Probabilistic Seismic Hazard Assessment: Combining Cornell-Like Approaches and Data at Sites through Bayesian Inference

Abstract

Abstract The societal importance and implications of seismic‐hazard assessment forces the scientific community to pay increasing attention to the evaluation of uncertainty in order to provide accurate assessments. Probabilistic seismic hazard assessment (PSHA) formally accounts for the natural variability of the involved phenomena, from seismic sources to wave propagation. Recently, increased attention has been paid to the consequences of alternative modeling procedures on hazard results. This uncertainty, essentially of epistemic nature, has been shown to have major impacts on PSHA results, leading to extensive applications of techniques like the logic tree. Here, we develop a formal Bayesian inference scheme for PSHA that allows us, on the one hand, to explicitly account for all uncertainties and, on the other hand, to consider a larger set of sources of information, from heterogeneous models to past data. This process decreases the chance of undesirable biases and leads to a controlled increase of the precision of the probabilistic assessment. In addition, the proposed Bayesian scheme allows (1) the assignment of a subjective reliability to single models, without requirement of completeness or homogeneity, and (2) a transparent and uniform evaluation of the strength of each piece of information used on the final results. The applicability of the method is demonstrated through the assessment of seismic hazard in the Emilia–Romagna region of northern Italy. In this application the results of a traditional Cornell–McGuire hazard model based on a logic tree are updated with the historical macroseismic records to provide a unified assessment that accounts for both sources of information.