Fast Spatial Prediction from Inhomogeneously Sampled Data Based on Generalized Random Fields with Gibbs Energy Functionals

Abstract

An explicit optimal linear spatial predictor is derived. The spatial correlations are imposed by means of Gibbs energy functionals with explicit coupling coefficients instead of covariance matrices. The model inference process is based on physically identifiable constraints corresponding to distinct terms of the energy functional. The proposed predictor is compared with the geostatistical linear optimal filter (kriging) using simulated data. The agreement between the two methods is excellent. The proposed framework allows a unified approach to the problems of parameter inference, spatial prediction and simulation of spatial random fields.