Introduction
Mineral Resource
A Mineral Resource is a concentration or occurrence of solid material of economic interest in or on the earth’s crust in such form, grade or quality, and quantity that there are reasonable prospects for eventual economic extraction (CIM - Definition Standards for Mineral Resources & Mineral Reserves, 2014).
A Mineral Resource is a concentration or occurrence of solid material of economic interest in or on the earth’s crust in such form, grade or quality, and quantity that there are reasonable prospects for eventual economic extraction (CIM - Definition Standards for Mineral Resources & Mineral Reserves, 2014).
Mineral Resource Estimation Workflow
Figure 1 represents a mineral resource estimation (MRE) workflow that begins with a multivariate dataset of continuous and categorical variables obtained from sample analysis and description (step 1). The samples must then be grouped into spatially continuous and statistically consistent domains to enable the geostatistical prediction of metal grades in unsampled areas while minimizing dilution from low-concentration sites. A good set of groups (domains) should exhibit high similarity among their elements, clear differences between them, and spatial continuity (step 2). The next step involves defining the boundaries or volumes of the groups (step 3), followed by constructing a geostatistical model to predict metal concentrations at discretized points within these volumes based on nearby samples that belong to the same group (step 4).
Figure 1 represents a mineral resource estimation (MRE) workflow that begins with a multivariate dataset of continuous and categorical variables obtained from sample analysis and description (step 1). The samples must then be grouped into spatially continuous and statistically consistent domains to enable the geostatistical prediction of metal grades in unsampled areas while minimizing dilution from low-concentration sites. A good set of groups (domains) should exhibit high similarity among their elements, clear differences between them, and spatial continuity (step 2). The next step involves defining the boundaries or volumes of the groups (step 3), followed by constructing a geostatistical model to predict metal concentrations at discretized points within these volumes based on nearby samples that belong to the same group (step 4).
Figure 1: MRE workflow. Where DDH (Diamond Drill Hole), RBF (Radial Basis Function), and gamma(h) (the variogram for a given distance or lag h).
Defining Spatial Continous Groups (Domains) for Grade Estimation Using Clustering
The definition of estimation domains is a critical and fundamental step in Mineral Resource Estimation (MRE) as it determines the location of mineralization beneath the ground by identifying groups of ore with the highest concentrations of the target metal. It is crucial to accurately select the appropriate mining method that accounts for the required selectivity to extract the mineralization without diluting it with host rocks.
Currently, the methodology for creating these groups involves manually digitizing polygons based on a cutoff grade applied to a primary continuous variable, which is time-consuming and subjective once different modelers can propose different groups that are spatially and statistically distinct. An error in this stage can lead to a wrong definition of the quantity of metal content and its position in space. To address these limitations, we tested three clustering methods to create groups that can provide more consistent, rapid, and less subjective domains.
The definition of estimation domains is a critical and fundamental step in Mineral Resource Estimation (MRE) as it determines the location of mineralization beneath the ground by identifying groups of ore with the highest concentrations of the target metal. It is crucial to accurately select the appropriate mining method that accounts for the required selectivity to extract the mineralization without diluting it with host rocks.
Currently, the methodology for creating these groups involves manually digitizing polygons based on a cutoff grade applied to a primary continuous variable, which is time-consuming and subjective once different modelers can propose different groups that are spatially and statistically distinct. An error in this stage can lead to a wrong definition of the quantity of metal content and its position in space. To address these limitations, we tested three clustering methods to create groups that can provide more consistent, rapid, and less subjective domains.