Copula Gaussian multiscale graphical models with application to geophysical modeling

Gaussian Multiscale graphical models are powerful tools to describe high-dimensional spatial data; they capture longrange statistical dependencies among distant sites by introducing coarser scales. However, such models are only applicable to Gaussian data. In this paper, a new class of copula Gaussian multiscale graphical models is proposed which possesses rich modeling capabilities and computational efficiency while eliminating the Gaussian assumption. Numerical results are presented for synthetic data as well as data from a few applications in geophysics - models of sea surface temperature and Asian rainfall patterns.