← Back to Sessions

Beyond the Waves: Integrating Rock Physical Properties for Deeper Seismic Understanding

Description: 

Technological advances are driving Earth Sciences into the era of big data. Machine learning (ML) techniques offer potential for the advancement of submarine paleoseismology because of the need to integrate large datasets made at various scales with physical properties offering microscale, sediment core offering mesoscale, and geophysical data providing macroscale insights. This data integration is vital for the goal of distinguishing between earthquake- and non-earthquake-generated event beds. At its simplest definition, an event bed can be treated as an anomaly for which anomaly detection ML can be applied to identify similarities or differences for faster analysis and characterization.

IODP (International Ocean Drilling Program) Expedition 386 collected a comprehensive dataset of cores, physical/chemical data, and geophysical profiles in the northern basins of the Japan Trench. Our goal is to test ML capabilities to complement traditional efforts to identify event beds. We focus on Site M0084 where an acoustically thick unit occurs near the top of the sedimentary sequence above a few other potential event beds that occur in otherwise relatively simple stratigraphy. First, we test ML application with synthetic seismograms to link cores to geophysical properties at this site as it is comprised of thick event beds, not prone to resolution issues. Next, we utilize convolution neural networks for anomaly detection of sediment core imagery to refine and incorporate synthetic trace results with the various datasets. In the future, this framework based on physical properties and geophysical data can be extended across the Japan Trench, incorporating more expedition sites for further event bed detection.

Session: From Faults to Fjords: Earthquake Evidence in Terrestrial and Subaqueous Environments - II

Type: Oral

Date: 5/1/2024

Presentation Time: 11:30 AM (local time)

Presenting Author: Rene

Student Presenter: Yes

Invited Presentation: 

Authors