Simulation of Categorical and Continuous Variables of a South-American Copper Deposit

Abstract

Planning a mining operation requires modeling the metal grades contained in the ore of the deposit based on sampling data, which are inevitably limited relative to the volume of the deposit. Unlike estimation, geostatistical simulation allows not only for the construction of a discrete three- dimensional model of the target grades, but also for the quantification of uncertainty through the generation of multiple equiprobable realizations.

This paper presents the construction of a three-dimensional block model of a porphyry copper deposit in South America. The study area covers approximately 1.4 km² to a depth of about 600 m, and the available data consist of approximately 40000 samples containing copper grade values and interpretive geological models provided by the data-owning company to define rock and mineral types. Since rock type is the primary control on the spatial distribution of metal grades, the work follows a hierarchical approach: simulating the geological domains first and then the grades within each of them allows this dependency to be accounted for and reduces the range of variability to be modeled within each domain. The geological domains are simulated using Truncated Plurigaussian Simulation, a method that allows for the reproduction of complex contacts between lithotypes; the total copper grade is then simulated within each domain separately using Turning Bands Simulation.

This paper has presented the methodology of how to obtain different block models of a South American copper deposit through the combined application of Truncated Plurigaussian Simulations for categorical variables and Turning Bands Simulations for continuous variables.

The geostatistical simulation approach offers clear advantages over traditional kriging: it reproduces the full variability of the deposit, avoids the smoothing effect, and provides multiple realizations that quantify uncertainty for mine planning purposes.