Exploration and Mining Geology, Vol. 18, Nos. 1–4, p.

13–23, 2009
© 2009 Canadian Institute of Mining, Metallurgy and Petroleum.
All rights reserved. Printed in Canada.
0964-1823/00 $17.00 + .00

Use of Twinned Drillholes in Mineral Resource Estimation

M.Z. Abzalov1,*,†
(Received January 13, 2009; accepted May 30, 2009)

Abstract — Drilling twinned holes is a traditional technique used for verification of intersections of
high-grade mineralization, testing of historic data, or confirmation of drillhole data during geological due
diligence studies. Twinned holes can also be used for special tasks such as correcting earlier data that are
recognized to be biased.
Successful implementation of the twinned-holes technique requires thorough planning. Experience
suggests that good practice is to drill twinned holes no more than 5 m apart. Many unsuccessful twinned-
holes programs could possibly have failed because the twinned holes have been drilled too far apart.
Twinned holes are best compared by mineralization intersections and, if the data and geological char-
acteristics of the deposit permit, by samples, equal-length composites, or geological units. Variables to be
verified by twinned holes should include the thicknesses of the geological units of interest (e.g., mineral-
ized thickness) as well as the presence of significant geological features (e.g., ore/mineralization contacts,
alteration, etc.).
A formal, rigorous analysis of twinned-hole data is essential. Repeatability of sampling, analytical
results, and bias must be analyzed and statistically quantified. The number of twinned holes required for
conclusive statistical and geostatistical analysis can be as high as 20–30, in particular where the studied
variables are characterized by high short-range (local) variability.
The efficiency of this approach is demonstrated by several examples of successfully applied twinned-
hole projects, including applications to orogenic gold, mineral sands, bauxite, and iron ore deposits.
© 2009 Canadian Institute of Mining, Metallurgy and Petroleum. All rights reserved.
Key Words: Twinned holes, QA, QC, Due Diligence.

Sommaire — Le jumelage de trous de forages est une technique traditionnellement utilisée pour la vérifi-
cation d’intersections minéralisées à haute teneur, la confirmation de données historiques ou la validation
de données de forage lors de l’exercice d’un mandat d’audit préalable. Le jumelage de trous peut égale-
ment être utilisé dans des buts particuliers, tel que la correction de données pré-existantes que l’on sait
être biaisées.
Une planification rigoureuse est nécessaire à un usage réussi de cette technique. Notre expérience sug-
gère que des forages jumelés ne devraient pas être distants de plus de cinq mètres. L’échec de plusieurs
programmes de forages jumelés pourrait avoir été causé par un écart trop grand entre les forages.
Il est préférable de comparer une intersection minéralisée à la fois dans les trous jumelés et, si les
données et les caractéristiques du gisement le permettent, une seule paire d’échantillons, de composites
de longueur égale ou d’unités géologiques. Les variables dont il faut tenir compte dans les trous jumelés
devraient comprendre l’épaisseur des unités géologiques d’intérêt (l’épaisseur minéralisée, par exemple)
ainsi que la présence d’éléments géologiques significatifs (contacts avec la minéralisation, ou zones d’al-
tération, par exemple).
Une analyse rigoureuse et formelle des données d’un programme de forages jumelés est essentielle.
La reproductibilité de l’échantillonnage, les résultats analytiques et le biais analytique doivent être quan-
tifiés et analysés statistiquement. Le nombre de trous jumelés nécessaire pour une analyse statistique et
géostatistique concluante peut être aussi élevé que 20–30, particulièrement quand les variables étudiées
présentent une forte variabilité à l’échelle locale.
L’efficacité de cette approche a été démontrée par de nombreux exemples de programmes de jumelage
de trous complétés avec succès; incluant des exemples provenant de dépôts d’or de type orogénique, de
placers minéraux, de gîtes de bauxite et de formations de fer rubanées. © 2009 Canadian Institute of
Mining, Metallurgy and Petroleum. All rights reserved.

1MASSA Geoservices, Mt. Claremont, Western Australia, Australia WA6010.

* Present address: Rio Tinto Exploration, PO Box 175, Western Australia, Australia WA6984.
† Corresponding Author: E-mail: [email protected]
14 Exploration and Mining Geology, Vol. 18, Nos. 1–4, p. 13–23, 2009

