ABSTRACT
Ordinary kriging (OK) and indicator kriging (IK) are probably the most frequently used kriging estimators in the mineral industry. OK is desirable when estimating a grade having symmetric distributions, but it tends to produce overly smoothed estimates for highly skewed grades. IK likely improves the variability of the grade estimates, but it tends to yield poor estimates of average grade for blocks or mining panels. Restricted kriging (RK), originally proposed to restrict the influence of erratic sample values, is generalized as a combination of OK and IK. Instead of using the ordinary unbiasedness condition, RK adopts a set of constraints based on the estimates of multiple indicator kriging. Of course, RK remains an unbiased estimator. The new approach restricts the parameter search space in such a way that the allocation of coefficients observes the characteristics of local grade distributions. Hence, RK combines the advantages of both OK and IK, and in effect, it reduces over-smoothing effects. A synthetic example is presented to illustrate how RK works in practice. Finally, a case study is given to demonstrate the effect of the new technique in reserve estimation.
