Using the gh distribution to model extreme wind speeds

Near-surface wind data of the global ocean are used for weather and climate prediction. It is important to be able to model the upper tails or extremes of wind speed data. In this paper, we present a method for modeling the upper tails of maxima wind speed data using gh distributions. This method provides a competitive and flexible alternative to the usual asymptotic distribution, the generalized extreme value distribution. The procedure is robust and enables us to identify outliers in the data. We illustrate the method with wind speed data from a mid-Atlantic region. (C) 2003 Elsevier B.V. All rights reserved.