Bayesian full-waveform inversion of tube waves to estimate fracture aperture and compliance

• Published in: Solid earth (Göttingen) 2020-04, Vol.11 (2), p.657-668
• Main Authors: Hunziker, Jürg, Greenwood, Andrew, Minato, Shohei, Barbosa, Nicolás Daniel, Caspari, Eva, Holliger, Klaus
• Format: Article
• Language: English
• Subjects: Acoustics; Algorithms; Aperture; Apertures; Bayesian analysis; Boreholes; Carbon dioxide; Carbon sequestration; Characterization; Compliance; Criminal investigation; Data acquisition; Earth science; Exact solutions; Forward problem; Fractures; Geothermal energy; Hazardous waste management industry; Hydrophones; Inversion; Markov chains; Markov processes; Mechanical properties; Mechanical tests; Monte Carlo methods; Open fractures; Probability theory; Radioactive waste disposal; Radioactive wastes; Setting (Literature); Statistical methods; Stochasticity; Velocity; Waste disposal; Waste management; Wave data; Waveforms
• ISSN: 1869-9529; 1869-9510; 1869-9529
• DOI: 10.5194/se-11-657-2020

The hydraulic and mechanical characterization of fractures is crucial for a wide range of pertinent applications, such as geothermal energy production, hydrocarbon exploration, CO2 sequestration, and nuclear waste disposal. Direct hydraulic and mechanical testing of individual fractures along boreholes does, however, tend to be slow and cumbersome. To alleviate this problem, we propose to estimate the effective hydraulic aperture and the mechanical compliance of isolated fractures intersecting a borehole through a Bayesian Markov chain Monte Carlo (MCMC) inversion of full-waveform tube-wave data recorded in a vertical seismic profiling (VSP) setting. The solution of the corresponding forward problem is based on a recently developed semi-analytical solution. This inversion approach has been tested for and verified on a wide range of synthetic scenarios. Here, we present the results of its application to observed hydrophone VSP data acquired along a borehole in the underground Grimsel Test Site in the central Swiss Alps. While the results are consistent with the corresponding evidence from televiewer data and exemplarily illustrate the advantages of using a computationally expensive stochastic, instead of a deterministic inversion approach, they also reveal the inherent limitation of the underlying semi-analytical forward solver.