Scoring and Testing Procedures Devoted to Probabilistic Seismic Hazard Assessment

This review addresses long-term (tens of years) seismic ground-motion forecasting
(seismic hazard assessment) in the presence of alternative computational models
(the so-called epistemic uncertainty affecting hazard estimates). We review the different
approaches that have been proposed to manage epistemic uncertainty in the context of
probabilistic seismic hazard assessment (PSHA). Ex-ante procedures (based on the combination
of expert judgments about inherent characteristics of the PSHA model) and expost
approaches (based on empirical comparison of model outcomes and observations)
should not be considered as mutually exclusive alternatives but can be combined in a
coherent Bayesian view. Therefore, we propose a procedure that allows a better exploitation
of available PSHA models to obtain comprehensive estimates, which account for
both epistemic and aleatory uncertainty. We also discuss the respective roles of empirical
ex-post scoring and testing of alternative models concurring in the development of comprehensive
hazard maps. In order to show how the proposed procedure may work, we also
present a tentative application to the Italian area. In particular, four PSHA models are
evaluated ex-post against macroseismic effects actually observed in a large set of Italian
municipalities during the time span 1957–2006. This analysis shows that, when the whole
Italian area is considered, all the models provide estimates that do not agree with the
observations. However, two of them provide results that are compatible with observations,
when a subregion of Italy (Apulia Region) is considered. By focusing on this area, we
computed a comprehensive hazard curve for a single locality in order to show the feasibility
of the proposed procedure.