Sei sulla pagina 1di 32

Practical Geostatistics for

Resource Estimation
Overview
Georges Verly
1
Last Update: April 2014
2014
g y
21400 21600 21800 22000 22200 22400
0
2
Creating a resource block model
Resource Modeling is
0.5
Wst
0.0
0.5
2.0
Au
4 g/t
21400 21600 21800 22000 22200 22400
2
0
0
-
2
0
0








-
1
0
0
-
1
0
0








0
0








1
0
0
1
0
0








2
0
0
2
0
0


2
DDH
Ultimate Pit
21400 21600 21800 22000 22200 22400
-
4
0
0
-
4
0
0








-
3
0
0
-
3
0
0






-
2
0

PARADISEINC.
SECTION: 9200 N
CO & CM
MEAS/IND RESOURCES
lc_013a
Mar-2010
georges.verly@amec.com May 2014
21400 21600 21800 22000 22200 22400
0
2
When information consists of drill hole samples
that have been logged and assayed.
Resource Modeling is
0.5
Wst
0.0
0.5
2.0
Au
4 g/t
21400 21600 21800 22000 22200 22400
2
0
0
-
2
0
0








-
1
0
0
-
1
0
0








0
0








1
0
0
1
0
0








2
0
0
2
0
0


3
DDH
Topo
21400 21600 21800 22000 22200 22400
-
4
0
0
-
4
0
0








-
3
0
0
-
3
0
0








-
2
0

PARADISEINC.
SECTION: 9200 N
DDH SAMPLES
lc_011
Mar-2010
Content
Data Collection
Resource Modelling Steps
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
M d l V lid ti
4
Model Validation
Classification
May-2014
Course / Notes Format
Appendix 1: Exercise solutions
Other
georges.verly@amec.com May 2014
Content
Data Collection
Resource Modelling Steps
Drilling, logging
Sampling, assaying, QA/QC
Database
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
5
p y y
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
Model Validation
Classification
May-2013
Data collection is most important
because:
It is the first step in resource
Data Collection
It is the first step in resource
modeling
There is a significant risk
associated to data collection
Poor data collection practice may
result in huge loss of value
We have to collect the right data
and the data must be of good and the data must be of good
quality!
6
Oct-2013
Brief overview of sampling and
QA/QC in this course
georges.verly@amec.com May 2014
Sampling QA/QC
Sample values must be accurate and precise
Poor sampling introduces:
R d i
Sample values must be accurate and precise
Poor sampling introduces:
R d i
f.163
Accurate
Precise
Not Accurate
Precise
Accurate
Not Precise
Random noise
Reflected in variogram nugget effect
Can be compensated to some degree
Bias
Cannot be compensated
Difficult to identify if data are inaccurate
7
Random noise
Reflected in variogram nugget effect
Can be compensated to some degree
Bias
Cannot be compensated
Difficult to identify if data are inaccurate
Octo-2013
The database must be clean
Good database management
The database must be clean
Good database management
Database
Good database management:
Process for capturing data in place
Process for prioritizing entries when multiple measurements are made
on same sample
Avoid manual or repetitive entries
Verification steps implemented
Security measures in place
Good database management:
Process for capturing data in place
Process for prioritizing entries when multiple measurements are made
on same sample
Avoid manual or repetitive entries
Verification steps implemented
Security measures in place
Data capture and
management not part of
this course.
Security measures in place
Documentation (e.g. purpose of data collection)
Etc.
Security measures in place
Documentation (e.g. purpose of data collection)
Etc.
8
May-2012
Data capture and
management not part of
this course.
georges.verly@amec.com May 2014
Content
Data Collection
Resource Modelling Steps
Drilling, logging
Sampling, assaying, QA/QC
Database
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
9
p y y
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
Model Validation
Classification
May-2013
The drill hole logs are interpreted
and a geological model is
created
The drill hole logs are interpreted
and a geological model is
created
Geological Model
Geological Model
0.5 0.5
2.0
Au
4 g/t
21400 21600 21800 22000 22200 22400
0
-
1



