Dhirendra Pratap Singh

M.Tech.(Mineral Exploration)
Indian School Of Mines ,Dhanbad.
GEOSTATISTICS - PAST, PRESENT AND
FUTURE
AND CASE STUDY ON BAUXITE
OUTLINE
IAMG (INTERNATIONAL ASSOCIATION OF
MATHEMATICAL GEOSCIENCES- 1968)
PROBLEMS/OBJECTIVES
WHAT WAS USED BEFORE
MATHERON, GANDIN and MATERN
COMPUTING
OTHER PEOPLE
WHERE IS IT GOING?
CONFERENCES
SOME BOOKS


PROBLEMS/OBJECTIVES
Applications in
Mining
Hydrology
Petroleum
Soil Science
Ecology
Epidemiology
Environmental monitoring and assessment
Agronomy
Atmospheric Sciences
Any discipline with spatial data
Complications with spatial data
Often expensive to collect and/or difficult
Point data vs non-point data
Hard data vs soft data



Georges Matheron(1930-2000)
Trait de Gostatistique applique, tome 1 (1962), tome 2
(1963). Paris: Editions Technip.
Centre de Geostatistiques et Morphologie Mathematiques,
Ecole des Mines de Paris (Fontainebleau).
The theory of regionalized variables and its applications.
Paris School of Mines publication, (1971)
The intrinsic random functions and their applications. Adv
Appl Prob 5: (1973)439-68.
Random Sets and Integral Geometry. J. Wiley. (1975)
Also did fundamental work on flows in porous media
Strong ties with mining industry, petroleum industry
Hydrology research group at Fontainebleau
Was acquainted with Matern and Gandin and their work
Influenced by work of D. Krige (South Africa)
Mathematical Morphology group became separate



OTHER PEOPLE
Krige, Danie G. 1919-
"A statistical approach to some basic mine valuation problems on the
Witwatersrand". J. of the Chem., Metal. and Mining Soc. of South
Africa 52 (1951): 119-139.
Tribute to Krige.pdf
Richard Webster
Rothamstead Research Center (UK)
Four seminal articles about applications in soil science, 1979-1980
Geostatistics for Environmental Scientists (2001) with M. Oliver
Noel Cressie
Statistics for Spatial Data (1993)
Brian Ripley
Spatial Statistics
Contributor to R

Geostatistics
Geostatistics:The original purpose of geostatistics
centered on estimating changes in ore grade within a
mine.
The principles have been applied to a variety of areas in
geology and other scientific disciplines.

A unique aspect of geostatistics is the use of
regionalized variables which are variables that fall
between random variables and completely deterministic
variables.
Geostatistics
Regionalized variables describe phenomena
with geographical distribution (e.g.
elevation of ground surface).

The phenomenon exhibit spatial continuity.
Geostatistics
It is notalways possible to sample every location.
Therefore, unknown values must be estimated
from data taken at specific locations that can be
sampled.

The size, shape, orientation, and spatial
arrangement of the sample locations are termed
the support and influence the capability to
predict the unknown samples.

Kriging
Kriging Method
Kriging is a geostatistical method and a probabilistic method,
unlike the others, which are deterministic. That is, there is a
probability associated with each prediction. Kriging has both a
deterministic and probabilistic component, respectively
Z(s) = (s) + (s), where both are functions of distance
Assumes spatial variation in variable is too irregular to be modeled
by simple smooth function, better with stochastic surface
Interpolation parameters (e.g. weights) are chosen to optimize fn
Assumes that variable in space can be modeled as sum of three
components: 1) structure/deterministic part, 2) random but spatially
correlated part and 3) spatially uncorrelated random part
Kriging Method
Hence, foundation of Kriging is notion of spatial autocorrelation,
or tendency of values of entities closer in space to be related.
This is a violation of classical statistical models, since
observations are assumed to be independent.
Autocorrelation can be assessed using a semivariogram, which
plots the difference in pair values (variance) against their distances.
Where autocorrelation exists, the
semivariance should increase until certain
distance where SV= variance around mean,
so flattens out. That value is called a sill.
The sloped area, or range is where values
are related to each other. Intercept is nugget
Kriging
5
conditioning
points
1
conditioning
point
Simulation
5
conditioning
points
1
conditioning
point
KRIGING METHODS
Hand Fit Method




Non-linear Least Squares Fit Method
Point Kriging Cross-Validation Method
Ordinary Kriging
Ordinary kriging is the simplest form of kriging.

It uses dimensionless points to estimate other
dimensionless points, e.g. elevation contour
plots.

In Ordinary kriging, the regionalized variable is
assumed to be stationary.
Kriging is named after the South African engineer, D. G. Krige, who
first developed the method.

