A Bayesian hierarchical model for climatic loads under climate change

In the mid-term future, climate change could determine significant alterations of the frequency and magnitude of climate extremes, so affecting the design of new structures and infrastructures, and the reliability of existing ones designed according to the provisions of present or past Codes. In this work, a Bayesian hierarchical model for the characterization of climate extremes under non stationary climate conditions is presented starting from the analysis of an ensemble of future climate projections. The Bayesian Hierarchical Model is formulated through the classical three-level formulation, in which the standard extreme value representation at each site is combined with a spatial latent process, and collects the main sources of uncertainties regarding climate projections. A Metropolis Hastings algorithm within a Gibbs sampler is implemented to update model parameters, and from the posterior probability density functions of the extreme value distribution parameters, return levels that serve as basis for structural design are estimated. The implementation of the model in different time windows combined with the Bayesian framework allows the probabilistic assessment of time evolution of extreme value parameters and return levels. The results obtained for a relevant case study demonstrate the possibilities of the proposed methodology to describe climate extremes under climate change and to provide guidance for potential amendments in the current definition of climatic actions on structures.