<title>Estimation and correction of wavelet dispersion in GPR data</title>

Abstract

Wavelet dispersion caused by frequency-dependent attenuation is a common and significant problem in ground-penetrating radar (GPR) data. In the radar image, it is displayed as a characteristic 'blurriness' that increases with depth. Correcting for wavelet dispersion in GPR data is an important step that should be performed before either qualitative interpretation or quantitative determination of subsurface electrical properties are attempted. Over the bandwidth of a GPR wavelet, the attenuation of electromagnetic waves in many geologic materials is approximately linear with frequency. For this reason, the change in shape of a radar pulse as it propagates through these materials can be described using one parameter, Q*, which is related to the slope of the linear region. Assuming that all subsurface materials can be characterized by some Q* value, the problem of estimating and correcting for wavelet dispersion in GPR data becomes one of determining Q* in the subsurface and deconvolving its effects through the use of an inverse Q filter. A method for the estimation of Q* from GPR data based on a technique developed for seismic attenuation tomography is presented. Essentially, Q* is determined from the downshift in the dominant frequency of the GPR wavelet with time down a trace. Once Q* has been obtained, we propose an inverse Q filtering technique based on a causal, linear model for constant Q as a means of removing wavelet dispersion. Initial tests on field data indicate that this technique is very effective at enhancing the resolution of the GPR image.