Extreme wind speed predictions are often based on statistical analysis of s
ite measurements of annual maxima, using one of the Generalized Extreme Val
ue (GEV) distributions. An alternative method applies one of the Generalize
d Pareto Distributions (GPD) to all measurements over a chosen threshold (p
eaks over threshold). This method increases the number of measurements incl
uded in the analysis, and correspondingly reduces the statistical uncertain
ty of quantile variances, but raises other important questions about, for e
xample, event independence and the choice of threshold. Here an empirical s
tudy of the influence of event independence and threshold choice is carried
out by performing a GPD analysis of gust speed maxima from five island sit
es in the north of Scotland. The expected invariance of the GPD shape param
eter with choice of threshold is utilized to look for changes of characteri
stic wind speed behavior with threshold. The impact of decadal variability
in wind on GEV and GPD extreme wind speed predictions is also examined, and
these predictions are compared with those from the simpler Gumbel and expo
nential forms.