SPE 68704

Methodology to Incorporate Geological Knowledge in Variogram Modeling

A. Bahar, SPE, H. Ates, SPE, Kelkar and Associates, Inc., and M. Kelkar, SPE, University of Tulsa and M. Al-Deeb,
ADCO

Copyright 2001, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the SPE Asia Pacific Oil and Gas Conference and
Exhibition held in Jakarta, Indonesia, 1719 April 2001.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
Geological knowledge is an important ingredient in a
successful reservoir characterization process. Geoscientists
and engineers have used variogram extensively as the tool to
quantify the spatial relationship of various attributes, e.g.,
facies/rock type, porosity, and permeability. Proper variogram
modeling is a key factor to obtain a geologically-sound
reservoir characterization model. This paper discusses the
difficulty that is commonly encountered by many practitioners
in modeling the variogram and proposes a way to incorporate
geological knowledge as the soft information to improve
variogram model.
Common difficulty in variogram modeling is the
calculation of horizontal variogram. The averaging technique
that uses combination of geological knowledge and analogy in
geophysical literature about frequency data analysis is
implemented to solve the difficulty in calculating horizontal
variogram. This technique has produced results that are
agreeable with geology of the reservoir.
The art of incorporating the geological knowledge in
variogram modeling lies in the fact that geological knowledge
is a qualitative measure whereas variogram is a quantitative
measure. The methodology to combine these two measures
presented in this paper is as follows. First, interpreting various
geological aspects of the reservoir in detail. These include, but
not limited to, the interpretations of geological environment,
sequence stratigraphy, pore-space characteristics, iso-chores,
iso-porosity and iso-permeability maps. From these
interpretations, a summary table, that includes the major
continuity direction, lateral extension and anisotropy index of
each attribute, is prepared. Second, calculating experimental
variogram using the Averaging Technique. Third, modeling

the experimental variogram considering the information

obtained from the first step.
The procedure presented above has been implemented as a
routine procedure in several reservoir characterization studies
for both carbonate and sandstone reservoirs in the Middle East
and in the USA. For illustration purposes, comparison of the
realization results, taken from carbonate field study, between
the model with variogram derived purely from the hard data,
i.e., well log data, and the variogram derived from both hard
and soft data, i.e., geological knowledge, is presented. It is
concluded that the incorporation of geological knowledge has
improved the confidence level of the results and should always
be part of any reservoir characterization study.
Introduction
Geoscience data sets are distinguished from other types of data
sets in one important aspect: they exhibit spatial relationship.1
In simple terms, neighboring values are related to each other.
This relationship gets stronger as the distance between two
neighbors becomes smaller. In most instances, beyond a
certain distance the neighboring values become uncorrelated.
This type of qualitative information needs to be defined in a
suitable form so that it can be used to estimate values at
unsampled locations. The most common statistics used to
describe spatial relationship is variogram. It is the most widely
used tool to investigate and model spatial variability of various
reservoir attributes.2 The success in modeling the spatial
relation, via the variogram, will provide higher chance of a
successful reservoir characterization study.
The main role of variogram is to reflect our understanding
of the geometry and continuity of reservoir properties, which
can have an important effect on the predicted flow behavior
and reservoir management decisions. It is the measure of
"geological variability" versus distance; it increases, as
samples become more dissimilar. Therefore, it is clear that
geological knowledge in reservoir charactersization is
incorporated through variogram model. Therefore, thorough
interpretation of geological knowledge should be done prior to
modeling the variogram relationship. The interpretation that is
of interest to the variogram modeling is anything that can lead
to the information of lateral extent and major/minor continuity
directions.
In this paper, the result of the variogram model is used as
the input for simulating geological rock type/facies, porosity

