First Steps in Seismic Interpretation [D.A. Herron, 2011] @Geo Pedia .pdf

Geophysical Monograph Series

Number 16

First Steps in Seismic Interpretation Donald A. Herron

Rebecca B. Latimer, managing editor

Tulsa, Oklahoma

Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/

SEG-SEISMIC-11-0601-FM.indd 1

05/12/11 7:44 PM

ISBN 978-0-931830-56-3 (Series) ISBN 978-1-56080-280-8 (Volume) Society of Exploration Geophysicists P.O. Box 702740 Tulsa, OK 74170-2740 © 2011 by Society of Exploration Geophysicists All rights reserved. This book or parts hereof may not be reproduced in any form without written permission from the publisher. Published 2011 Printed in the United States of America

Cover background image courtesy of Thomas H. Wilson

Library of Congress Cataloging-in-Publication Data Herron, Donald A., 1949 First steps in seismic interpretation / Donald A. Herron ; Rebecca B. Latimer, managing editor. p. cm. -- (Geophysical monograph series ; no. 16) Includes bibliographical references and index. ISBN 978-1-56080-280-8 (volume : alk. paper) -- ISBN 978-0-931830-56-3 (series : alk. paper) 1. Seismology. 2. Geophysical surveys. I. Latimer, Rebecca B. II. Title. QE534.3.H47 2011 551.22--dc23 2011047720



23/12/11 1:55 PM

Contents About the Author. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

v vii ix

Chapter 1: Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

Chapter 2: Seismic Response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9

Chapter 3: Seismic Attributes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Amplitude. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Coherence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Inversion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

21 22 28 30

Chapter 4: Velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sonic logs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Well-velocity surveys. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Seismically derived velocities. . . . . . . . . . . . . . . . . . . . . . . . Velocity anisotropy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Time-depth conversion. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35 36 38 41 56 57

Chapter 5: Migration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

Chapter 6: Resolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75

Chapter 7: Correlation Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 First look. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 Horizons versus faults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 Multiple reflections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 Manual versus automatic tracking . . . . . . . . . . . . . . . . . . . . 96 Artifacts and interpretation pitfalls. . . . . . . . . . . . . . . . . . . . 105 iii Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:44 PM

iv

Chapter 8: Correlation Procedures. . . . . . . . . . . . . . . . . . . . . . . . . . . Getting started. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Loop tying. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jump correlation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Correlations in depth-migration projects . . . . . . . . . . . . . . . Visualization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Interpretation processes and work flows. . . . . . . . . . . . . . . .

115 115 120 133 140 145 149

Chapter 9: Data Quality and Management . . . . . . . . . . . . . . . . . . . . 153 Data quality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Data management. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 Nomenclature systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Chapter 10: Other Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 Gridding and contouring. . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 4D seismic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 Seismic modeling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Interpretive judgment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Curiosity and interpretive thinking. . . . . . . . . . . . . . . . . . . . 170 The interpretation paradox. . . . . . . . . . . . . . . . . . . . . . . . . . 174 Approximations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 Uncertainty and risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 The workstation environment. . . . . . . . . . . . . . . . . . . . . . . . 178 Ergonomics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 Presentations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 Career development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 Advanced interpretation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 Time spent and value added. . . . . . . . . . . . . . . . . . . . . . . . . 185 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193



05/12/11 7:44 PM

About the Author Don Herron received a bachelor of science degree (with honors) in geological sciences from Brown University in 1971 and a master of science degree in geological sciences from the California Institute of Technology in 1973. He enjoyed a career as a seismic interpreter at Texaco (1973–1977), Gulf (1977–1984), and most recently Sohio/ BP (1984–2008). Since retirement in 2008, he has worked as an independent geophysical consultant for Petroleum Geo-Services (PGS) as a geosciences advisor, and with several oil companies as a seismic interpretation instructor. At Gulf and Sohio/BP he taught in-house courses in seismic interpretation and was co-instructor for the SEG Continuing Education course “Seismic Interpretation in the Exploration Domain” (1995–2007). He was a member of the Editorial Board of The Leading Edge (2002–2007, chairman in 2006–2007) and is author of the bi-monthly “Interpreter Sam” column in The Leading Edge. He is an active member of SEG, AAPG, and Sigma Xi.

v Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:44 PM



05/12/11 7:44 PM

Preface This book begins with an introduction that is more philosophical than technical, followed by five chapters on fundamentals of reflection seismic (titled Seismic Response, Seismic Attributes, Velocity, Migration, and Resolution). The gist of what I really have to say about the correlation of seismic records is in Chapters 7 (Correlation Concepts) and 8 (Correlation Procedures). Chapter 9 (Data Quality and Management) certainly should not be glossed over, and Chapter 10 (Other Considerations) contains my thoughts on several worthy topics that do not fit neatly into any of the preceding chapters. In large part, this book is a compilation of notes from seismic interpretation courses that I’ve had the good fortune to teach over the past three decades. Because I’ve assumed that readers are familiar with basic concepts and principles of geology and reflection seismology, the book is best viewed as a synthesis rather than a fundamental treatment of those concepts and principles. When I use the expression “geologically reasonable” to qualify interpretation results, which I do throughout the book, I mean “reasonable” in the sense of “analogous to known geology” or “consistent with known geology or sound geologic models” or “within the context of expectation or realization of some geologic concept or model.” I certainly don’t intend this book to be the definitive primer on interpreting reflection seismic data or a comprehensive treatise on the latest in correlation tools and techniques; rather, I’m seeking to give voice to a concern about “this particular art” that I’ve had ever since my first foray into interpretation in the early 1970s. My concern is founded on a statement by a man from whom I had the privilege to learn about exploration geophysics in the classroom and in the field. In his own book he wrote that “the correlation procedure itself is of such a nature that it can hardly be adequately described in a book.” Well, with the utmost respect for that man, here goes.

vii Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:44 PM



05/12/11 7:44 PM

Acknowledgments I thank Rebecca Latimer, Bill Barkhouse, Bruce Hart, and John O’Brien for their constructive reviews of my manuscript and also BP (Amal Ray and Tim Summers), PGS (Nathan Oliver), TGS (Tom Neugebauer), and WesternGeco (Lee Hooper) for permission to include data and images from their companies in this book. I thank Mike Schoenberger for sharing his characterization of seismic data quality with me; it’s the most concise and practical description of data quality I’ve ever known, so I’ve used it to set context throughout the book. I extend my thanks also to members of the SEG publications and graphics groups in Tulsa, in particular Jennifer Cobb and Kathy Gamble, without whose skill and patience this book could not have come into being. I’m especially grateful to Kathy Pile and Gary Stewart, whose editing gave my text the clarity and consistency it needed. In creating this book, I’m indebted to countless geoscientists, old and young alike, from whom I’ve learned so much over the years. Among all those talented men and women, I owe the most to Tim Smith, perhaps the most insightful interpreter I’ve ever known and an excellent teacher as well, with whom I’ve had the distinct privilege numerous times to share the front of a classroom.

ix Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:44 PM



05/12/11 7:44 PM

Interpretation is telling the geologic story contained in seismic data. It is correlating the features we see in seismic data with elements of geology as we know them. The story is read from a book having many chapters, some of which are either illegible or unintelligible, and others are lost or yet to be written. And although the story doesn’t always have a happy ending, only in its telling do we expand our knowledge. —Interpreter Sam



05/12/11 7:44 PM



05/12/11 7:44 PM

Chapter 1 Introduction Accurate interpretation of geophysical data — in particular, reflection seismic data — is one of the most important elements of a successful oil and gas exploration program. Despite technological advances in data acquisition and processing and the regular use of powerful computers and sophisticated software applications, you still face a tremendous challenge each time you begin to reconstruct the geologic story contained in a grid or volume of seismic data — that is, to interpret the data. On occasion, this interpretive tale can be clearly told; but most of the time, each page of each chapter is slowly turned, and rarely is the full meaning of the story completely understood. Where the correlation of one reflection record with another is very easy, little needs to be said. Almost anyone can understand such a correlation. On the other hand, this is a rare occurrence. The usual thing is for the correlation to be so difficult as to be impossible. It is for this reason that correlation procedure can hardly be described in words (Dix, 1952). Although Dix is speaking about the correlation of individual reflection records, which were used routinely before the advent of continuous common-depth-point (CDP) profiling, he clearly recognized the essence of interpretation as the considered extraction of geologic information from indirect geophysical measurements. His words are no less relevant and applicable now than they were 60 years ago, even in view of the high standards of data quality made possible by advances in seismic acquisition and processing, to say nothing of accompanying developments in interpretation technology. In the modern interpretation environment, you still face correlations that are “so difficult as to be impossible” because these correlations

1 Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/

SEG-SEISMIC-11-0601-001.indd 1

05/12/11 2:04 PM

2 First Steps in Seismic Interpretation

define the frontiers of opportunity, the ones posing the sternest challenges and ultimately leading to the greatest rewards. The primary aim of this book is to describe Dix’s correlation procedure in terms of the science, data, tools, and techniques now used in seismic interpretation in the oil and gas industry. As an individual geoscientist, you develop and apply your own approach and style when interpreting seismic data. You continually revise and refine correlation procedures during the course of your career and expand them as you complete different interpretation projects. With experience, you learn to check and recheck the validity of your procedures to fully understand the rules of evidence that govern their use: • What are the physical laws that control the phenomena you observe and consider as evidence? • What are the uncertainties in your evidence? You must have a good understanding of seismic acquisition and processing principles as well as fundamentals of geology before beginning to collect interpretive evidence and solve interpretation problems correctly. Continuing from Dix, then, you must also know when enough interpreting is enough: The threshold of impossibility is reached by different interpreters at different levels. The important thing is for each interpreter to understand his limitations. Obviously it is foolish to go ahead and correlate when no correlation is possible. This involves giving a definite interpretation that is almost sure to be misleading and therefore very expensive (Dix, 1952). The primary goal of seismic interpretation is always to describe geology, and all aspects of interpretation facilitate and support this goal. The products of seismic interpretation are an important subset of the indispensable elements used by geoscientists to define and develop oil and gas prospects. Although seismic interpretation is a very important part of the explorationdevelopment-production stream, it is only one of the elements used when integrating all available data to build a geophysically consistent and geologically reasonable picture of subsurface structure and stratigraphy. Drawing this picture accurately is a critical factor in successful identification of drillable prospects and exploitation of known hydrocarbon accumulations. Interpretation, the description of geology, depends critically on seismic data quality: The better the quality, the more accurate and reliable the interpretation. In the most general terms, quality is the degree to which



05/12/11 2:04 PM

Chapter 1: Introduction 3

something fulfills its intended purpose; because you use seismic data for different purposes, depending on where you are in the value stream (e.g., exploration versus production), you know that data quality appropriate and acceptable for one project may not be for another. For example, the quality of a high-resolution seismic survey used to detail the shallow subsurface and identify potential drilling hazards would be completely inadequate and essentially useless for deep exploration. In the same way, a 3D survey purposely acquired and processed to image deep subsalt targets would have little or no value for shallow hazards assessment. At the same time, quality may be less than optimal owing to problems in data acquisition or processing, and you need to be able to recognize these shortcomings, seeking advice from acquisition and processing specialists as needed, accounting for the shortcomings during interpretation, and making appropriate recommendations for improvements. There are three primary elements of seismic data quality: detection (signal-to-noise), resolution (temporal and spatial), and image fidelity (focusing and positioning). All efforts in seismic data acquisition and processing are designed to optimize data quality and “interpretability.” You are responsible for assessing data quality for each of your interpretation projects and for communicating this assessment as part of any presentation of project results. Seismic interpretation is, by the nature of seismic data and the earth itself, nonunique and highly subjective. You bring your perspective and powers of observation to bear on the interpretation problem at hand, the effects of which cannot be clearly identified in or separated from your maps and calculations — and yet are a controlling factor in your results. Stephen Jay Gould recognizes and appreciates the importance of talent for observation in naturalists, which can easily apply to interpreters: All field naturalists know and respect the phenomenon of “search image” — the best proof that observation is an interaction of mind and nature, not a fully objective and reproducible mapping of outside upon inside, done in the same way by all careful and competent people. In short, you see what you are trained to view — and observation of different sorts of objects often requires a conscious shift of focus, not a total and indiscriminate expansion in the hopes of seeing everything. The world is too crowded with wonders for simultaneous perception of all; we learn our fruitful selectivities (Gould, 1993). Although acquiring, processing, and analyzing seismic data are mathematically intensive and now almost exclusively digital, interpretation



