Max-stable processes for modelling extremes observed in space and time

TL;DR

This paper extends max-stable processes to model space-time extremes, providing new representations, explicit formulas, and analysis of extremal dependence for improved statistical inference in spatial-temporal data.

Contribution

It introduces novel space-time max-stable process constructions, extends existing models to the spatio-temporal domain, and analyzes extremal dependence using covariance functions.

Findings

Extended max-stable models to space-time domain.

Derived explicit bivariate distribution functions.

Linked covariance functions to tail dependence coefficients.

Abstract

Max-stable processes have proved to be useful for the statistical modelling of spatial extremes. Several representations of max-stable random fields have been proposed in the literature. For statistical inference it is often assumed that there is no temporal dependence, i.e., the observations at spatial locations are independent in time. We use two representations of stationary max-stable spatial random fields and extend the concepts to the space-time domain. In a first approach, we extend the idea of constructing max-stable random fields as limits of normalized and rescaled pointwise maxima of independent Gaussian random fields, which was introduced by Kabluchko, Schlather and de Haan [2009], who construct max-stable random fields associated to a class of variograms. We use a similar approach based on a well-known result by H\"usler and Reiss and apply specific spatio-temporal covariance models for the underlying Gaussian random field, which satisfy weak regularity assumptions. Furthermore, we extend Smith's storm profile model to a space-time setting and provide explicit expressions for the bivariate distribution functions. The tail dependence coefficient is an important measure of extremal dependence. We show how the spatio-temporal covariance function underlying the Gaussian random field can be interpreted in terms of the tail dependence coefficient. Within this context, we examine different concepts for constructing spatio-temporal covariance models and analyse several specific examples, including Gneiting's class of nonseparable stationary covariance functions.