A. BAHAR, H. ATES, M. KELKAR, AND M. AL DEEB

and permeability. The technique used to simulate these

attributes is conditional simulation, which simultaneously
generates petrophysical properties, i.e., porosity and
permeability, consistent with the underlying geological
description, i.e, geological rock-type/facies. The details of this
conditional simulation technique are discussed elsewhere.3
The first step in performing the spatial analysis is to
estimate the value of the variograms using the hard data, i.e.,
well data. These variograms are commonly referred as the
conditioning or experimental variograms. Estimating the
conditioning variograms in practice requires great care and
caution due to some problems such as lack of data pairs at
certain lag distance, e.g., due to well spacing, and biased
sampling, e.g., due to selective well location. Once the
conditioning variogram is estimated, the next step is modeling
the variogram to present it in a convenient format. Certain
restrictions exist in modeling the conditioning variogram.
These restrictions are imposed by the requirement that the
uncertainty in estimation process be represented by positive
variance.
Common difficulty in calculating conditioning variogram
is the calculation of horizontal variogram. This is due the
sparseness of the data in the horizontal direction, which is
governed by well spacing. In this methodology, the averaging
technique is implemented as a solution to solve the difficulty
whenever this problem occurs. This technique has been
discussed in detail previously by Bahar and Kelkar3 and will
be reviewed briefly in this paper.
Modeling variogram involves some uncertainties. One way
to capture this uncertainty is to incorporate the geological
knowledge in modeling the variogram. Before this information
can be incorporated in variogram modeling, a comprehensive
geological interpretation should be performed. The
methodology to incorporate geological information in
variogram modeling will be the most important aspect of this
paper.
The type of knowledge that should be incorporated is the
one that will help in guiding the continuity of the attribute
being investigated. This includes, but not limited to,
geological environment, sequence stratigraphy, pore-space
characteristics, iso-chores, iso-porosity, and iso-permeability.
Other geological information such as analog out-crop may
also be used. The interpretation of these informations should
lead to the direction of major continuity, lateral extension, and
anisotropy index.
For illustration of this methodology, variogram analysis of
several reservoir characterization studies have been conducted
using this methodology. Two case studies, one representing
the carbonate reservoir and the other for sand-stone reservoir,
is presented in this paper.
Approach
As a summary, the procedure to model the spatial analysis
presented in this paper is as follows:
1. Infer geological knowledge.

2.

3.

SPE 68704

Estimate conditioning variogram using hard data.

Apply the averaging technique for horizontal
variogram when necessary
Model the conditioning variogram considering the
geological information as the soft information.

Inferring Geological Knowledge. The art of incorporating

the geological knowledge in variogram modeling lies in the
fact that geological knowledge is a qualitative measure
whereas variogram is a quantitative measure. The
methodology in quantifying the geological knowledge is done
through analysis of the energy level of depositional
environment of the lithofacies. The general rule is higher
energy level corresponds to shorter continuity whereas lower
energy level corresponds to longer continuity.
Commonly, reservoir shows two types of deposition; first
type is the one that is deposited in quite water where mud is
deposited and remain in situ and and the second one is the one
that is deposited in higher energy conditions where mud is
washed out of the sediments, transported in suspension and
deposited elsewhere. The first type is characterized by low
energy condition whereas the second one is characterized by
high energy condition.
The energy level of the deposition can be determined from
the place where it was deposited, e.g., lagoons, bays, open
marine, etc. The shallower the water in each place the higher
the energy will be and vice versa. Therefore, the type of
environment where the lithofacies are deposited would lead us
with some relative information about its continuity. As a
guide, the following environment type summarizes the relative
continuity of lithofacies deposited in it.
Lagoons: an enclosed body of water, usually fairly
shallow (<50 m). This is a high energy environment
Thus, a shorter continuity.
Bays: generally quite well connected to the open sea,
no particular depth connotation. This is a medium
energy environment. Thus, a medium continuity.
Restricted Shelf: connected to the open sea, but with
sufficient barrier to restrict circulation. This is a
medium to low energy environment. Thus, a medium
to long continuity.
Open shelf: No restriction, connected to the main
oceanic mass, i.e., deeper water. This is a low energy
environment. Thus, a long continuity medium.
Additionally, the textural and fabric criteria such as
roundness, sorting and size of the grains may also provide
clues to depositional environment. Sorting is dependent on
grain size, grain shape/type rate of supply of sediments and
intensity of current of wave action. There is a general increase
in degree of sorting with increasing wave or current intensity.
Roundness also generally increases with increase in wave or
current intensity. Better sorting of the grain size will lower the
energy level of the deposition. Thus, it will have longer
continuity. On the hand, poor sorting will result in higher
energy level. Thus, a shorter continuity.

