Using Seismic Ambient Noise to Monitor Environmental Changes

• Main Author: Luo, Bingxu 1995-
• Other Authors: Dabkowski, Mieczyslaw K., Zhu, Hejun, Minkoff, Susan E., Lumley, David, Chen, Feng, Brikowski, Thomas H.
• Format: Thesis
• Language: English
• Published: 2023
• Subjects: Geophysics; Greenland
• Online Access: https://hdl.handle.net/10735.1/9956

Over the past decades, the Earth has suffered from a variety of environmental hazards, such as global warming, floods, drought, ice sheets melting and sea level rise. Scientists would like to monitor, interpret and predict these hazards by using different techniques. As a seismologist, I would like to generalize seismological techniques and apply them to explore these severe hazards. One of my research directions is to measure near-surface seismic velocity changes (dv/v) from ambient noise records, and then correlate them with independent environmental variables. The first topic is to measure decadal dv/v by auto- correlating noise records at each seismographic station deployed in Greenland. Our results demonstrate that dv/v for most stations have less than 3 months lag times in comparison to the surface ice mass change. These various lag times may provide us with constraints for the thickness of the subglacial till layer over different regions in Greenland. Moreover, we also observe a change in the long-term trend of dv/v in southwest Greenland, which is consistent with the mass change rate during the “2012-2013 warm-cold transition”. The second topic is to explore the potential of using regional averaged dv/v records to predict incoming flood events in the Yellowstone National Park (YNP). Over the past ten years, we find that the annual peaks of dv/v variations always have one to two months lead-time in comparison to anomalous water discharges in summers. In particular, there was a short- term high velocity perturbation around 62 days ahead of the 2022 historic flood in the YNP. We therefore suggest that the annual peaks of dv/v in springs are mainly driven by snow accumulation during winters, which might be the major contributor to the flood events in the YNP. This study demonstrates the potential of using seismic observables to predict incoming flood events one to two months in advance. The third topic is to use machine learning (convolutional neural networks) to build a complete aftershock catalog for the ...