Bivariate estimation of extreme wind speeds

Abstract

When wind speeds data exists but not with the length of record required to provide accurate parameter estimates, the error of the estimated return level can be very large and inefficient for design purposes. A way to reduce this error is by applying a joint estimation model, where information from nearby sites in the region may be combined with the record of inadequate length to increase information and to provide a regional at-site estimate of wind speed frequency. In order to achieve this goal, the Logistic model for bivariate extreme value distribution is proposed. The general procedure to estimate its parameters based on the maximum likelihood method is presented. Seven bivariate options were obtained by combining the marginal distributions: Gumbel, Reverse Weibull and General Extreme Value. A total of 45 sets, ranging from 9-year to 56-year, of largest annual wind speeds gathered of stations located in The Netherlands were fitted to univariate and bivariate distributions. At-site and regional at-site return levels were estimated and compared with those obtained in a previous study, which used the conditional Weibull distribution. A significant improvement occurs, measured through the use of a goodness-of-fit test, when parameters are estimated using the bivariate distribution instead of its univariate counterpart. Results suggest that it is very important to consider the bivariate joint estimation option when analyzing extreme wind speeds, especially for short samples.