TIME SCALES AND STATISTICAL UNCERTAINTIES IN THE PREDICTION OF EXTREME ENVIRONMENTAL-CONDITIONS

Accuracy and confidence levels that can be placed in the estimates of extreme environmental parameters are the keys for determining structural safety and reliability. Stochastic process modelling is shown to be efficient for introducing different physical time scales (stationarity, short and long term periods, storm situations). Then basic principles of the extrapolation techniques and of the uncertainty estimates are deduced from the theory of order statistics. In particular, the lack of knowledge of the underlying distribution is an important source of uncertainty which is investigated with special care. This question is unfortunately included in the class of ill-posed problems in the sense defined by Hadamard. This explains the need to introduce specific statistics of test and prior knowledge in order to estimate practical estimators and their related statistical uncertainties. Finally, the statistical package ST.EX.OM developed to assist engineers in simulation studies and decision making is presented.