Introduction and the issue could be examined by twinning the historic

The drilling of twinned holes (i.e., drilling of a new hole air-core holes with sonic drillholes.
or “twin” next to an earlier drillhole) is a traditional tech- Verification of earlier drilled holes by twinning them
nique used in exploration and mining geology for verifica- with new holes is a common technique used by geological
tion of mineralization grades. The importance of this tech- due diligence teams when reviewing third-party projects
nique is emphasized in the JORC Code, which specifically (Gilfillan, 1998). Drilling twinned holes becomes particu-
queries if twinned holes have been used for “verification of larly important where the reviewed project is based on his-
sampling and assaying” (JORC Code, 2004, p. 15). How- torical data that were collected without rigorously applied
ever, the experience at various operating mines and in min- sampling quality control procedures, or where the relevant
ing project studies is that, despite its practical importance quality assurance/quality control (QA/QC) documentation
and reasonable simplicity, the twinned-holes approach is is lacking. Where bias of historic data has been proven,
often misunderstood and not applied to full advantage in the data from twinned holes can be used for quantification
the mining industry. of this bias, and integrated into the resource model using
One of the most common myths is that the drilling of multivariate and/or nonstationary geostatistical methods
twinned holes is inappropriate in geological environments for correcting the resource estimates (Abzalov and Pick-
characterized by a large short-range variability of grade. ers, 2005).
This and other prejudices have limited the wider applica- It is unlikely that the new resource or reserve estimates
tion of this useful technique. Such limitations seem to re- based on geostatistically corrected data will be accepted by
sult from a lack of a clear description of principles of the a formal regulator, such as JORC Committee, and/or finan-
method, and an absence of representative case studies in cial institutions. However, such studies can provide a good
the technical literature. For example, recently published quantification of the risk or upside potential of the project,
guides on evaluation of mineral deposits (Peters, 1987; and can demonstrate the economic benefits of additional
Annels, 1991; Vallée et al., 1992; Sinclair and Blackwell, infill drilling, or, in some cases, complete redrilling of the
2002) only briefly mention the twinned-holes approach deposit.
and do not give an in-depth treatment of its limitations,
The number and locations of twinned holes, the drilling
strengths, and weaknesses.
methods, and the tested variables are selected to accom-
This paper partially addresses this deficiency by briefly plish the objectives of the proposed study. For example, if
describing the basic principles of the twinned-holes tech- the purpose of drilling twinned holes is to verify high-grade
nique, and providing examples of its application. The ex- intersections, new holes should be drilled next to previous
amples described represent different stages in the appraisal holes that reportedly intersected high-grade mineraliza-
of mineral deposits, ranging from scoping to feasibility tion. Conversely, if the objective is to test and statistically
studies, and different types of deposit, including orogenic quantify a possible bias in historic data, the twinned holes
gold, mineral sands, bauxite, and banded iron formation should be drilled in such a manner that assures that a wide
iron oxide deposits. range of grades is tested, including both low- and high-
grade mineralization. In this latter case, twinned holes
Method Overview should be distributed throughout the entire deposit, and the
Twinned holes should be thoroughly planned, starting number of twinned holes should be sufficient for statistical
with a clear definition of the objectives of the study, and analysis of the data.
progressing to choices of the number of holes, their loca-
tion, studied variables, and statistical procedures for treat- Statistical Treatment of the Results
ment of the resulting data. Statistical procedures applied to twinned-hole data can
vary depending on the objective of the study. In general,
Objectives of the Twinned-Holes Study two main errors are assessed from the matching pairs of
The most common application of twinned holes is for twinned holes: repeatability (precision) and bias. Repeat-
verification of grades reported by previous drilling. Usu- ability of twinned holes can be quantified by the average
ally, this involves verification of high-grade intersections, coefficient of variation (CVAVR) of the paired data (Equa-
including an assessment of grade variability. tion 1). This has been proposed as the universal measure of
Another common objective of drilling twinned holes is relative precision error in mine geology applications (Stan-
to verify historic data. Such a need usually occurs where ley and Lawie, 2007; Abzalov, 2008).
results from earlier drillholes are causing concerns either
by being noticeably different from other drilling campaigns 2 N  (ai − bi ) 2 
in the same area, or where the poor quality of drilling has CVAVR (%) = 100 × ∑
N i = 1  (ai + bi ) 2 
( 1 )
been revealed from previously documented parameters,
such as core recovery. In some cases, historic data might be This estimator is useful for variables characterized by
suspect (e.g., biased) because of suboptimal drilling tech- strongly skewed statistical distributions and where outliers
nology or inappropriate drilling parameters. For example, are present (Abzalov, 2008). Examples of application of
air-core drilling of a mineral sands deposit might be biased, this technique to twinned-hole data are presented in Table 1.
 Use of Twinned Drillholes in Mineral Resource Estimation • M.Z. Abzalov 15

The average coefficient of variation

values for the studied twinned holes
Reference to Diagram
vary from 0.9% to 51%. They are usu-
ally larger for variables characterized

Figs. 6, 7
by large nugget effects.1
Fig. 2

Fig. 5

Fig. 3

Fig. 8
Bias of the data can be diagnosed
by the graphic tools including Relative

–37.4 + 8.6

–21.5 + 9.7
RMA Parameters (y = ax + b)

Difference Plot (RDP) and Reduced

–0.6 + 0.3
–0.4 + 0.4
–0.1 + 0.8

–0.1 + 0.3

–1.2 + 0.9

–1.0 + 0.4
–0.5 + 0.4

–2.5 + 1.8
–3.4 + 1.6
–4.5 + 4.7
–1.5 + 1.8
–0.3 + 0.4
1.1 + 0.7

0.1 + 1.1
Major Axis (RMA) diagrams (Pitard,
b

1998; Sinclair and Bentzen, 1998;

Abzalov, 2008), and then assessed by
statistical tests (Davis, 2002). A com-
1.09 + 0.03
1.07 + 0.05

1.06 + 0.09
1.01 + 0.1
1.1 + 0.07
1.0 + 0.05

1.3 + 0.07
1.6 + 0.15
1.7 + 0.27
1.2 + 0.09
1.4 + 0.19
1.3 + 0.16
1.6 + 0.32
1.1 + 0.09
1.1 + 0.1

