Probabilistic Seismic Hazard Analysis using Nonergodic Ground-Motion Model in Central Italy

Abstract

Conventional Probabilistic Seismic Hazard Analysis relies on Ground Motion Models based on the ergodic assumption, whereby spatial variability observed in regional/global datasets is considered representative of temporal variability at a single site. This can inflate aleatory variability and hide local effects. This thesis investigates the transition effects from ergodic to partially and fully nonergodic GMMs in Central Italy. Using the OpenQuake Engine and the GMM of Sgobba et al., (2021),seismic hazard was evaluated at four accelerometric stations with contrasting site conditions and observational coverage: IT.LSS, IV.NRCA, IV.LNSS, and IT.SBT. The epistemic uncertainties in median predictions, location-specific adjustments, and aleatory variability were propagated through a logic tree. The results show that modeling repeatable source, path, and site effects reduces aleatory variance by approximately 61.8-69.5%, producing narrower conditional ground motion distributions, steeper hazard curves, and generally lower exceedance probabilities at high ground motion levels. At the 2,475-year return period, ergodic UHS exhibit a pronounced peak at T = 0.5 s due to higher aleatory variability, while nonergodic UHS shift the peak to T = 0.2 s at IV. NRCA, highlighting short-period amplification from the shallow sedimentary basin. Source-area definition significantly affects site-specific hazard. Compared with the ZS16 Umbria-Abruzzo zone, the newly developed smaller clustered area increases hazard at nearby sites, with diminishing effects at greater distances. Epistemic uncertainty is narrower at IV.NRCA due to dense local observations, but wider at outer-cluster stations, where regional records dominate. Thus, local records are crucial for constraining uncertainty. Overall, nonergodic PSHA provides a more physically consistent representation of site-specific hazard by separating spatial effects from aleatory variability while accounting for epistemic uncertainty.

Conventional Probabilistic Seismic Hazard Analysis relies on Ground Motion Models based on the ergodic assumption, whereby spatial variability observed in regional/global datasets is considered representative of temporal variability at a single site. This can inflate aleatory variability and hide local effects. This thesis investigates the transition effects from ergodic to partially and fully nonergodic GMMs in Central Italy. Using the OpenQuake Engine and the GMM of Sgobba et al., (2021),seismic hazard was evaluated at four accelerometric stations with contrasting site conditions and observational coverage: IT.LSS, IV.NRCA, IV.LNSS, and IT.SBT. The epistemic uncertainties in median predictions, location-specific adjustments, and aleatory variability were propagated through a logic tree. The results show that modeling repeatable source, path, and site effects reduces aleatory variance by approximately 61.8-69.5%, producing narrower conditional ground motion distributions, steeper hazard curves, and generally lower exceedance probabilities at high ground motion levels. At the 2,475-year return period, ergodic UHS exhibit a pronounced peak at T = 0.5 s due to higher aleatory variability, while nonergodic UHS shift the peak to T = 0.2 s at IV. NRCA, highlighting short-period amplification from the shallow sedimentary basin. Source-area definition significantly affects site-specific hazard. Compared with the ZS16 Umbria-Abruzzo zone, the newly developed smaller clustered area increases hazard at nearby sites, with diminishing effects at greater distances. Epistemic uncertainty is narrower at IV.NRCA due to dense local observations, but wider at outer-cluster stations, where regional records dominate. Thus, local records are crucial for constraining uncertainty. Overall, nonergodic PSHA provides a more physically consistent representation of site-specific hazard by separating spatial effects from aleatory variability while accounting for epistemic uncertainty.