Abstract: |
Geological domaining or modeling is a pre-requisite to resource characterization. Early in the exploration phase of a project, geologists integrate geological logs with drill hole assay data to generate a conceptual geological model. This facilitates geologic understanding and interpretation as more drill data becomes available. Later, when a resource model is initiated, these domains may be used to constrain grade estimation/simulation and often satisfy the need for stationary domains. In the mining sector, grade domains are often also considered to further control the distribution of grades during resource estimation. This is usually achieved by wireframe modeling on sections displaying grade assays or composites, indicator kriging, and/or via boundary modeling software such as Leapfrog. This paper proposes an alternative approach to conventional grade domaining, an approach that is based on MultiGaussian kriging. The method consists of estimation of grades using MultiGaussian kriging; however, instead of back transforming to obtain a grade estimate at each location, the probability to exceed certain grade thresholds are determined and grade domains are categorized accordingly. This permits uncertainty assessment of grade domains by post-processing for various grade thresholds. Other post-processing algorithms such as image cleaning to achieve reasonably continuous grade domains may also be required. The benefits of such an approach include: (1)sensitivity assessment on appropriate grade threshold(s) can be performed since this decision is made post-estimation; (2) inference of normal scores variogram is often easier than original units and/or indicators, and requires little additional effort; (3) extension to multiple grade domains, such as low, medium and high grade is not only easy but naturally progressive using this approach; (4) the method is computationally more efficient than manual wireframing, and; (5) the method is tractable and repeatable. An example to a mineral deposit is provided. Visual comparisons against indicator kriging for grade domaining and wireframed solids are shown. Grade distributions from the different domains are compared, between the different methods, to determine the statistical significance of the resulting domains. |
