Estimation of wind-induced extreme responses with various mean recurrence intervals considering model uncertainty due to limited data

The wind-induced extreme responses with various mean recurrence intervals (MRIs) can be estimated by accounting for the aleatory uncertainties in wind speed and extreme response coefficient in terms of their probability distributions. These distribution models when determined from limited samples of data also have uncertainties, referred to as epistemic uncertainties. This study addresses the epistemic uncertainties in probability distributions of wind speed and extreme response coefficient, and their influence on the estimated extreme responses with various MRIs. Closed-form expressions for quantifying the epistemic uncertainties in terms of coefficients of variation of the distribution model parameters are presented as functions of sample size. The influence of epistemic uncertainties on the predicted extreme responses with various MRIs is also estimated using closed-form expression in terms of confidence intervals. Based on the proposed formulations, a parametric study is carried out to shed light on the importance of consideration of various epistemic uncertainties. The results show that the epistemic uncertainty of wind speed generally predominates the predicted extreme response, while the epistemic uncertainty in the extreme response coefficient can also have considerable effect when the aleatory uncertainty in wind speed is low and that in extreme response coefficient is high. The proposed method is then applied to a long-span roof structure using wind loads determined from a wind tunnel test.