Date of Award
Bachelor of Science Honours
Faculty of Computing, Health and Science
Dr Ute Mueller
Although it can be time consuming and computationally more expensive to work with multivariate data, it is often desirable to exploit the relationships among and between the variables sampled across a study region. In most cases the availability of this secondary information can enhance the estimation of the primary variable(s). The aim of this research is to describe and then demonstrate the use of multivariate statistical methods in geostatistical analysis. The methods will be illustrated by application to a multivariate data set from an actual mineralisation. The data suite is known as MM22D and comes from the Murrin Murrin nickel mine near Laverton in Western Australia. Two four variable subsets of the MM22D data suite are used in the application. The first is MM22DHC4 and consists of nickel, cobalt, iron and zinc, chosen for their high correlations with each other. The second is MM22DTOP4 and consists of nickel, cobalt, magnesium and iron, chosen for their economic importance to the mining company. This thesis presents the theory and the process of the modelling and estimation of multivariate data. We demonstrate the use of principal component analysis in a geostatistical environment, in particular for the detection of intrinsic correlation. We illustrate the modelling and estimation of an intrinsically correlated data set (MM22DHC4) and estimate the variables in this data set using ordinary cokriging, principal component kriging and ordinary kriging. In addition we illustrate the derivation of a general linear model of coregionalisation and estimate the variables of this data set (MM22DTOP4) using ordinary cokriging and ordinary kriging. The grade control data from the MM22D data suite, which were considered reality, were used as a comparison and assessment of the accuracy of all of the estimates. As the data used in this study were isotopic it was anticipated that there would be little difference in the estimates obtained which was indeed the case.
Bandarian, E. (2003). Multivariate geostatistical analysis with application to a Western Australian nickel laterite deposit. Retrieved from http://ro.ecu.edu.au/theses_hons/338