HR: 16:15h
AN: H44A-02 [Abstracts]
TI: Inversion for the statistical structure of subsurface water content from ground-penetrating radar reflection data: Initial results and interpretation
AU: * Irving, J
EM: james.irving@unil.ch
AF: Institute of Geophysics, University of Lausanne, Lausanne, 1015, Switzerland
AU: Knight, R
EM: rknight@pangea.stanford.edu
AF: Geophysics Department, Stanford University, Stanford, CA 94305, United States
AU: Holliger, K
EM: klaus.holliger@unil.ch
AF: Institute of Geophysics, University of Lausanne, Lausanne, 1015, Switzerland
AB:
The distribution of subsurface water content can be an excellent indicator of soil texture, which strongly influences
the unsaturated hydraulic properties controlling vadose zone contaminant transport. Characterizing the
heterogeneity in subsurface water content for use in numerical transport models, however, is an extremely difficult
task as conventional hydrological measurement techniques do not offer the combined high spatial resolution
and coverage required for accurate simulations. A number of recent studies have shown that ground-penetrating
radar (GPR) reflection images may contain useful information regarding the statistical structure of subsurface
water content. Comparisons of the horizontal correlation structures of radar images and those obtained from
water content measurements have shown that, in some cases, the statistical characteristics are remarkably
similar. However, a key issue in these studies is that a reflection GPR image is primarily related to changes
in subsurface water content, and not the water content distribution directly. As a result, statistics gathered on the
reflection image have a very complex relationship with the statistics of the underlying water content distribution,
this relationship depending on a number of factors including the frequency of the GPR antennas used.
In this work, we attempt to address the above issue by posing the estimation of the statistical structure of water
content from reflection GPR data as an inverse problem. Using a simple convolution model for a radar image, we
first derive a forward model relating the statistical structure of a radar image to that of the underlying water
content distribution. We then use this forward model to invert for the spatial statistics of the water content
distribution, given the spatial statistics of the GPR reflection image as data. We do this within a framework of
uncertainty, such that realistic statistical bounds can be placed on the information that is inferred. In other words,
we attempt to address the question "what can we infer about the water-content statistical structure, given the
GPR data?", rather than "what is the water content statistical structure?". Results of applying our estimation
technique to simple synthetic models are positive, and give us hope that reflection GPR data can be used in
practice to better constrain knowledge of the nature of subsurface water content heterogeneity. If successful, this
type of approach could also be used with seismic reflection data to infer the statistical nature of velocity
heterogeneities.
DE: 1829 Groundwater hydrology
DE: 1835 Hydrogeophysics
DE: 3252 Spatial analysis (0500)
DE: 3260 Inverse theory
DE: 3275 Uncertainty quantification (1873)
SC: Hydrology [H]
MN: 2007 Fall Meeting