| Crosshole ground-penetrating radar (GPR) travel-time tomography is a popular geophysical technique for characterization of the shallow subsurface in environmental applications. With this technique, a critical factor determining the resolution of the velocity images obtained is the angular ray coverage of the subsurface region between the boreholes; when travel-time data representing a narrow range of ray angles are used for the tomography reconstruction, the resulting images contain undesirable directional smearing. Here, I investigate the problem that, even when the crosshole GPR survey geometry offers the potential for high-resolution imaging due to wide angular ray coverage of the inter-borehole region, two significant issues are commonly encountered when attempting to take advantage of this coverage. First, travel times corresponding to high-angle ray paths are often extremely difficult to pick because of low signal-to-noise ratios in the data. Secondly, even when high-angle travel-time data can be reliably determined, they often appear to be incompatible with the lower-angle data available, and tend to cause strong numerical artifacts when included in inversions.; To address the high-angle picking problem noted above, I develop a method for determining first-break times in crosshole GPR data using cross-correlations. High-quality reference waveforms for this technique are obtained from the data through the stacking of common-ray-angle gathers. To address the incompatibility issue with high-angle data, I first develop finite-difference time-domain (FDTD) numerical modeling codes that allow for the determination of realistic crosshole GPR antenna current distributions, and the modeling of transmitted and received waveforms in heterogeneous media. Using these codes, I then find that the high-angle incompatibility issue is likely the result of assuming that first-arriving energy always travels directly between the antenna centers; at high transmitter-receiver angles, this energy likely travels between the antenna tips. Using this knowledge, I develop an improved inversion methodology for crosshole GPR data. In addition to inverting for subsurface velocities, I estimate a small number of parameters that describe a travel-time correction curve as a function of ray angle. I then show the successful application of this improved inversion methodology to synthetic crosshole GPR data, and a data set collected at the Boise Hydrogeophysical Research Site. |
