PSI - Issue 44

Angela Chiecchio et al. / Procedia Structural Integrity 44 (2023) 11–18 Angela Chiecchio/ Structural Integrity Procedia 00 (2022) 000–000

12

2

1. Introduction It is well known that the probabilistic seismic hazard assessment (PSHA) involves consideration of all sources potentially contributing to the seismic shaking at the site, in a given time frame, and that the resulting uniform hazard spectra (UHS) are obtained by summing (aggregating) the exceedance probabilities from all such sources. As a consequence, UHS cannot be associated in a straightforward way to a single source, i.e., they cannot be directly associated to a single earthquake, characterized by a certain magnitude M w at a certain distance R . On the other side, many practical applications are specifically requiring the knowledge of the magnitude and distance ( M w - R ) pairs which is mostly contributing to the hazard at the site under study. For instance, it can be of interest to define one or more than one earthquake scenario at a site and to associate it to a proper probability of occurrence. Disaggregation procedures are commonly employed to rank the contributions to the PSHA results and to select, for a specific annual probability of exceedance and for a specific structural period, the “most representative” pair ( M w - R ), often called “design earthquake” (McGuire, 1995; Bazzurro and Cornell, 1999). However, such selection is affected by several ambiguities that make the disaggregation result non-unique and, because of that, of puzzling understanding and application for non-expert users. The selected ( M w - R ) pair depends on many choices that are often not transparent to the end user, such as: (i) the specific branch of the PSHA logic tree; (ii) the structural period, which moreover may vary consistently in time if the structure responds in a nonlinear field, (iii) the selected set of ground motion models (GMM), (iv) the choice of considering the modal, median or mean contribution, and (v) the number of standard deviations by which the logarithmic ground motion generated by a given couple of ( M w - R ) deviates from the median predicted ground motion value, also defined as ε (Barani et al., 2009). Therefore, more than a pair, the disaggregation results are actually a ( M w - R - ε ) triplet, where the spectra obtained from disaggregation may strongly deviate from the corresponding UHS, depending on the choice of the value of ε considered. For instance, Fig. 1 shows the UH spectra for 475 years for the site of L’Aquila (from MPS working group, 2004), compared to the spectra predicted with the Ambraseys et al. (1996) GMM, using the modal disaggregation ( M w - R ) parameters calculated by Barani et al. (2009) and different values of ε . The resulting spectral prediction changes considerably when, for the same ( M w - R ) parameters, a value of ε = 0 (i.e. the modal scenario), or a higher value obtained from disaggregation, is used. This of course poses a “warning” on the proper use of disaggregation results: with reference to Fig. 1, with T=0.2s, if the selection of accelerograms is based on M =4.8 and R =5km, without taking care of  =1.5, the spectral compatibility will be highly underestimated since the average response spectrum of the selected accelerograms will likely approach the median value (thin line in Fig. 1, top-left). Eventually, this will be fixed, in a more or less conscious way, by setting high and often unrealistic scaling factors for the selected accelerograms to approach the target spectral ordinates. An alternative, and simpler, paradigm to link in a non-ambiguous way a UHS to a design (or scenario) earthquake is presented in this paper: instead of looking, as in the disaggregation, for the “most likely” ( M w - R ) pair contributing to a UHS, the Best Matching Scenario Earthquake (BMSE), is introduced as the earthquake the median ground motion of which best approaches, in an arbitrarily broad period range, the target UHS. Although not a “panacea”, the BMSE may overcome several of the ambiguities previously associated to the disaggregation, especially when hazard is dominated by a specific nearby seismic zone or seismic fault, such as in the case of the Italian seismotectonic context. The paper is organized as follows. Section 2 introduces the method used for the definition of the best matching scenario, while sections 3 and 4 show an example of application of the BMSE and the corresponding results, obtained with reference to the current seismic hazard model of Italy. Finally, some conclusive remarks are given, aiming at pointing out the pros and cons of the approach. 2. Method Given the: (i) target UHS at a specific site, related to a certain return period, the (ii) ground motion model (GMM), and the (iii) seismogenic model of the sources dominating the hazard at the site, the BMSE is determined iteratively by searching for the ( M w - R - ε ) triplet, i.e., the scenario, that minimizes the difference between the target and the predicted spectrum, within a prescribed tolerance.