05/12/11 2:04 PM


activity per se is still primarily a visual (human and therefore fallible) process. Correlation of seismic records involves pattern recognition, depending heavily on the display of data and your knowledge and understanding of patterns in geology. Interpretation of any element of geology from seismic data involves answering the questions “What is it?” and “Where is it?” — answers that are rarely independent of each other. In other words, you often interpret what something is by where it is in relation to other features, or where and how large a feature should be because of what it is. Hence, we confirm the importance of migration of seismic data and, ultimately, the ability to visualize and reconstruct in depth what is only indirectly measured in time. Of course, it goes without saying that you will not be too terribly successful in the oil and gas business if you can’t accurately specify what, where, and how big your exploration targets are. Seismic acquisition, processing, and interpretation are related, as shown in Figure 1. Acquisition and processing can be thought of as forward processes in which acoustic-impedance contrasts in the subsurface produce measurable seismic responses (acoustic impedance [AI] and reflection coefficient [RC] are defined in Chapter 2). The interpretation of this response, which in Figure 1 is called “ideal” but often is very far from being so, is an inverse process that describes the original AI contrasts and ultimately the subsurface geology. Notice that the forward processes of acquisition and processing can give rise to different, nonunique responses, depending on the particular acquisition and processing techniques used. This is another way of saying that acquisition and processing determine data quality. The inverse process of interpretation can result in many different descriptions of geology, again because of varying data quality and also because the fundamental relationships among subsurface geometry, acoustic impedance, and geology are nonunique. In your better humors, you thank your good fortune for this nonuniqueness because it is an important factor contributing to your job security. Your domain of information in interpretation consists of facts (there may not be as many of these as you would like to believe), observations, inferences drawn from observations and their resultant models, and, of course, experience gained from having established facts, made observations, drawn inferences, and revised models over time. Taken together, these still represent a relatively small volume of your domain, the largest portion of which is the unknown. Accurate, well-integrated interpretations can reduce the volume of the unknown, but only if you maintain awareness of the distinctions among facts, observations, and models, all of which can be considered interpretive “evidence.” This awareness is a critical element in your assessment of technical risk in exploration projects, which, contrary to



05/12/11 2:04 PM

Chapter 1: Introduction 5 Acoustic Reflection Ideal Lithology impedance coefficient seismic response – + +

No depth scale implied

No time scale implied

Data acquisition and processing Interpretation Figure 1. The interrelationship of seismic data acquisition and processing with seismic interpretation. The former are forward processes, and the latter is an inverse process. AI = acoustic impedance; RC = reflection coefficient.

the way you would like it to be, is at best as subjective as the interpretation on which it is based. Perhaps the most common intellectual difficulty encountered in correlating seismic data is maintaining a clear distinction between observation and interpretation (see Figure 2). Observation is the essential foundation for meaningful interpretation; think of observation as “What do I see?” and interpretation as “What does it mean?” These questions can be easily and often unwittingly confused, allowing bias to enter an interpretation and resulting in premature or unwarranted interpretive conclusions. Experience does not guarantee that you will be able to keep observation and interpretation separate because there is a sense of urgency in the desire to explain observations and “get on with business” that can prevent you from devoting sufficient time to making an appropriate number of careful observations. Similarly, the lack of patience that often accompanies inexperience can lead to the same unfortunate result.



05/12/11 2:04 PM

6 First Steps in Seismic Interpretation Figure 2. The observe– interpret–test cycle when working with seismic data. We make observations on uninterpreted data, explain those observations in an interpretation (telling the geologic story contained in seismic data), and test conclusions with wells or additional data, leading to more observations and revised interpretation.

Observe Interpret

Correlate/explain/synthesize build model

Test

As said before, you can think of the objectives of seismic interpretation as seeking to answer two questions about subsurface geology: • “What is it?” — What elements of geology can you recognize (observe and explain)? • “Where is it?” — How accurately can you delineate elements of geology in three-dimensional space? To be of any use in a successful exploration program, your answers to these questions require that you understand how to accurately transform measurements and observations made in the reflection time domain into the depth domain. Except in the simplest cases, the inherent nonuniqueness of interpretation often allows your answers to “What is it?” to promote erroneous inferences about “Where is it?” or “How big is it?” — and vice versa. Which of these questions can or should be answered first, and the confidence with which either can be answered at all, clearly depends on the quality of available data, the tools at hand for analyzing those data, and your skill and experience as an interpreter. Often, prior knowledge of and experience in an area enable you to answer one of these questions with greater certainty than the other, and you effectively conduct a model-based interpretation, in which the course of the interpretation is guided by more than just observations and correlation of the data. There is nothing implicitly wrong with such an interpretation because you should incorporate all available information and experience into your interpretations. The peril lies in the possibility that prior knowledge can subconsciously (or otherwise) drive your interpretation, and so contradictory observations or correlations are downplayed or ignored because they don’t fit the model. In such cases, the objectivity essential to all interpretations is seriously at risk, and you may see only what you want to see.



05/12/11 2:04 PM

Chapter 1: Introduction 7

Your fundamental concern in seismic interpretation is recognizing that reflection seismic data do not (yet) provide one-to-one images of true subsurface geology. You must decide which features in the data are “real” and correlative and which are not, and you must always try to understand the differences between the two. At the same time, you need to determine how well resolved are the real features you see and how accurate are their spatial positions; hence, the importance of data quality and the ability to properly couch interpretation results within the context of that quality. In a philosophical sense, you should maintain healthy skepticism throughout your interpretations, using methodologies based on assumptions of doubt with the aim of gaining approximate or relative — but never absolute — certainty in your results.



