Advancements in seismic tomography with application to tunnel detection and volcano imaging

Abstract

Practical geotomography is an inverse problem with no unique solution. A priori information must be imposed for a stable solution to exist. Commonly used types of a priori information smooth and attenuate anomalies, resulting in 'blurred' tomographic images. Small or discrete anomalies, such as tunnels, magma conduits, or buried channels are extremely difficult imaging objectives. Composite distribution inversion (CDI) is introduced as a theory seeking physically simple, rather than distributionally simple, solutions of non-unique problems. Parameters are assumed to be members of a composite population, including both well-known and anomalous components. Discrete and large amplitude anomalies are allowed, while a well-conditioned inverse is maintained. Tunnel detection is demonstrated using CDI tomography and data collected near the northern border of South Korea. Accurate source and receiver location information is necessary. Borehole deviation corrections are estimated by minimizing the difference between empirical distributions of apparent parameter values as a function of location correction. Improved images result. Traveltime computation and raytracing are the most computationally intensive components of seismic tomography when imaging structurally complex media. Efficient, accurate, and robust raytracing is possible by first recovering approximate raypaths from traveltime fields, and then refining the raypaths to a desired accuracy level. Dynamically binned queuing is introduced. The approach optimizes graph-theoretic traveltime computation costs. Pseudo-bending is modified to efficiently refine raypaths in general media. Hypocentral location density functions and relative phase arrival population analysis are used to investigate the Spring, 1996, earthquake swarm at Akutan Volcano, Alaska. The main swarm is postulated to have been associated with a 0.2 km$\sp3$ intrusion at a depth of less than four kilometers. Decay sequence seismicity is postulated to be a passive response to the stress transient caused by the intrusion. Tomograms are computed for Mt. Spurr, Augustine, and Redoubt Volcanoes, Alaska. Relatively large amplitude, shallow anomalies explain most of the traveltime residual. No large amplitude anomalies are found at depth, and no magma storage areas are imaged. A large amplitude low-velocity anomaly is coincident with a previously proposed geothermal region on the southeast flank of Mt. Spurr. Mt. St. Augustine is found to have a high velocity core.