The spatial interpolation method of Kriging, originally developed for
estimation of ore reserves in mining and prediction in meteorology, ha
s become an important geostatical tool for assessment of contamination
in Superfund site characterization. Kriging uses an estimated spatial
covariance matrix, or alternatively the variogram, to find the best l
inear unbiased estimate (BLUE) of contamination at an unsampled locati
on. The modeling of the variogram (also called semi-variogram) is ther
efore a necessary step that must be performed before using Kriging for
spatial interpolation. In this paper, we propose and investigate a si
mple non-parametric procedure based on ranks for variogram modeling. S
ome simulated examples are used to illustrate the proposed procedure.
The proposed procedure is compared with the usual method of Ordinary K
riging for the Billings Smelter Superfund site data.