Intrinsic Random Functions and Parametric Covariance Models of Spatio-Temporal Random Processes on the Sphere

TL;DR

This paper introduces new covariance models for non-homogeneous and non-stationary spatio-temporal processes on the sphere, utilizing Intrinsic Random Functions to better capture real-world data dependencies.

Contribution

It develops novel parameterized covariance models for spherical spatio-temporal data that relax traditional assumptions of homogeneity and stationarity using IRF theory.

Findings

Models effectively capture non-stationary dependencies on the sphere.

Simulation and real data analysis validate model interpretability and accuracy.

Proposed methodology estimates parameters reliably.

Abstract

Identifying an appropriate covariance function is one of the primary interests in spatial and spatio-temporal statistics because it allows researchers to analyze the dependence structure of the random process. For this purpose, spatial homogeneity and temporal stationarity are widely used assumptions, and many parametric covariance models have been developed under these assumptions. However, these are strong and unrealistic conditions in many cases. In addition, on the sphere, although different statistical approaches from those on Euclidean space should be applied to build a proper covariance model considering its unique characteristics, relevant studies are rare. In this research, we introduce novel parameterized models of the covariance function for spatially non-homogeneous and temporally non-stationary random processes on the sphere. To alleviate the spatial homogeneity assumption and temporal stationarity, and to consider the spherical domain and time domain together, this research will apply the theories of Intrinsic Random Functions (IRF). We also provide a methodology to estimate the associated parameters for the model. Finally, through a simulation study and analysis of a real-world data set about global temperature anomaly, we demonstrate validity of the suggested covariance model with its advantage of interpretability.