Impacto da amostragem adicional na redução da incerteza associada à estimativa e classificação de recursos minerais

Abstract

Analyzing the uncertainty associated with the estimate is a fundamental step during the sampling mesh optimization process, where the regions of the deposit that require the collection of new samples are defined or, even, during the classification of resources and reserves. Both approaches are challenging and require the comparison of different methodologies to define the one that presents the best performance. On the one hand, during sample density campaigns, the responsible professional needs to adopt criteria for choosing the sampling pattern to be used (regular sampling versus sampling in regions of higher uncertainty), in addition to evaluating the uncertainty index that presents the best performance in defining the most erratic regions of the deposit. On the other hand, about resource classification, minimum classification criteria must be used, based on the most sophisticated resources and reserves classification codes. Faced with the problem described above, this thesis proposes using two approaches based on a database from a coal deposit located in the municipality of Figueira, Paraná, Brazil. In the first approach, based on virtual samples obtained by conditional simulation, a study related to the impact of additional sampling on reducing global and local uncertainty is presented, considering the conditional variance (VC) and interquartile range (IEQ) indices, demonstrating how this methodology can help in choosing the regions that require additional sampling and the number of samples to be added, which can represent savings in financial resources and time for mining companies. The second approach deals with a comparative study between resource and reserve classification methodologies, demonstrating the advantages and disadvantages of Maximum Estimation Error (MEE) and Superkriging in assessing the error associated with the estimate. More satisfactory results were obtained through Maximum Estimation Error in classifying resources contained in annual and quarterly production areas, while Superkriging presented high error percentages.