SPE 68704

METHODOLOGY TO INCORPORATE GEOLOGICAL KNOWLEDGE IN VARIOGRAM MODELING

As a general guide, the following list provides the energy

level based on the texture of the grain.
Mudstones and Wackestone
: Low.
Packstone
: Intermediate
Grainstone
: High
The energy analysis presented in the previous paragraphs
provides us with the relative distance of the continuity of the
lithofacies. The term relative has still to be used at this stage
since all of the information is still qualitative in nature. The
absolute distance can be obtained by combining this
information with the information obtained from the hard data,
i.e., well data. This will be explained later in Variogram
Modeling Section.
The directions of the maximum and minimum continuities
can be obtained through contouring of the well data. For
lithofacies, contour plot of the volume proportion at each well
may lead us with the conclusion of these directions.
For petrophysical properties, such as porosity and
permeability, the iso-porosity and iso-permeability maps can
be used for these purposes, i.e., relative distance and
major/minor directions of continuity.
Variogram Estimation. Variogram estimation is the process
to calculate variogram based on the hard data, i.e.,
conditioning variogram. Various geostatistical books and
software are widely available to calculate the conditioning
variogram. Ref. 4 is an example of book and software that can
be used for this purpose. Several commercial softwares are
also widely available. The basic definition of variogram will
not be presented in this paper since it has been defined in
many books and technical papers.
In principle, conditional simulation technique will require
the definition of spatial relationship, i.e., variogram, of each
variable being simulated in three main directions, namely
major, minor, and vertical directions. Both major and minor
variograms are in the horizontal direction.
Estimating vertical variogram is relatively a simple task
since abundant data are sampled in the vertical direction. On
the other hand, estimating the horizontal variogram has been
experienced as a difficult task by many practioners. In this
case, most calculations produce pure nugget variogram only.
Pure nugget variogram is the variogram that does not show
correlation/continuity for any lag distance. Modeling this type
of variogram will result in random distribution only. Since the
geoscience data sets exhibit spatial relationship, it is difficult
to adopt the pure nugget model in the geostatistics model. The
pure nugget variogram is often unavoidable due to the
sparseness of the data in horizontal direction. In other words,
this problem may occur since the data are not close enough to
each other.
In this paper, the Averaging Technique, presented by
Bahar and Kelkar,3 is used to solve this problem. The main
reason of using this technique is due to the fact that this
technique is based on the way the geologist infers the
continuity of an attribute, i.e., by considering the low

frequency information only and filter-out the high frequency

component of the data. In this technique the sill of the
variogram is taken from the vertical variogram whereas the
range of the variogram is evaluated after calculating the
vertical average of the data, which produced 2D data, and
calculating the variogram from these 2D data. The discussion
about the use of the sill calculated from the vertical data can
also be found in Ref. 2. Figure 1 shows an example of pure
nugget variogram and its corresponding result when averaging
technique is used.
Other technique that provides the distance or range of the
maximum continuity can also be adopted. An alternative to
this averaging technique can be found in Ref. 5.
Variogram Modeling. Variogram modeling is the process of
representing the conditioning variogram with an acceptable
mathematical model that can be input in the conditional
simulation technique being used. General procedure in
modeling variogram is presented in the literature. Ref. 1 and 2
are examples that can be used in modeling variogram in
general. This paper will concentrate on modeling variogram
when geological knowledge is available.
For many field case studies, horizontal variogram
calculation has produced conditioning variogram that permits
the modeler to draw infinite models. The geological
information presented previously may help us in reducing the
number of the possible models. The modeling technique
presented in this paper will be explained through an example
as follows.
Figure 2 shows an example where conditioning variogram
may lead the modeler to draw longer variogram range but the
geological information indicates that this lithofacies is
deposited in the shorter continuity medium. Therefore, a better
model would be the one that gives the shortest possible range
that is still supported by the hard data. This example shows
that the modeling is first done by considering the hard data
only. Then, using the soft information, the model is modified
to reflect the depositional environment of corresponding
lithofacies.
Case Study
The methodology described in the previous sections has been
implemented as a routine procedure for several field study.
For illustration purpose, one case study is shown in this paper.
The case study is taken from reservoir characterization
performed on carbonate reservoir from a reservoir in the
Middle East. The field is very heterogeneous and shows a
wide range of different limestone or dolomite facies that
present a complex vertical and lateral organization.
During the geological interpretation step, 11 lithofacies are
defined, namely lithofacies 1A, 1B, 2, 3, 4, 5A, 5B, 6A, 6B, 7
and 8, each being characteristic of a recognizable depositional
environment. These lithofacies have been organized into two
different depositional models or profiles that are developed
successively during deposition of the studied formation. The
two profiles are described as follows.

