A Study of Optimal Extreme Wind Estimation Procedures

We describe work aimed at improving procedures for the estimation of non-tornadc extreme windspeeds, regardlessof their direction, in regions not subjectedto hurricanes. Using the Genemlized Pareto Distribution (GPD) approach and the ConditionalMean Exceedance(CME) estimation method, we analyze 115 17-yearto 52-year sets of largest annual speedsand sets drawn from 48 15-yearto 26year records of maximum daily wind speeds. Based on this analysis we attempt an assessmentof the widely held belief that the Gumbel distribution with sitedependent location and scale pammeters is a universal model of extreme wind speeds. Some of our results suggest that the reverse Weibull distribution is a more appropriate model. This would result in more reasonable estimates of windinduced failure probabilities and wind load factors than the corresponding estimates based on the Gumbeldistribution. However, our assessment is so far only tentative owing to uncertainties inherent in our results. Future work based on lower thresholds (larger &ta samples) and alternative estimation methodsis planned.