Premium
Markov chain models for extreme wind speeds
Author(s) -
Fawcett Lee,
Walshaw David
Publication year - 2006
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/env.794
Subject(s) - extreme value theory , generalized extreme value distribution , maxima , tail dependence , econometrics , computer science , inference , bayesian probability , set (abstract data type) , wind speed , markov chain , extreme weather , bayesian inference , statistics , mathematics , meteorology , geography , climate change , multivariate statistics , machine learning , geology , artificial intelligence , art , programming language , art history , oceanography , performance art
Understanding and quantifying the behaviour of extreme wind speeds has important applications for design in civil engineering. As in the extremal analysis of any environmental process, estimates are often required of the probability of events that are rarer than those already recorded. Consequently, research has focused on the development of techniques that make optimal use of the available data. One such approach lies in threshold methods , which, unlike the more traditional annual maxima approach to the modelling of extremes, takes into consideration all extreme events, extreme in the sense that they exceed some high threshold. However, the implications of using all extremes in an analysis include problems of temporal dependence and non‐stationarity. Several pragmatic ways of circumventing the problem of temporal dependence have been developed, though these often include the deletion of many extreme observations, for example, filter out a set of independent extremes. This paper looks at another approach to inference—one which explicitly models the temporal dependence of the process and so can use information on all extremes—and investigates the appropriateness of assumptions of short‐term temporal dependence for wind speeds. We also examine the success of such methods at estimating some extreme events commonly studied for wind‐speed data. Throughout this paper extreme wind speeds are analysed within a Bayesian framework, which can be argued to be particularly advantageous for extreme value analyses. For example, the objective of an extreme value analysis is usually an estimate of the probability of future events reaching extreme levels—something which is handled quite naturally in a Bayesian analysis through predictive distributions . Copyright © 2006 John Wiley & Sons, Ltd.