[Skip to Content]
Banner
Menu
  • Home
  • Submit Abstract
  • Home
  • 2026 Annual Meeting Gallery
  • Machine Learning Detection of Offshore Tremor in the Cascadia Subduction Zone Using Multi-year Continuous Offshore and Onshore Seismic Data

← Back to Gallery

Machine Learning Detection of Offshore Tremor in the Cascadia Subduction Zone Using Multi-year Continuous Offshore and Onshore Seismic Data

In the Cascadia Subduction Zone, tectonic tremor provides important constraints on locked-zone geometry and slip processes along subduction plate interfaces. Current catalogs of tremor in Cascadia do not provide reliable detection in offshore regions, possibly due to either the absence of tremor or to the high levels of noise in ocean-bottom seismic (OBS) data. However, in tectonically similar settings such as the Nankai Trough, abundant updip tremor has been observed. Machine-learning techniques have recently revolutionized the search for earthquakes, but application to tremor remains less common. Machine-learning is thus a possible pathway to detecting tremor in challenging environments such as the oceans. This study adopts a machine-learning–based approach and analyzes multi-year continuous seismic records from 2017 to 2025 to investigate a long-standing yet fundamental question: whether tremor activity is present in the offshore region of the Cascadia Subduction Zone. This study integrates labeled onshore tremor records, onshore earthquake events, and offshore noise-only windows from Cascadia to construct and train a machine-learning framework for a ternary classification task (tremor, earthquake, and noise), using fixed-length windows of 150 s sampled at 100 Hz. The model is trained to differentiate tremor, tectonic earthquakes, and multiple types of oceanic and environmental noise based on their distinct signal characteristics. The network combines convolutional and recurrent architectures with time–frequency representations to jointly capture spectral and temporal characteristics of seismic signals. Model performance is evaluated under varying noise conditions to assess robustness and generalization. The trained network will be applied to long-term continuous OBS records in Cascadia to systematically assess the presence, absence, or potential scarcity of tremor in offshore environments, and to inform us as to the locking state of the Cascadia Subduction Zone.


Session: Linking Subduction Zone Processes and Cascading Hazards in Alaska, Cascadia, Chile and Beyond [Poster]

Type: Poster

Room: Exhibit Hall A+B

Date: 4/16/2026

Presentation Time: 08:00 AM (local time)

Presenting Author: Bochen Dong

Student Presenter: Yes

Invited Presentation: 

Poster Number: 125


Additional Authors

Bochen Dong

Presenting Author

bochen.dong@utdallas.edu

University of Texas at Dallas

Joseph Byrnes

Corresponding Author

joseph.byrnes@utdallas.edu

University of Texas at Dallas

Helen Janiszewski

hajanisz@hawaii.edu

University of Hawaii at Manoa

 

Machine Learning Detection of Offshore Tremor in the Cascadia Subduction Zone Using Multi-year Continuous Offshore and Onshore Seismic Data

Category

Linking Subduction Zone Processes and Cascading Hazards in Alaska, Cascadia, Chile and Beyond

Description