Earthquake catastrophe models provide probabilistic estimations of building portfolio losses that result from seismic hazards. Models that are designed to be used with large numbers of buildings often implicitly account for uncertainty associated with model components, but do not allow for an explicit comparison of different sources of uncertainty. We propose probabilistic methodologies that allow for explicit differentiation of the uncertainty in estimated portfolio losses as a function of the uncertainties associated with seismic hazard and building vulnerability. Our work is motivated by recent earthquakes in New Zealand and Japan, which highlighted the importance of understanding the sources of uncertainty that contribute to portfolio-level loss estimates. The proposed methodologies allow us to differentiate seismic risk between building portfolios as a function of the level of knowledge (epistemic) uncertainty associated with modeling seismic rates and the levels of aleatory variability and epistemic uncertainty associated with the buildings' vulnerabilities. A sample analysis incorporates epistemic uncertainty corresponding to Southern California's seismic rates into modeled losses for a Southern California building portfolio and then compares the impact of that particular source of uncertainty to the results from a portfolio in Northern California. For sources of uncertainty associated with buildings' vulnerability, we use our methodology to compare the impacts of aleatory and epistemic uncertainties on estimated portfolio losses as a function of the buildings' attributes, including their types of construction and numbers of stories. Component uncertainties (i.e., uncertainties corresponding to the seismic hazard and vulnerability) are assumed for our analyses so that the methodologies can be demonstrated.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below