Variogram Calculation

GEOMODELING AND RESERVOIR

MANAGEMENT

JAVID SHIRIYEV, PHD

Docent, Baku Higher Oil School
Introduction
• Geostatistics focuses on natural phenomena which are correlated in space, a feature of
all natural phenomena.
• Geostatistical modeling requires quantitative measures of spatial correlation for
estimation and simulation, and the most used tool for measuring spatial correlation is
the semi-variogram.
• For variogram calculation, it is essential to work with data that are free of outliers,
trends, and are oriented in an appropriate coordinate system.
• We will begin with a qualitative look at the variogram, followed by the parameters of the
variogram, then a quantitative look at variograms, and conclude with variogram
interpretation and modeling.

Mar-24 Petroleum Geoscience 2

Variogram is chart that converts distance to correlation.

Mar-24 Petroleum Geoscience 3

Semi-Variogram
• To quantify spatial continuity, we use variogram which is a
measure of dissimilarity vs. distance.
• In probabilistic notation, it is defined as the expected value:
2
2𝛾 𝐡 = 𝐸 𝑍 𝐮 − 𝑍 𝐮 + 𝐡

• The semi-variogram is one half of the variogram, and if we use

the definition of expected value semi-variogram becomes:
𝑁 𝐡
1 2
𝛾 𝐡 = ෍ 𝑍 𝐮𝑖 − 𝑍 𝐮𝑖 + 𝐡
2𝑁 𝐡
𝑖=1

• It is a function obtained by plotting the variance of the

differences between the properties at two locations separated
by a distance ℎ.

Mar-24 Petroleum Geoscience 4

Other measures of spatial correlation
𝐌𝐚𝐝𝐨𝐠𝐫𝐚𝐦: 𝐆𝐞𝐧𝐞𝐫𝐚𝐥 𝐑𝐞𝐥𝐚𝐭𝐢𝐯𝐞 𝐕𝐚𝐫𝐢𝐨𝐠𝐫𝐚𝐦:

𝑁 𝐡
1 𝛾 𝐡
𝑀 𝐡 = ෍ 𝑍 𝐮𝑖 − 𝑍 𝐮𝑖 + 𝐡 𝛾GR 𝐡 =
2𝑁 𝐡 𝑚2 𝐡
𝑖=1

𝐑𝐨𝐝𝐨𝐠𝐫𝐚𝐦: 𝐏𝐚𝐢𝐫𝐰𝐢𝐬𝐞 𝐑𝐞𝐥𝐚𝐭𝐢𝐯𝐞 𝐕𝐚𝐫𝐢𝐨𝐠𝐫𝐚𝐦:

𝑁 𝐡
1 𝑁 𝐡 2
𝑅 𝐡 = ෍ 𝑍 𝐮𝑖 − 𝑍 𝐮𝑖 + 𝐡 1 𝑍 𝐮𝑖 − 𝑍 𝐮𝑖 + 𝐡
2𝑁 𝐡 𝛾PR 𝐡 = ෍ 2
𝑖=1 2𝑁 𝐡
𝑖=1
𝑍 𝐮𝑖 + 𝑍 𝐮𝑖 + 𝐡
2

Mar-24 Petroleum Geoscience 5

How to Calculate the Variogram
• The number of data pairs decreases as lag distance
increases.
• Beyond ℎ = 𝑁Δ𝑥 Τ2, the reduction in data pairs
causes the variogram to fluctuate excessively with no
significance.
𝑁 𝑛!
• Therefore, the variogram is typically truncated = 𝐶 𝑛, 2 =
2 2! 𝑛 − 2 !
beyond ℎ = 𝑁Δ𝑥 Τ2.
𝑁
• If there are N data points, there will be data pairs
2
in the set.
• The variogram can be computed even for irregularly
spaced data.

Mar-24 Petroleum Geoscience 6

Directional Variogram
• If the spatial variance does not vary along with spatial
directions, this is called isotropy and, if does, this is
called anisotropy.

