Professional Documents
Culture Documents
net/publication/259480655
CITATIONS READS
6 1,154
3 authors:
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Luis Eduardo Souza on 07 February 2014.
This paper is part of a special issue on grade control and geostatistics of iron ore deposits
Sinclair and Vallée, 1998; Stephenson and Vann, 2001; There are definitions embedded in the Institution of
Souza et al., 2004). Mining and Metallurgy (IMM, 2001) code from the
Regarding the recent codes and systems of classification United Kingdom (Dowd, 1999). He emphasises the
of resources, the trend is not to regulate either the importance of correctly defining continuity and con-
technique used for classification or the method to estimate fidence levels for classifying resources and reserves. The
these resources. The codes emphasise the principles of standard established by the Canadian Institute of Mining,
transparency, materiality and also the responsibility of the Metallurgy and Petroleum, as well as its definitions, were
so-called competent person whose experience determines presented by the CIM (2000). South African code
which approach should be adopted for modelling any definitions (SAMREC, 2000) are discussed by Camisani-
given mineral deposit. In transferring the responsibility to Calzolari and Krige (2001). The geostatistical implications
the competent person, the codes allow a greater degree of involved with the implementation of the code are not
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy
freedom to choose methods and procedures for resource prescriptive with respect to the techniques that are used
assessment (Shaw et al., 2006). Various approaches that in the estimates of resources and reserves (Camisani-
have been historically used in the mining sector are still in Calzolari et al., 2000).
use, even with their limitation to provide an accurate error Most codes currently used for classifying mineral
definition. resources suggest the combination of numerous quanti-
Because the potential errors associated with an incorrect tative and qualitative criteria, such as the quality of data
or inappropriate estimate of grades can dramatically alter and geological and grade continuity. The codes should
an assessment of the mineable tonnage and because these not be prescriptive without proposing pre-established
estimates are completely dependent on the available rules (De-Vitry, 2003). He explains that these rules are
information, methods used for the estimation and difficult to apply to all situations, or even to use the
classification should provide a realistic risk model (Rossi, same rules for different types of ore within the same
2005). deposit. This is even though such rules are used
This article provides a review of some of the widely indiscriminately without a clear understanding of their
applied techniques to quantify the uncertainty asso- convenience or correct application. Conversely, some
ciated with the estimation of a mineral resource. The drawbacks in not having a clearly defined procedure are
methods were assembled in a software package devel- discussed in Shaw et al. (2006). They proposed an
oped using as initial template, i.e. the kriging program approach that would use not only the geological
(kt3d) presented by Deutsch and Journel (1998). Thus, knowledge but also the alternatives for quantifying the
different scenarios of resource categories (measured, error, highlighting the importance of incorporating such
indicated, and inferred) could be obtained for each information in classifying mineral resources.
method or for a given set of parameters chosen. To Due to the absence of standards prescribing a
determine its applicability and robustness, a compara- particular methodology, several studies can be found
tive study was conducted for the large iron ore deposit of showing different approaches to determine the uncer-
the Conceição Mine in central Brazil. tainty associated with estimates. Some of the traditional
methods and the geostatistical procedures used to
quantify and classify resources and reserves are pre-
Background of the topic sented in Annels (1991). A review of these traditional or
The historical evolution for international standards, classical methods and also the geostatistical framework
which regulate estimation, classification and reporting of available for the purpose, including conditional simula-
mineral resources and mineral reserves are presented by tion, is found in Mwasinga (2001), Sinclair and
Rendu (2000). Vallée (1999a, 1999b) examined the role Blackwell (2002) and Souza (2001, 2002). Annels
of the committee comprising the major international (1996) discusses the sources of error, their influence on
regulatory agencies (CMMI-Committee of the Con- resource classification and proposes a classification
gress of Mining and Metallurgical Institutions) from system (RRR, Resource Reliability Rating) based on
Australia, South Africa, United States, Canada, and the assigning a score to several factors.