0
0



1
0
0
1
0
0



2
0
0
2
0
0
created.
Most important step!
Mineralization controls must be
understood:
Controls must be at DH scale
Guides at district scales are of
little use
created.
Most important step!
Mineralization controls must be
understood:
Controls must be at DH scale
Guides at district scales are of
little use
Overburden
Ox. Saprolite
Sul. Saprolite
S k

21400 21600 21800 22000 22200 22400
1
0
0
1
0
0



2
0
0
2
0
0
Wst
0.0
DDH
Topo
21400 21600 21800 22000 22200 22400
-
4
0
0
-
4
0
0




-
3
0
0
-
3
0
0




-
2
0
0
-
2
0
0




-
1
0
1
0
0

PARADISEINC.
SECTION: 9200 N
DDH SAMPLES
lc_011
Samples logged and assayed
Geological model most important
Not part of this course
10
Saprock
Crb Lch Brk
Crb Stbl Brk
Dyke
Wst
0.0
2.0 Au
0.5
4 g/t
DDH
21400 21600 21800 22000 22200 22400 -
4
0
0
-
4
0
0




-
3
0
0
-
3
0
0




-
2
0
0
-
2
0
0




-
1
0
0
-
1
0
0




0
0



1
0


PARADISEINC.
SECTION: 9200 N
GEOLOGY + DDH SAMPLES
lc_012
Geological Model
Sep-2012
Geological model most important
Not part of this course
georges.verly@amec.com May 2014
Content
Data Collection
Resource Modelling Steps
Drilling, logging
Sampling, assaying, QA/QC
Database
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
11
p y y
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
Model Validation
Classification
May-2013
Data familiarization
A few simple:
Tools
Data familiarization
A few simple:
Tools
EDA Purpose, Tools
Purpose
Detecting possible errors
Identifying/confirming different
mineralizations
Answering questions such as:
What topcut values ?
Ordinary or indicator kriging ?
Mean and variance ?
Graphics
Statistics
Lots of results
Series of results per
geology domain
Unsure at start which result
i ht b f l
Detecting possible errors
Identifying/confirming different
mineralizations
Answering questions such as:
What topcut values ?
Ordinary or indicator kriging ?
Mean and variance ?
Graphics
Statistics
Lots of results
Series of results per
geology domain
Unsure at start which result
i ht b f l
Mean and variance ?
Providing information for
model validation
might be useful
Mean and variance ?
Providing information for
model validation
might be useful
12
Mar-2010
georges.verly@amec.com May 2014
EDA Graphs
30
Freq
(%)
Histogram of Au grades
P(2. < Au < 4.)
100
10
Log-Probability Plot
0
10
20
Au (g/t) 0 4 8 12 16
f. 3a
P(2. Au 4.)
y
99.99 99.8 99 98 95 90 80 70 605040 30 20 10 5 2 1 0.5 0.1 0.01
1.0
0.1
f.115
13
x
Scattergram
f. 89
Mar-2010
EDA Statistics
Location
Mean or average
30
Freq
(%)
Mean or average
Median (middle)
Dispersion (Spread)
0
10
20
Au (g/t)
f. 3f
0 4 8 12 16
Dispersion
Location
14
Variance
Standard deviation = Variance
Sep-2012
georges.verly@amec.com May 2014
EDA Statistics
Coefficient of Correlation
Smoothed Regression
S
R
E
J
E
C
T

