Evaluation of Kriging and Cokriging for Asbestos Ore Reserve Estimation at the Cana Brava Mine, Goiás, Brazil

Abstract

The ore-body at the Cana Brava Asbestos Mine in Minaçu, central-eastern Goias, Brazil, extends from north to south and dips 45° westwards. In the development of this mine more than 1000 vertical diamond drill holes were completed on a regular 35×35 m² grid. Representative samples from mineralized zones were analysed for total grade (TT), long fiber grade (TL), medium fiber grade (TM), and short fiber grade (TC), as well as the grades retained on different sieve sizes (T7, T14, T25, T50, and T200), length index (IC), and dust (PO). All variables show an asymmetrical distribution with average values of 3% for TT, 650 for IC and around 18% for dust.

This paper presents a comparative study between kriging and cokriging for asbestos reserve evaluation at the Cana Brava Mine. The main reason for using cokriging was the well-defined linear relationship between total grade (TT) and all other variables considered as secondary ones. The results have proved that both methods generate very similar estimates of reserves, confirming that when secondary variables are well correlated with the primary variable, cokriging gives results comparable to ordinary kriging. Despite of this fact we conclude that there is no need for cokriging when primary and secondary variables are well correlated and equally sampled. Therefore, in this case the ordinary kriging technique is sufficient for reliable reserve estimation. Subsequently, several heterotopic data compositions were made in order to study the influence of the primary variable over the secondary one, as well as to study the influence of the correlation coefficient between primary and secondary variables. Results illustrated that the correlation coefficient has a decisive influence on the precision of estimates rather than the contribution of the primary variable.