GNGTS 2019 - Atti del 38° Convegno Nazionale

GNGTS 2019 S essione 3.1 571 continental crust (Calvet et al. , 2013, Borleanu et al. , 2017, Mayor et al. 2016). The present study is devoted to extend, for the first time, the methodology on larger regional scale. Specifically, it aims to characterize the Tyrrhenian Sea by focusing on its lateral variations in terms of seismic attenuation and its variations with frequency. Data and method. The dataset includes velocity waveforms recorded by broadband stations in Italy and Corse. We collected data from the INGV (Istituto Nazionale di Geofisica e Vulcanologia) and the RESIF (Reseau sismologique et géodésique français) permanent seismic networks and LISARD (Lithosphere of SARDinia), a temporary seismic network which operated from July 2016 to November 2018. We selected data of the major events (M W >4) occurred in central Italy in the last ten years in order to achieve the best coverage in the Tyrrhenian Sea. The epicentral distances range from 400 km up to almost 800 km. Most of the data are characterized by focal depths in the range 1-20 km. The heterogeneous ray density is shown in Fig. 1. After the deconvolution of the instrument response, a fourth-order Butterworth bandpass filter is applied to each seismogram to obtain waveforms in the frequency bands 0.4−0.8Hz and 0.8-1.2Hz. Fig. 1 - Data coverage. The blue circles and the red triangles indicate respectively the epicentres and the seismic stations. According toAki and Chouet (1975), the energy envelope of seismic coda waves (calculated using the Hilbert transform of seismograms) decays as where E is the power spectral density, S ( f ) is a frequency-dependent source and/or site term, t is the lapse time, f is the frequency and Q c is the frequency-dependent quality factor of coda waves. At late lapse time in the coda, coda waves are composed of multiple-scattered waves and, after a certain time threshold, the coda can be modeled with the diffusion equation. In this assumption Q c is a measurement of seismic absorption. In the selected coda window, Q c –1 values are calculated through a linear regression using the energy exponential decay law. Only the envelopes that show a coda decay with correlation coefficient greater than 0.5 for lapse time 220- 320s are stored for successive analysis. In the diffusive approximation, we set up the inversion