Multivariate kriging
Résumé
Kriging is a special type of optimal linear prediction applied to random functions, with the particular requirement that the covariance structure of the random function should be known. Kriging may be extended to several variables by specifying a corresponding system of direct and cross covariance functions to be used in cokriging. Alternatively, secondary variables known at all sample and prediction locations can be included into (co-)kriging as external drift.