One of the major objectives for geologists undertaking feasibility studies on potential mineral deposits is the correct delineation of ore from waste. However this is an objective that can never be perfectly achieved because a geologist can not know the distribution of the mineralisation at all scales at all locations. Therefore the objective of a geologist is to minimise ore misclassification.
There are many factors contributing to ore misclassification in gold mines, one of which is random sampling error. It has been recognised for a time that random sampling errors experienced by the gold mining industry can be of considerable magnitude and that these random errors introduce a great deal of uncertainty when characterising the geologic properties of the mineralisation for resource modelling and mine planning.
This research reviews how random sampling error is measured and then quantifies the impact of prescribed random sampling errors on the anticipated profitability of a gold mine. An Archaean shear zone hosted gold deposit, the Yilgarn Star mine in Westem Australia, is used to demonstrate the impact of random sampling errors on mine profitability.
Firstly, a review of the performance of various measures of sampling error is undertaken. This review identifies the Pairwise precision measure as the most robust measure of the magnitude of the random sampling error. The magnitude of the sampling error of various mining data sets 1s then explored using the Pairwise precision and found to be in excess of ±50%.
A model of the Yilgarn Star mineralisation is built using a sequential indicator simulation algorithm conditioned to the exploration and grade control data. This model is considered to be an image of the true mineralisation and is sampled at a nominated sample density similar to many exploration drilling programmes. A random sampling error of a known magnitude is added to the resultant sampled reference model and a recoverable resource model of the mineralisation using the v multiple indicator kriging algorithm is built. This process is repeated for three different random sampling errors.
The financial impact of the three random sampling errors is quantified through the Lerchs-Grossman pit optimisation transfer function as applied in the Whittle 4D software. The maximum profit open pit shells derived from the various recoverable resource models show that random sampling errors can reduce the anticipated profitability by over 30% and misclassify over 20% of the economic ounces as waste.
The economic benefit of minimising the sampling error is demonstrated to be substantial. The expenditure of significant additional capital to correctly delineate the characteristics of the mineralisation before mining commences results in enhanced profits for the mining company and more efficient utilisation of the resource for the public good.