Chapter 5.

6
ORE RESERVE/RESOURCE ESTIMATION
A LAN C. N OBLE

Ore reserve estimates are assessments of the quantity and the potentially minable ore reserve. The most important mining
tenor of a mineral that may be profitably and legally extracted factors for consideration in evaluation of the ore reserve from
from a mineral deposit through mining and/or mineral beneficia- the resource are
tion. Estimation of ore reserves involves not only evaluation of 1. The range of likely cutoff grades.
the tonnage and grade of a deposit but also consideration of the 2. The degree of selectivity and the size of the selective
technical and legal aspects of mining the deposit, of beneficiating mining unit for likely mining methods.
the ores, and of selling the product. Thus a number of profes- 3. Variations in the deposit that affect the ability to mine
sional disciplines may be involved in ore reserve estimation in- and/or process the ore.
cluding geology, geostatistics, mining engineering, mineral pro- These mining factors often determine the degree of detail
cess engineering, mineral economics, land and legal issues, and that is required for the resource model and thus the degree of
environmental engineering. difficulty to develop a resource model for estimating ore reserves.
This chapter, however, addresses only the aspects of ore For example, a disseminated gold deposit may be continuous
reserve estimation that include determination of the tonnage, and regular in shape, if mined by bulk, open pit methods. The
grade, size, shape, and location of mineral deposits. Although same deposit may be discontinuous and difficult to estimate,
these are often referred to as ore reserve estimates, the term however, if mined by more selective underground methods at a
resource estimation is used rather than ore reserve estimation to higher cutoff grade. Such large differences in deposit shape due
emphasize that all aspects of the ore reserve estimate are not to variations in cutoff grade and mining method may require
being considered. different ore reserve estimation methods for different mining
methods.

5.6.1 RESOURCE ESTIMATION METHODOLOGY

A resource estimate is based on prediction of the physical
5.6.2 DATA COLLECTION AND GEOLOGIC
characteristics of a mineral deposit through collection of data, INTERPRETATION
analysis of the data, and modeling the size, shape, and grade of
the deposit. Important physical characteristics of the ore body Data that must be collected and compiled for the resource
that must be predicted include (1) the size, shape, and continuity estimate are as follows:
of ore zones, (2) the frequency distribution of mineral grade, 1. Reliable assays from an adequate number of representative
and (3) the spatial variability of mineral grade. These physical samples.
characteristics of the mineral deposit are never completely 2. Coordinate locations for the sample data.
known, but are inferred from sample data. The sample data 3. Consistently recorded geologic data that describe the min-
consist of one or more of the following: eralization controls.
1. Physical samples taken by drilling, trenching, test pitting, 4. Cross sections or plan maps with the geologic interpreta-
and channel sampling. tion of the mineralization controls.
2. Measurement of the quantity of mineral in the samples 5. Tonnage factors or specific gravities for the various ore
through assaying or other procedures. and waste rock categories.
3. Direct observations such as geologic mapping and drill 6. Surface topographic map, especially for deposits to be
core logging. surface mined.
Estimation of the resource requires analysis and synthesis of Although small deposits may be evaluated manually using
these data to develop a resource model. Methods used to develop data on maps and in reports, the amount of data required for a
the resource model may include resource estimate is often large, and data may be more efficiently
1. Compilation of the geologic and assay data into maps, evaluated if they are entered into a computer database. Computer
reports, and computer databases. programs can then be used to retrieve the data for printing
2. Delineation of the physical limits of the deposit based reports, plotting on digital plotters, statistical analysis, and re-
on geologic interpretation of the mineralization controls at a source estimation. Minimum information that should be in-
reasonable range of mining cutoff grades. cluded in a drillhole database are
3. Compositing of samples into larger units such as mining 1. Drillhole number or other identification.
bench height, seam thickness, or minable vein width. 2. Hole length, collar coordinates, and down-hole surveys.
4. Modeling of the grade distribution based on histograms 3. Sample intervals and assay data.
and cumulative frequency plots of grades. 4. Geologic data such as lithology, alteration, oxidation, etc.
5. Evaluation of the spatial variability of grade using experi- 5. Geotechnical data such as RQD (rock quality desig-
mental variograms. nation).
6. Selection of a resource estimation method and estimation Entry of data into a computer database is a process that is
of quantity and grade of the mineral resource. subject to a high error rate if not carefully controlled and
The estimation procedure must be made with at least mini- checked. Some procedures that may be used to ensure that the
mal knowledge of the proposed mining method since different data have been entered correctly are
mining methods may affect the size, shape, and/or grade of 1. Verification of the data using independent entry by two

