ALBERT

Description: A lab scale infiltration experiment was conducted in a sand tank to evaluate the use of time-lapse multi-offset ground-penetrating radar (GPR) data for monitoring dynamic hydrologic events in the vadose zone. Sets of 21 GPR traces at offsets between 0.44–0.9 m were recorded every 30 s during a 3 h infiltration experiment to produce a data cube that can be viewed as multi-offset gathers at unique times or common offset images, tracking changes in arrivals through time. Specifically, we investigated whether this data can be used to estimate changes in average soil water content during wetting and drying and to track the migration of the wetting front during an infiltration event. For the first problem we found that normal-moveout (NMO) analysis of the GPR reflection from the bottom of the sand layer provided water content estimates ranging between 0.10–0.30 volumetric water content, which underestimated the value determined by depth averaging a vertical array of six moisture probes by 0.03–0.05 volumetric water content. Relative errors in the estimated depth to the bottom of the 0.6 m thick sand layer were typically on the order of 2%, though increased as high as 25% as the wetting front approached the bottom of the tank. NMO analysis of the wetting front reflection during the infiltration event generally underestimated the depth of the front with discrepancies between GPR and moisture probe estimates approaching 0.15 m. The analysis also resulted in underestimates of water content in the wetted zone on the order of 0.06 volumetric water content and a wetting front velocity equal to about half the rate inferred from the probe measurements. In a parallel modeling effort we found that HYDRUS-1D also underestimates the observed average tank water content determined from the probes by approximately 0.01–0.03 volumetric water content, despite the fact that the model was calibrated to the probe data. This error suggests that the assumed conceptual model of laterally uniform, one-dimensional vertical flow in a homogenous material may not be fully appropriate for the experiment. Full-waveform modeling and subsequent NMO analysis of the simulated GPR response resulted in water content errors on the order of 0.01–0.03 volumetric water content, which are roughly 30–50% of the discrepancy between GPR and probe results observed in the experiment. The model shows that interference between wave arrivals affects data interpretation and the estimation of traveltimes. This is an important source of error in the NMO analysis, but it does not fully account for the discrepancies between GPR and the moisture probes observed in the experiment. The remaining discrepancy may be related to conceptual errors underlying the GPR analysis, such as the assumption of uniform one-dimensional flow, a lack of a sharply defined wetting front in the experiment, and errors in the petrophysical model used to convert dielectric constant to water content.

Description: A lab scale infiltration experiment was conducted to evaluate the use of transient multi-offset ground-penetrating radar (GPR) data for characterizing dynamic hydrologic events in the vadose zone. A unique GPR data acquisition setup allowed sets of 21 traces at different offsets to be recorded every 30 s during a 3 h infiltration experiment. The result is a rich GPR data cube that can be viewed as multi-offset gathers at discrete moments in time or as common offset images that track changes in the GPR arrivals over the course of the experiment. These data allows us to continuously resolve the depth to soil boundaries while simultaneously tracking changes in wave velocity, which are strongly associated with soil water content variations. During the experiment the average volumetric water content estimated in the tank ranged between 10–30% with discrepancies between the GPR results, moisture probe data, and 1-D numerical modeling on the order of 3–5% (vol vol−1), though the patterns of the estimated water content over time were consistent for both wetting and drying cycles. Relative errors in the estimated depth to a soil boundary located 60 cm from the surface of the tank were typically on the order of 2% over the course of the experiment. During the period when a wetting front migrated downward through the tank, however, errors in the estimated depth of this boundary were as high as 25%, primarily as a result of wave interference between arrivals associated with the wetting front and soil boundary. Given that our analysis assumed one-dimensional, vertical infiltration, this high error could also suggest that more exhaustive GPR data and comprehensive analysis methods are needed to accurately image non-uniform flow produced during periods of intense infiltration. Regardless, we were able to track the movement of the wetting front through the tank and found a reasonably good correlation with in-situ water content measurements. We conclude that transient multi-offset GPR data are capable of quantitatively monitoring dynamic soil hydrologic processes.