Technical note: Calibration of EVA methods for peak-over-threshold wind data using Bootstrapping

• Bootstrapping methods are used to calibrate extreme value analysis methods for peak-over-threshold wind (POT) speed observations.
• The Generalised Pareto methodology is demonstrated to be unsuitable for wind speed data.
• When the observations comprise Storm maxima or n-day maxima, the population of the missing values below the threshold is known.
• The most accurate analysis is obtained by fitting the POT observations to a left-censored distribution.

This paper uses Bootstrapping methods to calibrate the accuracy of estimating the design once in 50 year wind speed, V50, from peak-over-threshold wind speed observations. It confirms previous studies showing that the generalised Pareto distribution is not a satisfactory model for extreme wind speeds. It demonstrates that when n-day maxima are used the value of the total population of independent events is known. This allows the peak-over-threshold observations to be represented by a left-censored probability distribution, eliminating mean bias and greatly reducing the standard error in the V50 estimates.