A Study of Variance Estimators for Material Sampling Using Computerized Models of Contaminant Heterogeneity in Soil Stockpiles.

During the sampling of contaminated soil, sampling errors are unavoidable because of the spatial heterogeneity of the contaminant distribution. The variance is a convenient indicator for the potential magnitude of these errors. Four variance estimators are constructed for use in material sampling, all of which take account of the heterogeneity and the sampling design. Based on large scale three-dimensional computerized models of contaminant heterogeneity in soil stockpiles, these variance estimators are compared using a Monte Carlo simulation of different sampling designs. The Mean Squared Error (MSE) of each variance estimator is used to assess (and compare) the performance of each variance estimator: the lower the mean square error, the better its performance.