XIMIS, a penultimate extreme value method suitable for all types of wind climate

Abstract This paper introduces XIMIS, an extended version of the existing IMIS method. The new method is penultimate in that it does not rely on asymptotic results, which in turn depend on the rate parameter rT→∞. For input it requires a sample of a mutually independent data set drawn from the original parent data, but having the same annual maxima as that parent. Thus it can use independent storm data or m-day maxima from temperate storms, or thunderstorm or cyclone maxima. In the paper, temperate storm data from Boscombe Down, UK, and cyclone and thunderstorm maxima, respectively, from Onslow, WA, and Brisbane, QD, in Australia are analysed. It is shown that derivation of standard 1:50 yr design values needs a mild extrapolation, which does not require any sort of probability model. A simple power law transformation is used to assist the extrapolation by linearising the plot. Derivation of 1:10,000 yr values does require a model and it is shown that if the relevant working variable is used, then there is no case for using any model except Type I. It is then argued that the transformation used for linearisation has good claims for validity for gross extrapolation, and the linearised plots are used to estimate 1:10,000 yr values. It is concluded that XIMIS is a useful design tool.