In this paper a Kriging method is reviewed and a way of its application in numerical weather prediction is proposed. The basic principles of the Kriging are shown; the main advantage is its accuracy, but at the same time a disadvantage is its large computational complexity. The construction stage of the variographical model is highlighted, as it is the most important stage and has a significant impact on the accuracy of interpolation. The algorithm for the construction of the variographical model is described. Special attention is paid to averaging an experimental variogram by introducing a special interval, called "lag". Precisely this issue, according to the authors has a significant impact on the effectiveness of the practical application of Kriging for interpolation of meteorological parameters. The advantages of averaging an experimental variogram by the administration of lag are presented, and the error that arises in this case is estimated. A theoretical study for the determination of the optimal lag was conducted. The lag proposed for determination is guided by the criteria of accuracy and economy of computer time. The two-criteria problem is solved, and the formula, which makes it possible to determine the optimal lag on these criteria, is received. An example shown here is the application of the obtained results for solving the applied task associated with meteorological parameters forecast by the COSMO model.