Kriging uses the semivariogram, in calculating estimates of the
surface at the grid nodes.
Kriging Interpolation

Punctual (Ordinary) Kriging
If no drift exists and the weights used in the
estimation sum to one, then the estimated
value is said to be unbiased. The scatter of the
estimates about the true value is termed the
error or estimation variance,

=

=
n
i
E
G G
StoV
1
2
2
] * [
) ( o
Kriging Interpolation

The procedures involved in kriging incorporate
measures of error and uncertainty when determining
estimations.
In the kriging method, every known data value and
every missing data value has an associated variance. If
C is constant (i.e. known value exactly), its variance is
zero.
Based on the semivariogram used, optimal
weights are assigned to known values in order to
calculate unknown ones. Since the variogram
changes with distance, the weights depend on
the known sample distribution.
Punctual (Ordinary) Kriging
kriging tries to choose the optimal weights that
produce the minimum estimation error .
Optimal weights, those that produce unbiased
estimates and have a minimum estimation variance,
are obtained by solving a set of simultaneous
equations .
A fourth variable is introduced called the
Lagrange multiplier

WHY KRIGING IS USED?
Kriging is known by the acronym BLUE.
B- Best (because of minimum estimation
variance)
L- Linear (because of weighted arithmetic
average)
U- Unbiased (the weight sum to unity)
E- Estimator
INFLUENCE OF NUGGET EFFECT ON KRIGING WEIGHTS
1 2 3 4 5
+ + + + 100 m +

6 7 8 9 100 m 10
+ + + + +

11 12 13 14 15
+ + + +

16 17 18 19 20
+ + + + +

21 22 23 24 25
+ + + + +


13
+
Kriging weights Co= 0 10 20 30 40 50
C = 50 40 30 20 10 0
a
13
= .57 .37 .27 .19 .11 .04
a
8
=a
12
=a
14
=a
18
= .10 .12 .11 .09 .07 .04
a
7
=a
9
=a
17
=a
19
= .03 .05 .06 .06 .05 .04
a
3
=a
11
=a
15
=a
23
= -.01 .00 .01 .02 .03 .04
a
2
= a
4
= a
6
=a
10
= .00 .00 .01 .01 .03 .04
a
16
=a
20
=a
24
=a
22
= .00 .00 .01 .01 .03 .04
a
1
=a
5
=a
21
=a
25
= -.01 -.01 -.01 .00 .01 .04


Semivariance
Regionalized variable theory uses a related
property called the semivariance to express
the degree of relationship between points on a
surface.

The semivariance is simply half the
variance of the differences between all
possible points spaced a constant distance
apart.
Semivariance is a measure of the degree of spatial
dependence between samples (elevation(
Semivariance
semivariance :The magnitude of the
semivariance between points depends on the
distance between the points. A smaller
distance yields a smaller semivariance and a
larger distance results in a larger semivariance.
Calculating the Semivariance
(Regularly Spaced PointsRegularly Spaced Points(
Consider regularly spaced points distance (d) apart, the semivariance can be
estimated for distances that are multiple of (d) (Simple form):

=
=
+
h
N
i
h i i
h
z z
N
h
1
2
) (
2
1
) (
Semivariance
Z
i
is the measurement of a regionalized variable
taken at location i ,
Z
i+h
is another measurement taken h intervals
away d
N
h
is number of separating distance = number
of points Lag (if the points are located in a
single profile)

=
=
+
h
N
i
h i i
h
z z
N
h
1
2
) (
2
1
) (
Calculating the Semivariance
(Irregularly Spaced PointsRegularly Spaced Points(
Here we are going to explore directional variograms.
Directional variograms is defines the spatial variation among points separated by
space lag h.
The difference from the omnidirectional variograms is that h is a vector rather than
a scalar. For example, if d={d1,d2}, then each pair of compared samples should be
separated in E-W direction and in S-N direction.
Calculating the Semivariance
(Irregularly Spaced PointsRegularly Spaced Points(
In practice, it is difficult to find enough sample points which are separated by exactly
the same lag vector [d].
The set of all possible lag vectors is usually partitioned into classes
Variogram
The plot of the semivariances as a function of
distance from a point is referred to as a
semivariogram or variogram.
Variogram
The semivariance at a distance d = 0 should be zero, because there are no
differences between points that are compared to themselves.
However, as points are compared to increasingly distant points, the semivariance
increases.
Variogram
The range is the greatest distance over which the value at a point on the surface is
related to the value at another point.
The range defines the maximum neighborhood over which control points should be
selected to estimate a grid node.
Variogram (Models(
It is a model semi-variogram and is usually called the spherical model.
a is called the range of influence of a sample.
C is called the sill of the semi-variogram.