United Kingdom and compared its resolutions with The limited relevance of the kriging variance as an
those from the United Nations Council to define indication of the uncertainty, and a proposition named
international standards for evaluation and classification the combined variance (a combination of the kriging
of mineral resources and reserves and for the future variance with the weighted variance of the data used to
possibility of ISO certification. estimate a given block) are discussed by Arik (1999) and
One of the most advanced and organised proposals Heuveling and Pebesma (2002). Similarly, an alternative
to report ore reserves and mineral resources is the to the kriging variance is the variance of interpolation
Australasian code (AusIMM, 1999) and some of its (Yamamoto, 1999), which allows a measure of the local
historical aspects, implementation and enforcement are data dispersion. Most of the work in this area and even
presented in Stephenson (2000a, 2000b). Some reasons the latest codes used for classification, emphasise that
that led Rio Tinto, for example, to adopt the JORC geostatistics is one of the few methods able to predict the
code (or Australasian code for Reporting of Mineral uncertainty associated with the estimate because it is
Resorces and Ore Resenses) are discussed by Weather- able to address accuracy, i.e. the unbiased-ness and
stone (2000). The competent person role under the laws precision or spread of the error (Camisani-Calzolari and
ruling Australian corporations, and the role of both the Krige, 2001; Camisani-Calzolari et al., 2000).
technical staff and Board of Directors from a mining There is a need to create international standards and
company are discussed by Phillips (2000). The compe- for better geological and geostatistical methods to
tent person definition appears in all recent codes and is measure global and local continuity of a geological
well defined in the JORC code. model (Vallée, 1999b). He also points out the necessity
Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3 167
de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
168 Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3
de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy
2 Comparative results for resource classification at the Conceição iron ore mine using the criterion of minimum number
of samples to estimate a given block grade: a refers to samples found within a maximum search distance larger than
the range of the variogram, and b repeats the process, but forces the samples used to estimate block grade to be
within a maximum distance of 70% the range of the variogram
compare the results selecting a method adequate to their be found within this search leading to all blocks
aims. being classified as measured resources. A prac-
tical range is necessary to proceed with the
Number of samples method
(ii) the number of samples within the range of the
This method is based on the minimum number of grades relates to the fact that this range refers to
samples found within a search volume and used to the spatial continuity of the grades, not necessa-
estimate a given block. Generally, the search volume is rily the geological spatial continuity. For iron ore
defined as a percentage of the variogram range as these deposits, the geological risk associated with the
ranges indicate the distance where the data exhibit size of a hematite orebody for instance is larger
spatial correlation. The minimum number of samples is than the risk associated with its grades.
used within the mining industry despite it being
empirical and highly arbitrary in terms of defining the
number of samples used in each class. There are two
Standard error of the estimate
additional issues this method does not address: Ordinary kriging is probably the most used geostatistical
(i) by having zonal anisotropy, the sill of the method for mineral resource estimates due to its
variogram is reached at infinity, and conse- simplicity, robustness and capacity to provide an
quently the minimum number of samples would uncertainty measurement using the kriging variance
Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3 169
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
3 Maps with variances of the estimated grade (left) and the corresponding error calculated for 95% confidence limits for
the bench at 970 m elevation (right) in the Conceição iron ore mine. Plots a and b refer to kriging variance and error,
c and d for interpolation variance and e and f for combined variance
(Matheron, 1963; David, 1988; Isaaks and Srivastava, Slope of the regression line (p)
1989). The pros and cons in using the kriging variance as This method was originally conceived as a test to
an index to assess the quality of an estimate are evaluate the quality of the estimates, and it consists of
discussed by various authors (Arik, 1999; Yamamoto, calculating the slope of the regression between the
1999; Isaaks and Srivastava, 1989; Goovaerts, 1997). estimated values and the real value of each block. As the
The kriging variance (s2OK ) incorporates in its real values are unknown, the slope of the regression (p)
calculation only the geometrical characteristics from is derived using the covariances between the samples
the sampling pattern, including sample locations and the used to estimate a given block and the block. Thus, the
spatial continuity model (Journel, 1986). s2OK is not estimate is closer to reality (smaller error) the closer to 1
affected by the sample values, i.e. it does not recognise is the value of p.