A
U
ORIGINAL AU VERSUS REJECT
LC - TRENCHES
10.
100.
NB. OF DATA 477
XVAR: MEAN 5.814
STD. DEV. 6.110
YVAR: MEAN 6.021
STD. DEV. 7.264
CORRELATION 0.935
Au
Good
S
Poor
15
ORIGINAL AU
1. 10. 100.
1.
f107
Au
33b
Mar-2010
Sampling quality
Look at QA/QC results
Sampling quality
Look at QA/QC results
EDA Some Issues
Geological domains
Different mineralization characteristics?
Checking character of boundaries (firm?, soft?)
Compositing
Declustering
Top-cutting
Geological domains
Different mineralization characteristics?
Checking character of boundaries (firm?, soft?)
Compositing
Declustering
Top-cutting Top cutting Top cutting
16
Apr-2013
georges.verly@amec.com May 2014
EDA Quiz 1
2
1
2
8
6
2
1
6 Samples values (g/t Au)
f.184
17
Question: what is the average grade?
6 Samples values (g/t Au)
Mar-2010
2
EDA Quiz 2
1
2
8 2
1
6 Samples values (g/t Au)
f.184a
1
18
Question: what is the average grade?
6 Samples values (g/t Au)
Mar-2010
georges.verly@amec.com May 2014
Potential problem:
Clusters often located within
high-grade zones.
EDA Declustering
Clusters of samples are common
Potential problem:
Clusters often located within
high-grade zones.
Impact could be serious:
overestimation of the average
grade and variability
Solution Declustering:
Reduce the weight of each
clustered data
proportionally to the cluster
sampling density
Impact could be serious:
overestimation of the average
grade and variability
Solution Declustering:
Reduce the weight of each
clustered data
proportionally to the cluster
sampling density
f.116a
Clusters
N
sampling density
Calculate weighted
(declustered) statistics.
Declustering impact
Often a lower average grade.
19
Apr-2014
sampling density
Calculate weighted
(declustered) statistics.
Declustering impact
Often a lower average grade.
Potential problem:
Overestimation of mean grade
and of variability.
EDA Top-cutting
Outliers are common
0.3
Potential problem:
Overestimation of mean grade
and of variability. y
Solution (top-cutting):
Restrict high-grade zone of
influence during estimation; or
Top-cut high assay values
Combination of tools
Histogram, Probability plot
Indicator correlation plot
Metal at risk
?
0.1
0.2
0.3
No. Data 99688
mean 2.17
std. dev 4.95
330.0
0.100
Max.
Min.
f62
y
Solution (top-cutting):
Restrict high-grade zone of
influence during estimation; or
Top-cut high assay values
Combination of tools
Histogram, Probability plot
Indicator correlation plot
Metal at risk Metal-at-risk
Decile analysis
Impact
high-grade smearing reduced
Lower average grade
20
?
Important to record how much
metal removed by top-cutting
BH Au (g/t)
0.0 4.0 8.0 12. 16. 20.
0.0
f62a
Oct-2010
Metal-at-risk
Decile analysis
Impact
high-grade smearing reduced
Lower average grade
georges.verly@amec.com May 2014
Content
Data Collection
Resource Modelling Steps
Drilling, logging
Sampling, assaying, QA/QC
Database
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
21
p y y
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
Model Validation
Classification
May-2013
Suppose a vertical section through a mineralized zone as following:
Quiz 3
9m
Questions
3m
3m
9m
A B
D
Saprolite
Zone
f05_01
C
