Crosswell traveltime tomography in three dimensions

Crosswell direct arrival traveltime tomography has the potential to provide very high resolution images of the subsurface. The conventional approach, however, results in generally poor resolution for typical oil and gas well spacings (25 to 250 meters). The vast majority of published results indicate vertical resolution no better than 5 meters.; Evidence of limitations to resolution and uniqueness that are a consequence of the modeling scheme employed by the conventional unconstrained approach can be demonstrated by application of the Fourier Projection Slice Theorem. This conventional modeling---that implicitly assumes two vertical boreholes and a planar interwell region---further limits applicability to the real world of deviated wells and non-trivial out-of-plane structures.; This purpose of this dissertation is to introduce a new strategy for crosswell traveltime tomography that enables high vertical resolution (sub-meter) and provides a modeling framework that can handle fully three-dimensional problems.; The major change from the conventional approach involves model re-parameterization from pixel-based basis functions to discontinuous vertical layering with a 2-D Chebyshev polynomial representation. High vertical resolution is attained by using a small number of parameters for horizontal variability, consistent with the expected maximum vertical and horizontal resolution available from the data.; In addition to model re-parameterization, this approach limits the data used in the inversion to raypaths that do not travel nearly parallel to the geologic layering. In conventional tomography, these horizontal raypaths are essential for high vertical resolution. However, this research finds that for small vertical source-receiver separations there is a large inconsistency between forward modeling and traveltimes calculated from wave-modeled synthetic data. The highest resolution and most accurate inversions are therefore achieved by excluding these small vertical source-receiver separations.; Two of the key advantages of this 3-D modeling framework are the ability to handle 3-D geometries (e.g. steep dips and severe borehole deviations), and the ability to use data from multiple profiles simultaneously to recover a 3-D velocity distribution. This approach was used on both real and synthetic data, and attained sub-meter vertical resolution at typical oil and gas well spacings.