0.8 + 0.1
mon and valid approach is the con-
a

ventional statistical t-test. Other useful

Abbreviations: CVAVR = Average Coefficient of Variation of the paired data (Equation 1); RMA = Reduced Major Axis (Sinclair and Bentzen, 1998); THM = Total Heavy Minerals.
diagnostic tools and statistical methods
of estimation of precision and accur-
CVAVR
25.1
15.1
8.9
30.3
29.7

51.2
1.4
23.8
10.2
3.2
19.0
15.9
1.9
9.7
10.0
25.1
acy errors can be found in Vallée et al.
(1992) and Abzalov (2008). When bias
in certain data sets has been diagnosed,
(# Matching Pairs) Old Holes New Holes Old Holes New Holes
Standard Deviation

3.3
2.5
1.6
5.7
2.5

15.0
1.8
0.8
1.6
3.5
1.0
4.4
4.4
7.0
1.9
2.3
it can be further studied by geostatis-
tical methods (Abzalov and Pickers,
2005).
3.0
2.3
1.5
5.0
2.4

11.6
1.1
0.5
1.4
2.5
0.6
3.3
4.0
6.3
1.8
2.9

Distance Between Twinned Holes

In general, twinned holes should be
7.8
7.8
7.7
11.9
4.76

8.9
58.5
1.47
4.1
51.3
4.8
9.7
49.9
15.1
4.3
7.6

located as close as possible to the ori-

Mean Values

ginal holes to minimize the effects of

short-range variability in the studied
variables. In many geological environ-
7.7
7.7
7.7
9.4
4.74

7.9
58.8
1.40
4.0
51.3
5.1
9.3
49.9
14.9
4.4
9.5

ments, in particular when studying

gold deposits, it is impossible to avoid
Compositing Samples

12-m composite (169)

Composited to bench
Composited to bench
Composited to bench
3-m composite (669)

3-m composite (98)

the nugget effect, and advance know-

1.5-m composites

ledge of that parameter is essential

Intersection

Intersection

Intersection
Intersection
Intersection
Intersection
Intersection
Intersection
Intersection

when comparing results of old and new

drilling in a pair of twinned holes. In
(201)

practice, most twinned holes, including

*Only part of the Weipa deposit and related twinned-holes data were included in this study.

those listed in the Table 1, are drilled

Compared

Thickness, m
Thickness, m
Variable

at distances varying from 1 to 10  m

Al2O3, %

Al2O3, %

Al2O3, %
THM, %
THM, %
THM, %
THM, %
THM, %

SiO2, %

SiO2, %

SiO2, %

from the original holes, depending on

Au, g/t
Fe,%

Fe,%

the logistics of the drill sites, deposit

type, mineralization style, and studied
# Twinned
Table 1. Characteristics of the Studied Twinned-Holes Projects

Holes

variables. A well-known example of a

35

8
20

5
40

26

13*

62

(negatively) successful twinned-holes

program is the confirmatory drilling
Diamond (NQ) vs. diamond (NQ)
Diamond (triple tube) vs. aircore

done by Freeport-McMoRan at the

Driling Type: New vs. Old

Busang project of Bre-X Minerals Ltd.

Fraud was revealed by seven twinned
Diamond (PQ) vs. RC
Sonic vs. vibracore

holes (Lawrence, 1997). Each of the

Diamond vs. auger
Aircore vs. aircore

Aircore vs. aircore

RC vs. blast holes

seven new holes was drilled only 1.5 m

away from supposedly extensively
gold-mineralized holes, and all failed
Twinned-Holes Project

Iron ore deposit, Australia

Iron ore deposit, Australia

Bauxite deposit, Australia

Ti-sands project, Mada-
Ti-sands project, Africa

Ti-sands project, Africa

Orogenic gold, Canada

Bauxite deposit, Africa

1 Nugget effect is a geostatistical term defining

“an apparent discontinuity in the experimental
variogram near the origin caused by measure-
ment errors or by nested structures that have
gascar

Notes

ranges smaller than the sampling interval, or

both.” (Olea, 1991, p. 54.)
16 Exploration and Mining Geology, Vol. 18, Nos. 1–4, p. 13–23, 2009

