Integrating Machine Learning for Near-real-time Earthquake Monitoring and Public Notification
Description:
An earthquake early warning system is essential for providing timely information to the public during an earthquake. This information is valuable to various organizations, including stakeholders and research teams, where accuracy and speed are crucial. Because machine learning has emerged as a powerful tool for phase picking, event locating, and event discrimination, we developed a near-real-time earthquake monitoring system that uses machine learning for phase picking (PhaseNet) and event discrimination through audio parameter extraction from waveforms using Random Forest (RF) and Support Vector Machine (SVM) models. This study focuses on the evaluation of the method used in the event discrimination (SVM and RF) and discusses the application named SeismicBox2, a mobile application for notifications, as part of the monitoring system.
The system was tested with three months of continuous data (November 2023 to January 2024) from the Madagascar local seismic network. The created models (RF and SVM) were trained, validated, and tested using waveforms from 153 mining/quarry blasts and 2339 earthquake events. Both RF and SVM models achieved an average accuracy of 0.99, with precision, recall, and F1 scores of 0.99 for both quarries and earthquakes. The developed model successfully predicted event nature from the three-month data, leading to the discovery of new mining sites. Meanwhile, the SeismicBox2 application provided comprehensive information to the public, including the earthquake information, waveforms with the arrival times, USGS ShakeMap, exposed population numbers, and optional comments with GPS coordinates. While the system has shown promising results, its performance is limited by the number of monitoring stations. Expanding the network of stations is anticipated to enhance the overall effectiveness of the system.
Session: Advancements in Forensic Seismology and Explosion Monitoring - II
Type: Oral
Date: 4/17/2025
Presentation Time: 10:30 AM (local time)
Presenting Author: Andriniaina Tahina
Student Presenter: Yes
Invited Presentation:
Poster Number:
Authors
Andriniaina Tahina Rakotoarisoa Presenting Author Corresponding Author tahinarisoa@gmail.com Institute and Observatory of Geophysics in Antananarivo |
Hoby Razafindrakoto hobyraza@gmail.com Institute and Observatory of Geophysics in Antananarivo |
|
|
|
|
|
|
|
Integrating Machine Learning for Near-real-time Earthquake Monitoring and Public Notification
Session
Advancements in Forensic Seismology and Explosion Monitoring