← Back to Gallery

Session: Filling the Data Gap: Ocean-bottom Sensing with Fiber-optic Cables [Poster]

Type: Poster

Date: 10/9/2024

Time: 07:00 AM

Room: Stanley Park Ballroom

DeepDAS: Earthquake Phase Picker from Submarine Distributed Acoustic Sensing Data

The EU-INFRATECH funded SUBMERSE project will establish continuous monitoring of several oceanic telecom cables for landing sites in Portugal, Greece, and Svalbard. We develop tools for the automated analysis of these upcoming data sets as well as other submarine DAS data. Employing DeepLab v3, a cutting-edge deep neural network architecture renowned for its exemplary performance in semantic segmentation, this project aims to develop a specialized machine learning model for the detection of earthquakes and the identification of P and S waves using submarine DAS data. The inherently two-dimensional nature of our input data mandates the adoption of DeepLab v3.

Confronted with the distinctive challenges posed by submarine DAS data, which encompass varied oceanic noise environments and a spectrum of operational parameters including cable length, configuration, channel spacing, deployment conditions, and geographic diversity, we have chosen to implement a more expansive model to enhance detection capabilities. This project exclusively utilizes submarine DAS seismic records, now encompassing nearly three million earthquake records from multiple international locales. This approach ensures that our model is more effective to the unique characteristics of the submarine environment.

Our findings corroborate the model’s good capacity to detect seismic events and accurately categorize P and S waves using single-component multi-channel DAS data—a notable feature considering the absence of the three components typically essential for wave discrimination in conventional seismology. The model proficiently differentiates between P and S waves, even in scenarios where the order of these waves is reversed or their amplitudes are markedly altered, indicating that the model discerns complex patterns and features inherent in seismic data beyond mere amplitude analysis.

Presenting Author: Han

Additional Authors