MULTIVARIATE SIMULATION OF CHANNEL IRON ORE DEPOSITS AT BUNGAROO AND YANDICOOGINA, WESTERN AUSTRALIA

Abstract

Geostatistical conditional simulation has wide potential applications in the iron ore industry and is the favoured tool to assess variability and risk. Multivariate relationships are important in such simulation, for example between Fe and impurities such as Al2O3, SiO2 and P. Turning bands has been the main conditional simulation algorithm used in the Western Australian iron ore industry. In this thesis a more recent approach using minimum/maximum autocorrelation factors (MAF) and sequential Gaussian simulation (SGS) are used together and performance comparisons are made with turning bands at Yandicoogina, a channel iron ore deposit (CID) in Western Australia. MAF-SGS and turning bands algorithms both performed well in simulating Fe, SiO2, Al2O3 and P at Yandicoogina. Extensive checking of simulations showed both approaches could reasonably reproduce multivariate statistics and spatial continuity of composites including means, variances, histograms, quantile-quantile plots, scatter plots and variography in normal scores and data space, as well as in MAF space for the MAFSGS approach. MAF generated from transformed composites were largely uncorrelated and able to be considered independent for variography and simulation. Later back transformation from MAF space to normal scores space then to data space successfully reintroduced joint relationships seen in the conditioning data. While the MAF-SGS approach needs additional transformations compared with turning bands, a linear model of coregionalisation and hence the modelling of cross semivariograms is not required. If there are a high number of variables then construction of a linear model of coregionalisation becomes more difficult and the MAF approach may be preferred. In this study four variables were considered and a linear model of coregionalisation could be built. Turning bands transformations are from data space to normal scores space only, with no need to calculate or check any decorrelated factors. The two main methods currently used in the mining industry for determining optimum drillhole spacing with the use of conditional simulation were compared at the Test Pit area of the Bungaroo channel iron ore deposit, also in Western Australia. The method generates precisions for various drillhole spacing using two stages of simulation, whereas the approach calculates expected relative errors for different spacings via a simulation stage followed by an estimation step. Clear differences exist between the relative errors from the simulation-estimation method and the precisions calculated from the simulation-simulation method. The simulationestimation method appears more insensitive to the grid spacing with only moderate improvements in relative error as the drillhole spacing is tightened. The simulationsimulation method shows more marked improvement in precision with closer spacing and appears more realistic in this study. Al2O3 is the main variable to consider at Bungaroo when choosing a suitable drillhole spacing. SiO2 grades of composites are mainly below the SiO2 cut off grade for ore whereas the mean Al2O3 is grade is quite close to the Al2O3 cut off grade for ore. Hence although SiO2 has higher variability, it is not as critical as Al2O3 for determination of drillhole spacing. Fe and P have greater spatial continuity than Al2O3 and SiO2 and do not require such close drillhole spacing. Fifteen percent precision, based on a volume representing a quarter’s production and using the simulation-simulation method, may be regarded as acceptable for mine planning purposes. A resource evaluation drilling spacing of 150m along strike and 50m across strike appears appropriate for a precision of less than 15% for a quarterly mining volume in determining Al2O3 grade at Bungaroo.