← Back to Sessions

Implementation and Testing of EQTransformer to Detect Microseismicity Near the Alpine Fault, South Island, New Zealand

Description: 

The Alpine Fault in South Island, New Zealand poses a significant hazard to people and infrastructure in New Zealand. The likelihood of a large earthquake in the next 50 years is approximately 75%. Hazard analyses typically assume that the seismogenic thickness measured interseismicaly corresponds to the maximum depth of rupture of moderate and large earthquakes. Having reliable information on background microseismicity, and therefore of the seismogenic thickness, is important to accurately determine potential rupture areas through the Alpine Fault.

More than 80 permanent and temporary seismic stations are currently operating near the Alpine Fault. This large volume of data provides the basis for a new generation of earthquake catalogs. However, processing all data manually would be very time-consuming, and would likely result in a biased and inconsistent catalog. EQTransformer, a deep learning model used for earthquake signal detection and seismic phase picking, has proven effective and efficient at detecting microseismicity in diverse locations worldwide, so we use this model to process all data from the Alpine Fault.

To ensure that we develop a robust and highly-complete catalog using EQTransformer, we have performed a quantitative analysis of EQTransformer’s parameters and overall performance by comparing its output against a high-quality, long-duration microseismicity catalog from the Alpine Fault. We find that due to the prevalence in the training data set of P-arrivals occurring within 5–15s of each seismogram’s start-time, it is important to use large overlap lengths when applying EQTransformer to continuous data. We also find that the majority (92%) of EQTransformer picks are within 0.25s of the manual picks. EQTransformer is a recent advancement, and testing the effects of different parameters against existing high-quality picks will be key for obtaining high-quality microseismicity catalogs.

Session: Opportunities and Challenges for Machine Learning Applications in Seismology [Poster]

Type: Poster

Date: 4/19/2023

Presentation Time: 08:00 AM (local time)

Presenting Author: Olivia D. Pita-Sllim

Student Presenter: Yes

Invited Presentation: 

Authors