On the usefulness of lattice approximations for fractional Gaussian fields

TL;DR

This paper explores how lattice-based fractional Laplacian differencing can effectively approximate continuum fractional Gaussian fields, enabling efficient computations and accurate inferences in spatial modeling.

Contribution

It demonstrates the validity of lattice approximations for fractional Gaussian fields within certain parameter ranges, highlighting their computational advantages and practical applications.

Findings

Lattice approximations agree well with continuum models for specific fractional parameters.

Lattice methods enable fast, matrix-free computations.

Applications include sea surface temperature analysis in the Indian Ocean.

Abstract

Fractional Gaussian fields provide a rich class of spatial models and have a long history of applications in multiple branches of science. However, estimation and inference for fractional Gaussian fields present significant challenges. This book chapter investigates the use of the fractional Laplacian differencing on regular lattices to approximate to continuum fractional Gaussian fields. Emphasis is given on model based geostatistics and likelihood based computations. For a certain range of the fractional parameter, we demonstrate that there is considerable agreement between the continuum models and their lattice approximations. For that range, the parameter estimates and inferences about the continuum fractional Gaussian fields can be derived from the lattice approximations. Interestingly, regular lattice approximations facilitate fast matrix-free computations and enable anisotropic representations. We illustrate the usefulness of lattice approximations via simulation studies and by analyzing sea surface temperature on the Indian Ocean.