344
 ORE RESERVE/RESOURCE ESTIMATION 345
persons. This is a standard procedure at many commercial data- occurring prior to mineralization may have little or no effect on
entry shops that may dramatically reduce data-entry errors. the mineralization or may localize high-grade, vein-type mineral-
2. Manual comparison of a random sample of the original ization that must be modeled independently of a more uniform
data sheets to a print-out of the database. disseminated ore body. It is also important to determine whether
3. Scanning the data for outlier values. For example: drill the fault is a thin, well-defined structure or many smaller struc-
locations outside the project limits, high and low assays, and tures in a complex, wide shear zone. In the first case, the fault
sample intervals that overlap or are not continuous. is modeled as a simple surface with no thickness; in the second,
4. Comparison of computer-plotted data with manually plot- the fault zone must be defined and modeled apart from the
ted maps of the same data. Collar location maps and cross sec- adjoining rock units.
tions are especially useful to rapidly locate inconsistent collar Folding is particularly significant in sedimentary and stra-
locations and down-hole surveys. tabound deposits. Modeling of folding depends on whether fold-
Additional care and attention to detail and accuracy during ing happened before or after ore deposition, on the tendency of
data entry are essential. A database with a large number of errors the ore zoning to follow the stratigraphy, on any remobilization
may result in a resource estimate that is inaccurate and requires that occurred with the folding, and on the creation of traps or
a complete revision to provide defendable results. other favorable structures. In addition to defining the shape of
the folds, it is important to determine whether the mineralization
follows the contours of the folds or is independent of the fold
5.6.3 GEOLOGIC INTERPRETATION geometry.
Multiple phases of mineralization must be defined, particu-
The sample database represents a large three-dimensional larly where they complicate the ore zoning pattern through over-
array of point locations in a deposit. The sample data are quanti- lapping, discordant trends, and through post-mineral oxidation
tative and have been subjected to minimal reinterpretation after or leaching. Secondary enrichment and oxidation will almost
the original measurements. There is another body of geologic always require delineation of the modified ore zones.
knowledge, however, that does not fit this description. This is The character of the ore zone contact must be determined
the interpretation resulting from the geologist’s assimilation of and input into the resource model. A sharp contact will be
the large quantity of geologic data. These interpretative data handled as a discontinuity and the data used strictly indepen-
are often represented on plan maps or cross sections that show dently on either side of the contact. A transitional contact, how-
outlines of the extent of geologic features or iso-grade contours ever, is a broad, gradational boundary that may require data
that define ore zones. These interpretations combine to provide selection from zones of tens of feet (meters) to over 100 ft (30
an interpretative geologic model that is one of the most critical m) to achieve true differentiation between the different grade
factors in the resource estimation. Failure to develop an appro- zones. As a transitional zone becomes thinner, it will eventually
priate geologic ore body model is the most common reason for approach a sharp contact. For practical purposes, any transi-
large errors in the resource estimates. As shown in Fig. 5.6.1, an tional boundary thinner than the smallest selective mining unit
inappropriate geologic model may lead to errors greater than an will be modeled as a discontinuity.
order of magnitude. In addition to definition of these physical ore controls and
The geologist’s interpretation of the ore body should be used post-mineral modifications, a clear understanding of ore genesis
as much as possible in developing the resource estimate. There will always be beneficial in creating a resource model. In the
are, however, practical limits to the amount of complexity that simplest case, the ore genesis will give clues to the behavior of the
can be included in the resource model, and the geologic interpre- grade distributions and variograms; in other cases, the genetic
tation will be limited to critical inputs that define the shape and structure is so dominant that it can be used as a direct control
trends of the mineral zones at different cutoff grades and the in the estimation of mineral resources.
character of the mineral zone contacts.
Examples of geologic features that are often modeled include
1. Receptive vs. nonreceptive host rocks. 5.6.4 COMPOSITING
2. Alteration types that accompany mineralization or create
problems in beneficiation. Compositing is a procedure in which sample assay data are
3. Faulting, folding, and other structural modifications. combined by computing a weighted average over longer intervals
4. Multiple phases of mineralization. to provide a smaller number of data with greater length for use
5. Post-mineral features such as oxidation and leaching. in developing the resource estimate. Compositing is usually a
Changes in lithology are often important variables in re- length-weighted average. If density is extremely variable (e.g.,
source estimation because mineralization can vary due to physi- massive sulfides), however, compositing must be weighted by
cal or chemical attributes of the rocks. The differences may be length times density (or specific gravity).
distinct, such as the sharp contact between a skarn ore body Some of the reasons for and benefits of compositing include
and an unmineralized hornfels country rock. They also may be 1. Irregular length assay samples must be composited to
gradational, such as the gradual decrease in grade that is often provide equal-sized data for geostatistical analysis.
observed between a favorable and slightly less favorable host in 2. Compositing reduces the number of data and may signifi-
a porphyry copper deposit. Other important lithologic controls cantly reduce computational time, which is often proportional
include barren post-mineral intrusive rocks, nonreceptive shale to the square of the number of data.
beds, and other unmineralized materials that are contained 3. Compositing incorporates dilution such as that from min-
within the mineralized zone. ing constant height benches in an open-pit mine or from mining
The effects of faulting will vary according to whether the a minimum height/width in an underground mine.
faulting occurred before or after the mineralization, and to what 4. Compositing reduces erratic variation due to a high nugget
processes accompanied the faulting. A simple post-ore displace- effect caused by erratic high-grade values.
ment may create a discontinuity in the ore trends, preventing There are several different methods for compositing that
simple interpolation across the fault. The same type of fault may be used depending on the nature of the mineralization and
346 MINING ENGINEERING HANDBOOK

Fig. 5.6.1. Overestimation of ore reserves based on a geo-

logic model that is less continuous than the actual ore zones.

the type of mining. Common compositing methods are (1) bench grade to the waste population, resulting in underestimation of
compositing, (2) constant length compositing, and (3) ore zone ore grade and overestimation of waste grades.
compositing. Ore-zone composites are computed by first identifying the
Bench compositing is a method often used for resource mod- interval containing each ore zone in the drillhole. Each ore zone
eling for open pit mining and is most useful for large, uniform is then composited individually as follows: (1) the length of the
deposits. Composite intervals for bench compositing are chosen ore zone is divided by the desired length of the composite; (2)
at the crest and toe of the mining benches. Bench compositing this ratio is rounded up and down to determine the number of
has the advantage of providing constant elevation data that are composites that provide a length nearest the desired length when
simple to plot and interpret on plan maps. In addition, the divided into the length of the ore zone; and (3) the ore zone is
dilution from mining a constant-height, constant-elevation bench composited using length composites starting at the beginning of
is approximated by the bench composite. the ore zone and length as determined in the previous step.
Down-hole composites are computed using constant length A special case of ore-zone compositing is encountered in a
intervals starting from the collar of the drillhole or the top of vein or bedded deposit in that the width of the ore zone is
the first assayed interval. Down-hole composites are used when determined by a combination of minimum mining thickness
the holes are drilled at oblique angles (45° or less) to the mining (height) and assay limits. In these situations, composites must
benches, and bench composites would be excessively long. be recomputed for each combination of assay cutoff grade and
Down-hole composites should also be used when the length of minimum mining thickness.
the sample interval is greater than one-third the length of the Geologic codes are usually assigned to composites according
composite interval to prevent overdilution when the sum of the to the rock type, ore zone, or other geologic feature. This is often
lengths of the samples is much greater than the length of the a simple procedure, since most composites will be computed
composite. from samples taken from a single geologic unit. Assignment of
Ore-zone compositing is a method of compositing that is used geologic codes to composites that cross geologic contacts is more
to prevent dilution of the composite when the width of the complex, since the composite will be computed using data from
contact between waste and ore (or low grade and high grade) is multiple geologic units.
less than the length of a composite. Use of bench compositing If the geologic contact is transitional and does not separate
or down-hole compositing in this case may distort the grade contrasting grade distributions, it is appropriate to assign the
distributions by adding low grade to the ore population and high geologic codes according to the majority rule. If the composite
 ORE RESERVE/RESOURCE ESTIMATION 347
crosses a sharp boundary between contrasting grade distribu- for resource modeling if divided by a disconti-
tions, it is best to use geologic unit compositing or to assign the nuity such as a fault, or if variograms or grade
composite to the geologic unit with the most similar grade. trends are dissimilar.
If some sample intervals in the data are missing assays, it is Above 100% Grade distributions must be separated for
important to determine the reason for the missing data and modeling. Differences of 1000% or more may
account for it appropriately. Typical examples are be observed when barren, mineralized, and/or
1. The missing zone was not assayed because it was low high-grade populations are present.
grade or barren by visual inspection, or the sample was missing
because of poor core recovery in a barren zone.
Rules for analyzing coefficient of variation are as follows
Action: Composite using the average of the barren unit or
zero grade for the grade of the missing assay.
2. The sample was missing because of poor core recovery in COV Interpretation
a narrow post-mineral fault. 0% to 25% Simple, symmetrical grade distribution. Re-
Action: Ignore the missing interval when computing compos- source estimation is easy, many methods will
ites. The volume of the fault zone is small and the grade will be work.
similar to the grades in the country rock.
3. The sample was missing because of poor core recovery in a 25% to 100% Skewed distributions with moderate difficulty
vein that is higher grade and less competent than the surrounding in resource estimation. Distributions are typi-
country rock. cally lognormal.
Action: Ignore the missing interval when computing compos- 100% to 200% Highly skewed distributions with a large
ites, but retain the length of the interval for use in estimating the grade range. Difficulty in estimating local re-
grade of the vein. sources is indicated.
Above 200% Highly erratic, skewed data or multiple popu-
5.6.5 BASIC STATISTICS AND GRADE lations. Local grades are difficult or impossi-
ble to estimate.
DISTRIBUTION
Computation of basic statistics and evaluation of grade dis- Distributions with COV greater than 25% often have a log-
tributions are the first quantitative analyses of the grade data normal grade distribution, and the basic statistics will also be
and are basic tools to provide both feedback to the geologic compiled for the natural logarithms of grades. For a perfectly
analysis and input to the resource modeling. Important factors lognormal distribution, the lognormal statistics are related to the
in these basic studies include normal statistics as follows:
1. Detection of high-grade or low-grade outlier values.
2. Evaluation of the favorability of different lithologies as (5.6.1)
host rocks.
3. Differentiation of complex grade distributions into simple (5.6.2)
populations for resource modeling.
4. Identification of highly skewed and/or highly variable (5.6.3)
grade distributions that will be difficult to estimate.
Basic statistics should be computed for sample and/or com- where α is the average of the natural logarithms of grades; and
posite grades in each geologic domain that is suspected to have β is the standard deviation of the natural logarithms of grades.
different characteristics. This may include different lithologies, Close agreement between the mean, standard deviation, and
alteration types, structural domains, grade zones, or other group- coefficient of variation when estimated using both normal and
ing of data that has been recognized (or suspected) to have lognormal statistics is indicative of a lognormal population and
different grade distributions. Statistics that should be compiled is required to use lognormal statistics.
include
1. Number of data (samples or composites).
2. Average grade, thickness, etc. (mean) 5.6.6 GRADE DISTRIBUTION
3. Standard deviation (std. dev.) and/or variance.
4. Coefficient of variation (COV), the standard deviation The grade histogram and cumulative frequency distribution
divided by average grade. are used to study the relationship between the statistical grade
5. Histogram of grades. distribution and geologic parameters. The analysis is usually
6. Cumulative frequency distribution (probability plot). begun with a histogram of sample or composite grades. If the
The first item reviewed is the number of data; generally, at histogram is bell-shaped and symmetrical, a normal distribution
least 25 data are required to make comparisons between different is indicated, and the cumulative frequency will be plotted on
geologic domains. If sufficient data are available, average grades normal probability paper. Normal distributions are not usually
and coefficients of variation will be compared among the various found in mineral deposits except those with sedimentary origins.
geologic domains. General rules for evaluating differences in If the histogram is skewed to the right, a lognormal distribu-
average grade are as follows tion is indicated and the cumulative frequency distribution will
be plotted on lognormal probability paper. Lognormal distribu-
Grade tions are frequently observed in most hydrothermal precious and
Difference Interpretation base metal deposits.
Normal probability paper is a special graph paper in that
0% to 25% Grade populations that do not usually require the y-axis is a cutoff grade and the x-axis is the percentage of
differentiation for resource modeling. samples above (or below) the cutoff grade. The x-axis is scaled
25% to 100% Grade populations that require differentiation such that a normal distribution will plot as a straight line, the
 Percent Samples Above Cutoff

