Abstract:
The main causes of geological uncertainty in development of mines consist of inaccuracy from geological domains and uncertainty in metal grades (Emery, 2007). Because geological domains typically influence how the grade is distributed, these two types of uncertainty impossible to model separately. Creating geological domains and assessing the grade of metal inside them are often first steps in the resource estimation process in the mining sector. Legacy drill hole data samples taken decades ago contain a vast quantity of geological information, but there is no proof that accurate QA/QC was done on those specimens (Abulkhair and Madani 2022). In this case, it is important to build an appropriate block model that can be used as a representative training image for conducting a multiple-point geostatistics. Because complex, curved, continuous structures such as vein-type deposits unable to be defined with traditional two-point statistics, variogram-based stochastic modeling are frequently unable to represent the spatial distribution of gold veins.
This thesis proposes a unique method for stochastic simulation of a Mineral Resource in a structurally complicated gold deposit. Multiple-point geostatistics tries to overcome the variogram's limitations. While creating deterministic geological models of the subsurface, information from boreholes and expert subject knowledge are frequently used as cognitive or explicit geological modeling techniques. Such cognitive geological models, however, are unable to convey the ambiguity of layer boundaries. Also, there are typically only a few boreholes available throughout the research area, although it is still unclear where the facies boundaries exactly are. To quantify the uncertainty in these boundaries, however, probabilistic modeling enables the creation of numerous realizations of facies that can be thought of as equiprobable outcomes. We suggest a novel stochastic methodology that combines the efforts of probabilistic data integration to address the flaws in this approach.
Multiple-point geostatistics is based on the idea of going beyond two-point correlations between variables and obtaining (cross) correlation moments at three or more places at the same time utilizing training images to identify patterns of geological heterogeneity ( in our case veins). Since the Multiple-point statistics approach based on training image, also this work was investigated the appropriateness of using geological model produced in Leapfrog as a training image. Furthermore findings demonstrated that direct sampling (DeeSse) is a feasible multiple-point geostatistics approach for replicating the long-range connectivity and the structure of veins in gold deposit