GNGTS 2013 - Atti del 32° Convegno Nazionale

the research efforts on the area struck by the May-June 2012 Emilia seismic sequence, with special consideration for three sites: Mirandola (Modena), Casaglia (Ferrara), and Soncino (Cremona), the first two of which house accelerometer stations. As the study produced calculations both at regional and local scales, different methods were applied, as summarized in Tab. 1, and briefly described in the following. The simplest approach consists of using one or more ground-motion predictive equations (GMPEs) with parameter values tuned for specific ground profile categories, or soil classes (Level 0). In other words, this approach assumes that the soil conditions at the site of interest reproduce the average site effects at the seismic stations considered for the development of the GMPEs. Another widely used approach, which however cannot be considered as rigorously probabilistic, consists of multiplying the probabilistically evaluated response spectra on rock by a deterministic frequency-dependent (or, sometimes, independent) amplification function (or factor) derived, for example, from numerical soil response analysis (Level 2). In many applications, deterministic amplification functions can be replaced by average amplification factors taken from national building codes (Level 1) which, however, may account for actual local soil conditions only in a broad sense. A fully probabilistic method that allows for site-specific response is the so-called non-ergodic PSHA (Level 3) which relaxes the ergodicity assumptions implicit in the use of standard GMPEs by introducing the use of single-station standard deviation values (single- station sigma) in place of the ergodic sigma of the GMPE and by correcting the predicted median ground motion to account for site specific behaviour. In simple words, using traditional GMPEs implies accepting that the ground-motion variability computed from a global data set including recordings from multiple sites and from multiple earthquakes is an unbiased estimate of the variability of ground-motions at a single site (Anderson and Brune, 1999; Al Atik et al. , 2010; Rodriguez-Marek et al. , 2011). Repeatable and systematic effects of path and source and the effects of the same soil site conditions should in general make the ground-motion variability at a single site smaller than that computed utilizing records from other sites with similar soil conditions affected by other earthquakes with different paths and sources. Hence, part of the aleatory variability in the ground motion prediction can be translated into epistemic uncertainty. Tab. 1 – Classes of approaches to account for site effects in PSHA. SAF is for Soil Amplification Function. Hybrid probabilistic/deterministic Fully probabilistic Generic level Site-specific Generic level Site-specific LEVEL 1 PSHA at rock + SAF based on seismic norms LEVEL 2 PSHA at rock + SAF based on 1D analyses LEVEL 0 PSHA based on GMPE with site correction factors LEVEL 3 PSHA at site with single-station sigma applied LEVEL 4 PSHA on rock convolved with SAF conditioned to rock ground motion A further fully probabilistic site-specific approach is that proposed by Bazzurro and Cornell (2004a and 2004b) (Level 4). To the extent that soil amplification analyses provide an adequate representation of the seismic site response under the impact of strong ground-motion, the method of Bazzurro and Cornell accounts for local soil conditions in PSHA by combining in a probabilistically robust way (i.e., via the convolution of probability distributions) the hazard curve on rock at a site with the conditional probability distribution of the site amplification level obtained from numerical soil response analysis at the same site. If a linear predictive model for (log) in terms of (log) (where and indicate the amplification function and the spectral acceleration on rock at the oscillator frequency f , respectively) is appropriate, a simplified approach may be used (Level 4 simplified approach). The method consists of deriving closed-form equations to integrate directly into an existing rock 16 GNGTS 2013 S essione 2.1