note

The velocity records were first corrected for the instrumental response, the mean from the whole signal removed and differentiated to acceleration.

Then, time windows starting 0.5s before the S-wave onset and ending when 90 percent of the total energy after the S-wave onset has been released, with a minimum and maximum allowed duration of 5 and 15 s, respectively, have been extracted and tapered with a 5% Hanning-window.

For the selected windows, the acceleration FAS have been calculated and smoothed using the Konno and Ohmachi (1998) algorithm, fixing the smoothing parameter b to 40.

The spectral amplitudes are analyzed at 36 frequencies, equally spaced on a logarithmic scale over the range 0.5–30 Hz.

Pre-event noise windows of equal length as the signal windows were used to compute the signal-to-noise ratios (SNR), and, at each frequency point, only records with a SNR larger than 3 were retained.

Figure 3 shows examples of SNR calculated for three stations located in different positions over the study area and equipped with different instrumentation systems.

AQ06 and GFZ05 are velocimetric stations whereas MI01 is accelerometric.

Moreover MI01 and GFZ05 are installed on outcropping rock while AQ06 is located on the alluvial deposits of the Aterno river (see Fig. 1 and Bergamaschi et al. this issue).

The SNR are high with median values between 10 and 1,000 and only few frequency points with a ratio smaller than 3. For the whole data set, the number of selected spectral amplitudes, based on SNR > 3, spans from 1,124 at 0.5 Hz to 1,161 at 10 Hz.