← Back to Sessions

Building an Enhanced Earthquake Catalogue for Aotearoa New Zealand: Applying an Automated Workflow With Cutting-Edge Machine Learning Methods to Mine New Zealand’s Seismic Data

Description: 

The New Zealand nationwide seismic catalogue produced by GeoNet is primarily focused on ensuring fast hazard communication and response. The focus on hazard response, combined with the inconsistent nature of the New Zealand seismograph network, can mean events are mislocated or not located at all. We propose and develop a cutting-edge automated workflow to generate research-grade earthquake catalogues for Aotearoa New Zealand and have begun to apply this to New Zealand's seismic data. This workflow expands and improves the current catalogue, enabling further research into New Zealand's tectonic setting and a better understanding of present-day seismic hazard. We apply uniform identification, classification and location methods to the entire catalogue, reducing errors and uncertainties introduced by inconsistencies in the current catalogue.

We use the now well-tested EQTransformer AI seismic picker to efficiently and accurately detect and pick events in a uniform manor. We locate all events using the NZWide 2.3 3D Velocity Model (Eberhart-Phillips et al., 2022), producing accurate locations and robust uncertainties. We present the results of initial testing on five distinct regions around Aotearoa as well as the first years of the enhanced catalogue we have produced. We find differences in manual and automatic picking to be negligible in determination of final location for most events and that the improved velocity model with the removal of depth fixing has a much larger influence on final location. In all test regions, most events locate deeper than initially estimated by GeoNet, though smaller changes in location are observed in more densely monitored regions with less complex velocity structures. Future work will extend this cataloguing through time to develop a self-consistent high-quality earthquake catalogue that will have enhanced capabilities for capturing precursor signals important for earthquake forecasting across New Zealand.

Session: Towards Advancing Earthquake Forecasting and Nowcasting: Recent Progress Using Ai-Enhanced Methods [Poster Session]

Type: Poster

Date: 5/1/2024

Presentation Time: 08:00 AM (local time)

Presenting Author: Codee-Leigh

Student Presenter: Yes

Invited Presentation: 

Authors