Local optimization of observed points for Kriging with no need to make extra observations

For a second-order stationary random field, this paper presents a new formula, which strictly ensures the unbiasedness and optimality of the common Kriging, to calculate the unknown parameters in variogram or covariance function. Based upon this formula, an algorithm is developed to locate the observed points bearing no observations with the purpose of minimizing the maximal estimator variances on a networked domain.