CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Space-time extreme simulations for natural disasters Authors:  Fatima Palacios-Rodriguez - Universidad Complutense de Madrid (Spain) [presenting]
Gwladys Toulemonde - Université de Montpellier (France)
Julie Carreau - Institut de Recherche pour le Developpement (France)
Thomas Opitz - BioSP, INRA (France)
Abstract: Extreme events of natural phenomena not only can entail material damages but can also have devastating consequences for human societies and ecosystems. In order to assess the risks associated with destructive natural disasters, impact models can be fed with simulations of extreme scenarios for studying the sensitivity to temporal and spatial variability, but available data may be insufficient due to the sparse occurrence of extreme episodes. We introduce a methodology to simulate realistic spatio-temporal extreme fields stochastically. To this end, we use a moderate number of observed extreme space-time episodes to generate an unlimited number of extreme scenarios of any magnitude. The theoretical justification of our framework comes from the extreme-value theory, where generalized Pareto limit processes arise for threshold exceedances. For implementation and illustration of our methodology, we obtain extreme event simulations from hourly gridded precipitation data in Mediterranean France.