Boundary modeling with uncertainty

Abstract

Mineral exploration and mining are capital intensive and carry significant environmental and societal considerations. The feasibility of prospective mining operations hinges on a series of engineering decisions. The mineral resource estimation of the quality and quantity of resources greatly influences these decisions. Boundary modeling for defining subsurface geology is an essential aspect of the mineral resource estimation workflow. The proper spatial distribution of geological domains for further estimation is integral to having accurate and precise models. Uncertainty in the boundary and resource is quantifiable and is a result of sparse sampling and complex geology. Consequences of poorly defined boundaries include dilution of ore material, smearing of grade into uneconomical rock, and increased uncertainty in the deposit tonnage. These consequences directly impact the economic, environmental, and societal feasibility of operations. Boundary modeling workflows commonly use implicit techniques that automatically derive domain extents from data. These models are explicitly checked and edited to ensure the numerical model reflects known geological attributes. This deterministic approach generates a single model output and does not carry a measure of uncertainty. Stochastic approaches to boundary modeling capture short scale variability of the geology; however, imparting geological knowledge on the model is difficult. This thesis develops a new implicit methodology for boundary modeling that provides globally unbiased models with uncertainty. The indicator approach maps a field of probabilities and applies a threshold that results in an extracted boundary. Uncertainty assessment by varying the indicator thresholds provides eroded and dilated boundaries and a zone of uncertainty. Boundary definition is a critical and early step in resource estimation. The modeling of subsurface geology from sparse drill holes carries significant uncertainty. The results of this thesis provide a novel approach to boundary modeling with uncertainty. The geostatistical techniques and concepts are reviewed, the proposed framework outlined, and implementation, including case studies, are undertaken.