1) Which pair of samples probably shows the smallest grade difference?
A-B, A-C, or A-D ?
2) Same question for
22
Note: Grades such as Fe, MgO, SiO2, Ni
Sep-2012
georges.verly@amec.com May 2014
Variogram
The variogram is a function that quantifies the notion of
geological variability or spatial variability.
Why is it useful?
To improve accuracy of block grade estimates
To analyze block model selectivity (in situ
dilution)
To assess risk
For simulation
In fact, the variogram is essential
Why is it useful?
To improve accuracy of block grade estimates
To analyze block model selectivity (in situ
dilution)
To assess risk
For simulation
In fact, the variogram is essential
Distance
0
f.78
(dist)
S
i
l
l
No variogram No geostat
No guts No glory
(H. Parker)
No variogram No geostat
No guts No glory
(H. Parker)
23
Distance
0
0
Range
May -2013
24
georges.verly@amec.com May 2014
Contents
Data Collection
Resource Modelling Steps
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
Variogram: Spatial Variability
Block grade Distribution
Block vs. Sample grade distribution
Dilution Misclassification
25
Dilution, Misclassification
Estimation / Simulation
Model Validation
Classification
May-2013
Block Grade Distribution
Average grade
1.57 g/t
Small Blocks
Spread
Min/Max = 0.1/5.7 g/t
Variance = 1.96
Cut-off grade = 1.5 g/t
6 blocks (2x4m) > 1,5 g/t Au
f
2m
3.5
0.2
2.0
1.2
2.4
1.2
0.8
0.9
0 8
0.2 0.1 2.1 2.3
0 7 1 0 5 7
Proportion of ore
= 6/16 = 37%
Grade
= 3.00 g/t
4m
2m
f06_03a
0.8 0.7 1.0 5.7
1.5 g/t
georges.verly@amec.com May 2014
Average grade
1.57 g/t
Block Grade Distribution
Large Blocks
1.57 g/t
WAS
Spread
Min/Max = 1.00/2.23
Variance = 0.24
Cut-off grade = 1.5 g/t
2 blocks (4x8m) > 1,5 g/t Au
f
0.1 / 5.7
1.96
3.5
0.2
2.0
1.2
2.4
1.2
0.8
0.9
0.8
0.2 0.1 2.1 2.3
0.7 1.0 5.7
4m
1.83
1.23
1.00 2.23
Proportion of ore
= 50%
Grade
= 2.03 g/t
37%
3.00 g/t
8m
f06_03b
1.5 g/t
Block Grade Distribution
5
6 1
Grade/TonnageCurves
5
6 1
Grade/TonnageCurves
1
2
3
4
5
0.2
0.4
0.6
0.8
G
r
a
d
e
T
o
n
n
a
g
e
Tonnes (SB) Grade(SB)
1
2
3
4
5
0.2
0.4
0.6
0.8
G
r
a
d
e
T
o
n
n
a
g
e
Tonnes(SB) Tonnes(LB)
28
Feb-2014
SB = Small Blocks
LB = Large Blocks
0 0
0 0.5 1 1.5 2 2.5
CutoffGrade
Tonnes(SB) Grade(SB)
0 0
0 0.5 1 1.5 2 2.5
CutoffGrade
Grade(SB) Grade(LB)
georges.verly@amec.com May 2014
Contents
Data Collection
Resource Modelling Steps
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
C t
Only brief overview of
simulation in this
course
29
Concept
Classical Estimation Methods
Error of Estimation
Kriging
Model Validation
Classification
May-2013
Only brief overview of
simulation in this
course
Estimate grade at a given location using surrounding
information
Choice of method depends on deposit shape, mineralization
Estimate grade at a given location using surrounding
information
Choice of method depends on deposit shape, mineralization
Estimation - Purpose
Choice of method depends on deposit shape, mineralization
variability, data availability, etc.
Usually a weighted average of surrounding sample grades
Choice of method depends on deposit shape, mineralization
variability, data availability, etc.
