Evaluation of the Source Wavelet Estimation Procedure During Full-Waveform Inversion of Crosshole Georadar Data

High-resolution crosshole georadar and/or seismic tomographic imaging is becoming increasingly important for a wide range of environmental, hydrological, and engineering applications. This in turn has led to a significant interest in the use of full-waveform inversion approaches on these data, which hold the promise of providing images of pertinent petrophysical parameters at a spatial resolution comparable to that of borehole logs. A major issue related to the practical application of such methods, however, is the accurate estimation of the source signal. To address this problem, we have explored the viability and robustness of incorporating the source wavelet estimation procedure into the waveform inversion algorithm through an iterative deconvolution approach. We have developed this approach for crosshole georadar data, but the results are equally valid for corresponding seismic applications. Given the inherent linearity of such a deconvolution- based approach, one of the key questions that arise is its accuracy and robustness in the presence of non- linear effects, such as those related to strong scattering and the presence of significant amounts of ambient noise. Our results indicate that (i) our algorithm is remarkably accurate and robust even under quite challenging conditions, (ii) it is relatively insensitive to the first estimate of the source wavelet, and (iii) there is little or no trade-off in the inversion procedure between the wavelet estimation and the imaging procedure. Finally, we also find that, at least in non-dispersive or weakly dispersive environments, the proposed wavelet estimation procedure is largely independent of the underlying attenuation structure.