Uncertainty specification and propagation for loss estimation using FOSM methods

Probabilistic prediction of structural and nonstructural damage costs due to future earthquakes is one component of loss estimation currently being developed for use in performance-based earthquake engineering. Sources of uncertainty in this prediction include epistemic and aleatory uncertainty in the site ground motion hazard, the building response, the damage measures of each of the many building elements, and repair cost of each of the elements. These are inter- and cross-correlated random variables. Two desired results are the total uncertainty in annual losses, and the contribution of each uncertainty source to the total uncertainty. Monte Carlo simulation is a simple solution, but it can be computationally expensive. This study proposes an alternative approach using First-Order Second-Moment (FOSM) methods for all but the (dominant) ground motion intensity variable. Suggestions for characterization of correlations are presented. A procedure for applying FOSM methods in the calculation of total uncertainty is outlined. The proposed technique is very efficient, and easily used for sensitivity studies.