Integrating different generations of assays using multivariate geostatistics: a case study

Abstract

This paper summarises the results of a case study undertaken at a polymetallic deposit in South Australia. Multivariate geostatistical techniques were used to predict better the block sulphur grades. The necessity to utilise a multivariate geostatistical approach was dictated by the presence of two different generations of sulphur assays. The integration of these data sets was undertaken using the kriging with external drift and collocated co-kriging methods. Efficiencies of the multivariate methods have been tested using various combinations between data generations. Results of these tests have been compared with the conventional ordinary kriging estimates and reconciled against the drill hole data. This study shows that both tested multivariate methods – kriging with external drift and collocated cokriging – significantly improved the accuracy of grade estimation when compared with conventional ordinary kriging. In particular, a global average grade of the study area estimated by the multivariate methods was similar to expected true values even when the amount of new data available is less than 10% of all samples. The ordinary kriging technique when applied to such datasets containing less than 10% of the new samples and more than 90% of the old (biased) samples yields biased results which are largely affected by the more abundant old (biased) assays.