Combined effective wavelet estimation and full-waveform inversion of GPR data

Full-waveform inversion uses all information content of the measured data, which leads to higher resolution images compared to standard ray-based techniques. We developed a full-waveform inversion scheme for common-midpoint (CMP) GPR data that is based on a frequency domain solution of Maxwell's equations assuming a layered model of the subsurface. The full-waveform inversion requires a good start model of the subsurface medium parameters; permittivity and conductivity, which mainly influence the velocity and amplitudes, respectively. Traveltime inversion usually returns start values for the relative permittivity. However, reliable conductivity values are difficult to obtain. Erroneous conductivity values result in wrong amplitudes of the effective source wavelet that is needed for the full-waveform inversion, which in turn also results in erroneously inverted conductivity values, since the conductivity and wavelet amplitudes are coupled. To assure a good start model and an accurate estimation of the unknown source wavelet, we implemented a joint optimization of the effective source wavelet and the medium parameters using a combined global and local search algorithm. The full-waveform inversion was successfully applied to dispersive synthetic and experimental CMP datasets reflecting a single-layer low-velocity waveguide.