Fig. 5.6.2. Lognormal probability plotting paper.

slope of the line is proportional to the standard deviation of the Another common deviation from a straight line on the prob-
distribution, and the 50th percentile is the average grade. ability plot is a steeper slope at the upper end of the curve
Lognormal probability paper is similar to normal probability as shown in Fig. 5.6.4. This represents excess material in the
paper except that the y-axis is scaled according to the logarithm high-grade population and may be caused by two superimposed
of cutoff grade. The slope of the line is proportional to the populations, such as high-grade veins within lower-grade dissem-
standard deviation of logarithms of grade β, and the 50th percen- inated or stockwork mineralization. Other causes of excess high-
tile is the average of the logs of grades α. An example of lognor- grade assays include small zones that are highly favorable to
ma1 probability paper is shown in Fig. 5.6.2. mineralization because of higher permeability, favorable chemi-
The probability graph may be used to estimate the standard cal properties, secondary enrichment, or metamorphic remobili-
deviation based on probabilities from the normal probability zation. Since the high-grade mineralization usually has less conti-
distribution as follows: nuity than the lower-grade mineralization, the source of the high
grade must usually be identified and estimated separately from
the remaining mineralization.

which is based on 2 standard deviations, or

5.6.7 VARIOGRAM MODELING
The variogram is the fundamental tool used by the geostatis-
tician and geologist to measure spatial continuity of grade data.
which is based on ± 2.05 standard deviations.
The variogram is a graph of the average variability between
Often the probability graph is not a straight line, but will be samples vs. the distance between samples. A variogram is com-
composed of multiple straight lines or curves. A typical deviation puted by averaging the squared differences between pairs of
from a straight line is a downward curve at the low end of samples that are a given distance apart as follows
the graph as shown in Fig. 5.6.3. This curve represents excess
low-grade samples, and in porphyry systems is often attributed
to weakly mineralized late intrusions or to post-mineral, barren
dikes. On low coefficient of variation deposits, this type of graph
may also represent a normal distribution that has been plotted
on lognormal probability paper. The data should be examined to
determine the source of the low-grade material and to determine
whether that population has been or can be mapped geologically where N is the number of pairs at distance h, and h is the distance
and estimated separately. between the samples.
 ORE RESERVE/RESOURCE ESTIMATION 349
Cutoff Grade

Fig. 5.6.3. Deviation from a lognormal distribu-

tion that is caused by excess low-grade
samples.

Percent Samples Above Cutoff

Fig. 5.6.4. Deviation from a lognormal distribu-

tion that is caused by excess high-grade
samples.

Percent Samples Above Cutoff

The variogram function γ (h) is computed for a number of rate sampling and assaying of ore is difficult, or where poor
different sample distances, to provide an experimental variogram sampling and assaying techniques are employed. High nugget
that typically looks like the graph in Fig. 5.6.5. The most impor- effects are found in many gold deposits because of random gold
tant features of the variogram are the nugget, range, and sill. nuggets that cause large grade changes over small distances.
The nugget value is identified as the y-intercept of the variogram Similar high nugget values are often found in molybdenum de-
curve and represents random and short-distance variability fac- posits; these are caused by small pockets of pure molybdenite in
tors such as sampling error, assaying error, and erratic mineral- a disseminated or stockwork mineralization.
ization. High nugget values are commonly found in ore bodies Small nugget values indicate an ideal situation reflecting
where short distance variability is extremely high, where accu- good sampling techniques and locally continuous mineralization.
350

Fig. 5.6.5. Typical experimental variogram plot.

