Preferential sampling or clustering is frequently found in mining and earth sciences applications. Reliable statistics for a population are obtained when representative samples are available. Such representativeness can be achieved by a proper sample design covering evenly the area. This paper investigates two declustering methods to obtain unbiased statistics where clustered samples are available, namely the polygonal and the cell-declustering method. The impact of clustering is analysed for two different datasets. Polygonal method proved to be simpler as it provides an unique solution easily to be understood by the user. In relation to the cell-declustering method, a methodology to calculate the statistical entropy was implemented to help in determining the most appropriate cell size. The two methods lead to similar declustered statistics. However the final statistics showed a large difference when compared to the statistics obtained for the clustered dataset.
preferential sampling; declustering methods; statistical entropy
