Prediction of Design Wind Speeds by Extreme Value EstimationProcedures

点击翻译按钮获取中文摘要

9411642 Corotis Support is provided to continue research on extreme value estimation procedures for the prediction of design wind speeds. During the first phase several available techniques were tested and compared with some methods performing better than others, e.g., for data from Gumbel and reverse Weibull distributions the threshold approach performed better than the epochal approach but the reverse was true for data from a normal distribution. During the second phase of the research statistical analyses of data taken from maximum daily fastest mile speeds and Monte Carlo simulations based on the probabilistic models derived from these analyses will be used. The objectives of the work are: 1) to rank the performances of the Cumulative Mean Excedance and the de Haan-Dekkers-Einmahl estimation methods; 2) to account for the seasonality effects; 3) to verify whether more extreme speed data sets are in fact better fitted by reverse Weibull than by Gumbel distributions; and 4) to estimate confidence levels for "peaks over threshold" estimates of extreme winds based on short records. Long term wind speed data from United States weather stations will be used to compare epochal and threshold methods and to refine procedures for determining design wind speeds. The results from this work will improve the survivability of structures built in extreme wind areas and thus the safety of local population. ***