Abstract

Geostatistical estimation and simulation techniques are useful for characterizing the spatial distribution of soil properties. These techniques usually rely on a variogram model that measures the spatial variability of the property under study. However, uneven sampling designs often make the traditional sample variogram lose accuracy and tend to mask the spatial structure. In this paper, a methodology is proposed to weight data pairs in order to compensate for redundancies between clustered data values and to minimize the fluctuation variance of the sample variogram. A comparison between the traditional and weighted sample variograms is held through the analysis of a dataset consisting of heavy metal concentrations measured in a contaminated site. © 2007 Elsevier B.V. All rights reserved.