Collaborative kriging is a modification of kriging for multivariate problems, where the random field to be modeled is called the main variable and the other random fields involved in the modeling are called covariates. A collaborative kriging can have any number of covariates, but the main variable and the covariates must be correlated and both must be inherently smooth random fields that satisfy the isotropy assumption. The cross-covariogram can be used to estimate the cross-covariance function among random fields, but the latter is required to be symmetric.