Title: Trends in the extremes of storm environments
Authors: Jonathan Koh - EPFL (Switzerland) [presenting]
Erwan Koch - EPFL (Chair of Statistics STAT) (Switzerland)
Anthony Davison - EPFL (Switzerland)
Abstract: Concurrently high values of Convective Available Potential Energy (CAPE) and Storm Relative Helicity (SRH) are conducive to hazardous convective weather (tornadoes, hail and damaging wind) associated with severe thunderstorms. Hence, it is highly relevant to have probabilistic models for both variables' extremes that use relevant covariate information to account appropriately for their spatial and temporal dependence structures. We consider a large area of the Continental United States over the period 1979-2015 and use statistical extreme value models and appropriate multiple testing procedures. Various features of the data motivate a two-step model approach. In the first step, we perform pointwise modelling of each grid cell using a model for univariate extremes. Here, we show that there is a time trend in the location parameter of the generalized extreme-value distribution for SRH maxima in April, May and June and for CAPE maxima in April. We also show that there is a strong signal with the El Nino-Southern Oscillation for the SRH maxima in February. In the second stage, we perform spatial dependence modelling of each grid cell with a model for spatial extremes. We incorporate covariate effects and anisotropy into the dependence modelling and show that extreme SRH events are more spatially localised in the spring/summer. These results are consistent with what is expected under global warming; as such, they are of significant interest to the climate research community.