GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 2.2 GNGTS 2024 chart of ISPRA 1:100.000 and the classificaCon provided by Forte et al., 2019 ). On the basis of the resulCng correlaCons, we construct a geostaCsCcal methodology to model the spaCal dependencies observed in the site terms, tesCng both the spaCal staConary and non- staConary hypotheses; for the laber, we apply the Universal Kriging technique that captures the local variability and allows us to produce high-resoluCon maps of δS2S that are constrained to other parameters; i.e. the Vs30 data and the other site-related proxies. The need for dense informaCon in the region and low-uncertain esCmates of ground moCon parameters, moCvates the choice of central Italy as a case-study area. In this region, in fact, we dispose of a very high number of seismic events recorded in the last 20 years (more than 30.000 records of events with magnitude ranging from 3.2 and 6.5) and the calibraCon of specific regional GMMs in a completely non-ergodic framework ( Sgobba et al., 2021 ; Sgobba et al., 2023 ). Key-findings ● The staCsCcal correlaCon analysis shows that the δS2S esCmates at different spectral periods are mostly correlated with Vs30 measurements, especially at longer periods (from about 0.3s), while are less dependent on the other invesCgated proxies (κ0, slope, lithology, etc.); ● The best model of δS2S ( Figure 1 ) is obtained with a non-staConary parametrizaCon based on geographical coordinates (x, y) and a combined Vs30-map ad-hoc developed for the study . The parametric funcConal form (the same for all spectral parameters) is the following: δS2S ∼ β0 + β1x + β2 y + β3 (x*y) + β4 (Vs30-map) + model error. Note that the Vs30-map was constructed by combining the dataset of observaCons and esCmates inferred from the topographic slope (Vs30-WA), according to Wald and Allen's global empirical relaConship ( Wald and Allen, 2007 ), and the Italian map (Vs30- Mori) provided by Mori et al. ( Mori et al., 2020 ); indeed, we observed that neither Vs30-WA nor Vs30-Mori alone were able to capture the full range of variability of Vs30 observaCons; ● Non-staConary spaCal models including lithology were also tested but they are not able to significantly improve the predicCons of δS2S, suggesCng that the current lithological classificaCon or resoluCon may not be useful for construcCng spaCal models of the invesCgated parameters; ● The final opCmal model is able to capture the main trend of the empirical site response in central Italy, but it shows some limitaCons in reproducing the full range of local amplificaCons and the observed variability between different frequencies, suggesCng that addiConal key-proxies are needed, such as the fundamental frequency of soil deposit f0, the sediment thickness, as well as alternaCve maps of geo- lithological complexes.