Using Machine Learning to Denoise High-Frequency Ocean Bottom Data: Applications to Travel-Time Tomography
Abstract
The Seafloor Observation Network for Earthquakes and Tsunamis Along the Japan Trench (S-Net), installed and maintained by Japan's National Research Institute for Earth Science and Disaster Resilience (NIED), provides dense real-time data from 150 stations spanning the Tohoku Coast. The sheer amount of data that the S-Net network collects has vast implications for subduction zone science. However, the data from S-Net is inherently noisy due to its installation on the ocean floor. To overcome this issue, we have developed a machine learning algorithm that removes characteristic noise from the S-Net data with the goal of improving p-wave picking of teleseismic events. These denoised waveforms are picked and cross-correlated using the method of VanDecar and Crosson (1990), then compared with the results of cross-correlating land-based stations from the High Sensitivity Seismograph Network (Hi-Net), also operated by NIED. We hope to use this data to create a higher-resolution travel-time tomography of the Japan Arc, and compare the structure we observe in Japan with subduction zone structure in Cascadia and Alaska.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.T54D..04D
- Keywords:
-
- 4315 Monitoring;
- forecasting;
- prediction;
- NATURAL HAZARDS;
- 7230 Seismicity and tectonics;
- SEISMOLOGY;
- 8170 Subduction zone processes;
- TECTONOPHYSICS;
- 8488 Volcanic hazards and risks;
- VOLCANOLOGY