>
s
|
.
|

\
|

=
a h C
a h
a
h
a
h
C
h
where
where
3
3
2
1
2
3
) (
Variogram (Models(
Exponential Model
( )
a h
e C h

= 1 ) (
spherical and exponential with the
same range and sill
spherical and exponential with the same
sill and the same initial slope
Spatial Dependency
Geostatistics -
h = 1
( )
2
) ( ) (
2
1
) (

+ = h x z x z h
Spatial Dependency
Geostatistics -
h = 2
( )
2
) ( ) (
2
1
) (

+ = h x z x z h
Spatial Dependency
Geostatistics -
h = 3
( )
2
) ( ) (
2
1
) (

+ = h x z x z h
Spatial Dependency
Geostatistics -
h = 14
( )
2
) ( ) (
2
1
) (

+ = h x z x z h
) (h
h
Range of spatial
dependency
Functional Forms
Common Variogram Models
Interpolating a Surface From Sampled Point Data
Assumes a continuous surface that is sampled
Interpolation
Estimating the attribute values of locations that
are within the range of available data using
known data values
Extrapolation
Estimating the attribute values of locations
outside the range of available data using known
data values
Interpolating a Surface From Sampled Point Data
Estimating a point
here: interpolation
Sample
data
Interpolation
Interpolating a Surface From Sampled Point Data
Estimating a point
here: extrapolation
Sample
data
Extrapolation
Sampling Strategies for Interpolation
Regular Sampling
Random Sampling
Interpolating a Surface From Sampled Point Data
Interpolating a Surface From Sampled Point Data
Elevation profile
Sample
elevation data
A
B
If
A = 8 feet and
B = 4 feet
then
C = (8 + 4) / 2 = 6 feet
C
Linear Interpolation
Interpolating a Surface From Sampled Point Data
Non-Linear Interpolation
Elevation profile
Sample
elevation data
A
B
C
Often results in a
more realistic
interpolation but
estimating missing
data values is more
complex
Interpolating a Surface From Sampled Point Data
Global Interpolation
Uses all known sample points to estimate a value at an
unsampled location
Sample
data
Interpolating a Surface From Sampled Point Data
Local Interpolation
Uses a neighborhood of sample points to estimate a value
at an unsampled location
Sample
data
Uses a local neighborhood to
estimate value, i.e. closest n
number of points, or within a
given search radius
Trend Surface
Global method
Inexact
Can be linear or non-linear
predicting a z elevation value [dependent
variable] with x and y location values
[independent variables]


Trend Surface
1
st
Order Trend Surface
In one dimension: z varies as a linear function of x
x
z
z = b
0
+ b
1
x + e
Trend Surface
1
st
Order Trend Surface
In two dimensions: z varies as a linear function of x and y
z = b
0
+ b
1
x + b
2
y + e
x
y
z
Trend Surface
Geostatistical Case Study On Bauxite,
Surinam.
By:
Dhirendra Pratap Singh
M.Tech. Mineral Exploration
Indian School Of Mines,
Dhanbad.
INTRODUCTION
During 1970s exploration of bauxite was started in Moengo.
Sampling was done on 122x122

m
2
(400x400 ft
2
) grid, drilled with a
10 cm diameter auger drill.
For estimation of resource the traditional polygon-method was
used, based on square blocks of 61x61 m
2
or 30.5x30.5 m
2
f the silica content was more than 12% then it was considered as
waste material. In 1990s with the collaboration with the Adek
University Geostatistics was introduced and to apply Kriging to
61x61m
2
grid to estimate the 30.5x30.5 m
2
mining units.
The weights w
x
for each hole was calculated by Kriging based on the
omnidirectional semivariogram of various parameters. The aim of
the study is to quantify the vertical variation of the various
parameters and to introduce it in a geostatistical model
STUDY AREA
Surinam can be divide into
coastal area and Precambrian
Guiana shield.
The shield occupies 80% of the
country whose Precambrian
basement is highly weathered,
hills are covered by rain forest s.
Coastal area is divided into
10sedimentary formations.
The study area is shaded in the
map.
SAMPLING METHOD.
Sampling was done from 9962 regolith samples from
2412 drill holes
The data was gathered and analysed in the 1970s and
early 1990s by the geological department of Suralco
Sampling starts at the top 1m of the bauxite profile and
continues into the top 1m of kaolin underneath the
bauxite. The samples taken except the last one are
analysed for loss on ignition, and total % silica, titanium
oxide, iron ore, and aluminium oxide by XRf. These all
were done at the company laboratory at Paranam.
RESULTS OF DATA ANALYSIS.
In statistical study only those variables were
include those were considered to be important
for quality control purposes were included.
Accumulation = (z
i
l
i
) (%m)...................(1)
Where z
i
is the grade (%) of the composite drill section, l
i
is the length
(m) of drill section i.
Results were low skewed which indicate the
thickness and accumulation of aluminium oxide, and
positive skewed distribution for accumulation of
silica and iron ore oxide (table 1).

