Geostatistical Modelling of Gold Grades by Co-Kriging-Based Approach for Preserving the Outlying Values

Abstract

Geostatistical modeling of Gold grade (Au) is challenging since the presence of outliers makes the distribution long-tailed and impacts significantly the process of mineral resource evaluation, the mine design and financial optimization. Capping is a widely used technique consisting of truncating the data to some top-cut grade. However, this procedure is likely to omit the most important part of a deposit that probably is economically considerable. In this research, a co-kriging-based approach is applied in a gold deposit to preserve the upper quantile of the Au distribution while improving the precision of the estimation. The rationale of this idea is to divide the grade of Au into: truncated grade, a weighted indicator above the top-cut grade and a zero-mean residual. After this decomposition, the co-kriging is able to jointly estimate the truncated grade and the indicator. The benefit of this approach is to provide unbiased grade estimation and choosing the optimum top cut value while avoiding the outlying values for spatial continuity calculation and implementing the spatial prediction.