Crosshole Radar Tomography in a Fluvial Aquifer near Boise, Idaho

To determine the distribution of heterogeneities in the saturated zone of an unconfined aquifer in Boise, ID, we compute tomograms for three adjacent well pairs. The fluvial deposits consist of unconsolidated cobbles and sands. We used a curved-ray, finite-difference approximation to the eikonal equation to generate the forward model. The inversion uses a linearized, iterative scheme to determine the slowness distribution from the first arrival traveltimes. The tomograms consist of a layered zone representing the saturated aquifer. The velocities in this saturated zone range between 0.06 to 0.10 m/ns. We use a variety of methods to assess the reliability of our velocity models. Finally, we compare our results to neutron-derived porosity logs in the wells used for the tomograms. The comparison shows that the trends in porosity derived from the tomograms match the trends in porosity measured with the neutron probe. distribution. By comparing our results to neutron- derived porosity estimates from the wells used in the tomography experiment, we can use this sensitivity to porosity to validate the tomography model. Crosshole tomography provides an image of the porosity distri- bution in the subsurface. To determine the velocity distribution, we first establish that the tomograms are consistent with each other by inverting three well pairs that form a cross- section of the aquifer and comparing the results along the common wells. An important aspect of this study is appraising the solution. We look at the data residual distribution and approximations to the diagonal ele- ments of the resolution and covariance matrices. Because water strongly controls the EM velocity, we can relate the EM velocity to the porosity distribution in the saturated zone. We show the strong correlation between water saturated porosity and the EM velocity, finally displaying the radar-derived porosity section.