Geostatistics: A Review of
Basic Concepts
Univariate Statistics and
Variogram
Geostatistical Reservoir Characterization
Random Variable
l In the stochastic approach (as opposed to
deterministic approach), we treat reservoir
properties as a random variable
l A random variable, z, can take a series of
outcomes or realizations ( zi, i=1, 2, 3,.....N)
with a given set of probability of occurrences
(pi, i =1, 2,...N).
Geostatistical Reservoir Characterization
Distribution Function
Mean
Frequency of
Occurrence
Variance
zi
Geostatistical Reservoir Characterization
Histograms and Cumulative
Distribution Function
250 100.00%
90.00%
200 80.00%
70.00%
Frequency
150 60.00%
Frequency 50.00%
100 Cumulative % 40.00%
30.00%
50 20.00%
10.00%
0 .00%
0.005
10.00
46.42
100.0
215.4
464.2
1000.
0.046
0.215
0.464
1.000
4.642
21.54
0.001
0.002
0.010
0.022
0.100
2.154
Permeability Range, md
Geostatistical Reservoir Characterization
Producing Cumulative Distribution
Function from the Data
• Sort the data in increasing order
X 1 ≤ X 2 ≤ X 3 ≤ ... ≤ X N
• Assign a probability pi to the event ( X ≤ Xi )
• pi =(i-1/2)/N
• Plot Xi versus pi
Geostatistical Reservoir Characterization
PROBABILITY PLOT
Estimated Permeability Data Set 109 Data Points,
Mean = 0.36 Median = 0.10
Calculated Permeability Data Set
10
0.1
0.01
0.001
1 2 5 10 15 20 30 40 50 60 70 80 85 90 95 98 99
Probability , % Less Than
Geostatistical Reservoir Characterization
Statistics Review
Univariate Statistics: Basics
Expected value = Mean = Arithmetic
N
Average
∑ ∑z
1
E( z) = m = pi z i ≅ mˆ = i
N i =1
Variance = a measure of the spread of a
distribution about its mean
VAR( z ) = σ 2z = E ([ zi − m])2 = E ( zi2 ) − m2
N
∑
1
≅σz =
ˆ 2
( z i − mˆ ) 2
( N − 1) i =1
Geostatistical Reservoir Characterization
Coefficient of Variation
•Coefficient of Variation Cv is a dimensionless
measure of spread of the distribution and is commonly
Used to quantify permeability heterogeneity
σ̂
CV =
m
Geostatistical Reservoir Characterization
Cv for Different Rock Types
23
Carbonate (mixed pore type) 22
S.(4)
North Sea Rotliegendes Fm (6) 21
Crevasse splay sst (5) 20
Shallow marine rippled micaceous sst 19
Fluvial lateral accretion sst (5) 18 Very heterogeneous
Distributary/tidal channel Etive ssts 17
Beach/stacked tidal Etive Fm. 16
Heterolitthic channel fill 15
Shallow marine HCS 14
Shallow marine high contrast lamination 13 Heterogeneous
Shallow marine Lochaline Sst (3) 12
Shallow marine Rannoch Fm
11
Aeolian interdune (1)
Shallow marine SCS
10
9
Large scale cross-bed channel (5)
8
Mixed aeolian wind ripple/grainflow (1)
7
Fluvial trough-cross beds (5)
Fluvial trough-cross beds (2)
6
Shallow marine low contrast lamination
5 Homogeneous
Aeolian grainflow (1) 4
Aeolian wind ripple (1)
3
Homogeneous core plugs 2
Synthetic core plugs 1
0
0 1 2 3 4
Cv
Geostatistical Reservoir Characterization
Q-Q / P-P Plots
l Compares two univariate distributions
l Q-Q plot is a plot of matching quartiles
– a straight line implies that the two distributions have the same
shape.
l P-P plot is a plot of matching cumulative probabilities
– a straight line implies that the two distributions have the same
shape.
l Q-Q plot has units of the data,
l P-P plots are always scaled between 0 and 1
Geostatistical Reservoir Characterization
Q-Q plot of permeability vs. porosity
1.E+02
1.E+01
Permeability
1.E+00
1.E-01
1.E-02
1.E-03
0.001 0.010 0.100 1.000
Porosity
Geostatistical Reservoir Characterization
Data Transformation
Why do we need to worry about data transformation?
l Attributes, such as permeability, with highly skewed data
distributions present problems in variogram calculation; the
extreme values have a significant impact on the variogram.
l One common transform is to take logarithms,
y = log10 ( z )
perform all statistical analyses on the transformed data, and
back transform at the end → back transform is sensitive
l Many geostatistical techniques require the data to be
transformed to a Gaussian or normal distribution.
