1 of 23

What is Geostatistics? Why is it Special?

Geostatistics could be considered equal parts MATH /
SCIENCE / ART

MATH – because it is founded on statistics and

calculus

SCIENCE – because it incorporates physics, chemistry, and

geologic principles

ART – because subjectivity and opinion

are a requirement (QP has ultimate control)

2 of 23
Why Geostatistics?
Because thanks to Mother Nature. Geological features
are NOT considered random

Therefore mineral concentrations, related

to geological features, can be predicted

The underlying measureable continuity, allows for the

interpolation of values, into unsampled areas, using
available samples
3 of 23
Geostatistics Alone is not Enough
Five Fundamentals of Resource Estimation

1. Proper sampling of deposit

2. Integrity of the digital database
3. Understanding of the deposit geology and proper use
in resource estimation procedures
4. Use of appropriate estimation techniques
5. Use of appropriate classification methodologies

4 of 23
Geostatistics Alone is not Enough
Five Fundamentals of Resource Estimation

1. Proper sampling of deposit

1. Is the deposit drilled and sampled appropriately?
2. Are important geological contacts preserved in assay intervals?
1. Are grades “smeared” across contacts? If so, is it important for the scale of the
model being constructed?
3. Are appropriate and necessary geological data points captured
during logging?
4. Are appropriate analytical methods used for assaying?

5 of 23
Geostatistics Alone is not Enough
Five Fundamentals of Resource Estimation
1. Proper sampling of deposit
2. Integrity of the digital database
1. Has the digital data been validated?
1. Checked for assays greater than hole depth, overlapping intervals, erroneous
downhole deviation, appropriate collar locations, etc…
2. Does the digital database contain all available information, or
simply a predefined subset?
1. If a predefined subset, is the subset appropriate?
3. Are special fields appropriately identified and understood?
1. Below and above detection limits are accurately defined?
2. Are gaps or unsampled intervals understood? How should they be handled?

6 of 23
Geostatistics Alone is not Enough
Five Fundamentals of Resource Estimation
1. Proper sampling of deposit
2. Integrity of the digital database
3. Understanding of the deposit geology and proper use
in resource estimation procedures
1. Are geological controls of primary mineralizing events understood?
1. Lithological, alteration, structural, etc..
2. Are post primary mineralization controls understood?
1. Faulting causing displacement, volumetrically important barren intrusives,
weathering controls, etc…
3. What about geological controls, not “required” for grade estimation
but needed for geomet, geotech, density, etc…

7 of 23
Geostatistics Alone is not Enough
Five Fundamentals of Resource Estimation
1. Proper sampling of deposit
2. Integrity of the digital database
3. Understanding of the deposit geology and proper use in resource estimation
procedures
4. Use of appropriate estimation techniques
1. What estimation method is most applicable?
1. ID, OK, SK, Simulation
2. Is the chosen estimation method applicable to underlying grade
distribution, grade variability, spatial continuity, and account for
volume variance relationships?
3. Are you choosing the appropriate estimation parameters, to match
the estimation method?
8 of 23
Geostatistics Alone is not Enough
Five Fundamentals of Resource Estimation
1. Proper sampling of deposit
2. Integrity of the digital database
3. Understanding of the deposit geology and proper use in resource estimation
procedures
4. Use of appropriate estimation techniques
5. Use of appropriate classification methodologies
1. Are chosen confidence criterion applicable to deposit style
2. Do they appropriately account for the QP’s judgement of the quality
of sampling, database, geological continuity and understanding, and
grade estimation quality and continuity?
3. Are other necessary data points missing?
1. Density, Oxidation state, etc..

9 of 23
Geostatistics Alone is not Enough

If the fundamental inputs to resource estimation

are ignored, done incorrectly, or not understood…

They can never be compensated for, nor corrected

by geostatistics alone

Errors in the underlying data, database, geological

assumptions will be reproduced in the model
10 of 23
Where to Start
Interpretation
As geologists we are are tasked with interpreting the
data and “understanding” the geology of our deposits.

