Modeling and simulating spatial extremes by combining extreme value theory with generative adversarial networks

Article Article scientifique Article Modeling and simulating spatial extremes by combining extreme value theory with generative adversarial networks eng 2022-04-13 2022 author Boulaguiem Younes 0000-0003-0795-0714 854267 Département de santé et médecine communautaire (MECOM) author Zscheischler Jakob 0000-0001-6045-1629 author Vignotto Edoardo 884830 author van der Wiel Karin author Engelke Sebastian 881707 Recherche et enseignement GSEM (RESEM) 18 10.1017/eds.2022.4 2634-4602 https://www.cambridge.org/core/product/identifier/S2634460222000048/type/journal_article Modeling dependencies between climate extremes is important for climate risk assessment, for instance when allocating emergency management funds. In statistics, multivariate extreme value theory is often used to model spatial extremes. However, most commonly used approaches require strong assumptions and are either too simplistic or over-parameterized. From a machine learning perspective, generative adversarial networks (GANs) are a powerful tool to model dependencies in high-dimensional spaces. Yet in the standard setting, GANs do not well represent dependencies in the extremes. Here we combine GANs with extreme value theory (evtGAN) to model spatial dependencies in summer maxima of temperature and winter maxima in precipitation over a large part of western Europe. We use data from a stationary 2000-year climate model simulation to validate the approach and explore its sensitivity to small sample sizes. Our results show that evtGAN outperforms classical GANs and standard statistical approaches to model spatial extremes. Already with about 50 years of data, which corresponds to commonly available climate records, we obtain reasonably good performance. In general, dependencies between temperature extremes are better captured than dependencies between precipitation extremes due to the high spatial coherence in temperature fields. Our approach can be applied to other climate variables and can be used to emulate climate models when running very long simulations to determine dependencies in the extremes is deemed infeasible. Climate model simulations Extreme value theory Generative adversarial networks Spatial extremes Dewey 320 Dewey 310 Swiss National Science Foundation New metrics for constraining multiple drivers of hazard and compound hazards 179876 CH Swiss National Science Foundation Graph structures, sparsity and high-dimensional inference for extremes 186858 CH Swiss National Science Foundation Machine learning for detecting compound climate drivers of extreme impacts 189908 CH 117 Research Center for Statistics S5 24 Global Studies Institute 0DJ false Environmental data science 1 7fd21e5a-78db-4186-9f4c-edb6f2bdfcf9 Article (Published version) Public Article_scientifique_from_crossref.pdf 3143133 application/pdf CC BY-4.0 84a97ce52f7a51e77427e60a1fcedbd4f1cf232ec81b9f187a60acdde8b1e65e openAccess Faculté d'économie et de management / Research Center for Statistics Centres et instituts / Global Studies Institute

Title	Published in	Access level	OA Policy	Year	Views	Downloads

Modeling and simulating spatial extremes by combining extreme value theory with generative adversarial networks	Environmental data science			2022	5	5

Archive ouverte UNIGE

Vignotto, Edoardo