local variability (Goovaerts, 1997). It is not uncommon Commercial software packages have implemented the
to find blocks or grid nodes with the same estimated method and, currently, it is being used to analyse and
kriging variance even though the local grade variability optimise the search neighbourhood used for kriging
from the sample values in the local neighbourhood at (Rivoirard, 1987; Vann et al., 2003). Additionally, the
the two locations are significantly different. This method is used to classify blocks according to different
inability to recognise local variability is the major values of p.
weakness in using s2OK to measure uncertainty asso-
ciated with an estimate. Geometrical methods
Considering this limitation associated with s2OK , two Two variants of geometrical methods were investigated,
alternatives were used to calculate the error associated i.e. the area of influence, and the dilation-erosion which
with an estimate, i.e. the interpolation variance (Yama- is a type of mathematical morphology analysis to
moto, 1999) and the combined variance (Arik, 1999). distinguish between extrapolation and interpolation
These two alternative variances consider both the data zones. However, these methods do not provide a direct
spatial distribution and their local variability. measure of uncertainty as recommended by most codes,
170 Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3
de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3 171
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
172 Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3
de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy
Table 3 Number of blocks assigned to each resource Table 4 Number of blocks assigned to each resource
category using a geometrical method based on category using a geometrical method based on
the area of influence. Three values for the radius dilation-erosion of influence. Three values for the
of influence around a drill hole were chosen for radius of extrapolation were chosen for each
each resource category resource category
Resources Resources
Radius of influence Measured Indicated Inferred Extrapolation radii Measured Indicated Inferred
75, 125 and 150 22 129 1942 138 50, 100 and 125 14 956 8643 472
90, 140 and 170 23 215 947 82 75, 125 and 150 22 093 1925 191
100, 150 and 200 23 642 567 62 100, 150 and 175 23 585 554 120
Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3 173
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
Table 5 Resource category assigned to a selected block for various classification methods used and various selections
of controlling parameters
36261 Measured
56362 Measured
Minimum number of samples* 76563 Measured
96765 Indicated
116967 Inferred
36261 Measured
Minimum number of samples 56362 Measured
within 70% of the variogram range 76563 Measured
96765 Measured
116967 Indicated
36261 Measured
Minimum number of samples 56362 Measured
within 80% of the variogram range 76563 Measured
96765 Measured
116967 Indicated
36261 Measured
Minimum number of samples 56362 Measured
within 95% of the variogram range 76563 Measured
96765 Measured
116967 Measured
Kriging variance Measured
Standard error Combined variance Measured
Interpolation variance Measured
p150.95 and p250.80 Measured
Slope of the linear regression p150.90 and p250.80 Measured
p150.90 and p250.70 Measured
Area of influence (radii for 75, 125 and 150 m Measured
measured, indicated, inferred) 90, 140 and 170 m Measured
100, 150 and 200 m Measured
Dilation-Erosion 50, 100 and 125 m Indicated
(extrapolation distance) 75, 125 and 150 m Measured
100, 150 and 175 m Measured
*The numbers represent the minimum number of samples that a given block needs to be assigned into each of the resource categories
(measured, indicated or inferred).
174 Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3
de Souza et al. Comparative analysis of resource classification techniques: case study of the Conceição Mine,Brazil
Finally, two important aspects need to be remembered Matheron, G. 1963. Principles of geostatistics, Econ. Geol., 58, 1246–
1266.
being the subjectivity associated with the choice of
Mwasinga, P. P. 2001. Approaching resource classification: general
parameters used in the various methods for determining practices and the integration, Proc. 29th Int. Symp. on ‘Computer
resource inventories, and the relevance of testing various applications in the mineral industries’, Beijing, China, April, 97–
options and analysing the sensitivity of the choices with 104, A. A. Balkema.
respect to the resource classification results. Phillips, R. 2000. The liability of company directors and competent
persons for resource/reserve disclosure, in The Codes Forum,
Mineral Industry Consultants Association (MICA), Sydney,
Acknowledgements Australia, http://www.mica.org.au/pdf/phillips1.pdf
Rendu, J. M. 2000. International aspects of resource and reserve
The authors would like to thank CNPq (research agency reporting standards, in The Codes Forum, Mineral Industry
in Brazil) and Vale S.A. for supporting this study. Consultants Association (MICA), Sydney, Australia, http://
Published by Maney Publishing (c) IOM Communications Ltd and the Australasian Institute of Mining and Metallurgy
www.mica.org.au/pdf/rendu1.pdf
References Rivoirard, J. 1987. Two key parameters when choosing the kriging
neighbourhood, Math. Geol., 19, (8), 851–856.
