The impact of the choice of intensity measure and seismic demand model on seismic risk estimates with respect to an unconditional benchmark

Many methods for seismic risk assessment rely on the selection of a seismic intensity measure (IM) and the development of models of the seismic demand conditional on the IM. The individual importance of these two features to accurately assess seismic performance is well known. In contrast, this study aims to evaluate the impact that the combined selection of IM and the demand model has on risk estimates. Using a hypothetical seismic source model and a non-stationary stochastic ground-motion model, we present risk estimates for a mid-rise steel structure for 15 different IMs and five demand models derived by cloud analysis (four based on regression and a fifth based on an empirical binning approach). The impact of these choices is investigated through a novel method of model performance evaluation using a benchmark solution obtained via the unconditional approach (i.e., directly estimating demand exceedance frequencies from simulated ground motion time histories). The obtained results are also compared against traditional IM performance metrics, for example, efficiency and sufficiency. Finally, we demonstrate how risk estimate inaccuracies are propagated by performing a damage assessment on two example components. The results show that, for the scenario under investigation, Arias intensity combined with the binned demand model provides the best risk estimates, if sufficient samples are available, whilst ground displacement and duration-based IMs ranked worst, irrespective of the demand model. The findings highlight the importance and interconnectedness of the selection of the IM and the demand model when using cloud analysis and present a clear method of determining the most accurate combination for risk assessments.