Characterizing Seismic Events in an Industrial Corridor of the Chicago Area

Urban and industrial environments present a major challenge for seismic event detection and classification. Anthropogenic events can closely resemble tectonic signals and reduce their signal-to-noise ratios, leading to misclassifications of earthquakes and tremors. By detecting and characterizing these anthropogenic events, we can improve detection algorithms and motivate new applications in seismic monitoring and imaging. In this study, we develop a workflow to detect and cluster anomalous seismic events, recorded by a broadband seismic station installed in a unique industrial corridor of the Chicago area. Our workflow consists of (1) a power spectral density (PSD) misfit detector to detect anomalous events in continuous data and (2) a k-means clustering model to generate clusters of anomalous events. We discuss our workflow development, where we select parameters and generalizable model features to maximize the coherence and interpretability of generated clusters. When applied to two years of continuous seismic data, our workflow successfully identified several classes of events, including surface quarry blasts, underground blasts, and machinery operations. Using our results, we created a labeled data set of 1000+ man-made seismic events in the Chicago area. Our study demonstrates how a simple PSD detector and clustering model can be used to efficiently mine through a noisy multiyear data set and create an event catalog.