Geostatistical simulation of compositional data in the presence of several geological domains

This paper proposes a methodology for geostatistical simulation of compositional data in presence of several geological domains: After applying statistical analysis and compositionally compliant contact analysis on the data set, geological domains with the same compositional characteristics are merged. Plurigaussian simulation (PGS) is utilized to generate multiple numerical models of the remaining domains, with the aim of assessing the uncertainty in the domain boundaries and improving the geological controls in the characterization of compositional attributes. The resulting conditional realisations of geological domains are used to derive the posterior probabilities of occurrence of the domains over the deposit. The compositional data set is classified by geological domain and for each class realisations are generated across the target grid of the resource. The final mineral resource model is obtained by weighting the simulated mineral compositions associated with the different domains by the probabilities of occurrence of each domain. This approach accounts for the uncertainty in the geology of the ore body and spatial variability of mineral composition.

The approach is illustrated through an application to a nickel-cobalt laterite deposit located in Western Australia. A set of 8818 samples is considered in order to provide the conditioning data. Four rock types (Ferruginous, Smectite, Saprolite, and Ultramafic) are considered to define compositionally homogeneous domains. Three major (Fe, Al, and Mg) and two target (Ni and Co) elements are the variables of interest in this study. Since the data are compositional a filler variable is introduced to achieve closure and to retain the intuitive relation between each component and the mass of its associated element.