GNGTS 2022 - Atti del 40° Convegno Nazionale

GNGTS 2022 Sessione 2.2 315 TOWARDS AN ALTERNATIVE APPROACH TO DEFINE THE DESIGN EARTHQUAKE SCENARIO AT A SITE R. Paolucci, A. Chiecchio, M. Vanini Department of Civil and Environmental Engineering, Politecnico di Milano, Milano, Italy Introduction. The estimation of the characteristics of the ground shaking that occurs during damaging earthquakes is critical both for an efficient risk mitigation and for a correct design of seismic resistant structures (Douglas and Aochi, 2008). In this context, Probabilistic Seismic Hazard Analyses (PSHA) give an estimate of motion that structures and infrastructures need to withstand in terms of clear and communicable probabilities able to support and guide decision making in the mitigation process (Weatherill et al. , 2020). The primary advantage of a PSHA, over alternative representations of the earthquake threat, is the possibility of integrating over all possible earthquake occurrences and ground motions (McGuire, 1995). In this framework, the final output of such analyses is given in terms of a Uniform Hazard Spectrum (UHS), which represents the convolution of many different earthquake magnitude and distance ( M w - R ) pairs for a fixed annual probability of exceedance. Nevertheless, asmany practical applications require the knowledge of the earthquake scenario which is mostly contributing to the hazard at the site under study, in the last couple of decades it has become common practice to disaggregate the hazard into its scenario contributions: magnitude ( M w ), source-to-site distance ( R ) and epsilon ( ε ). This latter quantity defines the number of standard deviations by which the logarithmic ground motion prediction deviates from the median value (Barani et al. , 2009). Despite their extensive use, the disaggregation procedures developed in the last years suffer from a series of uncertainties which limit their practical application and often make results interpretation ambiguous. The selected ( M w - R ) pairs depend on different choices, some of which may be of crucial importance. Besides the dependence on the structural periods, ( M w - R ) analyses are developed for a specific branch of the logic tree used in the PSHA, so for a specific combination of ground motion models (GMMs), their values may be calculated from mean, modal or median contributions, evaluated with different procedures (e.g., different bin values). While modal values represent the most likely combination of ( M w - R ) pair, mean values are simpler and easier to be understood, but their results may be sometimes poorly related to the physical, known, dominating sources at the site. In order to shed light and, possibly, overcome these ambiguities, this work suggests an alternative approach to disaggregation that aims at defining a Best Matching Scenario Earthquake (BMSE), fitting the UHS at a specific site and for a return period of interest, at all structural periods, using one, or a combination, of GMMs predictions. In practice, for a given target spectrum, the method selects iteratively the ( M w - R - ε ) triplet that minimizes, with an acceptable tolerance, the difference between the “predicted” spectral amplitudes and the “target” ones. Method. An application of the method to a few sites of the Italian territory is illustrated herein, taking as reference (target spectra) the UHS calculated for the official Italian Seismic Hazard (SH) Map (Mappa di Pericolosità Sismica, MPS04 Working group, 2004). The GMM selected for the calculation of the prediction was the model of Ambraseys et al. (1996), that is one of the GMMs used in the framework of the MPS04 SH analysis. For these preliminary analyses we have used, for simplicity, a value of ε =0 (corresponding to the median spectrum prediction). The difference (Eq. 1) between the logarithm of the AMB96 prediction (Sa GMM ) and the logarithm of the UHS target (Sa Target ), for a specific ( M w - R ) pair, was therefore computed. The spectrum associated to the ( M w – R) scenario that minimized the maximum difference over all periods i (see Eq. 2) was considered as the Best Matching Scenario.