← Back to Sessions

Using Machine Learning for Near Real-time Monitoring in Utah and Yellowstone

Description: 

Machine-learning (ML) has rapidly changed the field of seismology by creating statistical models that can mimic analyst-quality decisions in label-rich scenarios such as signal detection, arrival time picking, and first-motion picking. In particular, ML methods seem to provide substantial value when characterizing periods of enhanced seismicity. To exploit this potential advantage over traditional methods, University of Utah Seismograph Stations has been actively incorporating these ML capabilities into a near real-time monitoring seismic system; Utah Real-Time Seismology (URTS). Since November of 2024, URTS has been monitoring seismicity in both of the UUSS’s authoritative Utah and Yellowstone National Park regions. In this presentation, we will provide a summary of the URTS implementation i.e., the detection, pick refinement, association, and location capabilities.Then, we will compare URTS to our production Advanced National Seismic System Quake Monitoring System (AQMS). Preliminary results from this activity indicate two main conclusions. (1) In Utah, there appears to be no clear benefit of URTS to our AQMS system. (2) In Yellowstone, URTS vastly outperforms our AQMS system in terms of detection and location of small, swarm-like earthquake activity.

Session: Network Seismology: Recent Developments, Challenges and Lessons Learned - II

Type: Oral

Date: 4/15/2025

Presentation Time: 10:45 AM (local time)

Presenting Author: Ben

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors