You are on page 1of 402

Petrel Quantitative Interpretation

Training and Exercise Guide


Petrel 2015

Schlumberger
Copyright Notice
Copyright© 2015 Schlumberger. All rights reserved.
This work contains the confidential and proprietary trade secrets of
Schlumberger and may not be copied or stored in an information
retrieval system, transferred, used, distributed, translated or
retransmitted in any form or by any means, electronic or mechanical, in
whole or in part, without the express written permission of the
copyright owner.

Trademarks & Service Marks


Schlumberger, the Schlumberger logotype, and other words or symbols
used to identify the products and services described herein are either
trademarks, trade names or service marks of Schlumberger and its
licensors, or are the property of their respective owners. These marks
may not be copied, imitated or used, in whole or in part, without the
express prior written permission of Schlumberger. In addition, covers,
page headers, custom graphics, icons, and other design elements may
be service marks, trademarks, and/or trade dress of Schlumberger, and
may not be copied, imitated, or used, in whole or in part, without the
express prior written permission of Schlumberger. Other company,
product, and service names are the properties of their respective
owners.
An asterisk(*) is used throughout this document to designate.a mark of
Schlumberger.
Table of Contents
Copyright Notice ..................................................................................................................... 2

About this manual .....................................................................................................................9


Prerequisites ...........................................................................................................................9
Learning objectives ................................................................................................................. 9
What you need .....................................................................................................................10
What to expect ..................................................................................................................... 11
Course conventions .............................................................................................................. 12
lcons ......................................................................................................................................13
More Petrel training courses ................................................................................................14
Summary ............................................................................................................................... 15

Module 1- Rock physics ....................................................................................................17


Learning objectives ...............................................................................................................17
Lesson 1 -The Ql global workflow ....................................................................................18
Lesson 2 -Introduction to Rock physics ............................................................................21
Lesson 3- Rock physics review .........................................................................................26
Density .......................................................................................................................... 27
Pand Swaves .............................. ................................................................................28
Velocity relationships ...................................................................................................31
Elastic Moduli: Hooke's Law .........................................................................................32
Fluid substitution background ......................................................................................38
Gassmann's equation ....................................................................................................38
Fluid mixture formulas ................................................................................................. .40
Effects of saturation .....................................................................................................42
Empirical rock property relationships: Pwave velocity and density, Vp to Vs ............ .43
Castagna mudrock line .................................................................................................43
Castagna-G reenberg .....................................................................................................44
What happens at a reflecting boundary? .....................................................................44
Zoeppritz equations ............................................................................................. ......... 46
Lesson 4- Rock physics workflow .....................................................................................47
Lesson 5- Basic rock physics studies .............................................................................. .49
Procedure- Access the Quantitative Interpretation .................................................49
Procedure- Set up a rock physics study for Elastic parameter estimation .............. 50
Procedure- Display the results of a rock physics study in a Well section window 55
Procedure- Edit a rock physics study with the Inspector ......................................... 57
Crossplots for elastic property estimates .................................................................... 58
Procedure- Crossplot elastic property estimates .....................................................59
Exercise 1 -Estimate and crossplot elastic parameters from a rock physics study.64
Quantitative Interpretation Table of Contents • 3
Procedure- Create classifications ............ ................................................................. 66
Exercise 2- Create an interactive cross plot ............................................................68
Procedure- Create a 30 crossplot ............................................................................ 69
Lesson 6- Fluid substitutions in rock physics .. ........ ........................................................ .71
Procedure- Perform a fluid substitution ..................................................................71
Exercise 1 -Perform a fluid substitution .................................................................. 80
Lesson 7- Shear velocity estimation from empirical relationships .................................. 82
Procedure- Estimate shear velocities from Pwave velocities ................................ 83
Exercise 1 (Optional)- Run a shear velocity estimation ........................................... 85
Results and discussion ...................................................................................... ........... 86
Lesson 8- Log blocking Backus workflow .........................................................................87
Procedure- Run the Log blocking Backus workflow .................. .................... ........... 88
Exercise 1 -Perform Backus log blocking ................................................................. 90
Review questions ................................................................................................................. 93
Summary ............................................................................................................................... 93

Module 2- AVO modeling .................................................................................................... 95


Prerequisites .................................................................................. ........... ............................ 95
Learning objectives ........ .. ..................................................................................................... 95
Lesson 1 -AVO modeling workflow .................................................................................. 96
Lesson 2- AVO approximations ............................ ............................................................ 98
Aki and Richards .. .. .....................................................................................................100
Shuey's 3-term and 2-term approximations ............................................................... 101
Verm and Hilterman ................... ....................... ................................. .... ..................... 103
Gidlow approximation ................................................................................................104
Fatti ............................................................................................................................ .106
Pan and Gardner ..................... ...................................................... .............................. 106
Summary .....................................................................................................................108
Lesson 3- AVO classes .... ................. ......... ............................................................... ...... 112
Lesson 4- Fitting of AVO approximations to seismic data ............................................. 114
Lesson 5- AVO Modeling study ...................................................................................... 115
Procedure -Import SEG-Y Prestack data .................................................................116
Procedure- Create an AVO synthetic model .............................. .......................... ... 118
Procedure- Create attributes from an AVO synthetic and from the observed
gather .......................................................................................................................... 124
Exercise 1 -Set up and run an AVO study ............................................................... 131
Procedure- Create crossplots of AVO attributes and amplitude versus angle ...... 134
Procedure- Create an amplitude versus offset crossplot.. ..................................... 135
Exercise 2- Crossplot attributes .................................... .... .. ............ ....................... 139
Exercise 3- Create AVO synthetics for different fluid saturations .... .................... 140
4 • Table of Contents Quantitative Interpretation
Observations on the effect of fluid saturations on the seismic response ........... ......143
Lesson 6- Wedge modeling ............................................................................................ 147
Procedure- Create prestack and poststack wedge models ................................. ... 150
Exercise 1 -Create a wedge model ........................................................................ 158
Review questions ............................................................................................................... 159
Summary .... ........... ... ...........................................................................................................159

Module 3- AVO reconnaissance ....................................................................................161


Prerequisites ........................ ...............................................................................................161
Learning objectives ............................................................................................................. 161
Lesson 1 -AVO reconnaissance too1 ...............................................................................162
Procedure- Set up and run the AVO reconnaissance workflow .............................165
Volumes of interest.. ................................................................................................... 172
Procedure- Create a VOl to co-visualize AVO attribute data with other seismic
data in the Interpretation window ..... ... ........ ............................................................. 174
Procedure- Display VOis in a 3D window ..............................................................177
Procedure- Create a crossplot of AVO attribute data from an AVO
reconnaissance VOI ....................................... .................................. ........................... 178
Exercise 1 -Apply the AVO reconnaissance at a prospect .....................................181
Results I Discussion ................................................................................................... 183
Review question ................................................................................................................. 185
Summary ............................................................................................................................. 185

Module 4- Simultaneous seismic inversion and the Inversion property builder .187
Prerequisites ....... ........ .................... .. .... ....................... .......................................................187
Learning objectives ........................................... .... .... ... .......................................................187
Lesson 1 -What is seismic inversion? .................................................... ........................ 188
Applications of simultaneous inversion ..................................................................... 191
Simultaneous inversion workflow .................................. .................... ........................ 192
Input data .... ........................... ....... ................. ............................................................. 192
Output .......................... ... ................... .... ........ ............................................................. 193
Key features ............................... ................................................................................. 193
Simultaneous inversion features and implementation in Petrel ...............................195
Penalty function ................ .. ................ .. ...................................................................... 196
User controls ..................................................................................................... .......... 198
Lesson 2- Inversion preparation: Well tie and wavelet extraction ................................ 202
Seismic trace alignment ......................................... .................................................... 203
Log conditioning/editing ....................................... ............................................ .......... 205
Procedure- Extract a deterministic wavelet.. ......................................................... 206
Predictability displays ............. ............ .............. ..........................................................214
Quantitative Interpretation Table of Contents • 5
General extraction guidelines .................................................................................... 215
Exercise 1 -Extract a deterministic wavelet ..........................................................217
Align seismic volumes with non-rigid matching (NRM) ............................................. 218
Exercise 2 (Optional)- Align data cubes using seismic trace alignment.. .............. 224
Procedure- Condition log data with the Log conditioning tool .............................. 225
Exercise 3 (Optional)- Use the Log conditioning toolbox ....................................... 229
Lesson 3- Low frequency model building and OC with the Inversion property builder .230
Procedure- Convert interpretations to surfaces .....................................................232
Procedure- Populate the model case .....................................................................233
Exercise 1-Create a low frequency model using the Inversion property builder .. 244
Procedure- Use a seismic Inversion OC plot ....... :.................................................245
Exercise 2- Create a seismic inversion QC plot .................................................. ... 248
Lesson 4- Simultaneous inversion and results ..............................................................249
Procedure- Run a Simultaneous inversion .............................................................249
Visualization and quality check of the simultaneous inversion output.. .................... 261
Exercise 1-Set up an inversion case and run a simultaneous inversion ............... 264
Resu Its ........................................................................................................................265
Procedure- Use the Geobody workflow ..................................................................269
Results ........................................................................................................................275
Additional diagnostics ......................................................................... ....................... 279
Lesson 5- Data input for AVOZ inversions .....................................................................284
Review questions ...............................................................................................................287
Summary ............................................................................................................................287

Module 5- Stochastic seismic inversion ......................................................................289


Learning objectives .............................................................................................................289
Lesson 1 -What is Stochastic inversion? ....................................................................... 290
Stochastic inversion background ...................... ..........................................................294
Lesson 2 - Stochastic inversion preparation ...................................................................300
Simple prior model building .......................................................................................301
Procedure- Create a simple prior model for Stochastic inversion ......................... 302
Exercise 1 -(Optional) Create a simple model for stochastic inversion ........ ......... 314
Variogram modeling .................................................._ ... :..............................................315
Exercise 2- Create a variogram map ...................................................................... 315
Procedure- Analyze data with a variogram ............................................................318
Exercise 3- Variogram analysis ..............................................................................323
Lesson 3- Set up and run a stochastic inversion ...........................................................324
Procedure- Set up the inversion run .......................................................................324
Exercise 1 -Run a stochastic inversion ................................................................... 332
Lesson 4- Stochastic inversion run evaluation ..............................................................333
6 • Table of Contents Quantitative Interpretation
Procedure- View the diagnostic output.. ................................................................333
Main diagnostics ........................................................................................................ 335
Diagnostic examples ..................................................................................................336
Model output .......................................................... ....................................................338
Lesson 5- Applications of Stochastic inversion results .................................................343
Workflows using stochastic realizations ...................................................................345
Procedure- Calculate the probability property .......................................................346
Lesson 6- Volume attribute modeling ............................................................................354
Procedure- Use volume attributes and reservoir AVO modeling to create
seismic models from stochastic inversion results ....................................................355
Procedure- Add spectrally shaped random noise .................................................. 360
Exercise 1 -Create a reservoir AVO model from stochastic inversion results
and add spectrally shaped random noise ................................................................... 362
Resu Its ........................................................................................................................ 362
Review questions ...............................................................................................................365
Summary ............................................................................................................................. 365

Module 6- Lithology classification ................................................................................367


Prerequisites ....................................................................................................................... 367
Learning objectives ................................................... ..........................................................367
Lesson 1 -Introduction to the workflow and classes definition .....................................368
Procedure- Create a classification using crossplots ..............................................370
Procedure- Create a classification using the Well log calculator ....................... ... 374
Procedure- Create a classification using Neural network ................ ..................... 377
Exercise 1 -Create a classification ......................................................................... 380
Lesson 2- Lithology analysis ..........................................................................................381
Visualization of the Litho Analysis results ................................................................. 382
The PDF confusion matrix statistics ...........................................................................386
Procedure- Generate the litho classification model. ............................................. .387
Exercise 1 - Generate the litho classification model and generate PDFs ..............389
Lesson 3- Lithology prediction .......................................................................................391
Procedure- Use the Litho prediction process .........................................................394
Exercise 1 -Use the Litho prediction process .........................................................397
Review questions ...............................................................................................................398
Summary .............................................................................................................................398

Appendix A- References ..................................................................................................399


Bibliography ........................................................................................................................399
Additional useful information .............................................................................................401

Quantitative Interpretation Table of Contents • 7


About this manual

This course teaches you how to use the different workflows in the
Petrel E&P software platform* Quantitative Interpretation (01) suite. It
gives you background information so that you can understand how the
processes work. It also shows you how to run the workflows, what
parameters are used, and how to interpret the output and results.
This course consists of six modules: Rock physics, AVO modeling, AVO
reconnaissance, Simultaneous seismic inversion, Stochastic seismic
inversion, and Lithology classification.

Prerequisites
To complete this course successfully, you must have
• experience working with Petrel
• familiarity with AVO concepts and fluid substitution as used in
E&P subsurface workflows
• completed training courses or have a good working knowledge
of the concepts taught in:
• Petrel Seismic Well Tie or Petrel Geophysics course or
equivalent
• Petrel Property Modeling course, especially familiarity
with geostatistics

Learning objectives
After completing this training, you will know how to:
• run the Rock physics workflows, identify the impact of fluids
on the well elastic paramete~s (computation and crossplots),
and quantify the effects of different fluids and saturation by
using fluid substitution
• run the AVO modeling workflow to generate synthetic
prestack seismic data and extract AVO attributes for use in
calibrating seismic data
• run the AVO Reconnaissance workflow and identify areas of
interest in a seismic data volume based on AVO models
Quantitative Interpretation About this manual• 9
• condition seismic data to use in inversion workflows using
seismic trace alignment
• run the Inversion Property Model Builder to create a low
frequency model for simultaneous inversion
• set up and run a simultaneous seismic inversion
• quality check the output using the inversion diagnostics and
the Seismic Inversion QC Plot tool .
• analyze the results using the Geobody workflow and the results
of rock physics studies
• build a simple pillar grid model and use this model to run a
Stochastic Seismic Inversion
• calculate properties from the results using the workflow editor
• run Reservoir AVO modeling to create synthetic seismic
displays from the output of stochastic inversions for quality
check purposes
• create wedge models
You also become familiar with:
• the concepts of rock physics including elastic moduli, fluid
substitution, and empirical relationships used in rock physics
studies
• AVO concepts and terminology
• different AVO approximations and their limitations and
understand how they are used
• primary and secondary AVO attributes and their uses in
analyzing seismic data

What you need


You must have this software to perform the exercises:
• Access to the training datasets
• Petrel 2015 and the 2015 Ql Extension installer
• Petrel licenses for Geosciences Core, Seismic interpretation,
Seismic Well-Tie, Data Analysis, Well Correlation, Quantitative
Interpretation, Multi-Trace Attributes, and Petrophysical
Modeling

10 • About this manual Quantitative Interpretation


What to expect
In each module in this training material, you encounter these types of
information:
• Overview of the module
• Prerequisites to the module (if necessary)
• Learning objectives
• Lessons, which explain about a subject or an activity in the
workflow
• Procedures, which show the sequence of steps needed to
perform a task
• Exercises, which allow you to practice a task by using the steps
in the procedure with a dataset
• Questions about the module
• Scenario-based exercises
• Summary of the module
You also encounter notes, tips, and best practices.

Quantitative Interpretation About this manual • 11


Course conventions
The content in this manual uses these conventions.

Characters in Bold Names of windows, dialog boxes, menus, and


tabs, such as "the Make/edit surface dialog
box."
Names of tools and icons, such as "Inspector
tool"
Schlumberger software application names, such
as Petrel, Tech log*, or ECLIPSE*
Characters in italics Names of sections and areas in windows and
dialog boxes, such as "Experiments area"
First use of important terms or concepts, such as
compositional simulation or "Safe mode is
used when ... "
Keyboard commands and action buttons, such as
"click OK'
Names of options, functions, lists, columns,
check boxes, and radio buttons
Characters in Code, data, and other literal text the user sees or
fixed-width enters, such as "Enter o . 7 3 2 3"
NOTE: Text you must enter is indicated in a fixed-width font or a
fixed-width font inside triangle brackets. Do not include the brackets
when you enter the required information.

12 • About this manual Quantitative Interpretation


Icons
Throughout this manual, you find icons in the margin representing
various kinds of information. These icons serve as at-a-glance
reminders of their associated text. See below for descriptions of what
each icon means.

Tips Notes Best practices

This icon JX>inls you to a tip This icon incicates that the This icon indcates the best
that wiD make your work following information is way to perform a gven
easier. particularly imJX>rtant. task when dfferent Of:XionS
are availalje.

Warnings Questions Lessons

This icon indcates when This icon identities the This icon identities a
you need to proceed with questions at the end of lesson, whch covers a
extreme caution. each lesson. particular topic.

Procedures Exercise Review questions

This icon identities the Tlis icon indcates that it's This icon identifies the
steps required to perform a your tum to practice the review questions at the
gven task. procedure. end of each modJie.

Prerequisites Learning oqectives What you need

This icon identifies any This icon identities any This icon indcates any
prerequisites that are leaning objectives set out ai=Piications, hardNare,
required for the course, or for the course, or for the data sets, or other material
for individual modules. current module. reql.ired for the course.

Quantitative Interpretation About this manual •13


More Petrel training courses
Sch lumberger SIS Petrel Training provides many courses for Petrel and
its functiona lities. The chart shows you how multiple types of training
can be combined and for what level they are built.
For more information, visit our web site: http://www.nexttraining.net.

I I
~
~
~
- i.__- . . ._ r-- -
II
,......- --...,
Petrel Petrel
Petrel Uncertamly Petrel Advanced
Rese!Volr and Advanced Uncertamly
Eng1neenng Oplilllllabon Wells and
(5 day cour~) Fundamentals Oplirmzahon
;::- =::! :=::::~~-=~ :=:.c:::~ - -
Petrel Petrel
ReseiVOir WorkHow
Petrel Well
Engmeenng Editor and
Design
(3 day course) Uncertamly
AnalysiS
~----~ ~ ~ ~ ~-~:~ ..==~
Petrel Petrel Well
Fundamentals Deliverabllity

~=~ r-- ....._ - ~ ~-=~ ~ ---


Petrel
Petrel User Petrel Geaogy Petrel Property Petrel Fracture Advanced
Ex penance Modeling Modeling Property
Modeling

'-------:------' !==-===!;:::.-=::::: ::::=-:~ . . _ r--

Petrel Shale Petrel VelOCity Petrel


Structural
Fundamentals Modeling
Modeling

Petrel
Petrel
Geophysics •
Petrel
Quantitative
INE.r.T blended
GeophysiCS COOI'S99USing
Advanced Interpretation Petrel

Stud10for Studio Plug-m rolated


Petrel Users Manager courses

Figure 1 Petrel training courses

14 • About this manual Quantitative Interpretation


Summary
In this introduction, we:
• defined the learning objectives
• described the structure of the manual
• outlined what tools you need for this training
• discussed course conventions that you encounter within this
material

Quantitative Interpretation About this manual•15


Module 1 - Rock physics

Despite good seismic quality and well control, the correct interpretation
requires quantifying the connection between geology and seismic data.
The Rock Physics module establishes this link by computing elastic
parameters. When pore fill and lithology are crossplotted, they can be
differentiated and potential reservoirs can be identified.
In addition, the fluid substitution assesses how changes in the fluid
content affect the seismic response.

Learning objectives
After completing this module, you will know
• the global Quantitative Interpretation (01) workflow
• basic rock physics concepts
• how to set up and run a rock physics study to:
• estimate elastic parameters
• perform fluid substitution modeling
• perform shear property estimation from Pwave velocity
logs
• perform log blocking by using Backus averaging
• how to crossplot results for analysis

Quantitative Interpretation Rock physics • 17


Lesson 1 -The Ql global workflow
This course presents one of the key technologies for characterizing
reservoirs; the Quantitative Interpretation of the seismic data.
Quantitative Interpretation involves the transformation of seismic
attributes (primarily, but not exclusively amplitudes) into physical
quantities that can be related to rock properties. Ql seeks to make
specific predictions, away from the wells, about quantities, such as,
lithology, pore fill, reservoir quality, and distribution.
In the first part, this course focuses on the technics to scan anomalies
and to understand the seismic signature of the lithology using AVO
Reconnaissance (if there are no wells) and Rock Physics and AVO
modeling when wells are present.
Because it can be difficult to interpret seismic amplitude quantitatively
as reflection coefficients that only represent contrast in impedance,
lithofacies can be identified based on inversion results, with absolute
values. By reducing the effect of the wavelet, lithology and fluid related
effects are therefore more interpretable. There are two types of
inversion available in Petrel, described in the second half of the
manual:
• Simultaneous inversion (or deterministic inversion) which
uses all the seismic input simultaneously. It is based on
minimization of the difference between the modeled seismic
. trace and the original one.
• Stochastic inversion which honors wells and seismic but
also include a statistical component, providing not one unique
solution but multiple equiprobable realizations. Each realization
is optimized to fit the seismic data based on the user criteria.
Finally, lithology analysis and prediction is a rock physics based
lithology prediction workflow that integrates well logs, seismic
inversion, and geological modeling and interpretation. It provides an
estimate of the most probable lithology and the uncertainty associated
with the prediction.

18 • Rock physics Quantitative Interpretation


The different modules can be classified as shown in Figure 1. They
cover all the modern seismic interpretation technics from the
Exploration to the Production stage.

AVO Reconnaissance (no wells}

El lc: para
Fluid predict & AVO Modelinc AVO Modeling (Module 2}
Utholo,. an on
AVO Reconnaissance (ModuJe 3}

Simultaneous inversion (Module 4)


Simultaneous inversion

and S ocha tic invenion

Figure 1 The different Q/ modules performed at different stages

Quantitative Interpretation Rock physics •19


The workflow (Figure 2) links the different processes of the 01 workflow.
On the left, it starts with AVO reconnaissance, which does not need any
well control. Reading the diagram toward the right, you will need more
and more wells to perform the different processes in the proper way
meaning that you will use them at different stages of development.

---
_______.,....,
Figure 2 Global Quantitative interpretation workflow using Petrel

'''ilutudta D Rock physrcs proce:s.s LIUWIOgy c.'a,s!-co ~:. p•:y,~:»eS

Qu,1/ofy control features


D 1\ VO Modeimg process h!Vi;f,<;iiJil ()(OCflSSES

D Result arwlysos features


D 1\ VO Rerot7naossance proce:s.s

20 • Rock physics Quantitative Interpretation


Lesson 2 -Introduction to Rock physics
The science that underlies 01 is the subject of rock physics. Rock
physics uses the principles of physics and the concepts of mechanics
and wave propagation. Rock physics is used to establish relationships
between rock and fluid properties (compressibility, rigidity, acoustic
impedance, and Poisson's ratio) and seismically observable quantities
(the speed at which seismic waves travel, and why and how they
reflect) to produce recordable seismic energy and information about
how seismic waves propagate. These rock properties have a
relationship to quantities, such as lithology, pore fill, and porosity.
These quantities are important for finding and developing hydrocarbons.
Rock physics is distinct from Petrophysics. Petrophysics focuses on
borehole measurements and their interpretation in terms of fluid
content and reservoir characterization. Rock physics shares the use of
downhole data with Petrophysics. However, rock physics concentrates
on:
• the physical relationships that link the fundamental properties of
the rocks to the quantities of interest to geophysicists that affect
seismic data
• how changes in the rock fabric and fluid content affect seismic
data
Consider the crossplot in Figure 3.

Quantitative Interpretation Rock physics • 21


Acoustic Impedance vs. Poisson's Ratio

cu
u
c
nJ
""C
cu
a.
E
u
~
=
0
u
<t

Gas Sands

Poisson's Ratio

Figure 3 Acoustic impedance versus Poisson's ratio. The trend lines show
where data from specific lithologies tend to cluster

This plot shows schematically how Pwave acoustic impedance and


Poisson's ratio relate for wet sands, shales, and gas sands. This graph
is powerful because it shows that these properties cluster differently,
depending on their lithology and pore fill. This means that you
potentially can discriminate different lithologies and pore fills by
calculating and crossplotting quantities such as acoustic impedance
and Poisson's ratio.
Seismic reflections from different lithologies and pore fills can be
differentiated. A reflection from an acoustically hard shale overlying an
acoustically soft shale occurs between two points on or near the shale
line. This reflection has a decrease in acoustic impedance, but an
increase in Poisson's ratio.
By contrast, a reflection from a gas (or light oil) sand should have a
decrease in acoustic impedance and (unless the shale is very hard) a
decrease in Poisson's ratio. Figure 4 illustrates these reflections
schematically. Figure 5 shows a real data example. Each of these cases
22 • Rock physics Quantitative Interpretation
has a unique signature that allows you to identify it if you can calculate
the changes in the two quantities in the plot.
The arrows in Figure 4 show how the acoustic impedance and Poisson's
ratio change differently for a reflection from a hard to soft shale
(Poisson's ratio increases, acoustic impedance decreases) and from the
hard shale to a gas sand (Poisson's ratio decreases, acoustic impedance
decreases).

Acoustic Impedance vs. Poisson's Ratio

ell
u
c
ns
"'C
ell
a.
E
u
-~
:s
0
u
<(

Gas Sands

Poisson's Ratio

Figure 4 Acoustic impedance versus Poisson's ratio to differentiate lithology


from the change in the acoustic impedance and/or the Poisson's ratio

Quantitative Interpretation Rock physics • 23


Figure 5 is an example from some real well data. A cloud of data points
sits offset from the main data trend in the lower center of the graph.
This small cluster is related to a gas filled sand penetrated by the well.
This crossplot confirms that on the real dataset, it is possible to
discriminate between pore fill and lithology by quantitatively calculating
properties such as Poisson's ratio and acoustic impedance. Analyses
that use log and seismic data as the input can be used to make
predictions about the subsurface.

-~~~~~~~~~~~~~~~~~~~

13 -

-
12-

~ 11-=

-~ 6 -
~
-;
s-j
3
0
Poisson rabo_Study 1 (Rock physics]

Figure 5 Acoustic impedance versus Poisson's ratio for actual well data with
points from a known gas sand highlighted in the polygon

The first part of this training class focuses on the Rock Physics process.
This process enables you to compute elastic parameters such as
acoustic impedance (AI), the Vp/Vs ratio, and Poisson's ratio from well
data. These physical rock properties then can be interrogated using
Petrel crossplots to determine which parameters can help identify
reservoirs and hydrocarbon bearing zones. You also look at the impact
of different fluids and saturations through fluid substitution exercises.
Then, you look at the AVO Modeling process. This process provides a
link to the seismic data by modeling the AVO response of your elastic
properties, derived from your seismic data. It uses the results of your
rock physics studies, such as fluid substitution, to characterize the
range of seismic response to different conditions. The objectives are to
24 • Rock physics Quantitative Interpretation
connect the seismic data to the well data and allow you to make
predictions about fluid content and reservoir presence and quality
where there is no well data.
After the AVO modeling process, you use the AVO reconnaissance
workflow. This workflow allows you to combine conventional data
screening for structures and amplitudes with on-the-fly calculation of
AVO attributes. AVO reconnaissance relies on Rock Physics and AVO
modeling processes to characterize what AVO signatures are useful to
identify potential hydrocarbon zones. Combining the search for potential
structures of interest with an analysis of the AVO response allows for a
more efficient screening and ranking of potential prospects.
Finally, you look at two types of inversion; simultaneous inversion
and stochastic inversion. Simultaneous inversion inverts seismic data
to calculate property volumes that contain acoustic and elastic
properties such as acoustic impedance, VpNs ratio, and density. You
then can use these results with the rock physics studies to identify and
map reservoir facies and fluid accumulations.
Stochastic inversion also inverts seismic data. It incorporates
geostatistics to create many alternative scenarios for the subsurface.
Each scenario is consistent with the well and seismic data. These
different scenarios can help you understand the distribution of the
reservoir in a particular formation and get a higher resolution. It also
allows you to calculate the uncertainty in quantities such as the gross
rock volume and reservoir quality.

Quantitative Interpretation Rock·physics • 25


Lesson 3- Rock physics review
Before proceeding, it is useful to review some basic material about how
seismic waves propagate in the subsurface and how properties (such
as, lithology, porosity, saturation, and fluid content) affect that
propagation and the reflections that you record in your seismic data.
Specifically, this lesson reviews
• what causes reflections
• different modes or propagation: Pand S waves
• velocity relationships
• definitions of some important elastic moduli
• fluid substitution
• mixing formulas
• impact of saturation
• empirical rock property relationships
This lesson starts with a basic seismic experiment as shown in
Figure 6.
Receivers

Impedance contrast causes


energy to be reflected

Reflections
Offset

Figure 6 A simple seismic experiment with one shot and an array of receivers

In this experiment, a source of energy is exploded at the surface and


seismic waves travel downward. If there is a change in the acoustic
impedance at a layer boundary; that is, a change in the product of the
velocity and densities of the layers, then some of the energy is reflected
back to the surface where it is recorded . These reflections make up the
seismic record. They are what you are going to try to use to calculate
properties of interest to tell you about the layers that caused the
reflections. A small sample of a seismic record is shown in the lower
art of Fi ure 6.
26 • Rock physics Quantitative Interpretation
As stated, a change in the acoustic impedance (the product of velocity
and density) is necessary to cause a reflection. You start by looking at
the density. Then, you look at the different ways seismic waves can
travel and how these different ways affect the velocity of the waves.

Density
If you look at a thin section from a porous rock, such as the one shown
in Figure 7, you see that the rock is made up of grains and spaces
between the grains. The grains form the rock matrix, and the spaces
form the porous portion of the rock volume.

Figure 7 A thin section showing the grains of the rock matrix with porous space
between

In the subsurface, the pore space is filled with fluids, usually brine. In
the places of interest to you, the pore fill can be oil or gas. Other rocks,
for example shales, have different structures, but in this course, you are
interested in rocks with pore space to trap oil and gas.

Quantitative Interpretation Rock physics • 27


Density, in the case of a porous rock, can be calculated in a
straightforward manner from the proportion of the constituent parts.
The formula is

In this formula:
• pb is the bulk density of a volume of the rock (including pores and
pore fill)
• pmis the density of the matrix, the material that forms the grains
• pwand PHcare the respective densities of the water (brine) and
hydrocarbons that fill the pore spaces
• <1> is the porosity or the fractional part of a sample of rock that is

pore space
• Sw is the water saturation in pore space expressed as a fraction
from 0 to 1
The velocity with which the seismic energy travels is more difficult to
calculate. First, you must consider how the energy is traveling.

P and S waves
Two modes in which sound can travel in rigid materials are the
longitudinal or P-wave mode and the S or shear wave mode. These are
known as body waves.
NOTE: Pand S stand for primary and secondary. These terms originated
in earthquake seismology and refer to the order in which the waves
arrive at the detectors.
These two modes are differentiated by the direction of particle motion
in the rock as the wave passes through.
In Pwave mode, the particle motion is along the direction of
propagation. In the shear mode, the motion is perpendicular. Figure 8
shows Pwaves on the left and Swaves on the right.

28 .. Rock physics Quantitative Interpretation


PWave SWave

direction of direction of
particle particle
motion motion

Figure 8 P and S waves with their respective directions of particle motion

Figure 9 illustrates further the type of motion involved in these two


modes. In this figure, Pwaves are shown on the left and shear waves
are shown on the right.

c
0
·usc

-..
Cl)
)( ·;c
·=
...
cg w


U)

;,;
-
....-------, c.J
(,)
·c
~

·~-·
E
.:0

>--
.§--
0
c
0
·;;
(I)

~
=-
E
0
• z
>
0

(..)

Direction of
wave propagation

Figure 9 Effects of P and S waves on a small sample of rock portion as the


wave propagates

Quantitative Interpretation Rock physics • 29


Figure 10 depicts these differences at the grain level in. This figure
shows an undeformed cube of rock on the left. When a shear wave
passes, the side to side motion changes the shape of the rock, but not
the volume of the rock. When a Pwave passes, the rock compresses (or
rarities), changing its volume.

Compressional
motion

Figure 10 Effects of shear and compressional stresses on the grains and fluids
in a small rock sample

From the illustration, you can see that these two processes are different
mechanically. In Pwave propagation, compressing a volume of rock
involves compressing the grains and the fluids in the pores. It involves
the mechanical properties of both constituent parts of the rock volume.
In S wave propagation, the grains and fluids are not compressed, so
shear waves are subject to a different set of mechanical properties.
There are two parameters that describe how the rock deforms in
response to the shear and compressional stresses shown in Figure 10.
These parameters are examples of what are called elastic moduli. The
amount by which the rock deforms when subject to a shear stress is
characterized by the shear modulus, usually represented by the symbol
Jl.
The amount by which the rock deforms when subjected to a
compressional stress is characterized by the bulk modulus, which
usually is represented by the letter K

30 • Rock physics Quantitative Interpretation


Velocity relationships
Two important relationships define the P(compressional) wave velocity
and the S (shear) wave velocity in terms of the elastic moduli.
The Pwave velocity is given by:

v.p =F+ip
p
The Swave velocity is given by:

v.
s
= ~
~p

In these equations:
• K is the bulk modulus of the combined rock and fluid system
• ~is the shear modulus for the combined rock and fluid system
• p is the density
You generally are familiar with typical ranges for velocities. Brine has a
Pwave velocity of around 1,490 m/s. High porosity gas filled sands can
be similar. On the other end of the spectrum, dense anhydrites, highly
cemented rocks, and volcanics can have velocities as high as 6,100 m/s.
Looking at these two formulas, you can see that
• Vp depends on the bulk modulus (and, therefore, on the fluids in
the pore space, because the fluids are compressed along with the
rock frame when a seismic wave passes). It also inversely depends
on the density. Thus, when pore fluid changes, both the density and
the bulk modulus contribute to a change in Vp.
• Vs depends on the shear modulus, which is not influenced by the
pore fluids (because they are not deformed by a shear stress).
Therefore, a change in pore fluids changes only the shear velocity
through its inverse relationship with the density.
• Shear waves travel more slowly than Pwaves.
The magnitude of these fluid-related changes is discussed later.

Quantitative Interpretation Rock physics • 31


Elastic Moduli: Hooke's Law
The elastic moduli relate how solids (or in some cases liquids and
gases) deform when stresses are applied. They take their form from
Hooke's Law. This law states that (under elastic conditions) the
deformation (strain) of a solid is linearly related to the applied force
(stress). The equation that describes this law is

a= C· e
In this equation:
• a is the applied stress
• Cis a constant of proportion, in reality a tensor
• e is the deformation or strain
When the material being deformed is isotropic, the tensor simplifies
considerably and there are only two independent quantities. Different
convenient moduli can be defined related to these two moduli. Tables
that show the relationships between the constants are available in a
number of publications (for example, Simm and Bacon, 2014, Seismic
Amplitude, An Interpreter's Handbook, p 12).
These elastic constants are of interest to seismologists:
• E=Young's modulus
• Jl =Shear modulus or rigidity
• k =Bulk modulus or incompressibility
• cr = Poisson's ratio
• 'A = Lame's constant

32 • Rock physics Quantitative Interpretation


Young's modulus
When a force is applied to a solid, the length of the solid changes.
Young's modulus is the constant of proportion that relates the stress to
the strain. It is defined in terms of the proportional change in length of
the solid, which is illustrated in Figure 11.

'"....., ,-··········=·~ } DJ

II~-·············<:
Figure 11 Change in length as a result of an applied stress

The relationship is written as

In this equation:
• a, is the stress expressed as a pressure
• !J.L/L is the fractional change in length
• E is Young's Modulus
The Young's modulus also can be calculated from Vp, Vs and P (inputs
in Petrel) as:
E = (1 + a)2tt
In this equation:
• a is the Poisson's ratio
• Jl is the Shear modulus
Some published values of Young's modulus are
• Quartz: 95.8 GPa
• Dolomite 116.6 GPa
In these values, a giga Pascal (GPa) is equivalent to 10 10 dynes/cm 2.
(Values are from Simm and Bacon, Seismic Amplitude: An
Interpreter's Handbook, 2014.)

Ouant1tat1ve InterpretatiOn Rock physics • 33


Shear modulus or rigidity
This modulus relates the shear stress to the shear deformation or
strain. It is expressed as

In this equation:
• as is the shear strain
• ~ Y/X is the proportional deformation
• Jl is the shear modulus also called the rigidity
The Shear modulus also can be calculated from Vs and p (inputs in
Petrel) as:

f1 = pVsz
In this equation :
• p is the density
• Vs is the S-wave velocity
Figure 12 illustrates these terms.

~y

. . :::·:::~··_··--~---r
as---!-"+
No volume X
change

Figure 12 Response of a solid to a shear strain

Some typical published values for Jl are:


• Smectite 7.5 GPa
• Quartz 45.0 GPa
• Dolomite 45.0 GPa
Values are from Simm and Bacon, 2014.

34 • Rock physics Quantitative Interpretation


Bulk modulus or incompressibility
This modulus measures the resistance to volumetric stress. It is written
as

In this equation:
• Pis the confining pressure
• L1 V/Vis the fractional change in volume
• K is the bulk modulus also referred to as the incompressibility
The Bulk modulus also can be calculated from Vp, Vs and (rho symbol)
(inputs in Petrel) as:
K =pVp2 - (4/ 3)pV/

In this equation:
• p is the density
• Vp is the P-wave velocity
• Vs is the S-wave velocity
As shown in Figure 13, confining pressures result in a shrunken volume.
The fractional change in volume associated with an applied stress is
related through the bulk modulus.

Figure 13 Volumetric change as a result of a confining pressure

Some published values for solids are


• Quartz 36.6 GPa
• Dolomite 94.9 GPa
These values are from Simm and Bacon, 2014.

Quantitative Interpretation Rock physics • 35


Some representative values for fluids are
• Kgas = 0.0435 GPa
• Koil = 0.3922 GPa
• Kwater = 2.68 GPa
These values are from Dvorkin, Gutierrez, and Grana, 2014.

Poisson's ratio
Poisson's ratio is one of the most important physical quantities used in
the 01 workflow, particularly in AVO analysis. You saw that when it is
combined with acoustic impedance, it can be used to discriminate
between reflections caused by different lithologies and fluid contrasts.
If you stretch (or compress) a solid, the length changes. The stretching
(or compression) also tends to change the cross sectional area of the
solid as shown in Figure 14. Poisson's ratio relates these two changes.

... hn~
~·o· ... .. ....... .. .. ... .. ........ ..... .. .. .. .. .. .. .. .. ......~o··. ····r!}·dh
· ...
~:·::~.................................................. _ .. :r.:=·
L
L+.6.L

Figure 14 Changes in length and width are related through Poisson's ratio

Using the terms in this diagram, Poisson's ratio is expressed as

In this equation:
• t1UL is the longitudinal strain
• !1h/h is the transverse strain
• cr is Poisson's ratio
Poisson's ratio can be recast in terms of the Pand S wave velocities,
which are more intuitive to seismologists. The relationship is

(?.)' - 2

a= 2[(~) 1] 2 -

36 • Rock physics Quantitative Interpretation


Figure 15 shows a graph of VpNs against Poisson's ratio.

Vp/Vs
0 .6

0.5 ,.......
! /
!
-: 0 .3
0 .4
j
v
0
Ill
JA
l. 0.2
I
0.1
j
0
0 2 4 6 8 10 12
Vp/Vs

Figure 15 Graphical depiction of the relationship between the VpNs ratio and
Poisson's ratio

From the graph, you see that Poisson's ratio varies from 0.5 at very high
VpNs to 0 when VpNs is equal to J2.
If you think of how Vp and Vs behave when you introduce gas into a
previously brine-filled porous sand, you expect Vp to decrease in
magnitude because gas is easier to compress than brine, so the rock
has a smaller bulk modulus. Also, you expect that the shear velocity is
little affected. Therefore, VpNs decreases, and, from the graph in
Figure 15, Poisson's ratio can be expected to decrease as well.
Poisson's ratio enters into several of the approximations for amplitude
behavior versus offset that you review later.

Quantitative Interpretation Rock physics • 37


Lame's constant
One last modulus that is often used in seismic analysis is Lame's
constant, A. It is not directly measurable. It is calculated from other
moduli. It is part of the relationship between the bulk modulus (K) and
the shear modulus J..l:
K=A + (2/3) J..l
In this equation, A with J..l is a combination often used to evaluate
lithology and fluid changes because J..l is relatively insensitive to fluid
changes and A depends on the bulk modulus. Goodway et al. introduced
the use of A and J..l in seismic analysis in 1997.

Fluid substitution background


Fluid substitution aims to predict how seismic velocity and impedance
change with pore fluids . When pore fluids change, the shear velocity is
affected only by its inverse relationship to the density. The density, as
you saw earlier, is easy to calculate. For Vp, you must deal with the fact
that by changing the fluid, the bulk modulus changes, so you need a
way to calculate the new bulk modulus to calculate the new velocity.

Gassmann's equation
Gassmann introduced the most frequently used way to calculate the
new bulk modulus. Gassman published the equations (Gassmann, 1951)
that describe how the Pwave velocity depends on the properties of the
various parts of a mixed system composed of a porous reservoir and
different fluids. Gassmann's formula for the Pwave velocity is a
modified version of the formula for velocity shown earlier and takes this
form:

38 • Rock physics Quantitative Interpretation


In this equation:
• Vp is the Pwave velocity
• p is the density
• K is the bulk modulus
• ~is the shear modulus
NOTE: The subscripts dry, m, and fl refer to the dry frame (that is, the
rock with empty pores), the mineral that makes up the frame, and the
fluid in the pores.
Compare the Gassmann formula to the previous expression for Pwave
velocity:

Y-
p
= F+ ~JL p

You can see that the first term in Gassman's equation really is the bulk
modulus for the saturated rock. What Gassmann has done is to give a
way to compute the bulk modulus from knowledge of the constituent
parts of the fluid filled rock.
Typically, the fluid substitution process consists of several steps:
• If Vp, Vs, and density are known, you can ca lculate the shear
modulus I""'
11, and with that value, you can calculate K
sat
• If you calculate Ksa t' then with a knowledge of both the porosity
and the fluids in the pores and an assumption about the mineral
that makes up the rock frame, you can calculate Kdry also.
• With this result, you can use the formula to calculate the new
velocity that results when a different fluid is substituted into the
pores.
• Values for the fluids (brine, oil, and gas) can come from laboratory
measurements or from empirical equations such as those published
by Batzle &Wang (1992). These equations are implemented in
Petrel.
• If shear velocity information is not available, you might have to
make assumptions to calculate~ -

Quantitative Interpretation Rock physics • 39


Some key assumptions underlie Gassmann's equations (Simm and
Bacon, 2014 and Dvorkin, Gutierrez, and Grana, 2014). Specifically,
these assumptions are:
• the rock is elastically isotropic and homogeneous
• the pore space is in perfect communication
• the pore fluid is frictionless and there is no coupling between the
rock and fluids
• the rock and fluids can be characterized by single bulk moduli and
densities
Gassmann's equation works well when the minerals and fluids are
distributed evenly throughout a rock volume. In situations where the
pore distribution is uneven or the pore sizes are irregular (as in some
carbonates, for example), it does not work well. Shaly sands, laminated
sands, and tight gas sands also are common situations where care must
be taken to ensure that Gassmann's equations are applied properly.
(Refer to Simm and Bacon 2014.)

Fluid mixture formulas


Gassmann's equation assumes that the fluids and the rock frame are
represented by single values for the densities and the bulk moduli.
However, many rocks actually are mixtures of several different
components. When hydrocarbons are present in the pore space, there
always is some residual brine saturation present as well. This means
that you need a way to calculate the equivalent or effective densities
and moduli from the constituent parts of the mineral and fluid mixtures.
The densities are easy to calculate and were discussed earlier. The bulk
moduli are more complicated. There are several models for calculating
them. Two common methods that are implemented in the Petrel Fluid
Substitution workflow are the Reuss and Voigt methods. For mineral
mixtures, these two methods give lower and upper bounds for the
actual value. They often are averaged together to give the Hill average
(Dvorkin eta I. 2014, p 14).
The Reuss average uses the harmonic mean:

~=fl +fz + ···


K K1 K2

40 • Rock physics Quantitative Interpretation


In this equation, K.I is the bulk moduli and fis the fractional amounts for
the different parts that constitute the mixture. If there are more than
two components in the mixture, additional terms are added to the right
side of the equation.
The Reuss average is the appropriate formula to use for fluid mixtures
where the individual fluid phases are in perfect hydraulic
communication (Dvorkin et al., 2014, p 17).
The Voigt average uses an arithmetic average of the respective moduli:
K = f1K1 + fzKz + ···
t.
In this equation, K. and have the same definitions as the Reuss
I

average. Again, if there are more than two components in the mixture,
additional terms are added to the right side of the equation.
As stated, the Hill average is the arithmetic average of the Reuss and
Voigt averages.

Quantitative Interpretation Rock physics • 41


Effects of saturation
With Gassmann's formula and a workflow to perform fluid substitutions,
you can investigate what happens to the Pand S wave velocities when
you substitute oil and gas in varying quantities for the normal brine pore
fill.
Figure 16 shows schematically what happens (for porous, uncompacted
rocks).
Vp and Vs versus Water Saturation Poisson's Ratio versus Water Saturation

High High

,-_JLD....._______,_) ....0
Ill
a:
-"'c
~
·a
a.
Vs

Low ' - - - - - - - - - - - - - - - - - - - - - _ _ . Low


~------------------~
0 Sw 1 0 Sw 1

Figure 16 The effects of different water saturations (that is, gas saturation) on
P and S wave velocities and on Poisson's ratio

Most of the effects are seen over a small range of saturation values
when the gas saturation still is quite small.
Vs changes little, except for a small increase in saturation because of
the inverse relationship with density. The introduction of gas lowers the
Pwave velocity substantially. Most of the decrease occurs with the
initial introduction of gas into the system (hence the difficulty in
determining low saturations from seismic measurements).
If oil is substituted instead of gas, the velocity would decrease as well,
but less so than with the introduction of gas. The oil gravity and the
Gas/Oil Ratio (GOR), or the amount of gas in solution, influences the
result for oil. Some oils with high GORs can produce a significant
velocity decrease. As expected, Poisson's ratio also drops given the
significant change in Vp and the small change in Vs.
In the case shown in Figure 16, Vs was affected only mildly by fluid
content, but shear velocity can be affected by lithology. Vp depends on
both fluid content and lithology. Thus, if you can isolate the Pwave and
S wave responses in a reflection, you have a potential recipe for
42 • Rock physics Quantitative Interpretation
discriminating between fluid and lithology changes. This process gives
you a powerful way to polarize anomalous seismic responses and
improve risk analysis of subsurface opportunities.

Empirical rock property relationships: P wave velocity and


density, Vp to Vs
Several empirical relationships have been derived over the years that
relate density to Pand Swave velocities. These relationships are useful
when actual measurements are unavailable, which can be an issue
when specific logs were not run or the log quality is poor. Lack of full
log suites can be an issue especially with older wells.
The empirical relationships also can be useful for modeling studies in
areas far from well control or in basins where there might be no well
control and velocity information is available only from seismic data.
Some widely used relationships that are available in the Rock physics
process are:
• Castagna mudrock line
• Castagna-Greenberg

Castagna mud rock line


As interest in AVO and shear studies grew, the relationship between Vp
and Vs was investigated empirically, and several relationships were
derived relating the two quantities. Castagna (Geophysics 1985)
published an early relationship, commonly used in the industry, from
work at ARCO. It is referred to as the Castagna mudrock line. It states
that
vp = 1.16~ + 1360m/s
This relationship was derived from water saturated sands, shales, and
siltstones.

Quantitative Interpretation Rock physics • 43


Castagna-Greenberg
Castagna's work was extended later with additional data and
relationships were derived for sands, shales, limestones, and
dolomites. These results were published in 1992 and are known as the
Greenberg and Castagna relationships. They are
Sandstone (kmjsec) : l's = -0.856 + 0.804~

Limestone (kmjsec ); }'s = -1.031 + 1.0171{,- 0.055v;,2


Dolomite (kmjsec ): ~ = -0.078 + 0.583~
Shale (kmfsec) : ~ = -0.867 + 0. 770~
All of these regressions assume that any porosity is brine filled. These
empirical relations apply only to the set of rock studied. Petrel gives
you the option to use the published values or user-determined project
specific values. As a good practice, these should be calibrated by an
rock physics expert if there is relevant data, for example, enough well
data with the same type of lithology, and same depth.

What happens at a reflecting boundary?


Look again at what happens at the reflecting boundary in Figure 17.

Impedance contrast causes


energy to be reflected

Time 1" Reflections


Offset

Figure 17 Simple seismic experiment with one shot and an array of receivers.
An example of the seismic data that would be recorded is shown in the lower
part of the figure

44 • Rock physics Quantitative Interpretation


As stated earlier, reflections occur when the velocity density product
changes across an interface. Look in a bit more detail at what happens
at a reflecting interface when a Pwave impinges on an impedance
contrast. Consider only incident Pwaves because most of the seismic
data that'you are analyzing is Pwave data (Figure 18).

Reflected S
lncidentP

Figure 18 The different components of the reflected and transmitted wave field
that result from an incident P wave

The incident Pwave creates both reflected energy and transmitted


energy. There are two components to the reflected energy, a Pwave
and a (converted) shear wave. Similarly, the transmitted energy has
both shear and Pcomponents.

Quantitative Interpretation Rock physics • 45


Zoeppritz equations
The equations that describe what happens at a reflecting boundary
were derived in 1905 by Zoeppritz. He derived the equations by applying
boundary conditions to propagating waves at an impedance contrast.
The full solution is a set of equations that bears his name, the Zoeppritz
Equations. They are
-]
-sin {)1 -cos rflt sin {)2 cos rA
Rp({)]) cos ()] -sin rflt cos {)2 -sin rA sin {)1
Rs({)I)
sin 2{)1
v
_E.!_cos 2rflt p2vs;vp] .
sm 2{)2
pV V
2 S2 PI COS 2¢.
cos ()]
2 2 2
Tp(B1 ) VSJ piVSIVP2 piVSI sin 2{)1
Ts({)I) -cos 2rflt VSI . 2¢.
-sm p 2VP2 cos 2¢ - p2 Vs2 sin 2¢2 cos 2¢1
1 2
VPI piVPI piVPI

Figure 18 describes the angle, velocity, and density terms in the


equations. The quantities RP(8 1) and R5(8 1) refer to the reflected P and
S wave amplitudes. The quantities T (8 1) and T)8 1) refer to the
transmitted Pwave and S wave ampfitudes.
These equations describe the amplitudes of reflected and transmitted P
and S components exactly, but they are not necessarily intuitive. For
this reason, many workers have derived linear approximations to the
full equation set that are easier to understand and calculate. These
approximations describe the angular dependence of the reflections and
the role of some of the physica·l parameters that have been discussed
so far.
Module 2 and ModuJe 3 discuss these approximations in more detail.

46 • Rock physics Quantitative Interpretation


Lesson 4- Rock physics workflow
The Rock physics workflow has a number of parts that complement
each other and provide a complete suite of tools to perform rock physics
studies. The flow diagram shown in Figure 19 illustrates the
components of the workflow and how they fit together.
A P-wave velocity (or sonic slowness) is the minimum requirement.
Most log suites include a density log, but if this is not available you can
calculate a synthetic density using Gardner's relationship. Similarly, if
not available, a synthetic shear sonic log can be calculated from the
compressional sonic (velocity) log with the available tools in Petrel.
When the log suite is complete, the Elastic parameter estimation
process can proceed. After the estimation process is complete, you can
crossplot the output to look for relationships that help identify
lithologies and hydrocarbon zones. Logs also can be blocked for
additional studies. Fluid substitution can be done to test alternative
fluid and saturation scenarios.
To assess the impact on seismic response, fluid substitution results can
be crossplotted or used in AVO modeling, which is discussed in
Module 2.

Quantitative Interpretation Rock physics • 47


Input data

AVO Modefillg process

Elastic paramelets O$limation {2) Rtsuft anatylis features


·Acoustic impedance
..S~Un~r impedaoee
.VpNsratio
.Poisson's ratio
·Young's modulus
·Bulk modulus
·Lame consl.ant
-Shear modulus
·lambda-Rho
·Mu-Rho Log blocking Backus (4)

·P·wave velocity. S·wave velocity and Density


·Oelta. Gamma and E~ilon

· PoW!IVCIV8Joci1y MW
.S.W.tve wtocity n.aw
·Oan.tilynew
·I< sat
·P-wwve ~e:oa1y dry
.S·wave veloe>1y dry
·OensMy<Jry
.f<dty
-&\eat mottulua
·OeiiSIIytnlneraJ
·Kmlnet~~l
-tlensaly :fluid
·K fiuid
·Oen !lily fluid II'IOC!IIIe<l
·K nuid modo!od

AVO Modeling (6}

Figure 19 Rock physics workflow

48 • Rock physics Quantitative Interpretation


Lesson 5- Basic rock physics studies
In Petrel, the Rock physics dialog box is accessed from the
Quantitative Interpretation tab. The results of rock physics studies
can be displayed in a Well section window and edited with the
Inspector.

Procedure -Access the Quantitative Interpretation


All of the workflows for Quantitative Interpretation in Petrel are on the
Quantitative Interpretation tab.
1. On the Home tab, click Perspective, then click Advanced
Geophysics.

Persped:~;e WindO•"' full Panes Reset Windo# Petrel Studio


filters layout y ;creen layo~i

View Insert Search

2. Click the Quantitative Interpretation tab.


All tools and workflows for Rock physics, Seismic well
calibration, Seismic conditioning, AVO, Seismic inversion,
Crossplot, and Lithol~y classification are under this tab.
q..,.,.,..,............,

l
r~

!!il ,,.,.., __, ... VI"""'"""""' •tt II


~··Ql:oob ~Wr'Aot.acitng r
..
~ O~n~
JJ """"'.....-. ]I 11!1 ,........, ........... :::s: ""'-'"""'"''...-
193 m [il'""""""' -
lrf't'c Att~.w.!J;•., 'Wcr1wtl l r'IIJN!NIUWW"!N; ~- rr11l'• G) :ot•:o....-..r""l\ :.~.;n AI/., t;I•'A · ~~ qr ~-JO ~("'*" 1 "*"" tJ ~QC~ ,_ cn~~
1W'tflolo .tiOf!lnwtt' ,.,,., kW r-cr, .. t ...'V• iWtdl;ar"'ICMf "nnltl l """"-M ..._;r
..,. _ _I - ,..,.,..n
Ql ,_ ~a..-.,.,.
Acc:.c.ol\<o?CS Scf;mewdlc;Jiitnocn • ~J:C:wirtjCf"lll!j I 1:\10 ~ntefJien I

Quantitative Interpretation Rock physics • 49


Procedure- Set up a rock physics study for Elastic
parameter estimation
You must enter all of the input data in the Rock physics dialog box.
Also, you can set the operation .type and the choice of output.
1. In the Rock physics group, click Rock physics.

The Rock physics dialog box opens.


2. Select Create new or Edit existing, depending on whether
you want to start a new study or edit an existing study.
• If you select a new study, name the study appropriately in
the box to the right.
• If editing an existing study, a list of previously created
studies is available from the list.
This procedure assumes that Create new is selected.

- -- - - - -
Study 1 [Rock physics]

Opertltion type: IElastic parameters estiml!tion


Compute el8sbc conslllnls

Us111g Vp. Vs end Density


0 Using L11mbde. Mu end Density

Input Output

yp P·Wl!Ve velooty. ~ [ ypvp

V. S·W!Ive velocity ~ [ V. vs

p Density· ~ [ ~1 RHOB

3. Select the well to be used in the study from the Well list or
insert the well to use from the Input pane.

50 • Rock physics Quantitative Interpretation


A depth range object can be used to limit the depth interval
where the calculation is run. To use this feature, complete
these steps:
a. Click {3Z
J to the right of the well .
b. Enter depth ranges in the dialog box that opens:

Ix Close I
NOTE: Setting depth ranges is covered later in the course.
4. Select the Operation type from the list.

You can choose one of these types:


• Elastic parameter estimation. This type allows you to
compute elastic constants for isotropic media.
• Shear velocity estimation. This type is used when shear
data is not available and you want to create synthetic
shear data from the Pwave velocity log.
• Log blocking Backus. This type performs Backus averaging
to create blocked logs.
• Fluid substitution. This type is used to estimate changes in
elastic properties caused by changes in pore fluids and
their saturations.
In this case, select Elastic parameter estimation.

Quantitative Interpretation Rock physics • 51


5. Select the option to compute elastic constants.
Two options are available:
• Using Vp, Vs and Density
• Using Lambda, Mu and Density
These options allow you to use two different methods to
compute elastic parameters. The option that you select
depends on the type of input data available.
In this case, select Using Vp, Vs and Density.
Compute elastic constants
@ Using Vp. Vs and Densibt
0 Using Lambda. Mu and Density

6. Open the Input tab and use the lists (or the blue arrows ) to
enter the well log curves to be used.
You can use velocity or transit time curves as input.
Input Output

Vj, P-weve velocity: [§)I Vj,Vp


Vs S-wave velocity ~( V.vs

Poensrty: ~~ ~1 RHOB

NOTE: If shear velocities are not available, you can create


them by using the empirical regressions discussed earlier. To
create shear velocities, choose Shear velocity estimation as
the Operation type in Step 4. This process is covered later in
the training.

52 • Rock physics Quantitative Interpretation


7. Open the Output tab and select the parameters to be
estimated.

~ Ch.«k aD ~ Elas3c Parameter Erination Study


~ Aoous6c impedance Acoustic irr.-pedance_Study 4 (Rock phys1cs]

fi] Shear impedanoe: Shear impedance_Study 4 [Rock ~ics)

~ VpiVs ratio: VpVs ratio_Sb..dy 4 {Rock ~h}'Sics]

~ Poisson's ratio: Poisson ratio_Study4 [Rock physics]

I'll Young's modulus: Young modL.io.B_Study 4 [Rock physics]

~ Bulk modulus: Bulk modulus_Sludy 4 [Rock ph~"Sics

~Lame constant Lambda_Siuctt4 [Rock physics]

~ Shear modulus: Shear modulus_Study ~ [Rode prpics)

11) Lambda-Rho Lambda-Rho_Study4 (Rock physics]

~ Mu-Rho: Mu-Rho_Slud-;4 [Rock physics}

Acoustic impedance, Shear impedance, VpNs ratio, Poisson's


ratio, Lambda-Rho, and Mu-Rho are common choices for
fluid-related seismic studies.

Quantitative Interpretation Rock physics • 53


NOTE: The Add suffix button allows you to add the same
suffix to all of the output from the same study. This
functionality can help you track results. Using this button is
recommended for project and data management.
8. Click Apply to calculate the parameters and leave the Rock
physics dialog box open or click OK to calculate the
parameters and close the Rock physics dialog box.
Outputs from the process are created in three different
locations in the Input pane.
• The suite of logs output for this study is created in these
locations:
• Wells> Global well logs> Study name folder
• Wells> Well> Well logs> Study name folder
• A study is created with the same name as the study used
in the process in the Wells> Well> Rock physics studies
folder.
This figure shows the Global well logs folder entry.
~ 0 Bastic Parameter Estimation Study
AI 0 Acoustic impedance_Study 4 [Rock physics}
Is 0 Shearimpedance_Study 4 [Rock physics)
~ 0 VpVs ratio_Study 4 [Rock ph~cs)
0 0 Poisson ratio_Study 4 [Rock physics!
lt*P 0 Lambda-Rho_Study 4 [Rock physics)
IJ•P 0 Mu-Rho_Study 4 [Rock physics}
M0 Young modulus_Study 4 [Rock physics]
K0 Bulk modulus_Study 4 [Rock physics]
A0 Lambda_Study 4 [Rock ph;•sics}
G0 Shea- modUus_Study 4 [Rock physics]

These figures show the well entries.


~ Bastic Parameter Estimation Study
AI 0 Acoustic impedance_Study 4 {Rock physics
Is 0 Shearimped~nce_Stud-; 4 [Rock physics)
~ 0 Vp Vs ratio _Stud-;4 [Rock physics
0 0 Poisson ratio_Stucrf 4 [Rock physics
7l*P 0 Lambda-Rho_Stuct,'4 [Rock physics)
IJ*P 0 Mu-Rho_Study 4 [Rock physics)
M0 Young modulus_Study 4 [Rock physics
K0 Bulk modulus_Study 4 [Rock physics
A0 Lambda_Study 4 [Rock physics}
G0 Shear modulus_Stuc:Pt4 [Rock physics]

~ ~ Rock physics studies


~ 8astic Parameter Estimation Study
54 • Rock physics Quantitative Interpretation
Procedure - Display the results of a rock physics study in
a Well section window
You can visualize the output data of the study in a Well section
window.
1. Click the Window in the Quick Access tool bar and select Well
section window from the Iist.

The settings dialog box for the new Well section window
appears.
2. In the X-section area of the window, select Create new
x-section.
3. In the Template area, select Create new template.
[g Select new well section window settings
n X-section - - - - - - - - - - - - - - -
~) Cre«e newx~ : X~

) Usel!ldstlr~!efnplate :

CJ AI'N71'{S do this for new wei section 1\'i'ldows

( ..- OK III(CMcel

4. Click OK.
An empty Well section window opens.

NOTE: The Rock physics operations can be found in the


Workflow editor. The Rock physics operations enable
multi-well calculations by using the different functionalities of
the Workflow editor.

Quantitative Interpretation Rock physics • 55


5. Select the check box next to the rock physics study in the
Input pane whose results you want to display.
Jj ~ Rock physics studies
~ ~ Study 1 [Rock physics] PAF
The input data and output data for the study are displayed in
the Well section window.
j~ ·.wu--. ·ll!ft!XNI'ISSTVOJ x

----~-•!!? . "-·.
~::~'G:_~._:_-4.'7;=.. ~~~: -~ ~~;:_;~~~ ......::- ·-~~Adll ~i":"~~.:::.~=~-=~~~ 3~ . ..
~f~~z.-... s...::~~c-:r~~l~"' ~ --.c:u_.~:.:...:~ ~ff~~~~~ .-:
'!:]:;:&J~1~~1:\"!"';~:..~...,:i... ~~-i~1.~.&...J.;J~Z~7~-·,~)-'f'";:"""~1 ~):':~_;,."L~~;-....,;:r.l!,~'!l"~_;..,.~r~"-'i '!J-; '!1§.
..
~t~tf~~!J'"".ai!r_t~~?i-~~ia:~\\M,~~~~,

1000

1200

I I\

You can customize the display by clicking Template settings


in the Window tool bar at the top of the Well section
window.

I~ Cs- IISSTV :_ [!( Well section !_e;.,pl ~ 1jJ

56 • Rock phys1cs Duant1tat1ve Interpretation


Procedure - Edit a rock physics study with the Inspector
An alternative method for editing a study is to use the Inspector.

1. In the Quick Access tool bar, click Inspector


The Inspector opens.

2. In the top right corner of the Inspector, click the Settings


button and change the Mode to Selection.

3. In the Petrel tree, select a study from the Rock physics


studies folder.
The Inspector tool is updated to show the settings for the
study that you select.

Quantitative Interpretation Rock physics • 57


4. To update the study, edit these settings.
~ ~ Rock physics studies

~~ ~5·~-11

Crossplots for elastic property estimates


Crossplotting rock properties is a common and useful way to assess
whether a given rock property can discriminate subsurface features of
interest, such as specific lithology types or pore fill (brine, oil, or gas);
usefully.
You saw one common cross plot earlier where Poisson's ratio was
plotted against acoustic impedance. This type of plot shows how
specific general lithologies and reservoir pore fills plot in this space
(shown in Figure 20).
So far you have used the Rock physics process to calculate elastic
parameters. These parameters can be visualized in the Well section
window along with seismic data and other logs. However, even though
the Well section window displays the data, it can be difficult to
understand relationships between different parameters by simply
inspecting the log curves. What you really need is a way to visualize the
results easily so that you can see the relationships between the
different constants and determine if they are useful in highlighting the
things in which you are interested.

58 • Rock physics Quantitative Interpretation


Acoustic Impedance vs. Poisson's Ratio Acoustic Impedance vs. Poisson's Ratio

Poisson's Ratio Poisson's Ratio

Figure 20 ·Elastic property crossplots

The Ql crossplot window allows you to analyze your data and


generate multiple selection to classify the information. In Petrel2015,
the Ql crossplot window supports these data:
• Welllogs
• Pre-stack and post-stack seismic data
• Horizon interpretation attributes
• Surface attributes
• Point attributes

Procedure - Crossplot elastic property estimates


This procedure describes how to create cross plots from the results of
an elastic parameter estimation.
1. On the Quantitative Interpretation tab, in the Ql Crossplot
group, click New QJ crossplot window.
~
[?iJ Collection _ •
~ Collec:ion settings
New Ql crossplot New
L _ window collection ~ Classncarion data
Ql crossplo·

An empty Ql crossplot window appears in the Petrel


window with a single empty collection.
NOTE: Crossplot collections are used to store a set of objects
to be plotted inside the crossplot canvas. Crossplot collections
apply a specific set of rules in an organized way. By default,
each crossplot window contains a collection when it is created.
Quantitative Interpretation Rock physics • 59
Additional collections can be added to plot additional objects.
You see an example of this process later in Module 2.
2. To display data in the crossplot, access the Settings dialog
box for the collection; On the Quantitative Interpretation
tab, in the Ql crossplot group, click Collection settings.
l1l] ~ Collection 1

New QI crossplot
New t.® Collection settings I
window collection lSi Classification data
QI crossplot

The Settings dialog box opens.


_G Settings for 'Collection 1'

X ( ~---------~-----~-~-------~-

y [~----------------------~
Z:

Pointattnbutes - - - - - - - - - -

.., Color --
__. "--- ----------

3. Open the Settings tab and then the Dimensions subtab.


4. In the Input pane, select the well whose data you want to
display in the cross plot by selecting the check box next to the
well name.
5. Select the data to display from the elastic properties listed
under the study node in the Global well logs folder.
The data that you select populates the Available objects area
of the Settings dialog box.
,. ~ Bastic Property Estimation Study
AI ~ Acoustic imped~nce_Bastic Property Estimation Study
Is [] =·
-J; [] VpVs ratio_Bastic Property &timation Study
cr ~ Poisson ratio_Bastic Property Estimation Study
M [] Young modulus_Bastic Property Estimation Study
K [] Bulk modulus_Bastic Property Estimation Study
A [] l.ambda_8l!stit Property Estimation Study
G [] Shear modulus_Bastic Property Estimation Study
ll*P [ ] Lambda-Rho_Elastic Property Estimation Study
1-'*P [] P4J-fflo_8astic Property Estimation Study
60 • Rock physics Quantitative Interpretation
6. In the Available objects area in the Settings dialog box,
select the check boxes for the properties to be plotted.
The lists in the Axes area are populated based on your
selections. Ensure that the correct data is selected for each
axis. If not, select the correct nnes from the lists.
7. To display the data, click Apply or OK
lJ Settings for 'Collection 1' '--~-

A Globol well logs

~ cr Poisson mtio_Study 1 [Rock p


~ AI Acoustic impedtmce_Study 1
0 sw X: lPoisson rmio_Study 1 [Rock physics)
.,. Wells
Y: lAcousbc impedance_Study 1 (Rockphysics)
0 ~ Jupiter-1
Z: SW
------------------------~
Point ottributes - - - - - - - - - - - - - - - - - - -

NOTE: It might be necessary to adjust the range of values on


the crossplot axes. From the Windows pane, access the
Settings dialog box for the Crossplot window. An example is
shown in the figure .

X: @ Lineor 0 Logorithmic

Y: @ Lineor 0 Logorithmic

Axis ronge
Min. Max.
X: 0.2 ~ 0.5 ~ EJ Lock
Y: 2 ~ 14 ~ 0 Lock

.t Apply ll ..1 OK II ~< Cancel

Quantitative Interpretation Rock physics • 61


The crossplot now displays the selected data.
8. If you do not want the data to be displayed with the same
color as the well, go to the Advanced tab in the Settings
dialog box for the collection and clear the Use well color as
point color option.
The crossplot in this figure shows a well-defined grouping of
data points not unlike the sand-shale trends in the stylized plot
shown in the second figure. There also is a cluster of data
points in the lower left of the plot. These data points have a
lower Poisson's ratio and lower acoustic impedance than the
rest of the data.

13

12
........
. . .:
"'··...
... ..... ~
~ni
0..
11
·. ...... . •.....'II',·:• ...
-~
'·~
~
w
~
10 .. .... ... . .
.c
Q.
.\ .,:1 .. ..
..><:
u
. : ~.r· ..
. ..-. .. .......
0
~ • ·.c ...
~ .:~,:-..
>.
'0
2
1/)
I
B
lij
'0
Q)

-~
~0
~

021 024 on ow on o• 0 .39 042 045


Poisson ratio_Study 1 [Rock physics]

62 • Rock physics Quantitative Interpretation


Acoustic Impedance vs. Poisson's Ratio

Poisson's Ratio

It would be helpful to know where these points are in the well


and whether they are associated with hydrocarbons or they
correspond to a specific lithology. First, you perform an exercise
to estimate the elastic properties and make a crossplot. Then
you see how to find out where these points are in the well.

Quantitative Interpretation Rock physics • 63


Exercise 1 - Estimate and crossplot elastic parameters
from a ·rock physics study
Perform an elastic properties estimation for the Jupiter 1 well using the
workflow described in this lesson. As an aid, the steps are described
briefly in this exercise. For more details, refer to the procedures that
you just reviewed, if necessary.
1. Create a new rock physics study; on the Quantitative
Interpretation tab (make sure the Advanced Geophysics
perspective is selected), click Rock physics.
2. Select Create new and name the study.
3. Choose the well (Jupiter 1).
4. Select Elastic properties estimation as the Operation type.
5. Select Using Vp, Vs and Density as input log data types.
6. Select the appropriate log curves to use for the calculations
from the list or from the Input pane.
7. Open the Output tab and select the desired output.
8. To run the estimation, click OK
9. Display some of the results from the estimation in a Well
section window.

10. Create a new crossplot window from the Ql crossplot group.


11. Click Collection settings from the Ql crossplot group to
open the Settings dialog box for the collection.

64 • Rock physics Quantitative Interpretation


12. Open the Settings tab and then the Dimensions subtab.
13. Select the well and the data for the crossplot from the Input
pane. For example, you can display the acoustic impedance
versus the Poisson's ration.
14. Choose the axes on which to plot the well and data.
15. Display the window side by side with the Well section
window.
What can you observe?

Poisson raboEiasbc Propeltf E~mabon study

Quantitative Interpretation Rock physics • 65


Procedure - Create classifications
To link the area of the cross plot with the logs in the Well section
window, you must learn how to create selections from the cross plot
display and classifications that you use to display the selection in the
Well section window. This procedure describes how to create
selections and classifications.
1. Click the Polygon selection tool in the crossplot window
tool bar.
When this tool is selected, you can click in the crossplot to
select an area of interest.

.... .. ~~.

·. :.-::.::.::...
. '·

~
·,.~· 'I'"
.. ..
. \i: .•
.,;

. .....

41,- r l -r> T
Oll
'T ' ' ' ' T r' '
0 24 027

66 • Rock physics Quantitative Interpretation


An entry for the selection appears in the Crossplot
selections folder in the Input pane.
~ ~ Crossplot selections
~ 0 Potential Gas Sand
~ 0 Crossplot selection 1
~ 0 Crossplot selection 2
~
.. 0 Crossplot selection 4

2. On the Window tab, on the Ql Crossplot tab, click


Classification data.

QI Crossplot

Class!icatlon
da a
Classification data

3. Give a name to the new multi-selection classification and


insert the crossplot selections from the Input pane. Edit
the colors if necessary.
4. Click OK The new log is located in the Rock Physics study
folder.
j- lnput • q.

4 111.$. ~ Wells
1-
4 ~ Global we/flogs
I> A· Log attributes
0 Rock Physics Input
0 Petrophysical Properties
4
0
~
AI
I. ..
Synthetic

~ Acoustic impedance_Study 1 [Rock physics


Is 0 Shear impedance_Study 1 [Rock physics]
-fi 0 VpVs ratio_Study 1 [Rock physics]
0" ~ Poisson ratio_Study 1 [Rock physics]
M 0 Young modulus_Study 1 [Rock physics)
K 0 Bulk modulus_Study 1 [Rock physics]
A 0 Lambda_Study 1 [Rock physics]
G 0 Shear modulus_Study 1 [Rock physics]
7t•P 0 Lambda·Rho_Study 1 [Rock physics]
~·P 0 Mu-Rho_Study 1 [Rock physics]
~uh 0 Collection 1 Multi-selection

Quantitative Interpretation Rock physics • 67


The selection appears in a track in the Well section window .
. _; __!_]

•. [
l
~
r[

f
I
L
I

Folsoon ratio_Study 1 [Rock physics]

The classification can be compared to the logs in the study and


to other logs (for example, resistivity) to understand what the
selection corresponds to. In this case, the red selection exactly
identifies the pay interval in the well.
The polygon selection can be edited interactively to see the
effect of including different parts of the crossplot.

Exercise 2 - Create an interactive crossplot


In this exercise, you create a selection and visualize it in the Well
section window. Refer to "Procedure- Create classifications" for
more detail, if necessary.
1. Select an area that could correspond to gas sands on the
crossplot that you created earlier using the Polygon
selection tool in the crossplot window tool bar.
2. Find the Crossplot selections folder in the Input pane.
3. Click Classification data and insert the selection just created.
4. Locate the Collection entry in the study in the Global well
logs folder and select it to display it in the Well section
window.
5. Compare the selection to the other log data.
6. Try editing the polygon and selecting new areas to see what
they correspond to in the logs.
68 • Rock physics Quantitative Interpretation
Procedure - Create a 30 crossplot
Crossplots easily can be extended from two to three dimensions.
Extending from two to three dimensions allows the data to be plotted
against a third parameter, which potentially can help to better
discriminate specific features in the well.
1. Create a 20 crossplot.
2. Click the 20/30 button in the cross plot window tool bar.
The crossplot expands to a third dimension.
~ ~ ~~~~~~~ ~ vj~l~
1

A.oouatic impedlmce_Study 1 [Rock pbysics][MPa.slm ]


4 5

/ •: -~-'--"-~ ---~- L-' --


6 1 & 9 10 II 12

I-·::\
JJ

,I oo._:; _- _ o~ ~·
10.33- - 0.33~
~0.30 - - 63() ~

6- 0~7 - - ()-2' \.

l
.E. 0.14 - - 0 .1.4 ' ~

r J.o

i
,- 0 . 11

Acoustic impedaoce_Siudy 1 [Rock physica)(Ml'a.slml

3. On the Quantitative Interpretation tab, in the Crossplot


group, click Collection settings.
...----------
& Pk,~ Collect ion 1

New Ql crossplot N ew
!"@
t: Collection settings I.
window collec ion [Si Classi'ica ion dat a
Ql crossplot

The Settings dialog box for the collection opens.

Quantitative Interpretation Rock physics • 69


4. Select a property from the rock property study by selecting the
check box next to the property entry in the Global well logs
folder.
The property is added to the Available objects area and is
plotted in the third dimension of the crossplot.
Jij Settings for 'Collection 1'

8 tnto-

A Global well logs Dimensions L.;,_,_-'..L,__.__ _ _ --~._Au_t_anced


_ _ _ _ ___, ,

ftl o Poisson r~~tio_Study 1 [Roc.lcp 1. Axes - - - - - - - - - - - - -


~ AI Acoustic Jm pedance_Study 1 I
[i] SW X: [POISsonmtio_Siudy 1 [RockphysicsJ

A Wells . y-
1 [Acoustic mpedence Ptudy 1 [Rock physics]
~ ~ Jupiter-1 1
IZ: [._S.¥;_'1_ _ _ _ _ _ _ _ _ _ _ .~

· Pointllttributes - - - - - - - - - - - [ }

[i] Color: .._[


s~_v_ _ _ _ _ _ ___,
• !~ [J

4 ' 6 7 t f H ll U U

70 • Rock physics Quantitative Interpretation


Lesson 6- Fluid substitutions in rock physics
When exploring for hydrocarbons, a key question often asked is: What
would the log and seismic response be if the pore fill in a particular
sand was different? If you have a gas reservoir, you might want to see
what the brine filled response would be, or vice versa, and what the
response would be to oil and different GOR scenarios.
You can explore different scenarios by using Gassmann's equations to
perform a fluid substitution. Gassmann's equations were discussed
earlier and underlie the fluid substitution workflows in Petrel.

Procedure- Perform a fluid substitution


This procedure describes how to perform a fluid substitution.
1. On the Quantitative Interpretation tab, in the Rock
physics group, click Rock physics.
{8 ~ Log .:,;tin· ator
ij Log cal::ulator
Rock
p hysics m ~I tool.;
Roclc p hysics

The Rock physics dialog box opens.


2. Select to create new study or edit an existing study.
For illustration purposes in this procedure, Create new is
selected.
3. In the Operation type field, select Fluid substitution from the
list.
-- --- -
--- I >)(

~--------------------------

_~ Create new: Fluid St.bsti u ·on 1

:_ Edit existi g '~ IEPI Study 2 [Rock ph:srcsJ T I


\~leU: § [.._~..:.._Ju_pit_e_r-1_ _ _ _ _ _ _ _ _ __...;1
Operation type: n ____________ _ _ _ ___,~
[.._A_ui_d_su_bs_ti_b_·o_ •]

Quantitative Interpretation Rock physics • 71


4. Select the Method.
There are three options
• Fluid substitution substitutes one type of pore fluid for
another
• Inverse relation calculates the dry response from the
saturated logs. Dry in this context means the response of
the rock with no pore fill.
• Simple relation calculates the saturated response from the
dry response

For illustration purposes in this procedure, Fluid substitution


is selected.
5. On the Input tab, select the appropriate logs.
For the Fluid substitution method, you would select the in-situ
log suite. If the Simple relation method was selected, you
would select the calculated dry case logs.
NOTE: The dry case can be selected as the output for a fluid
substitution exercise.

NOTE: The Rock physics process outputs the velocity after


substitution in the form of velocity logs (not transit time logs).
To help you compare the input and output logs, it is easier if
velocity logs are used as input, although transit time logs are
acceptable input.
lnp t ReseMJ,r CQnd,bQns

v., p·V{CV{j V{tfocsly ·I


V. S ·wave ve-.loctty ·I
p Oenstty ~ ~RHoe ·I
~ Porosi ~~ ~PH1E ·I

n • Rock physics Quantitative Interpretation


6. Open the Output tab and complete these steps:
a. Select the properties that you want to see modeled in the
substitution.
b. Select to calculate the P wave velocity sat, S wave
velocity sat, and Density sat.
rnput Resavoir conditions

[iJ Check aD [ Add suffix J Ruid Substitution 1

~ P-·,o,-ave velccit>J sat P-;uave 1Je!cci:t.y sat_Fluid Substiiution 1

rlJ 5-·....ave velccit-; sat S-VI't!Ye velocit'{ sat_Fluid Substitution 1

11] Dens~• sat

~ Ksat K sat_Fluid S bstitution 1

~ P-v;ave velcciF; dry: P-w-ave velacro; dry_Fiuid Subslitulioo

~ 5-v.ave velocro1 dry:

rlJ Den:sir; dry: Rho Gflf_ luid Su:bstitufun 1

111 Kdrr K dry_Huid Substifutic 1

~ Shear modulu~: M _Fluid Sub:stitLlticn 1

~ Density mineral: Rho minerai_Fiuid S bstitutioo 1

[l] K mineral K miroerai_Fhrid Substib...rtioo 1

[{] Densir; fluid: Rho fluid_ luid Substi:tuticn 1

~ Kftuid:

~ Oensir; fluid odelod: Rho fluid modeled_Ruid Subslitutioo 1

~ K fluid modeled: K fluid modeled_Fiuid Substitutio

Quantitative Interpretation Rock physics • 73


7. Open the Mineral properties tab.

0 Reuss

I~ Use presets T J
Name Mineral fraction Bulk modulus Density
~""--===--=---

Albite 100.00 % 5581 1.00 MPa -.. 2590.00 kg/m3 -..

8. To select preset values for different lithologies, use the Use


presets button on the right side of the tab.
In this case, select quartz. You can select several minerals if
the matrix has mixed mineralogy. You can enter the mineral
fractions manually or use mineral fraction logs, if available. If
you want to use logs to specify the mineral fraction, select the
Use mineral fraction check box.
9. If more than one mineralogy is present, select the mixing
method to be used.
NOTE: Remember, the Hill method calculates the average of
the Voigt and Reuss methods.

74 • Rock physics Quantitative Interpretation


10. Open the Reservoir conditions tab.
This tab is where reservoir conditions, in-situ fluid properties,
and fluid properties for the substituted fluids are specified.
fJI Rock physics ULJ
Rock pflysics I
@Ciliate ew· Auid substitution 1

I{) Edit 9X!Sbng· fC)l ([!lj Study l [Rack phys!CS] ·l


t wen. ~ (3:t Jupiler-1 ... ) ~
Operation type· IFluid substitutlon ·I
Method
@ Fluid sUbstitutioo 0 Inverse relaton 0 Simple relation [iJ
~ Reservtlil' condibons I Mineral properties I Output .j
MIXing method: Hil C) Voigt (() Reuss
Measured conditions 0 Mode!sd conditions
~
Fluid sutumtion

0 u~ solumon fogs I0 Use satui'ISOOn togs

fiJ I
~I
6 Gas 35.00% 90.00 %

~I
Oil 0.00 % [tJ 0.00 %

[tJ ..
~ Weter [{] 65.00 % 10.00 %
iiiiiil
Fluid properties 13
I
GIIS gravity (sg 8ir): 0.6000
0 0.6000
Oil gravity (API): -J 300 -[}-- 30.0
GOR (sm'/sm'):

J
-J 100.0
-o 100.0
Salinity {ppm) 3800.0
0 3800.0
LJ Include gas weter mlio 0 Include gas m~ler ratio
Reservoir condrttons

I~
P Pressure: D
--
6000.00 MPll ... 6000.00 MPll ...
T T~pereture: 0 so.oo ·c ... 80.00 'C ...

O rutomn D Aulo update r1l


1./ Apply n~ OK j[ J< caooe1 j

There are three main sections:


• The Fluid saturation section describes the Measured
conditions and Modeled conditions. Measured conditions are
the starting fluids and saturations (left side). Modeled
conditions are the conditions after fluids are substituted (right
side). The mixing method can be specified and there are check
boxes to allow saturation logs to be used.
• In the Reservoir properties section, the pressure and
temperature of the measured and modeled reservoir are
entered.
Ouant1tat1ve Interpretation Rock physics • 75
• In the Fluid properties section, gas gravity, oil gravity, and
GOR are specified as well as salinity for the brine fraction in
the reservoir. There are slider bars for both the measured and
modeled conditions. The values selected are displayed in the
boxes on the right side of the tab.
11. Enter parameters.
• Choose the desired Mixing method. Because you are
mixing fluids here, the Reuss average method might be the
most appropriate.
• Use the check boxes next to Gas, Oil, and Water to add or
remove a fluid from the starting conditions (left column) or
substituted conditions (right column).
• If you have saturation logs and want to use them, select
the Saturation log check boxes; otherwise, enter values
(or use the vertical slider bars) to adjust the saturations of
measured and modeled reservoirs.
• Enter the reservoir pressures and temperatures, or select
the check boxes to the left of the data entry windows if
pressure and temperature logs are available. Adjust the
units from the lists next to each check box.
• Adjust the values in the Fluid properties section as
necessary. Values can be entered or the slider bars can be
adjusted for both the measured and modeled values. There
are default values listed, but it is important to make good
selections because these values affect the modeled
response.
12. To limit the fluid substitution to a depth range of interest (for
example, a specific sand or pay interval), set a depth filter by

clicking this button [~] to the right of the We// field in the
upper part of the dialog box.
~ :X
- --- - - - - - - - - -- - - - -

Fluid Substitution 1

...) Edit existing. ., I


~/elf· ~( ~ upiter-1

Opera ·on type: (..._F_


Ju_id_su_bs_-ti_
tu_tio_n_ _ _ _ _ _ _ _ _ _ ___J
• I
76 • Rock physics Quantitative Interpretation
13. If a depth range has not been set already, complete these
steps.
a. Make the Well section window active.
b. On the Qualitative Interpretation tab, in the Rock
physics group, click Q/ tools.

Rock physics

c. On the 01 Tool Palette, click Create/edit well depth


range.
• Tool Palette 0 X

~ Create/edit well depth r ...


X

Quantitative Interpretation Rock physics • 77


d. Click in the Well section window at the top of the depth
range. While holding the mouse button down, slide the
cursor to the bottom of the depth range then release the
mouse button. The zone selected becomes highlighted.
(Check the Well section window settings if it is not
visible.)

The depth range appears as an entry in the tree in the


Filters folder.
~ - ~ R~;""""'"' . . . . . . . ,"
~ ~:
I .

e. Repeat to set additional ranges if desired.


f. Highlight this entry and click ~ to enter it into the Depth
range window of the Rock physics process. Enter as
many ranges as necessary to define the zones for the
substitution.
Depth range Window @
~j Depth range Top Base
~J 1 :; Pay zone 11915.61 (MD) 11924.92 (MD)
I
[J)

(7< Close I
g. When you are finished entering depth ranges, click Close.
78 • Rock physics Quantitative Interpretation
14. When all parameters are set, click Apply.
Entries appear in the Input pane in the Global well logs
folder, in the Rock physics studies folder. The entries also
appear in the Well that contains entries for the fluid
substituted logs.
This figure shows the Global well logs entry.
"' 0 Auid Substitution 1
vP 0 P-wave velocity sat_Auid Substitution 1
V: 0 S·wave velocity sat_Auid Substitution 1
P0 Rho sat_Fkid St.bstitwon 1
This figure shows the Rock physics studies entries.
"' ~ ~ Rock physics studies
~ 0 Bastic Parameter Estimation Stuct1
~ 0 Ruid Substitution 1 .

This figure shows the entry for the well.

yp P-'Nave velocit'J sat_Ruid Substitution 1


"V; S-'Nave velocity sat_Ruid Substitution 1
P Rho sat_Ruid Substitution 1
15. Repeat this procedure for as many studies as desired.
16. Display results in a Well section window. For details, refer
to Lesson 4, "Procedure- Display the results of a rock
physics study in a Well section window".
TIP: Make sure that all logs displayed in the same track have
the same scaling.

Quantitative Interpretation Rock physics • 79


Exercise 1 -Perform a fluid substitution
In this exercise, you perform fluid substitution, trying different
parameters to see their influence on GOR and oil gravity.
1. Perform a fluid substitution using the workflow described in
this lesson and the parameters in this table.

Reservoir Case Full of Case Full of


Conditions Base Case Gas brine
35% 100% 0%
Gas Saturation
Water 65% 0% 100%
Saturation
Pressure 6000 psi 6000 psi 6000 psi
Temperature 80°C 80°C 80°C
Mineral 100% Quartz 100% Quartz 100% Quartz
Properties
Use the same logs you used in the previous exercise as well as
the effective porosity.
2. Perform an additional fluid substitution case using the base
case from the table in Step 1 and substituting 90% oil and
10% brine.
3. To test the s·ensitivity to oil gravity and GOR interactively, use
the Fluid Properties slider bars.
How much of an effect do GOR and oil gravity have on the
result?
An example is shown in this figure. In the density and velocity
tracks, the base case is in black, with oil in green, gas in red,
and brine in blue.

80 • Rock physics Quantitative Interpretation


The red curves in the density and velocity tracks represent high
saturation gas, the black curves are the in-situ conditions, and
the green and blue curves represent oil and brine cases.

TIP: You can use settings in the Rock physics dialog box to
adjust the display interactively. In the Rock physics dialog
box, select Edit existing and the study that you want to edit.
The study parameters load into the window. To adjust the
saturation and fluid properties, use the slider bars. The plot
updates interactively.

The Inspector tool also can be used to perform the same


process. Select Inspector and set the Inspector Mode to
Selection. Select the study that you want to edit in the Rock
physics studies folder in the Input pane. The Inspector
loads the study parameters, which then can be edited.

Quantitative Interpretation Rock physics • 81


Lesson 7 -Shear velocity estimation from
empirical relationships
Not all wells have full log suites. It is common for older wells not to
have shear logs. It sometimes is necessary to estimate shear wave
velocity values to model potential AVO behavior. Much work has been
done to study the relationship between Pwave and shear wave
velocities and many companies have their own proprietary regressions.
Two sets of results commonly used in the industry were developed by
Castagna at Area (Castagna et al., 1985). Castagna and Greenberg
(Greenberg and Castagna, 1992) later extended these results.
The 1985 Area Mudrock line developed by Castagna is a regression that
expresses shear velocity as a function of Vp and takes this form:
Vs=(Vp-1360)/1.16 (m/sec)
The coefficients clearly are dependent on the sample set used in the
derivation. Best results are achieved if local calibration is done.
The Greenberg-Castagna relationships are a set of equations that relate
Vs to Vp for different lithologies. Regressions were developed for
shales, sandstones, limestones, and dolomites. In Petrel, you can use
these relationships in combination with another log and a cutoff value
to differentiate between lithology types (for example, a gamma ray with
a cutoff to differentiate between sands and shales). Again, local
calibration helps improve results.
The Greenberg and Castagna equations for sands and shales are
• Sands Vs=- 855.88t0.80416Vp (m/sec)
• Shales Vs=- 867.35t0.76969Vp (m/sec)

82 • Rock physics Quantitative Interpretation


Procedure - Estimate shear velocities from P wave
velocities
The shear velocity estimation requires a P-wave velocity log. In Rock
physics, two methods are available to compute the shear wave.
However, measured data is always preferable as modeled data
introduces a lot uncertainties.
1. On the Quantitative Interpretation tab, in the Rock physics
group, click Rock physics to open the Rock physics dialog
box.

2. Select Create new and select the well to use.


If desired, a depth window for the estimation can be set by
clicking the symbol to the right of the well entry.
3. Select Shear velocity estimation as the Operation type.
fB3 Rock physics

Q Cfeote 1ew: Shear Velccity Esli1116tion

' ,J EdiiE!XlSttng:

WeU: ~~....:.
[~ ~Jup:......iter_-1_ _ _ _ _ _ _ _ _......)
Operaton type· [ SheM veloctty estil116tion

Quantitative Interpretation Rock physics • 83


4. Open the Input tab.
5. Use the P wave velocity list (or ~)to enter the Pvelocity
log.
Transit time or velocity logs are acceptable input; however,
comparison of input and output is improved if a velocity log is
selected as input.
6. In the Compute shear wave section, select one of these
options: Mudrock line or Greenberg-Castagna.
The Greenberg-Castagna method requires selection of a
Filter log that is used to discriminate between lithologies. The
lithology calculated in turn determines which of the Greenberg-
Castagna relations are used to calculate the shear velocity. A
cutoff value must be selected to make the lithology selection.
7. Select the Filter log to use from the list or insert ~ a
selection from the Input pane.
8. To select a cutoff value for the two lithology types, use the
slider bar.
9. Select the regressions to use based on whether the regression
is to be applied to values greater than (Log>=) or less than
(Log<=) the cutoff.

l iv;,n~P-v.~.-e
l
0ul!l'Jt --------------~ 1
~~ ~
'-1

I I
[ Atp OT ?

Con·p1.~s earV>?Ie I
I
' M10drodc lu-e '
Vs= tvp· ~}!~_(n-/s ) 0

Fil:~~- ~( Y GH ·I
Culof- ----(}-- ~ ~1

Lop = Cuto;:< z IShale • I [QJ


VP -357~C>C(C<I . ' .Ti!OilCOC"Vp+ ((jOCCCr Vp"2( rr: s)

log <Cu:nH z ISandstone · I (Q]


\'s= =3ss~>CooC • .Sl):l20000 Vp+ COOC((C Vp"2( rr:s)

84 • Rock physics Quantitative Interpretation


10. Open the Output tab and ensure that the desired output is
selected.
The name of the output can be edited and a suffix appended to
help with data management.

~ Ct-eck aD [ Add suffix ] Shear Velocity Es!im<!!ion

~ 5-\J;ave velocitr s-·~.-ave vefocitl;_Shear VokJcit.J Estimation

11. To create the new shear wave log, click Apply or OK.

Exercise 1 (Optional) - Run a shear velocity estimation


In this exercise, you run a shear velocity estimation and compare it with
the existing shear log in the dataset.
1. Use "Procedure- Estimate shear velocities from Pwave
velocities" to create a new shear log (Vs) from the existing Vp
for the Jupiter well.
2. Display the results in a track of the Well section window
and compare the results to the existing shear log.
How good is the fit?
3. Adjust parameters and rerun if necessary.

Quantitative Interpretation Rock physics • 85


Results and discussion
An example of shear velocity estimation is shown in Figure 21. This
example uses the Greenberg-Castagna regressions and the gamma ray
log to determine cutoffs between sand and shale. The result generally
is good, but there is disagreement in the area of the pay sand, which is
highlighted in green. This disagreement is to be expected because the
regressions assume that the sands are brine filled. The estimated shear
log is displayed in red and the actual shear log is displayed in black.

Figure 21 Results of using a Greenburg and Castagna shear velocity estimation

86 • Rock physics Quantitative Interpretation


Lesson 8- Log blocking Backus workflow
Backus (1962) studied the effects of small scale layering on the velocity
of sound waves where the wavelength of the waves is much larger than
the individual layer thickness. He derived results that allow you to
calculate the effective velocity of such waves by averaging the
properties of the thinner layers. The method effectively replaces the
many thin layers with a thicker layer. This thicker layer has properties
that are equivalent to the thin layers as far as a long wavelength
seismic pulse is concerned. Backus averaging can be performed in the
Rock physics process of Petrel by using the Log blocking Backus
method.
Backus averaging (Backus, G. E. (1962)) can be useful when you want to
remove fine scale detail from well logs, especially where the detail is
well below seismic resolution. It creates blocked logs with layers that
represent the major geological units and with appropriate average
properties. Because of the layering, the resulting average layers are
anisotropic. The associated Thomsen parameters (Thomsen, Leon
(1986)) also can be calculated.
This workflow has many applications. For example, you can use it when
you want to have a simplified log for seismic modeling or to create
blocked logs that can be used for creating wedge models.
Inputs for the Log blocking Backus operation are Vp, Vs, and density.
Additionally, a blocking log is used to define the major units. These
units or blocks are defined by recursive median filtering of this log and
then by picking the largest discontinuities in the filtered log. These units
then are averaged with the Backus formula. The additional user-
selected parameters are
• Blocking factor: Specified as a fraction. The number of blocks
picked is equal to the Blocking factor times the number of input
samples.
• Minimum thickness: Redefines the blocks to remove those blocks
that are thinner than the minimum thickness specified. The blocks
generated are equal to or larger than the specified minimum value.
The output consists of the blocked logs and Thomsen anisotropies
(delta, epsilon, and gamma) for the blocks.

Quantitative Interpretation Rock physics • 87


Procedure - Run the Log blocking Backus workflow
The Log blocking Backus operation requires specific logs as input. You
can set some of the parameters and select the logs to be output.
1. On the Quantitative Interpretation tab, in the Rock physics
group, click Rock physics to open the Rock physics process
dialog box.
2. Create a new study and select the well to use. If desired, a
depth window for the estimation can be set by clicking the
symbol to the right of the well entry.
f!j Rock physics

~~ 1,'1/ell: § [..._~
.;.__
Ju_pit_e_r-1_ _ _ _ _ _ _ _ _ _ _...,-'1

Opera ·on type: [Log blocking Backus • J

3. Select Log blocking Backus as the Operation type.


4. On the Input tab, select the logs for blocking.
lnpu1 Output
~-----------------------------~

~: v; vp

~; v. Vs

~~ q 8 RHOB

~( AJ ~~_EP

Para eiers

Blocking f8Ctor
Q 0 100

M1mmum ttudcness 0··-- 000 m

5. Select a Blocking log to be used to define the major units in


the blocking.
6. Select an initial va lue for the Blocking factor and Minimum
thickness.

88 • Rock physics Quantitative Interpretation


7. On the Output tab, select the logs to be output.

~ Check all ( Add suffix ) Log Blocking Backus

[ll P-wave v~locity: P-wave velocity_l og Blocki g Backus

rlJ S-wave velocity: $-wave velocity_ og Blocking Backus

IZJ Rho: Rho_Log Blocking Backus

[1] Delta: Delta_Log Blocking Backus

rlJ Gamma: Gamma_log IBiocking Backus

rlJ Epsilon: Epsilon_l og Blocking Backus

B. Click Apply or OK

Quantitative Interpretation Rock physics • 89


Exercise 1 - Perform Backus log blocking
In this exercise, you perform a Backus log blocking operation. You then
display and interactively change the result to get the most suitable
parameters.
1. Run the workflow described in "Procedure- Run the Log
blocking Backus workflow" and display the results in a Well
section window.
2. To visualize the results, overlay the blocked and unblocked logs
in the same track for comparison. To help with the comparison,
make sure the minimum and maximum limits for the two logs
are set to the same range.

3. To assess when the best fit is obtained between blocked and


unblocked logs, interactively adjust the Blocking factor and
Minimum thickness.

90 • Rock physics Quantitative Interpretation


These figures shows examples of the parameters and results.

yp P-vFave velocit>f- ~ ( Vj, Vp

~ S-11.-ave velocity: ~ [ v; vs

P Density: ~( '?B RHOB

~ Blocking log: ~ [ AI Aooustic impedance Study 1 {Rock physics] PAF • J

Par.:meters --------------------------------u
Blocking factor: ~-------· ----- 0137

o.aa m

Quantitative Interpretation Rock physics • 91


These figures show an example using a different parameter
set.
Input

Vj, P-waYe velocity: ~( Vj,Vp

V: S-wave velocity: ~ [ v; vs

~( q8 RHOB
¥. Blocking log: ~ (AI Acoos6cimpedance_Sb.Jdy 1[Rock: physics] PAF "' J
Parameters

Blocking factor: -or---~~~~- 0.1&7

Minimum ihickness -of---------- 20.00 rr

92 • Rock physics Quantitative Interpretation


Review questions
• Name some elastic constants commonly used by geophysicists.
• What are the two types of seismic body waves?
• What happens to Vp and Vs when gas replaces brine in a
porous unconsolidated sand and why?
• What happens to Poisson's ratio?
• What workflow do you use to change the pore fill and
recalculate the Pwave velocity?
• What elastic parameters do Vp and Vs depend on?
• What are some well-known empirical rock property
relationships?
• What three methods can be used in Petrel to mix fluids or the
mineral composition of reservoir rocks? What elastic constants
are these methods calculating?
• Who was Backus and what did he do?

Summary
In this module, you learned how to run the various operations in the
Rock physics process and display and interactively crossplot results to
visualize the rock physics relationships. You performed these specific
operations:
• Elastic parameter estimation
• Fluid substitution
• Shear velocity calculation
• Log blocking Backus
You also learned about some basic theory regarding elastic moduli,
empirical rock physics relationships, and Gassmann fluid substitution.

Quantitative Interpretation Rock physics • 93


Module 2- AVO modeling

AVO modeling applies a variety of convolutional algorithms to compute


AVA responses. It uses the P-wave velocity, the S-wave velocity, the
density, and a wavelet as input to create synthetic gathers.
One approach is to perform fluid substitution before computing the
synthetic seismogram. The different scenarios and their seismic
responses then can be compared with the real dataset and used as a
reference in the future.

Prerequisites
To complete this module successfully, you must have completed the
previous module in the course.

Learning objectives
After completing this module, you will know about:
• different equations that describe the relationships between the
amplitudes of incident and reflected seismic waves including:
• Zoeppritz equations
• approximations that are used currently in Petrel for AVO
modeling and inversion.
• classification of AVO responses into different commonly used
classes
• how to set up and perform an AVO modeling study
• how to create crossplots to analyze the results
• how to create a wedge model

Quantitative Interpretation AVO modeling • 95


Lesson 1 -AVO modeling workflow
In this module, you learn about the AVO modeling workflow component
of the Quantitative Interpretation suite. Figure 1 shows a flow
diagram that provides an overview of the workflow.
Key inputs to the workflow are Vp, Vs, and density logs and a wavelet
for synthetic seismic modeling purposes. These inputs are used to
create gathers that model the AVO response. The gathers can be quality
checked in the Well section window along with the logs used to
create them.
The AVO response of the synthetic gathers and what class of AVO
behavior it shows can be evaluated and assessed event by event using
Z Levels. You can extract attributes and use them in crossplots to
identify pay zones and lithologies of interest.
This module discusses the AVO modeling workflow and how to run each
step.

96 • AVO modeling Quantitative Interpretation


- r-- r:-"' ~

I•i ~
I

~ i
g
VI
~~
5· ffi
GJ

~
0 :3 (6

t- c ~
0.
(') (p

I .__ ~ r- I ~ -[ (')
~
!; ~~

l
&
li: i'
!
1
!
0
3
r:;.) 5 '
....... :E
~
!
~

&.
Ill 0
1- ~
CD
.a; :§
:§ ~
c:
= I
(il ~ (/}
.__
0
8;a
~
1
~ )>
::: lc;
& ;i"'GJ g, 0
! c if~ <D
i5 ~ ~ Ql
{II
Ql
!,ltr.n
liil
..___
-~
~ ~.....-
I
~
fll §,iii
g :2.
:E~
-...

~
tO
J-· I

i
fll
~
:E I

0 DDD
::0
It> p~
:::;.
;§ %
~ s :i
.,
~
'< ~ ~
::;
~
a~ ~~ s
~
~- ~ r.Q
~ "C
,.,""
~
Q;)
~
~
!I>
~ ~

Figure 1 AVO Modeling workflow

At the end of the module is a discussion on one additional workflow


that can be used to create wedge models. You can use wedge models
to investigate tuning and thickness effects.
Before you begin to learn about how to use the various components of
the AVO Modeling workflow, you review some AVO theory background
material.

Quantitative Interpretation AVO modeling • 97


Lesson 2- AVO approximations
In Module 1, you were introduced to the Zoeppritz equations during the
discussion of what happens at a reflecting interface. The Zoeppritz
equations are shown here for reference along with the definitions of the
angles and velocities of the associated layers (Figure 2).
-I
-sin e. -cos tA sin e 2 cos tA
Rp(B1 ) cos e. -sin ¢. cos e2 -sin tA Sill o.
Rs(e1)
sin 2e1
v: cos 2¢. P2vs;vp • . p 2v.S2 vPl 2¢
cos o.
Tp(B.)
___fl_
2 Sill 2()2 2 COS
2 sin 2B1
VSl p.vs.vP2 PtVSl
Ts(e1 ) -cos 2¢. -VSl
S i l. l 2¢1 P2 VP2 cos 2¢2 - P2 VS2 sin 2¢2 cos 2¢.
VPl PtVPl PtVPl

Figure 2 describes the angle, velocity, and density terms in the


equations. The quantities Rp{9 1) and Rs(9 1 ) refer to the reflected Pand
S wave amplitudes, and Tp(9 1) and Ts(9 1) refer to the transmitted P
wave and S wave amplitudes.

Reflected S
lncidentP

Transmitted P

Transmitted S

Figure 2 An incident P wave and the resulting energy that is reflected and
transmitted. The terminology is the same as in the Zoeppritz equations.

98 • AVO modeling Quantitative Interpretation


While these equations provide a complete description of the amplitudes
of the reflected and transmitted waves, they are difficult to understand
intuitively. Because the publication of Zoeppritz results in 1919, many
workers have sought to derive linear approximations to the full equation
set that are easier to calculate and that still can describe the angular
behavior accurately.
Some of the approximations were derived to emphasize the role of
specific rock properties and their impact on the reflection amplitude.
Others were derived to demonstrate the angular behavior and which
properties were important at different angle ranges. Others are used
because they allow you to calculate a range of attributes that can help
characterize the reflections in terms of lithology and fluids.
This lesson discusses the various approximations used in Petrel, in
more or less historical order.

Quantitative Interpretation AVO modeling • 99


Aki and Richards
In 1980, Aki and Richards derived a linear approximation for the
reflection amplitude of a P-P (that is, incident and reflected Pwave),
Rpp(9). Their result is shown in this equation (refer to Petrel Help):

Rpp(8) = ~ (t.,.p) (1- 4p 2 V. 2 ) + 1


tJ.Vp - 4pzv.z tJ."('s
2 p s 2 cos 2 8 ~ . s l's
In this equation:
• R (9) is the reflection coefficient for an incident Pwave that also
r:flects as a Pwave
• 9 is the average of the incident and transmitted angle for Pwaves
• Vp and Vs are the average Pand S wave velocities across the
reflecting horizon
• p is the average of the densities
• p = sinq/Vp 1 where q1 is the incident angle
The L\'s are the differences between the properties in the lower and the
upper media.
As is typical of the approximations that you review in this module, the
formula expresses the Pwave toP wave reflection coefficient, RPP (9),
in terms of some coefficients, and a set of trigonometric functions that
govern the angular part of the relationship.
The Aki-Richards formula was written to stress the dependence of the
reflection amplitude on the contrasts in the density and the Pand S
wave velocities. It often is used where these quantities are the ones of
interest.
This formula is an important relationship for several reasons. It is a
good approximation to Zoeppritz across a large range of angles, and it
is widely used, mostly as the basis for later work on approximations.
The Aki-Richards approximation compares well with Zoeppritz results
for large angle ranges, if the zero offset reflectivity is small (Li et al.,
2003).

100 • AVO modeling Quantitative Interpretation


Shuey's 3-term and 2-term approximations
Starting with the results of Aki and Richards, Shuey published another
approximation in 1985. This approximation is a three term
approximation. It describes the Pwave reflectivity as shown in this
equation (from the Petrel Help):
1 (fla t:Jp) (fla 4 fl{3 2 11p) !Ja
Rpp(8)=- - + - + - - - - - -2- s i n2 8+-(tan 2 8-sin 2 8)
2 a p .2a y 2 {3 y p Za

In this equation:
• a = average Vp across the reflecting interface
• ~ = average Vs across the interface
• p = average density contrast across the interface
• 9 =average of the incident and transmitted angles
• d of each of these quantities are the changes in these properties
across the interface
• y = Vp/Vs
The advantage of this approximation is that it also can be represented
in terms of:
Rpp(9) =R(O) + G sin2(9) + F (siri (9)- tari (9))
It is often called the AVO equation where R(O) commonly refers to the
Intercept (Normal Incidence), Gto the Gradient or Slope and Fto the
Curvature (Far offsets).
This approximation also is a widely used result. Whereas the Aki-
Richards formula emphasized the changes in density and Pand S
velocity, Shuey's result emphasizes the angular dependence by dividing
the reflectivity neatly into several terms. These terms become
increasingly important as the incident angle increases.
The first term is a constant term that is the normal incidence reflectivity
when 9 = 0. As the angle increases, the second term starts to become
important. Its coefficient depends on both the contrast in Vp and Vs as
well as the ratio of Vp to Vs and the density.
The third term is small at small angles. As the angle increases, it
becomes significant and controls the shape of the angular re lationship
for large angles. Its coefficient involves only the Pwave velocity
contrast.

Quantitative Interpretation AVO modeling •101


Clearly if large angles (for example, greater than 30 degrees) are of
interest, the last term is important for modeling the correct behavior. If
large angles are not of concern (less than 30 degrees), then the last
term can be dropped leading to what is called Shuey's 2-term
approximation. This approximation often is written as:
Rpp(8) = P + G sin 2 8

In this equation, Pis referred to as the intercept term. From the full
3-term equation, it is equal to the normal incidence Pwave reflectivity.
The term G is called the gradient term. If Vp/Vs is taken as equal to 2,
then with a little algebra, it can be shown that the gradient can be
expressed in terms of the normal incidence Pwave and shear wave
reflectivity, as shown in this equation:
G = Rpp(O)- 2Rss(O)
This approximation is useful and helps you to visualize how the
amplitudes behave in terms of the Pand S wave reflectivities.
The terms Pand Gare examples of what are referred to as Primary AVO
attributes. They are quantities that can be determined directly from
fitting prestack seismic amplitudes to (in this case) the 2-term equation.
Quantities derived by combining primary attributes are called secondary
attributes.
Because the Shuey 2-term equation drops the third term, it should be
used with caution at angles above 30 degrees.
Although the 2-term approximation can be applied over a limited angle
range, it often is used when that limitation is met because of the wide
range of attributes that can be calculated from the primary attributes P
and G. These attributes are discussed later.

102 • AVO modeling Quantitative Interpretation


Verm and Hilterman
Another useful variation on the Shuey 2 term formula was proposed by
Verm and Hilterman (1995). They reformulated Shuey's ·equations with
the assumption that Vp/Vs=2 and that 8 is less than 30 degrees. This
variation has the form shown in this equation (from the Petrel Help):
Rpp(e) = Nl cos 2 e + PR sin 2 e
In this equation:
Nl =
a 2 p2 -a1 p1
a2 P 2 + a 1P 1
This equation is the standard normal incidence reflectivity where a and
pare the Pwave velocity and density.
U2- U1
PR=---
1- (J~ve
This equation is called the Poisson's ratio contrast. The sl terms in the
numerator of the equation are the Poisson's ratio in the upper and lower
media. The average in the denominator is the average of the values in
the two media.
Nl and PR are the primary AVO attributes that can be calculated from
this approximation. They also can be calculated from the Shuey 2-term
primary attributes, Pand G, as
NI=P

PR = w (P+G)

In these equations, Pand Gare the Shuey Intercept and Gradient


attributes.
NR*PR is a commonly used secondary attribute.
Verm and Hilterman noted in their paper that specific (Class 2)
reflections clustered in such a way that the NI*PR attribute would not
discriminate them well because Nl was close to zero. Rotating the
display, however, would transform the Class 2 behavior into a Class 3,
which would be identified better by the NI*PR attribute. Figure 3 shows
this transformation (after Verm and Hilterman 1995).

Quantitative Interpretation AVO modeling •103


45° rotation
Nl

Shales
R

Gas C)
sand . Wet
sand

Figure 3 Rotation of crossplotted attributes

The rotated attributes can be expressed as:


PR = PR(O) cos A- NI(O) sinA
NI = PR(O) sin A+ NI(O) cos A

In these equations, A is the rotation angle.


The Verm and Hilterman approximation has similar angle limitations as
Shuey's 2-term (up to around 30 degrees, depending on the properties
of the reflecting interface). The unrotated Nl attribute is similar to the
Shuey intercept attribute P. The PR attribute highlights the role of
Poisson's ratio in AVO behavior. In their paper, Verm and Hilterman
emphasized its use as a lithology indicator.

Gidlow approximation
In 1987, Smith and Gidlow derived an approximation also based on Aki
and Richards results (Smith and Gidlow, 1987). Their interest was in
creating weighted stacks to highlight Pand S wave reflectivity and Vp/
Vs contrast. By making assumptions about what the Vp/VS ratio was,
they were able to highlight the presence of gas by deriving a Fluid
Factor attribute.
This equation is their expression for Rpp in terms of P impedance
contrast, ~1/1, and shear impedance contrast, M/K (refer to the
104 • AVO modeling Quantitative Interpretation
Petrel Help). The approximation is shown in this equation:
tll tJK
2 cos 2 (8) Rpp(8 ) =I +2Ksin2 8 cos 2 e
In this equation:
• 8 is the average angle across the boundary
M
• [ = ZRpp (O) or twice the normal incidence Pwave reflection
coefficient
f1 K

• /{"is the Gidlow shear impedance contrast, which is related to the


S wave reflectivity by this equation:
/11( -8
K 2 Rss (O)
(1{, /Ys )
From this approximation, using Gardner's velocity density relationship
and Castagna's mudrock line equation (which is discussed shortly),
Gidlow derived what is known as the Gidlow fluid factor:

M = RP- 1.16 (~) R 5

This factor is designed to highlight events that arise from lithology or


fluid cases that depart from the Castagna mud rock line, which relates
brine filled Swave and Pwave velocities. If reflections are caused by
shales and brine filled sands that lie along Castagna's mudrock line,
then the fluid factor is constructed to be small. For events caused by
shale to gas filled sands, then the fluid factor is large.
Primary attributes for this approximation are the acoustic shear
impedance contrasts. The secondary attribute is the fluid factor.

Quantitative Interpretation AVO modeling •105


Fatti
In 1994, Fatti et al. rearranged the Aki and Richards result as shown in
this equation (formula quoted from Almutlaq and Margrave, 201 0):

Rpp(8) 1+
= - (1
2
----E- 4 P
tan 2 8) Ill
IP a 15 2
(1
!l/3- · -tan 2 8 - 2 · -
- sin2 8 -
a
(p) 2
sin 2 8 ) -!J.p
p

This formula can be simplified for smaller angles as


1 Ill j3 !J./5
Rpp(8) =- (1 + tan2 8) _1!__ 4 - sin2 8 -
2 !P a Is
In this equation:
• lp is the P impedance
• Is is the shear impedance
• a is the Pwave velocity
• ~ is the shear wave velocity
• p is the density
• d's are the contrasts in the values across the reflecting layer.
Similar to the Aki and Richards result, Fatti's approximation emphasizes
the contrasts in specific properties, but like Shuey's approximation, the
angular dependence is more explicit. Within Petrel, the Fatti
approximation is used primarily in inversion.

Pan and Gardner


Pan and Gardner (1987) reformulated Aki and Richard's approximation
to have the form shown in this equation (refer to the Petrel Help):

Rpp(e) .~c cos 2 e = ~ (6P + L\a) + (- 24 (~ L\a + L\{3) - ~ L\p) sin 2 e + ~2 (~ L\p + L\{3) sin 4 e
2 p a y 2 a p 2 p y 2 p {3
When angles are greater than 30 degrees, the observed variation of
amplitude with incident angle is close to a parabola. This approximation
was therefor created:
2 4
Rpp(8) .tt cos 2 8 = a + b sin 8 + c sin 8 (with a constant term
called a, a linear term called b, and a quadractic term called c).
In the first equation:
• e =average incident and transmission angle
• Rpp(8) is the Pwave reflectivity as a function of e
1116 • AVO modeling Quantitative Interpretation
• a, ~.and pare the average Vp, Vs and density above and below
the interface
• ~·s in these quantities are the differences between the values
above and below the interface
• y= VpjVs
This equation leads to three primary attributes, which are the
coefficients a, b, and c in the equation, and secondary attributes that
are derived from combinations of the primary attributes. These
attributes are
t!Jp
Densitv
• contrast: -=
p -2(b +c)

JiJVJ
The P wave o-r Uniaxial compressional m od-ulus contrast: - = 4a + 2(b +c) and M
M
= A+ 2J1
tlVp t::.a
The P ~·vave velocity contrast: - = - = 2(a + b +c)
vP a
If a value is assigned to VpjVs (y), additional secondary attributes can
be calculated:

VP ratio contrast: lly


~ y
= 2a + b + c
.
(1- Y
4
2
)

Poisson's ratio contrast: -


l1u
=
[2a + b + c (.1 - 4y2)]
( ? ) ·
(J y - - 1 (y 2 - 2)
fl/1 c
S \.-Vave modulus contrast - = - y2
tl 2

S wave velocity contrast


t::.p =
p b + c 1 +4
( y2)
S l.·vave impedance contrast vt::.v = -b - c 1-
(
4y2)

Quantitative Interpretation AVO modeling •107


Summary
As you can see, there are many different approximations to Zoeppritz
equations. They have different ranges of applicability in terms of the
angle range over which they work and the attributes that they highlight.
For reference, Table 1, Table 2, and Table 3 summarize where the
different approximations are implemented in Petrel and the attributes
that they yield.
Table 1 AVO Approximations in Petrel

AVO approximations used AVO approximations used


to create AVO models for attribute calculation
from data or models
Zoeppritz Shuey 2 term
Aki-Richards Gidlow
Shuey 2 and 3 Pan and Gardner
Verm and Hilterman
(available in wedge modeling
and reservoir modeling
attributes)
Fatti (available in wedge
modeling and reservoir
modeling attributes)

108 • AVO modeling Quantitative Interpretation


The attributes available for analysis depend on the approximation that
you use. Table 2 summarizes these attributes.
Table 2 Attributes Available for Analysis

Approximation Primary Secondary attributes


attributes
Shuey 2 Term Intercept Intercept*G radi ent
Gradient Gas Indicator
Fluid Factor
Reflection Coefficient
Difference
Sign(I)*Gradient
Pwave Reflection
Coefficient
Poisson's Ratio Contrast
NI*PRC (with rotation)
Gidlow Acoustic Fluid Factor
Impedance
contrast
Shear Impedance
contrast
Pan and Gardner Constant term Pwave Velocity contrast
Linear term Density contrast
Quadratic term VP Ns ratio contrast
PR contrast
S wave velocity contrast
Swave Impedance contrast
Shear modulus contrast
Uniaxial compressional
modulus contrast

Quantitative Interpretation AVO modeling • 109


Table 3 shows a summary of some of the secondary attributes, how
they are computed, and some notes on their uses.
Table 3 Summary of Secondary Attributes

Attribute Definition Comments/Uses


Intercept* P*G Useful for highlighting
Gradient Class 3 anomalies
Gas Indicator Px (lniGil Useful for highlighting gas
in Class 2 sands (small
positive R(O) that is
transitioned into negative
amplitudes with offset).
This attribute is designed
especially for the Gulf of
Mexico region. Worldwide
application requires local
adjustment (from the Petrel
Help).
Gidlow Fluid Highlights points that fall
Factor llF = RP -1.16 (~)Rs off the wet mud rock trend
in an intercept/gradient
Where y=Vp/Vs
crossplot. Rp and Rs are
calculated from the Gidlow
primary attributes (from the
Petrel Help)
Shuey Fluid (
2
1-y ) 0.29
AF= 1.6-0.232-y- P-YG
This attribute is similar to
Factor the Gidlow fluid factor, but
Where y=Vp/Vs is written in terms of the
Shuey intercept (P) and
gradient (G) (from the
Petrel Help)

110 • AVO modeling Quantitative Interpretation


Reflection R(O) + G Used as a gas-sand
coefficient Rp- Rs = 2 discriminator. The relation
difference is exact when VpNs = 2. It
is equal to zero for brine
sands capped by shale and
negative for gas sands (Rp
is fluid sensitive, but Rs is
not). Refer to Castagna and
Smith (1994).
Sign( intercept) sign(I)*G This attribute extracts the
*gradient Intercept polarity to be
multiplied with the
Gradient value.
Pwave Nl = P; PRC = 4/3(P+G) PR can be used as a
reflection PR =P~(O) cos A- N/ ( O)sin A lithology indicator and also
coefficient and Nl = PR(O)si n A NI(O)cosA ~ as a fluid content indicator
Poisson's ratio when displayed in a
contrast crossplot. Rotation of the
crossplot can help to
highlight Class 2 anomalies
(refer to Verm and
Hilterman).
Normal NI*PRC Similar to I*G, but more
Incidence* effective at highlighting
Poisson's ratio Class 2 anomalies when
contrast rotated.
There have been many studies that looked at the different AVO
attributes and whether one attribute is better that another. There seems
to be no universal attribute that stands out. The attribute that is best
for an area is best decided by local modeling and calibration .

Quantitative Interpretation AVO modeling •111


Lesson 3- AVO classes
As AVO analysis became more routine starting in the early 1980's,
specific AVO anomaly types were observed frequently and
classifications were proposed.
Rutherford and Williams (1989} first classified shale/gas sand ·interface
responses into three types: Classes 1, 2, and 3. Ross and Kinman (1995}
suggested that Class 2 responses should be divided further into two
separate classes that emphasized their differences at normal incidence.
They proposed a Class 2p (small positive normal incidence coefficient}
and Class 2 (small negative normal incidence coefficient}. Finally,
Castagna and Swan (1997) introduced an additional class of AVO
response: Class 4 characterized by a large negative normal incidence
coefficient and a positive gradient.
Table 4 summarizes the AVO classes and their characteristics.
Table 4 AVO Classes and Characteristics

AVO Zero offset RFC Gradient Absolute Possible


Class Amplitudes Lithology/Fluid*
1 Positive normal Negative Far< R(O) Tightly compacted
incidence (Positive R(O). reservoirs, a
reflection may have phase reservoir with a high
coefficient reversal dimming velocity/density
of reflection on
contrast, or both.
stack.)
Can occur at any
depth.
2P Small positive Negative Far> R(O) Mildly compacted
normal incidence (Positive R(O). sands
phase reversal
reflection and polarity flip,
coefficient dim spots or
weak negative
reflectors with
hydrocarbons on
stack.)

2 Zero to slightly Negative Far> R(O) Mildly compacted


negative (Low amplitude sands with gas or
at R(O).
reflection gas rich oil.
brightening of
coefficient reflection on
stack and
Polarity Flip.)

112 • AVO modeling Quantitative Interpretation


3 Negative normal Negative Far> R(O) Unconsolidated
incidence (Relatively high sands with gas pore
amplitude at
reflection R(O),brightening
fill or very gas rich
coefficient of reflection on oil.
stack (classic
bright spot).)

4 Large negative Positive Far< R(O) Shallow burial


normal incidence (Very high unconsolidated gas
amplitude at
reflection R(O). dimming of
sands or soft shales
coefficient reflection on overlain by hard
stack.) shales or carbonates
that have a higher
shear velocity.
• The causes of the different Class behaviors listed here are very general observations that have been seen in several different basins. Local

calibration of the AVO response using the sand and shale properties actually observed is a must for any definitive statements to be made.

Figure 4 shows graphically the amplitude relationships described in


Table 4.

Reflection coefficient R(9)

0.2

0.1

-0.1

-0.2

Figure 4 Reflection coefficient versus Angle of incidence for different AVO


classes

Quantitative Interpretation AVO modeling •113


Lesson 4- Fitting of AVO approximations to
seismic data
To analyze AVO quantitatively using any of the approximations
discussed, it is necessary to perform a fit between the seismic
amplitudes and the approximation being used.
Figure 5 shows this process schematically.
Offset

Rpp(O)

+
Offset
h Jterr:epl

Figure 5 Least squares fitting of amplitude versus offset data

If the seismic data that you are studying is clean and well behaved with
little organized or systematic noise contamination, then a conventional
least squares fit suffices. In Petrel, this fit is referred to as the
Standard method.
On real data, however, when organized or systematic noise (termed
coherent noise) is present (such as, multiples or other non-random
interference), a better fitting method is required that is not as sensitive
to what might be potentially large outlying amplitudes in the gather.
The Robust method uses a method proposed by Walden (1991) that
uses statistical tests to identify and avoid outlying high amplitudes.
Specifically, the data is divided into offset groups and median values of
the amplitude are calculated for each group. A fit is performed to these
median values and residuals are calculated. Outliers are weighted
down or eliminated and a fit is performed to the remaining values.
Because the median value is less affected by outliers than the average,
this method is less sensitive to outliers than a standard least squares
fit. The Robust method provides better results on real data when noises
are present.
114 • AVO modeling Quantitative Interpretation
Lesson 5- AVO Modeling study
In Module 1, you saw how the elastic parameters were affected by the
presence of hydrocarbons and how crossplots of different parameters
can be used to identify hydrocarbon zones in a well. You then saw how
it is possible to quantify the effects of hydrocarbon saturation on these
parameters by performing fluid substitutions and how different
saturations affected the rock properties.
The next step is to predict how seismic data responds to the presence
of different fluids and saturations. The objectives are to establish the
tie between the AVO response of the seismic and the well data at the
well location and to use modeled seismic responses to calibrate
potential prospects and appraisal well locations.
To accomplish these objectives, you perform AVO modeling to
characterize the AVO response of different saturations and different
fluids: oil, gas, or brine reservoirs in an area. This AVO modeling study
can be performed on both the actual conditions at the well and modeled
cases from fluid substitution.
When creating synthetic gathers, you can compare them to a prestack
seismic dataset you loaded in your project.
Two format of gathers can be imported to Petrel:
• 010 format allowing seismic data to be indexed for a quicker
data index. This format only is supported by the Prestack
Seismic Interpretation (PSI) plugin. This is a separate license
available on the Ocean store.
• SEG-Y format available to be directly imported in Petrel.

Quantitative Interpretation AVO modeling •115


Procedure -Import SEG-V Prestack data
This procedure describes how to import SEG-Y Prestack data directly
into Petrel.
1. Right-click on the Input pane and click Import file.
2. In the Import dialog box, select the Prestack SEG-Y dataset
(*.*)format.
3. In the Segy prestack loader, verify that the traces, the X and
Y locations have been mapped correctly, and the Angle or
Azimuth units have been defined properly. You can verify by
using the Browse ... button to look at your files and the
Examine button to check if the parameters (total number of
gathers ... etc) are coherent.
N Segy prestack loader ~
File headers

I EBCDIC header ...


II Binary header ...
I
Trace headers selection

Primary: 189 · 192 : trhlnlme ...


Secondary: 193 • 196 : trhXIine ...
- -- - - -- -
073 • 076 : trhSourceXCoordinate ...
X m

Y:
---
077 · 080 : trhSourceYCoordmate ... m

Angle/az1mulh unit em ...

II Header type mapping ...


II Browse ...
II Examine
~
Examine result

Total gather numbe r. 1110


--
Is full gndded True [30X37]

Primary header range. 3434 3492

2344
-- 24 16
---
Secondary header range.

10
- -- --
Maximum trace per gather.

Samples per trace: 1499


- - -
I Preview ... I
Import information

Name:
-
PrestackOataSet[SEGY]
-
Value range scan option: Random Scan ....
- -
·20.000.000
- --
15.513.578
-

1., OK
II~ Cancel II

116 • AVO modeling Quantitative Interpretation


4. A new Prestack dataset is located at the bottom of the Input
pane. Right-click on it and insert a line.
5. Open a Well section window and select the new line.
f> - Crossplot selec~ons
t> .,_ Filters
- 0 Z·leveJs
- VOis
~~ (fea~e seis ogram a :ribu:e
.. '!r Filte s folder
~ 1'!:;- Gur
4 bll ~~~~n· - - -

Quantitative Interpretation AVOmodeling •117


Procedure - Create an AVO synthetic model
This procedure describes how to create an AVO model.
1. On the Quantitative Interpretation tab, in the AVO group,
click AVO modeling.
f}} 1
' 11 ~ Volume Attributes
Wedge model
AVO AVO
modeling econnaissance II Qltools
AVO

Use the Avo modeling dialog box to enter parameters to set


up and run the modeling process.
~ AVO modeling
AVO modeling stucly

0 Cre8te ~====~============~~~
Avo modeling study 11 ~
• 0 Edit

Input [ 1 Algorithm j
Well: ~~......~:..;_·J_up,:..;_iter_-1_ _ _ _ _ _ _ _ _-J
Wavelet

Wavelet

Scele fllctor·

1 P-wevevelocity - - - - - - - - - - - - - - -

Log: ~.._I\fo...:....p_..:.Vp_ _ _ _ _ _ _ _ _ _ _J

S·wevevelocity - - - - - - - - -- - - - - -

Type: !Log
~==================~
Log. ~~IV:-=-•_
vs_ _ _ _ _ _ _ _ _ _--'
I Density

Type: !Log
Log: ~~~~=a=
RH=OO==================~

[J Auto-update l Canoel j

NOTE: A Reset parameters button is located next to the


Create study textbox. This is very useful when the mode is
changed from Edit to Create and if the remaining parameters
from the previous study are not going to be used.

118 • AVO modeling Quantitative Interpretation


2. Select Create or Edit, depending on whether you want to
create a new study or edit an existing study. If you select to
create a new study, enter a suitable name for it.
3. Open the Input tab. This tab is where the input data for the
study is specified.
You must enter these parameters:
• Well on which the study is being done
• Wavelet to be convolved with the seismic
• P-wave velocity (or transit time)
• Shear wave velocity (or transit time)
• Density
4. Insert ~the wavelet from the Input pane or select the
wavelet from the list.
NOTE: Initially, an analytical wavelet (for example, a Ricker
wavelet) can be used. If more rigor is required, a deterministic
wavelet derived from angle stack data might provide better
results. Deterministic wavelet extraction is discussed in
Module 4.
5. Similarly, enter the Pwave velocity log to be used.
6. Select one of these types for the shear wave velocity.
• Log: Select this type if a shear log is available and you want
to use it as input.
5-wave veloeity - - - - - - - - - - - - - - - - - - - -
Type:

Log:

• Constant value: Select this type if a single value is


appropriate and enter the value that you want to use.
S-wave velooit11- - - - - - - - - - - - - - - - - -

Type: (Con~tant value


Corostant value-:

• VpNs ratio: Select this type to use a constant VpNs ratio for
the modeling and enter the value for the VpNs ratio.
S-v,-ave velocity - - - - - - - - - - - - - - - - - ---l

Type: (VpN uatio

VpNs ratio: 1.5 :

Quantitative Interpretation AVO modeling •119


7. Select the density type. You can use log data or a constant
value.

Type Log

l09

8. (Optional) If you want the study to update when any of the


parameters are changed, select the Auto-update check box.
0 u~o-updare [ v' Apply I[
v OK I["' Cancet Jl

TIP: An AVO Modeling study can be updated on the fly using


the Inspector tool in the same manner as a rock physics study.
You must select the Auto-update check box; otherwise, the
Inspector tool provides you with only the parameter
information.
9. Open the Algorithm tab.
On this tab, you specify the AVO approximation to be used in
synthetic generation, the angle range and increments to be
modeled, and the prestack seismic dataset to which you want
to compare the results.
Algorithm
~------------------~

Synthetic - - - - - - - - - - - - - - - - - - - -

Seismic nome: Avo modeling study 1 synthetic (!1]


Output file: C:\Users\CFraser\Desktop\new gothers.segy - 0
Algorithm: [~
P-P-AV-0-.Z-oo-p-~
--~-------~
~ l

Sample interval: ~ o----~- 4 •.... ms

~ Use anti-alias filter

Anglerange - - - - - - - - - - - - - - - - - - -

Startangle : o ! degrees
End angle.
~--·-----~- - ------ 35 : degrees

Angle step
~----- ----- ----- 5 : degrees

Seismic display

Observed gather. ~

10. To run the AVO Modeling, you get the location of the modeled
pre-stack data by clicking D.
120 • AVOmodeling Quantitative Interpretation
11. In the Synthetic section, complete these steps:
a. Edit the Seismic name if desired. This value is the name of
the synthetic that is generated.
b. Select the Algorithm to be used from the values shown in
this figure:

s~~e~ ---------------------------------------
Seismic name: Avo modefing study l synthetic rt.:il
n:r•
Output file: -C-:\,Uset"S
-- ',C_Deslr _ _~
_ ebecql -esldo
-- pi~
-'Y'-
0-model
-- segy
-- o

A!gonthm: PP AVO. Zoeppitz ,


&l le intemsJ: b·~·~
· ~·~·~·~·~ ................ ms
PP AVO. Aki &Richards
PP AVO. Shuey 3 tenns
PPAVO. S 2terms

NOTE: Recall from the introduction that Zoeppritz provides


an exact solution; Aki-Richards and Shuey 3 term are good
approximations for larger angles. The Shuey 2 term is
increasingly less reliable as angles exceed 30 degrees. It is
important to think about the angle range to be modeled
and select an appropriate algorithm to use.
c. Use the slider to choose the sample interval (time) for the
synthetic and select the Use anti-alias check box to apply
an anti-alias filter before any well log resampling is done.

NOTE: Anti-alias filtering removes high frequency data from


the logs that are aliased in the resampling process. If the logs
are blocked, using this option is not necessary.
NOTE: To the right of the Seismic name field, there is an
button that currently is grayed out. After running the study, this
button becomes active, and you can use it to calculate AVO
attributes from the well AVO synthetic.

Quantitative Interpretation AVO modeling • 121


12. In the Angle range section, select the starting angle, the
ending angle, and the increment for calculation of the synthetic
traces.
~g~Bnge -----------------------------------­

Start angle: 0 l£l degrees


~

End angle:
Angle step:

NOTE: Again, be careful when choosing the ending angle. The


Zoeppritz, Aki-Richards, and Shuey 3 term algorithms give
better results for large angles than the Shuey 2 term algorithm.
13. In the Seismic display section, enter the prestack data to be
displayed in the Well section window (if available). The
gather nearest to the well head is selected for display. Click
to enter the dataset name.
Seismic display - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Observed gi!ther: ~'--------------

NOTE: To the right of the Seismic name field, there is an


button that currently is grayed out. After running the study, this
button becomes active, and you can use it to calculate AVO
attributes from the well AVO synthetic.

122 • AVO modeling Quantitative Interpretation


14. After you enter all parameters, click Apply or OK.
When the study is complete, the results are displayed
automatically in a Well section window using a default
template. This figure shows an example of the result.

"\fnelet

.[l
Q)
.:1
:~
Ci.
E
<(
1900
~

2!:<00 Time ms)

nr spectn
2100

l~J IJO

'quency (H

ase spectnJ

jj IJO

~quency (H

15. You can customize the Well section window. Click the Well
section window Settings button in the tool bar.
I ~ ~ I SS _ .. ~ ~Jell sectio;!empl .. ~

Quantitative Interpretation AVO modeling •123


The Settings dialog box opens.
[!; Settings for 'AVO_SynthGe n 4 '

B-:!. Vertical tracks (6)


41 Style
~ ;.. . 0 [JIndex track
8- 0 8 Vp Templates: \f., P-velocity • Filter template
! '-0 vPrll
B- 0 .8Vs 0 Show data with chos en template's measurement
1 -0 \f. Vs
8- 0 .l:\RHOB
' --0 \>8 RHOB
E?-· ~A Av o modeling stu dy 2 syn~ I Name W elllog

,.,
I~
· '--~ Avo modeling s1udy 2 S) '
1 B- 0 A Wavelet Vp_edited YP0Vp_edited
'
!--+4
'--01J.,.. Wavelet
Borehole markers ~
' ~ Background
~- L Deviated tracks (0) I[Q]
You can select the tracks to display in the Template objects
area of the dialog box and insert additional tracks. Select the
data to be displayed within a track from available data types
on the Definition tab in the Objects settings area.

Procedure - Create attributes from an AVO synthetic and


from the observed gather
After running the study, you can calculate attributes from the synthetic
and the observed (actual recorded) seismic data. This procedure shows
you how to calculate attributes only for the synthetic seismic data. The
process for calculating attributes is identical for both synthetic and
observed seismic data.
1. On the Quantitative Interpretation tab, in the AVO group,
click AVO modeling.
~ Volu me Attribut es
;f. ~~ AVO Seismic logs
AVO AVO
modeling rec onn aissance Wedge mod el
1

AVO

The AVO modeling dialog box opens.

124 • AVO modeling Quantitative Interpretation


2. Select Edit and select the study where you want to create the
attributes.

0 Create: : Avo modeling study 1


[ 'ii Avo I'TlOdeling study 1

3. On the Algorithm tab, click the button to the right of the


Seismic name field (or the Observed gather for the actual
seismic data).

Synthetic - - - - - - - - - - - - - - - - - - - -

Seismic name:

Output file:

Algorithm:

Sample interval:

[i] Use anti-alias Iii er

Well: ~( ~ Jupiter-1
Prest8clc dataset ~ ~ Avo-;odeling study 1 synthetic
Output folder name: Shuey 2-terrn_Avo modeling study 1 synthetic

AVO methods - - - - - - - - - - - - - - - - - - - - -

Algorithm: [ Shuey 2-terrn

Line fitting mode: [ Standard

4. On the Inputs tab, in the Inputs section, verify that the well
(where the result will be stored) and the Prestack dataset (in
this case the synthetic) are correct.
If the attributes are being calculated for the observed seismic,
then the real seismic dataset should be entered as the Prestack
dataset, and the gather closest to the well is selected
Quantitative Interpretation AVO modeling •125
automatically for attribute calculation. Use the blue arrows to
enter a different selection if desired.
The Output folder name is filled in automatically based on the
Algorithm selection (refer to Step 5). You can edit it if another
name is desired.
5. In the AVO methods section, select the algorithm to use to
calculate the attributes and the method to use to fit the data.
a. In the Algorithm field, select the method to use from the
list. This figure shows the available methods.

WO~s ------------------------------------
Algorithm Shuey 2-term

Line fitting mode: ~G~


idkl1~wm••••••••••··~~
Pan and Gardner

NOTE: The method you select depends on the AVO attributes


that you want to calculate for analysis. Different attributes are
available, depending on what algorithm you select.
b. In the Line fitting mode field, select the fitting method from
the list. There are two fitting methods available: Standard
and Robust.
~0~~$ ------------------------------------

line fitting mode.

On noise free synthetic data, performing a least squares fit


to the amplitudes sample by sample should be adequate to
calculate the primary attributes. Use the Standard method
for this data.
For data that has systematic noise issues (for example,
multiples or ground roll), a more robust method of fitting
that avoids contamination by outliers is needed. Use the
Robust method for this data.
c. If you select the Robust line fitting method, enter these
parameters:
• Scale factor: Used to determine the level of outlier
elimination.
• Overlap: Determines the overlap in the number of
126 • AVO modeling Quantitative Interpretation
traces between the segments in the first step of the
robust-fitting process.
If you enter 0, every trace falls into only one segment. For
any other value, the traces are shared by adjacent
segments. The segments are the groups whose data points
are filtered by a median filter prior to fitting.
After a fit is made, values are removed based on the scale
value to improve the fit.

~~mOOe -----------------------------11 1 11

t~ 1~----------~----------_J
~ ·I
Scale factor: 210fgj

Overlap: 0 ....
~

6. Open the Output tab and select the primary and secondary
attributes to be ca lculated.
NOTE: The attributes displayed on this tab change, depending
on the Algorithm that you select.
This figure shows the attributes for the Shuey 2 term algorithm.

Pmre~~~ -------------------------------------­
lntercept Intercept
Gr8dient
Second8~ 8tiJ:ibutes

0 Chedt811

0 lnlefcept • gr:xftent ltnterrepl'" grnd~er~l

0 Gas indicator: IGas indicator


0 Fluid flx:tor: IFluid faaor
Vp!Vs ratio:

0 I
Reftection coefficient difference: RerJEdicn coefficient difference

0 S[gn("111lerrept) • gradient ISignflllterrept} - gradient


[ ] P-wave reflection roefficient I~================~
P-wave refiedicn ooeffoe t

Poisson's ratio contf8st IPcisson's ratio contrast


Nl * PRC: ~================~
t•PRC

R.otabon angle: 0 ::

Quantitative Interpretation AVO modeling •127


This figure shows the attributes for the Gidlow algorithm.

Primcry attributes - - - - - - - - - - - - - - - - - - - - - -
Acoustic imped1snc:e contrast Acoustic impedance contrest

Shear impedance contrast


-Shear
-- ----------------
impedance contrast

Secondary attributes - - - - - - - - - - - - - - - - - - - - - - -
0 Checkall

[iJ Auid factor: j Fluid factor


VpNsmtio· 2 .0 :

This figure shows the attributes for the Pan and Gardner
algorithm.

Primcryattributes - - - - - - - - - - - - - - - - - - - -
Constont term: Const8nt term
Linear term. Lineerterm

QU8dra!JC term: Quadratic term

Second!lry ettributes
0 Checkall

D P-wllve velocity ocntnsst. ~ 0-m~ve velocity contrast


D Density contrast: I~==========~======~
Density contrast

[ ] VpNs mtio contr!lst I~==================~


VpNs ret1o contrast

Vo/Vs ratio.
====================~
20 :

D Potsson's mtio contrast I._P_oiss


_ on_·s__
re__
tio_co
_ n__
uas t ------------=-'
____

VpJVs ratio: 2.0 :


[ ] S--wave velocity contrast: IS·wave velocity contrast
VpJVsretio: 2.0 :

D S--wave IITipedence contrast j._s_-w_av_e_.m_~


_on_ce_c_on_tros
_t- - - - - - -----=-=-~'
VpNsralio: 20 :
LJ Shear modulus contrast IShear modulus contrast
VpNs ra!Jo.
====~=================
2.0 ;
0 Umaxiel compressional 1UniaXlel comp~essional
modulus contrast '-·- - - - - - - - - - - - - - '

Some of the secondary attributes require an assumption


regarding Vp/Vs to be calculated. Because Vp/Vs varies with
geology, lithology, and degree of compaction (among other
things), selecting this parameter requires a knowledge of the
local Vp/Vs behavior for the geologic section being
investigated. There is an area to enter this value where it is
needed.
128 • AVO modeling Quantitative Interpretation
7. Click Apply or OK
The output is stored in the Input pane under the Well entry
and in the Global well logs folder.
This figure shows the output in the Global well logs folder.
~ 0 Shuey 2-term_Avo modeli'lg study 1 syrthetic
tJ 0 Intercept
tl 0 ~diert
tl 0 Intercept • gradient
tJ 0 Gas indicator
tJ 0 Ruid factor
tl 0 Reflection coefficiert dfference
tl 0 Sign(interc~) ·~
tJ 0 P-'~'+-ave reflection coefficient
tJ 0 Poisson's ratio corbast
tl 0 NI.PRC

This figure shows the output for the well .


.. ~ 0 ~---1
"' ( Wl!ll'/og$
Petrophysical Properties
Rock Ph.,.sics Input
Synthetic
"' Shuey 2-term_Avo modeling study 1 synthetic
tJ Intercept
tJ Gradiert
tJ Intercept -gradient
tJ Gas indicator
tJ Auid factor
tJ Reflection coefficiert difference
tl Sign(lrtercept) . gradient
tJ P•Nave reflection coefficient
tJ Poisson s ratio contrast
tJ NI.PRC

Quantitative Interpretation AVO modeling •129


You can display these attributes in the Well section window.
An example of an AVO study after customizing the window and
including attributes is shown in this figure.

Bulk AVO gather


GR Nphi/Rho P sonic S sonic synthetic
density gather

130 • AVO modeling Quantitative Interpretation


Exercise 1 - Set up and run an AVO study
This exercise gives you a chance to work through the procedures
discussed in this lesson to create an AVO study and extract attributes.
1.On the Quantitative Interpretation tab, in the AVO section
group, click AVO modeling.
2. Create a new study.
3. On the Input tab, enter the well logs to be used
4. On the Algorithm tab, specify the algorithm to be used. To
start, select Zoeppritz as the algorithm. The sample rate and
angle range defaults work well for this exercise.
5. Enter the prestack seismic data in the seismic display section.
6. To run the study, click Apply.

10~

1quency (H
ase spHtru

.tJ
0

7. When the study is complete, click the Create seismogram


attributes from synthetic gather or Create seismogram
attributes from observed gather button on the Algorithm
tab to create attributes for the synthetic and the seismic data
(if prestack seismic data is available to use).

Quantitative Interpretation AVO modeling • 131


The Seismogram attributes creation dialog box opens.
8. On the Input tab, complete these steps:
a. Ensure that the Well and Prestack dataset are correct/
b. Select an AVO Algorithm. It is recommended that you start
with Shuey 2 term.
c. Leave Line fitting mode set to Standard for the noise free
synthetic.
9. On the Output tab, select the attributes that you want to
calculate. Include lntercept*Gradient in your selection.
10. Click Apply.
11. Customize the Well section window to include the poststack
data, zero offset synthetic and Intercept, Gradient, and I*G
attributes.

Bulk AVO gather


GR Nphi/Rho sonic 5 sonic synthetic
density gather

132 • AVO modeling Quantitative Interpretation


Questions
There is a pay sand in the well between 2200 and 2220 ms TWT.
• What kind of AVO response do you see at the top of the pay?
• What about the fluid level?
• How big of an angle do you need to observe the AVO response
at the top pay properly?
• What do the intercept and gradient traces do at each of these
levels?
• You can change the algorithm used to create the study. What
differences do you see if you change the AVO modeling
algorithm from Zoeppritz to Shuey 3-term and Shuey 2-term?
• What attributes from the suite calculated would seem to
discriminate the top pay event uniquely?

Quantitative Interpretation AVO modeling •133


Procedure - Create crossplots of AVO attributes and
amplitude versus angle
You can cross plot AVO attributes as described in Module 1, Lesson 4;
"Procedure- Crossplot elastic property estimates." That procedure is
described briefly here.
1. Create a new crossplot window by clicking the button in the
Crossplot group on the Quantitative Interpretation ribbon.
2. Select the collection settings.
3. Select the well from the Input pane whose attributes you
want to crossplot.
4. Locate the AVO study in the Global wei/logs folder in the
Input pane.
5. Select the attributes to crossplot.
6. Assign the attributes to the appropriate axes of the crossplot.
7. Adjust the scales of the crossplot axes if necessary.
An example of the Intercept plotted against Gradient for the
synthetic you just created is shown in the figure. The pay zone
is identified nicely in the Intercept* Gradient trace.
This figure shows an AVO crossplot with the selection shown
in the Well section window (right most track). Some
additional AVO attributes are displayed for comparison.
Intercept Vs Gradient
Crossplot

. . -';~.:.:- ..

t "" .- :.:.,:
..

134 • AVO modeling Quantitative Interpretation


;$~ lcollection 1 T

[@ Collection settings
New QI crossplot New
window collec ion ~ Classification data
'-------J
QI crossplot

2. Tile the new window so that it is visible with the Well


section window.
3. In the Ql crossplot group, click Collection settings for the
window.
Jf,~ ':_ollect~n ~ ~
[® Collection settings
New Ql crossplot New
window collection lSi Classiiication data
Ql crossplot

The Settings dialog box opens.


4. In the Settings dialog box, open the Settings tab and the
Dimensions subtab.
5. Locate the AVO modeling study in the Input pane.
6. Expand the study and select the synthetic gather collection (to
plot data from the model).
~ ~ ~ Avo modeling study 1 PAF Test
~ ~ltl ~- Avo modeling study 1 synthetic PAF Test
t> T.race Hear/ets
• e• ~ Avo modeling study 1 synthetic PAF Test collection

The selections appear in the Available objects area in the


Settings dialog box.

Quantitative Interpretation AVO modeling •135


7. On the Dimensions tab, complete these steps:
a. In the Available objects area, select the gather entry.
b. In the Axes area, select Amplitude for theY axis and Sin 28
for the X axis.

" Q,thef cotlectJons

ng study 1
I
~ -------------------------

" bl Avo modeling study 1 syntheti ~ Isin'(G)


M •: Avo mod y· Am-pllt
.-
1
- - ude: - - - - - - - - - - - - - ,

~ ~Z:============::::l
sitJ1.0)

PO!Pienn~ut~ ---------~

0 Color: ..___ _ _ _ _ _ __

"- TODI Pa~t?tte 0 X 1


~ Cr-eate/edit z-Jevel
,. QI tool!t X

10. Click a depth in the Well section window.


The Z level appears as a line in the window. An entry for the Z
level also is added to the tree.
Z-levels
&t Z-level
i}) Z-level2
.Q£ Z4evel3

136 • AVO modeling Quantitative Interpretation


With the Create/Edit Z level button active, the Z level can be
moved up and down by clicking it and dragging it up or down.
11. In the crossplot Settings dialog box, open the Filter tab.
12. Insert ¢ the Z level from the tree. Click Apply.

4 Galfler coBections
~ •: Avo modefing study

Quantitative Interpretation AVO modeling •137


13. Make sure the Well section window and the crossplot
window are still tiled. To plot the amplitudes interactively, drag
the Z level up and down.
NOTE: It might be necessary to adjust the horizontal and
vertical scale to set the ranges sufficiently to accommodate the
seismic amplitudes and the range of Sin 28. You set the ranges
in the Settings dialog box for the crossplot window entry
located in the Windows pane.
1£1 Settings for 'QI cros.s plot window 3' L1LJ
~ Settings I Collections !
Axis sccle

X @ Linear © l ogarithmic
w
Y: @ Linear 0 Logarithmic

Axis range [J
Min: Max:
X 0! 0.5 : E) l ock
- ---o.i ! -· -
Y: ...
0.1 4 [] lock

( ./ Apply II ..I OK II ~ CanceJ ,_ j

138 • AVO modeling Quantitative Interpretation


This figure shows the Well section window and the
crossplot displaying the amplitude at 2200 ms plotted against
Sin 28 at the top of the pay event.

This figure demonstrates the Class 3 behavior of the event at


the top of the pay section in the well. The graph is linear as it
should be when plotted against Sin 28.

Exercise 2- Crossplot attributes


Use "Procedure- Create crossplots of AVO attributes and amplitude
versus angle" and "Procedure- Create an amplitude versus offset
crossplot" to perform these tasks.
1. Crossplot two AVO attributes of your choice (for example,
Intercept versus Gradient. Select a region in the crossplot and
create a classification to plot the selection in the Well
section window.
2. Create an amplitude versus angle crossplot.

Quantitative Interpretation AVO modeling • 139


Exercise 3- Create AVO synthetics for different fluid
saturations
In this exercise, you use the fluid substitutions that you performed in
Module 1 to create AVO synthetics and observe the differences
between the AVO response to the different fluid types and saturations.
1. Create three AVO new modeling studies, one each for the brine
filled, 100% gas filled, and oil filled cases.
2. Edit the Well section window to display the three AVO
synthetics.
What are the differences?
NOTE: If you want to display the amplitude versus offset from
several AVO synthetics (for example, for different fluid content)
in the same 01 crossplot, create collections and set up the
different collections from the settings of the 01 crossplot.
mJ Settings for 'QI crossplot window 1' ULJ
0 I Settings I Collect ions
Info [

[;!]~~ [i
~ Neme Settings Active

~
[1] 0
15% Ges

Ill 50% Gas ®


ttl Base case @

I " Apply II ./ OK II ~< Cancel I

140 • AVO modeling Quantitative Interpretation


This figure shows an example of AVO synthetics for different
fluid content and saturations.

1 Seismic at well location


2 In situ synthetic
3 100% gas
4 40° high GOR oil
90% sat
5 100% brine

Quantitative Interpretation AVO modeling •141


The AVO behavior can be quantified by creating a crossplot of
the amplitudes versus angle for the different fluid
substitutions.
This figure shows AVO synthetics for different fluids and
saturations displayed next to a plot of amplitude versus angle.
The Brine case is shown in blue, oil in green. The in-situ well
conditions are in orange with the higher saturation gas case in
red.

tl)

~
a.. ••
• • • • •
E
<{
tl)
"0
.@
..
a.
E
<{ .. • • • •
tl)
"0
• •
a. ...
.@ • •
• • •
E
• •
<{
• •
tl)
"0
...

~
a.
E
<{
..

010 ! 0 18

1 Seismic at well location


2 In situ synthetic
3 100% gas
4 40° high GOR oil
90% sat
5 100% brine

142 • AVO modeling Quantitative Interpretation


Observations on the effect of fluid saturations on the
seismic response
The Intercept versus Gradient crossplot shows different AVO responses
for the different fluid types and saturations.
• As gas saturation increases from the in-situ value to 90%, there is
a subtle increase in the near offset amplitude, but it is small. As
discussed in Module 1, most of the effects of gas saturation occur
at low saturations and there is little change after the initial effect.
• The oil case shows a class 2 or class 2P responses. The near trace
is zero to weakly positive with a smaller negative gradient than the
gas cases. There should be a fluid level evident in the oil case. The
oil response is sensitive to the oil gravity and GOR. In the example
shown in "Exercise 3- Create AVO synthetics for different fluid
saturations", the oil gravity was set to 40° and the GOR to 1000.
• By contrast, the brine filled has a hard response, opposite in
polarity to the gas filled cases. The gradient appears slightly
positive. Thus you might expect to see a polarity flip as you go from
gas filled to brine filled across the fluid level.
The wet response predicted ties nicely with what you observe on the
seismic. The soft class 2 response should flip to a hard reflection below
the fluid level, which is exactly what you observe as shown in Figure 6.

Quantitative Interpretation AVO modeling •143


Figure 6 AVO synthetics compared to the stacked data.

144 • AVO modeling Quantitative Interpretation


You also can calculate the AVO attributes for the fluid substitution
cases. Figure 7 shows an example using the Intercept* Gradient
attribute.

Figure 7 AVO synthetics compared to the Intercept* Gradient attribute

1 In situ synthetic
2 In situ I*G
3 90% gas
4 90% gas I*G
5 90% oil
6 90% oiii*G
1 100% brine I*G
8 100% brine I*G
Quantitative Interpretation AVO modeling •145
The models show these responses:
• A strong positive I*G anomaly associated with the top of the pay in
both the in-situ and high (90%) saturation gas cases. In the high
gas saturation case, there also is a positive anomaly at the fluid
level.
• I*G anomalies also are present for the top and base of the hard
overlying unit, but are of opposite sign to the pay.
• There is an I*G response to the brine case, but it is smaller in
amplitude.
• I*G has almost no response in the oil case. However, there still is a
significant amplitude and waveform change going from oil to brine.

146 • AVO modeling Quantitative Interpretation


Lesson 6- Wedge modeling
When interpreting seismic, we assume that strong amplitudes define
layer boundaries and represents change in acoustic impedance, that is,
change in pore fluids or in rock types. This statement is true if the layer
is relatively thick; however, a thin layer can cause tuning, an
interference between the seismic pulse representing the top of the
layer and the one representing the base of the layer. Tuning generally
happens if the layer thickness is less than a quarter of the wavelength.
At tuning thickness, the amplitude reaches a maximum (Figure 8). For
reservoir thinner than the amplitude thickness, the amplitude will
actually decrease (limit of seismic visibility).

50 ID.§._

G:l
-o
::3
..,..,
.~ G.J
0.. c
~
E v
<( ::c
....
....,
G.J
+J
·u;
0
c. ...ccv
nJ
E c.
0 c.
u <(

Figure 8 Amplitude vs thickness and concept of tuning

Quantitative Interpretation AVO modeling •147


The idea is to analyze the maximum amplitude above which it is not
representative of lithologic changes in the reservoir but of tuning effect.
The Wedge modeling process in the AVO group of the Quantitative
Interpretation tab has this functionality. There are two parts to the
wedge modeling workflow: creating the model and calculating the
seismic response to that model.
Wedge modeling uses well tops to define a depth interval in a well that
is thickened and thinned to create a wedge. The size of the model is
specified and determines how quickly the wedge changes shape. You
decide where to locate the well in the model. Blocked logs (such as the
ones that you created in Module 1, Lesson 7: "Exercise 1: Perform
Backus log blocking") are most easily used to populate the properties of
the wedge. The blocks in the logs are increased (or decreased)
proportionately as the wedge thickens or thins.
A variety of methods are available to calculate the seismic response.
Seismic output can be a seismic cube for a particular angle or prestack
gathers for a range of angles. Reflections are modeled according to the
algorithm selected. If you select Zoeppritz, you can model pre- or
post-critical reflections, or both. When the model is created, the seismic
response is generated in a second step.
Figure 9 shows an example of the model wedge and the associated
seismic response.

148 • AVO modeling Quantitative Interpretation


Figure 9 Examples of wedge model geometry and the model

Quantitative Interpretation AVO modeling •149


Procedure - Create prestack and poststack wedge
models
This procedure shows you how to create wedge models.
1. On the Quantitative Interpretation tab, in the AVO group,
click Wedge modeling.
.., ~ Volume Attributes
AV~ ~ AVO Seismic logs
reconnaissance I Wedge model I
AVO

The Make wedge model dialog box opens.


2. Fill in the Define wedge model tab as shown in the figure.

Grid name:

1/IJell

Select a well : ~ :~ Jupiter-1

Markers

Select a •Nell tops folder:


~I Well tops

Select a wedge zone : Moo< Pay

Grid size and position


----
Number of grid cells: I: J: 6

Grid node at well location : I: 75 J: 3

Grid increment: Xinc : 25 m Y inc : 25 m

NOTE: For Petrel to able to create a proper wedge model, two


zones (above and below) and the area of interest (for example,
reservoir) are required. Therefor, the Well tops folder should
contain at least four well tops.

150 • AVO modeling Quantitative Interpretation


3. In the Grid size and position section, enter this information:
• values for the number of grid cells in I and J
• grid node at which to locate the well
• X andY increments that define the model grid
NOTE: Some experimentation might be necessary to get the
wedge geometry and the well placement the way you want it.
This figure shows a top view of a wedge model and defines the
I and J parameters. These are the recommendations:
• Number of grid cells: Should care more about I than J (I>
J)- you want a large I coverage and a small J coverage (J
= 1/1 for example).
• Grid node at the well location: It determines where you
want to place your well in the model. It is, however,
recommended to place the well in the middle of the wedge
model; therefore, around half the number of grid cells.
• Grid increment: Should be the same as the seismic.

The grid is defined with I = 1 located at the thin end of the


wedge. I increases as the wedge thickens. In a model with 1DO
grid cells in I, placing the well at grid number 75 places it
three-quarters of the way down the wedge from the thin end.
J is the model index orthogonal to I. The model is simply
repeated J times in this dimension.
The X andY increments define the cell dimensions, so they
define the overall length and width of the model.
Quantitative Interpretation AVO modeling •151
4. Open the Elastic properties tab and complete these steps:
a. In the Layering section, enter the log to define the model
layers in the Define layering from block log field. It is
recommended that this log be a blocked log. Also, enter a
value for the Minimum layer thickness.
b. In the Logs section, select the logs that will supply the
elastic properties from the lists.

Minimum layer thickness : ms

~gs -------------------------------------

Vp : vp P-itoave velocity_Backus Average Study1 [Rock p ..,.

Vs : \fa 5-wave veloci~·-B~kus Average Study1 [Rock p ..,.

Rho : P Rho_Backus Average Stu<ry 1 [Rock physics PAF ..,.

NOTE: The wedge is created with layers. The thickness of


those layers is controlled by the log that you enter in the Define
layering from block log field, subject to the Minimum layer
thickness value that you enter. To avoid excessive numbers of
layers and consequently long run times, it is recommended to
use blocked logs for the properties.
When different combinations of logs are used, these rules
define which logs are used and how:
• When blocked logs are used to define the layering, the
median value over each block of any unblocked log is used
to build the model.
• When unblocked logs are used to define the layering and
the other logs are blocked, cells are populated with the
appropriate value from the blocked logs.
It is recommended to create a blocked log that is blocked at the
level of the events you wish to look at. For example: For one
reservoir divided into two clear zones embedded between two
shaly sequences, your blocky logs should have four blocks
showing these changes.
152 • AVO modeling Quantitative Interpretation
5. Open the Synthetic seismic modeling tab.

AJgorithm:
Wavelet

Select a wavelet : ~ ~'--W_\. W


_a_vel_et_ _ _ _ _ _ _ _ _ _ ___,

Scale:

Angle range

Angles defined as: !§) Average angles ,:_y Incident angles

Start angle: degrees

End angle : 25 degrees

Bfedive angle: 11!i.321 I degrees


hlgle step :
========-' degrees

~ ~--------------------

~~ Preaitical reflections 0 Post~cal reflections J Both

Sample interval : 4 ms

Seismic t'jpe : (-: Post~ack Cube I~ Pre-Stack Gathers

Seismic name : 'NedgeSynthetic [E§ Create new synthetic seismic


[ •· P.ecalculate S'in!hetic se1~mic 1

D .Automatical~/ calculate new s'lnthetic seismic on Apply/ OK

6. From the Algorithm list, select the algorithm to use to


calculate the we e model seismic res

1
Navelet
PP AVO. Zoeppritz
Select a wavelet: r::J;:l PP AVO. Aki & Richards
~ PPA\'O. Fatti
Scale : PP AVO . Verm & Hilterman
PP AVO. Shuey 2terms
PP AVO. Shuey 3terms

These choices are available.


• Post stack, normal incidence: Creates a model with no
offset behavior, calculated from the normal incidence P
wave reflectivity.
• PP AVO, Zoeppritz: Uses the full Zoeppritz equations to
calculate the reflectivity, which can include pre- and
post-critical reflections.

Quantitative Interpretation AVO modeling •153


The remaining options are different approximations for AVO
behavior.
7. In the Wavelet section, enter the wavelet to be convolved
with the reflectivity sequence.
You can apply a wavelet scale factor if desired.
Wavelet - - - -- - -- - - - - -- - -- - -

Select a wavelet: ~ ;..._1.1\.


_ W _a_vel_et_ _ _ _ _ _ _ _ _ ____,

Scale:

8. In the Output section, complete these steps:


a. Select whether to output Pre-critical reflections, Post-
critical reflections, or both. (You must select the Zoeppritz
algorithm to use these options because this algorithm is
the only one that can calculate these reflections.)
b. Enter a value for Sampling interval.
c. Select the output type: Post-stack cube or Pre-Stack
gathers.
If you select a poststack cube, you have the option to
specify the angle for which to calculate the reflectivity in
the Angle range section.
d. Provide a name for the seismic output in the Seismic name
field.
O~pu - - - - - - - - - - - - - - - - - -- --
·~ P~ical reflections 1..) Post-oitical reflections _:, Both

Sample interval : 4 ms

Seismic type : •_ Post~ack Cube ~ Pre-Stack Gathers

Seismic name : WedgeSynthetic [ ~Create new s;rihetic seisnic ]

1·~ Recalculc:te syrthetic seismic J

D A.Jtomatically calculate new synthetic seismic on Apply/O K

154 • AVO modeling Quantitative Interpretation


9. In the Angle range section, enter the Start angle and End
angle.
• If the output is a poststack cube, the Effective angle is
calculated automatically for the volume. This output
simulates an angle range stack for the range of angles
chosen.
• If the output is prestack gathers, you must specify the
Start, End, and Angle step fields .
.Allglerange - - - - - - - - - - - - - - - - - -

ngles defined as: '~' Average angles >'.') Incident angles

Start angle: 5 degrees

End angle: 25 degrees

Bfective angle: •1 ~- ~:!3 degrees

.Allgle step : degrees

If you select the Zoeppritz, Aki-Richards, or Shuey (2-term or


3-term) algorithms, you can define the angle as average or
incident. For the remainder of the algorithms (Fatti and Verm
and Hilterman), the angles are hard wired to be average
angles.
NOTE: Selecting Average angles means that the angle
specified is treated as the average of the angle of incidence
and the angle of refraction for incident and transmitted P
waves. Some AVO approximations use this definition for the
angles in their equations.
If the velocity contrast is small, the difference between the
incident and refracted angle also is small. Selecting Incident
angle implies that the values supplied are interpreted as the P
wave incident angle.
Effective angle reports angles that corresponds to the RC
averaged over the start-end angle range. For example, Start
angle= 5; End angle= 22, Effective angle= 14.202.
Effective angles are calculated from the formula:
• 2 1 . 2 • 2 . .
Sin 8eff = 3(Slll 8mtn +Sill 8max + SID8mtn Slll8max)

10. To create the model, click Apply.

Quantitative Interpretation AVO modeling •155


11. To calculate the post or prestack seismic cube, click the
Create new synthetic seismic button at the bottom.
hlgle range

Aslgles defined as : ~) Average angles Incident angles

Start angle: 5 degrees

End angle: 25 degrees

Effective angle: 11 .an I degrees


Aclgle step : degrees

The model appears as an entry in the Models pane.


~ ~ Wedge Models
~ Wedge model2
!> ~ Skelelon
~ Faui~
,. 0 Ho.rfzons
0 Top Reservoir
• 0 Base Reservoir
If Ed~
!> :J !ntMfK:Iiom
,. IJ 0 Properties
yp 0 P-velocily (Velocity)(P-wave velocit'/_Backus Average Study 1 [Rock physics] PAF]
P0 Density (Density}[Rho_Backus Average Study 1 [Rode physicsJ PAF]
V: 0 S-velocit>1 (Velocity}(SJNave velocit-1_8ackus Average Stud'/ 1 [Rock physics] PAF]
lp 0 P~pedance (A.coustic_lmpedance) [Calculated]
Is 0 S-impedance (P..coustic_lmpedance) [Calculated]
~ 0 P/ S velocity ratio (P.atio) [Calculated
0 0 Poissons ratio {Ratio) [Calculated)
C' ~ Fauf filer
t> ~ ZDne l+fer
~ Segmt!fl!filt!!"

The synthetic seism ic appears in the Input tree.


This figure shows the poststack cube entry.
~ 0 Wedge model 2a
~ f; 0 WedgeSynhelic
0 lnline 51
0 XUne4
_J 0 Z =-2416J)()

This figure shows the prestack data entry.


~ 0 Nedge model 2 P.A.F 1
A ~l~ 0 1NedgeS;'T1thetic 2 PAF
t> Q TRJCe Heade!s

156 • AVO modeling Quantitative Interpretation


You can visualize the model in the 3D window as shown in
this figure.

You also can display the gathers in the Well section window
and in a Prestack window if the Prestack Seismic Interpretation
(PSI) license is available.

Quantitative Interpretation AVO modeling • 157


Exercise 1 - Create a wedge model
In this exercise, you create a wedge model based on the results of the
Backus log blocking operation in module 1.
1. Use the results of the earlier Backus averaging exercise to
create a wedge model.
2. Choose to output a postack cube.
3. Display and then mouse over the seismic result. If necessary,
change the color scale of the seismic output for a better
visualization.

4. Observe the interference effect relative to the top and the


base.
5. To avoid seismic misinterpretation, determine:
• the thickness at which the bottom of the wedge begins to
interfere with the top of the wedge (limit of resolution)
• the tuning thickness, that is, the thickness at which
amplitude interference is a maximum and the value of this
amplitude maxima.

158 • AVO modeling Quantitative Interpretation


Review questions
• Who was Zoeppritz and what did he derive?
• What are some well-known AVO approximations?
• What are primary and secondary AVO attributes?
• Over what angle range is the Shuey 2 term approximation
valid?
• What is a wedge model and why might you want to make one?

Summary
In this module you learned about:
• different formulations that describe the relationships between
incidents and reflected seismic energy including:
• the exact Zoeppritz equations
• several approximations in current use and used in Petrel
AVO modeling
• classification of AVO responses into different commonly used
classes
• setting up and performing an AVO modeling study
• creating crossplots to analyze the results
• performing wedge modeling

Quantitative Interpretation AVO modeling •159


Module 3-
. AVO
reconnaissance

AVO reconnaissance identifies and classifies AVO anomalies in an


exploration context. Input consists of poststack and prestack data to
generate primary and secondary attributes.
Using a Volume of Interest (VOl), the generated attributes can be
co-visualized with the stacked seismic data. You also can use
cross plotting for a quick AVO analysis.

Prerequisites
To complete this module successfully, you must have completed the
previous modules in the course.

Learning objectives ~
~
After completing this module, you will know how to run the AVO
reconnaissance process including how to:
• create a Volume of Interest (VOl) for interactive display and
co-visualization of AVO attributes
• make interactive crossplots of the attributes displayed in the
VOl

Quantitative Interpretation AVO reconnaissance •161


Lesson 1 -AVO reconnaissance tool
In the previous modules, you took the results of the rock physics studies
and used the AVO modeling process to calculate the seismic response
to a variety of fluid types and saturations. You used crossplots to
analyze the amplitude versus offset behavior and classify the AVO
response into classes.
You also looked at a variety of different AVO attributes to understand
how they responded to different fluids and saturations, and how they
can be used to analyze seismic volumes. You visually inspected the
attributes and used crossplots to identify and isolate anomalous
responses. So far, you have concentrated on a single well location.
To assess the range of seismic responses to fluids and saturations
across a range of reservoir types, qualities, and geologic settings, it is
standard procedure to perform rock physics studies and AVO modeling
on all relevant well data as part of an exploration play or prospect
appraisal. Even if there is no relevant well data, you can use the
empirical relationships discussed to make a set of general models to
use. In addition, you can assess which AVO attributes were the most
useful in identifying possible hydrocarbon accumulations. In this way,
you have a portfolio of possible responses to reference as prospects are
screened.
The next task, then, is to take what you learn from these well specific
studies and incorporate it into your evaluations as you screen data for
leads and prospects. By screening data for structural opportunities as
well as for AVO responses, you are better able to rank your
opportunities.
The AVO reconnaissance workflow provides the tools and workflows to
perform this task. Figure 1 shows an overview of the AVO
reconnaissance workflow.

162 • AVO reconnaissance Quantitative Interpretation


~
Q)
(..)
c
.,
"'
.,
'co
§
~
0
~
Pre-etaek

AVO RaoonnBlesan<.:e
I Stwey(1) II GldloW(2) II Panand~~~3~~~J
P rlmal'l/' attnbut>ea.r: Primary a11rlbtl1ea· Primacy- at1 rrbutes:
- loter<:ept (I} -A.cou&11C l m~e nce ()Q ntr.a~l -Lil'!e,ar ~arm
..(;f'81dlen i {O~ -Shear Impedance con1ru1 ..QI!l;adrellc term

Se~ry aHrlbUI.e s· Secor:tdary alllibutes: Secondary oltrlbuiM


-r· G ·Fiuiid lector . 11-wavo 'I'Cioclly oon!rasl
-G e·s lndiC.Qtor -Oon&ll.')' oontmst
-Re lleetbn COO'fllclenl (RC) -VpN$ ratio conl•ns1
<lltf*!'ence ·-PRC
-Sign(I •.Q) . s .wavo wtloeity conwut
- P - W iiiV()I RC tNt}
-Pots$0n, Nl11o contrast tPJ"lC}
. s .wn...n i m pQ(~;mco cootmM
-She.a r moou11rs e<>ntr.u t
D fnpvt c1u ta

~PR C•Nt '-u niaxial oompre1"10no1


1!11odul\l' COI"' I!R$ 1
CJ A VO R&conmm;sance process

Quality control
D Ouality co,JitoJ (()atvros
~
c
D Resttll analysts te.atures '+::::
t:
Attribute display using
VOis (4)
6athers disp-lay in
Interpretation window (5)
~
Q)
t.)
c::
ctl
(/)
-!!2
ctl
c::
AVO Cross-plot analysis {6) c::
ct.) c
0
·;::;
~ ~

Seismic anomaly hunting ~


":::(
i
EQ)
>
·;::;

( '
Geologteal and Geophysicallnterpret!ltlon I ~
:::;,
~
·~
~ a
Inputs to the workflow consist of angle stacks or prestack data.
Attributes are calculated from the data on the fly using several AVO
approximations. Attributes can be co-visualized with the stacked
seismic data using Volumes of Interest (Figure 2). Attributes from a VOl
can be crossplotted for further analysis. You can select among several
different AVO approximations according to which attributes are of
interest.
Figure 2 shows an example where a virtual AVO gradient section is
overlaid using transparency on a stacked section. This functionality
allows the interpreter to connect AVO attributes directly with the
structural and stratigraphic information in the stacked seismic volume.

Figure 2 An Interpretation window showing a seismic section overlaid by a


volume of interest that displays the gradient attribute

This module covers the procedures necessary to run the AVO


reconnaissance workflow. It is divided into three parts.
• how to create the virtual attribute volumes
• how to visualize the attributes
• how to create cross plots of the virtual attributes

164 • AVO reconnaissance Quantitative Interpretation


Procedure- Set up and run the AVO reconnaissance
workflow
This procedure shows you how to start the AVO reconna issance
workflow and create the basic volumes with which to work.
1. On the Quantitative Interpretation tab, in the AVO group,
click AVO reconnaissance.
'!jJ! ~ Volume Attributes
&JJ AVO Seismic logs
AVO
•reconnaissance I Wedge model
AVO

The AVO reconnaissance dialog box opens.


~ AVO reconnaissance

AVO Reconnaissance I
AVO method: (Shuey 2-term ·I
Input f Outputl
1@ Post-stack () Pre-stack
~
Seismic list
§] Seismic Effective angle :

~
LJ
~
IT'

Line fitting mode

Method: !standard ·l

l.t Apply II ,f OK I ~< Cancel J

Quantitative Interpretation AVO reconnaissance •165


2. Select one of these AVO methods from the list:
• Shuey 2-term
• Gidlow
• Pan and Gardner
~ AVO reconnaissance

Input Ou ~idlow -
o Post-stack Pan and Gardl'er

NOTE: As discussed earlier, each method uses a different


approximation to the Zoeppritz equations to model the AVO
behavior. Using different approximations allows you to
calculate different suites of AVO Attributes.
3. Open the Input tab and select the data to be used in the
reconnaissance process.
You can use these two types of input data:
• Poststack data, which refers to angle stack data
• Prestack data, which refers to prestack angle gathers
Select the appropriate option according to the data type to be
used.
~ AVO reconnai;sance

Pre-staok

Seismic l.st: Seismic

4. If you select poststack data, locate the data in the Input pane
and click next to the Seismic list to enter different angle
volumes.

166 • AVO reconnaissance Quantitative Interpretation


NOTE: Effective angles must be entered manually unless the
angle range for the partial stack has been entered in the Meta
data of the Settings tab for the cube. The formula to calculate
the effective angle is
. 2 1 . 2 . 2 . .
sm 89 ft = 3 Csm ()min + sm ()max + sm()min sm ()max)

This figure shows an example of angle stacks entered together


with the appropriate effective angle displayed. The data
volumes do not have to be in order.
I Input [outPUt]
I@ Post-stack 0 Pre-stack

Seismic list:
~. Seismic Effective angle

I~ g; Near_S-20 13.11 :

[Q] mMid_18_33 25.61 :

~ ~ Far_31-46 38.32 :
I

5. If you select prestack data, locate the data in the Input pane in
the Seismic folder and click ¢ to enter the dataset. Select
the trace limits in the Trace section as shown in this figure .

• Pre-stack: ~ ~~3DT_SEC_ORD_CE CTR!'3DT_PRI· _ORI )'(

T~re --------------------------------

Minimum:

Maximum:

Step:

The Trace section lists the first and last trace and the
increment. You can adjust these values if you want to limit
which traces are used in the calculations. Limiting traces can
be useful if the angle in the data exceeds the useful range for
the approximation being used, or if there are noises or
acquisition imprints that affect the near or far traces that might
bias the calculation.

Quantitative Interpretation AVO reconnaissance • 167


6. In the Line fitting mode section, select one of these methods
to use to fit the data from the list.
• Standard: As discussed earlier, this method performs a
normal least squares fit to determine the coefficients in
the approximation that are the primary AVO attributes. For
noise free synthetic data and good quality real data with
well-defined primary reflections, the Standard method
should be adequate.
• Robust: If the data contains noise, such as multiples,
residual ground roll, or other systematic noises, a better
fitting method is required that will not be as sensitive to
what might be potentially large outlying amplitudes in the
gather. The Robust linear fitting follows the approach of
Walden (1991 ).
The method first estimates the median angle, and the
amplitude of all of the groups of traces fits a line through
the points. Based on the distance from the data point from
the fitted line, outliers are weighted down or eliminated
according to the Scale factor. The Overlap parameter
defines the overlap between the different groups of traces
used in the median calculations.

Line fitting mode - - - - - - - - - - - - -


Method: 1~.-.Robust
_ _ _ _ _ _ _ _ _ _----J
Scale f!ctor. - 2.10 ffi
Overlap 0~

There are a minimum number of required angle stacks or


prestack angle traces, depending on the approximation used
and the fitting method applied:
• Standard fit- Shuey and Gidlow: three angle partial stacks
or live traces in gather
• Standard fit -Pan and Gardner: three angle partial stacks or
live traces in gather
• Robust fit- Shuey and Gidlow: five angle partial stacks or
live traces in gather
• Robust fit- Pan and Gardner: six angle partial stacks or
live traces in gather

168 • AVO reconnaissance Quantitative Interpretation


7. Open the Output tab and select the Primary and Secondary
attributes that you want to have calculated. The options on
this tab change, depending on the algorithm that you select.
This figure shows the Shuey 2-term attribute list.
~AVO reconnaissance

--------------------------~
Intercept
Graa~ant

Second8ryettnbutes - - - - - - - - - - - - - - - - - - - - - - 1
[] Check ell

D Intercept • gmdient: iIntercept • §<adient


D G3s indicator. ~cater
D Ruid factor. 1Fluid factor
2.0 ;-

[ ] Reflection coefficient difference Re~n coorndert difference

0 Sign(intercept) • gn~dient 1Sign(i terce:Jt}~ grodJ?nl


D P-wave reflection coefficient ~==~============~
P...,taVe re~ed:!on coefficient

Poisson's ratio contrast I~================~


Poisson's ratio comrasl

Nl • PRC:
~================~
: Nl • PRC

Rctalicn angle
~========~========
0 :

OK j [1t Cancel J

This figure shows the Gidlow attribute list.


~ ·x-
-~---~-~·.: __ -.. ~ _,-_ ~ -- --- - ---~~-·

Acoustic impedance contr~ Acoustic impedance contrast


Shear impedance contrast Shear Impedance amtrast

Secondary si!Jibutes - - - - - - - - - - - - - - - - - - - - -
0 Check all

[] Fiuidmctor:
Vp/Vsrano· 2.0 :

Quantitative Interpretation AVO reconnaissance •169


This figure shows the Pan and Gardner attribute list.
I
Primoryettributes - - - - - - - - - - - - - - - - - - - - - - - -
ConJIAnl term: Constont term
Lin~rterm. L1neorterm
Quedretic term. O uodrei"Jc temi"
Secondory ettributes - - - - - - - - - - - - - - - - - - - - - - - 1
[J Check oil
[J P-weve velodty contrast·
0 Density c:ontrtiSt
[J VpNs rotio controat:
VpN rollo 2.0 :
D Poisson's ratio contrest:
Voll/r:-rl3ilo 20:
t:J S-weve velocity centres I: IS·W3Vi: velocity contrast
VpNsr!llfO 2.0 ':!

0 S"W!!v• tmpedonce cantrosr.. S·wave tmpedcn ~Ires ::====--=:-=:====


Vo,Vs rct1o 2.0 :
0 Shear n1odulus contrast: [ Shear modulu~conui!st -~=~-=-~~-=~-=
VpNs ro1kl 2.0 -;

Apply I[" OK ] [ ,. Cancel

8. Click Apply or OK

170 • AVO reconnaissance Quantitative Interpretation


Entries for the attributes appear in the tree as seismic volumes .
.. litO~~
:> fl 0 ~..J ~G;'
, em o .tplw{).fd183..,~
.. tJ 0 JuplerfNeirl20}
0 lnt~ne ~&5
0 XI.Jne 2408
.J 0 [~;~~::.-~.]
" ESi 0 AVO ntercept and gmdiert
.. tJ O~
Olnline3458
0 XUne 23.83
" tJ 0 Intercept • gmdiert
0 lnline 3458
0 XUne 238J
"~0 AVO reflection coefficient and Poisson's ratio contrast
" tJ 0 P-.~~'fiJ'dhction~
0 lnline 3458
0 XUne 2383
" tJ 0 Poisson i r6lio conlnJst
0 lnline 3458
0 XUne 2383
"tJ D NI.PRC
0 lnline 3458
0 XUne2383
.. tf o~
0 lnline 3458
0 XUne 2383

The entry shown here is for a poststack calculation . A similar


entry appears under the prestack data name if the calculation
is run on prestack data.
NOTE: The AVO reconnaissance attributes are created below
the first angle stack in the Input pane in the Seismic folder.
The attributes are listed in a hierarchical order with secondary
attributes nested between the primary attributes used to
create them.

Quantitative Interpretation AVO reconnaissance •171


The volumes created in this procedure can be visualized in the
Interpretation window and in the 30 window like any
standard seismic volume. An example is shown in the figure.
The Intercept is in on the left and the gradient on the right.

.. ..

Volumes of interest
In an exploration context, initial data screening is likely to be done on
stacked data volumes received from a contractor. Stacked data likely
has a higher signal-to-noise ratio than attribute volumes. It also is likely
to be the preferred dataset among members of an exploration team
who are looking for potential leads and prospects.
A useful way to combine working with the stacked data and attributes
at the same time is by creating a Volume of lnterest(VOI). A VOl is a
tool that allows you to define the extent of a volume to be used in the
AVO analysis. The VOl displaying the AVO attributes can be overlaid on
a stacked seismic section in an Interpretation window. You can
visualize the two datasets together to obtain a qualitative
understanding of the relationship between them.
The vertical and horizontal extent of a VOl can be increased or
decreased from the Interpretation window directly. In addition, the
spatial extent (that is, inline and cross line dimensions) can be adjusted
in a 30 window.
112 • AVO reconnaissance Quantitative Interpretation
The VOl can display all of the seismic volumes available within the
same seismic survey. You can change the volume displayed in the VOl
interactively by using the Inspector tool, which is discussed later in
the training.
Figure 3 shows an example of a VOl window that contains the gradient
attribute overlaid on the stacked seismic volume in an Interpretation
window. The VOl can be panned around the section and its size
adjusted to see how the attribute appears at any point of interest.

Figure 3 An Interpretation window showing a seismic section overlaid by a


volume of interest displaying the gradient attribute

Quantitative Interpretation AVO reconnaissance •173


Procedure - Create a VOl to co-visualize AVO attribute
data with other seismic data in the Interpretation window
This procedure shows you how to create a VOl and display on the fly
AVO attributes.
1. Open an Interpretation window and display the seismic
in line or cross line where you want to view the AVO attributes.
2. On the Quantitative Interpretation tab, in the AVO group,
click Ql tools.
~ Volume Attributes
~~ AVO Seismic logs
AVO
reconnaissance Wedge model
AVO

The AVO Tool Palette opens.



~ Select

3. Click Select in the Interpretation window tool bar.

~ I ~~ [Sfl~
4. Click Create/edit volume of interest in the Tool Palette.

~ Create/edit volume...
-. Qltools X

5. Draw the VOl region in the Interpretation window (pink


rectangle in the example shown in the figure).

174 • AVO reconnaissance


The VOl appears as an entry in the tree as shown in this figure.
~ ~ VOis
fl) ~ Volume of interest 1
C.J ~ Vofume ofinterest2
ieW,l ~ Volume offnterest3
i!!Jl! ~ Volume of Interest 4
fiJ ~ Volume of interestS
There are several ways to proceed with the next steps. Using
the Inspector tool is a quick and convenient way to perform
these steps.

6. Open the Inspector tool

7. Click in the VOl (Select mode .1[:a. must be on).\


The Inspector displays options for the VOl.

Quantitative Interpretation AVO reconnaissance •175


8. Select the dataset to display from the list under
Reconnaissance tools and adjust the transparency as
necessary.
NOTE: Only seismic data volumes from the same survey are
available for display.
~~~~

9. To pan the VOl, make sure that the Tool Palette is active.
a. Click Create/edit volume of interest.
- Tool Palette 0 X

~ Create/edit volume ••.


., Qltools. x

W ~l fit ~ ~ ~
b. Click in the VOl to drag it to another location or click the
sides to resize it.

By covisualizing the data and AVO attributes, the AVO


signature can be evaluated at the same time as the structural
configuration. AVO responses can be ranked appropriately.

176 • AVO reconnaissance Quantitative Interpretation


Procedure - Display VOis in a 30 window
VOis also can be manipulated in the 30 window with results displayed
simultaneously in the Interpretation window. This procedure shows
you how to display and manipulate VOis in a 30 window.
1. Open a 30 window and make it the active window.
2. Select the VOl and the seismic data volumes of interest in the
Input pane.
The VOl appears. An example of the 30 window display is
shown in the figure on the left. An Interpretation window
displaying the same data is shown in the figure on the right.)

...---
.
. ......... .......
.. - .. --......:
-~
..

--

3.

4. Select Pick mode.

Quantitative Interpretation AVO reconnaissance •177


5. Click one of the corners of the VOl and drag it to resize it.
The VOl in the Interpretation window resizes to match the
selection in the 30 window.

.,... •,1. -- .... .


-:-~
...... .........
. ........

-2300

-2400

-2500

Procedure - Create a crossplot of AVO attribute data


from an AVO reconnaissance VOl
You saw earlier that a useful way to analyze AVO attributes is to
cross plot them. Crossplotting allows you to isolate areas that are
anomalous with respect to the background reflections. When searching
datasets using the AVO reconnaissance process, it is useful to crossplot
data in a VOl when something of interest is seen. This procedure shows
you how to crossplot data in a VOl.
1. Open an Interpretation window and display the seismic data
and an AVO attribute in a VOl.
2. Click the Create/edit volume of interest button from the 01
analysis Tool Palette.
3. Adjust the area of the VOl using the Create/edit VOl button
on the AVO Tool Palette to include the data of interest.

178 • AVO reconnaissance Quantitative Interpretation


4. Right-click in the VOl and select Create new crossplotfrom
the menu.

Sra.·, tr~ ~ e l~ item in tree


·····-·------- ---------------.--··--· ··--··· .I

The Create new crossplot dialog box opens.


5. Set the horizontal and vertical axis and click OK to finish .
Create new crc~sp ot ~

C-os.; pl o: def ~ l :ion -------------

Se-ismi~ coll ection. 1Js ~"eJ 1

1-o-i=ntal a;ci:s; [!}[ ~ J~r{Far3146}


~------------------
... ,

{I Ve.rti::a an s: §JI tJ ter-cept


~--~====~~~--
•I
!" ot< I[.x Cannel I
6. Select the datasets to crossplot on the horizontal and vertical
axes by inserting ? or selecting them from the list.
Create ner.v cro:;sp ot

C•oss pic: de f.,i:i on - - - - - - - - - - - - - - -

Vo wre o ' interest (::i:}Vol me of · :eres: :.J

Seismic co lroi::m

tl f- orizcrml axis:

Quantitative Interpretation AVO reconnaissance • 179


A new crossplot window appears, showing the selected
attributes for the data filtered by the VOl.

Intercept

You can pan the VOl around the section. The crossplot updates
as the VOl changes location.

180 • AVO reconnaissance Quantitative Interpretation


Exercise 1 - Apply the AVO reconnaissance at a prospect
In this exercise, you run AVO reconnaissance using three angle ranges
as input to generate attributes. Using VOl, you compare the
lntercept*Gradient (I*G) attribute with the original seismic volume. The
results are open to discussion.
1. On the Quantitative Interpretation tab, in the AVO group,
click AVO reconnaissance.
The AVO reconnaissance dialog box opens.
2. Select the Shuey 2-term option for the algorithm.
3. To enter the three angle range datasets, click
The effective angles for the angle ranges are shown in this
table.
Stack angle range (degrees) Effective angle (degrees)
5- 20 13.1
18-33 25.6
31-46 38.3
4. Display the intercept, gradient, and other attributes in the 3D
window and look for any anomalous responses.

Quantitative Interpretation AVO reconnaissance • 181


Use the 01 tools to draw a VOl in the Interpretation window.

........
To choose a data volume to display, use the Inspector.
'-::::-----:::::;;;;~~::-. ·~

7. Adjust the transparency and move the window over the section
to observe the response in different locations.

Questions
Do you see any consistent behavior in the attributes on and off
structure consistent with the AVO modeling in Module 2?

182 • AVO reconnaissance Quantitative Interpretation


Results I Discussion
Figure 4 shows a series of strike-oriented seismic lines from the seismic
volume along with the lntercept*Gradient attribute (calculated from the
angle stacks).
+

~· :

XL 2346

....
-- ~·· · ··-~
.s.- - ~
i.# .. ~..:.; =-=~.....~"! ~
-
~: . 0- · ~
:.:-:- ~
~~
- · :-:-·

Figure 4 Seismic appears with stacked data on the left and the Intercept*
Gradient attribute sections on the right

Quantitative Interpretation AVO reconnaissance • 183


The lines run from a position east of the structure across the crest and
then are downthrown to the fault and off the structure again. The first
cross line is off the structure to the east. The attribute shows high
amplitude events. These events are regional in nature and have a
negative I*G attribute value. As the location of the data moves on
structure, patches of reflectivity with the opposite sign appear
consistent with the sign expected for I*G from the AVO modeling.
The data is patchy, which might relate to underlying acquisition
problems. The structure also is cut by many small scale faults and is
overlaid in the crestal area by a shallow near seafloor gas anomaly that
affects data quality in the crestal region. Despite all of this information,
there does seem to be a consistent pattern between the on structure
and off structure AVO attributes that is consistent with the modeling
and the presence of gas pay in the structure.
A line location map is shown in Figure 5. The attribute in the map is the
stack amplitude of an event near the top of the pay in the well.

Figure 5 Location map showing the location of several crossline sections


through the structure

184 • AVO reconnaissance Quantitative Interpretation


Review question
How does the AVO reconnaissance process help evaluate data?

Summary
In this module, you learned about:
• running the AVO reconnaissance process
• creating VOis for interactive display and co-visualization of
AVO attributes
• making interactive crossplots of the attributes displayed in the
VOl

Quantitative Interpretation AVO reconnaissance • 185


Module 4 - Simultaneous
seismic inversion and the
Inversion property builder

Created by Schlumberger PTS for inversion consulting services,


simultaneous seismic inversion inverts the seismic data to acoustic and
elastic properties using a linear approximation to the Zoeppritz
equation. First, you run the Inversion property builder, which uses
the well and horizon data to generate a model to fill the low frequency
gap of the seismic bandwidth. This Low frequency model (LFM) and the
angles stacks and their associated wavelets are the inputs of the
simultaneous inversion.

Prerequisites
To complete this module successfully, you must have completed the
previous modules in the course.

Learning objectives
After completing this module, you will know
• some basic theory regarding Simultaneous inversion and
review in brief some key workflows, such as Wavelet
extraction, Trace Alignment and Log Conditioning
• how to use the Inversion property builder to create a low
frequency model for use in simultaneous inversion
• how to set up and run a Simultaneous inversion
• how to check the quality of the output using the simultaneous
inversion diagnostics and the Inversion QC plot tool
• how to analyze the output using crossplots and geobodies

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 187
Lesson 1 -What is seismic inversion?
The class introduction discussed how changes in the acoustic
impedance at layer boundaries in the subsurface created reflections as
seismic energy propagated through the earth. Figure 1 illustrates this
concept. The left side of the figure shows a column of different rock
types. Adjacent to that column is a schematic blocky impedance profile.
Each change in impedance is a source of reflected seismic energy.
On the right is a scaled up view showing perhaps several thousand feet
of the subsurface. The velocity and density logs shown here have many
contrasts each of which can give rise to a reflection. In addition to the
many layers in the logs, what also can be seen is a clear trend in the
log values with depth. While the reflection coefficients only depend on
the local contrast in impedance, the rock properties have absolute
values, and these values vary with depth.

Org Acoustic Reflection


Sonic Impedance coefficients

Figure 1 A lithologic column (left) showing impedance contrasts at the layer


boundaries. A sonic and density Jog on the right can be used to calculate the
acoustic impedance, and from there the reflection coefficient sequence

When you do a seismic survey, you propagate a wave whose reflected


value is recorded. Figure 2 illustrates this concept. It shows the pulse of
energy (the seismic wavelet) reflected at each spike in the reflectivity
sequence.
The sum of these reflected energy pulses is the seismic record.
Mathematically, you can simulate the seismic trace by multiplying the
wavelet by each reflection coefficient, placing the result at the time of
the coefficient and adding up all the results. This process is termed the
188 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
convolutional model. You use this model when you create seismic

b
synthetics as you did in Module 2.

Reflectivity •
Series Multiply each wavelet S~thebc
by the size ofthe Sewsmogram
reflection coefficient

Zero Phase
Wavelet

~.~.~.~~.~~-~· · -1t~~~~~.~-~
-
Figure 2 Convolutional model for calculating a seismic trace from a wavelet
and a reflection coefficient sequence

One consequence of having a finite pulse of energy is that it lacks both


low and high frequencies. Figure 3 illustrates this concept. This figure
shows a frequency spectrum with the blue curve that represents the
frequencies recorded in a possible seismic dataset. Both low
frequencies and very high frequencies are absent.

.s

-10

il "
·2S

~~--~--------~--~--~----~--~--~
20 JO ~ so 70

Ft~IHZI

Figure 3 Portion of the frequency spectrum occupied by seismic data and the
low and high frequency gaps that are present

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder •189
Lacking high frequencies means that there is a limit to the detail that
you can record and resolve. Therefore, very thin beds are difficult to
resolve and detect. Lacking low frequencies means that you cannot
recover the absolute values of the impedance directly from your data. If
you want to interpret absolute values for subsurface properties, you
must supplement the higher frequency detail of your seismic data with
a model for the low frequency portion of the earth spectrum.
The AVO responses that you worked with earlier told you a lot about
what created the reflections that you observed. However, they were
based on reflections, so they told you only about the contrasts that
caused those reflections. You did not measure absolute values. What
you want to do with inversion is to go beyond the reflectivity contrasts
and use the seismic data along with other data sources to calculate the
absolute properties of the underlying earth model.
To fill the low frequency gap, you use a model for the low frequencies
derived from well data. The higher frequencies come from the seismic
data. You use the convolutional model to make sure that any results
contain the correct high frequency detail by making sure that they
predict the seismic data to within a sufficient tolerance.

Low frequency
model from well
data

Figure 4 Portion of the frequency spectrum that is filled by well data in the low
frequency model

In this module and in Module 5, you look at two types of inversions that
are available in Petrel: simultaneous inversion and stochastic
inversion. In this module, you learn about simultaneous inversion.
In simultaneous inversion, you calculate (or invert for) a single earth
model that contains absolute property values. The low frequency model
provides the guide for the absolute values of the properties. Detail of
190 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
the interfaces and smaller scale layers come from the seismic
reflectivity. You do not try to extend the model to thin layers for which
the seismic data has no information.
In stochastic inversion, you add data to your model at a scale beyond
what is contained in the seismic data. Because the seismic data cannot
help you with this, you use statistical models of how the subsurface
varies to create alternative realizations that are consistent with the
well-based low frequency trends and seismic data. In this kind of
inversion, you invert for many possible solutions, not just one solution,
that contains a range of possible properties. Each result should be
consistent with the statistical models, the seismic, and the low
frequency properties from well data.

Applications·of simultaneous inversion


Simultaneous inversion does not add detail beyond what is resolved
already by the seismic bandwidth. There is improvement in interface
positioning and in amplitude estimation because wavelet effects are
removed. However, the resolution of the inversion results is normally
higher than that of the input seismic because of the reduction of thin
bed tuning effects due to the wavelet removal. The situation is the
same for lateral heterogeneity. You are constrained by the lateral
changes that are resolved by the seismic data.
Given these characteristics of simultaneous inversion, if an exploration
play or a field development involves very thin, poor, or unresolved
reservoirs, then simultaneous inversion might not help significantly in
characterizing the subsurface. However, if the objective is not
significantly below the seismic resolution reservoir unit (for example,
the Norphlet formation in the Eastern Gulf of Mexico), and you want to
characterize how if varies vertically and laterally, then simultaneous
inversion is an appropriate tool to use.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder •191
With this in mind, you can apply simultaneous inversion in situations
such as the ones listed here:
• Characterizing and observing the reservoir using integrated
seismic and well data
• Mapping elastic properties (for example, acoustic impedance,
shear impedance, and density) to perform these tasks:
• classify lithology
• estimate porosity or saturation
• predict pore pressure
• perform geomechanical deformation
• Interpreting converted wave data

Simultaneous inversion workflow


Figure 5 shows the simultaneous inversion workflow.

Figure 5 Simultaneous inversion workflow

Input data
The input data consist of:
• seismic data in the form of angle range stacks
• wavelets for each angle range stack
• low frequency models for the inversion parameters (for
example, acoustic impedance, VpNs, and bulk density)

192 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Output
The output depends on the data type:

For PP data • Acoustic impedance and shear impedance


• Acoustic impedance and VpNs
• Acoustic impedance, VpNs, and density
• Shear impedance and density
• Acoustic impedance, shear impedance, and
density
• For time lapse data, acoustic impedance, VpNs,
density, and monitor ratio of change

For PP and Acoustic impedance, shear impedance, and density


PSAVO
data

For AVOZ Absolute impedance, slow shear impedance, fast


acoustic shear impedance, azimuth of fast shear impedance,
impedance and density

Key features
Key features of the Petrel Simultaneous inversion are
• Full integration with Petrel data, models, processes,
Workflow editor, and graphics
• Time variant wavelet capabilities
• Global Simultaneous inversion methods using the Aki and
Richards and Fatti approximations
• Acoustic and elastic output volumes include Acoustic
Impedance, Shear Impedance, VpNs ratio, and density.
• Joint PP and PS mode
• Azimuthal Horizontally Transverse Isotropic (HTI) mode, with
output volumes for acoustic impedance, slow and fast shear
impedance, azimuth, and density
• Poststack and prestack 40 modes
• Multi-threading to leverage multi-core, multi-socket CPUs in
modern workstations

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder •193
If a good inversion is to be achieved, data preparation is extremely
important. These steps are important in the data preparation process:
• The seismic data must be processed through a flow that is
suitable for quantitative interpretation. This requirement
implies proper amplitude recovery and preservation and
wavelet deconvolution.
• Logs must be edited and conditioned properly. This task should
be done prior to synthetic generation and wavelet derivation,
but before using the logs in the low frequency model. Good
well ties also must be established.
• The datasets must be aligned. This task can be done through
the Petrel Seismic trace alignment process or other similar
process.
• Wavelets must be derived with care for each input data stack.
• A prior low frequency model must be constructed that includes
the well data and other input (for example, velocity models)
that shape the low frequency behavior of the model.
NOTE: The Seismic trace alignment, Log conditioning, and Wavelet
extraction processes are covered in more detail later.
After the inversion has run, calibration of the inverted properties to
parameters of interest is necessary to make predictions about
parameters such as reservoir porosity, fluid presence and type, and
lithology. These parameters must be based on suitable rock physics and
petrophysical studies so that you can make the relationships for fluid
prediction between parameters, such as acoustic impedance and
porosity or acoustic impedance and VpNs.

194 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Simultaneous inversion features and implementation in
Petrel
Key features of Simultaneous Inversion as implemented in Petrel are:
• It is a globally optimized deterministic inversion that uses a
simulated annealing approach to model perturbations in the
search for a solution. It is a model based inversion that uses a
low frequency prior starting model and outputs absolute elastic
parameters and optionally the forward modeled synthetics,
reflectivity and residuals.
• It simultaneously inverts multiple datasets, for example, angle
stacks, PP-PS data volumes, or 40 data sets.
• It has robust noise reduction that uses trace-to-trace continuity
to reduce sensitivity to noise.
• It can take into consideration time variant absorption effects.
Simultaneous inversion iteratively updates the elastic parameters of a
subsurface model until an objective function is minimized. The objective
function contains terms that measure the misfit with respect to the
seismic data, sample continuity, and other values to be discussed in
this training.
The optimization used in this inversion is a simultaneous inversion,
meaning that it uses all of the seismic input (for example, all the angle
stacks) simultaneously. It does not invert separately for each angle
stack.
It is a global optimization, meaning that the cost function is optimized
for the entire dataset, with no subdivision. This feature is important
because inversion techniques using global optimization can use
complex objective functions with many local minima corresponding to
realistic statistical models for subsurface prediction from seismic data.
The global optimization ensures that the subsurface model does not
become trapped at a model corresponding to one of the many local
minima in the objective function, but that the final optimum model
corresponds with the global minimum. Figure 6 illustrates this concept.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 195
c
0
;:
(.)
c
.2
~
0

Global search covering all possible solutions

Figure 6 Potentia/local minima in the cost function highlighted by the red dots.
The inversion seeks to find the global minimum highlighted by the green box.

The optimization uses a process of simulated annealing to search for


solutions to the inversion. The process has 15 main cooling stages,
optimizing a full trace at a time.
The implementation is deterministic; that is, it produces the same
output again, if the same input is used.

Penalty function
Inversions such as the simultaneous inversion discussed in this training
operate by trying to minimize a function that measures various
quantities such as the misfit with respect to the seismic, the
smoothness of the model, and the number of reflectors. This function is
termed the cost function or the objective or penalty function.
The cost function for Petrel simultaneous inversion contains these
terms:
• Penalty for differences between the seismic data and the
synthetic seismic determined from the estimated property
models by convolutional forward modeling
• Penalty for horizontal variations in the estimated property
models
• Penalty for deviation of the estimated property models from the
low frequency model
• Penalty for the presence of significant reflectors. Significant
reflectors are places in the estimated property models where
the reflection coefficient exceeds a predefined threshold.
• Penalty for vertical changes in properties between the
significant reflectors.
196 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
This term means that the final model must fit the seismic data. It also
means that the model tries to avoid rapid lateral variations in
properties. It tries to keep updates close to the low frequency model,
and it tries to avoid having too many large reflectors with big property
changes between them. Significant reflectors are not forbidden per se,
but the penalty avoids letting the inversion place too many large spikes
in the solution unless needed to fit the seismic data properly.
Expressed as a formula, the objective function is shown in Table 1:
Table 1 Terms for Objective Function

Term lnterpretati on
E = f(SNR) X "L(Si- di)2 Penalty for not fitting the
seismic
Penalty for horizontal
variation
Penalty for deviations from
the low frequency mode
+ R1 (#significant reflectors) Penalty for number of
reflectors that exceed
threshold

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder •197
Table 2 defines the parameters of the objective function.
Table 2 Objective Function Parameter Definitions

Parameter Definition
SNR User-defined signal-to-noise ratio
di Seismic data at sample i
Si Synthetic seismic at sample i
Ra User-defined horizontal standard deviation parameter
ci Continuity at sample i
P. Property value at sample i
pi±1 Property value at the four neighboring samples to
sample i, which might be determined by dip
Rcr User-defined standard deviation between the current
inversion result and the low frequency model
plfm Property value of the low frequency model at sample i
R1 User-defined threshold for significant reflectors
#significant The total number of significant reflectors
reflectors

User controls
The functions before each term in the objective function contain these
user controlled weights:
• The SNR or signal-to-noise ratio term that weights the fit to the
seismic data
• The horizontal standard deviation parameter Ru
• The standard deviation between the current inversion result
and the low frequency model Rcr
• The reflection coefficient threshold for significant reflectors R1.
This threshold controls how many reflectors are counted as
significant for inclusion in the cost function.

198 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
SNR parameter: f(SNR) x L:(Si- di)2
To assess the fit between the inversion result and the seismic data, the
inversion result is convolved with the wavelet and compared sample by
sample with the seismic. You want the inversion to calculate the
underlying rock properties, and, as much as possible, avoid modeling
any noise that might be in the seismic data. The SNR parameter
controls how close the match needs to be to allow for the fact that an
exact match requires the inversion to match both signal ~nd noise. If
S/N is set very high (that is, assuming there is very little noise in the
seismic), the inversion tries to match the data in much greater detail
than if it is set at a low value.
The parameter has a usual working range of 1 to 20 with a good default
being 5 +/- 1. It is not measured in db and should be evaluated as a
"relative goodness of fit" parameter and not as an absolute signal to
noise ratio for each stack. It is set separately for each input volume.
Trials with different parameter values are needed to determine a good
value for the dataset at hand.

Continuity parameter: f(1/ Ra) x L:C..I (P.-


I
P.l:t 1) 2
The horizontal continuity (Ra) parameter controls to what degree
horizontal variations in the property models are penalized. Ra is the
standard deviation of neighboring property traces. The larger the value
given to Ra, the smaller the penalty for lateral changes, and the
inversion optimization algorithm imposes less horizontal continuity. A
smaller value increases the penalty for horizontal variations and so
imposes more lateral continuity.
A usual working range is 0.01 to 0.20. Lowering the number yields an
increasing amount of horizontal continuity. Again, testing is necessary
to determine the optimal value.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder •199
Tie to LFM parameter: f(1/Rcr) x L(Pi- Plfm) 2
The tie to the LFM (Rcr) parameter controls to what degree deviation of
the estimated property model from the LFM is penalized. The greater
the Rcr value, the lesser the penalty, so the further the estimated
acoustic or elastic property model is allowed to deviate from the
associated low frequency model.
A usual working range is 0.01 to 0.20. Lowering the number forces
greater tie to the low-frequency model. In AVO inversion there is a
separate tie to LFM parameter for each of the input domains.
NOTE: The inversion requires that the value of Ra, which controls the
degree of horizontal continuity, must be smaller than the value for Rcr,
which controls the tie to the LFM. This requirement ensures a greater
emphasis in the inversion on horizontal continuity over similarity to the
LFM, and it helps mitigate a tendency toward a single trace inversion in
extreme cases.

Reflection coefficient threshold parameter: R1 (#significant


reflectors)
The threshold for reflection coefficients (R1) controls the definition of
significant reflectors in the inversion. All points in the estimated
property models with a reflection coefficient greater than the threshold
are interpreted as significant reflectors.
The penalty for the presence of significant reflectors is controlled by the
value of R1. As R1 increases, the penalty for each significant reflector is
increased and fewer significant reflectors are present in the inverted
property.
If a reflection is larger than required by R1, the process allows inversion
to enter a layer interface. It also allows a potential significant property
value change over the interface, thus avoiding smearing over
boundaries.
• For poststack data, a value of 0.015 (1.5%) is a recommended
starting point. The usual parameter range is 0 to 0.03.
• For AVOZ and time-lapse data, the unavoidable misalignments
of seismic usually make it necessary to lower the number
significantly, for example, to 0.00015 for three input seismic
data sets.
200 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
• For quality control, observe the output of the Relative number
of reflectors field from the inversion case. Bracket the reflection
threshold up or down until the output of Relative number of
reflectors is around 0.2 (20%), or until other factors visible in
the output datasets indicate the desired output quality. If more
than 25-30% of reflectors are significant reflectors, the
inversion run time increases and there is a tendency to model
more noise in the inversion than necessary.
The next lessons discuss the other three important steps in the
inversion workflow.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 201
Lesson 2 -Inversion preparation: Well tie and
wavelet extraction
In the inversion process, the intermediate results are compared to the
seismic to assess the degree to which the inversion result at that stage
correctly reproduces the input seismic data. This process is done by
convolving a wavelet with the reflectivity sequence computed from the
inversion result. Because there are multiple seismic angle volumes
involved, a wavelet appropriate to each volume must be supplied. In
addition, the well data must be tied carefully to the seismic so that the
LFM is registered correctly and the low frequency model properties are
associated with the correct geologic horizons.
Workflows to perform well ties and wavelet extraction are contained in
the Seismic well calibration group on the Quantitative
Interpretation tab. A complete review of the Seismic well
calibration process is beyond the scope of this course. It is assumed
that people taking this course already are familiar with Seismic well
calibration and its functionality and have completed Petrel training,
such as Seismic Well Tie or Petrel Geophysics, where this material
is covered in detail. For this reason, this training only briefly reviews the
wavelet derivation process recommended for generating wavelets for
inversion purposes.
It is recommended that wavelets be determined by means of the
Deterministic wavelet process. Deterministic wavelet derivation has
several key features that make it the method of choice:
• It determines the amplitude and phase of the wavelet.
• It correctly estimates the amplitude of the wavelet so that the
seismic prediction from the inversion is scaled properly for
comparison to the actual seismic.
• The deterministic process includes a diagnostic that estimates how
well the reflectivity and the underlying seismic reflectivity of the
well match. This diagnostic is useful if the well location is
uncertain or if there are data positioning uncertainties because of
migration errors, which are common in many datasets. The
measure of match between the two reflectivity sequences is
termed Predictability. It helps to select the best location for
wavelet extraction

202 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Seismic trace alignment
Because the inversion develops a model that is consistent with the
event amplitudes in the seismic data, it is critical that seismic events
be aligned properly so that the estimated elastic and acoustic
properties are the result of geologically-based AVO and are not biased
by event misalignment.
Figure 7 shows an example of data that is not aligned properly. Each
panel in the figure is a suite of traces from a different angle volume. In
this example, the data does not invert well because the traces are
aligned poorly and in some cases (as highlighted in the red boxes), the
events on the first trace are almost aligned with troughs on the last
trace.

>~H 1 1)~~:~-.~::·~~-·.,- ;-->. \'fi.


"'"~ ... "
_........,..._
..
~ - _,__.----

.. .
._.._.........___..,..----r-- 7:'..,........
,. ~ .... ' ..... ~

Figure 7 Example of poorly aligned data cubes


Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 203
Using the Seismic trace alignment workflow, it is possible to apply
residual, data-driven flattening to further enhance seismic data before
its use in AVO and 40 workflows. Seismic trace alignment uses
Non-Rigid Matching (NRM) (Nickel M., Sonneland L., 1999) to flatten
offset gathers or angle gathers to optimize the data for AVO and
inversion workflows.
Figure 8 shows the Seismic trace alignment workflow.

Seismic Trace Alignment Workflow

Legend: 811111

Figure 8 Seismic trace alignment workflow

The inputs are the seismic datasets to be aligned, one of which is


nominated to be the reference cube. The NRM process estimates the
displacements that can be viewed and OC'd as diagnostics when the
process completes. The displacements can be filtered to remove
unnecessary high frequencies in the shifts. The data then is stretched
to result in the flattened and aligned datasets. This step is followed by
a check of the quality of the volumes and inspection of the correlation
surfaces.
204 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Figure 9 shows the process by which NRM develops the displacements.

Cube 1
CumShlftl
matched

~------ CumSMt3

- - - CumShlft4

CumSM tS '

Figure 9 Sequence in which shifts are determined

Log conditioning/editing
As with any process that depends on log data (for example, AVO
modeling or Rock physics), the logs must be properly QC'd. Bad data
must be edited and if necessary replaced by data derived from
relationships, such as, Gardners or Greenberg (as described previously)
or from a neural network. It is important to engage the petrophysical
community in this process.
The Log conditioning workflow that is available in the Seismic-well
calibration group allows you to make adjustments to edited logs for
use in inversion and model building. This workflow offers tools to
perform operations such as blocking, depth shifting, interpolation,
smoothing, filtering, and trend calculation. This workflow and its
capabilities are discussed later.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 205
Procedure - Extract a deterministic wavelet
To run a simultaneous inversion, you must have a wavelet for each
angle range stack. In the Wavelet extraction toolbox, the
deterministic method is recommended. The Wavelet extraction
toolbox window is where you indicate the input data to use and set
the Verticalization type if necessary, the Taper parameters, the ion
parameters, and the desired outputs.
1. On the Quantitative Interpretation tab, in the Seismic-
well calibration group, click Wavelet extraction.

Seismic- ••ell calibrati-on

The Wavelet toolbox opens.


~--~~~--~-- r=-~~--~~~~--~.~ ~~~=z~~~·~------~
WIYtiel

!:'] .,... T11got


[I
[.j
Spocin_lnvPB_ful
Spodro_ISIS_Full
·I
~~
~cu
fl Rci<crl
fl Rlclaor2
1::1 Riclco<l

t:J ~-· .•r.~--bol<n_~


!:'] 1.\. \•llvoieis
Well
• • t:l:M ~ •
Soisn>c:
fl ... Wn•leisl<lo-_bol<n_~
[J 1/o. Vla>iOitiii.W•••I&JMl.,._~
D ... Vho.,le:s.'.'l......... _bofore_~
~ ln.lne. 3-'61
r:'J ~ We.,..le<s'Wawlels.FoMI_.U.
El l"l...,le!s,V/aYOieta_FoMI_!\11 Xino 2li8
e:J Vil>elo:a~'/o·.-_fi1lai.Aft lnlir! e o~~~ 10
., A.sed~"""'*
lelone 'Window
ACcalaklion.l'ndlcd
10 Al'roCiclailllly tnfo-

-
-t.t~AC»~
~
-
SioNt.,~
--
-
•ooo ....
23.10 ~2361
JS30) ~-
0.544
ooo-
3<!0

.
W.wlotdoooloy ~
s-. ..... 011_......c. ZltZ•o~ 2000 0•
Oporoocns s.t~CftRC 2021J O-
-~
~t1/-...._,.t 1100"""

I
P'-mon~tion · ~~ Power tC1KW:m in dB tC.tl
Phooo_, .. -~ ........
-..~
·1Sl 0
I
, t ~Wtt.;wt I

1 Wavelet list
2 Wavelet extraction settings
3 Wavelet displays
4 Predictability displays
There are several sections to the window. On the far left side is
a list of existing wave lets. By selecting the check box next to
several wavelets, you can make comparative plots of the
wavelets.
206 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
NOTE: It might be necessary to expand the display by clicking
the arrow on the far left of the window to expose the
Wavelets list.
Parameters

Length:

Sample rate:

Central frequency:

~ Nome
C [:, Target
D ' Spectra_lnvPS_Full [Anal~ ·I
C Spectra_ISIS_Full 1~ ·I
[£ ~ Ricker l
2] Alr.o-oo!culote
0 Rclsr2
C ~ Rid<er3
~-· k-----------------~
c Rid<er• 128 ms
L lo\. \."Javeiers\._,.Ja~et&_before_~
{l l.\. ':/avelets\\.'Ja.,-e&ets_befofe_l\ 2 m£
C 1.\. \•Jov..le!oi\•Jo....,_befc<e_~ ------
25 Hz
L 1.\. w.v.,le:s\Yiovelels _befc<e_~
E t ;"'"'"""'"'....,_Hnoi_All • O tnvettpolarily
[S \r-lel:elet51Ylavelet5_Ftn~I_Ait
E \'iovell!U\Wovei& _FNI_Aft
Operatioms
Phaoe menipul!ltion l Torne ohtll Honmng file<

-180 180
------------( r-- - -
Rot.te phase 0 00 ffi deg . R:-:-
1 .--; obie
-:-:to
- ....,
---:-ph...,
---,jl WaveleldisplayopticM - - - - - - - - - - - - -

§
Relo'lveemplitude
Phase: 0006 deg I Convert iD:zerophase I Po.....- opednsm m dB ocale
Phase opectnrm in WI"!Piled ongles
Show envelope

To the right of the Wavelets list is the wavelet extraction


settings area. In this area, all of the parameters for the
extraction are specified. Next to this area is where the
extracted wavelet and its power and phase spectrum are
displayed. The last area is where the Predictability maps are
displayed. These maps show slices through the predictability
attribute.
NOTE: The Predictability displays are present only when using
Deterministic extraction methods.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 207
2. To run a new extraction, select Create new, the Method
{Deterministic is recommended), and the Algorithm.
The available algorithms are Extended White, ISIS frequency,
and ISIS time.

@ Creote new: Extended White 1

~ Extended White 1
Method:
-----·-..
Deterministic
-
Algorithm: Extended White
-- ____ _:j
~ Auto-ca lculate

3. On the General tab, complete these steps:


a. Specify the Well and the Seismic volume.
b. Change the inline and cross line location of the well and
adjust the window around the well where wavelets and
the predictability are calculated by entering values for
In line, Xline, lnline window, and Xline window.
c. Select the RC calculation method.
Because you are interested in wavelets for angle stacks,
you must calculate the reflectivity appropriate to the
angles in the data volume that you are working with. So,
select Aki and Richards PP.
d. Select the well logs for the reflectivity calculation from the
lists and enter the angle range values for the seismic data.
The Effective angle is calculated automatically from the
Min angle and Max angle values specified.

208 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
General

Welt

Seismic:

TOR from : Jupiter-1

D Use SWT Study temporary TDR

D Allow wells outside survey

Inline: 3461 ---~o~--~


Xline: 2378 ~--101--~

In line window: 10

Xfine window:
- ----
RC calculation method: Aki and Richards PP

I Sonic or vefocity: ~ AtpDT


Density: ~ \)gR~O~
Shear sonic: ~ lt 8 ~TS
~ Advenced settings
..._,
Effective angle: 0 deg

Min angle: 0 deg

Max angle: 0 deg

Smoothing factor. 0

Filtering options - - - - - - - - - - - - - - 1

~ Use anti-alias filter

4. Open the Verticalization tab.


This tab controls how the seismic along a deviated well is
interpolated for the extraction.

lnline radius:

Xline radius:

Time smoothing winoo.v.

Square weights:

Quantitative Interpretation Simu~taneous seismic inversion and the Petrel Property Model Builder • 209
5. Open the Taper tab.
This tab allows you to specify the length of the cross
correlation window to be used in the wavelet extraction and
how it is tapered.
NOTE: The actual position of the window is specified on the
Extract tab. You can change the default parameters by
selecting the User defined check box. A white noise level can
be adjusted.

Parameters - - - - - - - - - - - - - - - -
Length of extraction xcor:
Length of extraction xcor taper:
\'1/hite Noise - - - - - - - - - - - - - - - -
Percentage of white noise:

(Reset )

The definition of these parameters is shown graphically in this


figure.
= '-'-1=
I GentAI I Vttticaliulion l Taper !...,Elcnct Ou!put
~l_ _ __
l'ltameten - - - - - - - - - -
leng1h of~ xcor.
lenglh of extraction Xeot taper.

v~e~ -----------------­
length of extraction xcor taper Percentage of~ite noise:
IReset I
90ms
length of extraction xcor

248ms

210 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
The Length of extraction xcor value and the Length of
extraction xcor taper can have a big effect on the extracted
wavelet. Some examples are shown in this figure.
Extracted Wavelet

;_L..., . ...\-. --- _:.. ..LU'-'--


~ .61) -+o -<J w <e a, t:l
T.m• -

Length d extraction xcor: 500 ~of ex!radJon xcor: 500 l..er'4h of ex!radicrl xcor. 200
L~ of e:4radion xcor taper. 250
---- Lefl!jh of elllnlc!ion xcor taper: 100 Length of eltradJcn xc:ortaper. -100
-:-----
----

J
·300 ·200 100 200 300 10

~---

l~ngth of ~action ltCOr: :9a ms Len¢h o extractbn xcor: .68 ms


length o extraction xc~r ta~r. 99 ms l ength of extraction xcor taper. 30 ms

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 211
6. Open the Extract tab.

Stort (a): 202() ms

Length (b): 692 ms

End (c): 2712 ms

RC window scon - - - - - - - - - - - - -

Offset to center (d): 0 ms

Length (e): 4() ms

Center {a+d): 2020 ms

Start: 2000 ms

End: 2040 ms

Toper - - - - - - - - - - - - - - - -

Taper: None

Length: 10 i @ Percent
~-------------~
()Time

This tab has two main sections. The top section labeled RC
window contains two parameters.
• Start (a): The time at the top of the window from which
the reflection coefficient data is taken for the cross
correlation on the Taper tab. It is used to calculate the
Predictability parameter discussed earlier.
• Length (b): Length of the window. It is shown graphically in
this figure.

T
e

212 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
The second section is labeled RC window scan. This section
controls what portion of the seismic trace is matched with the
window already selected on the well log reflectivity to derive
the wavelet and calculate the predictability.
The seismic window is the same length as the reflectivity
window. It starts at a distance controlled by the parameter
Offset to center (d) above or below the reflectivity window.
The window then is progressively shifted up and down relative
to the reflectivity window by half the Length (e) parameter. For
each shift, a new wavelet and the predictabil ity are calculated.
It is possible to edit these parameters graphically in a Well
section window if one is open during the extraction process.
7. Open the Output tab.
This tab allows you to save the Reflectivity, Acoustic
impedance, and seismic Dephase operator if desired.
Click the save icons to the right of the parameters that you
want to save.
General I Verticaftzation I Taper I Extract I Output I' - - - - - - - - - - - ,
Output parameters
RC Reflectivity:
,-----------~
New reflectivity
kl
~

AI Acoustic impedance: New acoustic impedance


~
~ Dephase operator: New dephase operator g]

r
L

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 213
Predictability displays
The Wavelet section of the Wavelet toolbox dialog box graphically
shows the extracted wavelet with its frequency spectrum and phase. It
is self-explanatory.
The Predictability displays are on the far right side of the window.
Predictability is calculated from this formula.

In this equation:
• Xcor(t) is the cross correlation of the seismic window and the
reflectivity window
• Acor1 (t) is the auto correlation of the reflectivity
• Acor2(t) is the auto correlation of the seismic data
They are weighted automatically by a hard wired cosine weighting
function. The maximum lag in the correlations also is hard wired.
Predictability values can range from 0 to 100, where 1DO is a perfect
match.
A useful characteristic of predictability is that it is a measure of the
similarity of the underlying reflectivity. It is independent of the seismic
wavelet and fairly insensitive to amplitude scaling differences as well
as to wavelet phase uncertainty between the two time series.
Figure 10 shows an example of some predictability displays. There are
several predictability displays available.
,. Maximum predictability (side view)
~

Figure 10 Example of predictability displays in the Wavelet toolbox window

214 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Remember, predictability really is a cube. It (and corresponding
wavelets) is calculated on a grid of inline and crossline locations and
for a series of shifts between the reflectivity and the seismic (the
Extraction tab parameters).
The initial display in the process when first run shows an in line/
cross line slice (left) located at the shift value that gives the maximum
predictability (right). This point is marked by a black X. You can click in
any of the grid cells and the wavelet display updates to show the
wavelet calculated at that location.
Examples of wavelets from different locations in the grid have been
shown on the inline/crossline display on the left.

General extraction guidelines


These general extraction guidelines are based on experience:
• The extraction window (RC window length) should not be less
than 500 ms; otherwise, the wavelet is not representative.
• A good run of thumb window= 4* wavelet length.
• The wavelet length should be at least -80 to 80 ms long.
• If there is enough well data available, choose a window at or
near the key zone of interest in the inversion. However, you can
choose not to cut though the reservoir due to fluid effects on
the log data. If possible, try to avoid very large dominant
seismic events because they dominate the extraction.
• You can select a location that is not at the well head as the
best wavelet to use, but you should be consistent with the
other angle stacks and select the same location for all of them.
• You are looking for a wavelet that has most of the energy in
the central peak and not too much in the side lobes, although
there always is some. For deviated wells, it is recommended to
use the Verticalization option to match the seismic along the
borehole. However, wells with deviations of more than 45
degrees can give poorly matching wavelets. This is a result of
two effects:
• at high deviations the wireline tool no longer measures
one formation at a time at formation boundaries
• the seismic wavefield is traveling vertically, but we are no
longer measuring vertical properties in the borehole
Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 215
Figure 11 shows an example of a well behaved wavelet suite with
associated power and phase spectra.

Wavelet Power spectrum Phase spectrum

Figure 11 Example of wavelet suite

Having a good deterministic wavelet can help significantly in matching


the seismic and the synthetic as shown in Figure 12. An example of a
well tie with a deterministic wavelet compared with an analytical
Ricker wavelet is shown. The time ties are better and the amplitude
match is improved, which are critical for good inversion results.

i ~~~ 'lllll~m~~·. ;~~~.Ji. pj~ii~


.~ ~~~ ~·£4too~~

·~~·

~~ • J

... ~

·w
.....:
A, "
~ J

~~~ ·0 ) ~
..,
Time. (ms) Time (ms)

26 SA
9122/2014

Figure 12 Comparison of deterministic wavelet and analytical Ricker wavelet


for a well tie
216 • Stmultaneous setsmic inverston and the Petrel Property Model BUilder Ouant1tat1ve Interpretation
Exercise 1 - Extract a deterministic wavelet
This objective of this exercise is for you to gain famil"iarity with
deterministic wavelet extraction and with the parameters and
diagnostics involved.
1. Use "Procedure- Extract a deterministic wavelet", the well
data, and the NRM seismic data -angle volumes to extract
some sample wavelets. You can use the same input data as in
this procedure.
2. Compare your extractions to those supplied in the Input pane
in the Wavelets folder.

o-~
--~ -
~~r-TI
ffi=-\-1
~
o.o.m...oc

=--- .
!t:!t:j
w.. . --
TCR-

I·-
""""-
._. U..&o'rs..dylltrll!>O'Ot!'TOR
t:....o

><61
Malt .... euiMe...,.,

)
@~
l1

AC-.-..- -6io!'>Y-- - - -
71-"""'""'"..
!«end-"!'
_0(,_, @) IJ,OT 0-..-.... da"""'
Oonocy ~ 'l. AH06 [] ---"'"'""~*~
D S~-.­
onglos

~ ll,ors
ill_,_,.,.._

3. Experiment with the parameters to see how close you can


come and what parameters affect the comparison.
NOTE: You can see plots of those wavelets by accessing their
Settings dialog box, or, alternatively, you can use the
functionality described at the beginning of "Procedure-
Extract a deterministic wavelet" to make comparison plots.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 217
Align seismic volumes with non-rigid matching (NRM)
Seismic volumes must be aligned properly for the inversion program to
invert the data properly. Alignment assures that the seismic events that
represent layer boundaries are all located at the same time. If they are
not, the inversion associates incorrect amplitudes with those
boundaries and calculates incorrect properties for the layers above and
below.
NOTE: After the Seismic trace alignment, the wavelets must be
extracted from the new aligned angle stacks.
This procedure shows you how to run the Seismic trace alignment
process.
NOTE: The test run ran in 25 minutes on the cropped volumes.
1. On the Quantitative Interpretation tab, in the Seismic
conditioning group, click Seismic trace alignment.
~=~ Seismic resampling
Seismic trace fS) S/N es':ir-1a~iol"l
~gnment · Wavele: eq~.oal i;;:a~lo r.
Seismic conditioning

The Seismic trace alignment dialog box opens.

Reference ~ismic:
D Output displaceMents
Qualitycheck - - - - - - - - - - - - - 1
[] Ou!put correlation su~
Surface time: ms

Corre~alion interval !._d_OO_ _ _ _ _ _ _ _ _ _ _: ms

I./ ~ply j (,_..r_ o_ __,l I~ Cancel

218 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
2. On the Settings tab, insert the seismic volumes located in
the Input pane that you want to align.
The volumes can be reordered and deleted, and the pane can
be reset using the buttons to the left. The 20 seismic and 30
cubes can be entered. Data must be in the time domain and in
the same survey.
NOTE: You can remove or change the order of the seismic
volumes in the seismic to be aligned fields, but keep ln mind
that the displacement fields are applied cumulatively based on
the order of the cubes or lines.
3. Select a cube to use as the reference cube from the list in the
Reference seismic field.
4. You can output the displacements, which are the relative shifts
between the data volumes. Select the Output displacements
check box to output these cubes.
They are stored in the same survey as the seismic data.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 219
5. As a check of the results, you also can output correlation
surfaces. These surfaces show the correlation of the datasets
relative to the reference seismic both before and after
alignment. Select the Output correlation surfaces check box
to write this output.
• The Surface time parameter specifies the correlation gate
center.
• The Correlation interval specifies the correlation shift to be
used.
TIP: A good practice is to set these parameters to encompass
the key zone of interest in the data. Correlation surfaces are
stored in an interpretation folder at the bottom of the data tree.
6. Open the Advanced tab. This tab contains parameters that
control filtering and smoothing of the displacement fields.
'I Seismic trace alignment ~ I
~~ A~aooro ~
~ ~-·n_~~~--------------------~
Parameters: - - - - - - - - - - - - - - - - - - - - - - - - - - -

High cut frequency. 30 tf1 Hz


Slope: dB/Octave

Amplitude level:

Lateral filter radius: traces

Vertical filter radius: ms

Time shift limit ms

AGC window length: ms

l" Apply I[ OK [ /(. Cancel

220 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Enter values for these parameters:
• High cut frequency and slope: These parameters control
the shape of a bandpass filter to be applied to the
displacement volumes to avoid excessive high frequency
jitter in the displacement values prior to applying them to
the data.
• Amplitude attenuation: The algorithm estimates the time
shift for each sample in each trace. Time shifts estimated
from weak reflectors are honored less than time shifts
coming from strong reflectors. This parameter specifies the
amplitude level where reflectors should be considered to
be weak. It often is preferable to run this method on data
with some sort of AGC applied (for example, a sliding
window AGC of 200 ms). With AGC applied, an Amplitude
attenuation equal to 1.0 usually gives robust estimates.
• Lateral and vertical filter radius: After each iteration, the
intermediate time shifts are filtered using a 30 mean filter.
This parameter specifies the lateral (or vertical) radius of
the mean filter in the inline and the cross line (or vertical)
direction.
• Time Shift Limit: This parameter is the maximum absolute
shift allowed up or down. Twenty milliseconds is a large
shift. Normally, 10 ms is more appropriate; otherwise,
cycle skipping can occur.
• AGC: Trace AGC (automatic gain control) scales the
instantaneous amplitude value with the normalized RMS
amplitude over the specified window. This default should
ALWAYS be changed to represent somewhere between
500-1000 ms.
7. When all parameters have been entered, click Apply or OK to
run the process.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 221
The program creates entries in the tree for the displacements,
the shifted volumes, and the diagnostics. The entries for the
data volumes include the inter-volume shifts, the cumulative
shifts, and the matched volumes. A portion of the tree with the
output data volumes is shown in this figure.
~ tl Mid_18_33 [MATCHED]
lnline 3458
XLine2383
Z=-2600.00
~ fl Far_31-46 [MATCHED]
lnline 3458
Xline2383
Z=-2600.00
,. fl! Near_5-20x Mid_18_33 [NRM]
lnline 3458
Xline 2383
Z=-2600.00
,. tiJ Mid_18_33xFar_3l-46[NRM]
lnline 3458
Xline2383
:J Z=-2600.00
~ ~ Mid_18_33 [Cumulative NRM]
lnline 3458
XLine2383
Z=-2600.00
,. ~ Far_31-46 [Cumulative NRM]
lnline3458
Xline2383
J Z=-2600.00
~ til Mid_18_33 [MATCHED]
In line 3458
Xline 2383
J Z=-2600.00
~ fll Far_31-46 {MATCHED]
lnline 3458
Xline 2383
Z=-2600.00

The correlation surfaces include correlations before and after


the application of the shifts. This figure shows an example of
how they appear in the tree. In this example, Jupiter[Near520]
is the reference volume.

222 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
~ ,. .,. 0 Seismic Trace Alignment Correlations
t> ~ ~ JD Hetp V;oL'51'on filers
~ I!! 0 Correlation Horizon Jupite([Near520 X Jupiter[Mid 1rn
~ t8 0 Survey 1
tl 0 TWT
~! 0 ~er"[Near520]x ~erfMd 1 833)
!>
~
'1) 0
0
...- ·-.
Seismic Trace Ahgnment Correlations
P. ~ ~ JD intetp inclusion £letS
A ~ 0 Correlation HoriZon Jupiter[Near520}xJupiter{Mid1&..1J U'IIA.TCHEDJ
A I:B 0 .Survey 1
tl 0 T'l\/T
{) t 0 .lqjter{Near520] X .ktpiler{Md1833] (MATCHED]
I> " 0 Seismic Trace Alignment Correlations

For displacement volumes, the correlation shifts generally are


small because the datasets are well aligned already.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 223
With correlation surfaces, again the differences between the
input data (top) and data after alignment are subtle because
the data already is well aligned.
Near to Mid correlation surfaces

Exercise 2 (Optional)- Align data cubes using seismic


trace alignment
In this exercise, you apply the Seismic trace alignment workflow using
the Jupiter raw cube and check the diagnostics from the outputs.
1. Run the Seismic trace alignment workflow using the raw data
cubes (Jupiter raw).
2. Select Output displacements and Output correlation
surfaces for diagnostics.
3. Visualize the diagnostics and the data cubes.
Do the results look reasonable and useable?

224 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Procedure - Condition log data with the Log conditioning
tool
Log data that is used in the inversion process must be in good condition
and properly edited so that you can rely on it for use in the inversion
process. Quantitative interpretation offers log conditioning tools that
can help you prepare the logs for use in the inversion process.
NOTE: Log conditioning is part of the Petrel Geoscience core license.
The Ql Extension enables some of the specific operations.
This procedure shows you how to use the Log conditioning tool.
1. Open a Well section window and make it the active
window.
2. On the Quantitative Interpretation tab, in the Seismic-
well calibration group, click Log conditioning.
- ._f.. W~atraction
Seismic If~ too conditionfng I
well t~ J.JI Well tie editing
Sei.smlc-well alibration

The Log conditioning toolbox opens. This toolbox contains a


number of functions that you can apply to log data.
T1,.: v. . ·. . ,,.,..

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 225
3. Click the Selections tool on the left side of the tool bar.
This tool allows you to highlight a region of the log that you
want to work on by drawing a box around it in the Well
section window.

Start End Log Well Enable Delete

The gray area in the log track in this figure shows the selection
made.
...---- - '- ..... --r - - ..~---- -- ....- ...
·
... ~··

226 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
A list of log curves appears below the toolbox.
NOTE: Some processes limit how many logs can be processed
simultaneously.
4. Scroll to the right side of the list of curves to expose the check
boxes.
5. In the Enable column select the Ia u want to condition.
End log Well Enable Delete
16n.06 1888.07 Vs Jupiter-1 til
1710.09 1889.91 Acoustic impedance Jupiter-1 D
2152.29 2328.44 Vp Jupiter-1 0
SSTVD (m) 1433.03 1563.30 RHOS Jupiter-1

6. In the Log conditioning toolbox, select the button for the


process that you want to apply to the selection.
A panel appears with the parameters that control the process.
In this figure, the Eliminate spikes operation has been selected.

Number of standard deviation: 1.38


-o--
~--~o samples

m
7. To change the parameters for the process, use the slider bars.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder •12.7
The Well section window updates interactively to show the
impact of the selections. The log display in this figure shows
where spikes to be eliminated have been identified.

8. When the parameter selection meets your criteria, click


Eliminate spikes to apply the selection.
If the parameter selection does not meet your criteria, click the
undo arrow in the tool bar and change the parameter selection.
When the result meets your criteria, click the Save button to
save the result.
9. To condition the logs as desired, perform as many operations
as necessa .
228 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Exercise 3 (Optional)- Use the Log conditioning toolbox
This exercise gives you the opportunity to work with the Log
conditioning toolbox and become familiar it.
1. Display a few of the Jupiter well logs in a Well section
window.
2. To activate the Log conditioning toolbox, click the Log
conditioning button.
3. Select a region of interest in a log and select the process to
apply.
4. Adjust the parameters and apply the process.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 229
Lesson 3- Low frequency model building and QC
with the Inversion property builder
The low frequency model is used in the inversion process to provide the
low frequency part of the properties of the earth. It fills the region of
the frequency spectrum that lies below the region contained in the
seismic data. Figure 13 shows an example of a low frequency model.

Figure 13 Example of a low frequency model

This figure illustrates these components of the model: the surfaces used
to build the LFM model, the wells used, and in the background, the
resulting LFM property model.
This example gives an idea of the scales involved. You can compare the
log scale property values that are highly detailed against the structural
surfaces at a normal seismic interpretation scale and, by comparison,
the low frequency component of the property.
The point is that a model with frequencies from DC to around 8Hz is
very smooth and devoid of most of the detail you are used to seeing in
either seismic or well log displays. However, it does capture the key
trends that allow you to convert relative changes in quantities in the
seismic bandwidth to ·absolute values.

230 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
The Inversion property builder is the tool you use to create low
frequency models for simultaneous inversion. The Inversion property
builder can be used to perform these tasks:
• Interpolate well logs into a 2D line or 3D cube, creating a
model suitable to use as the initial model for simultaneous
inversion.
• Extrapolate logs throughout the volume in a horizon consistent
manner. Where multiple logs are present, the log data is
weighted log data. The default weighting is inversely
proportional to the square of the distance from the well.
• Create a Guide model (typically, a velocity cube) to add spatial
information between wells .
• Specify a Depth trend.
There are four basic steps to building a model:
1. If necessary, convert seismic interpretations to surfaces.
2. Populate the model case.
a. Name the model.
b. Provide geometry from the seismic cube and, optionally,
the guide cube.
c. Add Global well log properties.
d. Add wells to use and define the weighting if you are using
multiple wells. ·
e. Add surfaces and set the depositional mode.
f. Decide on advanced parameters for log extrapolation.
3. Check the quality of the input data.
• Blind well test
• Well weightings
4. Run the model.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 231
Procedure - Convert interpretations to surfaces
First, the seismic interpretations must be converted into surfaces, one
of the main components of the LFM.
1. On the Seismic Interpretation tab, in the Utilities group,
click Make surface.

~ :il
:e
Polygon
ru
editing il

The Make/edit surface dialog box opens.

Makesumc~

~~: ~
~ ~~
~~=------~
M-. if1lU: ~ ~.-Itfl:_
.~ l_nts_p__, _ ___,! ~ ~ @ lnterp_1

~ o Name: --IHelp _1_ _~


--:....;;;.

~ Automatic from inJx.t dataiboU1dar(}


1Get a1 Rtm1gl han eeletted II Get his from lelected 1

Xmin: 759i4i.7R754 ~
y min: i83S478.2229 ~
X mcn: 764341.53778 'Ndh: 4593.7502450 ~ j91 &pend I
Y max: 7&50003.2229 Heqt: 14525 ~ [liiii stmk I
[{] Rot!tion: §B ·91 .143998

GOO~ ------------------------------
~ Xinc: 18.7500 Yinc: ~ j ~lodes: 24Gx582

Boundary
EJ 1\Aa<e boundar'/ from inplt and extend i 'Aih U nodes.
Note: If togOed on. the boundary in the inpU data wtl not be used.

0 Sa·1e computed bour.daryfor: l.__


a&_ed e. _______________,•j
.....:g'-

Iv' ~ II " OK

232 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
2. Open the Make surface tab.
3. Locate the horizon to convert in the Input pane and insert
in the Main input field.
~ ' 0 Stochalltic Model Make surface Hints
I> ~ 0 JDiretp inci.iYcn lms ~~--------------------~ n

I> ~ 0 Horizon4 i1put data: ,..,


Raeult
,.--
surf
_ace
_:______,
~ ~ 0 Horizon 5 Main 01put: ~ '-!t~-·l_nter_p__l -----'~ ~ . @ lnterp_l
!> ~ 0 Horizon6 I!Jl! LJ Ni!llle: 1'--
ht_..:..erp;::_
_l_----'
t> ~ 0 irtefJJ_2
I> ~ 0 lrterp_3
I> l{lc 0 lrterp_4
~>t& O ..

4. Open the Geometry tab.


5. Select the seismic volume in the tree that will form tbe grid for
the model.
6. Select User defined and click Get all settings from
selected.

Procedure - Populate the model case


This procedure shows you how to set up the model for a 30 case. The
20 case is covered later in the training.
1. On the Quantitative Interpretation tab, in the Seismic
inversion group, click Inversion property model.
• ·;.., '1 Inversion property builder
f.,. H I'K1
. It "'-If\ Inversion QC plot
Srmu aneous
inversion Stochastic inversion
Seismic inversion

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 233
The Inversion property builder dialog box opens.
This dialog box has several areas for parameter input. It also
has advanced settings. A tab at the top contains hints that
provide background information about the process.
[nversion property builder 1_1Ll
-~

Case editor Hints


l --
•L @ Create new: INew Name I
i/ 0 Edit exis1ing: ~I
t
0 P11rnmeter input Adv~:~nced I
I
·-· -
.,. Geomeby models
G "J
.,. ..6
I--
Seismic date defining geometry_ _ I --
Guide mode
- -----

_,
• GkJbalwell logs

"
G

Global well log


- -- - - - - - - -- --

.,.. Wells O --"~

..... 6
'Nell name
--- I Wetght ..._ Inverse dtstance exponent

...,. layers and horizons


I
.,. 6 . -
Top Honzon I Bottom HoriZon 1 Depositional mode
-- I Co~rm-;t;:
:i;i rop ~~ B~:~se ---

• Output filter
::::J
-
0 Apply a high cut filter to the output of: 8 Hz with a slope of: 30 dB/oct~:~ve

Runtime settings [ ij Ru lvJ [ •. Apply


II _ OK
II 1{ Cancel
i
2. Select Create new or Edit existing, depending on whether
you are creating a new model or editing an existing model. If
you create a new model, enter a model name in the field next
to the Create new option.

234 • Simultaneous se1sm1c mvers1on and the Petrel Property Model Bmlder Ouant1tat1ve InterpretatiOn
3. Open the Parameter input tab.
4. The geometry for the model is defined by an existing seismic
volume. In the Geometry models section, select the seismic
volume from the Input pane that represents the geometry to be
used for the model and click to insert it.
A volume can be removed using the button with the red arrow.
The folder symbol next to the blue arrow is used for 20 data
entry, which is covered later in the training.
• Geometry models

5. (Optional) To use a guide model (such as a velocity model),


enter it in the area to the right of the seismic data cube entry.
NOTE: If no guide model is used, vertical trends away from
well data are determined by the trends in the extrapolation
weighted well data. If the depth of an interval changes
significantly, there is nothing to adjust in the trend for that
depth change. By using a guide model, the trend in the model
away from well locations is adjusted by the local trend in the
guide model.
Specifically, before extrapolating the well data around the
model, the local trend from the guide model is removed at each
well location. These trend-removed logs are used to populate
the model through stretching and squeezing the properties
across the layers. Then the local trend from the guide model is
added back to the extrapolated properties at each location.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 235
This figure illustrates how using a guide model can influence
the model properties. There is a significant difference between
the depth trend of the acoustic impedance property between
the left and right side of the model. The difference in depth
trend across what is a large fault is the result of the
differences in the trend in the guide model.

6. Expand the Global wei/logs section of the window and click


to enter the global well log properties needed for the
model.
The properties that you can enter depend on the properties
selected to be inverted in the inversion process. To remove logs
from the Iist, use the red X.
..,. Globe! well logs

Globe! well log

236 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
7. Expand the Wells section of the window and click to enter
the wells that you want to use. Use the blue derrick symbol to
add all wells within a survey area. Wells that lie outside of the
seismic cube can be used as long as the input surfaces extend
through all the well locations.
... Wells r-
!_.,

Q~ ~
Well name I Weight Inverse distance exponent
+li Well-1 Ill 12
8. Specify the Weight to be given to the wells and the Inverse
distance exponent that controls how rapidly the weight
decays away from the well.
The weight applied to each well log is
W=k*(Wu /dn)
In this equation, Wu is the user-supplied Weight entered in
the well table, d is the distance away from the well, and n is
the Inverse distance exponent entered in the well table.
K is a correction factor to correct for cluster effects. You can
enable it on the Advanced tab.
For example, in this figure, without cluster weighting, the four
wells in the top left would receive equal weighting with the
well in the lower right. Because the four wells are close
together, though, their number tends to bias the estimation at
the blue location in the center toward the cluster. The Well
cluster weighting enabled correction attempts to
compensate for this effect.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 237
9. Expand the Layers and horizons section.
10. Select the first surface to add from the Input pane, and click
~ to insert the surface. Repeat for additional surfaces.
NOTE: Surfaces are most easily added in order.
These surfaces define the layers in the model. You can change
the order of surfaces and delete them using the buttons at the
top left of the section.

Top Horizon Bottom Horizon Trend [l£1Ls/m I ms}


Hcrizon4
Horizon S
Hccizon 6

NOTE: The surfaces must not overlap and should be entered so


that they are in order by depth, shallowest to deepest.
11. Specify the Depositional mode for intermediate layers.
The Depositional mode is fixed for the top and base layer.
Select one of these options:
• Offlap: The deposition follows the bottom horizon
• Conformable: The deposition follows both horizons
• On lap: The deposition follows the top horizon
In this figure, the diagram on the right shows what happens to
a log curved from Well A when it is translated away from the
well in an Offlap zone. There is no stretch and squeeze in this
case, and the log is just mathematically translated following
the bottom horizon.
If part of the translated log ends up above the top horizon, it is
truncated. When the translated log does not fill the zone
completely, the first sample in the log curve is copied upward
as a constant value (Offlap).

238 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
When a layer is set to Conformable, the log curve is stretched
and squeezed to fill the space between the layers as it is
moved away from the well.

The On lap case is similar to the Offlap case, but it follows the
top horizon. If part of the translated log ends up below the
base horizon, it is truncated. When the translated log does not
fill the zone completely, the first sample in the log curve is
copied downward as a constant value to the bottom boundary.
.. -= II -.

This figure shows an example using model data.

12. Specify depth trends.


The depth trend fields allow you to specify a depth trend that is
used to adjust the extrapolated log values for depth variations.
If no depth trend is specified, the log values between two
horizons are stretched and squeezed (subject to the
Depositional mode parameter) to fill the model between the
Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 239
horizons and copied to fill the model.
If some parts of the model are at very different depths (for
example, if the horizons cross a large fault), unrealistic values
can result. If a depth trend is specified, after the logs are
stretched and squeezed, the values are increased or decreased
according to the difference in depth between the model
location and the well. The depth trend is expressed as a delta
of the property value per vertical unit (millisecond, meter, or
foot).
If a guide model is used, the depth trend adjustment is made to
the extrapolated log values before the guide model trend is
added back.

NOTE: Trends must be calculated outside of the Simultaneous


inversion process. One workflow that you can use to calculate
trends is to use the well calculator to create a TWT log for the
desired property, and then use a function window to estimate
the trend by calculating a regression for the interval of interest.
This figure shows an example of a wedge layer with and
without depth trend compensation. As you can see on the right,
applying the depth trend progressively increases the property
value with depth at the bottom of the wedge as the wedge
expands and decreases it as it thins.

240 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
13. Expand the Output Filter section.
This section allows you to filter the model with a high cut filter.
Filtering the model allows you to limit frequency content of the
model to fit the frequency gap between the output of the
Inversion property builder and the low end of the seismic

r·0
~-
spectrum.
---~---··--
Ap~ly es~ igh ~ filter to the output of: 8 Hz with e slope of: 30 dB/octave

NOTE: You can use the Inspector tool to create an interactive


spectral plot of the frequency content of the seismic data.
14. Open the Advanced tab.
This tab allows you to control how the logs are extrapolated
above and below the first and last values.
I Pcremetet inputJ Adv!lnced L
ExtrapolatiOn options for 'overburden' and 'below TO':

~ Apply global 'overburden' extrepoletion settings to ell globe! well logs

~ Apply global 'below TO' extmpolation settings to ell globe I well logs

Gklblll well log: AI


·--------------------------~

Overburden [Giobaa - - - - - - - - - - - - - - - - - - - -

@ Consmnt using first log wlue


0 Con...~nl value of: G !<Pa.s/m

0 Uneer regression using window length of: 300 ms

0 linear value from first log value to 0 lcPa sfm at 0 ms

~m[~~q----------------------------------------

@ Constontusing last log value


0 Constant value at 0 kPa.s/m

(() Unecr regression using window length of: 300 ms

0 linear vakle tram last fog value to: 0 '<Pa sfm at 0 ms

SutVeywideoptions - - - - - - - - - - - - - - - - - - - -

~ WeD duster weighting enebled

The check boxes at the top allow you to specify if all logs are
treated the same regarding how their properties are handled.
Clearing these check boxes allows you to specify the settings
for each Global well log individually.
Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 241
This figure shows examples of some of these methods.
Default - constant User defined Constant
User defined :
firstllast values up 4700, down 7200 Regression extension
uno.5
~7T--r-~---r--,_--

2100

2200

2300

~+---~--H-~---+---

3100 3100

5. All the parameters now are set. Click the Run button to access
options that can help you to assess and check the quality of
the model before running it.

QC plot of w eightings Alh Q


Create model(s) Alt•R

Select one of these options:


• Blind well test: This option requires more than one well. It
computes the results at any well location using all other
wells in the survey. In this way, the computed properties
can be compared to the well log to see how well the
model is predicting properties.
The figure shows an example. The Green curve is the well
log, and the red curve is the model log. By comparing the
curves, you can adjust depth trends, inclusion of guide
models, and extrapolation of parameters, depositional
modes, and well weightings.

242 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
I I

II
L

• OC plot of weightings: This option produces a weight map


for each well showing, in map form, its weighting across
the model area. The results can be viewed in the 3D
window or in an Interpretation window. This figure
shows an example.

QC plot rJ weqtings Al+Q


Create model(s) Al+R

8 · :J lnmsion propetly buidef weilj'll maps


8 ::J lnvP8_30_Pre_Stack
zT O z
- ill 0 Wel-l
' iE O Wel-2
ill 0 \1/el-3

16. When the parameters are set, click Run> Create model(s).

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 243
The completed model appears as an entry in the Input pane,
with subentries for each property.
~ 0 Test Property Model
~ £I 0 Test Property Model AJ
0 ~3498
0 XUne2383
0 Z=-2600.00
~ fill 0 Test Property Model RHOB
0 ln1"1ne 3458
0 XUne 2383
0 Z=-2600.00
~ ill 0 Test Property Model VPVS
0 lnline 3458
0 XUne2383
0 Z-=-2600.00

Exercise 1 - Create a low frequency model using the


Inversion property builder
"Procedure- Populate the model case" described how to create a low
frequency model to use in a Simultaneous inversion run. Follow the
steps in that procedure to create a low frequency model that you use
later to run a Simultaneous inversion.
1. To open the Inversion property builder dialog box, click the
Inversion property builder button in the Seismic inversion
group.
2. To provide the Geometry, use the near angle NRM stack
volume. Leave the Guide model input blank.
3. Enter lp (AI), VpNs, and RhoB as the Global well logs.
4. Use the Jupiter well (Well1 ).
5. To build the layers, use the Surfaces _30 MEM surfaces
(Horizon 4, 5, and 6). Leave the depth trend input blank for the
moment.
6. To obtain the low frequency portion of the model, apply an 8Hz
high cut filter.
7. Accept the default advanced parameters.
8. Run the model.

244 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
9. Inspect the output in the 30 window (or Interpretation
window).

Procedure- Use a seismic Inversion QC plot


The Inversion QC plot functionality provides a useful way to visualize
and check the quality of the model results. Figure 14 shows an example.

Figure 14 Example of an Inversion QC plot

This procedure shows you how to create an Inversion QC plot.


1. On the Quantitative Interpretation tab, in the Seismic
inversion group, click Inversion QC plot.
Inversion property builder
ni ~v~rsion QC plot
1
. · •
S1multaneous -
inversion Stochastic inversion
Seismic inversion

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 245
The Seismic inversion QC plot dialog box opens.
I
[ttl Seismic irrmsion QC plot ~
!
1.-.ionQC F1 - -
[...Je o...new: lw.a..:tionwmdow J DBp~aJ--.,ror:
!/ OEdittiJOStlng: +
I• L""lld from .....son C4H! . Use defaults 1SMa dbpley Mlling ~I
v
... lrwwsian ,_..
u ...
Cuslom se.-ngs.

Rasle<izetion
_, dl ~ .
Plattyp. Tree:• colors
PnmeoyaJbe Globol-'l log Property lrend cube
I~·
I
O'•'llggle~a ~ -Gray

... w...
0Pcsll:vefill
O~fiO
1-Gray I~
D 1-Groy
H
·'
Wells :
0 v~rl8ble den511y
[il lntecpoletad dens•ty
0 PusiiJve wier ~U
Plot polarity
@NofT111!1
0 Reverse
- - --

QNegatJvecololfiD
... log liltw [J :J Lobe colorfiU
iD Apply lilar.
Numbercftm<:es. : 100 I
LaNcutfilter 0 ~ Hz •rlh e slope of
---
30.00 dB/octave
Decmollon specmg
,, I
Hgh<'.JI r 87 5 ~ Hz wlheslol"!of 3000 dB/oc::bYe
- Clip lunitsilraces)
'4 I
... Crostl«tton [J
... Scale
·~ lnltne () Crossline
@118dr81i~ j
• Track wiclh

SeismiclracJc 3.000 an log insert 1.000 an l oq triiCk:


[J
3.000 an
lnrerp01al100 type

~maxJlumiS).
,, I
-
... ElctriiC1ion Q
0 Alcng - bore trojl!doty
·@ VIIItJcalatwellheed~

OVertJCalar X: j o I Y: ~[Get From 1slweiiF 1


9> 'fji ~'
. Ceptu~ Ccp!Ure'M!Ils , !.I~~ L..... Apply ~~ [J< Cancei l

2. Select Create new or Edit existing, depending on whether


you want to create a new plot or edit an existing plot. If
creating a new plot, provide a name in the adjacent space.
3. In the Inversion results section, click to enter:

the model property cubes in both the Primary cube and
Property trend cube fields
• the associated Global well logs in the Global well log field
You can alter the order of the entries by using the up and down
arrows and delete them using the red arrow button.
4. In the Wells section, enter the wells used in the model.
5. In the Log filter section, enter values consistent with the high
cut filter to be applied to the model when created. If you want
to see the unfiltered model, leave this section blank.
6. In the Cross section area, select lnline or Crossline for the
display.
. .
246 • Simultaneous se1sm1c 1nvers1on and the Petrel Property Model Builder Ouant1tat1ve Interpretation
7. In the Track width section, adjust the scales.
8. On the right side of the dialog box, in the Display settings for
section, ·select the settings that control the appearance of the
display in the Well section window. The settings in this
pane apply to the property selected in the Inversion results
section on the left.
• To change the Well section window setting for a
property, select a property.
• Adjust the seismic display style by selecting the Custom
settings option.
Additional areas allow you to customize the Rasterization,
Scale, and Verticalization, similar to those in the Well section
window settings.
This figure shows an example of a populated Seismic

•l.h
/
lnvenoon QC
inversion QC plot dialog box.
!~ s~ismic inversion QC plot
~
Cn!ate r.ew:
EdrteXJstJng:
IWell section window
+I
~~--b:
Teet p-operty RHOB
- ..,
llfil tmd from ""'STlllm C>S3
@t!oedofaults. Deneily
0 Custcm lllltir.gs
,.. lnversJOn results
t:J ,;
• Rasierizaticn
-~ nl &fl r.
Pnmlll)'albe

~~ TeslpropertyAI I ~llpAI
Globe! well lag

I +I
Propertyhndalbe
J
I '..d.
Plot type
c '.'f<J'Ja 11103
QPo!!ll\'e~
,.Gray
Tracecolol's

i•GtBr
I·'I
j.lI
+rJrestpropertyVPVS I +fHVPVs I..,, I ~ 0 ~tiveliU
I•GtBr ... I
0 Vanable dens y
I
-
+I Testp~RHOB I ti~RHOO 1+1 l rtf
~~~~de'!Sty Plctpoll!rrty
..-Wells

·' - (i il 0 ?os:tM;aJlorfil
Q \aga:Nec.olot J
::::J tecc'Orfill
Q
>;o.·rtnl
0\'HS&

Wells
-
J
-~

~I!Well-1
.-Logfiller
~of!lliCE:s 1100 I
~nsp!IOI'g ll J
0 Apply fillac:
O•o lrlils\tmcm) 14 I
Lawcut~~er

H1ghcut . iLor 375 :


o; Hz

1-'.z
wrtlla~ot.

Wltl': a siopeot.
- -- ~M-·
:JU.OO dll,'ochve
I .. Salle

..- Cross 3edion


Gi
l~rype. Ou&i"ll:ic ~ l
l~lnJine () CrossliM
Scale(ma7,1lim1ts - ll I
,.. Trackwdh E) jl
Selsmiclmdc 3.000 an logtnsart: 1 tnl an loglmdc 3.0011 an '-"
- - -
.. Extmc:ban G)
Io Along wea bore tsajedory .. a, i' lli
• Captureal 1 :• Cap!tvewe ,t\ Export ' I ~~q~y J ~ (x esa~ j

Ouant1tat1ve Interpretation Srmultaneous sersmrc mversron and the Petrel Property Model Bwlder • 247
9. To create the Well section window display, click Apply or
OK This figure shows an example.

1 AI
2 Rho
3 VpNs

Exercise 2- Create a seismic inversion QC plot


Use "Procedure- Use a seismic Inversion OC plot" to display an
Inversion QC plot of the model and filtered logs.

248 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Lesson 4- Simultaneous inversion and results
Now that the seismic volumes are aligned, the logs are conditioned,
and the low frequency model is in place, you are ready to run the
inversion. This lesson contains the procedures that show you how to
set up and run a Simultaneous inversion.

Procedure - Run a Simultaneous inversion


The Simultaneous seismic inversion dialog box allows access to
several inversion methods. The seismic volumes and their wavelets
(which can vary with time) and the LFMs and their associated
parameters are entered here as input.
1. On the Quantitative Interpretation tab, in the Seismic
inversion group, click Simultaneous inversion.
"UJ ' . · f
lnversion property builder
..• lnversion QC plot
Simultaneous
inversion S ochaS'tic Inversion
Seismic inversion

The Simultaneous seismic inversion case setup dialog


box opens.
2. Open the Inversion case setup tab.

•O' Oeate new: l5m.JltZ!neous Inversion

I) Edit trAlstilg:
Method: IPost stack, normal incidence ·acoustic impedance
Surve-1: i 30
20 ~·~---------------------------

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 249
3. Select Create new or Edit existing. If creating a new case,
enter a name for the case in the adjacent field. Select an
inversion Method from the list.
IPost st~ck. normal incidence- acoustic impedance
Post stack. normal incidence -acoustic imped~ce
Tme lapse post stack. normal incidence- baseline acoustic impedance and monitor ratio of change
PP AVO. constant coefficients -acoustic impedance and shear imped~
PP AVO. constant coefficients- acoustic i~ and l Ns
lijill\'ltW-'D~t;,l$¥f·%1Mit4,,,J.g.&~DJM·'
; PP AVO Tme l~pse. Aki &Richards- acoustic impedance. VpNs and density baseline and monitor ratio of change
PS AVO . .Aki & Richards - shear impedance and density
w
PP AVO . .Aki & Richards- acoustic impedance. shear impedance and density
PP AVO . Fatti - acoustic impedance and VpNs
PP AVO . Fatti - acoustic imped~nce and shear impedance
• PP AVO . .Aki &Ridlards- acoustic impedance. VpNs. density and angle model update
PP & PS AVO. Aki & Richards - acoustic impedance. shear impedance and density
PP AVOZ. Aki & Richards - acoustic impedance. slow shear impedance. fast shear impedance. azimuth of fast shear impedance. density

The types of methods can be divided into poststack and


prestack.
• Poststack includes
• Inversion for acoustic impedance
• Inversion for time lapse properties
• Prestack includes inversion of
• Pwave to Pwave AVO data for impedances, Vp/Vs, and
density
• Pto S wave converted data for impedance and density
• Simultaneous PP and PS data for impedances and density
• Pwave to Pwave AVOZ
NOTE: Angle model update attempts to solve for the listed
properties and invert for the optimal angles for the seismic
data. The updated angles are an additional output. Because of
this additional output, the inversion takes longer to run (but is
the most accurate).
There are several different approximations used, but they are
hard wired according to what method is the most appropriate
for the desired output.
4. For Survey, select 2D or 3D and click to enter the parent
folder that contains the data from the Input pane. In the
General settings section, open the Input tab.
5. To enter the aligned seismic volumes in the Seismic stack
column and the associated wavelets in the Wavelet column,
click

250 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
6. Enter the value for the SNR parameter.
Remember from the introduction that the cost function contains
a number of terms and that the first term in this formula
measures the match between the seismic data and the seismic
predicted from the inversion model (refer to Table 1).
The SNR parameter controls the weighting of that term and, in
turn, how close the match needs to be. If S/N is set very high
(that is, there is very little noise in the seismic), the inversion
tries to match the data in much greater detail than if it is set at
a low value.
The parameter has a usual working range of 1 to 20 with a
good default being 5 +/- 1. It is not measured in db. It is set
separately for each input volume.
NOTE: Simultaneous inversions run quickly (in a matter of
minutes) on small volume such as the ones used in this class.
So it is possible to run a number of tests on selected
subvolumes to assess the impact of changing this parameter
and to achieve reasonable results.
7. Enter the Min angle and Max angle values for the stacks.
NOTE: If the names of the seismic volumes follow a specific
format convention, these values are populated automatically.
Stack volume names that contain the character'-' with
numbers before and after the character are parsed to strip out
the Min angle and Max angle ranges automatically for
insertion into the table.
The format used is xx.x-xx.x or xx-xx, so a floating point number
can be specified for angle ranges. For situations in which the
file name contains more than one'-' character, it is the LAST
occurrence of the '-' character in the file name that triggers this
behavior.
This figure shows an example of the populated table in this
section.
0

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 251
8. Select the check box Use time-varying wavelet.
Selecting this option enables two tabs that allow access to
two processes. One process is a workflow for Frequency
independent attenuation compensation, and the second
process is for Frequency dependent compensation. A third
tab accesses the ac plots generated by these two processes.
9. On the Frequency independent attenuation tab, select Use
frequency independent attenuation and enter the
parameters. This method is aimed at correcting an incorrect
gain function where the effect of transmission losses in the
seismic data is compensated for.
Frequency independent attenuation compares the time-domain
well log reflectivity data with co-located seismic data and
calculates any residual amplitude decay not present in the
earth reflectivity sequence. The type of well log reflectivity you
use must match the seismic data.
For example, for angle-stack seismic data, you should use a
matching angle-dependent reflectivity log curve, and the
reflectivity sequence must be in time.
For shear wave data, you should use a shear reflectivity log
curve. The time-domain well log reflectivity can be output
together with the extracted wavelet from the Seismic-well
Calibration process.
There are different algorithms available to perform the
estimation in the Attenuation approach list. The suggested
option is Constant energy, fixed zero time.
NOTE: For attenuation approaches that are not fixed zero time,
there can be a time-shift on the inversion output compared to
the input data. This shift arises if the energy midpoint of the
input wavelets is not equal to the zero time of the wavelet.
Such a shift occurs even if you do not select either attenuation
method, or you set both to 0 (zero). Therefore, ensure that the
extracted wavelets have energy midpoint at zero.
In the case of multiple input seismic angle stack datasets with
different angle ranges, best practices recommend that you
verify that similar values are estimated for the different angle
stacks in a dataset. A mean or median value can be used as
the user-specified attenuation value.

252 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
This figure shows an example of an input parameter set
Frequency independent attenuation.
/ (~ ~ altenuatlonii Frequency dependert altenJation 'j:ocJ
@ Use frequency lndepec~dent attenuation

Select seismic SiliciC INRM_NearCrop_5-20 1· 1


Time range ; §::=J :...........................................................................................................P ~ ms
-2000 -3200
QC plot filer length : 100 ~

Global reflectivity well l.Dg: ~~~~.=N=ew=~==-=~========================================~


Wells: I Weft-1

EsUnated attenuation: lo006143on I I Emae I


Ll User ~ed attenuation:
~~~oooo
~n=ooo============================================~l

This figure shows an example of the associated QC plot.


_,./ Frequency independent attenuation r
Frequency dependent attenuation l @§l

@ Frequenc-1 independent attenuation {) Frequency dependent attenuation

Estimated frequency independent attenuation ... NRM_NearCrop_S-20

2000
2080
2160
2240
2320
2400
Vi' 2480
E. 2560
~ 2640
I- 2720
2800
2880
2960
3040
3120
3200~----------~~~----------~~~~------------~~---------------~---------

Log(Seismic energy/Energy of reflectiv ity log)

NOTE: The tooltips provide a good summary of the process and


data needed.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 253
10. On the Frequency dependent attenuation tab, select Use
frequency dependent attenuation and enter the
parameters.
Frequency dependent attenuation estimates frequency decay
from the seismic data. The method of calculation assumes an
attenuation model where high frequencies from deep reflectors
are increasingly attenuated, resulting in a decrease in mean
frequency of the seismic data with depth.
The result of the model is that the seismic frequency
attenuation of the wavelet in dB is proportional to the product
of the frequency and the depth in TWT, or the estimation of the
attenuation factor (termed RATTA is based on estimating a
linear model between one over mean frequency and depth in
TWT.
Measured Decay (db)= RATTF*frequency*TWT
so that
RATTF =Measured decay/(frequency*TWT)
The factor RATIF is related to the Q factor by this equation:
rr
RATTF = Q
2

This figure shows an example of an input parameter set for


Frequency dependent attenuation.

"1 Use frequency dependent attenuation


Select se~smic st!lclc

-2000 -3200

QC plot filter length• 100

% oflroces: 1 000

Estmated al!enuallon: 0 011351608


0 Use specified al!enu!11Jon 0 OOOOOOCOO

254 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
This figure shows an example of the associated OC plot.
I Fraquencyindependentllttenuation I Frequencydependentllttenuation !oc
~----~----------------------------------
0 Frequency independent attenuation @ Frequency dependent lltlenuation

Estimated frequency dependent altenualion._ Near§20

';i
;;;

~2600

,2800 -+----1-----+----+----+--4--H~:...f-.::JIP

0.005 O.Ql 0.015 0.025 0.03 0.035 0.04 0.045


1/Mean Frequency Jsl

NOTE: The tooltips provide a good summary of the parameters


and process.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 255
This figure shows an example of the impact that attenuation
can have. Proper attenuation compensation resulted in an
improved estimation of the properties deep in the section as
you can see by the improved well tie.
No Attenuation With Attenuation

Extraction window{

11 . In the Select seismic stacks section, enter the Significant


reflector Reflection threshold.
This parameter controls the R1 parameter, which is the weight
of the significant reflector term in the cost function (refer to
Table 1).
When using the Reflectivity threshold parameter, keep these
points in mind:
• All points in the estimated property models with a
reflection coefficient greater than the threshold are
interpreted as significant reflectors.
• As R1 increases, the penalty for each significant reflector
increases and fewer significant reflectors are present in
the inverted property.
• If a reflection is larger than what is required by R1, the
inversion enters a layer interface. Also, a potential
significant property value change takes place over the
interface, thus avoiding smearing over boundaries.
• For poststack data, a value of 0.015 (1.5%) is a
recommended starting point and the usual parameter
range is 0 to 0.03.
• For AVO and time-lapse data, the unavoidable
256 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
misalignments of seismic usually make it necessary to
lower the number significantly, for example, to 0.00015 for
three input seismic data sets.
• For quality control, observe the output of the Relative
number of reflectors from the inversion case (double-click
the inversion case in the data tree after the run finishes
and open the QC tab). Bracket the reflection threshold up
and down until output Relative number of reflectors is
around 0.2 (20%), or until other factors visible in the
output datasets indicate the desired output quality.
12. In the Select low frequency models section, enter the low
frequency model, the terms for cost function weighting of
Horizontal continuity, and the Tie to the LFM.
The Property type field lists the properties required for the
inversion. These properties are determined by the inversion
method that you select. The inversion method defines the
parameters to be inverted.
To enter the low frequency model property volumes, select Use
cube and use the blue arrows in the Prior model cube/
constant column to select a model to use. These property
volumes are located under seismic in the Input pane .
.A g Test Property Model
• &I Test Property Model AI
lnline 3458
XUne23&3
Z=-2600.00
• ~ Test Propert}' Model RHOB
lnline 3458
XJJne 2383
Z=-2600.00
• &; Test Property Model VPVS
lnline 3458
XUne21&3
Z=-2>SOO.OO

Alternatively, you can enter values in the column. Constants


and property cubes can be intermixed.
... Select low frequency models
J
Property type IUse cube Prior model cube/consmnt Horizontal continuity T~eloLFM

P-imped8nce (kP8.s/m~ LiJ 9 tJ Test property AI 0.1 2 0.14 I

P/Svelocity ratio[] I
[lJ to? tl Test property VPVS 0.1 2 - 0.14
Oensity[g/cm3J [[] ~ I Test property RHOB [0.1 2
I
Q14
II
I
Ouant1tat1ve Interpretation Simultaneous se1sm1c mvers1on and the Petrel Property Model Bu1lder • 257
13. Enter the Horizontal continuity weight. Recall that this
weight affects the degree of horizontal continuity in the cost
function (refer to Table 1).
The horizontal continuity (Ra) parameter controls to what
degree horizontal variations in the property models are
penalized. Ra is the standard deviation of neighboring property
traces. The larger the value given to Ra, the smaller the
penalty for lateral changes. Also, the inversion optimization
algorithm imposes less horizontal continuity. A smaller value
increases the penalty for horizontal variations and so imposes
more lateral continuity.
A usual working range is 0.01 to 0.20. Lowering the number
yields an increasing amount of horizontal continuity.
NOTE: The value of Ra must be greater than 0 and smaller
than the value for the tie to the LFM.
14. Enter the lie to LFM weight.
The tie to the LFM (Rcr) parameter controls to what degree
deviation of the estimated property model from the LFM is
penalized (refer to Table 1). The greater the Rcr value, the
lesser the penalty, so the estimated acoustic or elastic property
model is allowed to deviate further from the associated low
frequency model.
A usual working range is 0.01 to 0.20. Lowering the number
increases the similarity to the LFM.
NOTE: The value of Rcr always must be greater than 0 and
larger than Ra. Although the default values are the same in
practice, the Ra values should be different from each property
type.
15. In the Specify the processing volume of interest section,
specify the VOl for the inversion.
There are two options that you can select based on how you
want to specify the VOl.
• To specify the VOl by IL and XL range, select Use inline,
crossline range. Select the ranges by entering values in
the associated fields or using the slider bars.
• To specify the VOl by polygon, select Use polygon and
click to enter the polygon. The time range also can be
adjusted.

258 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
This figure shows an example of this section if you select Use
inline, cross/ine range.

Use 1llline. crossf111e mnge 0 Use polygon

l't] lnline. 3355 3580 slap: 1


3355 3580

~----------------------------------~ ~~
"" Crossline: 2132 slep: 1

o----------------------------------1
2132 2634

Fll r me: ~ ·2000 ·3200


·3200
slap: 4 ms

This figure shows an example of this section if you select Use


polygon.
• Specify 1he processing vohme of interest

() Use m~ne. crossline renge @ Use polygon

~~oo: ~ ~------------------------~
lnline. begin: 3355 Md: 3580 step· '1

Crossline. begin: 2132 end: 2634 slap· 1

~ n ne: ·2000 ·3200 step: 4 ms


·2000 ·3200

Regardless of how the area and depth of interest is specified,


the inversion can take place only at locations where all inputs
(that is, seismic and LFM) are present.
NOTE: The inversion process automatically excludes traces
that are constant or all zero. However, there is no check for
zero values in a trace that is not all zeroes. Regions where
there is no data (such as, blank areas where there has been
undershooting, in the water column above the sea bottom, or
at longer travel times for larger angle stacks) are not excluded.
Including these blank zones in the inversion might cause poor
results because the inversion tries to fit the zero values as real
data. You must exclude these zones manually from the
inversion by adjusting the VOl.
If the IL, XL, and time range, or the area specified by a polygon
extends beyond where all inputs are present, the output is
truncated.
Similarly, if the trace spacing in the seismic is larger than in
the model, output is generated only where all traces are
present; no output is generated where traces are not present.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 259
The Run time settings verify that the inversion job can run (CPU
cores, memory/RAM, available disk space) when the job starts.
If you run other jobs at the same time as Simultaneous Seismic
Inversion, the processes might exhaust the CPU/RAM
resources or the other process might consume the available
disk space. Best practices recommend not to run other large
CPU, RAM, or disk-consuming jobs at the same time.
It is important to change the working directory to a suitable
location.
16. On the Output tab, select the desired output.
This tab specifies what output is written by the inversion and in
what format.
Select Same as input, SEG- Y, or ZGY as the format and
select the check boxes in the table for the output that you want
to create.
• The Output Property is the main inversion result. It
represents elastic properties in the specified unit as set by the
Petrel templates and in the grid and sample rate of the input
seismic. It is selected by default.
• Reflectivity is the reflectivity of each property calculated
sample by sample.
• For each input seismic dataset, you can select a set of
Synthetics and Residual seismic data sets as output.
Residual= Seismic- Synthetic.
~ lr~-~~d Output
Output form~l @ S~me cs ilput {) SEG·'r ZGY

IrMIISion result Output property


?-impEdance u
PIS velocity rabo _0_
Densty _j~]_

-~~---Se1-5m-icsl-----~:
Far)l 45
zlck -----R-es_Jdu-~l ----~.R~[ ----Syn__
I
lheli·cs ----li
260 • Simultaneous se1smic invers1on and the Petrel Property Model Bu1lder Ouant1tat1ve Interpretation
17. With all parameters and output now specified, click one of
these buttons.
• Run: If enabled, runs the inversion. If there are problems with
the input, the inversion does not run, but the case can be
saved to the tree.
• Apply: Applies the parameters and saves the case to the tree.
• OK: Saves the case and closes the dialog box.
• Cancel: Cancels changes made since the last time OK or
Apply were clicked. The window closes.
Visualization and quality check of the simultaneous
inversion output
When the inversion has run successfully, the output is stored in the
Input pane below the entry for the Inversion case. Several volumes are
placed in the tree, depending on the output selected. The volumes can
include property volumes, the associated reflectivities, and the
residuals for each of the properties inverted (Figure 15).
Test Property Model Simul Inversion
~ lJ 0 Test Property Model Simul lnvet'$ion
~ ~ 0 Test Property Model Simul Inversion P1mpedance
0 lnfine 3457
0 Xl..ine 2383
0 2=·2&04.00
I> fl 0 Test Property Model SiiU Inversion P-ftpedance Reftec:tivly
I> f!l 0 Test Property Model Simul lnvefflion P/ S velocity ratio
I> ~ 0 Test Property Model Simullnversion P/ S velocity ratio Reflectivity
t> t:J 0 Test Property Model Simul Inversion Density
C> f:l 0 Test Propert'f Model Simul Inversion Density Reflectivit'f
I> ~ 0 Test Property Model Smul lnversion NRM_Nea!Crop_5-20 Residual
I> ~ 0 Test Prcperty Model SinU lnYetsicn NRM_Neaerop_5-20 Synhetic
I> ~ 0 Test Property Model Smullnversion NRM_MidCrop_1S.33 ~idul!l
I> ~0 Test Propert'J Model Smullnversion NRM_MdCrop_1S.33 Synthetic
I> ~0 Test Property Model Simul Inversion NRM_FarCrop_31 -45 Residual
t> ~0 Test Property Model Simul Inversion NRM_FarCrop_3H5 Synthetic

Figure 15 Simultaneous inversion output in the Input pane

Each of these outputs can be viewed in the 30 window and


Interpretation window.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 261
Additional information about the run is captured on the QC tab of the
Inversion case. Double-click the inversion case in the data tree after the run
finishes and select the QC tab in the Settings dialog box.

~-
Settings for 'Inversion'
, ~·-·
- <: _,
- "''~ -- ·-···-~-" ,, .-. - ,- ~ .. - ·-· - l

0 Info I c- I Opeoation I ac 1People 1 a Plug1n 1


Lllst 8li8Clltion B!Drt del& ! 612£r& IS 3.l1.14 PM I
Last 8liBClltion end date: 16/261'2015 3 1725 PM I
Resuls

I Nllme
£1 New NGme Pimpedonce
I [J. New Neme PIS Ylllocity ratio
Cl New N~~tne o-ily

I
~
Slllb91K: I V11lue
Ralauv. mmfit (Ob) ·58+49173
Relative misfit c; En~) 26.0~ -
Relobve runber of Atfledcra ('t. ) ,6.4341507

1
~

Log ~le

Simultllneoua s81S1Tlic 1nversoon &ngme finiShed OK !#


=~~~~~ to li
hhmmss.!ISs- 151134.364
SPARAM: LOG_ALE filePeth•CWsers\COestrebecq\AppDaiii\L,ocai\Temp\SimSeislnv_l506261510344968\111Yi!f'SD1lag 1Xt
SPARAM: PROGRESS_FILE ~lePeiii.C:\Users\CD8Slr9llecqi,App0eb!\l,oc81\Temp\S•mSellllnv_1506261510344968\lnVV!IIOIIPrcgreeafie_IXt
SPARAM: DRYRUN• F
SPARAM: VERBOSE• F
SPARAM: TEMP_SPACE•5 174 1
SPARAM: OUTPUT_SPACE•51741
SPARAM: SAME_DISK• T
SPARAM: NCORES_MAX-4
SPARAM: RAM_MAX•13011
SPARAM: OPT_LEVEL•3
SPARAM: NDATA•l
SPARAM: DATA ID-1 filePath•C:\Users\COe6trebecq\AppOsb!\Lociii\Temp\SimSeislnv_l506261510344968\Da!ll_1\pllramfiletxt
~=:S~~c)p~~~~~lh-G.\Uwrs\CO.SnbecaiADoDallllloc:aHemo\SJmSeislnv 1506261510344968\Date 1\orowrlvfile.txl

Figure 16 The DC tab of the settings for a specific inversion case

This tab contains data about when the inversion was run and the
volume created and written as output. The Statistics section contains
statistics about the fit and the number of (significant) reflectors.
Additional data about the run is written to the Log File displayed in the
lower part of the window.
Observe the output of the Relative number of reflectors from the
inversion case. Bracket the reflection threshold up or down until the
output for Relative number of reflectors is around 0.2 (20%), or until
other factors visible in the output data sets indicate the desired output
quality.
262 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
The Relative misfit field is a global measure of the energy in the
residuals. Ideally, this value would be close to zero; however, the
seismic data contains noise and the inversion should leave this noise
behind in the residuals. To judge the Relative misfit parameter, you
must balance the value of the misfit number with a view of what is left
in the residuals. The question to ask is whether there is clear signal in
the residuals, or is just random noise left behind?
The Case tab contains the settings used to run the inversion. These
settings can be edited and the inversion rerun if desired .
• Settings for..,_,.,.

.o •m..J c- l ()pMlian 1 ac I P.,.. I B ~ 1

I >----,
j lnpJt Ouq,ut .

s.i""""stack SNR Min a~ IMX'B11gle WIVMI


+t'JNur_5-20 r4 5 20 + L'>- JP1 NoatRZ_5-20
. . {tMid_18_33 ::4 18 33
- +f· JP1_MidRZ_18-33
+l1Far_31_45 ,4 31
--
46 +t JPl_FarRZ_3 H 6

I Select laW "-""'Y modolo D


I
y

PtQ!letiYtype Uaecub6 Pnor modelcubolc:Gntlent HOfiZDIIIIII""'"'"IIIIy L r ooto LFM


. j 1>-impodonce[l<Pa.o/, . __ rii +-t'J-;-Test--:-p-rapeny___,.,AI------=----"-'---'===~(o.12 -- - 0 14- - -

I! PISwlocily ret.a(l :il +liT""P"'I*tYIIWS 10.12 0.1 4

~ ~-:=--~:~--2 ______________....__:-~~~: : - :~ I'

II -

Figure 17 The Case tab of the settings for a specific inversion case

Ouant1tat1ve Interpretation Simultaneous se1sm1c mvers1on and the Petrel Property Model Bu1lder • 263
The Run button is located on the Operation tab.

Unlock

Figure 18 The Operation tab used to rerun the case inversion

Earlier, you learned about the Seismic Inversion OC plot function, which
can be used to inspect the output properties and the logs at the well
locations.

Exercise 1 -Set up an inversion case and run a


simultaneous inversion
In this exercise, you set up a Simultaneous inversion case and run the
inversion. You become familiar with the input panels and options and
have the opportunity to change various parameters and observe the
results.
1.Use "Procedure- Run a Simultaneous inversion" to set up a
Simultaneous inversion case. Use (at least initially) this input
data for this exercise so you get reliable results:
• The three NRM angle data stacks as seismic input
volumes
• The LFM you created earlier in this module
• The three wavelets supplied for the near, mid, and far
angle datasets (that is, the wavelets listed as wavelets
after NRM and labeled JP1_NearRZ_5-20, JP1_
MidRZ_18-33, and JP1_FarRZ-31-46)
Later you can change parameters to see what effects they
have.
2. For the first run, select the check boxes to write out all
available output so you can have it available to visualize.
3. Visualize the output in the 30 window.
4. Create an Inversion QC plot for the output and see how the
inversion result compares to the log data at the well.
NOTE: The inversion should run in a few minutes.
264 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Results
Figure 19 through Figure 22 show the inversion result displayed in the
30 window. The input seismic is displayed with the property overlaid
with some transparency applied.

Figure 19 Results of a Simultaneous inversion. P impedance is at the top and


Vp/Vs at the bottom

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 265
Figure 20 and Figure 21 show the Inversion QC plot of the results. In
Figure 20, the low frequency data at the well location is displayed in
the center tracks of each seismic panel. In Figure 21, the detailed log
curves are displayed.
In each case, the log that appears to the right of the seismic displays
show the low frequency property curve (green), the inversion result
(blue), and the input log curve (red). The inversion clearly has captured
the key features of the seismic reflectivity and has honored the low
frequency trend from the model. Of interest is how well it has captured
the pay section. This section was absent from the low frequency model
and its inclusion in the inversion result is because of the seismic
contribution.

Figure 20 Inversion DC plots of the inverted data. The log tracks embedded in
the seismic show the low frequency model. The log that appears to the right of
the seismic displays show the low frequency property curve (green), the
inversion result (blue), and the input (red)

266 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Figure 21 Inversion DC plots of the inverted data. The log tracks embedded in
the seismic show full bandwidth log data. The log that appears to the right of
the seismic displays are the same as in Figure 20

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 267
In the Rock physics section, you created a crossplot of Vp/Vs plotted
against acoustic impedance. This crossplot is reproduced in Figure 22.
This plot shows that the pay section can be identified by where it is
plotted in this space. It is the isolated cloud in the lower left of the plot.

14
I• ft'"•l (Collocticn

r
Ill
Q. 12
~
u..
<
Q.

11
f
c..
10
.><
u
0
!;.

12 16 20 2.4 28 32 36 4.0 0 9

VpVs ratioStudy 1 [Rock physics] PAF

Figure 22 Crossplot showing VpNs plotted against acoustic impedance. The


properties in the plot were created in a rock physics study

To analyze the inversion result further, it is useful to create such a plot


from the inverted results. This plot should allow you to identify the
extent of the pay as characterized by the AI and Vp/Vs relationship. You
can create this plot by using the Geobody workflow.

268 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Procedure - Use the Geobody workflow
This procedure shows you how to extract a geobody from a selected
area in a crossplot.
1. On the Home tab, in the Perspective group, select the G&G
Perspective.

2. Activate a 30 window.
3. On the Seismic Interpretation tab, in the Volume
interpretation group, complete these steps:
a. Click Insert and select Box probe.

~-___.interpretatio n
1
Vo lume interpret ation r:;.

A Box probe appears in the 30 window.

b. Click Geobody interpretation.


The Geobody interpretation Tool Palette opens.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 269
4. In the Tool Palette, click the first button on the left to activate
the Manipulate probe function.
-::--= Tool PaJette 0 X
r!/ Manipulate probe
X

" I c?'- t. r::ji( ct<


Blue handles appear at the edges of the probe volume as
shown in this figure.

270 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
5. To contain the seismic volume, use the blue handles to resize
the probe.

6. Access the Settings dialog box for the Box probe by right-
clicking on the Box probe entry in the Input pane.

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 271
7. In the Settings dialog box, open the Volumes tab.
;i:J Settings for 'Box probe 1' LJL
8 Info l ~ooo Stmistics I People I Style I Volumes IOpacity I Extraction I
Seismic input
1st cube

~ t; Inversion P-impedonce
2nd cube
I ~
~ ,II; Inversion P/S velocity ~io I I.JW
3rdcube

~I I ·~ .
Co-rendering
Mode
ICROSSPLOT: Color: Color table l Opacity: Crossplot selection · IE:
Threshold
CL-- 7~21505 /~

Geobody masking
Geobody

~I I
Mode
IDisable/None ... ,

I Apply
II ./ OK II ~< Cancel I
B. In the Seismic input section, use the blue arrows to enter the
two volumes you want to crossplot and click Apply.
9. Open the Opacity tab and select Crossp/ot.

272 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
The data appears in the crossplot window as shown in this
figure.
stol

AI

l.8 2.0 2.2 2.6 2.8 3.0 3.2 3.4 3.6

10. To select a region of the crossplot to define the Geobody, use


the tools on the right side of the plot.

Selection -
@ Opacity

~
{) Color
10
~
9 •••
AI • •o
ools -

PoiU

Seal at LOD:

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 273
You can display additional volumes or co-visualization.

274 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Results
The result of selecting the region in the polygon in the plot in
"Procedure- Use the Geobody workflow" is shown in Figure 23.
Except for some outliers (acquisition artifacts), the Geobody nicely
isolated the area above the fluid level as seen in your data.

Figure 23 Geobody created by selecting an area of low acoustic impedance


and low VpNs in the crossplot window

Figure 24 and Figure 25 show some additional views.

Figure 24 Top down view of the geobody showing the main pay area in the
center and a smaller (unpenetrated} area with similar properties on the left

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 275
Figure 25 Geobody overlaid on a time slice through the data near the pay level;
note the excellent structural conformance

Except for the pay in the well, which you know about already, there is a
Geobody to the left of the volume downthrown to the main fault. Is this
a legitimate result and perhaps a new prospect? Figure 26 and Figure
27 show some seismic cuts through this area.

276 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Figure 26 In line sections through the unpenetrated anomaly highlighted in the
Geobody extraction

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 277
Figure 27 Crossline sections through the unpenetrated anomaly highlighted in
the Geobody extraction

The area of this second geobody clearly corresponds to a small fault


trap with hints of a flat event such as a weak version at the Jupiter fluid
level. Because this area is somewhat deeper than the well, you might
want to consider rerunning the inversion with some different
parameters to test with different models for the rock properties in that
area.

278 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Additional diagnostics
Additional diagnostics are available to check the quality of the
inversion. Specifically, the inversion outputs synthetic seismic; that is,
the seismic response calculated from the final model corresponding to
the angles associated with the input seismic volumes. It also outputs
the residuals, the difference between the real and synthetic volumes.
You can visualize these outputs in the 3D window and Interpretation
window. Figure 28 through Figure 31 show examples.

Figure 28 Synthetic seismic on the left and the actual seismic on the right;
there is a good match where the lines tie

1 Synthetic
2 Actual

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 279
Figure 29 Actual seismic in both traverses, which can be compared to the
synthetic section in Figure 28

280 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Figure 30 Actual seismic on the right and the residual seismic on the left, which
can be compared to Figure 28 and Figure 29

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 281
NOTE: The Inversion property builder and Simultaneous
inversion both work with 20 data, Simultaneous inversion
using the 20 output from the Inversion property builder.
In the Inversion property builder, the new column Top
horizon Marker appears with the 20 input and must be
defined.
You can visualize the model properties in the 30 window
(Figure 31 ).

1 AI
2 VpNs
3 Density
Figure 31 20 simultaneous inversion results

282 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
The Inversion QC plot tool also works for 20 data as shown
in Figure 32.

Figure 32 Seismic Inversion QC plot for 20 lines

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 283
Lesson 5- Data input for AVOZ inversions
Data loading for AVOZ follows a similar pattern to loading for 20 data.
Stacks for AVOZ inversions are loaded by highlighting the folder that
contains the data and clicking ¢.
Likewise, you can enter wavelets from a folder (Figure 33).
~ .._.___._._..__

· ..l
...]

,........ ...._ ~
lHR ~'H~Cf9! ..........
IV
l5l ""'
351 ~~
H.}.Ziw~-CICL1$NI. 7
!I_AZl_--~187-lV
H_;IZI___~IB7-ll7
lZ
12 "
;z v:·
+I. •..AZ·--~··-1-217
~~I
181 20.1 H,loi:5_;,,P.02.111·111.7
25.3 lU Mfo'l........_ro_;u=
- II Ol H_IZ<_,I"""'-Ql~:).ll.~

,.,
12
,.,
22 I H_Rl_,...._lnJ5..J»J

. ''"'
Jl
12
:!<.3
c
22
,:-,t ......_~!Ul51
to' I •.t-.-Jil.l53-l!.3
'~i

'
Hfll..-t.np;6._04_:r.:~
H_AU~£M_l!.c
H_Rl__ _ OUZ-'2 ·I

~
~
"'
;)17
lS.3 :l$1 I !Vi:~O<_l1.Q
c • !.: ·-~..IM.l!-<2
"' ';

--
.J
i~IOlft.t
f12
011

Figure 33 Selection of the associated wavelets

You can enter the low frequency models, azimuthal sector values, and
other parameters.
NOTE: For AVOZ cases, the Significant reflector constraint needs
special attention. For AVO and time-lapse data, it is necessary to lower
the number significantly, for example, to 0.00015 for three input seismic
data sets. For AVOZ it must be even smaller.
There also is an extra parameter relating to the threshold for computing
the anisotropy. Figure 34 shows the location of th is parameter and its
explanation.

284 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Figure 34 Minimum anisotropy cutoff parameter

Figure 35 through Figure 37 show some sample output from AVOZ


inversions.

Figure 35 Inversion DZ plots from a 20 inversion run

1 Acoustic Impedance
2 Density
3 Fast Shear
4 Slow Shear

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 285
Figure 36 3D window display showing the fast to slow shear ratio

Figure 37 3D window display showing the fast to slow shear ratio azimuth
vectors

286 • Simultaneous seismic inversion and the Petrel Property Model Builder Quantitative Interpretation
Review questions
• In what kinds of settings is simultaneous inversion appropriate
to use?
• What information is contained in the low frequency model?
Why is this information important to the inversion process?
• Why are good wavelets and well ties important to getting good
results?
• What kind of wavelet extraction process is recommended for
inversion?
• What is predictability?
• What kind of diagnostics can you visualize to check the quality
of the inversion process?

Summary
In this module, you learned about:
• basic theory regarding Simultaneous inversion and key
workflows, such as Wavelet extraction from Seismic-Well
Ca Iibration
• using the Inversion property builder to create a low
frequency model for use in simultaneous inversion
• setting up and running a Simultaneous inversion
• quality checking the output using the simultaneous inversion
diagnostics and the Inversion QC plot tool
• analyzing the output using crossplots and geobodies

Quantitative Interpretation Simultaneous seismic inversion and the Petrel Property Model Builder • 287
. . 5. - Stochastic
Module .
se1sm1c 1nvers1on

The Stochastic seismic inversion brings a solution to the resolution


issue in reservoir modeling due to the seismic resolution. By honoring
the well and the seismic data, it adds variogram parameters to provide
high frequencies to your model. Because this inversion is based on a
stochastic algorithms, the non-uniqueness of the process provides
multiple equi-probable realizations. These results then can be analyzed
and used to create, for example, porosity cubes from the acoustic
impedance.

Learning objectives
After completing this module, you will know:
• some of the differences between simultaneous and stochastic
inversions
• how stochastic inversions are used
• how a stochastic inversion works
• how to create a simple model for use in a stochastic inversion
• how to set up and run a Stochastic inversion
• how to check the quality of the output
• how to calculate properties from the inversion output
• how to create a reservoir AVO model from stochastic inversion
output, add random noise to the model, and visualize the
results

Quantitative Interpretation Stochastic seismic inversion • 289


Lesson 1 -What is Stochastic inversion?
The seismic data you use has limited spatial sampling and limited
frequency content. This limitation means that there are lateral
variations and vertical layers in the earth that the data cannot resolve.
Such fine scale details, especially in the vertical layering, are of great
interest when it comes to modeling how reservoirs perform when put in
production. Will the reservoir produce like a sandbox with uniform,
average reservoir properties, or will it exhibit discontinuities and
heterogeneities that impede fluid flow and reduce production and
ultimate recovery? Such questions cannot be answered from seismic
data alone.
In simultaneous inversion, you inverted the seismic data using a low
frequency earth model to create property models that honored both the
seismic and well data to within constraints. By contrast, in stochastic
inversion, you expand the inversion results to honor the wells and the
seismic, but you also include details beyond seismic resolution by
adding a statistical component to the inversion process. By adding this
component, you create a suite of equi-probable results that honor all
the known data and are consistent with the statistical variability in the
geology, as seen in the local well control.
To emphasize the difference between the seismic and reservoir model
scales, Figure 1 shows a seismic volume and a reservoir model
juxtaposed. The seismic scale clearly is much coarser.

290 • Stochastic seismic inversion Quantitative Interpretation


Figure 1 Contrast between the seismic scale (left) and the reservoir model
scale (right)

1 Seismic volume
2 Reservoir model

Quantitative Interpretation Stochastic seismic inversion • 291


In Figure 2, a stochastic inversion is compared with the result of a
simultaneous inversion similar to what you created in Module 4. The
added detail, which derives from the statistical component of the
inversion, is clear when compared to the simultaneous inversion.

Figure 2 Stochastic versus simultaneous inversion results. Note the difference


in scale

292 • Stochastic seismic inversion Quantitative Interpretation


Figure 3 illustrates the different parts of the frequency spectrum and
where data to fill those parts comes from in a stochastic inversion. As
in simultaneous inversion, the low frequency portion is filled with a
prior model that is constrained by well control. The frequency band
above the low frequency model is filled by the seismic data.
Beyond the seismic data, you are free to construct models with fine
detail if, when they are filtered to the seismic bandwidth, the result is
in agreement with the seismic. You use statistics based on the well
data to populate this high frequency region with statistically valid
realizations of the fine scale geology.

Frequency (Hz)

Figure 3 Frequency spectrum and how different data types fill the different
frequency regions

The schematic in Figure 3 shows:


• Low frequencies controlled by prior model (black)
• Intermediate frequencies controlled by bandwidth of seismic
amplitudes (gray)
• High frequencies controlled by vertical variogram model
(orange)
• Localized high frequency contribution around wells
This lesson first discusses the statistical element of the data input to
the inversion.

Quantitative Interpretation Stochastic seismic inversion • 293


Stochastic inversion background
The statistical component comes from the variogram that describes the
spatial variation in the model. The Sequential Gaussian simulation
algorithms populate the cells of the model based on this variogram
information and other inputs.

Variograms
It is intuitive that earth properties (for example, lithology, porosity, and
permeability) are more likely to be similar between two points if those
points are close to each other than if they are widely separated. The
degree to which things are dissimilar and the distance over which they
become so is captured in a statistical measure called the variogram.
The parameters used by stochastic inversion to perform the statistical
portion of the inversion property estimation come from a variogram
analysis of local well data.
Figure 4 shows an example of a variogram. On the horizontal axis is the
distance between samples and on the vertical axis is the variance, or
the degree of difference, and, therefore, the variance between
properties. At close distances, the difference usually is small. At large
distances, the difference usually is large. Therefore, the variogram rises
with increasing separation distance from a small value to a large value.

:lO 40 60 80 100
;:; ;:;
-ir Sample Variogram

Sill
~ C)
~
C)
~
~

~ C) C)

'~·"' t:>. t:>.

~ ,.
C) C)


...,
C)
~Nugget ...,
C)

~Range

20 40 60 80 100

Separation Distance

Figure 4 An example of a variogram and the associated terms used to denote


them

294 • Stochastic seismic inversion Quantitative Interpretation


The horizontal axis is the distance between sample locations (also
called the lag). The vertical axis is the semivariance defined as

1 lL
y(h) = 2IN(h) " ' (zi - zj) 2
N(h)

In this equation,
• N(h) is the number of samples that are separated by distance h
• Z.I and Z. are samples where i- j=h, which is the separation
J

distance
Variograms have their own special terminology.
• The value for the semivariance as the sample spacing approaches
zero (that is, how much things vary when near each other) is called
the nugget.
• As the distance increases, the semivariance rises until it plateaus.
The plateau value is called the sill.
• The distance at which the semivariance reaches the sill (or for
some analytic variogram models, 95% of the sill) is called the
range.
• A variogram plot from actual data points is called an experimental
or sample variogram.
• The fit of an analytic representation of the variogram to an
experimental variogram is called the variogram model.
There are several common types of variogram models: Gaussian,
exponential, and spherical variograms. The formulas are discussed
shortly, but their key characteristics are shown in the plot in Figure 5.

Quantitative Interpretation Stochastic seismic inversion • 295


0 200 400 600 800 1000 1200 1400
.....
~ II,)

+ Gaussian + Exponential + Spherical

~ 0
00
0

~
00

~-
~ 0 0
0. 0.
"""
~-
~ 0
:,.
0
:,.

0 0
II,) II,)

0
Range 0

0 200 400 600 800 1000 1200 1400

Separation Distance

Figure 5 Different variogram types and their characteristic shapes

Spherical is the simplest and most common variogram. The


Exponential has the largest semivariance at shorter lag distances. It
has an asymptotic approach to the sill at longer distances. Finally, the
Gaussian is reserved for phenomena that show a high degree of
continuity at short distances, then begin to transition to more of an
exponential behavior at longer distances. A Gaussian variogram gives a
more continuous look within shorter distances. The exponential
variogram gives more variation within shorter distances.
To compute a variogram, you must have a lot of samples with varying
separation. To assess the lateral variability in a property, say porosity or
something related to it such as acoustic impedance, you clearly need
multiple penetrations in your zones of interest. Some subsurface
properties have distinct lateral trends. A large number of penetrations
is needed to estimate properly the direction of such trends and their
statistics.

296 • Stochastic seismic inversion Quantitative Interpretation


Figure 6 shows an example of a map of the variance of a property with
a distinct spatial trend. Experience has shown that it can take up to 15
or more wells to assess properly the lateral statistics and identify and
quantify asymmetrical behavior.
-3200 ·2400 - 1100 ...., 1«10 2CIO )21110
a r-~~~~~~~~~~~~~ a

R i
! i

' -3200 -2CIO -1100 ...00


a
"'"'
toO
...............
1000 111100

1:50DCIO
1000 25IIOm

Figure 6 An example of a variogram map

It is important to note that working with fewer wells carries risks that
the results of a stochastic inversion will not model the distribution of
properties correctly within a reservoir. You must be aware of the
uncertainties in the results that are a consequence of being under
constrained.
In stochastic inversion, the key variogram parameters, the sill, nugget,
and range (and major and minor axes in the case where the distribution
is not uniform} are needed for the vertical and lateral dimensions of the
model. This information tells the inversion program how to predict
appropriate statistical variations in the inversion properties that are
tried as possible realizations of the property model.
Later in the training, you learn in more detail how to compute a
variogram in Petrel.

Quantitative Interpretation Stochastic seismic inversion • 297


Gaussian simulation
Then how does stochastic inversion use the statistics from the
variogram and the other inputs (the seismic and the prior model) to
create realizations of the subsurface? The process is called Sequential
Gaussian Simulation. These steps show an overview of the process:
1. Randomly pick an empty cell location in the model.
2. Use kriging to estimate the variance in the property value at
the cell location using the well data and other already
estimated cells.
3. Knowing the variance in the distribution from step 2, draw a
random value from the kriging estimate of the distribution.
4. Populate the cell and treat the cell value as an additional
control point to estimate other empty cells.
5. Repeat the process until the grid is populated.
Figure 7 shows the process schematically.

6
= :=== := ~ ~

fj,

*~ ~ ~ +- t---
I
@]Well location
0 Simulated cell
* Un-simulated cell

Figure 7 An example of Gaussian Simulation

298 • Stochastic seismic inversion Quantitative Interpretation


Stochastic inversion methodology
When the grid is populated, further optimization is required to ensure
that the model meets the constraint to which it must reproduce the
seismic data within a specific tolerance.
Following sequential simulation, these steps are taken:
1. Generate a trial AI model from the simulated grid.
2. Compute a reflectivity series and convolve with a wavelet.
3. Compute misfit against observed seismic.
4. Iteratively perturb the trial AI model.
5. Retain the best matching AI model.
6. Go to the next trace.
Figure 8 demonstrates this process.

N =Number of iterations

Synthetic •truth•
- Best fit after N iterations
R=Correlation coefficient between observed and synthetic seismic traces

Figure 8 Process for matching the seismic data in the stochastic inversion
process

To achieve multiple realizations, the Gaussian simulation is repeated to


produce different property models. These models then are optimized to
fit the seismic data, which produces multiple equally likely realizations
of the subsurface at a fine scale, all of which fit the well and seismic
data.

Quantitative Interpretation Stochastic seismic inversion • 299


Lesson 2- Stochastic inversion preparation
Figure 9 shows the inputs to the stochastic inversion workflow.

Figure 9 Different inputs and outputs of a stochastic inversion .

The overall workflow includes these steps:


1. Build the structural model (fault modeling, zone creation, and
layering).
2. Upscale the well log property logs into the structural model.
3. Build the prior model.
4. Analyze and model the variogram.
5. Run the inversion.
6. Check the quality of the inversion.
7. Convert to porosity and other applications using the Workflow
editor.
As with simultaneous inversion, you also must edit and condition logs,
establish well ties, and derive wavelets.
This section discusses building the prior model, upscaling the well logs,
and finalizing the prior model followed by a discussion of determining
the variogram parameters.
300 • Stochastic seismic inversion Quantitative Interpretation
Simple prior model building
The Stochastic inversion requires a prior model. This model must
contain average model properties for the grid cells and well log data
upscaled to the grid cell dimensions.
NOTE: The processing time for a stochastic inversion is significantly
longer than for a deterministic inversion. It is therefore recommended
to focus only on the reservoir zone for stochastic inversion.
The model can be a simple pillar grid or a complex structural model,
depending on the reservoir complexity. In this module, you build a
simple model to progress quickly to running the inversion, but keep in
mind that under different circumstances, a more complex model might
be required. Building models is covered in detail in other Petrel
courses.
The procedure to build a simple model includes these basic steps:
1. Create the grid (in the Seismic to simulation perspective, go
to the Structural modeling group and click Make simple
grid).
2. Add horizons (In the Corner point gridding group, click
Make horizons).
3. Create layers (In the Corner point gridding group, click
Layering).
4. Upscale logs to create trend logs (creates average properties
for each zone).
5. Make synthetic logs from upscaled average properties.
6. Create detailed layers (In the Corner point gridding group,
click Layering).
7. Upscale well log properties to model with the fine layers
created in Step 4 (In the Property modeling group, click
Scale up wei/logs).
8. Enter properties for the full model (In the Property modeling
group, click Petrophysical modeling).

Quantitative Interpretation Stochastic seismic inversion • 301


Procedure- Create a simple prior model for Stochastic
inversion
To create the prior model, you first must create a simple grid with a
thick layering to create trend (or low resolution) logs. Then, the layering
is refined and the logs are populated into the fine grid using a basic
extrapolation algorithm.
1. Create a simple grid.
a. On the Home tab, change the Perspective to Seismic to
simulation.

b. On the Structural modeling tab, in the Corner point


gridding group, click Simple grid.
1 Fzt.lt me del ;:.bje::t Horizons, 'f®
+~ Fa1...tt rr ode! opeHtior-& Zones fP
Sift SF to f3t..lt rrodel e Edit fault model
Pill3r
gri~dirg L3yeri ng tt1
Cornfr point grt.oldfng

302 • Stochastic seismic inversion Quantitative Interpretation


The Make simple grid with dialog box opens.
c. Select Create new and enter a name for the model.
d. (Optional) If you want to limit the inversion area with a
polygon, select a polygon in the Input pane and insert ~
it in the Boundary field.
e. On the Input data tab, select Insert surfaces.
f. Use the icons to insert the horizons to be used in the
model.

~ lnputdata 11m Tartan grid

EE..J :__. Skeleton only


~' Insert surfaces

(J] ~ Horizon 4 (Conforn>able)


rca1 @ Horizon 5 (Confowable)
~ @ Horizo 6 (Confom able)
l.§CI @ Horizon 6 • 150n'i5 (Conforn able)

~
~

Horizon type: C r.fcr•nd c ...-j


------------------------~

[ ,.t .A4>ply II OK j[ 1< Cancel

Quantitative Interpretation Stochastic seismic inversion • 303


g. On the Geometry tab, select a seismic volume in the
Input pane that will be used in the inversion and click Get
all settings from selected.

0 Edit existing (cmert active)


Bot..ndaty: ~ O_ s_ma
.-1 _l_lP-o1-ygoo
- . - - - -- - ----.

0 Automatic from inpli datalbou'ldary)


@ User defined: ( Get aa seti1gs from se~eaec~ I 1Get limls from :teleded 1

~
Xmin: 759735 .47720 @

Ymin : 7835487.8456 ~
X max: 764347.97747 Wiclh: 4612..5002£54 ~ [BII Expand I
y m&: 7848062.8455 Height: 12574.999906 ~ !mm Stwink 1

rlJ Rotation: §] -91_1439982~


Gridincrement - - - - - - -- - - - - - -- -

§] X inc: 18.7500 Yinc: 24.999S INodes: 247x 504

[-1 ~ J! ..t OK
h. Click Apply.
The model is stored in the tree in the Models pane.
2. Add horizons.
a. In the Corner point gridding group, click Horizons.
Define model Fa ult model o bj ect ...

m Simple grid ~ Fa It model ope ratiors


Pill3r
, SF to fa It model ~ Edit fault model grid ding ~ Dom ai cor versio
Comer point gridding

304 • Stochastic seismic inversion Quantitative Interpretation


The Make horizons with dialog box opens.
b. On the Horizons tab, the horizons entered on the Input
data tab in the Make simple grid dialog box are
displayed. Click Apply.

Horizons

Hints for the table:


-

Horizon type:
- .~

I.Settings I Faults I Segments I \"/ell adjustment I ?


Conform to :
- Ur :cicrr t)

Use horizon.fau~ lines: Q


.....
I Hints I
Input:

g]B]~~~~Ei] ~ ~ ~ ~[EJ
Use
Horizon Horizon Conform to another .Smooth horizon
Index StatllS Well tops Input #1
name type horizon iteJ"ations -fault
lines
r-
1 Horizon4 Conformable No 1 ../ Done 0 [l] Yes ~ ~ Horizon4
2 Horizon 5 Conformable No 1 ../ Done 0 ~ Yes ~ ~ Horizon5
3 Horiz on 6 Conformable No 1 ../ Done 0 [{] Yes ~ ~ Horizon 6
4 Horizon 6_ Conformable No 1 ./ Done 0 0 Yes ~ ~ Horizon S

3. Create layers
a. In the Corner point gridding group, click Layering.
,-
'fa
~
Define model · f3 ul model object • I Horizons Hori;:on cl ear -up

l3l Simple grid r~ Fault model op erations


PilLar
"""' Zones ~ Grid refinement
tft SF to l'at.lt IT odel . . Edit fault model g ri ddrr g ~~ Layering ~ Do mai co m.er5i or
Corner point gridding

The Layering with dialog box opens. -

~J

Common settings ---------------------------- ~


Build along : Along the pillars 0 fjorizons •,-.·ith steep slcpes
lSi{ D Use minimum cell thickness: o===J I ~ Include proportionallfractrons . start from·

Zone specific settings

~ Zone division: Reference surface : Restore eroded : Restore base:


~~
Reference Re.store Restore
Name Color Calculate Zone division .Status
surface eroded base
ProportionBI N rrber of l~yers : D Yes D Yes New
Proportional Number of layers : [] Yes D Yes New
Proportional Number of layers : [] Yes L] Yes New

b. Leave the number of layers as 1 per interval.


c. Click Apply.

Quantitative Interpretation Stochastic seismic inversion • 305


4. To create a model with zone average parameters, upscale
logs.; on the Property modeling tab in the Data preparation
group, click Well log upscaling.

Well log
lki
Data
S. ""'rai ir.g in age

upscaling analysis ~ G~ometriol rend


Data pr~paration

lltll Show red in well section DJ J

Ecflt e:osting:
@ OverNrile
Replace and add new
0 Leave an other cells unchanged

Well logs

0
o: Poirts attrble

Sdect log I p AJ

[@ Set!ings
LJ Usebias - - - - - --
l.Jpscaled property: .... ,

l.Jpscaled from : None


Scale~ settings - - - - - - ~B--f:2]
-
..- -\•-
/e-
lls-~

Average method: Arithmetic ..• ! c::_j 1.•/ell-1

Treat log : A-s points ~ \•iell-2


-----~=1
Method: Neighbor eel ...

LJ Use facies wetghting


(J ~ - runber d poirt.s in eel:

(J Zone correction
rupw
0 WeB report
0 Reset output sheet
0 Zone corrected log
0 30 grid z.one log
0 30 gnd well tops

l ..t ~ ll..t oK

306 • Stochastic seismic inversion Quantitative Interpretation


c. Select Create new.
d. In the Input section, select the log to upscale from the
Select log list. In this procedure, three properties that will
be inverted are upscaled: AI, Vp/Vs, and Rho. ·
e. To select specific wells, click None in the Wells field and
select the wells you want to use in the tree.
f. Click Apply.
g. Repeat these steps for the remaining properties. Be sure to
select Create new each time and select the wells if
necessary.
Entries for the upscaled properties appear under the model
in the Models pane.
~ 0 OassGrid
I> ~ 0 Sksle/oo
Fat.b
::J Honzons
0 Edges
0Hetseeiions
0~
lp 0AI [UJ
p0 RHOB(U]
~ 0 VPVS [U]

h. Rename these properties so that they are not confused


with other properties later on. For example, you can
rename them as AI trend [U], RHOB trend [U], and VPVS
trend [U]. To rename these properties, go to the Info tab in
the Settings dialog box for each property. These
properties are the average zone properties for the layers
for the inversion.
~
I>
= OassGrid
E!lli 0Skt!lt!fon
Fauts
p. =::J HoriZOI7$
0 Edge$
I> 0 Intersections
~ 0 PmpttttNJs
lp 0 Al trend (UJ
p 0 RHOB trend [U]
-Ji 0 VPVS trend [U]

Quantitative Interpretation Stochastic seismic inversion • 307


5. Make synthetic logs from upscaled average properties.
a. Open the Settings dialog box for the well whose
properties you are upscaling.
b. Open the Make Logs tab and the From property subtab.
>{~ ~7:.~ I- x __

18 Info · lr® Settings I Staij,.tj.,.


'lr'i\ Operations

tl Tune I Report II Make logs


II Qual ill/ attributes

IJ From properl'f II: From zones. II- From other logs I


~ Properties: [J
8 . 0 Properties
f· lp 0 AJ trend [U]
·· p 0 RHOB trend [U}
~0 VPI/S trend [U}
I
I

[J ~er.Vrite existing synthetic well log !..; Make logs


I
1../ PwiY II./ OK 11 1( C5lcel l ,,

c. Select the logs and click Make logs.


6. Add detailed layering into the model.
a. On the Structural modeling tab, in the Corner point
gridding group, and click Layering.
De11ne model Fat.lt model objed Horizons 'f~ Hori=oncle~ n-tJp

liJ Sim ple g rid ·~ F3 It model oper:'ltions Zones Ef1 Grid refinement
""' 5F to b It mod el ~ Edit fa ult mod el Layerin;;~ ~ Do mai <Otl\iersior.
Comer po int gridd ing

308 • Stochastic seismic inversion Quantitative Interpretation


The Layering with dialog box opens.
~~---,~~~--
=layering with 'Comer_Point_Models/Ciass

1\.lake layers

Common settings ---------------------------------~

Build along : Along the pillars 0 Honzons •Nith steep slopes


Ei( LJ Use minimum cell thickness: II:==J 0 Include proportona!Jfrac!ions . start from .

Zone specific settings

~ Zone division : Reference suface:

Reference Restore
Name Color Calculate Zone division
.surface eroded
L] Yes

L] Yes

7.

Vv'elllog 03) - r::r:c.1


upscaling ::~r:a lysis rr od elir
Data preparation

Quantitative Interpretation Stochastic seismic inversion • 309


The Scale up well logs with dialog box opens.
b. Select Create new.
c. In the Input section, make sure that Well logs is selected.
d. Select the well log to upscale from the Select log list.
e. To select specific wells, click None in the Wells field and
select the wells you want to use in the tree.
f. Click Apply.
Re eat these ste s for each lo
·-
Make propert>/
I
!c tE) Create new
I lff:!l Show r'esul in wei section l[j
~ VPIJS[UJ ...
/ @ Edit existing:
@ Overwrite
0 Use local grid filter () Replace and add new GJ
[J Leave all other cells unchanged

~· ~· Wdllogs
I ~: I
({') Wel~op attributes

00:
¢0
0 Points attribute

I Select log ~ VPVS ...


lo. 5eeu
14' H~~li 'IT'SP.Pfwt


[@
0
Settings

Usebias
Upscaled propetty: I
= #if'-ss

..- I
It=
~eDs: 1~10
10
iti~..htey(

Use saved search


Upscaled from: None
~I l
Scale up settings
El--lfi!@J Wells
Averc5ge method: Arithmetic ... l ... .l ~ Well-1
Treat log: ~points ... L!it~i '___ ! 0 Well-2

Method: Neighbor eel . . Ua


0 Use facies •.veighting

10 Min. number of points in cell : ~


10 Zone correction G.4
Oliputs
0 Well report
0 P.eset output sheet
0 Zone corrected log
I 0 3-D grid zone log
I 0 3D grid well tops

Lt ~ ,][ .t OK l[ 1( Cancel l ...

310 • Stochastic seismic inversion Quantitative Interpretation


The model contains entries for Trend average logs and the
Upscaled log properties in the model tree.
.. 0 Propetfies
lp 0 Al trend [Ll]
p0 RHOB trend [U]
~0 VPVS trend [U]
0
lp AJ[U]
p0 RHOB [U]
~ 0 VPVS {U)

8. Propagate the Zone Average Properties throughout the model.


a. In the Property Modeling group, click Petrophysical.

Geometrical Facies P~trophysical

Property modeling

The Petrophysical modeling with dialog box opens.


with 'Comer Point Models/Class Grid' ~]

I p AJ trend [U]
Status: 1 ts 14>scaled

[ Common II Zone settings j~ [j Global seed: IL-B_7_ __JI


.--
,Zones
- :. . ,] -...,----H-oriz-on-4 --Hor-izo-n 5- ,. [E) mCEHEJ ~~~ [Q) [IDJ]

~·:~~-~~ ~:;;~ ImJ Gaussian random function simulation ..- ]

Quantitative Interpretation Stochastic seismic inversion • 311


b. On the Make model tab, select Edit existing.
c. Click the padlock to unlock additional options as shown in
this figure.
Petrophysical modJmgwith 'Corner_Point_Models/Ciass Grid'

1-.. Edit existing: lp AI trend [U)

~~[li)~ Status: j ls~

IConvnon ](Zone~ l~ Ll Global seed: Ia7 I !


~,~-
: ~J -~.~H-
ooz-
oo-
4--
Ho-
rQo-
nS-· ~ffi ~~~ ~~ ~D

IFacies: ] No conditioning to facies . The zone is modeled in one sin~ operation.

Trends
~ Variogram

[Q]
-
,_0
-- ¢~
Total sill:

Nugget: o.ooo1 ol~


Range

Variogram type : Spherical


Major dir: Minor dir: Vertical:
hlisotropy range: 555 555 3.4
Major direction orientation
-90
Azimuth : 0 Dip: 0 45
0
45
90
[ ] Local varying azimuth

r:J Smbox local azimuth correction

312 • Stochastic seismic inversion Quantitative Interpretation


d. Select one of the Zone average trend logs from the list at
the top right.
e. In the Method for zone/facies list, select Closest.
~i

I Ip AI trend (U]
statUs : j £S lijiSciiled

[] Global seed: IB 7 I [lJ


;::::::::::~ -==--:-----
. rE1 [IHI HEJ ~ ~ ~ [l]D
[ Facies: ] No conditioning to facies . The zone is modeled in one single operation.

[Q]
Total sill : 1.0

Nugget: 0.0001

8 Neura• net
-----1 dJ User defined algorithm

f. Select the next zone from the Zones list.


g. Change the Method for zone/facies to Closest for that
layer.
h. Repeat these steps until all zones have been set and click
Apply.
1. Select a new Zone average/Trend property and repeat Step
c through Step i for all remaining properties.
You now have a model with broad average properties for the coarse
layers defined by our horizons at each grid point, and the detailed log
properties from the well upscaled to the detailed grid at the well
location.

Quantitative Interpretation Stochastic seismic inversion • 313


Exercise 1 - (Optional) Create a simple model for
stochastic inversion
In this exercise, you create a model for stochastic inversion. Use this
data as input:
• Polygon boundary: Small Polygon
• Surfaces: Use the Horizons under the Surfaces_30Mem folder.
Use Horizons 4, 5, 6, and 6-150ms
• Seismic volumes: Use the seismic data with the preface NRM_
located in the 3D_MEM_Surveyfolder for geometry
1. Use the data listed and the steps in "Procedure- Create a
simple prior model for Stochastic inversion" to build a simple
model that you can use for stochastic inversion.
2. When the model is built, examine it by visualizing it in the 30
window and make sure everything ran properly.
This figure shows an example of a slice through a model made
with the data listed in this exercise.

314 • Stochastic seismic inversion Quantitative Interpretation


Variogram modeling
With the prior model created, you now are almost ready to start setting
up your stochastic inversion run. Before you set up the stochastic
inversion, though, you must analyze the data to estimate the variogram
parameters that will be used to guide the Gaussian simulation process
for estimating trial values of the properties during the inversion.
NOTE: It is recommended that you take the Petrel Property Modeling
course before you take this class, or that you at least are familiar with
the material as it relates to variogram analysis.
Data can vary both laterally and vertically. You must compute
variograms that describe how the data is correlated in both dimensions.
There are two steps in the variogram analysis:
• Compute a variogram map to assess whether there are spatial
trends in the data that you need to take into account when you
compute the variogram parameters. If properties are more
similar in one direction than another direction, you must
describe this characteristic to the inversion so that the correct
spatial variability is built in to the Gaussian simulation. An
example of a variogram map was shown earlier in the
variogram discussion.
• The actual variogram computation.

Exercise 2- Create a variogram map


In Petrel, you can make Horizontal variogram maps and Sample
variograms using the Variogram tab in the Settings of a property or
object attribute.
In this exercise, you create a variogram map from one unique log
upscaled and propagated into the grid. However, this does not have
enough information to base our statistical component from it.
Another option where we do not have enough wells would be to
resample the seismic data or previous inversion result into the grid and
use this for variogram map.
1. In the Models pane, locate the model.
2. Expand the Properties and select the settings for a property.
3. In the Settings dialog box, open the Variogram tab.

Quantitative Interpretation Stochastic seismic inversion • 315


4. From the Type list, select Classical for the variogram type.
5. In the Generate section, select Horizontal variogram map.
6. Open the XV range tab.
7. Supply values for these fields: Number of lags x, Number of
lags y, Search distance in X-direction, and Search distance in
Y-direction.
A search distance of 70% of the model dimension is a good
starting point.
8. To run the process, click Run.
- -- --- --- -- - -- - - - - - - - ----- ---
."'~~ ... -,

Variogram

Horizontal search distance

Number of lags x : 10 Number of lags y: 10

Search distance in X-direction: 3000

Search distance in Y.drection: 3000

The process writes the variogram map to the tree as shown in


this figure.
~ It: =
VAR
Variograms
0 Var.Mapfrom AI

316 • Stochastic seismic inversion Quantitative Interpretation


The variogram map must be displayed in a Map window. This
figure shows an example.

·3200 -2o400 · 11100 -800 800 11100 2o400 3200


s a
~ i

I i
~ il
j §

~ ~-~
3200~--~
2o400~--1~
CI00~~~~~~800~~1CI00~-2~
o400~~~ ~
0 500 1000 1500 2000 2600m
1M "' leeeeeee! . .
1:50000

The center of the map represents zero lag. The lag distance
increases from the center outward. It might be necessary to
adjust the parameters to get a good resu lt. From the map
display, a trend along a line with an axis of 20 to 210 degrees
is apparent. Along this axis, the data is most correlated and
least correlated orthogonal to that axis. These directions can
be used to define the major and minor axes for the horizontal
variogram analysis.

Quantitative Interpretation Stochastic seismic inversion • 317


Procedure -Analyze data with a variogram
There are several ways to access tools to perform variogram analyses
in Petrel. This procedure describes how to do perform a variogram
analysis using the Data analysis process. This process works when the
property being analyzed exists as a property in a 3D model as does the
upscaled well log.
1. On the Property Modeling tab, in the Data Preparation
group, click Data analysis.
\6 User-defined object

Training image
Velllog Data ~rend

ups-=alirg · analysis modeling ~ Geo metri~l trend


Data preparativn

The Data analysis dialog box opens .


•• r1- t•fl' '• _ . . . ·• J t

------
Property: ~ VPVS Trend [U]

::J Decimation Decimated number of data: l Cut:;G _ ==:=J


'"' ~ L ~[}J[E)~-= ~ ~ (Q]~

2. Specify the property to be analyzed by selecting it from the


Property list.

NOTE: In the inversion, the variogram parameters are common


to all properties. That is, each property (for example, AI, Vp/Vs,
and density) shares the same vertical and horizontal ranges.

318 • Stochastic seismic inversion Quantitative Interpretation


3. To unlock additional panes, click the padlock.
4. Open the Variograms tab.
,-:!{}' -_... t. • "I.,- .. )... " ._ .!, ~
- ------- -------
Propert'{: I ; AJ P.Jl
~()~ EJ [] ~ Oecinated rumerot ~= 1 :::: ;

[ Zonec: J · Hcriz~f" -'- H:::riz:Jr ; --- ~~~liJ(ffi~~~[QJ[Ij]


IFacies: J Ho coOOI:iar1ingto~. ~~~[Q]
~

[](C
::t1 Sear.:m or. r · g;.::;& Uil'1& [@] '~
Direction Azimuth Dip
Number Lag Sea reb Band Tolerance Lao
lags dirtaoc~ .adm. wfdtb antle tofenoce
Vertical t!A
~· s 25 2ti£1 5l) ~s 5li
Major :J 0 s .:S.!: lXoO UJ 1.5 5n
Kinor 173 I) L5l) 1XCI 2::0& .G5 5I:

V~model~

Nugget: ~ ~~ [Q]~
.Structure 1
Type Spr,ericai
·.SUI ~.SC...$9

Major range 5ZQ


Minor range 5!:•!)
'VertiCal range l ~t-

; Vertic~! l±ir~t:i" I Maier cirectioo I Mioor ::i rectioo I


~ ~ [iJ ! 1:.;en-1 ... ~ ~ - ~W P.egre~n Model--

t~--f1 ,_~ _~--tt1


1 2

A f-.
"-.:;....--"
___ ___
------~
:~t::
'---11--'"
12.5 t2.s 11 2.s 1sh 21is
Distance
-2t'C -100 --

Ouant1tat1ve Interpretation Stochastic se1sm1c 1nvers1on • 319


The Variograms tab is divided into three main sections:
• Experimental variogram computation (the upper
section): Contains all the parameters that describe the
search cone used to select the data that is used to
compute the experimental variograms
• Variogram model fitting (the middle section): Contains
all the parameters that describe the variogram model
• Graphs (the lower section): Three direction tabs
(Vertical, Major, and Minor) are available. On each tab,
two graphs are displayed:
• On the left, a plot shows the search cone graph and its
dimensions, as well as the location of the sample points
from which it is sampling.
• On the right, the variogram graph with both experimental
and model variograms is plotted.
It is possible to reset all variogram parameters on the
Variograms tab by clicking the Reset button D in the top
right corner of the tab.
5. Turn the Zones button on or off to allow computation by zone
or to allow data from different zones to be used together.
For well data, when zones are thin, a single zone might not
have enough data to permit meaningful analysis without
including data from other (similar) zones.
In this figure, lp is the selected property and the Zones button
is turned on to select a single zone for analysis.
Property: lp AI [U]

~ llJ [j] !CJ 0 Decimation Decimated number of data: 110 I


lz.anes: l ,. [E)8JITJ[ffi[~}~ [ii [Q][OJ
IFacie3: j No conditioning t o facies. ~g~ [j)

In this figure, the Zones button is turned off, so the data is


analyzed as one set.
Property: lp Al [U}
~I]~ LJ [] Decimation Decimated number of data: L-1
1 _ _ _------"

IZones: I All zones modeled together (foggle 'Together' in Petrophysical modeling)


[ Facies: ] No conditioning to facies.

320 • Stochastic seismic inversion Quantitative Interpretation


6. In the Experimental variogram computation section, adjust
the parameters as necessary.
The parameters that cannot be adjusted are dimmed. In this
case, there is only one well available, so parameters relating to
lateral property variation are not available.

NA Number lAg Search BaiiCI Tolerance tolei.A.!..,.. Thickness


.,.,. lags distance radiUJ width .angle ........
Vertical NA 90 50 5.8 290 50 45 50 NA
Major 0 0 250 2000 200 50 0.001
Hlnor 270 0 250 2000 200 45 50

For a single well case, these parameters are avai lable:


• Number of lags: The number of data bins that collect the
data that enters the variogram analysis. Combined with
the lag distance, this parameter defines the extent of the
search cone.
• Lag distance: The length (in project units) of a bin that is
used to select data for the variogram analysis.
• Search rad ius: Automatically computed by multiplying the
number of lags by the lag distance.
NOTE: The tooltip contains definitions of the Bandwidth,
Tolerance angle, and Lag tolerance.
7. In the Variogram model fitting section, click Type to select
the type of model variogram to be calculated.
You can select the type based on which type fits the data
better.
Variognn IDOdel fitting
Nugget: 0.0001 Total sill: 1 ~ ~ (!iJ
Structure 1
Type Spherical
s•
Major range
0.9999
500
Minorranoe 500
Vertk:al range _ _ 100-..~
__

Quantitative Interpretation StochastiC seJsmJc mvers1on • 321


On the Vertical direction tab is a graphical display of, in this case, the
well and the region over which the data is collected for the
experimental variogram. You can change the shape of the search area
by clicking and dragging the blue squares in the cone or edit the table
above.
Vertical di recti on Major direction Minor d
- ~ [jj) A \'/ell· l --

NOTE: The vertical variogram model is easy to estimate as there is


usually plenty of data from the logs. The nugget also is easy to find and
will be the same for all three direction.
A Horizontal variogram model (major and minor directions) usually is
harder to obtain because there is limited data. Ranges in these two
directions can be implied from the Variogram map. The major range can
be found from the Sample variogram in the direction indicated by the
map.
To the right of the well pane is the variogram pane. This pane shows
the points used to create the experimental variogram. It also shows a
calculated best fit curve in gray and a user-adjustable curve in blue.
To change the shape of the model variogram, drag the brown and blue
squares highlighted in red in this figure.
......
......
40 CJ
m
• ~ ......

•••••••
8£.
• • •••• w
OJ

••••••••
~~~~~~~7~5~~,+,~,~~14r;~
~ ~~,~~~~~2+19~~2Y5~S~~1+91~--~ os
Distance ::!.
S!.
322 • Stochastic seismic inversion Quantitative Interpretation
You can try different variogram types to see if you can improve
the fit. The values for the Sill and Range in the Variogram
model fitting section are recalculated. When a reasonable fit
is achieved, the sill and range parameters can be read for later
entry into the inversion parameter input.

••••••••

Exercise 3 - Variogram analysis


In this exercise, you use the variogram tool to analyze the statistics of
the Well1 (Jupiter) data in preparation for the stochastic inversion run.
1. Use "Procedure- Analyze data with a variogram" to access
the variogram analysis tool and use that tool to estimate the
nugget, range, and sill of the data to use in the inversion.
2. To calculate the results, use the acoustic impedance.
3. Experiment with the parameters to see what works best to
give a good analysis with a sufficient search radius (greatest
distance between samples included in the analysis).
What values for the variogram parameters did you calculate?

Quantitative Interpretation Stochastic seismic inversion • 323


Lesson 3- Set up and run a stochastic inversion
With the model in place and the variogram analysis completed, you
now are ready to set up and run the inversion.

Procedure - Set up t.~e inversion run


The original logs, the trend logs, and the angle stacks and their
wavelets are the inputs for the Stochastic seismic inversion. The
variogram parameters, the property distribution, and the number of
realizations can be defined in this process after a rigorous analysis.
1. Change to the Advanced Geophysics Perspective.
2. On the Quantitative Interpretation tab, in the Seismic
inversion group, click Stochastic inversion.
•: ',;J ~ Inversion property builder
J
~ Inversion QC plot
Simultaneous
inversion -Q Stochas·ic inversion

The Stochastic seismic inversion dialog box opens.


-Q Stochastic sei~mic inversion

Algorithm:

Select model grid :

rlJ .Acoustic lmped;mce:

Model property trends - - - - - - - - - - - - - - - - - - - -

Acoustic Impedance :

324 • Stochastic seismic inversion Quantitative Interpretation


3. Open the Make stochastic seismic inversion tab and
complete these steps:
a. In the Algorithm list, select the type of inversion to run. In this
procedure, Pre-Stack Stochastic Inversion for Acoustic
Impedance, VpNs and Density is selected.
b. Select the model to be used for the inversion in the
Models pane and insert ¢ it into the Select model grid
field.
c. Select the seismic survey from the Input pane that
contains the data that will be used in the inversion and
click 9 to insert it into the Select seismic survey field.
-<) Stochastic seismic in~on
Make stochastic seismic inversion

Algorithm: IPre-Stack Stochastic Inversion for koustic ~ce. Vp!Vs and Densiy
Select model grid:

Select seismic survey:

4. Open the Model Properties tab.


The properties on this tab vary based on the algorithm that you
select. For the algorithm selected in this procedure, fields
appear that allow you to enter the appropriate Acoustic
impedance, Vp/Vs, and Density.
a. In the Model properties upscaled from well data
section, select the check boxes and the upscaled
properties from the lists.
b. In the Model property trends section, select the
average trend model properties from the lists.

[{] koustlc Impedance ~ AI

!'{] 1/pN s· VFVS

Mode!property!Tends - - - - - - - - - - - -

kousllc Impedance AI Trend

\'pN s: VFVSTrerd

RHOBtrerd

Quantitative Interpretation Stochastic seismic inversion • 325


5. Open the Seismic Volumes tab and complete these steps:
a. Select a seismic volume from the list on the left.
b. Enter these parameters for the seismic volume that you
select.
• Optimization weight.
• Minimum incidence angle and Maximum incidence angle:
These parameters are the near and far angles for the
seismic volume that you selected. The Incidence angle
range and the Effective incidence angle are calculated
automatically.
• Wavelet to be used for the selected seismic volume from
the Input pane. Insert ¢ the wavelet into the Seismic
wavelet field .
c. Repeat Step a and b for each volume in the list to be used.
Model Properties tJ Seismic Volumes lei Zone Settings trJ Output Properties 1
::; Expert Settings

Seismic volume settings

[] HDVA_VelocitLAms
Optimization 'Neight:
[] Average_Velocity [Realized 1
EJ Near_5-20
r:J Mid_18_33 Minimum incidence angle: 5 degrees
[] Far_31_45
[1] NRM_MearCrop_5-20 M~mum incidence angle: 2{) degrees
[1] NRM_MidCrop_18-33
[1] NRM_Far0op_31-45 Incidence angle range: 15.00 degrees
0 Depth_wavelet_cubefjar
[] CC_Cube_50
Eifective incidence angle: 13.11 degrees
D CC_Cube
Seismic wavelet: fE31V-.JP 1_NearRZ_5-20
NOTE: Use the Optimization weight field to control the relative
influence of the current seismic angle stack volume in the
inversion. The weight controls the importance of matching
traces from the current volume in the calculation of the global
objective function. Use the default value if all volumes are of
good quality.
6. Open the Zone Settings tab and complete these steps:
a. Select or clear the Use settings by zone check box. This
option controls whether the parameters on this tab are to be
applied to the whole dataset or if separate parameters are
entered for each zone. You select or clear this option,
depending on how the variogram analysis was done.
326 • Stochastic seismic inversion Quantitative Interpretation
NOTE: Whether you select or clear this option depends on how
similar the statistics and the geology are for the different zones
in the model.
b. In the Variogram model section, take these actions:
• Select the Variogram type from the list. Be consistent with
the variogram analysis that was done.
• Enter the Sill, Nugget, and Vertical range from the
variogram analysis.
• Enter values for the Major range and Minor range based
on experience with other wells or as test values. You can
assess the impact of the test values in trial inversions
until a reasonable result is obtained.
• If structural information is available and was used in the
variogram calculation, enter Dip and Plunge. For a single
well, the Azimuth cannot be determined,
• To align the major and minor axes with the model axes,
click the Align button.
• To force the range parameters to be the same, click the
Isotropic button.
c. In the Correlation matrix section, the model parameters
(in this case AI, VpNs, and density) are correlated to some
degree, which means that the program cannot make
completely random choices for each parameter. The matrix
in the Correlation matrix section informs the inversion to
what degree the model properties are correlated.
Enter values or click the Estimate button to have the
program estimate the values from the data.
To ensure that the estimates are statistically valid, click
the Validate button. If necessary, the program adjusts the
matrix so that it is valid.
d. In the Distribution section, enter appropriate Std
Deviation, Min, and Max values for the properties, or, if
upscaled properties were entered on the Model
properties tab, you can click Estimate to have the
program calculate the entries.
You can adjust the Min and Max values manually after the
program estimates them to prevent non-logical values from
being allowed (for example, Poisson's ration greater than
0.5).

Quantitative Interpretation Stochastic seismic inversion • 327


Model Properties § Seismic Volumes !! Zone Settings Output Properties
~~~--~--~--~~--~~--
Expert Settings

[[] Use settings by zone

Variogram model· - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Variogram type: [ Exponential ... ]


Sill : Nugget effect: 0 Isotropic

Major range : 1500 Azimuth : 0 [n ft.Jign

Minor range: 1500 Dip : 0

Vertical range: 1500 Plunge: 0

Correlation matrix

.Acoustic Impedance: VpJVs : Density:

Acoustic Impedance: 1.0 [¢ Estimate


VpNs: -o.sn6 1.0 [):lValldate ...
Density: 0.8239 -0.5033 1.0

Distribution

Std. Devimion Min Max


Acoustic Impedance: S29.20 10 13559.64 1 u 16918.75 I [¢ Estimate 1

Vp1Vs: 0.18 L] 11.95 I rJ 13- I [¢ Estimate 1

Density: 73.53 LJ I 2. 8 I ti~::.:~~:l

328 • Stochastic seismic inversion


7. Open the Output properties tab and complete these steps:
a. Use the Output property folder field to specify the name of
the folder that contains all of the output elastic property
realizations. This folder appears in the Models pane under
the Properties folder of the model.
A 0 Oass Grid SSI Output
" 0 Acoustic Impedance
lp 0 Oass Grid SSI_.Acoustic lmpedance_1 [UJ
lp 0 Oass Grid SSI_.Acoustic lmpedance_2 [U]
lp 0 Oass Grid SSI_koustic lmpedance_3 [U]
lp 0 Oass Grid SSI_.Acoustic Impedance_4 [U]
lp 0 Oass Grid SSI_koustic lmpedance_5 [UJ
lp 0 Oass Grid SSI_Acoustic lmpedance_6 [U]
" 0 Vp/Vs
~ 0 Oass Grid SSI_Vp!Vs-.1 [U)
~ 0 Oass Grid SSI_Vp/Vs _2 [U]
Vi 0 Oass Grid SSI_Vp/Vs_3 [U}
~ 0 Oass Grid SSI_Vp/Vs_4 [U]
~ 0 Oass Grid SSI_Vp/Vs_5 [U]
~ 0 Oass Grid SSI_Vp/Vs_6 [U)
"- 0 Density
p0 Oass Grid SSI_Density_1 [U)
p0 Oass Grid SSI_Density_2 [U)
p0 Oass Grid SSI_Density_3 [U)
p0 Oass Grid SSI_Density _4 [U}
p0 Oass Grid SSI_Density_S [U)
p0 Oass Grid SSI_Density_6 [Uj

Quantitative Interpretation Stochastic seismic inversion • 329


b. Use the Prefix for output property names to control the
naming of the output elastic property realizations. The
properties are created in the output property folder.
c. Use the Random number generator seed field to control the
seed value used to initialize the random number generator
used in the stochastic inversion algorithm. The Random
number generator seed controls the property distribution in
each realization. A set of realizations can be reproduced
exactly by using the same seed (if all other settings remain
the same).
d. Use the Number of realizations to generate to specify how
many property realizations are created in the output
property folder.
e. Set the Number of realizations per path to a value greater
than 1 to enable efficiencies of scale that will improve the
overall performance of the algorithm.
NOTE: This improved performance comes at the cost of
increased memory consumption because of the larger
number of realizations that are held in memory at any one
time. The amount of memory consumed is in direct
proportion to the number of realizations per path.
Therefore, use of this setting for values greater than one
requires careful monitoring of the memory consumption.
Model Properties tJ Seismic Volumes iSS Zone Settings fl Output Properties
Output Setting~

Output propert'lfolder: Class Grid SSI Output

Prefix for output propert'f names: Class Grid SSI

Random number generator seed : ~I


Number of realizations to generate: [5 ~I
Number of realizations per path: ~I

330 • Stochastic seismic inversion Quantitative Interpretation


8. Open the Expert settings tab to show the expert settings.
Complete these steps:
a. Enter the value for the Convergence criterion. The cost
function must reach this value to achieve successful
convergence. The value can be between 0 and 1. A value
of 1 requires a perfect match to the seismic data, and a
value of 0 implies little to no match.
NOTE: A value of 0.8 means that the RMS difference
between the actual and predicted trace is 20% of the RMS
value of the trace itself.
b. Specify the Maximum number of iterations. This value
prevents the program from trying to achieve a match for
excessively long periods of time.
c. In the Diagnostics section, select the Output
convergence data as point set attributes check box to
have the program write out the convergence maps and
other diagnostics as point set entries in the tree.
TIP: These settings give good visual indications of where
and how well the inversion converged. They are key to u
understanding where the inversion potentially is reliable
and where it is not.
Model Properties (J Seismic Volumes E! Zone Settings fJ Output Properties ':' Expert Settings

Optimization Settings
-
Convergence criterion: 0.8

Maximum number of iterations: ~I


Diagnostics

~ Output convergence data as point set attributes

9. To launch the inversion job, click Apply or OK

Quantitative Interpretation Stochastic seismic inversion • 331


Exercise 1 - Run a stochastic inversion
You do not have enough time in class to run a full-fledged stochastic
inversion with enough realizations to be statistically meaningful. For
this reason, in this exercise, you set up a simple inversion run with a
small number of outputs. This exercise gives you experience in using
the interface, assembling input data, and visualizing the diagnostic
output.
For this exercise, use this data:
• Model: Use the model you created in the earlier exercise
• Geostatistics: Use the results of the variogram analysis that
you performed earlier.
• Seismic volumes: Use the seismic data with the preface
NRM_Iocated in the 3D_MEM_Survey
• Wavelets: Use the wavelets in the Wavelets_Finai_After_
NRM folder for geometry
• Where not specified, the default parameters work well .
1. Using the given data, set up and run a stochastic inversion.
Refer to "Procedure- Set up the inversion run", if necessary.
2. Select the default parameters that work well. The major axis
range can be set to around 1700 m and the minor range to
around 1500 m.
3. Set the inversion to generate five realizations
4. Select the check box to output the full suite of diagnostics.
5. The run should complete in about 40 minutes or less.

332 • Stochastic seismic inversion Quantitative Interpretation


Lesson 4- Stochastic inversion run evaluation
After running the inversion, the diagnostics should be examined and the
results quality checked to be sure that the inversions converged. The
input tree contains several diagnostic entries that can help understand
the inverted output. This data is in the form of a pointset.

Procedure -View the diagnostic output


If the option in the Diagnostics section of the Stochastic seismic
inversion dialog box is selected, the diagnostic attributes are stored in
the Input pane. You can customize the style of the output for better
visualization.
1. To display the diagnostics, locate the SSI Output folder in the
Input pane.
2. Open the folder and select an entry to view.
~ 0 Oass Gid SSI Output
~ TwT 0 Diagnostics
.. :S·0 Pttrtbutes
TwT 0 HYT auto
~o x
~OY
~ 0 I index
~ 0 J Index
~ 0 Defined
~ 0 Vertical
~ 0 Vertical and convex
• 0 lllndex
0 XL Index
~ 0 OassGridSSI_sequence_1
Q 011ss Grid SSI_cost_1
~ 0 Oass Grid SSI_IerZIIions_1
~ 0 011ss Grid SSI_converged_1

Quantitative Interpretation Stochastic seismic inversion • 333


3. Access the settings for the Diagnostics (top entry in the folder).
a. Open the Style tab and then the Common subtab.
b. To color the points with their attribute value, change the
Color selection to As attribute.
c. To improve visualization quality, change the Size.
d. Click Apply or OK

J3. Common
Color. As attribute
Symbol: a SimpUfied sphere

Siize: 5

~fateri:al: """ I

334 • Stochastic seismic inversion Quantitative Interpretation


Main diagnostics
The main diagnostics of interest are the Sequence, Cost, Iterations, and
Converged attributes. They are defined in this table.

<Prefix>_sequence_<seed> Order in which locations were visited


during the generation of a realization.
Lower numbers indicate locations that
were inverted earlier in the process
and larger numbers indicate locations
that were inverted later. The value
varies from 1 (the first location) to the
total number of defined locations (the
last location).Posting this attribute on
the point set reveals a structured,
multiresolution grid visitation
sequence that is specific to a
realization with this seed.
<Prefix>_cost_<seed> Correlation coefficient between
observed seismic data and the
synthetic seismogram generated from
the optimized model at this location.
Specific to a realization with this seed.
<Prefix>_iterations_<seed> The number of iterations performed in
the iterative optimization at this
location. Specific to a realization with
this seed.
<PrefiX>_converged_<seed> Boolean flag indicating whether the
iterative optimization has convergence
at this location (true) or has not fully
converged (false). Specific to a
realization with this seed.

Quantitative Interpretation Stochastic seismic inversion • 335


Diagnostic examples
Figure 10 shows the Sequence diagnostic. The color code reveals the
order in which the cells were inverted.

Figure 10 Sequence diagnostic

Figure 11 shows the Cost diagnostic. This diagnostic shows the value of
the cost function at each location. The vast majority were above the 0.8
threshold used as the value for this run, which indicates a successful
inversion over most of the area. The inversion had trouble fitting the
data where the dip is steep in the fault zone and where the shallow gas
anomaly interferes with data quality.

Figure 11 Cost diagnostic

336 • Stochastic seismic inversion Quantitative Interpretation


Figure 12 shows the Iterations diagnostic. This diagnostic is the number
of iterations that it took for convergence to occur or for the maximum
number of iterations to be exceeded. Generally, the number of iterations
is low and the program is able to fit the data without too much effort. In
areas where the cost function is low, many iterations were tried to get
to a converged status, but these iterations failed to reach convergence.

Figure 12 Iterations diagnostic

Figure 13 shows the Converged diagnostic. This diagnostic shows


where the program converged. This display has the same patterns as
the cost and iteration appears.

Figure 13 Convergence diagnostic

Quantitative Interpretation Stochastic seismic inversion • 337


There is one of each of these displays for each realization. If a hundred
realizations are created, a lot of diagnostics are created to review.
Testing on a small area with a few iterations can help fine-tune the
parameters before running a full job with many realizations.

Model output
After reviewing the diagnostic output, you should review the output
data.
Figure 14 through Figure 17 are examples of the output for one run with
a particular set of parameter input choices.

Figure 14 Input seismic data

338 • Stochastic seismic inversion Quantitative Interpretation


Figure 16 Example of the VpNs property

Quantitative Interpretation Stochastic seismic inversion • 339


Figure 17 Example of the density property

In this example, there were not enough wells to assess the spatial
variogram for lateral variation, so an isotropic value of around 2000 m
was chosen as a trial to assess the result (the vertical range was set at
around 50 m). The fine vertical layering is clear, as is a great deal of
lateral continuity in the beds. The hydrocarbon accumulation and flat
spot, however, are not very clear.
Compare this example to the simultaneous inversion shown in Figure
18.

340 • Stochastic seismic inversion Quantitative Interpretation


Figure 18 Simultaneous inversion result for acoustic impedance at the location
of the previous sections

It is possible that the Major range and Minor range of 2000 m was
excessive and did not allow enough flexibility to pick up the pay section.
To test this assumption, a second run was made with a range
parameter of 500 m in the lateral directions and 50 m vertically. The
resu It is quite different.

Figure 19 Example of the acoustic impedance property for a small variogram


range value

Quantitative Interpretation Stochastic seismic inversion • 341


Figure 21 An example of the density property for a small variogram range value

This result is noisier and with some patchiness, but the flat spot clearly
is visible in this result. It seems to be closer to representing what you
know the subsurface to be and with some further adjustment, what you
would like to see in the result.

342 • Stochastic seismic inversion Quantitative Interpretation


Lesson 5- Applications of Stochastic inversion
results
Applications of stochastic inversion include
• Petrophysical property modeling
• Dynamic flow simulation
• Volumetrics
• Well Planning
• Well connectivity
• Connected volumes
The key point of stochastic inversion is to create multiple realizations
rather than just a single best result. Because there are multiple
different but equi-probable results, you can analyze the results
statistically and derive likelihoods for various outcomes.
For example, if you have a regression between acoustic impedance and
porosity, the acoustic impedance volumes can be transformed into
porosity cubes. These porosity cubes then can be used to create
permeability cubes that can be used in flow simulation.
The volumes can be interrogated for the net rock volume that exceeds a
porosity threshold and the results ranked according to how often they
occurred. This distribution can be used as input to a volumetric
assessment.
Using geobodies, the amount of connected (and, therefore, potentially
producible) rock volume can be assessed, again with probabilities
based on how often they occurred in the realizations. This output also
can be used for well placement decisions.

Quantitative Interpretation Stochastic seismic inversion • 343


Figure 22 and Figure 23 show the transformation of an acoustic
impedance volume into multiple porosity models.

AI-PORO
Transform

Figure 22 Acoustic impedance with an AI to porosity trend. The trend can be


used to calculate porosity from acoustic impedance

Hi-Res, Equiprobable
Stochastic Simulations

Depth Volumes

Figure 23 Porosity volumes calculated from the impedance volume

344 • Stochastic seismic inversion Quantitative Interpretation


These volumes can be ranked and then representative models can be
selected for further analysis. The cumulative probability plot in Figure
24 allows P15, P50, and P85 models to be selected and analyzed
further.
0.9

0.8

0.7
(;
~ 0.6
:I
c:r
f 0.5
u.
u
i 0.4 -+- 1 Realization
:;
E 0.3
:I
u
0.2

0.1

334000000 330000000 342000000 346000000 350000000


Pore Volume, m3

Figure 24 Cumulative probability distribution calculated from the statistics of


the stochastic inversion results

Workflows using stochastic realizations


The many realizations resulting from a stochastic inversion give a sense
of the uncertainty or variability in outcomes that are consistent with
what you know from seismic and well control. However, to use the
results quantitatively, you must be able to calculate properties from the
realizations.
Examples of calculations that are commonly done include
• Calculating the probability of a specific property based on its
frequency of occurrence in the different realizations
• Ranking results
• Converting acoustic impedance to porosity
These calculations can be done using the Workflow editor in Petrel.
NOTE: You have not been able to generate enough volumes in class to
perform these calculations effectively in an exercise, so using the
Workflow editor is presented here as an example only.

Quantitative Interpretation Stochastic seismic inversion • 345


Procedure - Calculate the probability property
A pre-made workflow has been created to calculate the probability of
the different properties. You must enter the inversion results into the
workflow, run a dummy property, and set the cutoffs.
1. To access the Workflow editor, open the Workflows pane in
the lower left of the Petrel window.

r> lj v~
If& Woddlow 1
§) PrabiMiy_CalcdiUin_AI_VPVS T
Dl Probabilty_CalcUation_no_den5ity 0
Dl Porosity Calculation
~ High_Ranking
~ Copy of High-Ranking
§.) Porisity from AJ
~ L1hocube
§.) Woddlow 3
fiTl WO!kflow 2
~l 1Norkflow 4

mWorkflov.-s 110 Windows I


This pane displays a list of workflows that have been built.
New workflows also can be built.

346 • Stochastic seismic inversion Quantitative Interpretation


2. Double-click a workflow entry to open it in the Workflow
editor.
The workflow Probability_Calculation_AI_VPVS contains a
workflow to calculate the probability associated with cell
values that meet specific cutoff values. This figure shows what
the workflow looks like in the Workflow editor.

f'k.-mettee:l:PfCSSIOn SRHOS_CA • lSOO


Numonc...,.,siOII Sl • ,.'
0-'------------------

(Q F«oi!Oar.sm lo.t ~ U VanlllltA!IIIobol


8 Gt<~nome ~@i-;
l ij'"Vwill>lo
-.- AX-I;Ioboll
-.-~
8 Gt<ealew...,.name ~~ ilvJrill>le R. Icloboll
-;.""'< • .,.,., ~ ~ ~~[] YtoriobleAYI;Ioi»>J _ _
liiJ Prooertycalewlrlo< rJ U..fl• &pott;oncrile Jl'ld>.(JAJ <•SAI. c..t Mtl 11/PVS <- SV!'VS.CU. - ·l . -.;))
th.:mencexpt~1on S' • So-l
le Emiooo
li F'fooenyealeulolcr u.e•"' ~..-Jie. Sl'lob·-ISI

Quantitative Interpretation Stochastic seismic inversion • 347


The actual workflow appears in the panel on the right side of
the Workflow editor window.
l. m\•lith 30 grid ~ • Copy of Small Model PAF 1 Use: Specified grid
2 m> Set reference list ~ Q Variable A [global
3
4
5
6 Propert'f calculator r:J Use filter SProb=O
7
8 Nurreric expression SAI_Cut: 4500
9 ~ N rrericexpression -SVP
_ V_S_-Cut
_ _ _ _ _ _ __
2.7
l.O ~ Nurreric expression SRHOB_Cut 2500
1.1 f Nurreric expression Sl
------------- = 0
1.2 {Q For all icons tn list ~ • ([ Vc:riable A [global I
1.3 ~ Get calculator narre SAl ~ ~ Variable AA [global}
:14 ~ Get calculator name SVPVS ~ ~ Variable A!. [globaiJ
:15 Cle:l cal lat;::r nrre !sRtJOB I~ ~ Variable AY [global]
.16 tij Property calculator D Use filter ExpressiOn or file: SProb~{SAI ~ = SAJ_Cut And SVPVS <=SVPVS_Cut . SProb+1 . SProb...O)
17 NUJJ"eric expression Sl = Si~ 1

18 }(; End loop


19 iii Propero; calculator [] Use filter Expression or file: SProb=SProb/ SI

The steps in the workflow are predefined. Only a few entries


are needed to get it ready to run.
• The model must be specified in line 1.
• The actual realizations to use in the calculation must be
set in line 2.
• A probability property for the model must be defined and
initialized in line 4.
3. To enter the model to be used, click in line 1.

348 • Stochastic seismic inversion Quantitative Interpretation


4. Double-click Set reference list in line 2.
The Initialize reference list dialog box opens. Use this dialog
box to enter the inversion results to be used in the probability
calculation.
-~ ,
Set reference list Variable A {global I
~ Initialize ref-erence list

~ Variable AX [glo ~Iii variable AZ [glo ~ Variable AV [glo


-~
1 ~
2
3
4
s
6

5. Use the buttons at the top of the dialog box to manipulate the
table. Use the blue arrows in the table to enter rea lizations.
In this workflow, AX is assigned to acoustic impedance, AZ to
VpNs, and AY to density.
Enter the volumes that contain these property rea lizations into
the appropriate column as shown in this figure.
~ Initialize reference list

1 Class Grid SSI_Density


2 Class Grid SSI_Densito;
3 Class Grirl SSI_Cen.sity
4 Class Grid SSI_Density
5 ~~G~$1_~~~
6 Class Grid SSI_Densit-l

6. After all of the data is entered, click OK


WARNING: Be sure to clear any empty rows using the tools in
the tool bar. Empty rows in the input cause errors in the
inversion run because the program cannot find the empty input
data volumes.
7. In line 4, define a dummy property to contain the computed
probability.
Quantitative Interpretation Stochastic seismic inversion • 349
You can define this property using the Property calculator. To
access the Property calculator, right-click the Property entry
in the tree in the Models pane and select Calculator, or click
Calculator on the Property modeling tab, in the Property
interpretation group.

