HR: 0800h
AN: H51E-06    [Abstracts]
TI: Geostatistical Characterization of the Unsaturated Zone Using Cross-Borehole Ground Penetrating Radar
AU: * Looms, M C
EM: mcl@geol.ku.dk
AF: University of Copenhagen, Department of Geography and Geology, Oester Voldgade 10, Copenhagen K, 1350, Denmark
AU: Hansen, T M
EM: tmh@gfy.ku.dk
AF: University of Copenhagen, Niels Bohr Institute, Juliane Maries Vej 28, Copenhagen Ea, 2100, Denmark
AU: Nielsen, L
EM: ln@geol.ku.dk
AF: University of Copenhagen, Department of Geography and Geology, Oester Voldgade 10, Copenhagen K, 1350, Denmark
AU: Jensen, K H
EM: khj@geol.ku.dk
AF: University of Copenhagen, Department of Geography and Geology, Oester Voldgade 10, Copenhagen K, 1350, Denmark
AU: Binley, A
EM: a.binley@lancaster.ac.uk
AF: Lancaster University, Department of Environmental Science, Lancaster, Lancaster, LA1 4YQ,
AB: The water infiltration through the unsaturated zone is a determining factor for the quantity and quality of the underlying groundwater. An accurate understanding and description of the important processes in the unsaturated zone is needed in order to produce reliable decision tools for groundwater exploitation and protection. High resolution tomographic images obtained using cross-borehole ground penetrating radar may provide valuable information regarding the characteristics of the shallow subsurface. More specifically, geostatistical properties of hydrological state variables, such as moisture content, may be estimated from derived velocity distributions. Unfortunately, commonly used least-squares inversion techniques result in smooth, minimum variance estimates of the subsurface radar wave velocity structure, which may diminish the utility of these images for geostatistical inference. To address this limitation, we present here a recently developed stochastic inversion technique to infer the subsurface geostatistical properties using cross-borehole ground penetrating radar data alone. For a specific choice of prior covariance model, we evaluate how likely it is that samples of the posterior Gaussian probability density function (PDF) are samples of the prior Gaussian PDF. The properties are inferred without using the inversion images directly and are not affected by the excessive smoothing/damping often observed in previous literature. The velocity distributions and obtained correlation structures are compared with similar estimates found using traditional least-squares inversion using both synthetic studies as well as travel time data collected at a field site in Denmark. In comparison to the traditional inversion algorithm the stochastic inversion technique produces images containing a higher degree of spatial variability, in addition the subsurface structures are less connected and more contained. Small-scale changes in the velocity distribution may result from local moisture content changes, and these small-scale features may potentially be extremely important if small-scale flow patterns like fingering and preferential flow occur.
DE: 1835 Hydrogeophysics
DE: 1866 Soil moisture
DE: 1875 Vadose zone
SC: Hydrology [H]
MN: 2007 Joint Assembly