Unsupervised Clustering of Microseismic Events and Focal Mechanism Analysis at the CO 2 Injection Site in Decatur, Illinois

• Published in: Journal of geophysical research. Machine learning and computation Vol. 2; no. 3
• Main Authors: Willis, Rachel M., Yoon, Hongkyu, Silva, Josimar A., Juanes, Ruben, Williams‐Stroud, Sherilyn, Frailey, Scott M.
• Format: Journal Article
• Language: English
• Published: 01.09.2025
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000544

Characterization of induced microseismicity at a carbon dioxide () storage site is critical for preserving reservoir integrity and mitigating seismic hazards. We apply a multilevel machine learning (ML) approach that combines the nonnegative matrix factorization and hidden Markov model to extract spectral representations of microseismic events and cluster them to identify seismic patterns at the Illinois Basin‐Decatur Project. Unlike traditional waveform correlation methods, this approach leverages spectral characteristics of first arrivals to improve event classification and detect previously undetected planes of weakness. By integrating ML‐based clustering with focal mechanism analysis, we resolve small‐scale fault structures that are below the detection limits of conventional seismic imaging. Our findings reveal temporal bursts of microseismicity associated with brittle failure, providing insights into the spatio‐temporal evolution of fault reactivation during injection. This approach enhances seismic monitoring capabilities at injection sites by improving fault characterization beyond the resolution of standard geophysical surveys. Understanding how microseismic activity occurs during carbon dioxide () injection is important for monitoring subsurface changes, ensuring the stability of the storage site, and reducing the risk of induced seismic events. In this study, we use machine learning (ML) to analyze patterns in microseismic data collected at the Illinois Basin‐Decatur Project (IBDP). At the IBDP, one million metric tons of was injected over a three‐year period for the demonstration project. Our approach applies ML techniques to extract key features from seismic signals, reducing complex waveform data into simplified representations (or “fingerprints”). By clustering these fingerprints, we identify different patterns of microseismic activity linked to fault reactivation during injection. Our results show that both the size and timing of seismic events influence how they are grouped, revealing distinct subsurface responses to injection. We also analyze the slip behavior of these microseismic events to better understand fault movement. Using a combination of event characteristics, such as magnitude, timing, and faulting style, we develop a model of small‐scale fault structures that provides new insights into how injection‐induced seismicity evolves over time. This method enhances the ability to quickly and accurately interpret geophysical sensing data, improving monitoring and risk assessment for storage projects. Microseismic data reveal different types of subsurface poroelastic responses due to injection Characteristics of small‐scale faults can be estimated by applying spectral unsupervised feature extraction and focal mechanism analysis Unsupervised machine learning has the potential to enhance imaging of hidden faults during injection