GNGTS 2013 - Atti del 32° Convegno Nazionale

local geological characteristics this station should not present significant amplification effects on ground motion. A second instrument (AR4) was deployed in the alluvial valley (SA6) which corresponds to an area of possible future industrial expansion. Soft soil layers corresponding to the recent alluvial deposits characterize this recording site suggesting the presence of some amplification effects in a frequency range of interest from the engineering point of view. The other two instruments were installed on the Santopadre Formation, the first in the historical city centre the second on a hilly area representative of the countryside of the city. Our choice of installing two instruments (AR2 and AR3) on the same MOPS unit (SA2) was motivated by the extreme lateral and vertical variability of the characteristics of the formation in terms of lithotypes that range from conglomerates to siltstone. This feature cannot be easily represented in the geological map and does not allow associating one lithostratigraphic succession for the entire formation. Due to these considerations we decided to investigate the possibility of a different seismic response that cannot be caught by the MOPS map. Moreover the area were AR2 is located suffered the most severe damages in the strongest historical earthquakes and some damage in the February 2013 event. Seismic data analysis. As first step in our analysis of seismic data ambient noise waveforms were used for evaluating horizontal over vertical spectral ratio (HVNSR) following the approach proposed by Nakamura (1989). The continuous recording mode allowed selecting long time windows recorded at night to perform the analysis. This choice was necessary to reduce the effect of the strong sources of anthropic disturbs found during the day in some of the recording site. Moreover it was possible to remove those windows contaminated by the presence of micro-earthquakes, quite frequent at the time of the network operation. All the signals were processed with an antitrigger algorithm (SESAME, 2004) to remove noise windows containing disturb and transient phenomena; Fourier spectra were smoothed with a Konno and Ohmachi (1998) algorithm, and HVNSR was evaluated on the geometrical mean of the Fourier spectra of horizontal components. The resulting HVNSR is an average over a large number of selected 40 seconds long time windows varying from 50 to 100 according to the sites characteristics. This ensures a good stability in the results. 392 GNGTS 2013 S essione 2.3