The Gaussian RF model is unique in statistics for its extreme
analytical simplicity and for being the limit distribution of many
analytical theorems globally known as “central limit theorems”
The transform to any distribution (and back) is easily accomplished
by the quantile
transform
Geostatistical Reservoir Characterization
Normal Scores Transformation
l Many geostatistical techniques require the data to be
transformed to a Gaussian or normal distribution:
Geostatistical Reservoir Characterization
Standard Normal Distribution
z = (w-µ)/σ
1
Cum. Normal 0.9
pdf Normal 0.8
0.7
0.6
0.5 0.6827
0.4 0.9545
0.3
0.2 0.9973
0.1
0
-3 -2 -1 0 1 2 3
Geostatistical Reservoir Characterization
Exercises
l Univariate analysis of well log data
l Distribution Characteristics
l Heterogeneity Measures
Geostatistical Reservoir Characterization
Statistics Review
Bivariate Statistics
The Covariance and the Variogram are
related measures of the joint variation of
two random variables.
Geostatistical Reservoir Characterization
Statistics Review
Covariance
COV ( A, B ) = E ([ Ai − mA ][ Bi − mB ]) = E ( Ai Bi ) − mA mB
N
∃ 1
≅ CAB =
N
∑ ( a b ) − m∃ m∃
i =1
i i A B
>0 if A, B are positively correlated
CAB = 0 if A, B are independent
< 0 if A, B are negatively correlated
Geostatistical Reservoir Characterization
Statistics Review
Variogram
2γ ( A, B ) = E ([ A − B]2 )
N
1
≅ 2γ∃ =
N
∑ (a
i =1
i − bi ) 2
γÙ0 A is increasingly similar to B
γÙ∝ A is increasingly dissimilar to B
Geostatistical Reservoir Characterization
Spatial Variation
Assume:
Variation in a property between two points
depends only on vector distance, not on
location.
Model Variability:
Nh
Variogram γ ( h ) = 1
2N h
∑ i
[ζ (
i =1
x ) − ζ ( x i + h )]2
1 Nh
Covariance c( h ) =
Nh
∑ ζ ( x i )ζ ( x i + h )
− m2
i =1
Geostatistical Reservoir Characterization
Modeling Spatial Variation
l zi =z(xi) is some property at location xi
l Interpret zi as a random variable with a
probability distribution and the set of zi
to define a random function z.
l Assume the variability between z(xi)
and z(xi+h) depends only on vector h,
not on location xi*.
Geostatistical Reservoir Characterization
Modeling Spatial Variation
l Use variogram and/or covariance to
model variability
Nh
∑
1
2γ (h) = 2γˆ ( h) = [ z ( xi ) − z ( xi + h)]2
Nh i =1
1 Nh
COV ( h) = cˆ(h) =
Nh
∑
i =1
z ( xi ) z ( xi + h) − mˆ z2
Geostatistical Reservoir Characterization
Data Sources
l Lots of wells in subject reservoir
l Lots of wells in similar reservoir
l Outcrops
l Secondary and soft data (seismic, interval
constraints, expert judgement)
Geostatistical Reservoir Characterization
11600
Porosity Log Depth Porosity
11060 0.083
11060.5 0.074
11061 0.062
11500
11061.5 0.058
11062 0.061
11062.5 0.066
11400 11063 0.07
Depth, ft
11063.5 0.073
11064 0.078
11300 11064.5 0.079
11065 0.075
11065.5 0.072
11200 11066 0.072
11066.5 0.074
11067 0.075
11100 11067.5 0.077
11068 0.098
11068.5 0.129
11069 0.151
11000
0 0.1 0.2 0.3 0.4 11069.5 0.157
Porosity, fraction
Geostatistical Reservoir Characterization
Variogram Calculation
φ (u) φ (u+h) φ (u) φ (u+h)
0.083 0.074 0.083 0.062
0.074 0.062 0.074 0.058
0.062 0.058 0.062 0.061
0.058 0.061 0.058 0.066
0.061 0.066 0.061 0.07
0.066 0.07 0.066 0.073
0.07 0.073 0.07 0.078
0.073 0.078 0.073 0.079
0.078 0.079 0.078 0.075
0.079 0.075 0.079 0.072
0.075 0.072 0.075 0.072
0.072 0.072 0.072 0.074
0.072 0.074 0.072 0.075
0.074 0.075 0.074 0.077
0.075 0.077 0.075 0.098
0.077 0.098 0.077 0.129
0.098 0.129 0.098 0.151
0.129 0.151 0.129 0.157
0.151 0.157
Geostatistical Reservoir Characterization
0.35
Variogram Calculation 0.35
0.3 0.3
R 2 = 0.9812 R 2 = 0.8761
0.25 0.25
0.2 0.2
0.15 0.15
0.1 0.1
Lag=0.5 Lag=1.5
0.05 0.05
0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
0.35 0.35
R 2 = 0.7653 R2 = 0.352
0.3 0.3
0.25 0.25
0.2 0.2
0.15 0.15
Lag = 10
0.1 0.1
Lag=2.5
0.05 0.05
0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
l As the separation distance increases, the similarity between
pairs of values decreases
Geostatistical Reservoir Characterization
Variogram Definition
1.2
Sill - No correlation
Increasing variability
0.8
Variogram
Range
0.6
Model Fit
0.4 Experimental
Nugget
Effect
0.2
0
0 5 10 15 20 25 30 35 40 45 50
Distance
Geostatistical Reservoir Characterization
Variogram Model
Variogram improves with increasing:
- Number of data pairs at each lag spacing.
