Best practice in Multiple Indicator Kriging (MIK) – importance of post-processing and comparison with Localised Uniform Conditioning (LUC)

Multiple indicator kriging (MIK) has been used in the minerals industry for some decades. As one of the non-linear estimation methodologies, MIK has advantages related to resolving multiple or mixed populations, high variability data and strongly skewed distributions. Applying the MIK methodology requires a lot more effort when compared to other non-linear methodologies, for example localised uniform conditioning. This is due not only to the variogram modelling for multiple indicators, but the validation is also very important because the panel model is fundamental for follow-up work, including the generation of recoverable resources at the local or SMU scale. Apart from this, only a few reported case studies of MIK have documented post-processing of the estimated modelled distribution, representing the probability distribution at point scale at un-sampled locations. The MIK post processing should include the change of support from the point scale to the panel scale, and then the extraction of quantile values from the panel conditional cumulative distribution function (CCDF) for localisation of grades into the SMUs within each panel, based on a ranking estimate marking the SMUs from high to low. In this paper, we present a case study on a gold deposit with the MIK method with full-post processing implemented for recoverable resources. The MIK point estimate was carried out with an indirect lognormal change of support from point to panel scale, and localisation was done with a custom script. The localised uniform conditioning (LUC) method was also applied for comparison purposes. The comparison shows that MIK has advantages compared to LUC when comes to capturing the high-grades, especially when mixed or varying anisotropy is present within the mineralisation.