Silica and iron ore oxide accumulation show lack of
spatial correlation along the bauxite horizon (Table
2).






For a sampling distance of 61m, the semivariogram is
the characterised by a pure nugget or a relative
nugget of more than 90%. As a pure nugget effect is
rare for natural phenomena, a spatial correlation
should probably appear if the sampling distance is
reduced. The spatial correlation of the thickness and
the accumulation of LOI, titanium oxide and
aluminium oxide are characterised for the same
sampling distance by partial nugget effect

STATISTICAL ANALYSIS OF DATA
LOI and aluminium oxide have the skewness
coefficients of -0.7 and -0.9, respectively, mean
and median for LOI was 25.3 and 26.8% and 49.4
and 50.8% for aluminium oxide.
Silica and Iron oxide have positively skewed
distributions with skeweness coefficient o 1.7 and
1.2 respectively, mean and median is 10.4 and
4.8% for silica and 12.8 and 8.5% for iron oxide
Titanium oxide is highly skewed and has 0.3
coefficients of skeweness. It mean and median is
2.3 and 2.3% respectively.
It shows that probably the data set do not show
clear evidence of the existence of multiple
populations
When we compare the full data set the histogram and
probability graphs show distinct differences (fig 6c, 7
and 6a ,b)
The histograms of 1
st
, 2
nd
, 3
rd
, and 4
th
, differ from the first and
have greater smoothness and decreases gradually to the higher
values. Mean and median for silica decreases significantly from
5.3 to 20.7% and from 3.1 to 22.5%.
The skewness decreases from 2.52 to -0.1 (Table 3).


It can be observed that silica vs titanium has negative
correlation LOI vs silica, LOI vs iron oxide, silica vs
aluminium iron oxide vs aluminium oxide and iron oxide
vs titanium oxide has negative correlation +uncorrelated
pattern. Aluminium oxide vs LOI and titanium oxide vs
LOI have positive correlation.
For some correlations a significant difference was
noted, such as for first and second there is one bi-
variate distribution that shows a negative correlation
between parameters for second two bi-variate
distributions are present.

ANALYSIS
Calculations are based on the theory of
regionalised variables developed by
mathematicians Matheron (1970). Marechal
and Roullier (1972).
Under normal geological conditions, samples
taken close together are typically more similar
than samples separated by large distances
To quantify these dissimilarities
semivariogram is used.

Figure shows the scatter plots of averages vs standard
deviations from moving windows.


Discussions and Conclusion:

The bottom of a type one lateritic bauxite profile , to the
top, kaolinite progressively gives way to gibbsite , goethite
to haematite , whereas quartz progressively dissolve and
most completely disappears at the top
Titanium remains relatively immobile. In the deposit
studied, this zoning is best demonstrated by the increasing
iron and decreasing silica contents from the bottom to the
top of the profile
Strong decrease in silica is due to the increasing
replacement of kaolinite by gibbsite and the severe
dissolution of quartz.
The spatial variations of loi, titanium and aluminium oxides
do not differ much from that of the accumulations.

By restricting sampling and data processing to distinct
regolith horizons, the 2D spatial correlation between
samples is improved
The thickness of the successive slices was arbitrarily taken
equal to the length of the individual rods (1m).Although
the geological zoning does not corresponds to th1m
intervals or multiples of 1m, the vertical variations could
be quantified, as demonstrated here in a useful way. The
practical feasibility an easy implementation of considering
1m thick slices are in an extensive and industrial oriented
exploration campaign also an important criteria.
As there are three dimensional models is established, the
following study has been started to investigate the effect
on the ore evaluation and mine planning by the improved
geostatistical model a the performance of the model will
be assessed by comparison to producing data.

References
Geostatistical Modelling of lateritic bauxite ore
bodies in Surinam: effects of vertical dimension;
L.Diko, A.Vervoort, I.Vergauwen.
Google.com, Wikipedia.org.
Lecture notes given Prof.B.C. Sarkar for Mtech.
Geostatistics for Natural Resource Modelling:
Editor- B.C.Sarkar
Geostatistics: Concepts and Application In
Mineral deposit For Exploration and mining: By-
B.C.Sarkar