to substantiate the originally stated gold values (Lawrence, holes, and adjusting the drill mud and drilling rate to en-
1997). sure the best possible core recovery. Twinned holes require
more thorough monitoring by technical and geological per-
A study undertaken by Abzalov et al. (2007) at the Yandi
sonnel. In all cases, the sampling protocols and quality of
iron ore open pit has shown that CVAVR of Al2O3 grades
assays in the verification holes should be rigorously mon-
estimated from matching pairs of blast holes and reverse
itored and documented (Abzalov, 2008).
circulation (RC) holes increase from 23.8% to 35.7% when
the distance between twinned holes increases from 1 to 10 Comparison of Studied Variables
m. These results suggest that many unsuccessful twinned-
holes studies could have failed because the twinned holes Prior to comparing assayed grades in the twinned holes,
were drilled too far apart. it is necessary to compare geological logs, because the geo-
logical features are usually more continuous than grades
The close distances of 1 to 1.5 m employed for twinned (Sinclair, 1998). Twinned holes are initially compared by
holes at Busang (Lawrence, 1997) or to verify grades at mineralized intersections (Fig. 1a). Comparisons should
the Yandi mine (Abzalov et al., 2007) are likely to rep- not be limited to average grades of intersections, but should
resent extreme cases. At Busang, the confirmatory holes also include thicknesses and positioning of the contacts.
had to be as close as possible to the originals to satisfy The latter is particularly important when downhole con-
technical due diligence considerations. Similarly, the large tamination of the samples is suspected. Common examples
short-range variability caused by the presence of irregu- include the use of auger drilling for evaluation of bauxite
larly distributed clay pods at the Yandi mine required very or heavy mineral deposits, or RC drilling in gold projects.
close spacing of the twinned holes (Abzalov et al., 2007).
In general, particularly when the studied variable has less More detailed studies include comparison of twinned
spatial variation, the acceptable distance between twinned holes by matching intervals. Practically achievable out-
holes can be larger. However, personal experience of the comes are often obtained through grouping samples by
author indicates that good practice is not to drill matching geological units (Fig. 1b). This approach is particularly
pairs of holes more than 5 m apart. useful when twinned holes are drilled through stratified
mineralization or regularly intercalated low- and high-
Drilling Quality and Quantity grade zones.
When planning twinned drillholes, it is necessary to re- In particular cases, for example when twinned holes are
member that important decisions might have to be made drilled through mineralization exhibiting gradational zon-
based on a limited number of verification holes. Therefore, ing, comparison can be made directly between samples
it is necessary to ensure that their quantity and quality are where they are of an equal length; alternatively, compari-
sufficient to make a conclusive decision. The number of sons can be made between equal-length composites (Fig.
twinned holes depends on the aim of the twinning program 1c). Compositing of samples of small size into composites
and the range of variation in the studied variables. Where of a larger size smooths away the impact of outliers and
studied variables are characterized by large short-range converts samples into common length data, which are ne-
variability, the number of twinned holes should be suffi- cessary for geostatistical resource estimation (Pan, 1995;
cient for conclusive statistical and geostatistical analysis. Sinclair and Blackwell, 2002). In twinned-holes analysis,
In the author’s experience this can often be achieved by when comparing variables with high statistical variation,
drilling a relatively small number of holes. However, spe- grouping samples to larger composites is commonly ne-
cial studies that require 20 to 30 twinned holes are not un- cessary to minimize the noise exhibited by individual sam-
common (Table 1). ples. It is important to ensure that samples are composited
It is necessary to ensure that verification holes are of to geological boundaries.
the best practically achievable quality. This often requires An example of comparing twinned holes by composites
the use of a different drilling technique, commonly a more of different lengths is shown in Figure 2. In this project,
expensive one than that used for the previous drilling. For compositing from 3 to 12 m has significantly decreased
example, air-core holes in mineral sands projects are pref- data variance and scatter, and has decreased CVAVR from
erably twinned by sonic drilling, auger holes in bauxite 25.1% to 15.1% (Table 1, first example). Reduced Major
deposits by diamond core holes, and RC holes in coal de- Axis functions of two data sets of 3-m and 12-m compos-
posits by PQ diamond core holes. It is important to realize ites are very similar (Fig. 2a, Table 1). When the same data
that verification holes, even if they are drilled using the were composited to ore-grade intersections, CVAVR was fur-
most up-to-date technology, can still produce biased and/ ther decreased to 8.9% (Fig. 2b, Table 1). The RMA func-
or excessively variable results. For example, the quality of tion of these data coincides with the 1:1 line of the scatter
confirmatory diamond drilling can be marred by poor core diagram.
recovery, and sonic drilling can be adversely affected by
self-injection of sands into core barrels caused by litho-
static pressure. Case Studies
The risks of incorrect twinned-holes studies can best be Gold Deposits: Confirmation of High-Grade Intersections
minimized by optimizing drilling parameters, for example
using HQ or PQ drill bits for verification of NQ diamond Confirmation of high-grade intersections is essential in
 Use of Twinned Drillholes in Mineral Resource Estimation • M.Z. Abzalov 17

Fig. 1. Idealized diagram explaining various comparative approaches for twinned-holes data: a. Comparison by ore-grade intersections. b. Comparison
by geological units. c. Comparison by matching individual samples.

a b

Fig. 2. Twinned-holes study at a Ti-sands project, Africa (THM = total heavy minerals). a. Scatter diagram comparing matching pairs between twin holes.
Samples have been composited to 3-m (gray symbols) and 12-m intervals (black symbols); RMA = Reduced Major Axis. b. Scatter diagram comparing
the same drillholes by the average THM grades of the mineralized intersections.