A small nugget value on a variogram confirms that the assays

can be reliably used for geologic interpretation and resource
estimation. Low nugget values are typically found in many types
of deposits, including hypogene porphyry copper, iron ore, and Fig. 5.6.6. Angular and fixed-distance tolerance methods for select-
coal. High nugget values have also been found for each of these ing variogram pairs.
types of deposits so each deposit must be analyzed individually.
Most variograms increase in value from the nugget for some
distance and then level off to a constant value. This distance is
called the range of the variogram, and the variogram value is
called the sill. The range is equivalent to the geologist’s concept
of range of influence, that is, the distance beyond which samples
are not correlated with other samples and beyond which grade
trends should not be projected. The sill value is usually equal to
the sample variance. If the sill is higher or lower than the vari-
ance, zonal effects or multiple grade distributions are usually
indicated.
The slope and shape of the variogram often vary in different
directions, with the range increasing in the direction of greatest
continuity of the mineralization. This behavior is referred to as
a geometric anisotropy.

5.6.7.1 Computing an Experimental Variogram

Computing an experimental variogram from a set of ran-
domly spaced data involves finding pairs of data that are oriented
in the required direction, determining the distance between the
samples, then summing the squared differences of the grades.
Since the data are usually sparse, it is necessary to use a tolerance
when locating samples in the desired direction and to use a
distance increment to classify samples by distance. The direc-
tional tolerance is usually achieved with a window angle, or a
fixed distance, as shown in Fig. 5.6.6. These methods may be
combined and/or generalized into three dimensions as shown in
Fig. 5.6.7. The distance tolerance is a fixed distance increment
(cell size), selected so a reasonable number of samples fall in
Fig. 5.6.7. Composite and three-dimensional methods for selecting
each cell. Some guidelines to aid in computing experimental variogram pairs.
variograms are
1. Variograms must be computed within continuous zones of
mineralization. Do not cross contacts between different geologic
domains.
2. The maximum distance used should be less than one-half 3. The maximum search distance perpendicular to the direc-
the length of the mineralized zone in the direction of the vari- tion of the variogram must be less than one-half the range of the
ogram. variogram in the perpendicular direction.
 ORE RESERVE/RESOURCE ESTIMATION 351
0.70

Fig. 5.6.9. Experimental variogram modeled with a linear variogram

Fig. 5.6.8. Experimental variogram modeled with a spherical vario- model.
gram model.

4. The distance increment should be approximately equal to

the average spacing between samples in the direction of the
variogram.
5. At least 30 pairs of samples are required to compute a
valid variogram. More pairs produce a more stable variogram.
6. All samples must be the same size and should be obtained
by the same or similar sampling methods.
7. Data should be declustered before computing the vario-
gram. In particular, a few twin holes may give a misleading
impression of the nugget effect.
A model, or equation, is fitted to the experimental variogram
for further geostatistical evaluations such as kriging. The most
common variogram model is the spherical model shown in Fig.
5.6.8. This model has the equation

(5.6.7)

Fig. 5.6.10. Experimental variogram modeled with an exponential

(5.6.8) variogram model.

where Co is the nugget, C is the sill, and a is the range.

A spherical variogram model may be constructed graphi-
cally by drawing a horizontal line at the variogram value equal for that cell. The resulting variogram is known as a relative
to the variance of the samples. This value is equal to nugget plus variogram.
the sill C o + C. A line is drawn through the points at the
short-distance end of the curve. The nugget C o is estimated where
the line intersects the Y-axis, and the range a is estimated as 1.5
5.6.7.3 Lognormal Variograms
times the distance where the line intersects the variance. If data are clearly lognormal, a variogram may be computed
Other variogram models used in resource estimation include using the logarithms of sample grades. The resulting lognormal
the exponential, linear, hole effect, and various combinations of variogram is often less erratic and more easily interpreted than
“nested” structures. Examples of some of these variograms are the variogram of untransformed values. This variogram may be
shown in Figs. 5.6.9 to 5.6.11. used directly for lognormal geostatistics or may be transformed
to a relative variogram as follows:
5.6.7.2 Relative Variograms
(5.6.9)
Lognormally distributed data often exhibit a proportional
effect where the standard deviation of grades increases with (5.6.10)
grade. This results in variograms with higher values in high-
grade areas than in low-grade areas. This may be corrected by Caution must be exercised when using the lognormal vario-
dividing each cell in the experimental variogram by the square gram since small deviations from lognormality may have large
of the mean of the samples that were used in the variogram effects on the transformation to a relative variogram.
352

Fig. 5.6.11. Experimental variogram modeled with nested spherical

and linear variogram models.
Hole Grade Area

1 0.12 39.4 Total Area = 333.7

5.6.8 RESOURCE ESTIMATION (MODELING) 2 0.21 37.6
3 0.17 42.0 Total Area x Grade = 71.39
Methods for resource estimation or modeling are generally 4 0.50 37.7
divided into (1) traditional, geometric methods that are done 5 0.33 33.8 Average Grade = 0.2139
manually on plans or sections and (2) interpolation methods 6 0.05 50.1
7 0.26 46.8
such as inverse-distance-weighting and kriging that require the 8 0.15 46.3
use of a computer.
Fig. 5.6.12. Computation of an estimate using the polygonal method.
5.6.8.1 Geometric Methods
Manual resource estimations are usually done on plan maps
or cross-section maps that cut the deposit into sets of parallel
slices. Data plotted on the maps include drillhole locations, assay Polygonal and Cross-sectional Methods: Polygonal and
values, and the geologic interpretation of the mineralization con- cross-sectional methods are related methods in that each ore
trols. The geometric methods used are based on geometric interval is assigned its own polygon of influence. Tonnage and
weighting of assays and include area averaging, polygonal, cross grade is then computed using the same procedure as was used for
sectional, and triangular. the area-average method, except that the areas used to compute
Area Averaging: The area-averaging method is among the tonnage are the area of each individual polygon. Polygons are
simplest of all reserve estimation methods, involving only a geo- drawn on plan maps based on the perpendicular bisectors of the
logic interpretation of the shape of the ore and averaging of the line between each drillhole as shown in Fig. 5.6.12. The size and
grades within that shape as follows: shape of the polygons may be limited, if desired, by a maximum
1. Draw the outline of the ore body on each map; these are distance from each hole. On cross sections, the polygons are
the ore blocks and may be regular or irregular shapes. If several usually drawn one-half the distance from each drillhole as shown
ore zones or ore types are present, each is drawn individually. in Fig. 5.6.13. The distance from a drillhole may also be limited
2. Measure the area of each ore block (usually by planime- to a maximum distance in the cross-sectional method.
tering). Multiply the area times the thickness of the ore and A computer approximation of the polygonal method is the
divide the resulting volume (cubic feet) by the tonnage factor nearest neighbor estimation. This method requires superposition
(cubic feet per ton) to compute tons of ore; in SI units, multiply of a rectangular grid of blocks over the drilled area as shown in
the volume (cubic meters) by the density (tonnes per cubic meter) Fig. 5.6.14. The grade of the nearest sample is then assigned to
to compute the tonnes of ore. each block. This method will closely approximate the polygonal
3. Compute the average grade of samples within each block. method if the block size is no more than 25% of the average
4. Calculate the sum of the tonnage in the individual blocks. drillhole spacing.
Average grade is the tonnage-weighted average grade of the Triangular Method: The triangular method is similar to the
individual blocks. polygonal method except that areas of triangles are estimated,
Despite its simplicity, the area-averaging method provides and the grade of each triangle is based on the average of the
excellent estimates where the drilling pattern is uniform, grades grades at each of the corners of the triangles as shown in Fig.
are continuous, and ore boundaries are distinct and sharp. Prob- 5.6.15.
lems may arise, however, when the drill pattern is nonuniform. The geometric methods all have the advantage of simplicity
With a nonuniform drill pattern, a cluster of holes in a high- and ease of implementation. In addition, they will provide an
grade zone will cause overestimation of grade. Area-averaging unbiased estimate of the average grade of a deposit at a zero
methods also may be difficult to implement on deposits with cutoff grade. A resource estimate using a geometric method
discontinuous or spotty ore zones, especially if the ore contacts provides a quick, inexpensive check to verify nonbias of a more
are gradational, and multiple cutoff grades are desired. complicated, computer-generated resource model.
 ORE RESERVE/RESOURCE ESTIMATION 353

