← Back to Sessions

Automated Earthquake Detection and Location Applied to Local-Scale Seismic Monitoring

Description: 

Comprehensive catalogs of microseismicity are crucial to effective monitoring of induced seismicity. Due to the large quantity and low signal-to-noise of microseismic events, maintenance of such catalogs is labor-intensive. Recent advances in machine-learning workflows for detection and location have shown promise as a potential solution to this issue. Here, we use the BPMF package (backprojection and matched filtering, Beaucé et al., 2023) in concert with a pre-trained convolutional neural network model (Leifer et al., AGU 2022) to automatically detect and locate microseismic events induced by hydraulic fracturing within a local-scale seismic monitoring array. The array consists of 9 wideband seismometers installed in shallow postholes with approximately 1 km spacing. Conventional operations consisting of automatic, real-time processing in Earthworm, manual review, and template matching provide the basis for this deployment, producing a catalog of 1375 events in the upper 5 km spanning magnitudes from M-0.94 to M2.8 over the course of 32 days. The machine learning model used in this study was trained on a proprietary dataset containing data from seven local-scale arrays with similar characteristics. In comparing the automatically generated catalog to our conventional operations, we find that the machine learning workflow recovers more than 90% of the manually reviewed events. However, locations for the recovered events show higher residuals when compared to the manual locations due to small inaccuracies in the arrival times produced by the BPMF workflow. Additional fine-tuning of the parameters used in backprojection and matched-filtering in conjunction with improvements to the pre-trained picking model should enhance both the percentage of recovered events and location quality. These results demonstrate the potential of automated machine learning workflows to efficiently generate valuable microseismic catalogs in near real-time. Future applications of this workflow will aim to improve seismic monitoring capabilities of enhanced geothermal systems and carbon sequestration projects.

Session: Seismic Monitoring, Modelling and Management Needed for Geothermal Energy and Geologic Carbon Storage [Poster Session]

Type: Poster

Date: 5/2/2024

Presentation Time: 08:00 AM (local time)

Presenting Author: Alex

Student Presenter: No

Invited Presentation: 

Authors