Incremental Improvement of Grade Control Models Using Resource Data

"The mine geologist’s role in classifying ore from waste during grade control represents a critical link in the mining value chain. Any improvement in the ore/waste selection success rate (thus reduction in misclassification) which can be delivered by the mine geologist represents a value proposition. Improvement achieved without additional operating cost (for example extra drilling and sampling) will directly and immediately add value to the mining operation and therefore clearly must be pursued. Such improvement may be leveraged from existing resource drilling data which are typically less numerous but better quality than the grade control data. They are usually excluded from grade control estimates (termed the ‘base case’ approach here) because of the differences in quality.Various specialised geostatistical tools are available to account for differences in precision and accuracy between data sources during spatial estimation. This paper uses a conditional simulation (CS) case study to assess the benefit of integrating resource data into grade control estimates through ordinary kriging (OK) of pooled data sets, cokriging (CK), and ordinary kriging with variance of measurement error (OKVME) when the grade control data are:unbiased but increasingly imprecisebiased and increasingly imprecise. The premise assumed is that the resource data are unbiased and precise.The paper demonstrates that, if appropriately handled, resource data may be used to improve grade control estimates and thus improve ore/waste selection success without additional operating cost. Where grade control data are imprecise but unbiased it is recommended to integrate resource data using OKVME. This re-appropriates kriging weights from less precise to more precise data locations and improves the grade control estimation precision compared to the base case; it consequently provides an improvement in the grade control selection success rate. If grade control data are associated with significant bias, integration through CK is recommended as the biased data are zero sum weighted. CK consequently provides an unbiased estimate with some improvement in estimation precision compared to the base case and consequently improvement in the ore/waste selection success rate.CITATION:Cornah, A J, 2014. Incremental improvement of grade control models using resource data, in Proceedings Ninth International Mining Geology Conference 2014 , pp 387–396 (The Australasian Institute of Mining and Metallurgy: Melbourne)."