Abstract:
|
Extremes are of great interest in climatology and, hence, extreme value statistical methods are frequently used in climate science. As an alternative to fitting generalized Pareto distributions to exceedances beyond some threshold, I propose specific parametric models for univariate distributions that can mimic the behavior of any one generalized Pareto distribution in the upper tail and the negative of any other generalized Pareto distribution in the lower tail. The resulting density has nice analyticity properties, is easy to compute in closed form and has parameters that are directly interpretable in terms of the behavior of the distribution in both tails. Applications to observational and computer model climate records will be considered.
|