GNGTS 2013 - Atti del 32° Convegno Nazionale

data, does not generally provide very accurate estimates, mainly because GPR velocity often decreases with depth, unlike seismic velocity. Therefore some assumptions used in standard velocity analysis, such as assimilating RMS velocity to stack velocity, produce inaccurate results when applied to GPR data analysis (Becht et al. , 2006). We developed and tested a new procedure to estimate the EM velocity field from Common Offset (CO) GPR data, which is still the most common and less demanding acquisition scheme. Current velocity analysis methods on CO GPR data. There are various techniques to estimate the EM velocity distribution from CO GPR data by taking into account the diffractions recorded on the GPR sections, which are caused by objects having dimensions comparable with the mean wavelength of the radar signal incident on them. The most common are: 1) the Diffraction Hyperbola Fitting and 2) the Migration Velocity Scan. 1) The diffractions in a GPR section have a hyperbolic shape, whose convexity depends on the RMS velocity of the radar wave between the ground surface and the scattering object. Therefore the velocity distribution can be estimated by fitting the diffraction hyperbolas at different depths. This method is very simple and can be applied to any hyperbolic event within a GPR section. Moreover, in situations where there are no reflection events, such as in some urban environments, this is the only available method. However, the hyperbola fitting method has different problems, which limit the accuracy of the final results. In order to have truly hyperbolic diffraction events, the analyzed system must have plane, parallel and homogeneous layers. In a real system, lateral variations of the velocity distribution cause distortions of the hyperbolic shape and consequent errors in velocity estimation. Another cause of distortions is the shape of the scatterer since, if it has one dimension longer than the others, it does not behave as a point source and the recorded event results from a combination of reflection and diffraction. Further limitations to such procedure depend on the actual presence and regular distribution of scatterers in the GPR section, able to produce well-defined diffraction hyperbolas, free from interferences with other reflection or diffraction events. The lower the number of diffractions is, the lower will be the resolution of the velocity field, both vertically and horizontally. Finally, in a GPR section the diffraction hyperbolas are often spatially limited to an interval around their vertex. This fact, together with the non-impulsivity of the radar signal, adds further uncertainty to velocity estimation because hyperbolas with very different curvatures can fit the same data. 2) Migration is used to collapse diffractions and to move dipping reflectors to their correct position in a GPR section, and it requires an accurate velocity distribution as input parameters. Therefore, by iteratively applying the process of migration and by analyzing the results, it is possible to estimate the actual velocity distribution. If the distribution is accurate, all the diffraction hyperbolas will collapse around their apexes, otherwise there will be either an under-migration for lower velocities or an over-migration for higher velocities. Theoretically, dipping reflectors can also be used to analyze the results of the migration, if independent information on their real position is available. However, also such method depends heavily on the actual presence, abundance and regular distribution of clearly visible diffraction hyperbolas in the entire GPR dataset. New method . The new proposed procedure is a generalization of the one developed by Forte et al. (2013) and estimates the EM velocity distribution from each recorded trace, in a 2-D bistatic CO GPR profile, using the principles of geometrical ray theory. The method makes four assumptions: 1. In the proximity of each trace position the system is horizontally layered. 2. Each layer is lossless, non-dispersive, homogeneous, and isotropic. Such conditions are not generally respected by earth materials, but they are good approximations for many media in which the GPR signal propagates efficiently (Davis and Annan, 1989). 3. The radar propagates a plane EM wave. 113 GNGTS 2013 S essione 3.2