Study on the propagation of uncertainties in a probabilistic seismic risk model for portfolio loss assessment

Probabilistic earthquake loss models, similar to any other catastrophe model, are structured by three main components, namely a hazard input model, an exposure layer and a vulnerability model. Seismic risk is commonly defined as the convolution between the aforementioned elements. Uncertainties are associated with each of the components, and they are originating from the input of numerous parameters that define the seismicity, the ground motion intensity, the exposure characteristics of a building inventory (e.g., taxonomy, spatial resolution) and the seismic vulnerability. Thus, the final loss estimation inevitably carries a high degree of dispersion, with different assumptions in the modelling process leading to considerably different results. The treatment of uncertainty in risk analysis from a conceptual point of view has been the topic of a wide discussion in literature. The study discussed in this manuscript lies in this perspective: in the context of a seismic risk assessment model, the main goal of the work is to create a solid framework allowing to (i) understand the importance of the different parts of the model in a first stage; (ii) have a deeper comprehension on how the uncertainty are propagated throughout the latter; (iii) give estimate of the outputs’ distributions, i.e., portfolio losses. The work makes strong use of Monte Carlo simulation and it will rely on parallel computing process through the use of high-performance computers (HPC) from CINECA within the HPC-TRES project.