05/12/11 2:04 PM



05/12/11 2:04 PM

Chapter 2 Seismic Response Seismic response is measured by the reflection generated at an acoustic impedance boundary according to the properties of the layers above and below the boundary and the nature of the seismic pulse impinging on that boundary. Referring to Figure 1, the equation below defines acoustic impedance (AI) as the product of compressional-wave velocity V and bulk density r: AI = V ρ The following equation defines the reflection coefficient (RC) in terms of AI for normal incidence of a seismic pulse at an AI boundary:

RC =

(V2 ρ2 − V1ρ1 ) ( AI2 − AI1 ) = . (V2 ρ2 + V1ρ1 ) ( AI2 + AI1 )

The Zoeppritz equations define the reflection coefficient for nonnormal angles of incidence of a seismic pulse at an AI boundary; these equations generally are applied in a simplified form (e.g., Shuey, 1985). For the purposes of this text and defining seismic as “having to do with elastic waves” (Sheriff, 2002), here we describe seismic response in terms of compressional-wave (P-wave) reflections but do not discuss shear waves (S-waves) or mode conversions in detail. You can initially and most easily describe seismic response with reference to an isolated impedance boundary and can further develop understanding of the composite response from multiple, closely spaced boundaries by way of the convolutional model (discussed later in this chapter). You need



05/12/11 2:05 PM

10 First Steps in Seismic Interpretation Incident pulse

Upper layer

Reflected pulse

V1, r1

Lower layer V2, r2

V = compressional-wave velocity, r = bulk density Figure 1. Definitions of acoustic impedance (AI) as a rock property, defined as the product of compressional-wave velocity V and bulk density ρ. The contrast in AI between two layers of rock gives rise to a seismic reflection when a seismic pulse impinges on the boundary between the layers.

to be familiar with a mathematical description for a waveform in terms of its frequency, amplitude, and phase characteristics, being especially careful to define phase and polarity as used in describing the shape or “character” of a reflection. The confidence with which you identify and correlate a reflection from an acoustic impedance boundary, which interpreters call a seismic event or horizon, based on its appearance or character depends on seismic data quality, on simple and well-known impedance relationships, and, perhaps most importantly, on correlation of seismic data to available well data via well ties. The importance of horizon identification increases as you move along the value stream from wildcat exploration through appraisal and development to production because this movement is toward greater detail of description in telling your geologic story. When interpreting and mapping in a frontier area, it may not be important to know whether a particular reflection corresponds to the top of a sand or a shale. But for a production project in the same area many years and millions of dollars later, it could be crucial to understand the seismic response for the top of a reservoir sand when choosing well locations and calculating reserves — hence, the importance of understanding seismic response in identifying horizons for interpretation. Understanding the seismic response to an AI boundary requires knowledge of the seismic pulse incident to that boundary and the behavior of the



05/12/11 2:05 PM

Chapter 2: Seismic Response 11

pulse as it propagates through the earth. The seismic pulse causes particle motion in the subsurface through a medium treated as elastic in response to stress applied in the form of an impulse (e.g., detonating a charge of dynamite or firing an air gun). Dix (1952, his Figures 11.4 and 11.5) presents schematic diagrams illustrating these particle motions for positive and negative reflection processes. A seismic waveform is a description of this particle motion as a function of time, which can be treated as a composite of many individual functions of time for the different frequency components present in the waveform; the analytical representation of a seismic waveform as the sum of individual sinusoidal functions is called Fourier analysis (Sheriff, 2002). For the sake of clarity and proper use of terminology, you should always be careful to distinguish between a reflector and a reflection: the former is a surface or boundary across which there is an acoustic impedance contrast, and the latter is a measurement of the particle motion caused by impingement of a seismic pulse upon the former. Keep in mind that you observe reflections and interpret reflectors (that is, elements of geology) from your observations — in that order. Maintaining a clear distinction between reflections and reflectors will help you remember that no seismic line or volume, no matter how carefully acquired and processed, is a completely accurate representation of true subsurface geology. A seismic pulse propagates through a subsurface that is not really elastic, so you can’t expect the pulse to retain its exact shape as it travels from the seismic source to a receiver. The change in shape of a wavelet, which is to say in the amplitude and phase characteristics of its different frequency components, because of propagation through a nonelastic earth is called attenuation. The physical properties of the subsurface of the earth cause the higher-frequency components of a wavelet to be preferentially reduced in strength, primarily because of converting the energy of particle motion to the heat of friction. In general, the farther or longer a signal travels, the more it is attenuated. Attenuation correction of seismic data, which can be done probabilistically (based on measurements of the data themselves) or deterministically (based on correlation with other physical measurements) is an important step in a seismic data-processing sequence. The change in shape of a wavelet as a result of attenuation suggests that, all other things being equal, you should not expect to see the same seismic response to the same impedance boundary that occurs at two different depths. A modeled product such as a synthetic seismogram, which usually is generated with an invariant wavelet, will therefore be better for making an accurate well tie in that portion of the seismic section where the wavelet used for the synthetic seismogram is a good approximation for the actual wavelet in the data. This is why wavelets are extracted from seismic data



