Ordinary Kriging

Abstract

The following introduction to the method of ordinary kriging will first present the kriging theory, give a small example, examine the effect of the variogram attributes on kriging, and finally provide a comparison to the other estimation methods for the Walker Lake data. The prediction methods discussed earlier in the course (polygonal declustering, triangu-lation, distance-based methods) were all based on weighted averages of some subset of the sampled points. The weights in these cases were based solely on the locations of the sampled points. They do not utilize any information about how similar the sampled values y(t i) = y i are expected to be to y(s 0) = y 0 , the predicted value. The kriging estimator incorporates the covariance structure among the Y i 's into the weights for predicting Y 0. In this way, the ordinary kriging estimator, like the other estimators studied, is also a weighted average: Y (s 0) = Y 0 = n i=1