Simulation of Gaussian random field in a ball

Abstract

The presented paper is devoted to statistical modeling of Gaussian scalar real random fields inside a three-dimensional sphere (ball). We propose a statistical model describing the spatial heterogeneity in a unit ball and a numerical procedure for generating an ensemble of corresponding random realizations. The accuracy of the presented approach is corroborated by the numerical comparison of the estimated and analytical covariance functions. Our approach is flexible with respect to the assumed radial and angular covariance function. We illustrate the effect of the covariance model parameters based on numerical examples of random field realizations. The presented statistical simulation technique can be applied, for example, to the inference of the 3D spatial heterogeneity in the Earth and other planets.