05/12/11 2:05 PM


over windows or intervals of specific interest and then are used to generate synthetic seismograms for correlation only in that interval. Where possible, these extractions are done at or near points of well control so that log data can be used in the extraction process. In the time domain, a periodic function for a single frequency can be described as a sinusoidal wave, as with the cosine wave illustrated in Figure 2. The general form of the equation for this cosine wave as a function of time is y(t ) = A cos (2π ft + φ ), where A is the amplitude, f the frequency, t the traveltime, and φ the phase of the waveform. The value φ is the angle, measured in degrees (where 360° = 1 cycle), that represents the lead (the amount of time the waveform is advanced) or lag (the amount of time the waveform is delayed) with respect to a reference starting time. Phase φ is defined as the negative of phase lag (Yilmaz, 2001), which is to say that a negative time shift (time delay) corresponds to a positive phase value and a positive time shift (time advance) corresponds to a negative phase value. For example, Figure 3 shows that a cosine wave lags a sine wave by /2 or 90°:

π π  π π  π sin   = cos  −  = cos(0) = 1, sin ( 0 ) = cos  0 −  = cos  −  = 0,. . .  2   2 2  2 2 or

π π  π π π cos   = sin  +  = sin(π ) = 0, cos ( 0 ) = sin  0 +  = sin   = 1,. . . .  2   2 2  2 2 T

A

t

Figure 2. A simple sinusoid defined as a cosine wave. The shape of this waveform is determined by its amplitude A, frequency f, and phase φ. T is the period of the waveform.



05/12/11 2:05 PM


t

–p/2 0

p/2

p

3p/2 2p

cos(0) = sin(0 + p/2) = sin(p /2) = 1 sin(0) = cos(0 – p/2) = cos(–p /2) = 0 Figure 3. Phase relationship between a sine wave (red) and a cosine wave (blue). The sine wave leads the cosine wave by 90°, and the cosine wave lags the sine wave by 90°.

The waveforms shown in Figures 2 and 3 are infinite, single-frequency sinusoids; however, all of the wavelets with which you work in practical seismic interpretation are finite and have limited bandwidth. They are the summation of discrete sinusoids, each with its own amplitude, frequency, and phase characteristics. This is the basis of Fourier analysis. An example of a finite, band-limited wavelet and its component sinusoids is shown in Figure 4; in this example, the amplitude and phase of the components are constant (phase = 0) and only the frequency of the individual sinusoids varies. Knowledge of the phase of a waveform is important in Fourier analysis because this angle sets a reference for the starting time (zero time, effectively) for each component waveform defined by its own frequency and amplitude. An illustration of phase rotation of a simple band-limited wavelet symmetric about t = 0 through one full cycle from 0° to 360° for 90° increments is shown in Figure 5. As expected, phase rotations of 180° and –180° are identical. The wavelet in the center trace in Figure 5 is symmetric about t = 0, meaning that it literally describes particle motion that occurs before t = 0, which is physically nonrealizable. For this reason, the wavelet is called a noncausal wavelet (see Figure 6). Because of its symmetry, it is also referred to as a zero-phase wavelet; each of its component sinusoids is zero phase, and each is uniquely defined by its own amplitude and frequency according to Figure 2. In terms of signal processing, a zero-phase wavelet has the shortest time duration (pulse width) for a given bandwidth (frequency range). The



05/12/11 2:05 PM


Finite band-limited wavelet

t 40 Hz 30 Hz 20 Hz 10 Hz 5 Hz Figure 4. Illustration of a finite, band-limited wavelet as the summation of five component sinusoids. All of the components have the same amplitude and phase (phase = 0).

–180°

–90°

0°

+90°

+180°

Time

_

+

Figure 5. Phase rotation of a zero-phase wavelet (center trace) through a full 360° in increments of 90°. The display convention used in this figure is described in Figure 7.



