CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Simulation of spatial hazard events using extremal principal components Authors:  Christian Rohrbeck - University of Bath (United Kingdom) [presenting]
Daniel Cooley - Colorado State University (United States)
Abstract: Hazard event sets, which correspond to a collection of synthetic extreme events, are an important tool for practitioners to analyse and manage future risks. We summarise our method for generating hazard event sets over a set of spatial locations. We start by deriving a set of extremal principal components, with the first components describing the large-scale structure in the extremal dependence structure. The simulation framework is then based on an approximation of the distribution of the extremal principal components. The approach is dimension-reducing in that it distinctly handles the first and remaining extremal principal components. We conclude by discussing some of our recent extensions and a comparison with other existing methods. All aspects are illustrated by applying our method to study extremal behaviour in an environmental setting.