deposit styles characterized by high nugget effect of metal grade shoot must be assessed and/or confirmed. The first
grades. It is particularly important where a deposit contains task can be achieved by twinning the original hole with a
erratically distributed high-grade shoots, for example some new hole drilled as close as possible to the original hole.
styles of gold deposits (e.g., Vallée et al., 1992; Abzalov, However, such closely located twinned holes do not allow
1999, 2007; Abzalov and Humphreys, 2002). proper assessment of the size of the shoot, and therefore do
In such cases, confirmation of high-grade intersections not fully accomplish the total objective for verification of
must include two objectives. First, the high-grade assays high-grade zones.
returned by a previously drilled hole must be verified, and, The second part of the objective requires drilling several
second, the geological significance and size of the high- holes at different distances from the first high-grade inter-
18 Exploration and Mining Geology, Vol. 18, Nos. 1–4, p. 13–23, 2009

cause the project included several drilling campaigns, the

project management elected to verify some of the earlier
drilled high-grade holes by twinning them with the new
holes (Table 1). Structural interpretations and close-spaced
drilling in the detailed study areas have shown that high-
grade shoots have 50- to 70-m strike lengths and widths of
approximately 25 m (Fig. 3a). Twinned holes were drilled
within 1 to 5 m, and in some cases up to 10 m, from the ori-
ginal holes. This was considered acceptable given the prior
knowledge of the structure of the gold lodes and dimen-
sions of the high-grade shoots (Fig. 3b). The twinned holes
overall confirmed the previously reported grades, although
significant variation (CVAVR = 51%) between twinned-hole
grades was revealed (Table 1). This level of variation was
considered acceptable because the deposit is characterized
by high short-range variability. The relative nugget effect
of the gold grade is approximately 40%.
The second case (Fig. 4) was observed at an orogenic
gold deposit in the Norseman-Kambalda greenstone belt,
Western Australia. The first phase of resource definition
drilling at this deposit delineated resources potentially vi-
able for economic extraction by a conventional open pit
mine. However, it was recognized that the main risk of in-
correct estimation of the resources was related to results
Fig. 3. Confirmation of high-grade intersections in an orogenic gold de-
posit, Canada, by twinned holes. a. Long section (looking north) showing for two high-grade shoots, in particular one in the north-
location of the twin holes; b. Cross section (looking west) showing twin ern part of the deposit where the length-weighted average
holes drilled through high-grade gold shoot. Numbers denote the sample grade of a key drillhole intersection totalled 51.4 g/t. To
grades (Au, g/t). mitigate this risk, additional drilling was undertaken. This
high-grade zone was first confirmed by drilling a twin hole,
which returned 41.9 g/t (Fig. 4), and after that, the exten-
sion of the high-grade shoot was studied by drilling several
holes along the strike of mineralization.
Delineation holes, drilled at 15- to 20-m intervals, show
that the high-grade shoot extends less than 30 m along
strike and is approximately half of this size in the across-
strike direction (Fig. 4). When these findings were incor-
porated into the revised resource model, it caused signifi-
cant changes in the open pit design, production schedule,
and project economics.

Twinned-Holes Studies in Iron Ore Deposits

During evaluation of an iron ore deposit in the Pilbara
region of Western Australia, the quality of RC holes was
tested by diamond holes of PQ core diameter. Results pre-
sented in Table 1 show that PQ holes have verified the RC
Fig. 4. Delineation by closely spaced step-out drilling of the high-grade
zones at an orogenic gold deposit, Norseman-Kambalda greenstone belt,
drilling. The small differences between Al2O3 grades of
Western Australia. the PQ core samples and the corresponding RC samples
were statistically insignificant, so the results from the RC
drilling technique were accepted for the project feasibility
section. These holes are not twinned holes senso stricto study.
and represent a related technique, namely delineation of Another twinned-holes study was undertaken at the
the high-grade zones by closely spaced step-out drilling. Yandi open pit iron ore mine located in the eastern part of
The difference between these cases is shown by Figures the Pilbara region (Abzalov et al., 2007). The grade con-
3 and 4. trol approach used at the mine was based on sampling of
The first case (Fig. 3) is represented by an orogenic production blast holes on a 6 × 6-m spacing. An alternative
gold deposit in Canada. The presence of high-grade zones approach of using RC holes distributed on a 25 × 25-m
was recognized by the project team, and were partly de- grid was considered. In order to assess representativity of
lineated by closely spaced infill drilling. However, be- the blast-hole samples, they were initially compared with
 Use of Twinned Drillholes in Mineral Resource Estimation • M.Z. Abzalov 19

