PSI - Issue 78

Iunio Iervolino et al. / Procedia Structural Integrity 78 (2026) 1553–1560

1559

requires first selecting one return period value. The lowest value from literature, that is, 700 years, is adopted in this preliminary analysis.

Fig. 4. Probability density functions of the interarrival times according BPT, SPM and HPP model (a), probabilities of observing at least one earthquake in a time interval of 200 years, from to +200 , given that no earthquakes occurred between zero and (b). The parameters of the Weibull distribution in Eq. (9) are then calibrated to produce a coefficient of variation equal to 0.5, matching the value recommended by Akinci et al. (2017) for the BPT model. The resulting values of and are 1.2667E-03 and 1.9522, respectively. Note that the relationship between interarrival time and earthquake magnitude is not considered here, as this section focuses only on the occurrence of earthquakes, without explicitly accounting for magnitude-dependent effects. Finally, Fig. 4a also includes the PDF computed under the HPP assumption with = 739 years. While BTP and SPM distributions exhibit similar shapes, the HPP distribution follows the well-known exponential decay. Fig. 4b presents the corresponding probabilities of observing at least one earthquake in a time interval of 200 years, from to +200 , given that no earthquakes occurred between zero and , Eq. (6). When is lower than approximately 300 years, both non-homogeneous models yield lower probabilities than the HPP, the results of which are constant with respect to the last event. Because the last modelled earthquake on the source occurred in 1908, the elapsed time is 117 years. For this value of , indicated via the dotted line in the figure, the HPP model returns a probability equal to 0.237, whereas the BTP and SPM yields lower probabilities of 0.060 and 0.134, respectively. These results are consistent with the findings of Akinci et al. (2017) and suggest that, for the fault under consideration, using non-homogeneous models for earthquake occurrence leads to less severe hazard outcomes than the HPP assumption, when the models are calibrated to have comparable return periods. 4 Final remarks This study preliminarily addressed two aspects of seismic hazard analysis that are deemed relevant for the site of Messina, in southern Italy, located in proximity of the fault responsible for the 1908 M7.1 earthquake. First, the effects of pulse-like ground motions were investigated through near-source PSHA, conditional on the occurrence of an earthquake on the same fault that generated the 1908 event. Second, the study explored two alternative models for earthquake occurrence on the Messina Strait fault, to represent stochastic earthquake occurrence which depends on the seismic history of the source. The following is worth recalling. • For vibration periods between 3 s and 10 s, the mean spectral acceleration, conditional on the occurrence of one event with magnitude = 7 ± 0.3 on the considered fault, from NS-PSHA exceeds that from classical PSHA, by more than 50%. For vibration periods up to 1 s, hazard increments are lower than 10%. • The largest hazard increase due to pulse-like effects is equal to 64% for = 4 s , and it is larger than 60% for periods ranging from 3 s to 5 s, this interval including the median pulse periods conditional on magnitude in the = 7 ± 0.3 interval. However, absolute differences between NS-PSHA and PSHA do not exceed 0.08 g, for any vibration period. • The considered time-variant models for earthquake occurrence, yield lower probabilities of at least one earthquake occurring within the next two hundred years compared to the corresponding result obtained using the HPP. More specifically, today, that is 117 years since the last earthquake, the probabilities from BPT and SPM are 0.060 and 0.134, respectively, whereas the result from the HPP is 0.237. The BPT and SPM models begin to return probabilities larger than the HPP after approximately 300 years from the last event.