GNGTS 2019 - Atti del 38° Convegno Nazionale

GNGTS 2019 S essione 3.3 743 illustrated by Fig. 3, where the differences between individual EQ andAN dispersion curves are presented in the form of histograms at different periods. At each period, whereas the standard deviations have the same order of magnitude, the mean difference obtained after correcting is significantly closer to zero than before the correction. We infer that most of the discrepancy reported in literature between AN and EQ can be explained by the neglect of wavefront distortion. Accounting for the arrival azimuths results in a > 300% decrease of the velocity bias at most of the investigated periods. By means of a Kolmogorov-Smirnoff test (e.g. Press et al. , 1992), we found that the probability the histograms of EQ and EQ-corrected in Fig. 3 correspond to the same statistical distribution is ~4% at 20 s, and systematically below 1% at longer periods, indicating that the improvement brought by arrival-angle correction is statistically significant. Fig. 3 - Histograms of phase-velocity differences between EQ and AN at the periods of 20, 25, 30, 35, 40, and 45 s. Measurements obtained with/without the correction for the arrival azimuths are indicated in red and grey, respectively.