt of Idaho, I compare aftershock sequences related to the Sulphur Peak, Challis, and Stanley earthquakes. Using machine learning, I locate low magnitude events with low signal to noise ratio, enhancing the spatiotemporal distribution of the aftershock sequence for each region. To assist hypocenter location, I test a range of velocity models to determine the best hypocenter locations and lowest misfit. I then observe the results of an automated detection and phase picking algorithm in conjunction with appropriate velocity models to compare them to traditional detection methods and handpicked phases. This increases in the number of detected events compared to the USGS catalogs from: 1,946 to 52,125 for Stanley, 551 to 2,170 for Sulphur Peak, and 189 to 2,845 for Challis. The results for the Challis sequence reflect the work of Pang et al., (2018) but automated detection provides an increase in the number of high-quality events while maintaining the same average vertical error. The automated catalog of aftershocks for Sulphur Peak reflects the results of Koper et al., (2018), while providing insight into the machine learnings inability to detect aseismic driven seismicity. The results for Stanley show that the mainshock that occurred on March 31st, 2020, did not occur on the Sawtooth fault but a NS trending 30km long normal fault with a westward dipping plane. Using the relationship between subsurface rupture length and estimated magnitude by Wells and Coppersmith (1994), I support this claim. The moment tensor solutions from the Saint Louis University moment tensor determinations database are relocated using the machine learning locations. In doing so I show that left-lateral strike-slip with oblique thrusting is a driving mechanism behind seismicity in the Sawtooth Valley. This suggests that left-lateral motion could be responsible for greater right lateral extension, extending southwest of Sawtooth Fault, within the Trans-Challis-Fault System, to the northern extent of the Beaverhead fault (Payne et. al., 2013). The resultant catalogs and interpretation show by using machine learning to detect and pick seismic phases, paired with an adequate local (<50km from epicentral region) seismic network, one can enhance seismic detection and aid in determining the driving mechanisms responsible for coseismically driven earthquake sequences.

Advisor: Lee Liberty

Co-Advisors: Dylan Mikesell, Blaine Bockholt