GPR waveguide and full-waveform inversion for the hydrogeophysical characterization of soils and aquifers

Abstract:

Ground-penetrating radar (GPR) is a powerful tool for characterizing hydrologic processes. Coupled hydrogeophysical inversion can be used to invert time-lapse GPR data to obtain soil hydraulic properties. However, the inversion will fail if the hydrological model is not appropriately conceptualized, or when the ray-based methods that are often used to limit the computation time are not appropriate. Ray-based approaches cannot be used in the case of wetting fronts in the vadose zone due to precipitation/infiltration or thawing events, where low-velocity layers with high water content can trap the GPR waves and act as a waveguide such that multiple internal reflections cause dispersion. Utilizing the dispersion, manifested as a singled appearance of arrivals on multi-offset data, we can invert for waveguide properties. Single- or two-layer waveguide inversion approaches return average water contents but are incapable of representing the gradational nature of the water content distribution of the shallow subsurface. Recently, a shuffled complex evolution algorithm was implemented that used a piece-wise linear function to closely match the shallow gradational water content profile present for early-time infiltration events for different n-parameter values of the Mualem–van Genuchten equation using HYDRUS-1D. Similar waveguide phenomena can arise in the presence of high porosity layers in saturated aquifers for crosshole GPR. These high porosity layers result in significant late arrival high-amplitude elongated wave trains that were detected in crosshole data from the Widen, Boise and Krauthausen aquifers. Full-waveform inversion of these data that uses an accurate forward model to explain the measured data is able to return decimeter scale resolution of the porosity based on the GPR velocity. With increasing computer power, it is expected that these advanced processing can be soon included in the coupled inversion approaches.