Kriging and Prediction of Nonlinear Functionals

Abstract

The prediction of a nonlinear functional of a random field is studied. The covariance-matching constrained kriging is considered. It is proved that the optimization problem induced by it always has a solution. The proof is constructive and it provides an algorithm to find the optimal solution. Using simulation, this algorithm is compared with the method given in Aldworth and Cressie (2003).