Triangle Grade Area

Fig. 5.6.13. Computation of an estimate using the cross-sectional

method.

Fig. 5.6.15. Computation of an estimate using the triangular method.

some higher-grade material. The resulting mined grades are dif-

ferent from the predicted distribution; for cutoff grades below
the average grade of the deposit, the mined grade will be lower
and the tons will be higher. If the cutoff grade is significantly
higher than the average grade of the deposit, however, both
the mined grade and tons can be lower, resulting in a severe
overestimation of contained metal.

5.6.8.2 Moving Average Methods

The moving average methods, inverse-distance weighting and
kriging, are the most widely used procedures for computer-
assisted resource estimation. The basic procedure for both of
these methods is as follows
1. Divide the ore body into a matrix of rectangular blocks
Hole Grade # Blocks as shown in Fig. 5.6.16.
2. If geologic controls are present and will be used to control
Total Blocks = 114 or modify grade assignment, a geologic code must be assigned
to each block.
Sum Blocks × Grade = 24.21 3. Estimate the grade of each block by searching the database
for the samples surrounding each block and computing the
Average Grade = 0.2124 weighted average of those samples. The weighted average is
computed using the following equation:

(5.6.11)
Fig. 5.6.14. Computation of an estimate using the nearest-neighbor
method. where g* is the estimated grade, gi is the grade of sample i, wi is
the weight given to sample i, and n is the number of samples
selected.
The most common problem with geometric methods is that
they may imply more selective mining than may be achieved by 5.6.8.3 Practical Considerations for Moving
the mining method. This results from estimating the resource Average Methods
from samples the size of a drillhole but mining larger, less selec-
tive volumes. High-grade blocks usually include lower-grade ma- The determination of the block size, anisotropies, and the
terial when they are mined, and low-grade blocks usually include sample selection criteria are common considerations for either
 MINING ENGINEERING HANDBOOK

Fig. 5.6.16. Example of a geologic unit coded into a

block model.

Block Model
Geologic Interpretation
Representation

kriging or inverse distance estimation. These factors are often

the most critical factors in developing a satisfactory resource
model since the geologic controls are introduced via these mecha-
nisms. Poor selection of these parameters commonly results in
an unsatisfactory resource model with significant overestimation
or underestimation of ore tonnage or grade.
Block Size: Factors that must be considered in determining
block size include the size of the resulting model, drillhole spac-
ing, mining method, and geologic controls. General rules for
block size determination are
1. The largest block size possible should be used to minimize
the size of the block model and reduce computational time and
disk storage requirements.
2. The block is normally one-half to one-fourth the average
drillhole spacing. A bigger block may be used if the drilling is
extremely dense. Smaller block sizes provide minimal improve-
ment in the estimation unless strong geologic controls are Fig. 5.6.17. Use of an oriented search ellipse to approximate the
present. structure of an ore zone.
3. The block size must be at least one-half the size of the
smallest geologic feature that will be modeled. Larger blocks will 2. The maximum radius should be at least equal to the
destroy the location and/or size of small features. distance between samples to prevent discontinuities in the
4. Block size may be related to a proposed mining method. weighted average as samples drop in and out. (For a square grid,
The block height is usually the same as bench height in a deposit the maximum radius is the diagonal.)
that will be mined by open pit methods. 3. The maximum number of samples is usually on the order
5. Most commercial software packages for resource estima- of 10 to 20. More than 20 samples rarely improves the estimate;
tion allow each dimension of the block to be a different size and fewer than 10 samples may cause discontinuities in the estimated
may allow rotation of the entire model to an orientation other grades.
than north-south. 4. A minimum distance to the nearest sample may be used
6. These rules are often contradictory, and the best solution to prevent excessive extrapolation.
will be a compromise that will vary on a case-by-case basis. 5. A search ellipse or other anisotropic pattern may be used
Anisotropies: Strong anisotropies, or trends, are often ob- to align the search with trends in the ore as shown in Fig. 5.6.17.
served in mineral deposits that have ore zones with greater conti- The axes of the search ellipse should be oriented parallel to grade
nuity in favorable orientations, commonly shown by vario- trends. The length of the ellipse axes should be proportional to
grams with longer ranges in certain directions. the range of continuity in the respective directions. The vario-
Sample Selection: The purpose of the sample selection step gram ranges and visual appraisal of the grade zones on plans
is to provide a subset of the data that is representative of the and sections are both used as guides to determining the orienta-
region around the block. Weighted moving average methods tion and length of the search axes.
may be very sensitive to sample selection and time spent in 6. Three composites are usually the maximum required from
analyzing the sample search pattern and how it relates to the a single drillhole. More than three provides redundant data and
data is essential. Some rules to assist in defining a sample selec- may cause strange kriging weights.
tion search are 7. Search patterns may be modified to select data with quad-
1. Samples must be selected from geologic domains similar rants or other geometric limit as shown in Fig. 5.6.18. Use of a
to that of the block. quadrant search will improve estimations if data are clustered.
 ORE RESERVE/RESOURCE ESTIMATION 355
5.6.8.5 Kriging
Kriging is the geostatistical estimation method developed to
provide the “best linear, unbiased estimate” for grade based on
a least squares minimization of the error of estimation, or kriging
error. Kriging and its variants have had much theoretical devel-
opment and are well described in the geostatistical literature.
For a more detailed discussion of kriging, the reader may refer to
David (1977, 1988) or Journel and Huijbregts (1978). Important
factors in the kriging estimate are
1. The average of the estimates should not be systematically
higher or lower than the true value; this is established mathemat-
ically by setting the sum of weights equal to zero, that is,

