Declustering and Debiasing

Strategic project decisions are based on the distributions global variables, for example, total mineable resource, or recoverable oil volume. These global variables distributions are very sensitive to rock type proportions and the histograms of continuous variables. Representivity of the input one point statistics is important in all spatial models. The process of assembling representative one point statistics is complicated by sample clustering and spatial bias in the data locations. Explanation is provided on the source of nonrepresentative sampling and the need for declustering and debiasing. This work addresses some key implementation details on declustering. Standard declustering is not always able to correct for sampling spatial bias. Two methods for correcting bias in the one point statistics: “trend modeling for debiasing” and “debiasing by qualitative data” are reviewed and demonstrated with a poly metallic data set.