GNGTS 2015 - Atti del 34° Convegno Nazionale

defined “rectilinearity” (Schimmel and Gallart, 2004), can be used as an index of how much instantaneous signal is close to a linear ( rl =1) or circular ( rl =0) polarization. The methodology of ambient noise analysis proposed in Del Gaudio (2013), consists of passing the noise recording through narrow-band filters with varying central frequencies ν c and then calculating instantaneous polarization characteristics from the analytic representation of the filtered time series in order to obtain: 1) instantaneous values of ratios H max /V between the amplitudes of the maximum horizontal component of ground motion and the vertical one; 2) recognition of a preferential signal polarization, possibly reflecting directional site amplification; 3) identification of single data sample whose polarization is compatible with Rayleigh wave particle motion; 4) identification of data sample whose polarization is compatible with Love or SH-waves; 5) calculation of the H max /V average over data samples of Rayleigh type. The last calculation, carried out for different ν c values, produces mean H/V ratios derived from instantaneous polarization (named HVIP values), which can be assumed as an estimate of the ratios between horizontal and vertical component of Rayleigh waves at difference frequencies. The implementation of this procedure requires the choice of a filtering type and of some “threshold” parameters, to identify Rayleigh/Love waves, i.e.: i) the maximum admissible deviations of the planarity vector from horizontality and of vectors a →  ( t ) and b →  ( t ) from vertical/ horizontal directions ( diplim ); ii) the maximum rectilinearity rllim to distinguish elliptical (Rayleigh waves) from linear (Love waves) polarization; iii) the minimum number nmin of consecutive data samples that identifies a wave packet of coherent type. The last criterion is justified considering that, within a large number of samples, single isolated cases satisfying the Rayleigh/Love identification criteria could be purely casual. Thus, a more reliable identification of Rayleigh or Love waves require the presence of a certain number of consecutive samples with coherent type of polarization, so that “packets” of Rayleigh/Love waves are identified, rather than just single data samples. Test implementation. In order to optimize the analysis results, a series of tests were arranged using synthetic signals consisting of time series of 1000 s, sampled with a frequency of 100 Hz. These time series were generated synthetizing signals simulating Rayleigh and Love waves arriving at a recording station from sources randomly located around the station at distances between 100 m and 1 km. For each source, signal includes 50000 harmonic components spaced in frequency by 0.001 Hz, whose spectral amplitudes and phases, assumed equal at the source, are modified as function of wavelength, simulating an anelastic attenuation through a medium with a low quality factors (25-30). To give a transient character to these signals, each of them is modulated through a cosine window having a duration variable between 0.5 to 5 s. These transient signals emerge from a casual background noise of Gaussian type of mean 0 and standard deviation varied, among different synthetics, to obtain different signal-to-noise ratios. This choice was motivated by the results of preliminary tests conducted on real noise recordings (see Del Gaudio, 2013), which showed that only a small fraction of the recordings has a well defined polarization of Rayleigh or Love type. This probably occurs because, for most of the noise recording, signals of different polarization overlap and a signal with a specific kind of polarization can be identified only when it has much more energy than the others. Rayleigh wave ellipticity of each harmonic component was attributed according to an H/V curve simulating the presence of 40 m thick soft layer with an S-wave velocity of 300 m/s overlying a stiffer bedrock, of 800 m/s. This velocity contrast generates an H/V peak value equal to 3.37 at 1.9 Hz. With regard to ground motion direction, two kinds of situations were simulated, one characterized by polarization controlled by wave propagation direction (i.e. with horizontal component parallel and transversal to this direction for Rayleigh and Love waves, 76 GNGTS 2015 S essione 2.2