Stationary max-stable fields associated to negative definite functions

Abstract: Let $W_i,i\in{\mathbb{N}}$, be independent copies of a zero-mean Gaussian process ${W(t),t\in{\mathbb{R}}^d}$ with stationary increments and variance $\sigma^2(t)$. Independently of $W_i$, let $\sum_{i=1}^{\infty}\delta_{U_i}$ be a Poisson point process on the real line with intensity $e^{-y} dy$. We show that the law of the random family of functions ${V_i(\cdot),i\in{\mathbb{N}}}$, where $V_i(t)=U_i+W_i(t)-\sigma^2(t)/2$, is translation invariant. In particular, the process $\eta(t)=\bigvee_{i=1}^{\infty}V_i(t)$ is a stationary max-stable process with standard Gumbel margins. The process $\eta$ arises as a limit of a suitably normalized and rescaled pointwise maximum of $n$ i.i.d. stationary Gaussian processes as $n\to\infty$ if and only if $W$ is a (nonisotropic) fractional Brownian motion on ${\mathbb{R}}^d$. Under suitable conditions on $W$, the process $\eta$ has a mixed moving maxima representation.

• The paper establishes a novel connection between stationary max-stable fields and Gaussian processes through negative definite functions.
• It introduces a rigorous criterion for Brown–Resnick stationarity by subtracting an appropriate drift term to ensure translation invariance.
• The research offers practical representations by linking variogram-based and mixed moving maxima models for extreme value analysis.

Overview of Stationary Max-Stable Fields Associated with Negative Definite Functions

The paper presented by Zakhar Kabluchko, Martin Schlather, and Laurens de Haan focuses on the study of stationary max-stable fields associated with negative definite functions, providing both theoretical insights and practical representations for these stochastic processes. The research explores the connection between these processes and Gaussian processes exhibiting stationary increments, and it proposes novel representations for max-stable processes, underpinning their applicability in modeling extreme values.

The central object of study is the Brown–Resnick process, a max-stable process that exhibits translation invariance and has standard Gumbel margins. It arises from Gaussian processes with stationary increments, emphasizing the role of stationary random processes in understanding extremal events. The paper meticulously establishes that subtracting an appropriate drift term from such Gaussian processes results in a Brown–Resnick process, providing a rigorous criterion for determining when a process manifests this stationarity.

Key Theoretical Contributions

1. Brown–Resnick Stationarity: The authors define a stochastic process as Brown–Resnick stationary if the resulting process from a Poisson point process combined with the given process is statistically invariant with translations. This introduces a novel class of stationary max-stable processes that are pivotal in the theory of extremes.
2. Variogram-Based Representation: The process is characterized by a variogram which must be negative definite. The variogram serves as a tool to construct Gaussian processes whose increments are stationary, thus enabling the derivation of the underlying max-stable process.
3. Mixed Moving Maxima Representation: Under certain conditions on the asymptotic behavior of the associated Gaussian processes, the Brown–Resnick process can be represented as a mixed moving maxima process, highlighting its potential application in spatial models for extreme value theory.
4. Domains of Attraction: The paper explores the domains of attraction for max-stable processes, elucidating the conditions under which the maxima of independent Gaussian processes converge to a Brown–Resnick process.

Empirical and Numerical Results

The research provides theoretical assurances for the convergence of these processes, supported by insightful derivations rooted in the properties of Gaussian processes and their tail behaviors. Notably, the authors extend results to support the applicability of their findings in spatial processes, underscoring the robustness of the max-stable frameworks they propose.

Implications and Future Directions

The results of this paper have broad implications for applied probability and extreme value theory, particularly in areas demanding accurate models for extreme events such as meteorology and environmental sciences. By establishing a bridge between Gaussian processes and max-stable processes, the paper paves the way for future research into more complex and multidimensional stochastic models.

Additionally, the representation of these processes via negative definite functions encourages further exploration of alternative max-stable models, potentially incorporating anisotropy or leveraging other functional forms of variograms. This could lead to more nuanced and flexible applications in geostatistics and risk management.

Overall, the paper provides a comprehensive theoretical framework for stationary max-stable fields, with robust mathematical foundations and a clear trajectory for future theoretical and applied advancements in stochastic modeling of extremes.