The management of uncertainty in three-dimensional (3D) geologic models has been addressed by researchers across a range of use cases including petroleum and minerals exploration and resource characterization, as well as hydrogeologic, geothermal energy, urban geology, and natural hazard studies. Characterizing uncertainty is a key step toward informed decision-making because knowledge of uncertainty allows the targeted improvement of models, is indispensable to risk analysis, improves reproducibility, and encourages experts to explore alternative scenarios. In the minerals sector there is not a unified approach to uncertainty characterization, nor its mitigation.
Assessing and mitigating uncertainty in 3D geologic models is a growing field but quite compartmentalized among different subdisciplines within the geosciences. By comparing uncertainty analysis as implemented for three modeling scenarios: basins, regional hard-rock terranes, and mines; at different stages of their respective workflows, we can better understand what a future “complete” modeling platform could look like as applied to the minerals industry.
We analyze uncertainty characterization during the different steps in building 3D models as a generic workflow that consists of (1) geologic and geophysical data acquisition followed by processing and inversion of geophysical data, (2) the interpretation of a number of discrete domains boundaries defined by stratigraphic and structural surfaces, (3) homogeneous or spatially variable properties infilling within each domain, and finally (4) use of the models for downstream predictions based on these properties, such as resulting gravity field, gold grade distribution, fluid flow, or economic potential.
Although regional- and mine-scale modelers have much to learn from the basin modeling community in terms of managing uncertainty at different stages of the 3D geologic modeling workflow, perhaps the most important lesson is the need to track uncertainty throughout the entirety of the workflow. At present in the minerals sector, uncertainties have a tendency to be recognized within discrete stages of the workflow but are then forgotten, so that at each stage a “best guess” model is provided for further analysis, and all memory of earlier ambiguity is erased.