05/12/11 2:05 PM

Chapter 2: Seismic Response 15 Noncausal wavelet

Causal wavelet

0

Time

_

+ Figure 6. Noncausal and causal wavelets. The causal wavelet involves particle motion only after time = 0, whereas the noncausal wavelet involves particle motion before time = 0, which is not physically realizable. The display convention used in this figure is described in Figure 7.

seismic response for a zero-phase wavelet also is easier and more intuitive to visualize because its maximum amplitude corresponds exactly to the position of the reflecting interface (see Figures 5 and 6). Displays that show the amplitude and phase characteristics of the sinusoids for every frequency component of a wavelet are called the amplitude (amplitude as a function of frequency) and phase (phase as a function of frequency) spectra. Given these amplitude and phase spectra, a resultant wavelet can be uniquely constructed by summing individual frequency components having the characteristics defined by these spectra. Figures 5 and 6 use the same display convention, i.e., they represent seismic response in the same way with reference to a standard impedance configuration. The display convention most commonly used by SEG is the positive standard polarity convention (Figure 7), in which polarity means positive or negative trace deflection. When discussing or presenting your work, you should state the phase of your data, to the degree it is known, and the display convention you are observing. Similarly, you should ask about wavelet phase and the display convention being used in any discussion or presentation involving seismic data if that information is not communicated or clearly annotated on seismic displays. Figure 8 illustrates the four different display formats for reflection seismic data. Of these, the most common used on workstation displays is variable density, often with user-defined or customized color schemes. Wiggle traces superimposed on a variable density background is also a popular display format.



05/12/11 2:05 PM

16 First Steps in Seismic Interpretation Acoustic impedance Low

Reflection coefficient –

Wavelet

+ 0

High Figure 7. The SEG positive standard display convention for reflection seismic data. “For a zero-phase wavelet, a positive reflection coefficient is represented by a central peak, normally plotted black on a variable area or variable density display” (Sheriff, 2002).

In virtually all cases, reflection seismic data represent a composite response to many closely spaced impedance boundaries, some of which are sharp and distinct and others of which are gradational. This composite response actually is the result of constructive and destructive interference of the discrete responses to individual impedance boundaries, described by the so-called convolutional model. Convolution is a mathematical operation that, in simplest terms, involves multiplication, shifting, and summation of two functions of the same variable (for seismic data the variable is traveltime t). You can think of convolution as simulating the propagation of a seismic pulse through a layered earth. The output of a 1D convolution, such as the convolution of an RC series calculated from an AI log (which has been converted to the time domain) with a seismic wavelet to produce a synthetic seismogram is probably much easier to visualize than to describe in words or to understand from exacting mathematical language. In Figure 9, the RC series consists of four coefficients, each corresponding to an AI boundary; the coefficients are not evenly spaced, and they do not all have the same magnitude and sign. This RC series will be convolved with the zero-phase wavelet shown to the left of the series, and both must have the same sample rate. Note that this wavelet is a wiggle trace that uses the SEG positive standard polarity convention. In the convolutional model, the seismic response to a given RC is created by reproducing the seismic wavelet scaled to the magnitude and sign of that RC. As shown in Figure 9, the scaled wavelet is reproduced as the seismic response for each of the four RCs, and the final convolution output or composite response is



05/12/11 2:05 PM

Chapter 2: Seismic Response 17 Variable density

Variable area

Wiggle

Variable-area wiggle

Figure 8. Four display formats for reflection seismic data. Display formats are independent of the polarity convention used for a given data set.

the sum of the individual scaled responses. There is both constructive and destructive interference between individual seismic responses in the composite response. This interference is substantial when the effective width of the seismic pulse is greater than the interval between adjacent RCs. For purposes of this discussion, consider the pulse width to be the breadth of the central peak or peak/trough. Notice also that there is no individual seismic response for any points in the RC series where RC = 0, that is, where there is no impedance contrast. The differences between the composite responses in Figure 9a and 9b indicate that your interpretation of geology from seismic data depends critically on the wavelet in your data. Knowledge of wavelet phase is important because it relates seismic response to geology in terms of the characteristics of the source wavelet (pulse) as defined in Figure 2, that is, the reflection seismic response to a given geologic boundary or feature changes for different source wavelets. The phase of the wavelet contained in any seismic data set can vary laterally and vertically (temporally) and is estimated most accurately by deterministic methods using well control. In the absence of well control, you can



05/12/11 2:05 PM

18 First Steps in Seismic Interpretation Reflection coefficients

a)

Individual responses

Overlay responses

Composite response

Overlay responses

Composite response

– + Input wavelet

b)

Reflection coefficients

Individual responses

– + Input wavelet

Figure 9. (a) The convolutional model. The individual responses of each reflection coefficient to the input seismic wavelet, scaled to the magnitude and sign of the reflection coefficient, are summed to generate the composite seismic response. There are destructive and constructive interference of the individual responses in producing the composite response. (b) Convolution of the reflection coefficient series shown in (a) with a different source wavelet. The differences between the composite responses for the two wavelets show that accurate interpretation of these responses depends on knowledge of the source wavelets.

