The variogram is an important function in geostatistical applications such as kriging and various conditional simulation methods. In general, the actual variogram function is unknown for a particular application, and inferred from the experimental data. In the same way as first order statistics, the experimental variogram is also affected by preferentially clustered data. The irregularity of sampling has a well-known destructuring effect on the sample variograms, but it is rarely considered in their experimental calculation and modeling. This paper presents a study where a declustering method based on the data polygon of influence was applied in order to weight data pairs in experimental variogram calculations. The sample weighted variograms were appropriately modeled and used in sequential indicator simulations of coal thickness in a large Brazilian deposit. (C) 2012 Elsevier B.V. All rights reserved.