- Number of lags with data.
è Lots of data required for statistically
significant variogram.
Geostatistical Reservoir Characterization
Variogram Terminology
l Sill
– the variance of the data (1.0 if the data are normal scores)
– The plateau that the variogram reaches at the range
l Range
– As the separation distance between pairs increases, the corresponding variogram
value will generally increase. Eventually, an increase in the separation distance no
longer causes a corresponding increase in the averaged squared difference between
pairs of values.The distance at which the variogram reaches this plateau is the range
l Nugget effect
– natural short-range variability (microstructure) and measurement error
– Although the value of the variogram for h=0 is strictly 0, several factors, such as
sampling error and short term variability, may cause sample value separated by
extremely short distances to be quite dissimilar. This causes a discontinuity from the
value of 0 at the origin to the value of the variogram at extremely small separation
distances
Geostatistical Reservoir Characterization
Variogram Characteristics
γ γ
h h
Low Spatial Correlation High Spatial Correlation
γ α1
α2
α3
h
Anisotropic
All geological inference is buried in the variogram.
Geostatistical Reservoir Characterization
Variograms
Modeling Spatial Correlation
l The shape of the variogram model determines the spatial continuity of the random
function model
l Measures must be customized for each field and each attribute (φ,Κ)
l Depending on the level of diagenesis, the spatial variability may be similar within similar
depositional environments.
Geostatistical Reservoir Characterization
Variogram and Covariance
l Assuming second order stationarity, the
following relationship applies.
γ (h) = var(z)− cov(h)
⇒ cov(h) = cov(0)−γ (h)
l These are important relationships to be
used during kriging using variograms.
Geostatistical Reservoir Characterization
Variogram Interpretation
Geometric Anisotropy
Same shape and sill but different ranges
Geostatistical Reservoir Characterization
Variogram Interpretation
Cyclicity
1 3 Sill
1
2 4
Distance
φ
1 2
3
Depth
Geostatistical Reservoir Characterization
Variogram Interpretation
Cyclicity
Geostatistical Reservoir Characterization
Variogram Interpretation
Zonal Anisotropy
Both sill and range vary in different directions
1
Variability ‘between wells’
‘Within well’ variability
Positive correlation over large distance
Well 1 Well 2 Well 3
Geostatistical Reservoir Characterization
Variogram Interpretation
Zonal Anisotropy
Geostatistical Reservoir Characterization
Variogram Interpretation
Trend
Negative
1 correlation
Positive
correlation
Distance
Trend » non stationarity
Depth
the mean is not constant
φ
Geostatistical Reservoir Characterization
Variogram Interpretation
Vertical Trend and Horizontal
Zonal Anisotropy
Geostatistical Reservoir Characterization
Vertical Well Profile and
Variogram with a Clearly Defined
Vertical Trend
50
45
Regression:
y = -1.5807x + 51.611
40
35
30
Depth
25
20
15
10
0
0 5 10 15 20 25 30
Porosity
Geostatistical Reservoir Characterization
Vertical Well Profile and
Variogram after Removal of the
Vertical Trend
50
45
40
35
30
Depth
25
20
15
10
0
-8 -6 -4 -2 0 2 4 6 8
Re siduals
Geostatistical Reservoir Characterization
Methodology for Variogram
Interpretation and Modeling
l Compute and plot experimental variograms in what are believed to be the principal
directions of continuity based on a-priori geological knowledge
l Place a horizontal line representing the theoretical sill.
l Remove all trends from data.
l Interpretation
– Short-scale variance: the nugget effect
– Intermediate-scale variance: geometric anisotropy.
– Large-scale variance:
• zonal anisotropy
• hole-effect
l Modeling
– Proceed to variogram modeling by selecting a model type (spherical, exponential,
gaussian…) and correlation ranges for each structure
Geostatistical Reservoir Characterization
Exercises
l Vertical variogram calculations
l Areal variogram calculations
l Variogram modeling
l Inference of spatial
variation/correlations
Geostatistical Reservoir Characterization