GNGTS 2013 - Atti del 32° Convegno Nazionale

Site locations are shown in Fig. 1. Each dataset was analyzed to obtain the curve of H/V spectral ratio as a function of frequency. The curves were obtained by averaging the H/V spectral ratios obtained by Fourier transforming 30 seconds long time-windows with 30% overlap. All time windows were 5% tapered before spectral analysis. As is well known, raw spectra can show very strong variations between contiguous values, so that they can result to be almost unreadable. Because spectra are best represented in two-logarithmic reference axes, smoothing of raw spectra into evenly spaced frequency bands of constant width on a logarithmic scale facilitates their reading, without relevant loss of meaning. All raw spectra were smoothed using the “Konno-Ohmachi” filter W (Konno and Ohmachi, 1998) expressed as below Eq. (1): (1) where f is current frequency, fc the central frequency where the smoothing is performed and b is the bandwidth coefficient. The generally selected value of 40 for b was used throughout. Details concerning station names and the relevant peaks are reported in Tab. 1. Some examples of the obtained H/V curves are reported in Fig. 3 where the last image reports all the obtained H/V curves. The analysis of the H/V spectral curves indicate the presence of a resonance frequency (F 0 ) peaks ranging between 0.6/0.71 Hz and 1 Hz associated with H/V amplitude between 2 and 3.4. The average value of this peak is around 0.82±0.13 Hz (Fig. 1). A second peak (F 1 ), observed at some stations, characterized by lower frequencies (0.4 Hz - 0.59 Hz) and comparable average H/V amplitude to F 0 (2.1-3.0). A third more energetic peak of very low frequency is encountered in several stations with frequencies ranging between 0.21 Hz and 0.29 Hz and spectral amplitudes ranging between 2 and 7. The analysis of the spatial distribution of the resonance peak (F 0 ) is clearly not related to the main geomorphological features. However, this variation may be explained by variation in depth of the first sensible acoustic impedance contrast some 80/120 meters b.g.l., strong enough to give a resonance peak. This depth is estimated by assuming an effective value of 200/340 m/s for V S in the upper layer. The second peak (F 1 ) is related to a second discontinuity, whose depth can be estimated to be around 170/200 m m b.g.l, assuming an effective value of V S for the second layer of 350/400 … m/s. The assumed values of V S are in agreement with those obtained from the analysis of several passive seismic antennas (ESAC) conducted a long a 20 km long profile running between Cento (SW of Ferrara) to Bondeno (NW of Ferrara) (Abu Zeid et al. , 2013). Concerning the lowest encountered peak of around 0.21/0.29 Hz more investigation is needed to explain its origin. Conclusions and future work. Quantitative interpretation of HVSR data is inherently ambiguous because frequency of each peak depends on two parameters, that cannot be resolved simultaneously. Therefore the presented results should be considered as preliminary, as they must be integrated by independent information. We have recently assembled a 48-channel seismograph, complete with forty eight 3-component seismometers with proper frequency of 4,5 Hz. With this equipment we plan to collect ESAC data so that joint inversion of both data sets will be possible, to get a model of shear waves velocity V S to a maximum investigation depth of 200-250 m. Further measurements using BB seismometer shall be performed in order to check the reliability of the lowest encountered H/V peaks. In this way we plan to get the first deep quantitative models below the town of Ferrara, where for obvious reasons, geophysical information is lacking. 171 GNGTS 2013 S essione 2.2