GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.2 GNGTS 2024 between instrument cost and results. To fully review this statement, more test sites are expected to be analysed in such a way. Second experiment: how the data quality of GPR 2D sectons and 3D data volume may be improved by the stacking of adjacent channels. The previous observatons bring to the second experiment: since, in a typical test site, all the dipoles were not strictly necessary, an idea to make beter use of them is to stack adjacent dipoles. This is meant to try to improve the signal-to-noise rato, potentally useful at depth or when the data are noisy for various reasons. The second dataset was acquired at the archaeological park of Tindari in the municipality of Pat, Sicily, Italy. This site is recognized to be very challenging with respect to the GPR wave propagaton given the high signal atenuaton due to the presence of clay formatons at the site (De Domenico et al., 2006). Fig. 2 shows the improvement from non-stacked radargrams and radargrams obtained from the stacking of adjacent dipoles. The major improvements (red rectangles in Fig.2) were notced in the 15-30 ns range, and this improvement is maintained in the 2D sectons afer the other processing steps (not shown here). The stacked 2D radargrams were then used to create 3D GPR data volumes, as in the frst experiment. Fig. 2 – Comparison between GPR sectons before and afer stacking. a1): afer dewow, move startme, and background removal. a2) : a1) + stacking of adjacent 5 channels. An energy decay gain flter was then applied to both sectons to allow for meaningful visual readability. In Fig. 3, the a-panels refer to a 60-cm depth-slice, while the b panels display a 40-cm depth slice. Looking at Fig. 3, it is possible to recognize that the stacking of adjacent radargrams can improve the overall quality of 3D data volumes, when sources of noise, or atenuaton, or other problems, challenge the readability of depth-slices even at shallow depths. The noise gets reduced, which makes bigger features more visible. In the data displayed and available, the improvement was stll not able to clearly distnguish underground features which were not already visible without the stacking. In our opinion, the stacking of adjacent channels (or other processing steps based on this philosophy) is promising as