Heimdall: A Graph-Based Seismic Detector and Locator for Microseismicity
Description:
The application of Machine Learning (ML) has significantly transformed traditional tasks in observational seismology such as phase picking and association, earthquake detection and location, and magnitude estimation. Despite progress, the use of ML-based classical workflows in microseismic data analysis still poses challenges. We present HEIMDALL, a grapH-based sEIsMic Detector And Locator for microseismicity, powered by the latest advancements in Deep Learning (DL) methodologies. HEIMDALL utilizes an attention-based, spatiotemporal graph-neural network for detecting seismic events. It also implements a waveform-stacking approach for event location, leveraging output probability functions over a dense regular grid. We tested HEIMDALL using a month-long waveform dataset from December 2018, obtained during the COSEIMIQ project (active from December 2018 to August 2021) at the Hengill Geothermal Field in Iceland. This dataset is optimal for testing seismic event detection and location algorithms due to its high seismicity rate (over 12,000 events in about two years) and the presence of burst sequences with very short interevent times (e.g., less than 5 seconds).
We evaluated HEIMDALL's performance by comparing the catalog we generated with those produced by two different DL methods, as well as one manually created by ISOR for the same timeframe. The DL algorithms tested included (i) MALMI, a waveform-based location algorithm, and the recent (ii) GENIE graph-neural-network algorithm. To optimize GENIE for our specific study area, we repicked continuous waveforms using the PhaseNet algorithm and trained a new model. We discuss the advantages and limitations of each method and explore potential enhancements for detecting and locating microseismic events, with a focus on monitoring induced seismicity at Enhanced Geothermal System (EGS) sites.
Session: Seismic Monitoring, Modelling and Management Needed for Geothermal Energy and Geologic Carbon Storage - III
Type: Oral
Date: 5/2/2024
Presentation Time: 02:30 PM (local time)
Presenting Author: Matteo
Student Presenter: No
Invited Presentation:
Authors
Matteo Bagagli Presenting Author Corresponding Author matteo.bagagli@dst.unipi.it University of Pisa |
Francesco Grigoli francesco.grigoli@unipi.it University of Pisa |
Davide Bacciu davide.bacciu@unipi.it University of Pisa |
|
|
|
|
|
|
Heimdall: A Graph-Based Seismic Detector and Locator for Microseismicity
Session
Seismic Monitoring, Modelling and Management Needed for Geothermal Energy and Geologic Carbon Storage