← Back to Sessions

Developing Machine-learning-based Seismic Data Processing Tools for a Carbon Storage Site

Description: 

Continuous seismic hazard monitoring is critical for ensuring the safety and security of carbon storage sites. However, the large volumes of data generated by passive seismic monitoring pose significant challenges for conventional processing techniques. Machine learning algorithms offer a promising alternative to streamline labor-intensive and time-consuming workflows. A key limitation in developing such algorithms is the limited labeled data. Transfer learning has proved effective in address this issue, successfully adapting machine learning models to smaller spatial scales, as demonstrated in geothermal applications. Building on this concept, we developed a machine learning-based tool, TL-Picker, to process passive seismic waveforms and accurately extract signal arrival times. Additionally, we created a complementary machine learning tool to locate seismic events by utilizing signal arrival times and seismic station information. Trained with synthetic arrival time data, this machine learning tool achieved performance comparable to a physics-based method. Finally, we implemented an unsupervised learning tool to infer fault planes from seismic event locations. Our suite of tools has the potential to advance the automation and accuracy of seismic monitoring, offering enhanced capabilities for ensuring the safety and operational efficiency of carbon storage sites.

Session: Building and Decoding High-resolution Earthquake Catalogs With Statistical and Machine-learning Tools [Poster]

Type: Poster

Date: 4/15/2025

Presentation Time: 08:00 AM (local time)

Presenting Author: Chengping

Student Presenter: No

Invited Presentation: 

Poster Number: 33

Authors