Three-dimensional Inversion of High Resolution Ground-penetrating Radar for the Stochastic Structure of Velocity Heterogeneity of a Fluvial Aquifer

Abstract:

Knowledge of the spatial correlation structure of the hydraulic properties in the shallow subsurface is critical for the detailed characterization of near surface aquifers and realistic simulation of flow and transport processes. While the vertical correlation structure is often well constrained by borehole analysis, the lateral correlation structure is often poorly estimated, due to unreliable interpolation between sparsely spaced boreholes and methods that only capture bulk aquifer properties (e.g. pump tests, slug injections). Ground-penetrating radar (GPR) velocity measurements are highly sensitive to changes in water content. In the saturated zone, water content is equivalent to porosity, making ground-penetrating radar a promising tool for extraction of lateral correlation lengths of important hydrologic properties in the near surface. Previous work has shown that there is a simple relationship relating the 2D autocorrelation of a GPR reflection image and the underlying velocity field, allowing for a Monte Carlo inversion strategy to generate sets of parameters describing the autocorrelation of the velocity field, that are consistent with recorded GPR reflection data.

We extend this method to the three-dimensional case and apply it to a realistic synthetic GPR dataset. Results obtained from applying this inversion strategy to the 3-D case are similar to the results from previous work done on the 2-D case, and indicate that the inverse solution is non-unique. Multiple combinations of vertical and lateral correlation parameters describing the velocity heterogeneity can generate GPR reflection images with the same 3-D autocorrelations. Despite the non-uniqueness, the aspect ratios of the extracted parameter pairs are the same, indicating that the aspect ratios of the velocity heterogeneity can be reliably extracted. We are currently applying this method to a densely sampled GPR dataset collected at the Boise Hydrogeophysical Research Site (BHRS). The BHRS is an experimental well field located near Boise, Idaho, consisting of 13 boreholes that provide experimental control.