My Advice:
Build your concept with paper, computers, your
knowledge

Don’t let the computer build your concept

You are the operator. You are in Control. Make the

computer do what you want!

11 of 23
12 of 23
 Porphyry Copper / Supergene Enrichment
Mineralized Porphyry Copper Deposit
Intrusive Stock
Rain water & Ground water precipitating through ground, channeled by fractures

Copper Mineralogical Zonation

Leached Cap Leaching of Primary Copper by

metals removed, perched zones
Ground Water
Oxide Minerals
(chrysocolla, brochantite, etc..)
Supergene Enrichment
(chalcocite, covellite > chalcopyrite)
Water Table Precipitation of Secondary Copper
Transition
(chalcopyrite > chalcocite)
Hypogene
(chalcopyrite, bornite – no visible
secondary copper minerals)

13 of 23
Data Statistics Change by Geology

Total Copper Sequential Copper

10.000 10.000

1.000 1.000

0.100 0.100

0.010 0.010

0.001 0.001

0.000 0.000
leach oxide enr trans prim all leach oxide enr trans prim all
Q1 0.03 0.21 0.39 0.32 0.18 0.09 Q1 0.02 0.14 0.30 0.11 0.02 0.03
Min 0.00 0.01 0.01 0.01 0.00 0.00 Min 0.00 0.01 0.00 0.01 0.00 0.00
Median 0.06 0.31 0.67 0.46 0.33 0.29 Median 0.04 0.23 0.56 0.15 0.04 0.08
Mean 0.09 0.44 0.85 0.51 0.36 0.38 Mean 0.08 0.35 0.74 0.21 0.05 0.28
Max 3.48 8.35 8.48 5.82 6.50 8.48 Max 3.39 7.36 7.00 2.69 2.75 9.80
Q3 0.10 0.54 1.12 0.64 0.49 0.51 Q3 0.07 0.42 0.99 0.24 0.06 0.32
NSamples 17,425 2,836 14,836 6,673 47,815 95,964 NSamples 6,759 2,663 14,431 6,476 19,548 51,950
CV 1.42 1.12 0.77 0.60 0.67 1.09 CV 1.95 1.27 0.84 0.94 1.19 1.66

14 of 23
Data Statistics Change by Geology

Acid Soluble Copper Cyanide Soluble Copper

10.000 10.000

1.000 1.000

0.100 0.100

0.010 0.010

0.001 0.001

0.000 0.000
leach oxide enr trans prim all leach oxide enr trans prim all
Q1 0.01 0.11 0.05 0.02 0.01 0.01 Q1 0.01 0.02 0.24 0.08 0.01 0.01
Min 0.00 0.00 0.00 0.00 0.00 0.00 Min 0.00 0.00 0.00 0.00 0.00 0.00
Median 0.02 0.17 0.08 0.03 0.01 0.02 Median 0.01 0.03 0.47 0.12 0.02 0.05
Mean 0.03 0.27 0.10 0.04 0.01 0.05 Mean 0.03 0.08 0.64 0.17 0.03 0.22
Max 1.93 6.52 1.55 1.14 0.29 6.52 Max 3.15 3.07 6.59 2.24 2.63 6.59
Q3 0.03 0.32 0.12 0.04 0.02 0.06 Q3 0.02 0.06 0.87 0.20 0.04 0.22
NSamples 10,785 2,700 14,704 6,562 21,282 58,790 NSamples 7,255 2,663 14,433 6,478 19,966 53,157
CV 2.11 1.31 0.85 0.88 0.82 2.07 CV 3.22 2.34 0.90 1.04 1.47 1.88

15 of 23
 Data Statistics Change by Geology
Why is Domaining Important?
Contact Profile
Lcap CuT Ox CuT Enr CuT Trans CuT Prim CuT Lcap Scu Enr Scu Ox Scu Trans Scu Prim Scu
1.20

