Global Earthquake Detection with Machine Learning: Exploring Array and Network Based Detection

Session: Leveraging Advanced Detection, Association and Source Characterization in Network Seismology [Poster]
Type: Poster
Date: 4/30/2020
Time: 08:00 AM
Room: Ballroom
Description: 

The U.S. Geological Survey (USGS) National Earthquake Information Center (NEIC) is continuing to develop the ability to leverage deep-learning tools into its operational systems. Initially, models were developed to classify source characteristics of waveforms surrounding standard short-term-average / long-term-average (STA/LTA) picks. Instead of associating picks that only contain arrival-time information, picks with added information, such as phase-type and source-distance, are associated into event detections. This additional information improves hypocenter location estimates and reduces false associations. Still, relying on STA/LTA picks does not increase NEIC’s overall detection abilities. Therefore, NEIC is exploring the use of machine learning detectors to directly detect earthquakes from continuous waveforms. Here, we explore multiple strategies of detection. First, we explore detecting earthquake arrivals using three component data using models trained on a global dataset. Next, we explore leveraging seismic arrays to not only detect events, but also to estimate slowness and back azimuth. Lastly, we attempt to perform network wide detections of surface waves of moderate magnitude events lacking nearby observations, such as earthquakes in the mid-Atlantic. We discuss how each of these tools could be incorporated into global earthquake cataloging, along with potential uses and pitfalls.

Presenting Author: William Yeck

Authors