Mixed-phase Deconvolution of Ground-penetrating Radar Data

Ground-penetrating radar (GPR) surveying has the potential to provide detailed images of the shallow subsurface. However, these images are distorted due to effects of the source wavelet. Standard deconvolution procedures, which are routinely used to increase the vertical resolution of seismic data by removing the source wavelet, are often not effective when applied to GPR data. It is commonly suspected that this problem is linked to the fact that standard stochastic deconvolution algorithms are based on the assumption of a minimum-phase source wavelet, whereas GPR source wavelets generally are mixed-phase. We propose a new approach to deconvolve GPR data that is based on the observation that a mixed-phase wavelet corresponds to the convolution of a minimum-phase wavelet and a dispersive all-pass filter. Consequently, the successful deconvolution of GPR data involves estimating two inverse filters, one that cancels the minimum-phase wavelet and one that removes the effects of the all-pass filter. We demonstrate the potential of this approach using synthetic and field GPR data. The results show that our deconvolution approach significantly increases the vertical resolution compared to standard processed GPR data.