Estimating 10000-year return values from short time series

The distribution of outliers is used as a tool for finding the extreme value distribution of meteorological parameters and to provide return values for large return periods from short records. Its potential is demonstrated for five cases. For extreme winds in the Northern Hemisphere (NH) the method shows that appropriately transformed annual maximum wind speeds can be described by a Gumbel distribution; for extreme waves it rejects the proposed adoption of an exponential distribution and points to a Gumbel distribution; for extreme daily European precipitation R it confirms the theoretically predicted value k = 2/3 in its Weibull distribution and it also justifies the application of the Gumbel distribution to R2/3 up to return periods of about 50 000 years; for seasonal precipitation in the Netherlands it highlights enhanced extreme precipitation in the coastal area in December-January-February (DJF) and failure of the k = 2/3 hypothesis outside June-July-August (JJA); for sea levels in the Southern North Sea it points to the Gumbel distribution and provides improved estimates for the 104-return value of the sea level at coastal stations, which is elaborated for the Dutch tidal station Scheveningen. Copyright  2009 Royal Meteorological Society