1.00
Total Copper and Sequential Copper

0.80

Leach cap oxide enriched transition primary

0.60

0.40

0.20

0.00
Distance From Contact

16 of 23
The Model
Cross Section View through the deposit
Oxide

Enr

Trans

17 of 23
Truth (Domained) vs UnDomained
Two estimations were
completed using the
Scenario Comparison Grade / Tonnage
same estimation 500,000 1.5

Thousands
parameters (search,
450,000 1.4
sample count, etc..)
400,000 1.3
First estimation
350,000 1.2
considered geological
domains 300,000 1.1
Tonnage

Grades
Second estimation 250,000 1

considered NO geological 200,000 0.9

domains
150,000 0.8

At 0.15% Scu cut-off the 100,000 0.7

Undomained model
predicts -6.5% less metal 50,000 0.6

- 0.5
At $2.50/lb Cu this 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
equates to ~$950M SCu Cutoff
Difference
Tonnage No Domains Tonnage Domains Scu No Domains Scu Domained

18 of 23
The Model Compared to Grade Shell
Cross Section View through the deposit
Oxide

Enr

Trans

19 of 23
Truth (Domained) vs Grade Shell
Here we compare the
original domained model,
Scenario Comparison Grade / Tonnage
to an implicit grade shell 500,000 1.50

Thousands
model
450,000 1.40

A 0.15% Scu cut-off grade 400,000 1.30

shell was produced via
350,000 1.20
implicit techniques
300,000 1.10
At 0.15% Scu cut-off the
Tonnage

Grades
grade shell model 250,000 1.00

predicts 8.5% more metal 200,000 0.90

At $2.50/lb Cu this 150,000 0.80

equates to ~$1.3B 100,000 0.70

Difference
50,000 0.60

- 0.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SCu Cutoff

Tonnage GradeShell Tonnage Domains Scu GradeShell Scu Domained

20 of 23
Undomained vs Grade Shell
To show the difference
between the geology
Scenario Comparison Grade / Tonnage
model (although 500,000 1.50

Thousands
undomained) and the
450,000 1.40
gradeshell
400,000 1.30
At 0.15% Scu cut-off the
350,000 1.20
grade shell model
predicts 15% more metal 300,000 1.10
Tonnage

Grades
At $2.50/lb Cu this 250,000 1.00

equates to ~$2.2B 200,000 0.90

Difference
150,000 0.80

100,000 0.70

50,000 0.60

- 0.50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
SCu Cutoff

Tonnage GradeShell Tonnage Undomained Scu GradeShell Scu Undomained

21 of 23
Proposed Ideas Going Forward
Include write-up on Geological / Domain model
validation
many reports contain very little back up / justification to
parameter choices in geological model, and / or domain choices
Include volume / tonnage sensitivity information
test multiple methods (explicit, implicit parameter option A,
implicit parameter option B, etc…)
Include a comparison of the block proportion summary
from 3D geological model to a NN declustered model
of data
has any volume bias been introduced?

22 of 23
Conclusion
Geostatistics requires an artistic component

Geological features are NOT random 10.000

Grades can be interpolated IF geological

features are understood 1.000

Computers & Software should not be

expected to do all the work, the input and 0.100
guidance must come from the geologist

BUT the use of computers & software gives 0.010

geologists power and control, like we have
never seen before
0.001
Ore grades often change considerably from
zone to zone, so overall contained metal will
be directly correlated to volume
representation of the high grade geological 0.000
features leach oxide enr trans prim all
Q1 0.02 0.14 0.30 0.11 0.02 0.03
Min 0.00 0.01 0.00 0.01 0.00 0.00
Test multiple methods to understand the Median 0.04 0.23 0.56 0.15 0.04 0.08

uncertainty associated to model Mean 0.08 0.35 0.74 0.21 0.05 0.28
Max 3.39 7.36 7.00 2.69 2.75 9.80
Q3 0.07 0.42 0.99 0.24 0.06 0.32
Models are expected to be reproducible, so NSamples 6,759 2,663 14,431 6,476 19,548 51,950
parameter choices must be disclosed CV 1.95 1.27 0.84 0.94 1.19 1.66

23 of 23