PRELIMINARY COMPARISON BETWEEN EXPERIMENTAL AND THEORETICAL HVSR CURVES USING DIFFERENT MODELS.

The ambient vibration Horizontal to Vertical Spectral Ratios (HVSR) (Nakamura, 1989) is a widely used technique to identify the seismic resonance phenomena induced by the presence of seismic impedance contrasts at depth. Moreover, the HVSR curve can be used to constrain the shear wave velocity (Vs) profile in numerical inversion procedures: for this purpose, different HVSR forward modeling were developed in the last decades, which differ from each other both for the basic theoretical assumptions related to the ambient vibration wavefield simulation and for the phases of the involved seismic waves. In order to explore if significant differences between these modeling exist, Tanzini et al. (2022) and Lunedei et al. (2022) performed some comparisons using the results of large sets of numerical simulations obtained by considering realistic Vs profiles. In particular, these HVSR modeling were taken into account: based on body waves propagating vertically (Herak, 2008); based on ellipticity of the fundamental mode of Rayleigh waves (Wathelet et al., 2020); where ambient vibrations wavefield results from the contribution of random sources uniformly distributed at the surface under the assumption that surface waves dominate the wavefield (Lunedei and Albarello, 2009) and considering the full wavefield (Lunedei and Albarello, 2016); where ambient vibrations constitute a diffuse random wavefield (Garcia-Jerez et al., 2016) considering the full wavefield and surfaces waves only. Both studies show that strong similarities between the resulting simulated curves exist. In view of these conclusions, in this work, a comparison of the different theoretical HVSR modeling with experimental HVSR curves has been performed. HVSR measurements were carried out at test sites belonging to VEL (Valutazione Effetti Locali) down-hole database of the Tuscany Region (https://www.regione.toscana.it/-/banca-dati-vel). In particular, more than 50 sites with Vs profiles characterized by the presence of the seismic bedrock (Vs≥800m/s) and strong impedance contrasts were selected. Velocimetric acquisitions were carried out using the three-directional 24-bit digital tromograph Tromino™, produced by Moho SRL (https://moho.world/); the ambient vibrations were acquired for 20 min with a sampling frequency of 128 Hz. The HVSR curves were computed according to the procedure described by Sesame (2004) and Picozzi et al. (2005). In particular, the spectra of the single components were computed by averaging 20-s-long non-overlapping windows; a detrend and a 5% cosine taper were applied to each window, and the spectra were smoothed by using a triangular moving window with a frequency-dependent half-width (10% of central frequency). The horizontal components were combined with the geometric average. Three measurements were performed at each site and the best quality one was chosen following the criteria proposed by Albarello et al. (2011). Theoretical HVSR curves were simulated considering the models mentioned above and taking into account the Vs and Vp profiles of the considered down-holes; density values were deduced from Vp values following the relationships proposed by Brocher (2005) and, for not purely elastic models, damping values for Vp and Vs are assumed equal to 0.01 for all the layers. Finally, these curves were compared with the respective experimental ones in order to evaluate the differences in terms of frequency and peak amplitude as well as of overall trend.