Uncertainty Quantification in Off-Ground Monostatic Ground Penetrating Radar

The characterization of the subsurface can be performed by full-waveform inversion of electromagnetic data. The modeling process relies on the ability to retrieve the scattered field Green´s function from the measured data using sets of antenna characteristic global reflection and transmission coefficients. Crucial for a successful implementation of this technique is the understanding of uncertainties involved in the acquisition of the antenna calibration and survey measurements, and how these propagate in the parameter estimation results. We find that averaging a number of possible Green´s functions obtained from one measurement with several antenna characteristic coefficients sets works remarkably well in reducing the uncertainties. The accuracy of the inversions improves using characteristic coefficients acquired as close as possible to the measurement conditions. A clear relation between dynamic range and system resolution is highlighted, based on the number of effective bits contained in the data.