Relative Kriging Errors - A Basis for Mineral Resource Classification

Abstract - During geostatistical estimation of block grades, the kriging process produces a measure of error, the kriging variance or the kriging standard deviation. Estimation of the block error depends on the semi-variogram model for a deposit, which reflects fundamental characteristics of a deposit such as (1) nugget effect, (2) range of influence of grades, (3) different local spreads of data as a function of average grade (proportional effect), (4) nested structures, and (5) underlying anisotropies in the data; this model may be affected by sampling or assaying limitations, however. Because the kriging error incorporates so many features of both the deposit and data to be used for estimation, it may offer a substantial improvement on block classification over traditional procedures such as data density. Practice indicates that the relative kriging standard deviation (RKSD) is particularly useful for block classification in measured, indicated and inferred categories. Plots of RKSD in plan and section also can be used to demonstrate where extra drilling and sampling would be most advantageous for upgrading the class of some blocks. Moreover, the use of grade and RKSD filters can show where material near cut-off grade is poorly defined, indicating sampling target areas to improve block classification in the vicinity of the ore-waste boundary.