No label defined (Q4677141)

Let \(\xi=\{\xi(t),t\in{\mathbf T}=[0,T]^{d}\}\) be a centered Gaussian random field which can be represented in the form \(\xi(t)=\sum_{n=1}^{\infty} \xi_{n}f_{n}(t)\), where \(f_{n}(t),n\geq1\), are continuous functions, \(\sum_{n=1}^{\infty}f^2_{n}(t)<\infty\), where \(\xi_{n}\), \(n=1,2,\ldots\), are independent Gaussian random variables, such that \({\mathbf E}\xi_{n}=0\), \({\mathbf E}\xi_{n}^2=1\). A field \(\widetilde{\xi}_{N}(t)=\sum_{n=1}^{N} \xi_{n}f_{n}(t)\), \(t \in [0,T]^d\), is called a model of the field \(\xi(t)\). The model \(\widetilde{\xi}_{N}(t)\) approximates \(\xi(t)\) with given reliability \(1-\nu\), \(\nu\in (0,1)\), and accuracy \(\delta>0\) in the Banach space \(C^1_{[0,T]^d}\) if the following inequality holds: \[ {\mathbf P}\left\{\sup_{t\in {\mathbf T}}| X_{N+1}(t)-{\mathbf E} X_{N+1}(t)|>\delta\right\}\leq \nu, \] where \(X_{N+1}(t)=X(t)-\widetilde{X}_N(t)\), \(X(t)=\xi^2(t)+ \sum_{i=1}^{d} ({\partial\xi(t)\over \partial t_{i}})^2,\) \(\widetilde{X}_N(t)= \widetilde{\xi}_{N}^2(t)+ \sum_{i=1}^{d}({\partial\widetilde{\xi}_{N}(t)\over \partial t_{i}})^2.\) The approximating models of the Gaussian fields with given reliability and accuracy in Banach space \(C_{[0,T]^{d}}^1\) are constructed. For Gaussian random fields, isotropic in wide sense, on the unit sphere the theorem for simulation of the fields on the sphere is proved.