full blast-hole cone grades, and with RC samples com- RC holes were composited to bench heights and com-
posited to bench heights. In total, 6 blast-hole cones were pared with the blast-hole grades. Results show that average
processed. Comparisons revealed that blast-hole sample blast-hole grades are consistent with the RC (and/or dia-
grades of Al2O3 and SiO2 were slightly higher than their mond) grades when the distance between them is less then
values in the full-cone assays, whereas the Fe grade was 1 m (Table 1). When the distance increases, the variations
lower (Table 2). Comparison of the RC composites with between grades of the matching pairs of holes also increas-
full-cone assays showed a similar trend; however, the mag- es. Coefficients of variation of Al2O3 grade increased from
nitude of the differences was larger than that of the blast- 23.8% at 1-m separation, to 30.1% at 5 m, and 35.7% at 10
hole samples (Table 2). m. Based on these results, it was concluded that use of RC
The study was continued by drilling additional RC holes drilling for grade control at this mine would not necessarily
that twinned blast holes. In total, 12 pairs of RC and blast provide better results than blast-hole samples, and, there-
holes were obtained. The new results showed similar rela- fore, the grade control procedures remained unchanged
tionships to those presented in Table 2, but with larger vari- (Abzalov et al., 2007).
ability of the results. In order to investigate the causes of
excessive variability between the twinned holes, all avail- Mineral Sands: Validation of Historic Drilling
able pairs of RC (and/or diamond) holes and blast holes Air-core drilling is a common technique used for re-
separated by a distance less than 1 m were recovered from source definition of heavy mineral sands deposits of ilmen-
the deposit drillholes database. ite, rutile, zircon, and leucoxene (Abzalov and Mazzoni,
2004). However, it is the author’s experience that in some
Table 2. Comparison of Six Full-cone Grades With Corresponding Blast- cases, depending on the mineralogy of the deposit, depth
hole and Composited RC Samples, Yandi Iron Mine, Australia of water table, clay content of the sands, and other fac-
tors, air-core drilling can generate biased results by sys-
Fe, wt% SiO2, wt% Al2O3, wt% tematically underestimating the THM (total heavy min-
Full cone 58.18 6.38 1.14 erals) grades of the mineral sands. The potential bias can
Blast-hole sample 58.59 6.16 1.06 be tested and quantified by drilling twin holes, historically
using triple-tube diamond core drilling and, more recently,
RC sample 59.014 5.80 1.02
by sonic drilling.
Note:
Abzalov et al., 2007. A case study for a Ti-sands project in Africa is presented

a b

Fig. 5. Twinned-holes approach applied to verify historic data

at a Ti-sands project in Africa (THM = total heavy minerals).
a. Air-core hole twinned by triple-tube diamond drillhole; hori-
zontal distance (<5 m) between twinned holes is not to scale.
b. Scatter diagram of the THM grades of diamond drillholes
versus air-core holes (RMA = Reduced Major Axis).
20 Exploration and Mining Geology, Vol. 18, Nos. 1–4, p. 13–23, 2009

in Figure 5 and summarized in Table 1. Mineralization is tacts of the bauxite. The latter is necessary to assure good
distributed between several stratigraphic units with a com- control of mining losses and dilution, because many baux-
bined thickness exceeding 100 m. Historic air-core holes ite orebodies are narrow, with an average thickness of less
were tested by twinning them with triple-tube diamond than 3 m.
core holes (Fig. 5a). All assays were composited to 3-m When twinned holes are used as part of the routine
intervals. The matching pairs of samples are plotted on a QAQC procedures, the second hole is not drilled immedi-
scatter-diagram in Figure 5b, and statistically analysed. ately after the first hole. This assures that twinned holes
Results clearly show that historic air-core holes tend to provide a good control of drilling quality, and supplement
underestimate THM grade. Bias of air-core samples is the traditional QAQC approaches (Abzalov, 2008) based
clearly seen by the noticeable mismatch between the RMA on duplicate samples and reference materials.
line and the first bisect (1:1 line) of that diagram (Fig. 5b). An example of twinned holes drilled at the Weipa baux-
This twinned-holes analysis has revealed that, in addition ite deposit in Australia is shown in Figure 8, where twinned
to variations related to lithology and mineralogy of the holes fully confirm previous drilling results: estimated
mineral sands, the bias of the air-core samples increases RMA functions of the studied variables coincide with the
when sampled units are located below the water table. 1:1 line on scatter diagrams. However, thickness and SiO2
Another example was taken from a Ti-sands project are characterized by larger variability than Al2O3. Coeffi-
in Madagascar (Figs. 6, 7; Table 1). The deposit was first cients of variation of thickness and SiO2 are 12.5% and
drilled using a vibratory drillhole built in-house by the pro- 9.7%, respectively, whereas CVAVR of Al2O3. is 1.9% (Table
ject team. At the feasibility stage, the quality of the data 1). These results are consistent with variography analysis
was validated with twinned holes drilled using a commer- of these variables. The Al2O3 variogram is characterized by
cially available sonic drill rig. Sonic drilling is a drilling a small nugget effect (10.5%), whereas thickness and SiO2
technique based on sonic vibratory technology (Oothoudt, values exhibit large short-range variability manifested by a
1999). This technique aims to collect representative undis- large nugget effect on their variograms: 39.3% and 24.4%,
turbed samples of nonconsolidated sands, as evidenced by respectively.
the well-preserved textures of recovered materials (Fig. 6).
In this example, sonic drilling fully confirmed the stratig- Discussion and Conclusions
raphy and grades of the heavy mineral sands as earlier de- The twinned-holes approach does not replace other
fined by the historic vibratory drilling (Fig. 7). validation techniques, such as bulk sampling or compara-
tive analysis of data from drilling campaigns related to dif-
Bauxites: Use of Twin Holes as a Routine Control of ferent stages of infill drilling. Each of the validation meth-
Drilling Quality
ods has its specific areas of application. Comparison of
In many bauxite deposits in Australia, Africa, and Asia, drillholes from different drilling campaigns is commonly
twinned holes are used as a routine quality control tech- used for testing exploration data prior to resource/reserves
nique. Besides conventional tasks such as diagnosis and estimation. This approach is most effective when different
quantification of the biases of assays, twinned holes are drilling generations evenly cover the same area, allowing
commonly used to verify thicknesses of bauxite seams and their comparisons by global statistics. Such comparisons
to assure accurate positioning of the upper and lower con- are usually made at a first stage of data testing, entirely

