Extreme value statistics for North Atlantic cyclones

Abstract

Extremes of the cyclone intensity measures geopotential height (z 1000 ), mean horizontal gradient (∇z), cyclone depth ( D ), and relative vorticity (ζ 850 ), are analysed in re-analysis data (ERA40) and model simulations (ECHAM5/MPIOM) in the North Atlantic region for extended winter seasons. Generalized Pareto distributions (GPD) are estimated for model validation and climate change assessment. Covariates, linear trend and North Atlantic Oscillation (NAO) are included to analyse the dependancies of the extremes. In ERA40 no significant linear trend can be detected, while evidence for a NAO impact on z 1000 , ∇z and ζ 850 extremes is found. Model validation yields good agreement with consistent scale and shape, but a shift to lower values is notable. Like in ERA40 no trend is found in the simulation. The evidence for an NAO impact on cyclone extremes is less corroborated in the simulation, pointing to sample size effects. In the warmer climate scenario (A1BS) extreme value statistics shows an intensification for all variables. Significant differences in GPD are obtained through testing for lower (higher) parameters. In contrast, considering all cyclones an increase is only present for z 1000 , while a decrease is found for ∇z and ζ 850 and no change for D .