• Corpus ID: 240354445

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

  title={Modelling and simulating spatial extremes by combining extreme value theory with generative adversarial networks},
  author={Younes Boulaguiem and Jakob Zscheischler and Edoardo Vignotto and Karin Wiel and Sebastian Engelke},
Modelling 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… 
1 Citations

Figures from this paper

Simulating flood event sets using extremal principal components
Hazard event sets, which correspond to a collection of synthetic flood events, are an important tool for practitioners to analyse and manage future flood risks. In this paper, we address the issue of…


Gradient boosting for extreme quantile regression
This work proposes a gradient boosting procedure to estimate a conditional generalized Pareto distribution by minimizing its deviance, and shows that it outperforms classical methods from quantile regression and extreme value theory, especially for high-dimensional predictor spaces and complex parameter response surfaces.
A climate database with varying drought‐heat signatures for climate impact modelling
Extreme climate events, such as droughts and heatwaves, can have large impacts on the environment. Disentangling their individual and combined effects is a difficult task, due to the challenges…
Added Value of Large Ensemble Simulations for Assessing Extreme River Discharge in a 2Β Β°C Warmer World
The assessment of return periods of extreme hydrological events often relies on statistical analysis using generalized extreme value (GEV) distributions. Here we compare the traditional GEV approach…
Extremal behaviour of aggregated data with an application to downscaling
&NA; The distribution of spatially aggregated data from a stochastic process X may exhibit tail behaviour different from that of its marginal distributions. For a large class of aggregating…
Graphical models for extremes
A general theory of conditional independence for multivariate Pareto distributions is introduced that enables the definition of graphical models and sparsity for extremes and it is shown that, similarly to the Gaussian case, the sparsity pattern of a general extremal graphical model can be read off from suitable inverse covariance matrices.
Identifying meteorological drivers of extreme impacts: an application to simulated crop yields
<p>Compound weather events may lead to extreme impacts that can affect many aspects of society including agriculture. Identifying the underlying mechanisms that cause extreme impacts, such as crop…
Pareto GAN: Extending the Representational Power of GANs to Heavy-Tailed Distributions
The Pareto GAN leverages extreme value theory and the functional properties of neural networks to learn a distribution that matches the asymptotic behavior of the marginal distributions of the features.
Dependence properties of spatial rainfall extremes and areal reduction factors
An Introduction to Statistical Modeling of Extreme Values
  • E. Smith
  • Computer Science
  • 2002
Stuart Coles’s book on the modeling of extreme values provides an introductory text on the topic, a modeling-oriented text with an emphasis on different types of data and analytical approaches, meant for individuals with moderate statistical background.
Spatial modeling of extreme snow depth
The spatial modeling of extreme snow is important for adequate risk management in Alpine and high altitude countries. A natural approach to such modeling is through the theory of max-stable…