A method of spatial prediction, named by Matheron in honour of Krige. The values of some quantity, s, are known for n points in space. Estimates for intervening locations are required—for example, we might wish to derive contour lines from a series of spot heights. This is also called point kriging. If, instead, the aim is to estimate the total amount (of e.g. oil, or gold) in a region, rather than at a point, then that is called block kriging. The predictions are to be calculated as the weighted sum of the observed values. The question is what weights (see weighted average) to use so as to minimize the average squared difference between the predicted value and the true value.

If, at each spatial location, data are collected on several variables, then the interpolation of values for the variable of prime interest may be improved by noting the variation in the other variables—this is called cokriging.

Other variants include disjunctive kriging, which uses linear combinations of functions of the data, indicator kriging, which uses binary indicators (e.g. presence/absence) in place of the observed data, and universal kriging, in which the surface being estimated is an unknown linear combination of known spatial surfaces.