GNGTS 2019 - Atti del 38° Convegno Nazionale

742 GNGTS 2019 S essione 3.3 We address this issue by developing a two-station method that allows measuring the frequency-dependent arrival azimuths of the Rayleigh waves at the receivers. The implementation of the method relies on the fact that the radial and vertical components of a Rayleigh wave are phase-shifted by π/2, and therefore the Hilbert transform (e.g. Claerbout, 1985) of the radial component is approximately in phase with the vertical one (e.g. Stachnik et al. , 2012; Ensing and van Wijk, 2018). Therefore, the arrival azimuth at each receiver can be estimated by seeking the angle for which the squared difference between the vertical and Hilbert-transformed radial component is minimum. At each period, the two arrival azimuths thus retrieved (one per station) are then averaged and used for calculating the arrival angle of the wavefront with respect to the interstation great-circle path, similarly to Foster et al. (2013). Finally, the average arrival angle can be used to apply a correction for retrieving the true phase velocities, rather than the apparent ones, with the direct effect of decreasing the measured phase velocities. This correction also accounts for both slight misalignments of the epicenter and stations and errors in source localization. Application to Central-Western Mediterranean data. The effects of correcting for arrival azimuth are evaluated on a data set consisting of 443 shallow teleseismic events (depths ≤ 50 km and epicentral distances between 20° and 150° ) with magnitudes between 6.5 and 8.5, recorded from January 2005 to January 2019 by 361 public stations distributed across the Central-Western Mediterranean, resulting in ~52000 triplets of stations and sources approximately lying along the same great circle path (we set a maximum azimuthal deviation of 5°) and ~16000 station pairs to be used for comparing the dispersion curves obtained from EQ and AN. This dataset allowed us to retrieve ~12000 EQ interstation dispersion curves corrected for the arrival azimuth and ~15000 ones without applying the correction; the lower amount of EQ- corrected dispersion curves is attributable to the grid-search criteria employed for this study in the research of the arrival angle at the receivers, which brings to reject a greater number of measurements than EQ without correction. The discrepancy between results from AN and EQ without correction is roughly ~1% of the AN phase velocity, similar to previous results (e.g. Yao et al. , 2006; Kästle et al. , 2016; Kästle et al. , 2018), and significant e.g. from the point of view of tomographic imaging and geodynamic interpretation. Fig. 2(b) shows a comparison of the mean EQ-AN discrepancy before and after arrival-angle correction. The correction leads to a significant improvement of the EQ-AN fit at periods of 20 to 50 s. The same result is Fig. 2 - (a) Dispersion curves obtained for four selected station pairs (station acronyms and interstation distance as indicated), and (b) differences between the dispersion curves obtained from the AN and EQ methods, averaged over all pairs. The black, grey and red curves show the results of the AN method, and the EQ method without and with arrival-angle correction, respectively. The dashed blue lines show reference velocity values (PREM, Dziewonski and Anderson, 1981).