One of the methods used for the interpolation of spatially correlated data is known as kriging, which is a regression technique. Consider a model of the form \(Z(x)\), a random function defined in \(p\)-dimensional space satisfying a weak stationarity condition as follows: \[ \begin{aligned} & E[Z(x + h) - Z(x)] = 0 \text{ for all } x, h;\tag\text{i}\\ &\gamma(h) = 0.5 Var\{Z(x + h) - Z(x)\}\text{ exists and depends only on }h.\tag\text{ii}\end{aligned} \] The objective is to estimate \(Z(x_ 0)\) or \(Z_ v\) where \(Z_ v\) is a spatial average over a \(p\)-dimensional volume given by data \(Z(x_ 1),\dots,Z(x_ n)\). The estimator is a linear function of the data \(\sum \beta_ i(x_ 0) Z(x_ i)\), the coefficients being determined by requiring that the estimator be unbiased and that the estimation error has minimum variance resulting in a linear system of equations: \[ \begin{aligned} & \sum \beta_ i(x_ 0) \gamma(x_ i - x_ j) + \mu(x_ 0) = \gamma(x_ 0 - x_ j);\quad j = 1,\dots,n,\\ & \sum \beta_ i (x_ 0) = 1.\end{aligned} \] The minimal variance is computed from the variogram model, which, in turn, must be estimated and modeled from the data. In the case of the estimation of a spatial average the terms on the right- hand side are replaced by averages.