GNGTS 2019 - Atti del 38° Convegno Nazionale

414 GNGTS 2019 S essione 2.2 USE OF 3D PHYSICS-BASED NUMERICAL SIMULATIONS FOR PROBABILISTIC SEISMIC HAZARD AND RISK ANALYSIS M. Infantino 1 , R. Paolucci 1 , M. Stupazzini 2 1 Department of Civil and Environmental Engineering, Politecnico di Milano, Milano, Italy 2 Munich Re, Munich, Germany The Probabilistic Seismic Hazard Assessment (PSHA), (Esteva 1967,1968; Cornell 1968), is the approach generally adopted to evaluate the seismic hazard and, therefore, it is paramount to assess the seismic risk in large urban areas. PSHA aims at providing at a specific site the variation of a selected ground motion parameter in function of the return period by a suitable combination of seismotectonic information, frequency-magnitude relationship and ground motion propagation models. Here the attention is given to the last of the previous components of PSHA, i.e., the influence of different techniques used to describe ground motions at the site of interest. On this regard, the classical approaches rely on empirical ground motion models (GMMs), which, however, roughly take into account the specific tectonic and geotechnical conditions in which the urban area lies and tend to be poorly calibrated in the near-source region of large earthquakes. On the other hand, in the recent years, boosted by the increasing availability of computational resources, 3D physics-based numerical simulations (PBSs) of earthquake ground motion including a full 3D seismic wave propagation model from the source to the site, have gained increasing consensus thanks to the validation exercises (e.g. Razafindrakoto et al., 2018; Evangelista et al., 2017; Taborda et al., 2016; Goulet et al., 2015; Paolucci et al., 2015; Taborda and Bielak 2013; Pilz et al., 2011), so that they are expected to become, in near future, an alternative or complementary tool at GMMs to provide realistic ground motion estimations. Nevertheless, although the advantage of a Physics-Based PSHA, there are few examples in literature of practical applications of PBSs for PSHA (e.g. Convertito et al., 2006; Hutchings et al., 2007; Graves et al., 2011; Villani et al., 2014; Stupazzini et al., 2015; Tarbali et al., 2019), especially because of the large computational efforts required to carry out numerical simulations. In this work two approaches have been considered referred hereinafter as GAF and footprint respectively. The GAF approach is the one proposed by Villani et al. (2014) and implemented in the CRISIS software (Ordaz et al. , 2013) through the generalized attenuation functions (GAF). It consists of extrapolate the lognormal distribution fitting the frequency histogram of the PBSs at each site (Fig. 1,a) and replace the statistical moments of such distribution into the PSHA in place of the statistical moments of GMMs. However, for its intrinsic definition the lognormal probability model is unbounded on the upper side, differently from the frequency histogram which turns out to be naturally limited. This poses a well-known issue in PSHA (e.g. Bommer et al., 2004), since the integration along an unbounded range of variability of the ground motion Fig. 1 - Physics- based PSHA according to the GAF (a) and Footprint (b) approaches.