It Is Not Our Fault Yet: Multi-Attribute and Machine Learning Study For Improved Upper Basement Fault Detection In An Area Of Carbon Capture Utilization and Storage: Decatur, Illinois, USA

Abstract

Identifying and interpreting faults is critical in many geological prospects, from traditional well planning and operations to energy transition ventures such as Geothermal and carbon capture, utilization, and storage (CCUS). Effective geohazard analysis can reduce the operation time, cost, and uncertainty. More importantly, with new technologies such as CCUS, where public perception and support are sensitive, it is important to decrease analysis uncertainty and prevent large-scale reactivation events.To improve the accuracy of fracture delineation, a machine learning (ML) and multi-attribute approach was employed in Decatur, Illinois— where microseismicity has been induced in the rhyolitic basement related to CCUS. Due to the poor seismic imaging resolution of the basement and the potential presence of sub-seismic faults, traditional geometric attributes (e.g., coherence and curvature) are not sufficient alone. Structure-oriented filter (SOF) is utilized before the application of the multi-attributes which resulted in increased fault confidence and connectivity. Seismic attributes candidates are gray level co-occurrence matrix entropy (GLCM), fault enhancement of energy ratio similarity (ERS), most positive curvature (k1), most negative curvature (k2), and aberrancy. For ML, a pre-trained convolutional neural network (CNN) was utilized while generative topographic mapping (GTM) was calculated using the aforementioned candidate attributes. Our findings show ERS and CNN can effectively map larger-scale fault patterns where curvature and aberrancy are better for faults seen as flexures or folds. However, only some faults detected from ML and attributes coincided with the location of microseismic events. This is most likely due to a combination of factors such as the poor quality of seismic image at the basement, high viscoelastic attenuations at deep depths, and small vertical displacement of sub-seismic faults.