New dimension to account for sample error and volume support in resource estimation - SME Transactions 2011

Data used to build geostatistical models come from a wide variety of sources with potentially differing sample errors and volume supports. Typically all available sample measurements are pooled together for statistical analysis irrespective of the quality of the data. The sample data is used to assess spatial continuity, make local estimates and assess uncertainty. A new methodology is proposed to simplify accounting for samples of differing quality. The result is improved estimates with higher weights assigned to high-quality samples, while low-quality samples receive less influence in resource estimation. A new dimension, d, is introduced and calibrated to account for sample error and/or volume during estimation. The addition of an extra dimension can be easily incorporated in existing kriging or inverse distance workflows. The only additional parameters required are the sample error and volume for each datum. The methodology is demonstrated and the improvement in local accuracy is documented. Extending the dimensionality of the domain to four (three spatial dimensions and the additional d dimension) requires a valid 4D variogram model; therefore, the spherical variogram model is generalized to arbitrary n-dimensions.