Loading...
Automating near-surface characterization using ambient noise distributed acoustic sensing data
Punithan, Nikhil
Punithan, Nikhil
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2025
Date Submitted
Keywords
Collections
Files
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
As urban populations grow, geohazards that affect the shallow subsurface increasingly impact more people and infrastructure. These hazards can impact large regions, and occur on a wide variety of time scales. This motivates a large-scale, repeatable time-lapse methods to monitor the subsurface. Typical near-surface seismic exploration methods fall short as land access for geophone deployment may be impossible in developed areas, and active seismic data acquisition can be both logistically challenging and expensive to repeat.
Distributed Acoustic Sensing (DAS) using existing telecommunications infrastructure mitigates the data acquisition problem, as buried fiber optic cables are abundant in dense urban environments. Since these installations are semi-permanent, ambient seismic data can be collected over long time scales, enabling the use of seismic interferometry to retrieve surface wave Green's function approximations. Large-scale near-surface characterization can be conducted using Multichannel Analysis of Surface Waves (MASW) at many locations, followed by stitching together neighboring 1-D shear wave velocity profiles. Since processing and analyzing large amounts of seismic data may be infeasible for small teams of geotechnical engineers and infrastructure managers, an automated workflow can be employed to facilitate large-scale near-surface characterization with limited resources.
In this thesis, I explore an automated approach to near-surface characterization using passive DAS data. I automatically pre-process passive DAS data to remove unwanted coherent signal. I then use seismic interferometry between successive channel pairs to create noise correlation functions along the fiber array. Surface wave dispersion analysis is conducted automatically through a combination of threshold filtering and density-based spatial clustering. Lastly, I invert each dispersion curve for a 1-D shear wave velocity profile to create a pseudo 2-D shear-wave velocity section. By applying this workflow to two dark fiber arrays, I show that large scale shear wave velocity interfaces can be resolved, however clusters of 1-D profiles show unrealistic lateral heterogeneities that can be attributed to challenges in mode picking. I also compare results between two subsets of data from each fiber to show the time-lapse repeatability of the automated processing workflow.
Associated Publications
Rights
Copyright of the original work is retained by the author.