GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 2.2 GNGTS 2024 Capturing spaHal variability of empirical site amplificaHon funcHons δS2S: a case-study in Central Italy S. Sgobba 1 , C. Feliceja 1 , G. Lanzano 1 , F. Pacor 1 , T. BortoloU 2 , A. Menafoglio 2 1. Is9tuto Nazionale di Geofisica e Vulcanologia, INGV (Milan, Italy) 2. MOX, Department of Mathema9cs, Politecnico di Milano (Milan, Italy) Among the techniques for predicCng seismic shaking for damage scenarios or regional hazard and risk assessment, the empirical methods based on the use of a ground moCon model (GMM) are probably the most rapid and widespread. However, to produce the esCmates at ground-level, we need a key-proxy to describe the effect of the local site response, which is usually represented by the shear wave velocity in the first 30 m depth (Vs30) from in situ geophysical measurements or inferred from other proxies (geology, topography, etc.). Although Vs30 has the advantage of being a syntheCc predictor, on the other hand it has been shown that it is not truly representaCve of site response ( Castellaro et al., 2008; Luzi et al., 2011; Bergamo et al., 2021 ), thus, alternaCve site proxies called site-to-site terms (δS2S) are being explored; they represent systemaCc deviaCons of observed amplificaCon at a site from the median values predicted by a parCal or fully non-ergodic GMM calibrated on a set of reference rock sites ( Lanzano et al., 2022; Kotha et al., 2020 ). To date δS2S can be considered the most reliable representaCon of empirical site response if esCmated from a sufficient number of site-specific ground-moCon observaCons ( Bard et al., 2020; Loviknes et al., 2021 ). To increase the use of δS2S for site characterizaCon studies, there is a need to map this parameter at locaCons in space that lack direct observaCons. GeostaCsCcal methods, such as the simple Kriging technique, are commonly used for this purpose, but limitaCons arise in less densely sampled areas or in complex regions (e.g. characterized by the presence of different geological and tectonic structures, alluvial basins, mountain chains, etc. within a few kilometers) due to a too coarse interpolaCon. To address this issue, we explore the staCsCcal correlaCons of δS2S in terms of PGA and elasCc spectral acceleraCon (SA) up to 2s, with other mappable site proxies, in order to calibrate a parametric model capable of constraining the spaCal predicCons to other proxies or geological/lithological informaCon that are available in the area. In our case the proxies are: Vs30 measurements, high-frequency abenuaCon decay parameter κ0 ( Anderson and Hough, 1984 ), topographic slope from the Tinitaly DEM at 40m ( Mascandola et al., 2021 ) and local descripCons of the litho-straCgraphic units of the region from large-scale maps (i.e. the