Fig. 6. Ti-sands project, Madagascar. a. Drill core of nonlithified sands recovered by sonic drilling. b. Close-up view of drill core specimen of sands.
Cross-current bedding is clearly discernable on the cut surface of the core.
 Use of Twinned Drillholes in Mineral Resource Estimation • M.Z. Abzalov 21

a b

Fig. 7. Validation of historic exploration data by twinned holes,

Ti-sands project, Madagascar (THM = total heavy minerals).
a. Historic hole drilled using vibratory technology, twinned
by sonic hole. b. THM grades of sonic drillhole samples ver-
sus samples from twinned vibratory holes (RMA = Reduced
Major Axis).

based on data subsets extracted from a drillhole database. sufficient for a conclusive decision regarding confirmation
When comparisons of the data generations reveal sys- of the variables of interest. Common practice is to drill
tematic differences, more detailed analyses are usually several twinned holes, but usually less than ten. However,
undertaken, including drilling of twinned holes. Bulk sam- when studied variables are characterized by large short-
ples are taken at the advanced stages of a mining project range variability, the number of twinned holes required for
study and are used mainly for metallurgical test work. An- conclusive statistical and geostatistical analysis might be
other objective of the bulk sampling program is validation larger. In the case studies described in this paper, the num-
of reserve models by comparing estimated values in the ber of twinned holes drilled varies from 5 to 62 (Table 1).
block model with bulk sampling results (Annels, 1991). Twinned holes should be located as close as possible to
A twinned-holes approach, representing the drilling of the original holes in order to minimize the effect of short-
a relatively small amount of verification holes, is usually range variability in the studied variables. Good practice in
applied for quick and conclusive confirmation/refutation the twinned-holes approach is to drill them no more then
of high-grade intersections (e.g., the Bre-X Busang pro- 5 m apart. In special cases, the distance between twinned
ject), verifying historic data, and testing third-party mining holes can be as close as 1 to 1.5 m, as previously discussed
projects. Given the typically small number of verification in the Busang project where the imperative was to con-
holes drilled, it is necessary to thoroughly plan the program clusively confirm or reject previous drilling results (Law-
to ensure that the confirmatory holes are of the best prac- rence, 1997).
tically achievable quality and value. They are often drilled At the Yandi project, where the studied mineralization
using a more expensive but more reliable technique than is characterized by large short-range variability, compara-
that used for the previous drilling. tive analysis of the matching pairs of blast-holes and RC
The quantity and quality of the twinned holes should be holes has shown that repeatability of the results signifi-
22 Exploration and Mining Geology, Vol. 18, Nos. 1–4, p. 13–23, 2009

information provides an important guide for choosing

matching intervals in twinned holes, and facilitates inter-
pretation of the data.
Comparison by mineralized intersections is the most
common and practical approach for comparing twinned
holes. Length-weighted average grades of the matching
intersections and thicknesses of mineralization are calcu-
lated and compared, and good practice is also to compare
locations of the geological contacts determined by twinned
holes. When the thickness of mineralization is large, and
in particular if the mineralization has a zoned structure, it
becomes necessary to compare twinned holes by geologic-
al units or samples grouped to equal length composites.
Grouping samples to larger composites is commonly ne-
cessary to minimize the noise exhibited by samples taken
at smaller intervals.
Statistical techniques used for analyses of twinned holes
results are not different to other QAQC approaches (Val-
lée et al., 1992; Pitard, 1998; Sinclair and Bentzen,1998;
Stanley and Lawie, 2007; Abzalov, 2008). Repeatability of
the twinned-holes results can be quantified by the average
coefficient of variation. Bias in the data can be diagnosed
by graphical tools, in particular RMA and RDP diagrams,
and then assessed by statistical t-tests.
In summary, the planning of twinned-holes studies
should consider choice of the optimal drilling technique for
verification holes, clear definition of the number of holes
and their location, optimal distance between twinned holes,
choice of the studied variables, and statistical techniques
for data analysis. This approach has been successfully used
in twinned-holes studies in many different deposit styles
and commodities, including orogenic gold, mineral sands,
bauxite, and banded iron formation iron ore.
Objectives of the studies varied from confirmation of
high-grade intersections at gold deposits, to quantification
and correction of bias in historic air-core data at Ti-sands
projects. These case studies demonstrate that a popularly
held belief that the twinned-holes approach is impractical
and/or inapplicable in geological environments charac-
terized by large short-range variability is unjustified, and
might commonly simply reflect prejudice rather than fact.
Fig. 8. Validation of grade and thickness of bauxite by twinned holes,
Weipa deposit, Australia (RMA = Reduced Major Axis). a. Al2O3; b. SiO2;
c. Thickness (m). Acknowledgments
The author expresses his sincere gratitude to geologists
from Rio Tinto, in particular J. Dumouchel and J. Bower
cantly worsens when the distance between twinned holes
for geological contributions and many stimulating discus-
increases from 1 to 10 m (Abzalov et al., 2007). This result
sions. The manuscript has benefited from critical reviews
suggests that many apparently unsuccessful twinned-hole
by G. Broadbent, A. Sinclair, and an anonymous reviewer.
studies might have failed because the twinned holes were
Helpful editorial comments by J. Richards are gratefully
drilled too far from the original holes.
acknowledged.
Geological interpretation should be an important part of
twinned-holes analysis. It has been shown (Postolski and
Sinclair, 1998; Sinclair, 1998) that robust understanding of References
geological features and their control of the grade distribu-
tion allows considerable refinement of spatial auto-correl- Abzalov, M.Z., 1999, Gold deposits of the Russian North
ation models of grade values, and therefore improves the East (the Northern Circum Pacific): Metallogenic over-
quality of numeric resource/reserves estimates. Geological view: Australasian Institute of Mining and Metallurgy,
 Use of Twinned Drillholes in Mineral Resource Estimation • M.Z. Abzalov 23