Mar-24 Petroleum Geoscience 7

Calculating Experimental Variogram
• Lag distance
• Lag tolerance
# If the data is irregularly spaced, there is no way we
can get pairs of values exactly stated by lag distance.
Defining a lag tolerance ensures that we can get
enough pairs for a particular lag distance.
• Lag number
# It is chosen such that the total distance is about one
half of the reservoir size in the direction being
considered.
• Anisotropy
# Direction angle
# Angle tolerance
# Bandwidth

Mar-24 Petroleum Geoscience 8

Choice of lag Tolerance
• This choice is appropriate when too • This choice is appropriate when too
many data is available: few data is available:

Mar-24 Petroleum Geoscience 9

Calculating Experimental Variogram
• Example: looking at lag
distance of 4
• Start at a given node and
compare value to all nodes
which fall in the lag and
angle tolerance.
• Repeat computation at next
node until all nodes have
been computed.

Mar-24 Petroleum Geoscience 10

Calculating 3D Directional
Experimental Variogram
• Dip angle
• Dip angle tolerance
• Dip bandwidth
• Azimuth angle 𝜑
• Azimuth angle tolerance Δ𝜑
• Azimuth bandwidth
• Lag distance Δℎ
(can be direction specific)
• Lag tolerance Δℎtol
(half lag distance)
• Number of lags
(max distance / lag distance)

Mar-24 Petroleum Geoscience 11

Experimental Variogram – Calculation
Non-Uniform Spatial Distribution

Mar-24 Petroleum Geoscience 12

Experimental Variogram – Ex 1
0 10 20 30
• In the table, total organic content (TOC) from 16 equally
spaced wells is summarized. First column and row provide 0
30 10 12 9 8
spatial data of the TOC.
10 20 17 10
20 7
• Calculate the variogram in the north-south direction at lag
distance of 20 ft. 20 24 20 16 12
10
• Calculate the variogram in the northeast-southwest direction 30
0 32 28 12 18
at lag distance of 20 2 ft.
0 10 20 30
0 10 12 9 8
𝑁 𝐡
1 2
10 20 17 10 7
𝛾 𝐡 = ෍ 𝑍 𝐮𝑖 − 𝑍 𝐮𝑖 + 𝐡
2𝑁 𝐡 20 24 20 16 12
𝑖=1

30 32 28 12 18

Mar-24 Petroleum Geoscience 13

Experimental Variogram – Ex 1
0 10 20 30
• In the table, total organic content (TOC) from 16 equally
spaced wells is summarized. First column and row provide 0
30 10 12 9 8
spatial data of the TOC.
10 20 17 10
20 7
• Calculate the variogram in the north-south direction at lag
distance of 20 ft. 20 24 20 16 12
10
• Calculate the variogram in the northeast-southwest direction 30
0 32 28 12 18
at lag distance of 20 2 ft.
0 10 20 30
𝛾 20, N − S 0 10 12 9 8
1
= ሾ 10 − 24 2 + 20 − 32 2 + 12 − 20 2
10 20 17 10 7
2∙8
+ 17 − 28 2 + 9 − 16 2 + 10 − 12 2 20 24 20 16 12
+ 8 − 12 2 + 7 − 18 2 ሿ = 44.69
30 32 28 12 18

Mar-24 Petroleum Geoscience 14

Experimental Variogram – Ex 1
0 10 20 30
• In the table, total organic content (TOC) from 16 equally
spaced wells is summarized. First column and row provide 0
30 10 12 9 8
spatial data of the TOC.
10 20 17 10
20 7
• Calculate the variogram in the north-south direction at lag
distance of 20 ft. 20 24 20 16 12
10
• Calculate the variogram in the northeast-southwest direction 30
0 32 28 12 18
at lag distance of 20 2 ft.
0 10 20 30
0 10 12 9 8
𝛾 20 2, NE − SW 10 20 17 10 7
1
= ∙ 9 − 24 2 + 8 − 20 2
+ 10 − 32 2
+ 7 − 28 2
20 24 20 16 12
2∙4
= 161.75 30 32 28 12 18

