Estimates of Extreme Wind Distribution Tailsby J. A. Lechner, Natl Inst of Standards and, Technology, Gaithersburg, United States,
S. D. Leigh, Natl Inst of Standards and, Technology, Gaithersburg, United States,
E. Simiu, Natl Inst of Standards and, Technology, Gaithersburg, United States,
Abstract: The assumption that extreme winds have a Gumbel distribution appears to yield unrealistically high failure probability estimates for wind-sensitive structures. For this reason we study the question of whether extreme winds are better fitted by distributions with finite or shorter tails. We report progress in the following areas: (1) use of the Pickands and Conditional Mean Exceedance methods for estimating the tail length; (2) application of these methods to epochal (largest annual) wind speeds; (3) comparisons of results so obtained with results obtained from simulated sets of data; (4) application of these methods and the Dekkers-Einmahl-de Haan method to non-epochal extremes. Preliminary results suggest that for most U.S. weather stations not prone to hurricane winds, reverse Weibull distributions (which have finite tails) fit extreme wind data sets better than the Gumbel distribution (which has infinite tail). However, the significance of these results remains to be evaluated, and more precise methods for estimating tail length need to be used. Also needed are additional studies based on non-epochal and simulated data.
Subject Headings: Probability distribution | Failure analysis | Structural failures | Probability | Wind speed | Comparative studies | Hurricanes and typhoons
Services: Buy this book/Buy this article
Return to search