Random-variable models are frequently applied to recorded sequences of
hydrological extremes. However, even if the recorded extremes behave
like random variables the underlying probability distribution still re
mains unknown. It follows that better extrapolations of extreme hydrol
ogical events will never be achieved by comparing permutations of the
latest estimation techniques and specified probability distributions.
Yet such estimation/distribution comparisons continue to proliferate t
hrough the hydrological literature in the vain hope that some 'best' e
xtrapolation method will emerge in time. The questionable value of the
whole comparison process calls into question the worth of objectivity
as a desirable attribute in techniques for analysing hydrological ext
remes.