The r largest order statistics model for extreme wind speed estimation

The paper presents the statistical estimation of extreme wind speed using annually r largest order statistics (r-LOS) extracted from the time series of wind data. The method is based on a joint generalized extreme value distribution of r-LOS derived from the theory of Poisson process. The parameter estimation is based on the method of maximum likelihood. The hourly wind speed data collected at 30 stations in Ontario, Canada, are analyzed in the paper. The results of r-LOS method are compared with those obtained from the method of independent storms (MIS) and specifications of the Canadian National Building Code (CNBC-1995). The CNBC estimates are apparently conservative upper bound due to large sampling error associated with annual maxima analysis. Using the r-LOS method, the paper shows that the wind pressure data can be suitably modelled by the Gumbel distribution.