2. The error of estimation σΚ2 , which is expressed as variance

(G – G*), or

(5.6.14)
Fig. 5.6.18. Selection of data using a quadrant search and an ellip-
tical search pattern.
is minimized, where G is the actual grade, G* is the estimated
grade, σB , D2 is the variance of blocks the size being estimated in
the deposit, the σB , X i are the covariances between each sample
8. Sketch the search pattern on plan maps and/or cross- and the block being estimated, and the σ Xi,Xj are the covariances
sectional maps in both well-drilled and sparsely drilled areas. between the individual samples.
This will aid in visualization and assure that the search pattern The kriging variance in Eq. 5.6.14 is minimized using the
is appropriate. Lagrange principle to create a modified equation that satisfies
the nonbias constraint in Eq. 5.6.13 as follows:
5.6.8.4 Inverse Distance Estimation
Inverse distance weighting, one of the earliest used interpola-
tion methods, is based on an empirical observation that the (5.6.15)
weight of each sample in Eq. 5.6.11 is proportional to an inverse
power of the distance from the location of the estimate to the where µ is the Lagrange multiplier. This equation is differenti-
sample. The inverse distance estimate is thus a weighted average ated with respect to each of the wi and µ, resulting in a set
with the individual weights computed as an inverse power of of simultaneous equations that may be solved for wi and µ as
distance as follows: follows:

number of samples (5.6.12)