Mar-24 Petroleum Geoscience 15

Experimental Variogram – Ex 2
• Plot the experimental variograms in the E-W, N-S, NE-SW and NW-
SE directions. Perform the calculations up to and including a lag
distance of 4 units. The samples are on a square grid with 1 unit 26 22 19 14 16 19 16 14
size.
23 20 17 20 14 23 21 17
• Do the variograms show any evidence of anisotropy in the porosity
data? 22 17 18 19 18 25 20 19
21 15 20 18 20 20 18 13
19 18 15 15 18 23 22 20
18 16 10 16 14 18 20 18
17 14 10 13 13 15 14 17
15 13 11 10 17 16 15 11

Mar-24 Petroleum Geoscience 16

Experimental Variogram – Interpretation
• Sill
# The variance of variable of interest, the
maximum dissimilarity
# In stationary case the sill equals to the
priory variance.
• Range covariance
# The lag distance to reach the sill
• Nugget effect
# The nugget effect is due to measurement
error and geological small-scale structures,
that is, features occurring at a scale smaller
than the smallest data separation distance.

Mar-24 Petroleum Geoscience 17

Experimental Variogram – Interpretation

• Cyclicity
# It may be linked to underlying geological
periodicity
# Could be due to limited data

Mar-24 Petroleum Geoscience 18

Experimental Variogram – Interpretation
• Trends
# All geological processes impart a trend in the
petrophysical property distribution.
# For example, the sedimentation process presents a
systematic decrease in reservoir quality from proximal to
distal portions of the depositional system.
# Presence of trends in the data leads to failure of the
stationarity assumptions, hence traditional geostatistical
models (kriging, etc.) cannot be used.
# Systematic trend in values (e.g., linear dependence in
one direction, or variogram increase above the sill) must
be removed prior to variogram modelling and estimation.
# Trend is added back to the estimated or simulated values
at the end of the study.

Mar-24 Petroleum Geoscience 19

Experimental Variogram – Trend Effect
Data Variogram
Geostatistical models with trend removal:
Raw data
 Kriging with trend (universal)
with trend  Kriging with External drift (external variable)
 Locally varying mean (LVM)
 Moving window residual kriging (local predictions)
 Intrinsic Random function of order k (higher moments)
 Residual methods with non-linear trend models
(Artificial Neural Networks, etc.)

Residuals after Periodicity becomes

removing a visible after the trend is
linear trend removed

Trend removal methods:

• polynomial
• spline
• non-linear models

Mar-24 Petroleum Geoscience 20

Experimental Variogram – Interpretation
• Anisotropy
# Geometric anisotropy
# Zonal anisotropy: the range of
correlation in one direction exceeds the Geometric
field size.
# Zonal anisotropy - areal stratification
# Zonal anisotropy - vertical stratification
or layering

Zonal

Mar-24 Petroleum Geoscience 21

Experimental Variogram – Interpretation
Areal Stratification
• Anisotropy
# Geometric anisotropy
# Zonal anisotropy: the range of
correlation in one direction exceeds the
field size.
Layering
# Zonal anisotropy - areal stratification
# Zonal anisotropy - vertical stratification
or layering

Mar-24 Petroleum Geoscience 22

Variogram Models
Pure Nugget model Sources of Nugget:
• No spatial correlation; • small-scale variability
• measurement errors
• Interpolation has
• positioning errors
NO significance.

 0, h=0
 ( h) = 
co , h0

Mar-24 Petroleum Geoscience 23

Variogram Models
𝛾 0 = 0 for all models
𝛼
1) 𝛾 ℎ = 𝐶o + 𝐶 − 𝐶o ℎ 0≤𝛼≤2
1. Power Model
3 ℎ 1 ℎ 3
2. Spherical Model 𝐶 + 𝐶 − 𝐶o − 0< ℎ ≤𝑎
2) 𝛾 ℎ = ቐ o 2 𝑎 2 𝑎
𝐶 ℎ >𝑎
3. Exponential Model
ℎ
3) 𝛾 ℎ = 𝐶o + 𝐶 − 𝐶o 1 − exp −3 𝑎
4. Gaussian Model
ℎ 2
4) 𝛾 ℎ = 𝐶o + 𝐶 − 𝐶o 1 − exp −3
5. Hole Effect 𝑎

sin ℎΤ𝑎
5) 𝛾 ℎ = 𝐶o + 𝐶 − 𝐶o 1 − ℎΤ𝑎

Mar-24 Petroleum Geoscience 24

Variogram Models
Power model
• reflects correlation at all scales
- fractal behaviour
• characterises trends

 0, h=0
 ( h) =  
ch , 0    2, h0

Mar-24 Petroleum Geoscience 25

Variogram Models
1. Power Model

2. Spherical Model
# It is a commonly encountered variogram shape.
# Shows linear behavior near the origin.
# Reaches its sill with zero derivative.

