A short memory condition for infinitely divisible random fields (Q7015124)

The authors consider the following definition of short/long range dependence:\N\NDefinition 1: A measurable stationary random field \(X=(X(t))_{t\in\mathbb{R}^d}\) is called short range dependent if \N\[ \int_{\mathbb{R}^d}\int_\mathbb{R}\int_\mathbb{R} \left| \text{Cov}\left( \mathbf{1}_{\{ X(t)>u\}}, \mathbf{1}_{\{ X(t)>v\}} \right) \right|\mu(du)\mu(dv)dt <\infty \] \Nfor all probability measures \(\mu\) on \(\mathbb{R}\). If the above integral is infinite for some \(\mu\), the field \(X\) is called long range dependent.\N\NThe authors provide a sufficient condition of short range dependence for a measurable stationary infinitely divisible moving average random field \(X\) in terms of its kernel function and characteristic exponent. They demonstrate how to check this condition in the special case of a Gaussian or an \(\alpha\)-stable symmetric integrator measure. Furthermore, the authors provide an example showing that the proposed condition is not necessary.