where wi is the weight computed for each sample i, each d i is the

distance between the location being estimated and sample i, and
po w e r
is the inverse distance weighting power. Care must be taken
with an inverse distance estimate to ensure that none of the
distances d i are very small or equal to zero, resulting in division
by zero, or floating point overflow. This problem may be allevi- (5.6.16)
ated either by adding a small constant to each distance or by
assigning the value of the closest point to the estimate if the The foregoing system of equations (5.6.16) is usually solved
distance is less than some threshold. using simple gaussian elimination to determine the weights and
The degree of smoothing, or variance reduction, in the in- the Lagrange multiplier µ. The kriging error of estimation is
verse distance estimates may be controlled by changing the
then computed based on the following relation:
weighting power and search parameters. A lower power, large
search radius, and/or greater number of points used in the esti-
mation result in more continuous estimations and a greater re- (5.6.17)
duction in the variance of the estimated values. Higher weighting
powers, smaller search radius, and/or fewer points result in less which is based on the equation for error of estimation (5.6.14)
reduction in the variance of the estimated values and more erratic and the relationship between µ and the kriging equations
appearance of the estimated values. The appropriate parameters (5.6.16),
may only be found through trial and error and experience to
achieve the desired distribution of the estimated values and (5.6.18)
trends, or contours, to match with production results, or to
produce trends in the estimated values that visually match with The individual covariances and block variances are com-
the empirical expectations of the geologist. puted from the experimental variogram as follows:
356 MINING ENGINEERING HANDBOOK
(5.6.19) ing with exogenic drift. A method of universal kriging that uses
the geologists’ interpretation of grade-zone trends as the exogenic
(5.6.20) drift is zoned kriging. Cokriging is the method of kriging that
accounts for the correlation of a primary variable with a second-
(5.6.21) ary variable, for example, gold with silver or molybdenum with
copper, etc. When cokriging is used with qualitative secondary
where σS , D 2 is the variance of samples in the deposit, γ (Xi,Xj ) is variables such as alteration, rock type, or other geologic features,
the value of the variogram function between samples Xi and Xj, it is known as soft kriging. Disjunctive kriging is a method used
AVE (γ (B,Xi )) is the average value of the variogram between the which attempts to estimate not only the local block grade but
block and sample Xi, and AVE(γ (B,B)) is the average value of also the shape of the tonnage-grade distribution within the block.
the variogram between all points within the block.
Lognormal Kriging: Lognormal kriging is a method of non-
linear kriging that was developed to improve estimation when
5.6.8.6 Volume-variance Effects and Recovery
the underlying data are distributed according to a lognormal Functions
probability distribution. The basics of lognormal kriging include:
The volume-variance effect refers to the inverse relationship
(1) the variogram is computed using the natural logs of the data,
between the distribution variance and the volume of blocks. The
(2) the kriging system is solved to provide a weighted average of
volume-variance effect is characterized by Krige’s relationship
the natural logs of the data, and (3) the kriged log average is
as follows
then transformed back to normal values using lognormal trans-
formation similar to that shown earlier in Eq. 5.6.1. The mathe-
matics of lognormal kriging are complex and are discussed in (5.6.23)
Rendu (1978) and Journel (1978, 1980).
Complications in the practical application of lognormal krig- where σ 2B , D is the variance of blocks in the deposit, S2 S,D is the
ing are many, including a strict requirement for a lognormal variance of samples in the deposit, and S2 S,B is the variance of
distribution and a variogram which is stationary over the field samples in the block. The variance of samples in the block may
of estimation. Serious local and global biases may occur if either be estimated from the variogram as follows
of these conditions are not met. In addition, there is a tendency
for lognormal kriging to overestimate the high-grade end of the (5.6.24)
population when the coefficient of variation is greater than 2.0.
Lognormal kriging is recommended only for special purposes where γ ( S, B ) is the average of the variogram for samples within
where the results can be monitored closely and adjusted to pre- a block with the size and orientation of the mining block.
vent biases. The volume-variance relationship is unimportant where the
Indicator/Probability Kriging: Indicator kriging and proba- entire deposit is above the cutoff grade or where the ore is mined
bility kriging are related methods that are used to improve esti- nonselectively. Generally, however, the cutoff grade is higher,
mation when ore zones are erratic and grade distributions are and only a portion of the mineralized grade distribution is selec-
highly variable and complex. Advantages of indicator kriging tively mined as ore. The shape of the grade-tonnage distribution,
include less smoothing of estimated grades than ordinary kriging as defined by the distribution variance, is then a critical factor
and robustness in handling nonstandard grade distributions. in determining the grade and tonnage above cutoff.
The first step in indicator kriging is to set one or more cutoffs For practical resource estimation purposes, the variance of
with which to define indicator variables. Given a cutoff gc, the mining blocks is generally larger than the variance of kriged
indicator variable is set to 1 if the grade is above gc or 0 if the resource estimation blocks. The variance of mining blocks is
grade is below gc (the order of the { 1,0 } coding may be reversed); generally smaller than the variance of resource estimation blocks
indicator variables are coded similarly for each desired cutoff. for polygonal estimation.
Variograms are modeled for each indicator variable and an ex- Polygonal estimation underestimates tons and overestimates
pected value for each indicator is estimated using ordinary krig- grade for low cutoffs. At higher cutoffs, tonnage and grade
ing and the appropriate indicator variogram. are both overestimated. Kriging tends to overestimate tons and
The resulting indicator estimates, which may be interpreted underestimate grade for low cutoffs. At higher cutoffs, tonnage
as either the probability that the block will be above the cutoff and grade are both underestimated.
or the percentage of the block that is above cutoff, are used to For polygonal estimation, the difference between estimated
estimate the grade of the block as follows and mined reserves is usually handled with dilution factors where
a fixed tonnage is added with a grade that is less than the cutoff.
(5.6.22) These dilution factors are adequate for correction of overall
reserves but are not accurate for smaller areas if local grades
where each I*j is the estimate for the indicator for cutoff j, gj is vary significantly from the average grade. Caution must also be
the estimated grade for the interval j to j+1, and n is the number observed since dilution factors will vary according to the cutoff
of indicator cutoffs. The interval grades gj are usually estimated grade, the population variance, and the amount of variance re-
as the average of the cutoff grades for the interval, or, if the duction between the polygonal and mine block distributions. It
interval is large, may be estimated from the kriged grade of those should be noted that polygonal reserve estimates may require
data in the interval j to j+1. The prior method is more precise dilution factors for both volume-variance effects and contact-
when a large number of indicator cutoffs are defined; the latter mining geometric effects.
is most often used for a single cutoff. Kriging reserves are corrected for volume-variance effects
Other Types of Kriging: Other types of kriging that are according to the distribution of mining blocks within the reserve
not widely used include universal kriging, cokriging, disjunctive block as (1) the variance and distribution of mining blocks within
kriging, and soft kriging. Universal kriging is a method to incor- the reserve block is estimated, and (2) the tonnage and grade
porate trends into the kriging equations. If the trends are defined above cutoff is estimated for the block. The mining block distri-
according to a secondary variable, it is known as universal krig- bution parameters are most effectively determined by compiling
 ORE RESERVE/RESOURCE ESTIMATION 357
to estimate minable blocks; the character of the ore boundaries
determines how grade will be estimated at the borders between
different grade zones; and the available time and money deter-
mine the detail and effort that will be expended on the estimate.
Considerations for selection of a resource estimation method are
summarized in Table 5.6.1.
Cost: Simple, manual methods such as polygonal and cross-
sectional estimations are the cheapest and quickest methods for
estimation of resources when the quantity of data is small. This
is usually the case for preliminary evaluations in exploration
stages. As the number of data increase and a more detailed
estimate is desired, computer-assisted methods should be used
in order to save time and money. The least expensive computer-
Fig. 5.6.19. Ore dilution and losses caused by mismatch between assisted methods are automated polygonal or nearest-neighbor
the mining geometry and the ore geometry. methods and the most expensive methods involve extensive defi-
nition of geologic controls in conjunction with the more complex
geostatistical methods. For further discussion of computer appli-
production statistics of the grade-tonnage curves, or recovery cations to ore estimation, see Chapter 8.4.
curves, for several grade ranges of estimated blocks. Alterna- Ore Boundaries: The appropriate reserve and dilution esti-
tively, a lognormal distribution may be assumed for mining mation method is determined by the character of the ore/waste
blocks within reserve blocks. The variance of the distribution contacts. Sharp, simple boundaries are modeled with linear out-
of mining blocks within reserve blocks may be estimated from lines defining discrete mineral zones. Individual estimations are
production data. If production data are not available, the vari- made for each mineral zone; dilution is estimated based on the
ance of mining block may be approximated by intersection between the shape of the mineral zones and the
shape defined by the geometry of a mining method. A sharp,
(5.6.25) irregular boundary is also described with linear boundaries defin-
ing mineral zones; the actual ore-waste contact is much more
where σ2 b, B is the variance of mining blocks in the estimated irregular than the interpreted boundary, and dilution must be
reserve block, σ2B, D is the variance of reserve blocks in the de- increased accordingly. Geometric methods are usually appro-
posit, σ2 S,B is the variance Of samples in the reserve block, σ 2 k , B priate for ore bodies with sharp contacts, although kriging or
is the estimation (kriging) variance of reserve blocks, and σ2 k , b inverse-distance methods may be used within the zones if sup-
is the estimation variance of mining blocks (based on grade- ported by sufficient data.
control samples). Gradational boundaries are handled as transitional between
different mineral zones; kriging or inverse-distance methods are
most appropriate to model ore bodies with gradational contacts.
5.6.8.7 Dilution and Mining Losses Sufficient dilution for a gradational contact is usually incorpo-
The estimated tonnage and grade must be adjusted for dilu- rated in the modeling method.
tion of grade and losses of tonnage that occur in the course of Extremely erratic, irregular boundaries are difficult to define
mining. Dilution is waste that is not segregated from ore during accurately and are most appropriately estimated using methods
mining, thus decreasing the grade of the ore and increasing the such as indicator kriging.
tons. Ore losses are due to the inability of the mining method to Deposit Geometry: Simple geometry is often found in tabu-
follow accurately and to segregate small isolated pods and small lar, stratabound deposits, veins, and structural zones. The geom-
irregular offshoots from the main ore body. Dilution is most etry of these deposits is easily described using two-dimensional
significant in deposits with sharp contacts between high-grade methods such as contouring of thickness and elevation. Few
ore and barren waste and least significant in deposits with grada- additional controls are required other than boundaries to limit
tional contacts between ore and waste. the lateral extent of the mineral zones.
Dilution tonnage is estimated according to the quantity of Deposits with moderately complex geometry include both
waste mined with the ore based on the mismatch between ore deposits with simple geometry that have been moderately folded
body and mining geometry, overbreak in blasting, or lack of or faulted and deposits with large, simple, massive shapes such
accurate location of the ore/waste contact as shown in Fig. as porphyry copper and molybdenum. Definition of deposit ge-
5.6.19. Care must be taken in estimating dilution that the actual ometry includes definition of fold axes, fault boundaries, and
ore/waste contact is not more irregular than the model since zoning of trends within the deposit. While these controls are not
dilution will be underestimated as shown in Fig. 5.6.20. Dilution usually difficult to define, their definition is necessary to provide
grade is estimated as the grade of the waste at the ore/waste accurate resource estimates.
contact. Mining losses and grades are estimated according to Deposits with very complex geometry are usually associated
similar procedures. with structural deformation and are folded, faulted, stretched,
and twisted to form extremely discontinuous shapes that are
difficult to describe and model. Multiple ore controls such as a
5.6.8.8 Selection of Resource Estimation Methods
combination of stratigraphic and structural controls or multiple,
Selection of an appropriate resource estimation method de- overlapping pulses of mineralization also commonly form very
pends on the geometry of the deposit, the variability of the grade complex shapes. Definition of deposit geometry requires detailed
distribution, the character of the ore boundaries, and the amount examination of structural geology and ore controls to provide
of time and money available to make the estimate. Deposit geom- cross sections or plan maps which define the shape and location
etry determines the amount of detail that must be interpreted of mineral zones. These sections or maps may then be used
and input to the reserve estimation; the variability of the grade directly for manual resource estimation or may be digitized to
distribution determines the amount of smoothing that is required provide control for a computer block model or three-dimensional
358 MINING ENGINEERING HANDBOOK
Overbreak