3. Exponential Model

4. Gaussian Model

5. Hole Effect

Mar-24 Petroleum Geoscience 26

Variogram Models

1. Power Model
2. Spherical Model
3. Exponential Model
# variogram reaches 95% of sill at
the effective correlation distance 𝑎

3. Gaussian Model
4. Hole Effect

Mar-24 Petroleum Geoscience 27

Variogram Models
1. Power Model

2. Spherical Model
3. Exponential Model

4. Gaussian Model
# represents very smooth behavior of functions at
short distances;
# if exhibits a nugget effect, it usually depict errors of
measurements;
# a small nugget c0 is usually added to the Gaussian
model to avoid numerical instabilities and artefacts

3. Hole Effect

Mar-24 Petroleum Geoscience 28

Other Variogram Models

Cubic model
• combination of a power models limited by
a constant beyond the correlation range 𝑎
• rarely used

c(7( ) − ( ) + ( ) − ( ) ), 0  h  a
h 2 35 h 3 7 h 5 3 h 7
 ( h) =  a 4 a 2 a 4 a

 c, ah
Mar-24 Petroleum Geoscience 29
Other Variogram Models
Hole effect model
• represents periodic structures (e.g. objects)
• acts in one direction only

 (h) = 1.0 − cos( h )

Dampened Hole effect model
• a product of the exponential covariance and the hole effect;
• more common than pure hole effect

 − ha 
 (h) = 1.0 − exp  cos(ha)
 d 
Mar-24 Petroleum Geoscience 30
Other Variogram Models (From 𝐃𝐮𝐛𝐫𝐮𝐥𝐞, 2003)

Mar-24 Petroleum Geoscience 31

 (From 𝐃𝐮𝐛𝐫𝐮𝐥𝐞, 2003)

Mar-24 32
Variogram Modelling Steps
• Scan through different direction (rose, sectors)
• Pick an appropriate direction
• Simple to model (consider orthogonal direction)
• Drift is negligible (directional)
• Choose the appropriate scale (vary lag distance and number)
• Quantify the goodness of variogram model fit
• Account for the number of pairs when fitting a model
• Use cross-validation to compare results with different variogram models
• Do not over-work the variogram model fitting
• Sensitivity analysis of the model parameters

Mar-24 Petroleum Geoscience 33

References
• Caers J., Petroleum Geostatistics, SPE, 2005.
• Jensen J.L., Lake L.W., Corbett P., Goggin D. Statistics for Petroleum Engineers and Geoscientists,
Gulf professional publ., 2000.
• Deutsch C.V. and Journel A.G. (1998) GSLIB. Geostatistical Software Library and User’s Guide. N.Y.,
Oxford University Press.
• Deutsch C.V. (2002) Geostatistical Reservoir Modeling. Oxford University Press.
• Goovaerts P. (1997). Geostatistics for Natural Resources Evaluation. N.Y, Oxford University Press.
• Dubrule, O.R.F., Thibaut, M, Lamy, P. et al. (1998) Geostatistical reservoir characterization
constrained by 3D seismic data. Petroleum Geoscience 4(2), 121-128.
• Kanevski, M., Maignan M., Analysis and Modelling of Spatial
Environmental Data. Dekker/PPUR, 2004, 304 pages, CD-ROM
• Kovalevskiy E. (2011) Geological modelling on the base of geostatistics, EAGE
• S. Strebelle (2002) Conditional simulation of complex geological structure using multiple-point
statistics. Math. Geology, v. 34, p. 1-22.

Mar-24 Petroleum Geoscience 34