Uncertainty management in resource models is inherently interdisciplinary. Collaboration among geologists, geostatisticians, mine planners, and geometallurgical specialists enables the translation of geological uncertainty into operational and economic variables.
Mineral resource models do not represent reality; they are approximations subject to multiple sources of uncertainty. Historically, conditional simulations have been applied only to continuous variables (e.g., element grades), leaving geological variability largely unaddressed.
The main sources of uncertainty in a resource model are:
- Sampling quality
- Data spacing (information effect)
- SMU size (support effect)
- Geological variability (lithology, mineralized zones, alterations)
- Grade variability
Simulating grades alone is insufficient. The geology of the deposit—lithology, mineralized zones, and alterations—defines the primary controls on mineralization and establishes the estimation domains. It also directly determines tonnages and, consequently, the metal content of a deposit. Without modelling geological uncertainty, total model uncertainty is systematically underestimated.
Several algorithms can simulate geology as a categorical variable, generating equiprobable scenarios of the deposit. The most widely used include:
- Distance-based simulation
- Latent Gaussian variable simulation
- Indicator simulation
Sequential indicator simulation often yields erratic results, failing to reproduce complex geology. Distance-based simulations produce geologically consistent realizations and are straightforward to implement, but their main drawback is inherent non-stationarity, which complicates variogram modeling. Plurigaussian simulations also provide geologically consistent results, but their implementation is more complex and supported by limited commercial software.
An efficient way to interpret the results is through entropy calculation derived from categorical realizations. This allows the generation of geological uncertainty maps and the identification of both under-sampled areas and zones of higher geological complexity.
Some practical applications are:
- Optimization of drill spacing for resource classification
- Input for stochastic mine planning
- Definition of priority areas for new drilling
- Identification of high-uncertainty zones within mine plans
- Operational risk mapping
This paper presents case studies that integrate geological variability using categorical variable simulation techniques. Their main advantages and limitations are comparatively analyzed, highlighting the importance of geological simulations across different deposit types.
Comparisons are presented between categorical–numerical simulation and purely numerical approaches across different deposit types—a skarn and two copper porphyries—where geology exerts distinct controls on mineralization. In all cases, tonnage and metal uncertainty range from 15% to 40%, strongly dependent on geological controls and the support level at which they are evaluated.
A comparison was conducted within a copper porphyry between classification derived from the SSDV (Spatial Sampling Density Variance) method, based on kriging variance, and classification obtained from conditional simulations (geology and recoveries). In this case, highly continuous geological units show no method differences in category; however, in more variable units, the difference reaches up to 35% in Category 1.
Models based solely on kriging—which tend to oversimplify variability and underestimate uncertainty—are compared with models that integrate conditional simulations of both geology and grades. The results are consistent: integrating both simulations within a stochastic planning framework leads to strategies with higher metal recovery and improved economic performance.