Geostatistical modeling of heterogeneous geo-clusters in a copper deposit integrated with multinomial logistic regression: An exercise on resource estimation

Resource estimation is the main and primary step in development of a mining project. Principally, it necessary to first identify geological domains through boreholes, model them at unsampled locations, then evaluate grade(s) interest inside each built domain. The traditional determination these categorical over sampling points suboptimal as considers mostly-one or two variables from core logging. This leads neglect influence other significant variables. To circumvent problem domain identification, spatially dependent clustering machine learning algorithms can be great help detecting such domains. However, one that may appear when using techniques resulting geo-domains (geo-clusters) obtained by technique might heterogeneous show non-stationary property. reason aim produce compact contiguous clusters, which are well suited establishing geo-domains. makes procedure modelling challenging necessitates use advanced geostatistical propagate geo-clusters locations. An algorithm presented this study employs sequential indicator simulation paradigm complex variability geo-clusters. Since spatial trends underlying required method, study, we propose multinomial logistic regression infer trends. was tested an actual case porphyry copper deposit Iran, where Cu, Mo, Au, Rock Quality Designation (RQD), mineralization zones, alteration types, rock types were employed entire deposit. results compared with conventional no trend used. examination maps several evaluation criteria including visual inspection realizations, probability maps, reproduction proportion geo-cluster, connectivity measures, analysis, showed findings proposed superior