Proceedings of the PACRIM ‘99 symposium, p. 701– ploration results, mineral resources, and ore reserves:
714. AusIMM, 20 p.
Abzalov, M.Z., 2007, Zarmitan granitoid-hosted gold de- Lawrence, M.J., 1997, Behind Busang: the Bre-X scan-
posit, Tian Shan Belt, Uzbekistan: Economic Geology, dal: Could it happen in Australia?: Australian Journal of
v. 102, p. 519–532. Mining, v. 12, n. 134, p. 33–50.
Abzalov, M.Z., 2008, Quality control of assay data: A re- Olea, R.A., ed., 1991, Geostatistical glossary and multi-
view of procedures for measuring and monitoring preci- lingual dictionary: Oxford University Press, New York,
sion and accuracy: Exploration and Mining Geology, v. Oxford, 177 p.
17, p. 131–144. Oothoudt, T., 1999, The benefits of sonic core drilling to
Abzalov, M.Z., and Humphreys, M., 2002, Resource esti- the mining industry, in Tailing and mine waste ’99: Pro-
mation of structurally complex and discontinuous min- ceedings of the sixth international conference on tail-
eralisation using non-linear geostatistics: Case study of ings and mine waste ‘99, Fort Collins, Colorado, USA,
a mesothermal gold deposit in northern Canada: Explor- 24–27 January, p. 3–12.
ation and Mining Geology, v. 11, p. 19–29. Pan, G., 1995, Practical issues of geostatistical reserve es-
Abzalov, M.Z., and Mazzoni, P., 2004, The use of condi- timation in the mining industry: CIM Bulletin, v. 88, p.
tional simulation to assess process risk associated with 31–37.
grade variability at the Corridor Sands detrital ilmenite Peters, W.C., 1987, Exploration and mining geology, 2nd
deposit, in Dimitrakopoulus, R., and Ramazan, S., eds., edition: Wiley, New York, 706 p.
Orebody modelling and strategic mine planning: Uncer- Pitard, F., 1998, A strategy to minimise ore grade recon-
tainty and risk management: Australasian Institute of ciliation problems between the mine and the mill, in
Mining and Metallurgy, p. 93–101. Mine to mill, 1998: Australasian Institute of Mining and
Abzalov, M.Z., and Pickers, N., 2005, Integrating different Metallurgy, p.77–82.
generations of assays using multivariate geostatistics: Postolski, T.A., and Sinclair, A.J., 1998, Geology as a basis
A case study: Transactions of Institute of Mining and for refining semi-variogram models for porphyry-type
Metallurgy, Section B: Applied Earth Science, v. 114, deposits: Exploration and Mining Geology, v. 7, p. 45–
p. B23–B32. 50.
Abzalov, M.Z., Menzel, B., Wlasenko, M., and Phillips, Sinclair, A.J., 1998, Geological controls in resource/re-
J., 2007, Grade control at Yandi iron ore mine, Pilbara serve estimation: Exploration and Mining Geology, v.
region, Western Australia—Comparative study of the 7, p. 29–44.
blastholes and Reverse Circulation holes sampling: Sinclair, A.J., and Bentzen, A., 1998, Evaluation of errors
Australasian Institute of Mining and Metallurgy, Pro- in paired analytical data by a linear model: Exploration
ceedings of the iron ore conference, 20–22 August, and Mining Geology, v. 7, p. 167–173.
2007, Perth, p. 37–43. Sinclair, A.J., and Blackwell, G.H., 2002, Applied min-
Annels, A.E., 1991, Mineral deposit evaluation: A practical eral inventory estimation: Cambridge University Press,
approach: Chapman and Hall, London, 436 p. 381 p.
Davis, J.C., 2002, Statistics and data analysis in geology, Stanley, C.R., and Lawie, D., 2007, Average relative error
3rd ed.: Wiley, New York, 638 p. in geochemical determinations: Clarification, calcula-
Gilfillan, J.F., 1998, Testing the data—The role of tech- tion, and a plea for consistency: Exploration and Mining
nical due diligence, in AusIMM, eds., Ore reserves and Geology, v. 16, p. 265–274.
finance seminar, Sydney, 15 June, 1998: Australasian Vallée, M., David, M., Dagbert, M., and Desrochers, C.,
Institute of Mining and Metallurgy, p. 33–42. 1992, Guide to the evaluation of gold deposits: Geo-
JORC Code 2004, Australasian code for reporting of ex- logical Society of CIM, Special Volume 45, 299 p.
24 Exploration and Mining Geology, Vol. 18, Nos. 1–4, p. 13–23, 2009