Characterization and Classification of Ambient Seismic Noise in Urban Environments

Abstract

This thesis presents the characterization and classification of ambient seismic noise in urban environments. Urban ambient seismic noise refers to the coherent seismic signal that comes from uncontrolled, or passive sources such as natural (e.g., wind, water waves, tides) and anthropogenic sources (e.g., traffic, pedestrians, machinery). Continuous urban ambient seismic monitoring can track citizen activities such as vehicle and pedestrian traffic, informing municipal planners and policymakers of how the urban environment is being used. As the proportion of the world population residing in urban environments increases, the applications of urban ambient seismic noise detection and monitoring grow. This is particularly evident in the emergence of smart city infrastructure that expands to bring together new types of sensors for monitoring. However, it is challenging to achieve coherent acquisitions due to the spatial, temporal, and spectral characteristics unique to urban environments, therefore new methods for characterization and classification are sought. An event detection workflow for characterizing ambient seismic noise using a pair of Tromino3G+ seismographs is presented, identifying coherent peaks in amplitude in time-frequency space as seismic events. Each event is defined by an amplitude, frequency, source azimuth, duration, and bandwidth. The workflow is applied to ambient noise acquisitions in an urban park environment, characterizing three major groups of events according to their seismic source. Recommendations are made regarding seismograph sampling frequency, sensitivity, and location at the site to inform a potential future long-term seismic monitoring system. The event detection workflow is also applied to a long-term (multi-day) and a short-term (20-minute) datasets in a residential area near a university campus. An unsupervised clustering workflow is applied to the event datasets to find similarities between events to establish an urban seismic noise baseline characterization. The clustering workflow is applied on an ambient seismic noise dataset acquired during a large pedestrian gathering (Queen’s University Homecoming celebrations) and results show that the anomalous activity is detectable using the developed procedure, making unsupervised clustering an effective technique for seismic noise classification. Both the event identification and unsupervised clustering workflows contribute to urban monitoring efforts, optimizing survey design and the interpretation of ambient seismic noise.