If you have access to OneMine as part of a member benefit, log in through your member association website for a seamless user experience.
|Common practice in mineral resource estimation consists of partitioning the orebody into several domains defined by grade intervals, prior to the geostatistical modelling and estimation/simulation at unsampled locations. This paper shows the pitfalls of grade domaining through a case study in which we compare the performance of several estimation schemes and demonstrate that the use of domains defined by grade cut-offs implies a deterioration of the resource estimates, mainly in what refers to precision and conditional bias. Then, several conceptual limitations of the grade domaining approach are stressed, in particular the fact that it does not account for the spatial dependency between adjacent domains and for the uncertainty in the domain boundaries. Also, this approach is shown to be sensitive to the cut-offs that define the domains, to provoke artifacts in the kriging maps, histograms and scatter grams between true and estimated grades, and to lower the kriging variance, a feature that may impact the mineral resource classification. An alternative approach is finally proposed to overcome these limitations, based on a stochastic modelling of the grade domains and a co-kriging of the grades using information from all the domains. Keywords: geostatistics, kriging, conditional bias, geological control, grade zoning.|