GNGTS 2013 - Atti del 32° Convegno Nazionale

thick sediments because they are overconservative at short periods and underconservative at long periods. Stewart et al. (2003) enlarged the database to other tectonically active regions and added to the California earthquakes events from Turkey and Japan, concluding that neither shear-wave classification nor detailed surface geology can provide an optimal predictive scheme when long periods are concerned (T>1 s). Castellaro et al. (2008) re-examined the original data by Borcherdt (1994) finding that, despite the conclusion of the author, the V s,30 appears a weak proxy to seismic amplification, which is too complex to be described by a single parameter. Lee and Trifunac (2010) also demonstrate that the average soil velocity in the upper 30 m should not be used to scale strong motion amplitudes and sustained the necessity of other parameters, which should be related to the response of sediments of large depths (i.e. 200 m), that they call “soil-site conditions”. The seismic response of a site can be described by other parameters than Vs,30. Examples are the frequency at which the first peak of amplification occurs (fundamental frequency/ period) or the frequency corresponding to the maximum peak of amplification (predominant frequency/period). During the 1980 classification schemes alternative to those based on Vs,30 were proposed, such as the Japan Road Association classification (1980, 1990), which, in parallel to the V s,30 , used the predominant soil period to discriminate among classes, accounting, although in an indirect way, for soil thickness, the parameter invoked by many authors as the principal deficiency imputable to V s,30 . Recently, a two-parameter classification approach has been attempted by Cadet et al. (2008), who propose to categorize soils according to the average shear wave velocity, V s,z , over the top z meters (z in the range 5–30), and the site fundamental frequency f 0 . A comprehensive analysis on about 500 sites from the KIKNET network, shows that f 0 is very poorly correlated with any of the V s,z values, thus providing independent, complementary information on the overall thickness and stiffness of sedimentary cover, and the surface stiffness. They obtained that the best combination was given by the couple of variables (V s,30 , f 0 ), but they also sustained that a simpler site classification, based only on f0,may lead to satisfactory results. The objectives of this study are: i) to find soil classes with a similar response to an earthquake; ii) to develop empirical amplification factors for 5% damped response spectral acceleration in the period range T 0.04-4s; iii) to identify proxies for site classification (V S,30 , stratigraphy, fundamental frequency, etc.). We propose a soil classification, which is not based on a priori subdivision, but it relies on the natural aggregation of empirical amplification function of the sites, obtained by normalizing the spectra of recorded motions by a reference (rock) spectrum from a ground motion prediction equation [GMPE, developed by Bindi et al. (2011), based on the ITACA 1.0 database]. The empirical amplification functions of sites are aggregated by means of cluster analysis. Data set. The data set, used for the analysis, consists in the strong motion recordings from 1972 to 2011 (ITACA 1.1; http://itaca.mi.ingv.it/ItacaNet/) , plus the recordings of the major events of the 2012 Emilia seismic sequence, of the national accelerometric network (RAN), operated by the Italian Civil Protection Department (DPC). In the framework of the projects INGV-DPC S6 2004-2006 (http://esse6.mi.ingv.it/ ) and S4 2007-2009 (http://esse4.mi.ingv.it/ ), several recording sites were characterized with different geophysical tests (down-hole, cross-hole, microtremor arrays, SASW or MASW). If we account for the studies conducted by ENEL after the Friuli 1976 and Irpinia 1980 seismic sequences, about 100 sites of the RAN are characterised by a shear wave velocity profile down to a minimum depth of 30 m. Moreover, the DPC promoted microtremor measurements at about 200 recording stations. The fundamental frequencies (f 0 ) of each station has been estimated through the horizontal to vertical spectral ratio computed from Fourier spectra of microtremor measurements and has been compared with the fundamental frequencies obtained 313 GNGTS 2013 S essione 2.2