GNGTS 2014 - Atti del 33° Convegno Nazionale

52 GNGTS 2014 S essione 2.1 For the choice of M max , the characteristic magnitude ( M W&C ) has been calculated according to the scaling law of Wells and Coppersmith (1994), which uses the area of the fault rupture. More precisely, regarding the 3 largest sources only a part of their length has been considered as rupture length following aspect-ratio considerations because no indications on their possible segmentation are available in the literature. The 3 fault planes that represent the Catanzaro source, however, have been taken as segments that can break along their entire length. The other dimension (width in depth) of the fault planes has been defined after an analysis of hypocentral depth distribution of recent seismicity, and a sensitivity test on hazard results; it never exceeds 35 km depth. M K&G has been calculated, as well. As in the previous case, we chose the higher value of the two magnitudes ( M W&C for the Rossano and the Crati sources and M K&G for Lakes and Catanzaro sources) and the related standard deviation has been adopted as uncertainty. Parameters characterizing the logic tree . The seismicity parameters applied for the calculation of seismic hazard in the used code [CRISIS by Ordaz et al. (2012)] are the λ and β values of the G-R relation and M max . The individual seismicity rates, used in the cumulative form in the calculation of the G-R parameters, have been estimated following the approach “higher not highest” (HNH) already applied to the seismic hazard maps of the Italian territory (Slejko et al. , 1998) taking into account of the earthquake catalogue completeness. The resulting linear fit obtained by linear regression analysis has been considered valid in the range of magnitudes between 4 ( M min ) and M max , with M max computed on geological (different for the two considered zonations) or statistical (Kijko and Graham, 1998) approaches. A logic tree (Fig. 2a) with 6 branches has been considered in the present study; it is composed of: • two source models, one is based on the composite sources of DISS (DISS Working Group, 2010) and the other is defined on the basis of alternative sources (sources debated, as reported in DISS) which have been reanalyzed in the present study (OGS branch in Fig. 2a); • one single approach (Slejko et al. , 1998) for the definition of the seismicity model (G-R parameters and M max ); • three groundmotion prediction equations (GMPEs) for horizontal peak ground acceleration (PGA) and spectral ordinates: the first one (C&F) is of global validity (Cauzzi and Faccioli, 2008), the second one (AMB) was calibrated on European data (Ambraseys et al. , 1996), and the third one (S&P) on Italian data (Sabetta and Pugliese, 1996). PSHAresults . The results, here presented as an example of the work done, refer to one of the sites hosting an operating dam. Seismic hazard has been estimated in terms of PGA and spectral acceleration (SA) for a 475-year return period, standard reference for seismic design, and for 101- and 1950-year return periods, respectively corresponding to the limit state of damage (SLD) and the limit state of collapse (SLC) in the Italian building code NTC08 (��������� ����� Ministero delle Infrastrutture, 2008�� ��� ������������ ���� ���� �������� ��� ����� ���� ������ �� ���� � ). The calculations have been repeated for three soil types: a) rock ( V S30 >800 m/s); b) stiff soil (360< V S30 ≤800 m/s) and c) soft soil (180< V S30 ≤360 m/s). Fig. 2b, for instance, illustrates the uniform hazard response spectra (UHRSs) obtained by the 6 individual branches of the logic tree, and the average spectrum (grey solid line) obtained as a weighted mean value of the individual spectra, with the related standard deviation. Note that here the mean value is represented, and not the median one (50th percentile) as in the MPS04 map (Stucchi et al. , 2011). Pros and cons of mean and median values are extensively debated in the literature (see e.g., ���������� ��� ������� ����� �������� ����� ������� Abrahamson and Bommer, 2005; McGuire, 2005; Musson, 2005�� ��� ���������� �� ��� ������ �� ����� ����� ���� ������ ���� ��� ������� ������ ��� ��� ). The dispersion of the curves is quite large with regard both the maximum values and the decay behaviour. In particular, the spectrum obtained by the application of the C&F GMPE is significantly higher than the others for the 0.1-s period. The UHRSs obtained for the six sites under study are compared in Fig. 2c, regarding only the 1950-year return period and stiff soil. Given the vicinity of the six sites, the UHRSs are quite similar with the maximum peak at 0.1 s.