GNGTS 2018 - 37° Convegno Nazionale

GNGTS 2018 S essione 3.3 759 x i ( t ). In this case, the ICA output is represented by three independent signals corresponding to the basis vectors s , while the estimated coefficient matrix A accounts for the amplitude of that independent wave at each location and for each direction of motion. In this way, the azimuth and incidence angle of the polarization vector are directly calculated from the ratio of the amplitude coefficients of the mixing matrix A . In the present work we propose two approaches based on the ICA that allow for event discrimination and wavefield decomposition. Specifically, we developed an automatic procedure to have a prompt and clear separation among the different sources in the seismic signals, such as meteo-marine microseism, anthropogenic noise, and VT earthquakes at Campi Flegrei volcanic area. Moreover, we introduce a coarse-grained variable, i.e. the Frequency associated with the Maximum Amplitude of the Power Spectral density of the ICs (FMPSDA). This parameter is sensitive to the variation in the frequency bands of interest (e.g. that corresponding to the corner frequencies of VT events) and can be used as marker of the insurgence of seismic activity (Fig. 1). In addition, we introduce a novel approach, the “ICA-based Polarization” (ICAP), which consists in the estimate of the polarization parameters directly from the ICs and it is based on the ICA ability in decomposing the earthquake wavefield into polarized basic sources. We applied the technique to both volcano-tectonic earthquakes occurred at Campi Flegrei, as well as tectonic seismicity occurred in the Amatrice area (Central Italy), and we estimated the direction of the ground motion for both the raw data and the ICs. The particle motion of the raw data was obtained for suitable time windows, containing the P and S-wave phases. On the contrary, the particle motion of the ICs was determined without a windowing procedure. The hodograms of the raw data and the ICs are in good agreement, proving that the ICA well decomposes the seismic wavefield in the vertical and horizontal planes (Fig.2 and 3). The ICs are clearly polarized and correspond to wave-packets containing the P- and the S- phases. The obtained results indicate that the ICA successfully discriminates and extracts different independent sources in continuous seismic signals. The coarse-grained procedure on massive data is able to detect the occurrence of even extremely low-energy VTs and the FMPSDA may represent a suitable parameter to monitor in the observatory practices. The approach can be employed for the prompt detection in massive data of other kinds of seismic signals such Fig. 1 - Time evolution of the FPSDMA, from 3 rd to 12 th October 2006 (no VT activity) and from 19 th to 28 th (occurrence of VTs). Upper plot: results for the raw signal, vertical component. Lower plot: results for the ICs, vertical component. The 12–16 Hz frequency band corresponding to VT activity is evidenced in yellow.