Usually a weighted average of surrounding sample grades
Estimation methods:
Polygonal
Inverse distance
Sectional
Estimation methods:
Polygonal
Inverse distance
Sectional
Block
Samples:
Au
1
= 2.3 g/t
Au
2
= 6.7 g/t
Etc.
Au
1
Au
2
Au
3
Sectional
Kriging
Etc.
Surrounding grades should have similar grade values
to that expected in block
Concept of stationarity in the selection neighbourhood
Surrounding grades should have similar grade values
to that expected in block
Concept of stationarity in the selection neighbourhood
30
Sectional
Kriging
Etc.
Apr-2013
Block
f.183a
Estimated Au
Block
= w
1
Au
1
+ w
2
Au
2
+ w
3
Au
3
georges.verly@amec.com May 2014
Suppose the following
situation:
Cu
0
grade is unknown ?
Cu
1
= 0.5
Cu
2
= 0.2
5m
f.141a
Best Continuity
Quiz
Cu
1
and Cu
2
grades are
known
Both Cu
1
and Cu
2
are located
5m away from Cu
0
Spatial correlation is better
horizontally than vertically
5m
( ) ( ) = +
*
0 1 1 2 2
Cu w Cu w Cu
The estimate at Location 0 is:
What is your choice for the weights W
1
and W
2
?
Example:
Cu
0
*= 0.5 Cu
1
+ 0.5 Cu
2
(W
1
= W
2
= 0.5)
Cu
0
*= w
1
Cu
1
+ w
2
Cu
2
Jun-2013
Every time there is an
estimate, there is an error
Estimation Error
Er
o
0
Error
Error = True Value
Estimated Value
The spread of the error (o
Er
) can be
calculated if the variogram is known.
Jun-2013
georges.verly@amec.com May 2014
Estimation Error
?
Cu
1
= 0.5
Cu
2
= 0.2
5m
5m
f.141a
Best Continuity
Different estimation methods
Different spreads of errors
Different estimation methods
Different spreads of errors
5m
( ) ( ) = +
*
0 1 1 2 2
Cu w Cu w Cu
W t
A bit Even
f.152c
Worst
+ 0
NN2
Cu
0
* = 0 x Cu
1
+ 1 x Cu
2
NN1
Cu
0
* = 1 x Cu
1
+ 0 x Cu
2
A bit
better
+ 0
ID
Cu
0
* = 0.5 x Cu
1
+ 0.5 x Cu
2
Even
better
+ 0
Jun-2013
Estimation Error
The variogram gives access to
the spread of the estimation error
0.90
0.95
1.00
1.05
1.10
1.15
1.20 STDV(Error)
Best estimate is that one that
minimizes spread of the error
Kriging.
"NN 1"
"I t iti
"Best"
ID
"NN 2"
34
Spread: Standard deviation in this case
May-2009
0.9
0.80
0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 1.0
0.1 1.0 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.9 0
0.85
Estimation Weights (w1 and w2)
f.152b
g g
"Intuition
W2
W1
georges.verly@amec.com May 2014
Kriging picks the set of weights such that the spread of the error is
Estimation: Kriging
Kriging picks the set of weights such that the spread of the error is
minimized.
Kriging is:
Best Minimize the spread (variance) of the error
Linear Weighted average of surrounding sample grades
Unbiased Mean of the error is 0
Estimator
minimized.
Kriging is:
Best Minimize the spread (variance) of the error
Linear Weighted average of surrounding sample grades
Unbiased Mean of the error is 0
Estimator Estimator
BLUE
35
Estimator
BLUE
Feb-2012
Estimation: Kriging
O di k i i (OK)
Most common krigings
Ordinary kriging (OK)
Mineralization well behaved
Top-cutting data is enough to control high-grades
Indicator kriging (IK)
Mineralization very variable
Continuity directions vary with cut-off
36
Co-Kriging
More than one variable
Mar-2010
georges.verly@amec.com May 2014
4 g/t
21400 21600 21800 22000 22200 22400







