GNGTS 2013 - Atti del 32° Convegno Nazionale

while the high SNR condition is necessary to make negligible the effects of the many local sources of noise active in the underground laboratory. Looking at Fig. 1 we see immediately that there is not any windows along the coda where the two conditions are verified in the whole frequency range we are studying. Therefore it is necessary to split the frequency range in different intervals that will correspond to different time windows along the coda. The boxes in spectrogram and coherence of Fig. 1 shows as an example the time windows and their corresponding frequency intervals adopted for that earthquake. After have selected the time window, Fourier spectra are computed for each component of any stations over a 20 s sliding window with 50% overlapping. The spectral amplitude ratio H/V and spectral power amplitude ratio are computed as a function of frequency, and finally the mean and standard deviation among the array stations are computed. At this point we consider only the result in the frequency interval corresponding to the selected time-frequency window. The result for a regional earthquake is shown in Fig. 2. The spectral ratios in Fig. 2 show at least two distinct trends in two different frequency band of analysis: • H/V is very close to unity (equal to 1.1 - 1.2) and PE hr is equal to 0.67 within error bars in the high/medium frequency range (from 7 to 22 Hz) identified by blue and partly black lines; • for the low frequency range (0-7 Hz, mostly in red), H/V depart from a constant value with oscillations around unity and PE hr assumes values between 0.4 and 0.8. The results for high frequencies band (7-22 Hz) is a clue of equipartioned coda in subsurface under the hypothesis of wavefield composed by body waves propagating in an homogeneous medium. Further observations can be performed for low frequencies part of energy partition by simulating energy ratios and comparing them with the observed ones (Margerin et al. , 2009). By considering a reliable velocity and density model for the area under investigation (Capuano et al. , 1998; Chiarabba et al. , 2009), energy partition ratios have been simulated by applying a method based on a spectral decomposition of the elastic wave in a stratified half-space (Margerin, 2009). The energy ratios have been calculated at different depths (between 1 and 2 km) and for three different assumptions on the origin of coda waves: 1) coda composed of body waves incident from below the crust only; 2) coda composed of surface and guided waves in the crust only; 3) the coda is the sum of the two previous contributions. The best qualitative Fig. 3 – Observed (red line with error band) and simulated values of PE hr (lines with other colors) obtained for different depths (see legend) and under the hypothesis of coda waves composed by surface and body waves. The best qualitative agreement is obtained for 1.4 km depth (green line). 67 GNGTS 2013 S essione 1.1