Resolving precipitation-induced water content profiles through inversion of dispersive GPR data

Abstract:

Ground-penetrating radar (GPR) has become a popular tool for monitoring hydrologic processes. When monitoring infiltration, the thin wetted zone that occurs near the ground surface at early times may act as a dispersive waveguide. This low-velocity layer traps the GPR waves, causing specific frequencies of the signal to travel at different phase velocities, confounding standard traveltime analysis. In a previous numerical study we demonstrated the potential of dispersion analysis for estimating the depth distribution of waveguide water contents. Here, we evaluate the effectiveness of the methodology when applying it to experimental time-lapse dispersive GPR data collected during a laboratory infiltration experiment in a relatively homogenous soil.

A large sand-filled tank is equipped with an automated gantry to independently control the position of 1000 MHz source and receiver antennas. The system was programmed to repeatedly collect a common mid-point (CMP) profile at the center of the tank followed by two constant offset profiles (COP) in the x and y direction. Each collection was completed in 30 s and repeated 50 times during a 28 min experiment. Two minutes after the start of measurements, the surface of the sand was irrigated at a constant flux rate of 0.006 cm/sec for 23 minutes. Time-lapse COPs show increases in traveltime to reflectors in the tank associated with increasing water content, as well as the development of a wetting front reflection. From 4-10 min, the CMPs show a distinct shingling characteristic that is indicative of waveguide dispersion. Forward models where the waveguide is conceptualized as discrete layers and a piece-wise linear function were used to invert picked dispersion curves for waveguide properties. We show the results from both inversion approaches for multiple dispersive CMPs and show how the single layer model fails to represent the gradational nature of the wetting front.