2
0
0
2
0
0
Estimation: Kriging
0.5
Wst
0.0
0.5
2.0
Au
DDH
Ultimate Pit
3
0
0
-
3
0








-
2
0
0
-
2
0
0








-
1
0
0
-
1
0
0








0
0








1
0
0
1
0
0



PARADISEINC.
SECTION: 9200 N
CO& CM
37
21400 21600 21800 22000 22200 22400
-
4
0
0
-
4
0
0








-
3
0

CO & CM
MEAS/IND RESOURCES
lc_013a
Output from kriging is a grade block model that needs to be validated.
Mar-2010
38
georges.verly@amec.com May 2014
Contents
Data Collection
Resource Modelling Steps
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
39
Model Validation
Classification
May-2013
Estimated block model must be reasonable Estimated block model must be reasonable
Validation
Why
Tools
Visual checks
Statistical checks
Block model VS. sample mean grades
Block model variability Selectivity
Local accuracy (kriging efficiency)
Grade profiles across geological contacts
Visual checks
Statistical checks
Block model VS. sample mean grades
Block model variability Selectivity
Local accuracy (kriging efficiency)
Grade profiles across geological contacts
Tools
Cross-validation
Reconciliation
With previous model
With production
Cross-validation
Reconciliation
With previous model
With production
40
Mar-2010
georges.verly@amec.com May 2014
Validation Visual Validation
22400 22500


Kriged Grade Map + Sample Values






1
1
4
0
0
1
1
4
0
0


41
1 2 3 4
Au (g/t)
0. 5.
22400 22500


xval_map_ok
Look at the trends,
geological contacts,
estimated grades vs.
sample values
Feb-2012
42
georges.verly@amec.com May 2014
Contents
Data Collection
Resource Modelling Steps
Geological Modelling
Block Grade Model Estimation
EDA: Exploratory Data Analysis
Variogram: Spatial Variability
Block grade Distribution
Estimation / Simulation
43
Model Validation
Classification
May-2013
Resources must have reasonable prospects for
economic extraction.
S t f i l ti it t ff d
Resources must have reasonable prospects for
economic extraction.
S t f i l ti it t ff d
Regulation / Classification
Some assessment of economics, selectivity, cutoff-grade
Resources must be categorized into Inferred,
Indicated, and Measured categories.
Depends on confidence in geological and grade
continuity.
Depends also on quality of information.
Some assessment of economics, selectivity, cutoff-grade
Resources must be categorized into Inferred,
Indicated, and Measured categories.
Depends on confidence in geological and grade
continuity.
Depends also on quality of information.
Resources must be signed off by a Qualified/
Competent Person.
Resources must be signed off by a Qualified/
Competent Person.
44
Jun-2013
georges.verly@amec.com May 2014
21400 21600 21800 22000 22200 22400
0
2
Resource modeling is creating a resource block model
Summary (1/4)
0.5
Wst
0.0
0.5
2.0
Au
4 g/t
21400 21600 21800 22000 22200 22400
2
0
0
-
2
0
0








-
1
0
0
-
1
0
0








0
0








1
0
0
1
0
0








2
0
0
2
0
0


45
DDH
Ultimate Pit
21400 21600 21800 22000 22200 22400
-
4
0
0
-
4
0
0








-
3
0
0
-
3
0
0






-
2
0

PARADISEINC.
SECTION: 9200 N
CO & CM
MEAS/IND RESOURCES
lc_013a
Sep-2012
21400 21600 21800 22000 22200 22400
0
2
Summary (2/4)
When information consists of drill hole samples
that have been logged and assayed.
0.5
Wst
0.0
0.5
2.0
Au
4 g/t
21400 21600 21800 22000 22200 22400
2
0
0
-
2
0
0








