Confidence limits for extreme wind speeds in mixed climates

A recent paper updated the methodology for the analysis of extreme wind speeds in mixed climates originally proposed in 1978 by Gomes and Vickery (GV) to take advantage of recent improvements in methodology and available data records. This involved the separation of independent sub-annual maxima by storm mechanism, determining the Weibull shape parameter k, preconditioning for fastest convergence using k, fitting each mechanism to the Fisher-Tippet Type I distribution and synthesising the composite extreme-value distribution. That work did not address the confidence of the composite distribution. In this paper, the "bootstrap" method was used to derive the confidence limits of the composite distribution for the revised methodology, as well as for the original GV methodology, illustrated by data from Onslow and Brisbane in Australia. It is shown that the composite distribution can be directly fitted to the observations, without prior separation by storm mechanism, when each mechanism is adequately represented in the observations. However, separation by causal mechanism prior to analysis should always be the preferred option when suitable data records are available. (C) 2003 Elsevier Ltd. All rights reserved.