A. BAHAR, H. ATES, M. KELKAR, AND M. AL DEEB

Profile 1: It is a very flat and low energy profile that can be

assimilated to a ramp setting. The cartoon of this model is
shown in Figure 3. This model was prevailed during the
deposition of the lower part of the reservoir unit. Due to the
flatness of this profile, energy is low. As a consequence of this
low energy level, sedimentary lateral variations are limited.
This is recognizable in the fact that parts of reservoir unit
shows the best facies homogeneity and is easy to correlate
between the wells.
Profile 2: It is a rimmed platform model, where a shoal barrier
separates the open marine domain from a lagoonal domain.
The cartoon of this model is shown in Figure 4. This profile
was prevailed during the deposition of the upper part of
reservoir units. It is a more irregular profile that is
characterized by the presence of a large and continuous shoal
belt all around the studied structure. They are potentially the
best reservoir facies. Due to the great irregularity of the
depositional profile, sedimentary lateral variations are very
high. Thus, a high energy profile. This is expressed in the
difficulty to correlate sub units within that part of the reservoir
units.
The eleven lithofacies defined in the model are only
known at some of the cored-wells. In order to fully use the
electric-logs, which are available at all wells, rock typing
process has been conducted to those lithofacies which resulted
in 8 different Dominant Rock Types (DRT), namely DRT-1
through DRT-8. The rock typing process is done using Electro
Facies Technique.
One of the challenges that is faced in this study is to find
the spatial relationship of the DRT, not the lithofacies, given
the depositional environment and grain texture of the
lithofacies. Therefore, the first step in quantifying the
geological knowledge is to link the DRT with the lithofacies.
Table-1 shows the link between the DRT and lithofacies for
this field. As we can see from this table, each DRT has been
assigned to a possible depositional environment and grain
texture. Combining the information of depositional
environment with the texture of each DRT we obtain the
relative distance of the continuity as shown in this table.
In addition to the analysis of the depositional environment
plot of vertical volume proportion of each DRT at each well is
also prepared to get the direction of the DRT. Figure 5 shows
an example of such plot. In this figure, each point represents
the volume of DRT-4 at each well for one of the reservoir
units. We can see that this DRT is continuous parallel to the
main structural axis of the reservoir.
Figure 6 shows an example of simulated DRT based on
variogram that is modeled using hard data only. The
corresponding result for the model with soft information
included is shown in Figure 7. In this example, we can see
that the soft information has smoothed the result. This
indicates that the soft information has increased the correlation
range of the DRT.
The example for the porosity and permeability are shown
in Figure 8 through 11. As in the case of DRT, soft
information for porosity has increased the correlation range.

SPE 68704

For permeability, we can also see that the direction of the

major direction is slightly different between the hard-data only
and the hard-and-soft model.
Conclusions
Several conclusions can be drawn from the study as follows:
1. Geological knowledge is incorporated into reservoir
characterization
process
through
variogram
modeling.
2. The procedure in variogram modeling consists of
three steps, namely inferring geological knowledge,
variogram estimation, and variogram modeling.
3. The geological knowledge that should be
incorporated is the one that leads to the information
of major continuity (lateral extension, anisotropy
index and direction).
4. The geological knowledge is inferred by analyzing
the energy level of the depositional environment.
Higher energy level corresponds to shorter continuity
and lower energy level corresponds to longer
continuity.
5. Other tool that can be used to build the soft
information is the contouring maps.
6. The averaging technique is implemented to solve the
problem of pure nugget variogram that is commonly
found in calculating horizontal variogram.
7. The use of geological knowledge, in modeling the
variogram, is kept within the limit where the hard
data can be supported, i.e., it should be used as soft
information only.
8. The purpose of incorporating geological information
is to reduce the uncertainty of the spatial relationship.
9. The methodology has been adopted in several field
studies.
References
1. Kelkar, M. Applied Geostatistics for Reservoir
Characterization, Class Notes, The University of
Tulsa, Tulsa, OK, 1996.
2. Gringarten, E., and Deutsch, C. V., Methodology for
Variogram Interpretation and Modeling for Improved
Reservoir Characterization, paper SPE 56654
presented ath the 1999 SPE Annual Technical
Conference and Exhibition, Houston, 3 6 October
1999.
3. Bahar, A. and Kelkar, M., Journey From Well
Logs/Cores
to
Integrated
Geological
and
Petrophysical Properties Simulation: A Methodology
and Application, SPE Reservoir Evaluation & Eng.,
Vol. 3, No. 5, Oct-2000, page 444.
4. Deutsch, C. V, and Journel, A. G, Geostatistical
Software Library and Users Guide, Oxford
University Press, New York, 1992.
5. Pizzaro, J. O. and Lake, L. : A Simple Method to
Estimate Interwell Autocorrelation, paper presented
at the Foourth Intl. Reservoir Characterization
Technical Conference, Houston, (1997).

