Damage-informed seismic fragility of non-structural elements in South Iceland: A Bayesian hierarchical approach

Abstract

This study assesses the seismic fragility of residential buildings in Iceland, examining both structural and non-structural damage resulting from the earthquakes of 2000 and 2008. Leveraging detailed loss datasets from these events, we apply a Bayesian Hierarchical Fragility Framework with a separate Bernoulli approach to derive empirical fragility functions for multiple damage states. This method integrates a spatially correlated latent intensity measure field inferred from ground-motion prediction equations (GMPEs), explicitly capturing both the spatial correlation and inherent uncertainty of GMPEs using Markov-Chain Monte Carlo (MCMC) sampling. By simultaneously modeling all damage states within a unified framework, while assigning distinct fragility parameters to each state, this approach leverages damage observations from all states to infer the latent IM field. This joint inference captures both the variability in the experienced IM at each building and the uncertainty in how that IM translates into observed damage responses. This research marks the first effort to specifically quantify the seismic fragility of non-structural damage in Icelandic buildings. Non-structural elements, which represent a significant proportion of total construction costs, contribute extensively to losses in global seismic events, often occurring even at low levels of ground shaking. The results underscore that non-structural components like interior fixtures and flooring are particularly vulnerable, more so than structural elements. Despite improvements in seismic codes, this study reveals a persistent vulnerability, underscoring the urgent need for targeted design and construction strategies that enhance the resilience of non-structural elements to ensure better overall building seismic functionality.