203
2 CLASSIFICATION METHODS
Mineral resource classification is important to mining companies, investors, and financial
institutions, as investment decisions are usually based on grade, tonnage and confidence
assigned to deposits. However, the approach used to perform classification remains subjective because of lack of formal standards. The various public reporting codes do not categorically recommend the method or threshold needed to be used for classification, as the process
depends on the judgement of the responsible CP or QP. The industry’s best practice for
classification is assessing and quantifying the uncertainty and risk associated with mineral
resource estimation. The two basic methods used to perform classification tasks are the geometric and geostatistical techniques. In the survey, the geometric classification method was
discovered to be mostly used in the mineral industry. Classification is commonly performed
on a block-by-block basis, but the volumes are chosen reasonably large and contiguous,
because one often believes that confidence in the grade should not change abruptly between
adjacent blocks, Deutsch et al (2016).
2.1 Geometric methods
The geometric methods of mineral resource classification consider the amount, proximity
and location of data available for estimation of a block. In the review of the recent classification reports, the geometric information used by the industry professionals included the
dimensions of ellipsoidal search (ES), number of drill holes (NDH), minimum number of
samples per estimate (NS), distance to nearest drill hole (DNDH), average drill hole spacing (DHS), and de-clustering by octant search (OS). The survey of the classification reports
showed that each QP unilaterally selected the information to be considered in assigning
confidence levels to blocks, as there are no standards to determine the needed information
to be used for a particular type of deposit. The quantity of information to be used for classification is unrestricted, hence various QPs in the industry make different assumptions
in assigning parameters to define the classification categories. Considering the minimum
distance between a block’s centroid and the composite samples for estimation, some QPs
assigned different percentages of the variogram sill range while others assumed different
values, based on their understanding of the deposits. Examples of the different assumptions
made by the QPs in the reviewed technical reports can be found in the survey results in the
next section. Generally, the geometric classification method does not consider the spatial
continuity of the data in characterizing uncertainty associated with the estimation of the
grades.
2.2 Geostatistical methods
The geostatistical classification methods are used to quantify risk on a given future production period. It is an effective and efficient method used to model geologic and grade
uncertainty in mineral deposits. According to Deutsch et al (2016), desire to have purely
probabilistic criteria based on sound estimates of uncertainty are understandable. The probabilistic techniques rely on classifying the blocks either by using KV directly or characterizing uncertainty of grade estimates, tonnages, and quantity of metals based on confidence
intervals, kriging variance, and/or conditionally simulated realization of grades. After reviewing the recent NI 43-101 public reports, it was evident that the industry players have not
embraced the geostatistical techniques. Although it makes sense, characterizing uncertainty
of grade estimates, tonnages and quantity of metals based on confidence intervals, kriging
variance and/or conditionally simulated realization of grades were applied by only few companies in the mining industry.
2.2.1 Kriging variance approach
Kriging is a minimum variance estimator which minimizes the squared error between
the estimated value and the unknown true value. The error variance generated from the
2 CLASSIFICATION METHODS
Mineral resource classification is important to mining companies, investors, and financial
institutions, as investment decisions are usually based on grade, tonnage and confidence
assigned to deposits. However, the approach used to perform classification remains subjective because of lack of formal standards. The various public reporting codes do not categorically recommend the method or threshold needed to be used for classification, as the process
depends on the judgement of the responsible CP or QP. The industry’s best practice for
classification is assessing and quantifying the uncertainty and risk associated with mineral
resource estimation. The two basic methods used to perform classification tasks are the geometric and geostatistical techniques. In the survey, the geometric classification method was
discovered to be mostly used in the mineral industry. Classification is commonly performed
on a block-by-block basis, but the volumes are chosen reasonably large and contiguous,
because one often believes that confidence in the grade should not change abruptly between
adjacent blocks, Deutsch et al (2016).
2.1 Geometric methods
The geometric methods of mineral resource classification consider the amount, proximity
and location of data available for estimation of a block. In the review of the recent classification reports, the geometric information used by the industry professionals included the
dimensions of ellipsoidal search (ES), number of drill holes (NDH), minimum number of
samples per estimate (NS), distance to nearest drill hole (DNDH), average drill hole spacing (DHS), and de-clustering by octant search (OS). The survey of the classification reports
showed that each QP unilaterally selected the information to be considered in assigning
confidence levels to blocks, as there are no standards to determine the needed information
to be used for a particular type of deposit. The quantity of information to be used for classification is unrestricted, hence various QPs in the industry make different assumptions
in assigning parameters to define the classification categories. Considering the minimum
distance between a block’s centroid and the composite samples for estimation, some QPs
assigned different percentages of the variogram sill range while others assumed different
values, based on their understanding of the deposits. Examples of the different assumptions
made by the QPs in the reviewed technical reports can be found in the survey results in the
next section. Generally, the geometric classification method does not consider the spatial
continuity of the data in characterizing uncertainty associated with the estimation of the
grades.
2.2 Geostatistical methods
The geostatistical classification methods are used to quantify risk on a given future production period. It is an effective and efficient method used to model geologic and grade
uncertainty in mineral deposits. According to Deutsch et al (2016), desire to have purely
probabilistic criteria based on sound estimates of uncertainty are understandable. The probabilistic techniques rely on classifying the blocks either by using KV directly or characterizing uncertainty of grade estimates, tonnages, and quantity of metals based on confidence
intervals, kriging variance, and/or conditionally simulated realization of grades. After reviewing the recent NI 43-101 public reports, it was evident that the industry players have not
embraced the geostatistical techniques. Although it makes sense, characterizing uncertainty
of grade estimates, tonnages and quantity of metals based on confidence intervals, kriging
variance and/or conditionally simulated realization of grades were applied by only few companies in the mining industry.
2.2.1 Kriging variance approach
Kriging is a minimum variance estimator which minimizes the squared error between
the estimated value and the unknown true value. The error variance generated from the