← Back to Sessions

Training a Convolutional Neural Network to Identify Earthquake Onset in Real-Time GNSS Data

Session: Advances in Real-Time Geophysical Network Operations and Data Analytics [Poster]

Type: Poster

Date: 4/23/2021

Presentation Time: 03:45 PM Pacific

Description: 

Global Navigation Satellite Systems (GNSS) data are an important source of information about earthquakes as they occur, as they do not saturate in the near field with high magnitudes as traditional broadband seismic data do. For this reason, integration of real-time GNSS data into earthquake early warning (EEW) systems is a subject of great interest. However, GNSS data have a much higher noise floor than traditional seismic data, particularly in real-time, making it difficult to identify the early onset signals of earthquakes. For this project, we will develop a machine learning algorithm for identifying earthquakes in real-time GNSS data by modifying an existing convolutional neural network (CNN) code designed for picking P-waves in seismic data. This CNN will be trained using synthetic earthquakes generated by FakeQuakes (Melgar et al., 2016), as well as real-time noise from 237 UNAVCO GNSS stations located within 300 km of the Ridgecrest Sequence. The algorithm will then be tested on UNAVCO position solutions spanning June 2019–July 2020 from the same 237 GNSS stations to see how well it can identify the onsets the major Ridgecrest earthquakes, as well as smaller foreshocks and aftershocks. These position solutions are the same types of data that an EEW system would receive, so the success of this algorithm could have implications on how real-time GNSS data is incorporated into EEW systems.

Presenting Author: Sydney N. Dybing

Student Presenter: Yes

Authors