On the treatment of uncertainties in the development of fragility functions for earthquake loss estimation of building portfolios

State-of-the-art methods for the assessment of building fragility consider the structural capacity and seismic demand variability in the estimation of the probability of exceeding different damage states. However, questions remain regarding the appropriate treatment of such sources of uncertainty from a statistical significance perspective. In this study, material, geometrical and mechanical properties of a number of building classes are simulated by means of a Monte Carlo sampling process in which the statistical distribution of the aforementioned parameters is taken into consideration. Record selection is performed in accordance with hazard-consistent distributions of a comprehensive set of intensity measures, and issues related with sufficiency, efficiency, predictability and scaling robustness are addressed. Based on the appraised minimum number of ground motion records required to achieve statistically meaningful estimates of response variability conditioned on different levels of seismic intensity, the concept of conditional fragility functions is presented. These functions translate the probability of exceeding a set of damage states as a function of a secondary sufficient intensity measure, when records are selected and scaled for a particular level of primary seismic intensity parameter. It is demonstrated that this process allows a hazard-consistent and statistically meaningful representation of uncertainty and correlation in the estimation of intensity-dependent damage exceedance probabilities. Copyright (c) 2016 John Wiley & Sons, Ltd.