Calculator Net.ra l
ne!
Property interpr:et.atton

IJ Property cakul~or for 'Cla.s ~ Grid'

~
~

S,ettings
~·~·---~----~-----·--------------------

Import (on selection), .. [] From fife: [][Run)


Select property Vl!Mble: llttach new to template :
ErJjt global color table
' lp AI Trend [U] • General
: p RHOB trend (U]
I~ r. ~ VPVS Trend (U}
Filtertemplates: [• Fu1c:tions
I
;··-lp AI [U]
0 Ut efilter
I· Geometry I
;. p RHOS [U]
;. -J! VPVS [U] O Rad [ ENTER I
Collapse (recursive)
rB Class Grid 551 Output G) Deg
() Grad
@]G]
Expand (recurs ive)
[j Hyp
[!]EJ
Sort by name.s
Sort by property temp lates
lO ~lV I Rotnd IIScJt 10 [!] 0 ~ Ihld I [!]
Sort by property templates/name ~~~0[}]0c:JW G
Sort by time
Auto color all (recursive)
~~~OJ00DOJEJ
Auto name all (re<:ursive} ~~~0@JD0Wc:J

350 • Stochastic seismic inversion Quantitative Interpretation


8. Enter the expression to initialize the probability property into
the window as shown in this figure and click ENTER.
Ill Property calculator for 'Class Grief
Show
@ 1-Wory
0 AeUt

[OJ~
JL____ _ _ _ _ _ _ _ ____,~ I
~~ l
[CJ Fromfile: .___ _ _ _ _ _ _ _ ____, []I Run ]
Jlttach new to template:
··lp AITrend(Uj ~ General
;p RHOB trend [U] Filter templates: T-Ftn::tions-.--,
,-I

;. ~ VPI/S Trend [U)


! lp AI [UJ 0 Use filer IT Geometry ]
! p RHOB [U]
Rad [ ENTER ]
:· ~ VPVS [U}
ffi - Class Grid SSI Output @]GJ
Grad [!] [ <= I
a ~~ 1fbnd 11 ScJt 1000~[ .aoo 18
~~08J00c:J0G
~~[BOJ00DOJ EJ
~~~CEJ@JDGJUJ 8
The property is placed in the Models pane tree with the value
assigned.
~ O blh~IM
lp 0 AI Trend [U)
~ 0 VPVS Trend [U)
p0 RHOS Trend [U)
lp 0 AI[U
~0 VPVS [U)
p0 RHOS [U)
t> - 0 SSI Output Small Model PAF 1
I> - 0 SSI Output Small Model PAF 2
~ - 0 SSI Output Small Model PAF 3
I> 0 koustic Impedance
I> 0 VpNs
t> - 0 Density
0 Probability

Quantitative Interpretation Stochastic seismic inversion • 351


This figure shows a probability property with a zero value in a
3D window.

9. Highlight the property in the tree and insert it into the


appropriate cell in line 4 in the Workflow editor.
~ r ~ I 111 Variable Z [global = Q I Probability

10. Set the cutoffs to be used to appropriate values according to


the results of any rock physics or petrophysical modeling
processes that were run.

8 rn Numeric expression SAI_Cut 4.7


9 rn Numeric expression $VPVS_Cut 2.35
10 rn Numeric expressio SRHOB_Cut 2500
11 rn Numeric expression $1 0

11. To apply the changes, click Apply.


12. To make sure that the workflow has no syntax errors, click
Test.
13. To run the workflow, click Run.

352 • Stochastic seismic inversion Quantitative Interpretation


An example of the output is shown in this figure.

Quantitative Interpretation Stochastic seismic inversion • 353


Lesson 6- Volume attribute modeling
Volume attribute modeling in the Seismic inversion group is a
process that you can use to create synthetic seismic volumes from
inverted elastic properties. You can create synthetic seismic volumes in
two ways:
• From inverted property cubes
• From a hybrid model where a pillar grid is embedded effectively in
a volume
In the first case, the inverted property cubes can, for example, be the
property cubes of a simultaneous inversion.
In the second case, the simultaneous inversion cubes can be used as
the background volume, and the pillar grid can be the stochastic
inversion output properties.
This process can be applied to stochastic inversion because it creates
synthetic seismic volumes from the stochastic inversion results. These
results then can be compared to the input seismic angle volumes as a
quality check measure. If the match is not good, parameter adjustment
in the inversion might be needed to achieve a proper match to key parts
of the seismic.
The synthetic volumes created by this process are noise free in the
sense that they are a straight convolution of the wavelet with a
property volume. Spectrally shaped random noise can be added in a
second step to create a more realistic seismic volume.

354 • Stochastic seismic inversion Quantitative Interpretation


Procedure - Use volume attributes and reservoir AVO
modeling to create seismic models from stochastic
inversion results
This procedure shows you how to run the Reservoir AVO modeling
process as applied to stochastic inversion output properties.
1. On the Quantitative Interpretation tab, in the AVO group,
click Volume attributes.
t}} 11 ~ Volurne Al:rb<:e<

AVO
AVO li\1 AVO Seismic logs
modeling r@Connaissance Wedge model
Al/0

The Volume attributes dialog box opens.


2. Open the Calculate attributes tab and take these actions:
• In the Category column, select Reservoir AVO modeling.
• In the Attribute column, select Isotropic synthetic.
~ Volume attributes

Caloulate attributes L:.H.:::.:in=.ts...t____ _ _ _ _ _ _ _ _ _ _ _ _ _ _-----.

Use Ca1r:gory AtlriJute


Ill Res~oir AVO modeling • lsouopic synlhetic

To the right of the Attribute column is a check box labeled


Realize. If you clear this check box, the modeling results are
created as a virtual volume. Creating the results as a virtual
volume allows you to assess the results quickly. If you select
this check box, the results are realized, which takes more time
to calculate.
3. In the lower part of the dialog box, open the Parameters tab.

Algorithm : [ PP AVO. '*i & Richards -acoustic impedance. )/pNs and density • ]

Gridproperties - - - - - - - - - - - - - - - - - -

[l] lndude elastic properties from grid

Selectagrid : ~ '--------------__J
Acoustic impedance:

Vp!Vs:

Density:

Quantitative Interpretation Stochastic seismic inversion • 355


4. From the Algorithm list, select the algorithm to use to model
the AVO.
If you are doing this step as part of a qua lity check process for
an inversion, choose the same algorithm that was used in the
inversion run.

Post stack. normal incidence · acoustic impedance


Grid properti• PP AVO. Zoeppritz · Vp. Vs and density
1'71 PP AVO. Zoeppritz - acoustic impedance. VpN s and density
~ Inc PP AVO. Zoeppritz - acoustic impedance. Poisson's ratio and density
Select PP AVO, Zoeppritz · acoustic impedance. shear impedance and density
PP AVO. Aki & Richards ·Vi . Vs and den ·

kous PP AVO, Aki &Richards · acoustic impedance, Poisson's ratio and density
PP AVO. Aki & Richarcts • acoustic impedance. shear impedance and density
Vi N s PP AVO. Fatti · acoustic impedance and shear impedance
P PP AVO. Verm &Hi~erman ·acoustic impedance and Poisson's ratio
. PP AVO, Shuey 2tenns • Vp. Vs and density
Densit PP AVO. Shuey 3 terms - Vp . Vs and density

For this procedure, PP AVO, Aki &Richard-acoustic impedance,


Vp/Vs and density is selected (this algorithm was selected for
the stochastic inversion).
5. If you wish to model with the hybrid type model, leave the
Include elastic properties from grid check box selected.
If it is cleared, the modeling is performed using only the
properties entered on the Input/output tab in the Volume
attributes dialog box. The remainder of the Grid properties
section is dimmed.
In this procedure, becal1se you want to model the seismic from
a stochastic inversion, leave the box selected.

Algorithm : IPP AVO. Aid & Richards • acoustic impedance, VpNs and density ... J []

Grid properties
~ lndude elastic properties from grid
~--------------------------~

Select a grid: @:) '-------------------------------'


.Acoustic impedance:

VpN s:

Density:

356 • Stochastic seismic inversion Quantitative Interpretation


6. In the Select a grid fie ld, insert ~ the model grid from the
Models pane whose properti es you wish to model.
7. Select the remaining property values, Acoustic impedance, Vp/
Vs, and Density, from the lists.
These properties should be the inverted properties that you
want to use to model the angular reflectivity that you convolve
with the wavelet to create the synthetic seismic.

Algorithm: IPP AVO, ftki .& Richards- acoustic impedance, Vp/Vs and density •I
Grid properties
[iJ lndude elastic properties from grid
Select a grid: Oass Grid _ _ =>=]
Acoustic impedance: fJ Class Grid SSI_Acoustic lrrpedance_1 [Uj
Vp/Vs: Class Grid SSI_Vp.rVs_1 [U]

Density: Class Grid SSI_Density_1 [U]

Wave!fl. - - - - - - - - - - - - - - - - - - - -

Select a wavelet : ~ k\ ... J~RZ_5-20


Scale: 1

Angle range

Angles defined as: ~~ Average angles ~1 Incident angles

Start angle : 5 degrees


End angle: 20 degrees
Effective angle: jn 112 degrees

Output
@ Pre~ntlcal reflections Post~cal refl ect1cns Both

8. Fill in the remaining fields on the tab with the wavelet to be


used in the convolution and the angle range you want to
model.
• Define angles as Average angles for most algorithms.
• For all Zoeppritz algorithms, define angles as Incident
angles.
• Specify an angle range for modeling. The effective angle is
reported in a read-only field.
9. 0 en the In u out ut tab.
Quantitative Interpretation Stochastic seismic inversion • 357
10. Insert ¢ three property volumes for Input 1, Input 2, and Input
3.
The SSI properties are embedded in these volumes. In the
example shown in the figure, the properties from a
simultaneous inversion have been chosen.
NOTE: If the Include elastic properties from grid option had
not been selected, only the property volumes on the Input/
output tab would be used for the model.

Input ~ tl Test Propert)• Model Sirnul lnversion P-impedance X


Input 2 ~ ~ Test Property Model Simul Inversion P/S velocity r. X

lnput3 ~ £1 Test Property Model Simul lnv.ersion Density X

Output: ~ Class G ·oReservoir AVO \1ode' X

11. To run the model, click Apply or OK


When the process completes, an entry for the synthetic volume
appears in the Input pane under the first property cube
entered.
~ ~ Test Property Model Simul Inversion
~ t; ~ Test Property Model Simullnversion P~mpedance
0 lnline 3457
0 XUne1383

" - ~ Oass Grid Reservoir AVO Model


0 lnline 3457
0 .XUne 2383
I> 0 Test Property Model Simul Inversion P-impedance Reflectivity
I> ~ 0 T§/Prop~ Mochl Simd nvtm:r7 P/5 m'ooty r~io
I> £II 0 Test Property Model Simul Inversion P/S velocity I?Jtio Reflectivity
~> f.;IO
I> 0 Test Property Model Simul Inversion Density Reflectivity

The (virtual) data can be viewed in any standard viewer. After


you close the Volume attributes dialog box, the entry in the
Input pane for the volumes persists and can be viewed at any
time by selecting it.

358 • Stochastic seismic inversion Quantitative Interpretation


The Reservoir AVO Model settings that created the volume can
be accessed from the Settings dialog box for the data volume.
View them on the Make virtual attribute tab.
Meta data J Op:aclt-; Make virt12l attribute

Attribute: Isotropic synlhetic


Output Output (1 of 1)

Algortthm: ( PP AVO. Aki & Richards - acoustlc impedance. VpNs and densly

Glid properties - - - - - - - - - - - - - - - - - - - -
0 lndude elastic properties from grid

Select a grid : ~I OassGrid

Acoustic impeda"lce: fJ Class Grid SSI_Acou.stic lrrpea-ar.ce_1 [U


VpN s: fJ Class Grid SSI_VptVs_1 [U:
Density: fJ Class Grio SSI_Density_1 (UJ
Wavelet

Select a wavelet: ~ 1 ~t\-. JP1_NearRZ_5-2 I


Scale:
~ ====~----------~
1
·
Allglerange - - - - - -- - - - - - - - - - - - - -

.Angles defined as : ~ Average angles E) Incident angles

Start angle : 5 degrees

End angle: 20 d~

Effective angle: In 11.! I degrees

Output
,·::> 0 re"\.'TlliCal reflectiOI"'S Post -critical reftectJor;s Both

You can apply further Volume attribute processing to the


existing virtual cube by running another Volume attributes
process using the pre-existing process (even if virtual) as input.
For example, the Volume attributes process allows you to add
spectral ly shaped random noise to provide a more realistic
seismic volume. Adding different amounts of noise can be used
as a way to test the detectability of subtle seismic signatures.

Quantitative Interpretation Stochastic seismic inversion • 359


Procedure- Add spectrally shaped random noise
This procedure shows you how to add random noise to the Volume
attributes process.
1. On the Quantitative Interpretation tab, in the AVO group,
click Volume attributes to open the Volume attributes
dialog box.
2. Select Reservoir AVO modeling as the Category and Spectrally
shaped random noise as the Attribute.
3. Select or clear the Realize check box as desired.
4. On the Input/output tab, complete these steps:
a. Insert ¢ the pre-existing virtual volume from the Input
pane to which you want to add the noise.
b. In the Output field, enter a name for the output data.
Use Category Albilule Reaize_

[{! Reservoir AVO modeling • Spectrally shaped rarcom noi~ • D

Input ~ tfl Class Grid ReseJVoir AVO '•1odel X

Output ~ Class Grid Reservoir AVO Model v.1th Ncise X

360 • Stochastic seismic inversion Quantitative Interpretation


5. Open the Parameters tab and complete these steps:
a. The noise generated by this process has a flat spectrum. In the
Noise spectrum section, select the Shape noise spectrum
using wavelet check box to filter the generated noise with a
wavelet. This wavelet can be derived from the data. In
practice, it can be any wavelet that you want to use to provide
shape to the noise spectrum.
b. Select a wavelet from the Input pane and insert it.
c. In the Noise amplitude section, enter values for the RMS
value of the signal (seismic) data and the noise.
Alternatively, the values can be estimated from the data
and the wavelet by clicking the Estimate buttons.
d. The S/Nfield is calculated automatically. Enter a desired
S/N ratio in this field and click in the RMS signal or RMS
noise field to recalculate the RMS noise level to be applied
as shown in this figure.
Input/output Parameters L-----------------
Noise spectrum

~ Shape noiee spectrum using wavelet

~ Ikt\.. JP1_NearRZ_5-20

Noiseamplitude - - - - - - - - - - - - - - - - -

RMS signal: 1$3.851519596144 ('¢=~ Estinatefrom i1U


RMS noise: 46.1814137485972 i¢=1 e.tinate from wavelet
SINr!!tio : 12 dB

e. Choose a random number Seed (or accept the default).


If you want to create alternative results when you click the
Apply button without entering a new value, select
Automatically generate seed. However, these runs are
not repeatable.
~~ ---------------- - - ----
LJ hrtomatJcaUy Qenerate seed
Seed:

~ -----------------------
~ Combine random noise with inptt data

Quantitative Interpretation Stochastic seismic inversion • 361


Oass Grid Ae;,ervoir AVO Model

Exercise 1 - Create a reservoir AVO model from


stochastic inversion results and add spectrally shaped
random noise
Use "Procedure- Use volume attributes and reservoir AVO modeling
to create seismic models from stochastic inversion results" and
"Procedure- Add spectrally shaped random noise" to create a
Reservoir AVO model from the stochastic inversion results and to add
random noise to the results.
Results
Figure 25 through Figure 27 show the results from performing the
procedures in this lesson.

Figure 25 Acoustic impedance property from a simultaneous inversion

362 • Stochastic seismic inversion Quantitative Interpretation


NOTE: The panel on the left in Figure 26 shows the very smooth SSI
result (lateral range about 2000 m). The panel on the right is from the
SSI result with very small range, about 500 m.

Figure 26 Simultaneous inversion acoustic impedance overlaid with the SSI


acoustic impedance result

Figure 27 Modeled seismic

Figure 28 Original seismic data

Quantitative Interpretation Stochastic seismic inversion • 363


Figure 29 and Figure 30 show the results of the noise modeling with
random noise added. S/N was set at 12 db.

Figure 29 Noise free synthetic seismic

Figure 30 Synthetic seismic with noise added at 12db SIN

364 • Stochastic seismic inversion Quantitative Interpretation


Review questions
• Explain the difference between simultaneous and stochastic
inversions.
• What extra information does a stochastic inversion provide?
• How might that extra information be used?
• What does a variogram measure?
• Explain the meaning of Nugget, Range, and Sill in a variogram.
• What information is needed to run a stochastic inversion?

Summary
In this module, you learned about:
• some of the differences between a simultaneous and
stochastic inversion
• how stochastic inversions are used
• how a stochastic inversion works
• creating a simple model for use in a stochastic inversion
• setting up and running a stochastic inversion
• quality checking the output
• calculating properties from the inversion output
• creating a Reservoir AVO Model from stochastic inversion
output, adding random noise to the model, and visualizing the
results

Quantitative Interpretation Stochastic seismic inversion • 365


Module 6 - Lithology
classification

Lithology Analysis and prediction is a rock physics based lithology


prediction workflow that integrates well logs, seismic inversion and
geological modeling and interpretation. It provides an estimate of the
most probable lithology and the uncertainty associated with the
prediction.

Prerequisites
To complete this module successfully, you must have completed the
previous modules in the course.

Learning objectives ~
~
After completing this module, you will know how to run the set of
processes performing lithology classification including:
• creating classification log
• lithology analysis
• lithology prediction

Quantitative Interpretation Lithology classification • 367


Lesson 1 -Introduction to the workflow and
classes definition
Reservoir properties estimation requires the following:
• Accurate 30 seismic inversion using the Simultaneous
inversion process (covered in Module 4). The seismic data is
inverted to acoustic and elastic properties.
• Rock physics integrated with geological interpretation to
generate prior estimate and relate the reservoir parameters to
the seismic parameters along the well using the Lithology
analysis process.
• Generation of lithology prediction volume and its associated
uncertainty using the Lithology prediction process.
These steps are represented in the flowchart in Figure 1.
Input.,. Md qu.liry control
."·.:ms~l~ic : .·,
" ~ . ~ \ .
', .l WeUlog~ ' ~
illll'ersion Vo1umes

.·~' ·Geolog:e~f': :-~


inter pretation

Creating c/as.siflc:,ation
data using weH log.s

Figure 1 The lithology classification workflow


368 • Lithology classification Quantitative Interpretation
After performing the inversion, Litho Analysis establishes the relation
between seismic related elastic measurements and physical rock
properties using Rock physics modeling. Petrophysical and elastic log
attributes are used to define lithology and/or fluid classification.
Both types of logs are used to generate statistical Probability Density
Functions (PDFs) which can be displayed in the PDF viewer window.
PDFs describe quantitatively our knowledge about the targeted rock
properties, the relations between rock properties, and the inherent
uncertainty (Figure 2).
'~lkPI•l
.. a II
.w
u 1S

...
-t
. .. 17 1.7 '""-
·I
f: 14t
0.) ...
- 11
~
... I!
"' - IJ

II
I
a ,.. r.W
I

P~p:pulil)

Figure 2 3D and 2D PDF viewer

It is common practice to establish the facies definition (here, termed


Jithoclass log) with crossplots between different logs looking for
cluster separation. However, additional methods can be used, such as
threshold-based classes using the Well log calculator or neural
network. These classes depend on the target; that is, the classes could
represent different facies, fluid types, fractures versus non-fractured
rock, etc.
The facies identification generally is done with one or a few key wells
where the data (such as, logs, cores, production data) and interpretation
are most complete. The following procedures describe different ways to
build this lithoclass log.

Quantitative Interpretation lithology classification • 369


Procedure - Create a classification using crossplots
This procedure shows you how to define lithology classes by analyzing
elastic properties (here, Vp, Vs and acoustic impedance (AI) logs) and
crossploting sediment properties such as porosity (PH IE), water
saturation (Sw) and clay content (VCL).
1. In a Well section window, display Well-1 and the logs as
shown in the figure.
2. Open the Tool Palette and identify the interval of interest by
creating a depth range (or a marker range).
t

P
t
· 0 MlrX4111
· ~~ Wtllt
Eil S.11mie
Fillar•foldl!
---~~~~~~·
• 0 Waveltlt
• el Su!factt_lOMtm
1!1 Hori.w"4 • 150ms
.., Horlzon-4
eJ Horl.ron 5 izon 4 Horizon
eJ Honzon6
eJ Horizon 6 •150ms
If Sm111 Polygon
•n ® CI'Oit IIC6ont
II·... 0 WtiH...NAM_NtarCrop_s-20
11-. 0 Wtll·l...NAM..MidCrop_l8·33
' 0 Wtli·I_NAM_FarCrop_31-15 izon 5
' eJ F1HtrJ

~ CniiVecllt well cllpth r... lzon 6 Hori zon

• Will conwlallon

... 9 11 150ms Horizon

370 • Lithology classification Quantitative Interpretation


3. If necessary, open a 01 crossplot, select Well-1, the AI, and the
VpNs ratio logs to separate the water saturated rock from the
gas saturated rock to identify a potential pay and, hence, the
pay interval. You can use Sw, PH IE, and VCL as the third
attribute to show that the sediments are not saturated in
water and have high porosity and a poor clay content.
Change the scale in the settings of the cross plot window if
necessary.

4. Open a new 01 crossplot, select Well-1 and display PH IE versus


VCL with Sw as the third attribute to identify the lithology
classes.
5. In the Collection settings, select Sw as the color code and
insert the depth range just created.
:::::~__,.\;;;;=;;;;;;;;;;;;lr=;p;;;:;;;;;i;;;;::;=;;;;;;;;~---,' Oimentions ( Fkl ~ ! kJ~

X: (PHIE
Y (va.

L ~~~~~----------------------'1
Pooi~ -------------------- 1

[t)Color [sw

Quantitative Interpretation Lithology classification • 371


6. From the Window tool bar, select a cross plot selection and
draw it on the crossplot if the different classes can be
separated distinctively.
NOTE: In this example, we use one crossplot to make the
classification, but you can use different types of crossplots to
define the different lithologies and fluids along the well.
~ u .».)lllt~
' . OS.iw)O n~ t »
~n e.M<Wt 'MI-I(CGII:icll1) 1
lblorl410'
~
t hi;
- wqlllil
l6fd
//#Wid
J 'f Fin ~
ft~N
t Dlpfl~1
~ B~lllllcliW

~It]
~~
~0
~0

~ qu

) ~
~&•­
i ~
~s-c~
Vc~~nt-...
~SJIIOflll'tl.ltl
03
tkDNicWili'ICIIYI
M,.•l!lld!l PH1Efm31m31

7. Change the names and colors of the selections, for example,


instead of Crossplot selection 2, name it Shale and choose
red as the color.
8. Highlight the crossplot selections just created, right-click on
them and select Create classification log.
NOTE: You also can highlight the selection directly from the
cross plot window and right-click to create the Classification
log.

372 • Lithology classification Quantitative Interpretation


9. In the settings of the new discrete log, on the Info tab, change
the name of the classes according to the crossplot selections.

Color.
Type· Global wei log (disc.)
Template: - 1111 Generol - . -, re1
.. ·dis - e- - - - - .

Color lable. Geneml discrele

He sand
.__ Date.

U ' uNOEF 1· 1 1· 1
Change selected rows
Pat!Bm. ~ Line color. [- ... I
Petrel filename: (obi';ct can1 be saved separately)
Orig. filename: (Made by PetreO I ./ Apply II ./ OK 11 1< Cancel

I . . Apply II
10. Quality check the new lithoclass log in a Well section
window.

Quantitative Interpretation Lithology classification • 373


Procedure - Create a classification using the Well log
calculator
This procedure explains how you can also establish a classification by
using empirical threshold for PH IE, VCL and Sw using the Well log
ca leu Ia tor.
1. In the Templates pane, under the Color tables folder, create
your own facies template by right-clicking on the Discrete
color tables folder and selcting Insert new discrete color
table.

" - Color tables


Seismic
Seismic .atlnbutes
Seismic inversion
Interpretation to imaging
Geological data
Other
" Bodies §to Import (on setec ion) , ..
r> • C~if~~~-;(;~~~~~~)---~~--~u ·~-- ·-
4 4

Boolean
Cased ,
Ce!l adivity ! Expand (recursive)
Completion~ ~I Sort by names
Connected Insert new cont inuous color -table

• Default disc ]
n;n rl::~~;t:ifir --
Insert new discrete color table
- -- - - - --

2. Edit the colors and names.

U lu NOff

Change selected rows


Pattem: 0::::r::: line color:

I ¥' Apply II v' OK II X Cancel

374 • Lithology classification Quantitative Interpretation


3. Now you must create the log template using this specific color
table. Still in the Templates pane, right-click on Discrete
property template and select Insert new property template.
Give it a proper name (for example, Li thoclass
template) and select the color table you just created for
the default color table.
General discrete c. • Collapse (recursive)
Facies ., !
Expand (recu rsive)
Facies Jupiter Sort by names
Ruvial facies
Sort by property templates
Bodies
Connected vola Sort by property t emplat es/ na me
Lithologies Au t o color all
Time stratigraph ~ Au•o name all
Zo~e log from A· -
[nsert new property template
Sbmgs
Main zones
Zones (hierarch
Zones ~ rr=:===r-::==--'""'-""- - - -- -, I
Layers
Fluids
Horizons
Faults
Segments
JJK Segments
,..t
JJKiayers
Boolean
Region U•e =n
1

Wei/Index I lc:an oolor

Cell activity I
Oab ult t .,lart:lblo
Dip classmcabon colllt':lenc
Q~lo/ ~--------------~
AVOCiass '1
Material Region
1
Default discrete 1 L_
Unti1!ed 1 l l l'---'---;:::::====~====:::;-.::===::::::;----' 1
I J /IWf ll ll CaAt•l

4. Open the Well log calculator and create your classes using the
Lithoclass template. In this example, we will use these
classes:
Shale= if ( VCL >= 0.3,1 ,0)
Class1_sand =if ( VCL < 0.3 And SW >= 0.8, 2,0)
Class2_sand = if ( VCL < 0.3 And SW < 0.8 And PH IE<= 0.22, 3, 0)
HC_sand =if ( VCL < 0.3 And SW < 0.8 And PH IE> 0.22, 4, 0)

Quantitative Interpretation Lithology classification • 375


5. Combine the classes and make a final class log. In this
example:
classlog=ShaletCiass1_sandtCiass2_sandtHC_sand

6. Quality check the new log in a Well section window.


W II 1 [TWT)
. • : :rl
. .

376 • Lithology classification Quantitative Interpretation


Procedure- Create a classtfication using Neural
network
Artificial Neural Nets are tools for automatically finding relationships
between multiple known parameters and a single unknown parameter.
In most cases, a Neural net is an adaptive system that changes its
structure based on what goes through the network during the learning
phase. In particular, the Neural net classification allows you to create a
classification model object (for example, facies from well logs). This is
facilitated by the Train estimation process in Petrel.
In this procedure, Neural network is used to identify the relationship
between the water saturation, the clay content, the porosity, and the
gamma ray logs.
1. In the Litho classification group, click Neural net.

litho

~
ana lysis :i-
Litho cfassific .•.

2. In the Train estimation model dialog box, click


Classification and select Well logs as the data type.

Quantitative Interpretation Lithology classification • 3n


3. On the Input tab, select Well-1 and the logs which will help
the classification.
NOTE: Before running Neural network, it is important to make
sure that the input data is correlated. The Correlation analysis
shows the correlation coefficient between each of the selected
logs. If the correlation is very low (close to zero), there is no
relationship between the data and it will be difficult to achieve
a good mathematical model that can be used for prediction.
:1- Train estimation model ~~
Tlllin estimetlon model I
~========================~
@ Create new Neural net
(() Edit existing: r-1~-~~-N-u
e_ ra_fne
-· t-,- - - - - - - - - - - -....--.1
@ ClassifiC31ion
<() Estimation

Data type: Well I • Method: Neur~~ l n •

j L: Input !C® Settings I~ Output l


Wells:
[] Use saved search
J None J0 r-g-..·
Logs:
--=~=-G-fo_be_fwe!-1-Jogs---==--===;,.
~ ~
~I I >IpO AI
L!_j 1 ; -so CALl
a -.+.! ~ Wells :- atpO DT_C
'· • ~ w e11-1 ;--ar.o DTS_syn
~-- 1 ~ GR
~--•nO NPHI
~OwTO One-vmy time 1
:- ~~ PHIE
f-~0 PHIT_C
; ·· cr 0
PR_syn
L-R0 0 RESDEEP
: ·Rs 0 RESSHAL
;-- PO RHOB_C
~- lsO SU
.__. ~sw

:-TwT0 TWT
:-· '4H~ va.
-c I "' I
I Principal component analysis I [ Correlalion analysis j

I .t Apply II .t OK I[ 1( Cancel I

378 • Lithology classification Quantitative Interpretation

I
4. On the Settings tab, it is recommended to select Supervised.
Purely statistical facies identification (Unsupervised) also can
be done, but the results will be more reliable if manual
geologic interpretation is provided (such as, existing facies
interpretation) and used as input-output pairs. These data pairs
are used to make a model that estimates the correct output
data when presented with the given input data. The idea is
that the same model can then be applied to similar input data
to compute reasonable output. The error in the network is
assessed by passing the training data's input through the
neural network and comparing it to the original data.

<0 Unsupervised 12

@ Supervised Fccies_interpretation

Training

Max number of iterations: 20

Error limit 10 %
Cross validation: 50 %

Probability threshold: 0

5. Set up the training parameters:


• Max number of iterations: The maximum amount of
times the training is run
• Error limit(%): When the number of points classified
incorrectly is below this limit, the model is assumed to be
trained and stops.
• Cross validation(%, supervised only): This parameter
is the percentage of the input data used to test the result
and give the error. The remaining part is used to train the
model.

Quantitative Interpretation Lithology classification • 379


6. Click Apply. The output is a log and a Neural net output that
contains input logs {found under Attribute) and output classes
{Class filter) .
.11 1- ~ Global wei/logs
0 Facies_interpretation
1111 0 Facies_interpretation (Neural net 1]

.. ~ Neural net 1
., :S· Aflribute
;( GR
~ff PHIE
sw
~H VCL
Facies_interpretation
DYII Facies_interpretation [Neural net 1]
.. T Class liNer
:ee: shale
::: Class 1 sand
::: Class 2 sand
••
.,.e; He sand

Exercise 1 - Create a classification


In this exercise, you create your own classification log.
1. Use "Procedure - Create a classification using crossplot" to
create your lithoclass log. Name it classLog. Additionally
you can review "Procedure - Create a classification using the
Well log calculator" to create you lithoclass log.
2. Display the result in a track of the Well section window. If
necessary, manually edit specific zones by using the Paint
discrete log option from the Well correlation Tool Palette.

380 • Lithology classification Quantitative Interpretation


Lesson 2- Lithology analysis
Now, you will use the facies classification to classify the well log data
that affects seismic wave propagation in the subsurface: compressional
and shear velocities and density logs.
One of the biggest challenge is the intrinsic variability of rock properties
(significant change across facies) shown in Figure 3. The reason is that
each facies does not correspond to a unique type of rock. The log
response is affected by changes in fluid, saturation, mineralogy,
temperature, and so on. To represent the intrinsic variability, a Monte
Carlo simulation is performed.

Figure 3 Example of intrinsic variability between two sandy shale units

Lithology analysis is a fully Bayesian approach that integrates different


measurements at different scales. Bayesian statistics has the unique
feature of requiring the specification of prior distributions for any
unknown parameters; here, the lithology layers contribution ratio from
the lithoclass log. Bayesian statistics re-evaluate the probabilities of an
event based on empirical observations.

Quantitative Interpretation Lithology classification • 381


Visualization of the litho Analysis results
The outputs of the Litho analysis process are the probability density
functions (PDFs). A PDF can be defined as a shape which describes the
probability of an event occurring within a portion of a continuous space.
This displays the facies from the lithoclass log versus the well logs
input.
In other words, analyzing the example in Figure 4, the event named HC
Sand has the highest probability of occurring around 4000 kPa.s/m in
the Acoustic impedance (or P impedance) space.

l Ex lO
-I
'-
'
'
-~­
doss2snl
• hcsnl
·I
14><10

-I
12lll0

·I
1Cx10

t
f sexJ0 2
,;:

·l
6Cx10

·l
4Cx10

·l
2Cx10

OCxlO
0

l
<•lx10
l
6Cx10
3
80x10
I
10.10
I
12x10
.
14x1C

Figure 4 Litho Analysis results of the P-impedance attribute in 10 plotting mode


(univariate)

382 • Lithology classification Quantitative Interpretation


The less overlapping you have between the different facies, the better
the attribute is able to identify the differentiated classes. By analyzing
the example in Figure 4, we can see that the Acoustic Impedance helps
us discriminate the HC sand property from the other data. In the VpNs
ratio space, we have a slightly different scenario where we can
differentiate two groups; the HC sand & sand and the shale & sandy
shale as shown in Figure 5.

t
ICxiO

Hx iO~

H-x10° l-.........::= ---==::::....___ _~_;:,.---------==----l


1
1!)10° 18.rl0: 1!b:1J~ 2.01:10 l'b.I0° HriOC 2.k l 2h l0° l b.!Oc 32x:ll'
PIS.W,IIIIo(J

Figure 5 Litho Analysis results of the Vp/Vs ratio attribute in 1D plotting mode

Quantitative Interpretation Lithology classification • 383


The results also can be interpreted in the 20/30 space (Figure 6) where
the data can be plotted against two or three attributes (multivariate).
In the example in Figure 6, the event HC sand has the highest
probability of occurrence. Around 4 kPa.s/m of Acoustic impedance and
between 1.9 and 2.1 of VpNs ratio the probability of having HC sand is
very high, more than 90% .

o.f1
~· J
o.A

'-,, oJ
' '
o-0

,,.
~ ,
.,. , ""
,....,.

'~
.....
"'').

"'~
..,.-.,
I•"'"'
. Httu.d I
Figure 6 Results in 20 plotting mode

384 • Lithology classification Quantitative Interpretation


However, viewing from above in the 20 plotting mode, we can see that,
in the area defined in the section above, it will have the probability to
belong to other lithology as well (shale in Figure 7). Hence, the less
overlapping you have between the different facies, the more confidence
you have in interpreting your data.
I



t

Figure 7 View from above of the 20 plot

Quantitative Interpretation Lithology classification • 385


The PDF confusion matrix statistics
After running the Litho analysis, you analyze the results and
performance of the Bayesian classification through the confusion matrix
table. You can identify if the algorithm has problems when defining the
different classes according to the input data; in other words, if the
process is getting confused while classifying the data.
In the Litho analysis dialog box, the QC tab contains two tables
(Figure 8):
• The first table represents the probability of classifying a
sample of belonging to class I, given that its true class is J
(Prediction I True).
• The second table represents the probability of the true class of
a sample being class I, given that it was predicted to have
class J (True I Prediction).
To get a high-confidence classification results, the values along the
diagonal of this matrix should be as large as possible.

~fin~ QC
!PDF confusion matnx statistics (Facies vs prediction)
.,.. P (Prediction I True)
~
Focies neme shale clllss 1 sand ci8SS2 sand he sand
shale
- f74.33%
- 144.39%
- 03.16% Too.63%
c~ss1 send 20.84% ,44.04% 00.90% 101 .91%

dcss2sand 02.95% 03.52% 75.61% 10.47%


hcst~nd 101 .88% j08.06% 20.32% 86.99%
. P (True I Prediction)
:J
Focies Mme shale clllss 1 sand ~2scnd he send
- ·- -
sht~le 60.67% 30.79% 03.19% 01 .60%
dt~ss 1 send 36.23% 65.06% 03.80"/o 06. 8r~o
dcss2scnd 02.58% t01.33% 81.69% 17.33%
he send 00.52% !02.82% 11.31% 74.19%
.,.. Classifier (and lithodcss) error and warning messages i]1
No errors detected - press ac to recompute

Figure 8 The PDF confusion matrix tab

386 • Lithology class1ficat10n Ouant1tat1ve InterpretatiOn


Procedure- Generate the litho classification model
In this procedure, you use the lithoclass log created previously to
generate PDFs from the cluster analysis of the log data as a
representation of the variability in the formation properties.
1. Open the Litho analysis dialog box from the Quantitative
Interpretation tab.
I Iii
litho fJ
analysis :!-
litho dassific ...

2. Define the name of your litho classification model.


3. On the Definition tab, select the Litho classification log to be
used, define the classes which are going to participate in the
classification and set up the prior probability values to modify
the lithology layers contribution ratio.
NOTE: Without geological knowledge, you would generally
leave the default parameters which give the facies an
equiprobable chance of occurring. However, it is recommended
to get information from seismic (for example, by performing
structural interpretation) to get the lateral information.
4. Select the well logs containing the elastic properties and
define:
5. The number of bins: How many bins will be used when
forming the histogram for the attribute. A high number of
bins can significantly slow the process.
6. The upscaling factor: The increase in resolution to apply
when interpolating histograms.
7. The inversion error factor: How much smoothing to
apply to the upscaled histograms.
NOTE: You must compromise between the discretization
(number of bins) and the smoothing. Too many bins would limit
the results to this specific sample when too much smoothing
will not capture variability and will affect discrimination. It is
recommended to run a number of tests to choose the
parameters.
8. At the bottom section, select the well from the Input pane and
the top and base markers defining the interval of interest.

Quantitative Interpretation Lithology classification • 387


~ ~,.00 acl
... Ctess definition [J
L11hod8ss log . ..;> ~ dasslog I
Closs name I
Prior probebility (relative) Enabled
shale 0.25 M
dass 1 sand 0.25 [ .{]
dass 2 sand 0.25 !0__
he sand 025 [.()
... Select global well logs [J
.JI ~-
Global well log_ ~mber of btns ~ling h2ctor I Inversion e~ ~~r
1~ lip Al 10 4 J7

~~fCVPVS 1 10 14 17

~ I P RHOB 1 10 14 17

... Select wells Cl


~
Weil-l I

Top merk:er. ~-Merk:er4 I


Sese merker. ~~Merker7 I

9. Click QC, and go to the QC tab to analyze the PDF confusion


matrix statistics. A PDF Viewer window is automatically
opened.
10. Select the facies you want to display from the Input pane.
" Litho Analysis
shale
class 1sand
class2sand
he sand

11. Under Window, on the PDF viewer tab, try the different
plotting mode. Depending on the plotting mode selecting,
different options are enabled.


View Cane'a
all align en: •
View

• 'X
3 a::ributes node

388 • Lithology classification Quantitative Interpretation


Exercise 1 - Generate the litho classification model and
generate PDFs
In this exercise you learn how to use the Litho analysis process, quality
check the results using the PDF confusion matrix, and use the PDF
viewer.
1. Use "Procedure- Generate the litho classification model".
2. Tile different PDF viewer windows and play with the different
plotting mode to analyze how the HC sand and the shale can
be differentiated with the different attributes .

• sliM
• sandyshllle
• gnd

• HcS!IId

J J 4
4.0x10 8.0x1.0 1.2x10
1'-ftpedance [kPa.s"m)

Quantitative Interpretation Lithology classification • 389


3. Go back to the QC tab in the Litho analysis dialog box and
analyze the percentage. Start with the P (Prediction I True)
table.

He sand

• How many of the samples belonging to Shale were


predicted as a HC sand ?
• How many of the samples belonging to sandy shale were
predicted as a HC sand ?
• How many of the samples belonging to sand were
predicted as a HC sand ?
• Looking at the HC sand prediction, do you think you can be
confident in the classification results?
4. Now observe the percentages for the P (True I Prediction)
table.
• P(True I PredK:ti:Jn)

shale sandy shale sand


60.62% 01.60%
3C .J1% 06.87%
(IL. ~S% 17.34%
00.51% 74.19%

• What is the probability of the samples belonging to HC


sand class for being classified as shale, sandy shale or
sand?
• Again, looking at the HC sand prediction, do you think you
can be confident in the classification results?

390 • Lithology classification Quantitative Interpretation


Lesson 3- Lithology prediction
The PDFs generated from the Litho analysis process are derived from
the well logs. The aim of the Lithology prediction is to apply the PDFs
to the elastic attributes (generated using the Simultaneous inversion)
and generate the classification volume and the associated probability
volumes for each class.
The Litho prediction process uses a classifier to form Maximum A Priori
(MAP) lithology predictions (Kay, 1993) on inversion data. In other
terms, the different data at one point is collected into a sample vector.
Then, it evaluates each of the different probability density functions at
the sample vector. The PDF that is the largest for this vector is chosen
as the predicted lithology class for that point.
The scales of the seismic attributes and well logs are very different.
The layer thickness is typically within 1/4 to 1/10 of the wavelength.
That is why the new PDFs derived by the well logs are downscaled to
the seismic resolution. The example in Figure 9 shows clearly the
difference in scale.

Figure 9 Difference in resolution between the porosity log and the seismic cube

Quantitative Interpretation Lithology classification • 391


Applying the classifier to the data generates one output object
representing the most likely classification (Figure 10). It also creates
one probability for every classes of the classifier. In case of seismic
input from Inversion results, we produce lithology or fluid prediction
volumes and their associated uncertainties.
The lithology and fluid prediction volume can be compared to the
original seismic.

Figure 10 Example of most likely classification from seismic input

You can apply the classifier to seismic cubes, wells, and pillar grids. The
input data must be collections of rock parameters of the same type.

392 • Lithology classification Quantitative Interpretation


The use of the Geobody interpretation probe capabilities is a good way
to understand and quality check the results. It allows you to analyze the
temporal and spatial extent of the various classes. You can extract
geobodies from the probability of each defined class or from the
classification cube (Figure 11 ).

Figure 11 Geobody filtered on the HC Sand class

Quantitative Interpretation Lithology classification • 393


Procedure - Use the Litho prediction process
This procedure shows how to apply the output from the Litho analysis
process to the Litho prediction process. The goal is to apply the POFs to
the elastic attribute cubes from the seismic inversion and to produce
lithology or fluid prediction volumes and their associated uncertainties.
1. Open the Litho prediction dialog box from the Quantitative
Interpretation tab.
lil
Litho I!J
analysis :f..
Litho classffic ...

2. Insert the classification model previously created in the Litho


analysis process, define the input type (in this case seismic),
and the template to use.
NOTE: The output cubes require a continuous template. The
template can be created by clicking the Create template from
Classlog used by classifier button (this will be saved in the
Continuous other template folder in the Templates pane.

Cl8ssifier: ~ New test


I
Input type: Seismic

Template: ~~==~~~~==========~=========-~~
classLogGenDisPetrel_Template

394 • Lithology classification Quantitative Interpretation


3. Select the input volumes (here the results from the
Simultaneous inversion) from the Input pane. The selection
should be performed according to the properties in the Litho
classification model.
4. If you want to apply the prediction only over a specific zone of
the seismic volumes, insert the top and base surface to use.
5. Click OK

Attribute mapping

Attribute Input
P-impedance +9! Slmlnv 30. P-impedance {Rel!lized} 1 I
P/S velocity ratio +lJ Slmlnv 30. P/S velocity ratio [Rel!lized} 1 I
Density I+ mJ Slmlnv_30. Density [Realized) 1 I

Zone

fl] Apply to range


Topsurfuce: m'"-H_o_Iiz_o_n_4_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ____,
~
Bottom surface: 91@ Horizon 6 - 150ms
[ ~ Apply ok ] [x eance1 ]

Quantitative Interpretation Lithology classification • 395


The results will be located in the Input pane under the first
seismic cube listed in the Attribute mapping table of the Litho
prediction process.
A ~ 3D~MEM~Survey
.. [)I~ Slmlnv~30.
..£I~ Slmlnv_JD. P-impedance [Realized] 1
D lnline 3457
D Xline2383
D Z=-2604.00
A f!ii ~ · Litho cube classification Newtest
.. tJ ~ C!assiffcationxx
~ lnline3457
~ Xline2383
"'tiD shale
D lnline3457
0 Xline2383
. . tl 0 sandysha/e
0 lnline 3457
·D Xline 2383
"'tiD sand
0 lnllne3457
0 Xline 2383
.. tiD Hcsand
0 lnline3457
D Xline2383

396 • Lithology classification Quantitative Interpretation


Exercise 1 - Use the Litho prediction process
The objective of this exercise is for you to gain familiarity with the Litho
prediction process.
1. Use "Procedure - Use the Litho prediction process" to
generate a litho classification cube and the probability cubes
derived from the PDFs created from the Litho analysis process .

.
>J0-6(Camo<eiNcf IAnYf X

-
.,_,_

~~

2. Identify the different output cubes and display the


Classification cube and the well in a 30 window.
According to the color code used in the Litho analysis
process, can you identify a pay section crossing the well?
New test
shale
sandy shale
sand
He sand

3. Now display the Shale cube.


What is the probability to find shale in the productive section
identified earlier?

Quantitative Interpretation Lithology classification • 397


4. Insert a Horizon probe from the Seismic interpretation tab.
In the settings of the Horizon probe, set it between Horizon 6
and Horizon 6 -150 ms for the new Classification seismic
volume.
5. Play with the opacity to extract the HC Sand class and display
the filtered geobody with the original seismic.

Review questions
• What is the most common way to establish classification and
differentiate facies from well logs?
• What is the main input of the Litho analysis process?
• Why do we use the Bayesian approach?
• Which window do you use to visualize the PDFs? How many
plotting modes are available?
• What is the use of each of the PDF confusion matrix?
• What is the main input of the Litho prediction process? How do
you use the different output cubes?

Summary
In this module, you learned about:
• the different methods to create classification from well logs
• running the Litho analysis process and quality check the results
with the PDF confusion matrix
• running the Litho prediction process and identify potential pay
from the results

398 • Lithology classification Quantitative Interpretation


Appendix A- References

These documents and the online help provide significant support for the
tasks detailed in this training.

Bibliography
1. Backus, M 1962. Long-wave elastic anisotropy produced by
horizontal layering: J. Geoph Res., 67, 4427.
2. Batzle, M. and Wang, Z. (1992) Seismic Properties of Pore
Fluids, Geophysics, 57, 1396-1408
3. Castagna, J. P., Batzle, M.L., and Eastwood, R. L. 1985,
Relationships between compressional-wave and shear-wave
velocities in clastic silicate rocks, Geophysics 50, 571-581
4. Greenberg, M. L. and Castagna J. P., 1992, Shear-wave velocity
estimation in porous rocks: Theoretical formulation, preliminary
verification and applications: Geophysical Prospecting, 40,
195-209.
5. Gardner, G.H.F., Gardner, L.W., and Gregory, A.A. 1974,
Formation velocity and density-the diagnostic basics for
stratigraphic traps, Geophysics 39, 770-780.
6. Gassmann, F., 1951,Uber die elastizitat poro sermedien:
Vierteljahrsschrift der Naturforschenden Gesellschaft in Zurich,
96, 1-23.
7. Bill Goodway, Taiwen Chen, Jon Downton, Improved AVO
fluid detection and lithology discrimination using Lame
petrophysical parameters; "A,p", "Jlp", &""-/11 fluid stack",
from Pand S inversions. SEG Technical Program Expanded
Abstracts 1997: 183-186.
8. Zoeppritz, K., 1919, Erdbebenwellen VIII B, On the reflection
and penetration of seismic waves through unstable layers:
Goettinger Nachr., pages 66-84
9. Aki, K. and Richards, P. G., J979, Quantitative Seismology, W.H.
Freeman and to. P 153
10. Castagna, J.P., and Swan, H. W., 1997, Principles of AVO
Crossplotting. The Leading Edge, 16, 337-342
Quantitative Interpretation References • 399
11. Fatti, J.L., Smith, G.C., Vail, P.J., and Levitt, P.R., 1994,
Detection of gas in sandstone reservoirs using
12. AVO analysis: a 3-D seismic case history using the Geostack
technique: Geophysics, 59, 1362-1376.
13. Pan, N. D. and Gardner, G.H.F., 1987, The basic equations of
plane eJastic wave reflection and scattering applied to AVO
analysis. Annual Progress Review 19, Seismic Acoustic
Laboratory, University of Houston.
14. Rutherford, S. R. and Williams, R. H., 1989, Amplitude-versus-
offset variations in gas sands: Geophysics, Soc. of Expl.
Geophys., 54, 680-688.
15. Shuey, R.T., 1985, A simplification of the Zoeppritz equations,
Geophysics 50,609-614.
16. Smith, G.C. and Gidlow, P.M., 1987, Weighted Stacking for
Rock Property Estimation and Detection of Gas, Geophysical
Prospecting 35,993-1014.
17. Gidlow, P. M., Smith, G. C., and Vail, P., 1992, Hydrocarbon
detection using fluid factor traces: Joint SEG/EAEG Summer
Research Workshop, Technical Program and Abstracts, 78-89.
18. Li, Y., Downton, J., Xu, Y., AVO Modeling in Seismic Processing
and Interpretation. Part 1: Fundamentals. CSEG Recorder,
December 2003, 42-52
19. Mahdi H. Almutlaq and Gary F. Margrave, 2010, Tutorial: AVO
inversion CREWES Research Report- Volume 22 (201 0)
20. Ross, C. P., and Kinman, D. L., 1995, Nonbright-spot AVO: two
examples. Geophysics, 60, 1398-1408
21. Verm, R. and Hilterman, F., 1995, Lithology color-coded seismic
sections: The calibration of AVO crossplotting to rock
properties. The Leading Edge No.8 August 1995, 847- 853
22. Walden, A. T., Making AVO sections more robust: Geophysical
Prospecting, European Association of Explorat~on
Geophysicists, val. 39, pp 915-942, October 1991

400 • References Quantitative Interpretation


Additional useful information
These texts are useful references for additional reading:
1. Seismic Reflections of Rock Properties, Jack Dvorkin, Mario
Gutierrez, and Daria Grana. Cambridge University Press, 2014
2. Seismic Amplitude, An Interpreter's Handbook. Rob Simm and
Mike Bacon. Cambridge University Press, 2014

Quantitative Interpretation References • 401

You might also like