GNGTS 2018 - 37° Convegno Nazionale

252 GNGTS 2018 S essione 1.3 vertical spectral ratios (H/V) in the 0.1-1 Hz frequency band. We used the GEOPSY package (Wathelet et al. , 2004) selecting at least one hour of noise signals and calculating, on 400 s moving time windows with 20% overlap, the spectral ratios (Fig. 3). The stationary results refer to the stability of the H/V ratios throughout the analysis: which means that the H/V ratios performed on a time-window should be as similar as possible to each other, especially with respect to the amplitude peak frequency. The continuous black curves represent the average H/V ratio, while the dashed black curves are representative of the standard deviation. The grey bands identify the frequency peak, or f 0 , with the relative standard deviation bandwidth. We can see that the stations placed near the coast (T1361 and IPSM) exhibit a maximum in the H/V ratios between 0.1 Hz and 0.2 Hz, while the other stations show the maximum at higher frequencies, between 0.4 Hz and 0.8 Hz. Inversion procedure and results. We applied the neighbourhood algorithm employed by Wathelet et al. (2004) (DINVER), which infers the best velocity model thorough a stochastic search in a multi-parameter space. One point of this space corresponds to one velocity model defined by S- and P-wave velocities, thickness, densities and Poisson’s ratios of the soil layers, and a synthetic dispersion curve is calculated by inverting the S-wave equation as function of the surface waves velocity. We performed a joint inversion of the autocorrelation coefficients (Fig. 2), the dispersion curve obtained from the f-k analysis and the mean H/V ratio obtained from all the station in the 0.2 – 1 Hz frequency band. The resulting models after the preliminary inversion step show a fairly good fit between experimental and theoretical curves using a model parameterization composed of two main layers over half-space with a shear-wave velocity increasing with depth. Conclusions. The results obtained in this preliminary study will be the starting point for a correlation study of all the stations installed on the Ischia island, that will be considered as a large array, to obtain a more accurate S wave velocity model in the first 2-3 km of depth. In fact, the depth resolution depends on the maximum resolvable frequency of the used ring arrays. The lower frequency limit is defined by the largest ring diameter, that was approximately 0.2 Hz. This means that, if we consider the maximum wavelength observable equal to about 8000 m, we can obtain a pseudo-depth of investigation equal to 3200 m (wavelength/2.5). The minimum wavelength, instead, is equal to about 800 m, consequently we can obtain the maximum resolution thickness equal to about 300 m. This value is higher than the local lithological heterogeneity, that are for the most part in the first 200 m (Vassallo et al. , 2017), and we can consider the obtained velocity model as representative of the whole Ischia island. Furthermore our result is in agreement with the velocity models achieved by Strollo et al. (2015). Fig. 3 - Noise H/V ratios as a function of frequency. The colored curves on the H/V vs frequency graphs represent the H/V ratios calculated on each time-window. The grey vertical band indicates the maximum in the H/V, or rather the f 0 .