GNGTS 2014 - Atti del 33° Convegno Nazionale

GNGTS 2014 S essione 2.1 71 This paucity of data recorded in near-field conditions dramatically affects the validity of almost all GMPEs, both for the vertical and horizontal components. Conversely, the recent Ferrara earthquake, together with its aftershocks, has provided a huge amount of data in the near field. These data, furthermore, have clearly shown that the vertical PGA can be similar to, or even larger than, the horizontal one. This aspect was already pointed out by the 2011 Christchurch earthquake (Kam and Pampanin, 2011). The literature review presented in Grimaz and Malisan (2014) confirms that the ratio vertical/ horizontal PGA is usually greater than 1.0 within 10-20 km from the epicentre (depending anyway from fault characteristics). Furthermore, Grimaz and Malisan (2014) highlighted as the vertical acceleration in the near field could be one of the epicentral effects capable of provoking serious and specific consequences on structures, being the vertical action simultaneous with the horizontal one. For this reason, they hypothesized the definition of a multilayer hazard map for considering also the specific effects in the near-field area, and among these, particular attention should be paid to the vertical component of seismic motion. In particular, they proposed to assess the near-field effect within an area where they are not negligible (named PEACH – Potential Epicentral-Area Contribution to Hazard). For the Italian territory, as a first glance, the proposal was to consider the individual seismogenic sources reported for example in the DISS project (DISS Working Group, 2010). In this work, following the above cited idea, recent GMPEs suitable for north-eastern Italy (the most seismic area of northern Italy, see Fig. 1a) have been applied for the computation of a regional seismic hazard map in terms of vertical PGA. Furthermore, the ground motion estimates related to an E-W oriented profile have been compared with those deriving from the direct application of the standards of the Eurocode 8 (CEN, 2004), i.e., scaling at 90% the horizontal estimates. These elaborations have been performed by considering different hypotheses about the applicability of the GMPEs in the very near field (outside their range of calibration) on the basis of a logic tree approach already applied for seismic hazard analysis in NE Italy (Slejko et al. , 2011). Methodology. In this work, the probabilistic approach for seismic hazard assessment proposed originally by Cornell (1968) has been applied by using the software Crisis 2012 (Ordaz et al. , 2012). The Cornell (1968) approach is based on two work hypotheses: 1) the earthquake recurrence intervals are exponentially distributed (Poisson process: made up by independent, non-multiple events, and the process is stationary in time) and 2) the magnitude is exponentially distributed [the Gutenberg - Richter (G-R) relation holds]. In addition, the seismicity is considered uniformly distributed inside the seismogenic zone (SZ). The Cornell (1968) method, then, needs the following input data: the SZ geometry definition, the seismicity models (in terms of average number of earthquakes per magnitude interval, and maximum magnitude), and the GMPE of the parameter chosen to represent the ground motion. The quantification of the uncertainties (McGuire, 1977) is a crucial point in modern probabilistic seismic hazard analysis (PSHA). Two kinds of uncertainties characterise the results in PSHA: the aleatory variability and the epistemic uncertainty (McGuire and Shedlock, 1981; Toro et al. , 1997). Aleatory variability is the natural randomness in a process. It is considered in PSHA taking into account the standard deviation of the relation describing the process. Epistemic uncertainty is the scientific uncertainty in the model of the process and it is due to limited data and knowledge. It is considered in PSHA using alternative models. The logic tree approach for PSHA (Kulkarni et al. , 1984; Coppersmith and Youngs, 1986) has been introduced for quantifying the epistemic uncertainties. Each node of the logic tree collects a series of choices, represented by each branch of the logic tree. The final aggregate result is obtained by weighting adequately the individual results coming from the different branches [see more discussion in Rebez and Slejko (2004a)]. The logic tree considered in this work is deliberately very simple and refers to that already used by Slejko et al. (2011): one zonation (FRI zonation, see Fig. 1a), one method for the