Modelling and filtering of surface scattering in ground-penetrating radar waves

Ground-penetrating radar (GPR) is a geophysical method based on the propagation, reflection and scattering of high-frequency (from 10 MHz to 1 GHz) electromagnetic (EM) waves in the subsurface (Daniels et al. 1988). This method is currently used to image the subsurface and has myriad applications; void detection; prediction of deterioration in railroads, airport runways and concrete; detection of archaeological objects; mapping of subsurface wastes and contaminants in environmental engineering (Owen et al. 1995); stratigraphic and bedrock mapping and hydrological applications (Beres & Haeni 1991; Van Overmeeren 1994); mapping faults hidden under vegetation cover (Meyers et al. 1996; Liner & Liner 1997). The depth of investigation depends on the frequency used and on the medium traversed by EM waves; for low-loss geological materials it does not exceed 50 m (Davis & Annan 1989). In some cases, GPR data show strong diffraction hyperbolae due to surface objects on the ground. EM waves propagate with little attenuation in air and conductive surface objects (power lines, metallic fences) and trees are strong reflectors, so reflections from above-ground features are often present in GPR data (Sun & Young 1995). Thus, it is very important to recognize and distinguish if the diffractions are from subsurface heterogeneities (Papziner & Nick 1998) or from surface scattering and not to confuse the latter with the subhorizontal geological reflections. A method of modelling and filtering such surface scattering is presented in this paper. Migration of radar data with the free-space velocity focuses the air diffractions producing large isolated amplitudes. Subsurface reflections are overmigrated (not focused) and random Gaussian noise is defocused and becomes more Gaussian (Harlan et al. 1984). After applying a filter in amplitude (treshold) to the migrated data, only the large amplitudes of the focused hyperbolae are preserved. Then the result is diffracted with the same velocity (free-space velocity) to produce synthetic air diffractions. Finally, a comparison of the synthetic air diffractions with the real data removes the surface scattering from the original data. The methodology described in this paper is tested and illustrated on two examples of GPR profiles which show strong surface scattering from objects on the ground.