GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.3 GNGTS 2024 Figure 3: High frequency inference on a secton of the Viking Graben dataset: A represents the low-fltered data, the same used in Figure 2A, B is the high frequency inference, and C represents the high frequency reference data. The amplitude spectra of the seismic line depicted in A (Orange), in B (Green), and C (Blue) are reported in D. Thanks to this, we can make a proper comparison between the input (Figure 3A), the predicton (Figure 3B) and the reference data (Figure 3C). If we focus on the strongly dipping refector marked with black and red arrows, we can see that while the predicton matches prety well the reference, this feature was not visible (black arrow) or barely visible (red arrow) in the NN input. Further focus should be put on the difracton paterns pointed out with the blue arrows: while the match is very good between NN predictons and reference data (i.e. Figure 3B and C), hyperbolas are difcult to be interpreted in the input lower frequency data (i.e. Figure 3A). In order to further evaluate the results of the two models, we focus on the area from 120-1250m and from 0.6-1.3s (see e.g. Figure 2A). This area is very interestng because it images a typical wedge structure. In Figure 4 we present a wiggle plot of the input (Figure 4A) and the high- frequency inference (Figure 4B).