Seismic Imaging of Underground Tunnels using Full-waveform Inversion

Abstract

There are a variety of challenges when applying full-waveform inversion to underground tunnels in the near-surface. One challenge is in efficiently handling large amounts of seismic data on high-performance computing machines. Big data-sets allow for more measurements to be made that can resolve structures at unprecedented resolution. Utilizing these data-sets in seismic inversion work-flows requires developing new technologies that are fast, adaptable, and reproducible. The adaptable seismic data format (ASDF) is a modern data format that offers an organized, standards-based, and provenance-enhanced tool for earthquake seismology.

A second challenge is generating accurate three-dimensional images of small subsurface voids near the surface. In advancing seismic imaging in the near-surface regime, three-dimensional elastic full-waveform inversion is applied to a known underground tunnel located in a desert geology at a depth of approximately 10 m. A set of successful experiments are presented that approximately image the underground tunnel in three-dimensions.

A third challenge is in estimating the source time function from field data. Since the recorded data includes the effect of the source, the more accurate the estimate of the seismic source, the more accurate the model of the inverted subsurface velocity structure. A linear inversion method is applied for inverting for the source time signature from recorded sensor data. The inverted source is verified by comparing it with the measured accelerometer data from the field site.

A fourth challenge is in incorporating anelastic attenuation into the inversion workflow. The range of possible parameter values for attenuation are constrained using the recorded data from the tunnel site. A set of 2D numerical experiments are presented that invert for attenuation directly in acoustic and elastic media.