High-resolution ground-penetrating radar monitoring of soil moisture dynamics: Field results, interpretation, and comparison with unsaturated flow model

Surface ground-penetrating radar (GPR) techniques have been used by a number of previous researchers to characterize soil moisture content in the vadose zone. However, limited temporal sampling and low resolution near the surface in these studies greatly impedes the quantitative analysis of vertical soil moisture distribution and its associated dynamics within the shallow subsurface. To further examine the capacity of surface GPR, we have undertaken an extensive 26 month field study using concurrent high-frequency (i.e., 900 MHz) reflection profiling and common-midpoint (CMP) soundings to quantitatively monitor soil moisture distribution and dynamics within the shallow vadose zone. This unprecedented data set allowed us to assess the concurrent use of these techniques over two contrasting annual cycles of soil conditions. Reflection profiles provided high-resolution traveltime data between four stratigraphic reflection events while cumulative results of the CMP sounding data set produced precise depth estimates for those reflecting interfaces, which were used to convert interval-traveltime data into soil moisture. The downward propagation of major infiltration episodes associated with seasonal and transient events are well resolved by the GPR data. The use of CMP soundings permitted the determination of direct ground wave velocities, which provided high-resolution information along the air-soil interface. This improved resolution enabled better characterization of short-duration wetting/drying and freezing/thawing processes, and permitted better evaluation of the nature of the coupling between shallow and deep moisture conditions. The nature of transient infiltration pulses, evapotranspiration episodes, and deep drainage patterns observed in the GPR data series were further examined by comparing them with a vertical soil moisture flow simulation based on the variably saturated model, HYDRUS-1D. Using laboratory-derived soil hydraulic property information from soil samples and a number of simplifying assumptions about the upper and lower-boundary condition, we were able to achieve very good agreement between measured and simulated soil moisture profiles without model calibration; this is a strong indication of the overall quality of the GPR-derived soil moisture estimates. The only notable difference between simulated values and GPR water content estimates occurred during extended dry soil conditions near the surface.