Annels, A. E. 1991. Mineral deposit evaluation: a practical approach, Rossi, M. E. 2005. Indicator simulations of categorical variables, Proc.
London, Chapman & Hall. 32nd Int. Symp. on ‘Computer applications in the mineral
Annels, A. E. 1996. Ore reserves: errors and classification, Appl. Earth industries’, Tucson AZ, USA, March–April, EUA, 247–252.
Sci. (Trans. Inst. Min. Metall. A), 105, A150–A156. SAMREC. 2000. South African code for reporting of mineral resources
Arik, A. 1999. An alternative approach to resource classification, Proc. and mineral reserves (The SAMREC Code), The South African
28th Int. Symp. on ‘Computer applications in the mineral Mineral Resource Committee (SAMREC) and South African
industries’, 45–53, Golden, CO, Colorado School of Mines. Institute of Mining and Metallurgy, http://www.saimm.co.za/
Armstrong, M. 1998. Basic linear geostatistics, Berlin, Springer. links/reporting-codes.
AusIMM. 1999. Australasian code for reporting of mineral resources Shaw, W. J., Godoy, M. C. and Larrondo, P. 2006. An approach to
and ore reserves – The JORC code, Joint Ore Reserves more objective classification of mineral resources, Proc. 6th Int.
Committee of The Australasian Institute of Mining and Min. Geol. Conf., 85–89, Darwin, Australian Institute of Mining
Metallurgy, Australian Institute of Geoscientists and Minerals and Metallurgy.
Council of Australia, http://www.jorc.org Sinclair, A. J. and Vallée, M. 1998. Preface – quality assurance,
Camisani-Calzolari, F. A., Krige D. J. and Dixon, J. R. 2000. The continuous quality improvement and standards in mineral
South African code for reporting of mineral resources and resource estimation, Explor. Min. Geol., special ed., 7, (1–2), iii–v.
mineral reserves and the geostatistical implications involved, Sinclair, A. J. and Blackwell, G. H. 2002. Applied mineral inventory
Proc. 6th Int. Geostat. Cong., Cape Town, South Africa, April, estimation, Cambridge, Cambridge University Press.
CD-ROM. Souza, L. E., Gambin, F., Costa, J. F. C. L. and Koppe, J. C. 2001.
Camisani-Calzolari, F. A. and Krige, D. J. 2001. The SAMREC code Estimativa de Incerteza na Classificação de Recursos Minerais
seen in a global context, Proc. 29th Int. Symp. on ‘Computer por Simulação Geoestatı́stica, Revista Escola de Minas (Revista
applications in the mineral industries’, Beijing, China, April, 39– Técnica-Cientı́fica Mı́nero-Metalúrgica), 54, (2), 143–148.
44, A. A. Balkema. Souza, L. E., Costa, J. F. C. L. and Koppe, J. C. 2002. A geostatistical
CIM. 2000. Resource and reserve definitions: CIM standards on contribution to the use of resource/reserves classification systems,
mineral resources and reserves – definitions and guidelines, Can. Proc. 30th Int. Symp. on ‘Computer applications in the mineral
Min. Metall. Bull. (CIM Standing Committee on Reserve industries’, 73–84, Phoenix AZ, USA, February, Society for
Definitions), 93, (1044), 53–61. Mining Metallurgy and Exploration.
David, M. 1988. Handbook of applied advanced geostatistical ore
Souza, L. E., Costa, J. F. C. L. and Koppe, J. C. 2004. Uncertainty
reserve estimation, in Developments in geomathematics 6,
estimate in resource assessment: a geostatistical contribution, Nat.
Amsterdam, Elsevier Scientific Publishing Company.
Resour. Res., 13, (1), 1–15.
