Almost sure convergence for the maximum of nonstationary random fields (Q2412514)

Let \(\mathbf{X}=\{ X_{\mathbf{n}}\}_{\mathbf{n} \in \mathbb{Z}^{2}_{+}} \), where \(\mathbb{Z}_{+}\) is the set all positive integers. Let \(\{ u_{\mathbf{n},\mathbf{i}}: \mathbf{i}\leq \mathbf{n}\}_{\mathbf{n}\geq \mathbf{1}}\) be a sequence of real numbers, where for \(\mathbf{i}= (i_{1},i_{2})\) and \(\mathbf{j}= (j_{1},j_{2} )\), \(\mathbf{i} \leq \mathbf{j}\) means \(i_{k}\leq j_{k}\), \(k=1,2\). Under the assumption, that \( \mathbf{X} \) is a nonstationary random field satisfying a coordinatewise-mixing-type condition \( D^{*}(u_{\mathbf{n},\mathbf{i}} )\), a local dependence condition \(D'(u_{\mathbf{n},\mathbf{i}})\) and some asumptions on index sets, the authors prove that, if \[ \sum_{\mathbf{i}\leq \mathbf{n}} P(X_{\mathbf{i}}>u_{\mathbf{n},\mathbf{i}}) \rightarrow \tau,\; \mathbf{n} \rightarrow \mathbf{\infty}, \] for some \(0\leq \tau <\infty\), where \(\mathbf{n}=(n_{1},n_{2})\rightarrow \mathbf{\infty}\) means \(n_{k} \rightarrow \infty\), \(k=1,2\), and \(\{ n_{1}n_{2} \max \{ P(X_{\mathbf{i}}>u_{\mathbf{n},\mathbf{i}}):\mathbf{i}\leq \mathbf{n} \} \}_{\mathbf{n}\geq \mathbf{1}}\) is bounded, then \[ \lim_{\mathbf{n} \rightarrow \mathbf{\infty}} \frac{1}{\log n_{1} \log n_{2}} \sum_{\mathbf{k} \in \mathbf{R}_{\mathbf{n}} } \frac{1}{k_{1}k_{2}} \mathbb{I}\{ \bigcap_{\mathbf{i}\leq \mathbf{k}} \{ X_{\mathbf{i}} \leq u_{\mathbf{n},\mathbf{i}} \} \} =\exp(-\tau) \; a.s. , \] where \(\mathbb{I}(A)\) is the indicator od the random event \(A\), and \(\mathbf{R}_{\mathbf{n}} = \{ 1,\ldots,n_{1} \} \times \{ 1,\ldots,n_{2} \}\). The obtained result is applied to Gaussian random fields.