-
1
0
0
-
1
0
0








0
0








1
0
0
1
0
0








2
0
0
2
0
0


46
DDH
Topo
21400 21600 21800 22000 22200 22400
-
4
0
0
-
4
0
0








-
3
0
0
-
3
0
0








-
2
0

PARADISEINC.
SECTION: 9200 N
DDH SAMPLES
lc_011
Sep-2012
georges.verly@amec.com May 2014
Resource estimation steps
1. Data collection
Resource estimation steps
1. Data collection
Summary (3/4)
Sampling
Database
Clean and complete database
Good sampling and QA/QC
2. Good geological model
Mineralization controls understood
3. EDA: Exploratory Data Analysis
Data checking
Clean and complete database
Good sampling and QA/QC
2. Good geological model
Mineralization controls understood
3. EDA: Exploratory Data Analysis
Data checking
EDA
Geology
Data statistics
Including analysis of geological boundaries
Information gathering used for the:
Estimation (topcuts, ...)
Model validation (mean, variance, ...)
Data statistics
Including analysis of geological boundaries
Information gathering used for the:
Estimation (topcuts, ...)
Model validation (mean, variance, ...)
47
0
10
20
30
Freq
(%)
Mean
Au (g/t)
f. 3d
Variance
0 4 8 12 16
Feb-2014
Resource estimation steps
4. Variogram: Spatial variability analysis
Resource estimation steps
4. Variogram: Spatial variability analysis
Summary (4/4)

Variogram
Quantify grade spatial correlation
5. Block grade Distribution
Grade distribution depends on block size
Selectivity, dilution, recovery issues
6. Estimation
Error of Estimation
Quantify grade spatial correlation
5. Block grade Distribution
Grade distribution depends on block size
Selectivity, dilution, recovery issues
6. Estimation
Error of Estimation
BLUE
GTCurves
Selectivity
Kriging
7. Model Validation
Estimated model must be thoroughly checked
8. Classification
48
Kriging
7. Model Validation
Estimated model must be thoroughly checked
8. Classification
0.9 3.7
2.8
3.4
8.2
2.1
?
Validation
Jun-2013
georges.verly@amec.com May 2014
Contents
Resource estimation steps
1. Database
2. Sampling p g
3. Geological Model
4. EDA: Exploratory Data Analysis
5. Variogram: Spatial Variability
6. Block grade Distribution
7. Estimation
8. Model Validation
9. Classification
49
Course / Notes Format
Appendix 1: Exercise solutions
May-2012
The Course
5. Block Grade Distribution
Will follow the resource estimation steps as in this
introduction.
1. Univariate statistics
Theory
EDA practice
2. Bivariate Statistics
Theory
5. Block Grade Distribution
Sample vs. block grade
distribution
Selective Mining Unit
(SMU)
6. Estimation
Estimation methods
Error of Estimation
7. Kriging
Ordinary Kriging
Content
50
y
EDA practice
3. Sampling, QA/QC
4. Spatial Variability
Variograms
Ordinary Kriging
Other krigings
8. Model Validation
9. Classification
10.Simulation
Apr-2013
georges.verly@amec.com May 2014
The Geostat Learning Curve
Monday Tuesday Wednesday Thursday Friday
Geology here
to give you a
break
51
May-2013
EDA
Sampling
QA/QC
Variogram
Block Grade
Distribution
Error of
Estimation
Kriging Model
Validation
Simulation
Classification
52
georges.verly@amec.com May 2014
The Notes Table of Content
53
Oct-2010
The Notes Usual Format
Will follow the following format
E l f ti
Usually four parts per section
1. Preliminaries
Content
Brief overview of material
covered
Format
3. Appendices
Demonstrations
Exercise solutions
4 Review
Example of sections:
Variography
Estimation
54
What you need to know
2. The section itself
With eventually some
recalls
Some exercises
4. Review
Oct-2010
georges.verly@amec.com May 2014
The Notes Special Section
One special section to show:
Link between course and estimation
process
Progress made during the course
Different kinds of variances
Formulas
Appendix 3 Formulas
Univariate & Bivariate Statistics
=
1
X i m x
N
Mean
Covariance
( ) ( ) ( ( =

1
COV X Y x m y m
Calculating un-weighted (nave) statistics
Appendix 2 Variances (1/6)
Variance
The variance is a measure of dispersion of
values around the average
For the same element (e g Cu) there are
Freq
(%)
f. 3f
Dispersion
Freq
(%)
f. 3f
Dispersion
55
Apr-2013
21
Nov 2012
( ) ( ) ( ( =

| |
( = = |
\ .