Deutsch, C. V. 1996. Correcting for negative weights in ordinary
Stephenson, P. R. 2000a. The JORC code – its operation and
Kriging, Comput. Geosci., 22, (7), 765–773.
application, Mineral Industry Consultants Association (MICA),
Deutsch, C. V. and Journel, A. G. 1998. GSLIB: geostatistical software
in The Codes Forum, Sydney, Australia, http://www.mica.org.au/
library and user’s guide, New York, Oxford University Press.
pdf/stephenson1.pdf
De-Vitry, C. 2003. Resource classification – a case study from the
Stephenson, P. R. 2000b. The 1999 JORC code and the mining
Joffre-Hosted iron ore of BHP Billiton’s Mount whaleback
professional, Proc. 4th Int. Min. Geol. Conf., Coolum,
operations, Appl. Earth Sci. (Trans. Inst. Min. Metall. A), 112,
Queensland, Australia, May, The Australasian Institute of
A185–A196.
Mining and Metallurgy, http://www.jorc.org/pdf/stephenson2.pdf
Dowd, P. A. 1999. Reserves and resources – continuity and levels of
confidence, Proc. 28th Int. Symp. on ‘Computer applications in Stephenson, P. R. and Vann, J. 2001. Common sense and good
the mineral industries’, 93–99, Golden, CO, Colorado School of communication in mineral resource and ore reserve estimation,
Mines. mineral resource and ore reserve estimation, in The AusIMM
Goovaerts, P. 1997. Geostatistics for natural resources evaluation, New guide to good practice, 13–20, Melbourne, The Australasian
York, Oxford University Press. Institute of Mining and Metallurgy.
Guimarães, M. L. V., Toledo, J. P. and Faria, C. A. 2004. Estimativa e Stone, J. G. and Dunn, P. G. 1996. Ore reserve estimates in the real
Monitoramento das Reservas de Minério de Ferro do Sistema Sul world, 2nd edn, Littleton, Society of Economic Geologists.
- CVRD – Quadrilátero Ferrı́fero – MG, Proc. III Cong. Vallée, M. 1999a. Toward resource/reserve estimation, inventory and
Brasileiro de Mina a Céu Aberto/III Cong. Brasileiro de Mina reporting standards, Proc. 28th Int. Symp. on ‘Computer
Subterrânea, Belo Horizonte, Instituto Brasileiro de Mineração. applications in the mineral industries’, 69–76, Golden, Colorado
Heuveling, G. B. M. and Pebesma, E. J. 2002. Is the ordinary Kriging School of Mines.
variance a proper measure of interpolation error?, Proc. 5th Int. Vallée, M. 1999b. Resource/reserve inventories: what are the objec-
Symp. of Spatial Accuracy Assessment in Natural Resources and tives?, Can. Min. Metall. Bull., 92, (1031), 151–155.
Environmental Sciences, 179–186, Melbourne, RMIT University. Vann, J., Jackson, S. and Bertoli, O. 2003. Quantitative Kriging
IMM. 2001. Code for reporting of mineral exploration results, mineral neighbourhood analysis for the mining geologist – a description
resources and mineral reserves (the reporting code), Institution of of the method with worked case examples, Proc. 5th Int. Min.
Mining and Metallurgy, European Federation of Geologists and Geol. Conf., Bendigo, The Australasian Institute of Mining and
The Institute of Geologists of Ireland, http://www.imm.org.uk/ Metallurgy, 1–9.
rescode/reportingcode.doc. Accessed at http://www.eurogeologists. Weatherstone, N. 2000. Rio Tinto’s adoption of the JORC code as a
de/images/content/panels_of_experts/minerals_and_their_sustainable_ world reporting standard, Mineral Industry Consultants
use/reporting_code.pdf. Association (MICA), in The Codes Forum, Sydney, Australia,
Isaaks, E. H. and Srivastava, M. R. 1989. An introduction to applied http://www.mica.org.au/pdf/weatherstone1.pdf
geostatistics, New York, Oxford University Press. Yamamoto, J. K. 1999. Quantification of uncertainty in ore reserve
Journel, A. G. 1986. Geostatistics: models and tools for the earth estimation: applications to Chapada copper deposit, State of
sciences, Math. Geol., 18, (1), 119–140. Goiás, Brazil, Nat. Resour. Res., 8, (2), 153–163.
Applied Earth Science (Trans. Inst. Min. Metall. B) 2010 VOL 119 NO 3 175