GNGTS 2013 - Atti del 32° Convegno Nazionale

error associated to the velocities in the first layer has the major influence on the uncertainties of the final results. Nevertheless, this error can be reduced by combining different independent measurements, like CMP gathers, TDR measurements, or dedicated trans-illumination experiments. The assumption of 1-D model in the proximity of each trace position, as well as the small spread approximation used in the algorithm, are acceptable for most of the real GPR Fig. 1.– Results of the inversion program. A) Velocity distribution in the six layers, with the mean uncertainties in each layer equal to 0.20, 0.19, 0.20, 0.21, 0.20 and 0.20 cm/ns, for increasing depths. The value and uncertainty in the first layer are constant and defined by the operator, as required by the inversion program. Where layers are not defined, their graph merges with the nearest ones. For clarity, the velocity values were smoothed with a 101 trace window, and the error bars were not shown. B) Geometrical location and shape of the six picked layers; with the mean uncertainties in each underground interface equal to 0.01, 0.03, 0.08, 0.19, 0.51 and 0.52 m, for increasing depths. 117 GNGTS 2013 S essione 3.2