Deterministic deconvolution for GPR data in the t-f domain

Deconvolution methods encounter difficulties in increasing the temporal resolution of GPR data due mainly to non-stationarity of the records. GPR wavelets are typically mixed phase, which is additionally a major failure of standard deconvolution methods. Here, we propose a deterministic deconvolution method for GPR data, implemented in the t-f domain, which utilizes narrow time windows and sets spectral balancing as a precondition. A reference wavelet is estimated experimentally for the calculation of a time varying deconvolution operator. Its phase variation is extracted from the spectrally balanced deconvolved GPR trace. The algorithm, tested on synthetic and real data, produces very promising results. In particular the deconvolved GPR section acquired over sands exhibits better temporal resolution and reveals reflected waves travelling through high loss media.