COV , ,
1
i X i Y
XY i i X Y
X Y x m y m
N
s x y m m
N
Variance
Coefficient of Variation
Correlation Coefficient
( )
| |
= = |
\ .

2 2 2 2 1 1
X i X i X s x m x m
N N
Standard Deviation
= Variance X s
=
X
X
X
s
CV
m
( ) = =

COV ,
COR ,
XY
X Y
X Y
X Y r
s s
Note:
Sample variance: /(N-1)
Population variance: /N
Variance of sample grade values
Section Univariate Statistics
The sample grade variance is
obtained at the stage of EDA statistics
VAR( ) Z
2 2
Other notations: s ,
Z Z
o
13
For the same element (e.g. Cu), there are
several variances
Values
Variance of sample grade values
Section Univariate Statistics
The sample grade variance is
obtained at the stage of EDA statistics
VAR( ) Z
2 2
Other notations: s ,
Z Z
o
Oct-2009
56
georges.verly@amec.com May 2014
Appendix 1
Quizzes and Solutions
57
Mar-2010
Appendix 1 Quiz 1
2
1
2
8
6
2
1
6 Samples values (g/t Au)
f.184
58
Question: what is the average grade?
6 Samples values (g/t Au)
Mar-2010
georges.verly@amec.com May 2014
Appendix 1 Quiz 1 Solution
Nave average
1
2
8
6
2
1 2 8 6 2 1
Declustered average
1
8 2
1
Question: what is the average grade?
6 Samples values (g/t Au)
f.184
1 2 + 8 + 6 + 2 + 1
3.3
6
+
=
8 + 6
1 2 + + 2 + 1
2
2 6
+
59
About -20% change from naive to declustered!
2
2.6
5
=
Mar-2010
2
Appendix 1 Quiz 2
1
2
8 2
1
6 Samples values (g/t Au)
f.184a
1
60
Question: what is the average grade?
6 Samples values (g/t Au)
Mar-2010
georges.verly@amec.com May 2014
2
8 2
1
Appendix 1 Quiz 2 Solution
Average (all data)
1 2 8 1 2 1 1
8 2
1
Question: what is the average grade?
6 Samples values (g/t Au)
f.184a
Average (outlier deleted)
1 2 + 8 + 1 + 2 + 1
2.5
6
+
=
1 2 + 1 + 2 + 1
1.4
5
+
=
61
About -40% change when highest value is deleted!
5
Mar-2010
Suppose a vertical section through a mineralized zone as following:
Quiz 3
9m
Questions
3m
3m
9m
A B
D
Saprolite
Zone
f05_01
C
1) Which pair of samples probably shows the smallest grade difference?
A-B, A-C, or A-D ?
2) Same question for
62
Note: Grades such as Fe, MgO, SiO2, Ni
Sep-2012
georges.verly@amec.com May 2014
3m
9m
Situation
A B C
Appendix 1 Quiz 3 Solution
Q1) Which pair of samples probably shows the smallest grade difference?
A B A C A D ?
3m
A B
D
Saprolite
Zone
f05_01
C
Likely grade
profiles
Likely grade
profiles
A D
A-B, A-C, or A-D ?
A1) A-B because separating distance is:
Smaller than A-C
Same as A-D, but A-B is the direction of better continuity
63
Oct-2013
Variogram Solution 1 (3/3)
3m
9m
Situation
A B C
Appendix 1 Quiz 3 Solution
Q2) Which pair is next?
A C A D ?
3m
A B
D
Saprolite
Zone
f05_01
C
Likely grade
profiles
Likely grade
profiles
A D
64
Oct-2013
A-C or A-D ?
A2) Probably A-C if grade continuity is much better along A-C direction
than along A-D
Otherwise it could be A-D
Depends on mineralization anisotropy.
georges.verly@amec.com May 2014

Potrebbero piacerti anche