← Back to Sessions

Unsupervised Spectral Clustering and Spectral Ratio Analysis of Earthquakes in Cushing, Oklahoma

Description: 

Oklahoma has been at the center stage for the increase in midcontinent seismic activity during the last 15 years due to the production of oil and gas and the associated injection of fracking fluids and wastewater disposal. While there were notable large earthquakes, such as the M5 Cushing and Pawnee events, there are a vast number of smaller events; the spectral characterizing of these events can provide valuable insight into faulting mechanisms, which can lead to a better understanding of how fluid injection impacts the evolution of the subsurface. Two hundred fifty small earthquakes were detected using a combination of machine learning and beamforming algorithms with a dense nodal array composed of 130 three-component stations deployed in Cushing, Oklahoma, from November 5th, 2019, to December 15th, 2019. Here, we use SpecUFEx to perform the unsupervised spectral clustering of these earthquakes to investigate their spectral content's spatial and temporal variations. SpecUFEx uses non-negative matrix factorization and hidden Markov modeling to create a fingerprint of an earthquake's normalized spectral content; these fingerprints are then used as features to perform k-means clustering of the events.

Analysis of the spectral content centered around the P-wave arrival on the vertical component of all stations in the array using SpecUFEx identified two spectral clusters. One cluster is enriched in energy across a higher and broader frequency band; the other cluster appears narrow-banded. The main Cushing fault exhibits events only in the broad cluster enriched in high-frequency content, while two faults outside the array are almost entirely narrow-banded. The clustering results suggest that the spectral content recorded at the array stations may be sensitive to the hosting fault structure or the event's path to the array. To determine the path effect on the clustering results obtained from SpecUFEx, we will perform further analysis by exploring the spectral ratio of the events in each cluster and calculating the empirical Green's function to obtain the corner frequency and stress drop.

Session: Innovative Applications of Seismic Nodal Technology for Hazard Mitigation and Earth System Monitoring [Poster]

Type: Poster

Date: 4/15/2025

Presentation Time: 08:00 AM (local time)

Presenting Author: Braden

Student Presenter: Yes

Invited Presentation: 

Poster Number: 13

Authors