Dilution

Stope Design Line

Fig. 5.6.20. Underestimation of dilution because of an irreg-

ular ore-waste contact.

wire frame model. Deposits with complex geometry are prone to Clark, I., 1979, Practical Geostatistics, Elsevier, Amsterdam, 129 pp.
large estimation errors due to misinterpretation of deposit geome- David, M., 1977, Geostatistical Ore Reserve Estimation, Elsevier, Amster-
try and ore controls; order of magnitude errors are common. dam, 364 pp.
Grade Variability: Deposits with low variability may be esti- David, M., 1988, Handbook of Applied Advanced Geostatistical Ore Re-
serve Estimation, Elsevier, Amsterdam, 216 pp.
mated with many methods. Common methods include automatic
Davis, J.C., 1973, Statistics and Data Analysis in Geology, Wiley, New
contouring and polygonal methods with cross-sectional estima-
York, 550 pp.
tion or area-averaging techniques for more complex geometry. Guarascio, M., David, M. and Huijbregts, C., 1976, Advanced Geostatis-
Weighted averaging methods (kriging and inverse distance) tics in the Mining Industry, Reidel, Dordrecht, Holland, 491 pp.
are most commonly used for estimation of deposits with moder- Journel, A., 1980, “The Lognormal Approach to Predicting Local Distri-
ate variability although polygonal or cross-sectional methods are bution of Selective Mining Unit Grades,” MathematicaI Geology,
also used. Weighted averaging methods may require recovery Vol. 12, No. 4, pp. 285-303.
functions and polygonal methods may require dilution to com- Journel, A. and Huijbregts, C.J., 1978, Mining Geostatistics, Academic
pensate for volume-variance effects, although in most cases the Press, London, 600 pp.
adjustments are small, on the order of 5 to 15%. King, H.F., McMahon, D. W., and Butjtor, G.J., 1982. A Guide to the
Weighted averaging methods are most commonly used for Understanding of Ore Reserves Estimation, Supplement to Proceed-
estimation of deposits with high variability. Other appropriate ings No. 281, Australasian Institute of Mining and Metallurgy,
methods may include indicator kriging, polygonal, and cross- Parkville, Vie., 2 1 pp.
sectional methods. Volume-variance effects are often large with Rendu, J.-M., 1978, An Introduction to Geostatistical Methods of Mineral
these deposits and must be compensated for with recovery func- Evaluation, Monograph, South African Institute of Mining and
tions for weighted averaging methods and large dilution of polyg- Metallurgy, Johannesburg, 1978, 100 pp.
onal and cross-sectional reserves. For coefficients of variation Rendu, J.-M., and Readdy, L. 1982, “Geology and the Semi-
above 2.0 or 3.0, local grade estimates are extremely difficult Variogram—A Critical Relationship.“, Proceedings, 17th Sympo-
sium on Applications of Computers and Operations Research in the
and must be tempered with judgment and caution.
Mineral Industry, AIME, New York, pp. 771-783.
Sinclair, A.J., 1976, “Applications of Probability Graphs in Mineral
REFERENCES Exploration,” Special Volume No. 4, Association of Exploration
Geochemists, 95 pp.
Barnes, M., 1980, Computer-Assisted Mineral Appraisal and Feasibility, Verly, G., et al., eds, 1984, Geostatistics for Natural Resources Character-
SME-AIME, New York, 167 pp. ization, Reidel, Dordrecht, Netherlands, Vols. 1-3.
 ORE RESERVE/RESOURCE ESTIMATION 359
Table 5.6.1. Selection of Estimation Method Based on Deposit Geometry and Variability
Low Variability, Moderate Variability, High Variability,
COV < 0.25 COV > 0.25 < 0.75 COV > 0.75

Simple Geometry

Deposit description Tabular, continuous grade and Tabular, large ore pods Tabular, small ore pods
thickness Moderately variable grade Highly variable grade
Flat or constant dip

Example deposits Evaporite Stratiform copper Gold veins

Sedimentary Iron Mississippi Valley lead Gold placers
Limestone Simple porphyry copper, molyb- New Mexico uranium
Coal denum Alluvial diamond

Estimation methods Grade and thickness using any 2- 2-dimensional methods. Inverse 2-dimensional methods. Inverse
dimensional method: polygonal, distance or kriging. Polygonal or distance or kriging with recovery
contouring, inverse distance, krig- cross section with 5 to 15% di- functions. Polygonal with 15 to
ing. Geometric controls for bound- lution 35% dilution
aries of ore zone, faults, and fold
axes

Moderately Complex Geometry

Deposit description Simple, bedded. Uniform grade Simple three-dimensional geome- Simple three-dimensional geome-
but erratic thickness, gentle fold- try. Moderately variable grade try. Two-dimensional with
ing, or simple faulting smaller, more erratic ore pods.
Simple folding, faulting

Example deposits Bauxite (variable thickness) Porphyry copper Stockwork and Carlin-type gold
Lateritic Nickel (variable Porphyry molybdenum Volcanogenic base metals
thickness)
Salt Dome

Estimation methods Estimate grade, thickness and el- Inverse distance or kriging with Inverse distance of kriging with re-
evation using any 2-dimensional external controls to define the covery functions. Polygonal or
method shape and grade trends. Polygo- cross section with 15 to 35% di-
Must define structural geology nal and cross-sectional methods lution
(faults, fold axes) may be used but will require dilu-
Variability of thickness may be dif- tion/volume variance correction
ficult to predict

Complex Geometry

Deposit description Otherwise simple deposits that Complex geometry due to fault- Deposits with extremely variable
have been severely folded and ing, folding, or multiple mineraliza- grade and highly contorted, com-
faulted tion controls. Moderately variable plex ore shapes. Typically little
grade continuity between individual ore
zones. General mineral envelope
definable but with 50% or less
ore

Example deposits Talc Tungsten skarns (folding/fault- Archean gold deposits

Gypsum (deformed) ing). Base metal skarns (erratic Roll front uranium
shape). Copper porphyry com-
bined with local skarns or replace-
ments (multiple controls)

Estimation methods Cross-section methods with de- Cross-sectional methods with de- Estimation very difficult. Size,
tailed definition of structural geol- tailed input to describe structural shape, and grade not locally pre-
ogy. Difficult to define geometry geology and or zones. Geostatisti- dictable. Cross-section, area-out-
for 3-dimensional block models cal methods may be appropriate line methods, indicator kriging ap-
and geostatistical methods but difficult to implement be- plicable.
cause of geometric complexity Errors of 50 to 100% typical. Ton-
nage often overestimated be-
cause of incorrect geologic
model