SPE 68704

6.

7.

METHODOLOGY TO INCORPORATE GEOLOGICAL KNOWLEDGE IN VARIOGRAM MODELING

Robertson
Petroleum
Training
Center:
Characterisation of Carbonate Reservoirs, Short
Course, 13 17 May 2000.
Tinker, S. W., Building the 3-D Jigsaw Puzzle:
Application of Sequence Stratigraphy to 3-D

DRT

RESERVOIR
UNIT

POSSIBLE
LITHOFACIES

R3

1A, 1B

R1

R3

1A

R1

6B

R2

R3

6A

R2

R1

5B

Table-1

TEXTURE
Mudstone /
Wackstone
Wackestone
Mudstone
Wackestone
Packstone
Wackestone
Packstone
Wackestone

R2,R3

5A

Wackestone

R1

6B

Wackestone

R2

R3

Packstone Grainstone
Nodular
Wackestone

Reservoir Characterization, Permian Basin, AAPG

Bulletin, V. 80, No. 4, April 1996. P-460-485.

ENVIRONMENT

CONTINUITY

Open Marine - Lower

Offshore
Lagoonal
Open Marine - Lower
Offshore
Lagoonal - Sub Tidal
Open Marine - Tidal
Influence
Lagoonal - Sub Tidal
Open Marine - Tidal
Influence

Short Range

Porous Limestone

Short Range
Medium to Long
Range

Porous Limestone

Lagoonal - Sub Tidal

Short Range

Porous Calcareous Dolomite

Short Range

Porous Calcareous Dolomite

Short Range

Porous Dolomite

Open Marine - High

Energy
Lagoonal Domain High Enery
Open Marine - High
Energy
Open Marine

Medium to Long
Range
Short Range
Medium to Long
Range
Short Range

GENERAL
LITHOLOGY
Tight Limestone
Low Porosity Limestone
Low Porosity Limestone
Low Porosity Dolomite

Porous Limestone

Short to Medium
Range

Highly Porous Limestone

Medium Range

Nodular Carbonates

Depositional Environment of Lithofacies and Possible Lithofacies of Each Dominant Rock Type

A. BAHAR, H. ATES, M. KELKAR, AND M. AL DEEB

SPE 68704

Hard Data
Hard & Soft Data

a . C o n v e n tio n a l T e c h n iq u e

Figure 2 Two Possible Variogram Models based on Hard

Data and Hard & Soft Data.

b . A v e ra g in g T e c h n iq u e

Figure 1
Conditioning Variogram calculated using
Conventional Technique (a) which produces pure nugget
variogram and Averaging Technique (b) which improves the
structure of the variogram especially at the shorter distance.

Figure 3 Cartoon of Depositional Model for Ramp Platform

Model.

SPE 68704

METHODOLOGY TO INCORPORATE GEOLOGICAL KNOWLEDGE IN VARIOGRAM MODELING

5A

5A
5A

6B

5B

6B

3
5B

5B

6B
1B

5A

DRT-2

1B

5A
3

Sea Level

1A

1A
Fair Weather
Wave Base
Lithofacies

6B

5B

5A

6B

Storm Wave Base

5A

DRT-3
DRT-6

1B
back shore
LAGOONAL DOMAIN

Lagoonal Domain

shore face

upper shore

lower shore

Figure 6 DRT Distribution based on Hard Data Variogram.

OPEN MARINE DOMAIN

Open Marine Domain

Figure 4 Cartoon of Depositional Model for Rimmed

Platform Model.
Structural Axis

DRT-2
DRT-3
DRT-6

Figure 7 DRT Distribution based on Hard and Soft Data

Variogram.

Figure 5 Vertical Volume Proportion of Lithofacies.

A. BAHAR, H. ATES, M. KELKAR, AND M. AL DEEB

SPE 68704

Figure 10 Permeability Distribution based on Hard Data

Variogram.
Figure 8 Porosity
Variogram.

Distribution

based

on

Hard

Data

Figure 11
Permeability Distribution based on Hard and
Soft Data Variogram.
Figure 9 Porosity Distribution based on Hard and Soft Data
Variogram.