Waveform tomography and its application at a ground water contamination site

This thesis develops and applies seismic waveform tomography to solve the unique problem of imaging complicated shallow sub-structures with high resolution. Shallow sub-structures are commonly characterized by seismic reflection/refraction imaging, georadar and seismic travel time tomography (e.g., Steeples, 1998; Carcione et al., 2000 and Azaria, 2002). Their resolving power or applicability is often limited. In contrast, waveform tomography, a full wave field inversion technique, resolves sub-structures at a resolution that is a fraction of the illuminating wavelengths. Forward modeling in waveform tomography is based on a finite difference solution to the acoustic wave equation in the space-frequency domain. During inversion for model parameters, the technique efficiently calculates the gradient of a misfit function with respect to model parameters by correlating back-propagated and forward modeled wave fields, avoiding the forbidding task of explicitly computing Frechet kernels. Part of this study compares travel time and waveform tomography in a synthetic cross-well test. The two tomographic approaches are found to be complementary if data contains no significant low frequency spectra. I then apply waveform tomography to two datasets from a ground water contamination site at the Hill Air Force Base (HAFB) to sample formation heterogeneities and to map the 3D geometry of a buried paleo-channel where DNAPLs (Dense Non-Aqueous Phase Liquids) were dumped. The first is a VSP-surface seismic experiment. The final velocity model from waveform tomography applied to the VSP dataset generally correlates well with lithology logs, depth migrated 2D/3D reflection data and a velocity model from 3D travel time tomography. Large velocity variations vertically and laterally (200m/s) occur in a distance as short as ∼1m. The model is interpreted geologically and petrologically. Scale features down to ∼1.5m were recovered. I then apply waveform tomography to 45 2D seismic profiles extracted from a 3-D surface seismic experiment at HAFB, and recover the 3D geometry of a buried paleo-channel acting as a trap for DNAPLs. By combining the identified cross-sectional geometry, the 3D geometry of the channel is reconstructed. The subsurface map could be used to plan injection/extraction well placements with good precision and low cost in the on-going ground water remediation program