Abstract
In order to be able to perform a spatial interpolation according to the kriging method, one must know the generalized covariance of the underlying intrinsic random field. Recently, P. K. Kitanidis (1983) described how this generalized covariance can be estimated by making use of the maximum likelihood method. Here, it will be described how this estimation can be simplified if the model for the generalized covariance is linear with two parameters. The computations are based on a particular set of generalized increments, the so‐called principal increments. These increments can also serve as a starting point for a study of the mathematical properties of the maximum likelihood estimators.