GNGTS 2019 - Atti del 38° Convegno Nazionale

GNGTS 2019 S essione 3.2 683 domain, respectively. The non-uniform sampling of the wavelengths is evident in Fig. 2b, where the large-wavelength data points are more distant from each other compared to the short- wavelength data points. Starting from the initial model (Fig. 1a) and the true average DCs corresponding to the paths (Fig. 2a), we performed an inversion without considering any weight for the data points. In Figs 3a and 3b, we compare the real VS model with the estimated VS model from non-weighted inversion. We performed a second inversion, where we assigned weights to data-points of the true DCs. The weights were obtained as the distance between the wavelengths of the data points (Fig. 2b). In Fig. 3c, we show the estimated VS model of the weighted inversion. Comparing the two inversions (Figs 3b and 3c), the non-weighted inversion shows a low resolution in recovering the true VS model between depths 30 to 40 m, and almost no resolution below 40 m, while the estimated VS model from the wavelength-based weighted inversion shows very good resolution up to 60 m. Below 60 m, the weighted inversion shows almost no changes compared to the initial model. Conclusion. The SWT is inherently a very powerful tool to estimate VS with high resolution, if an adequate coverage of the data is available. The investigation depth of the inversion can be enhanced, by assigning wavelength-based weights to the data points prior to the inversion. Fig. 3 - The real VS model compared with the estimated VS models using non-weighted and weighted inversions (a) The real model. (b) The estimated VS model using non-weighted inversion. (c) The estimated VS model using weighted inversion.