Selection of Empirical Green's Functions by Waveform Similarity Analysis: An Approach to Predict Ground Motion in Areas with Saturated Records

Abstract

Abstract This work is focused on a procedure based on selecting suitable empirical Green’s functions (EGFs), able to predict ground motion for moderate earthquakes, in the case where no records are available due to saturation phenomena. The aim of the article is to generate synthetic seismograms for the 24 November 2004, M L 5.2, Salo earthquake (northern Italy), an event capable of saturating all velocimetric stations installed within the first 100 km from the epicenter. The proposed approach uses a waveform similarity analysis, based on the normalized cross-correlation technique, and it is able to identify EGFs that represent doublet events of a target. The ground motion was finally simulated using the method proposed by Irikura (1986). In this case, due to the saturation of near-source velocimetric instruments, the normalized cross-correlation matrix was calculated considering the first not saturated velocimetric station (ASO2, 108 km northeast to the epicenter), including the same selected portion of signals (on the base of signal-to-noise ratio) related both to the target and to 11 events, with M L ranging from 2.2 to 3.0, that occurred in the same area. The similarity analysis, performed through the bridging technique, allows us to detect an M L 2.9 aftershock characterized by a meaningful degree of similarity (70%) compared to the target. Given as a fact that if two events are similar for a far-field station they have to be similar also for a near-source one, it was in this way possible to use the selected aftershock (doublet), recorded in near source to reproduce the target for the saturated near-source stations. The results of the simulations were compared with ground-motion values predicted by empirical ground-motion prediction equations (GMPEs), calibrated using both Italian and European data.