Scale Model Experimental Validation and Calibration of the Half-Space Green's Function Born Approximation Model Applied to Cross-Well Radar Sensing

Efficient forward models that describe the physical nature of the geophysical problem are desired for subsurface sensing and reconstruction of a contrasting contaminant pool volume. An analytical model to approximate sensing with radar is developed and implemented in the frequency domain in terms of the half-space lossy dyadic Green's function. The Born approximation is employed as a linear forward model, which will eventually be used for tomographic inversion for object detection. The forward model is compared with measurements generated by a cross-well radar (CWR) experiment in a controlled soil test tank using broadband borehole antennas. Soil parameter (dielectric constant and loss tangent) variance with frequency is represented by a quadratic polynomial. Calibration for soil parameters is performed via CWR data using an iterative nonlinear parameterized inversion technique. With the appropriate calibration, good agreement is obtained with wideband experimental measurements for several different borehole antenna placements, confirming the accuracy of the model.