visually estimate wavelet phase by observing certain reflections that may be present in your data (see Table 1). Using reflections from any of the boundaries listed in Table 1 assumes that the boundary can be identified conclusively, that there is a well-known and consistent acoustic impedance contrast across it (the algebraic sign of the reflection coefficient across the boundary is known), and that it is isolated



05/12/11 2:05 PM


Table 1. Subsurface boundaries that can be used for visual estimation of wavelet phase. No single boundary is absolute or foolproof. Best: Seafloor Hydrocarbon/water contact (seismic flat spot) Use with care: Top of salt/volcanics Base of salt Basement

from other nearby boundaries so that its character is not a composite reflection response. In marine settings, the seafloor reflection is commonly used to check wavelet phase because the impedance contrast between seawater and sediment is almost always positive. Similarly, a hydrocarbon/water contact, which appears as a seismic flat spot in a reservoir that is thick enough to be resolved seismically, can be used confidently to estimate wavelet phase (see the discussion of seismic resolution and tuning in Chapter 6). A seismic flat spot occurs because the presence of hydrocarbons as the pore-filling fluid lowers the AI of the hydrocarbon-bearing portion of a reservoir below that of the nonhydrocarbon-bearing or brine-filled portion of that reservoir. Not all flat spots are perfectly flat because velocity effects in time imaging can tilt or distort them and because some hydrocarbon/water contacts are not truly horizontal. A flat spot can occur only for reservoirs in which the hydrocarbonbearing portion of the reservoir is seismically resolved because the seismic response from a hydrocarbon-bearing interval whose thickness is below a certain value called the tuning thickness will be a composite of responses from the top and base of the interval that will not directly represent wavelet phase. The flat spot indicated by the arrow in Figure 10 shows a well-defined, symmetric peak (black). According to the accepted polarity standard and display convention for this image, within the visual acuity of the observer to see asymmetry in the waveform, the phase of the data is zero. Note that near the right-hand edge of this flat spot is a high-amplitude trough-over-peak amplitude response; this point marks the tuning thickness of the low-impedance, hydrocarbon-bearing portion of the reservoir. Continuing to the right, the decrease in the amplitude of the trough-over-peak signature reflects the decrease in thickness of the hydrocarbon-bearing portion of the reservoir. Note also that the top of the reservoir is not marked by a single, sharply defined reflection (a trough or a peak) along its full extent, suggesting that the top of the reservoir interval might be gradational in some places.



05/12/11 2:05 PM


t

Figure 10. Example of a well-imaged seismic flat spot, denoted by the yellow arrow, on time-migrated data. This image suggests that the seismic data are zero phase (courtesy PGS).

The problem with using boundaries such as top and/or base of salt, top of volcanics, and basement (which can take on a variety of geologic and economic meanings) for estimating wavelet phase is that these boundaries often are gradational and poorly defined, so their seismic responses are effectively composite responses to multiple, closely spaced impedance contrasts rather than to a single, well-known impedance contrast. At the same time, the impedance properties of the materials above and below these boundaries, especially for basement, are not necessarily well known or regionally consistent; so neither the magnitude nor the sign of the impedance contrast across such boundaries can be inferred confidently without well control. Most interpreters prefer to work with zero-phase data, for which a seismic event or horizon is symmetrically disposed about its correlative impedance boundary and thus is most easily and intuitively visualized. Knowledge of wavelet phase and the display convention of your data should enable you to draw geologically reasonable conclusions when correlating a given seismic response to a particular AI boundary. At the same time, you should recognize that a given impedance boundary can give rise to different seismic responses, depending on the phase of your data. This knowledge is critical for accurate interpretation of seismic attributes, as discussed in the next chapter.



05/12/11 2:05 PM

Chapter 3 Seismic Attributes By definition, a seismic attribute is a measurement based on seismic data (Sheriff, 2002). In the strictest sense, then, two-way traveltime, also known as horizon time, is perhaps the most important and frequently used seismic attribute, although it isn’t usually considered an attribute. Brown (1996) includes horizon time in his list of 66 different attributes and indicates that an attribute is “necessarily a derivative of a basic seismic measurement.” He presents a generalized classification scheme that breaks attributes into four categories: time, amplitude, frequency, and attenuation. Brown also poses two questions that all interpreters must address when analyzing seismic attributes: 1) What do they all mean? 2) When do we use one and when another? In the same vein, a paper with the delightful title “Redundant and Useless Seismic Attributes” by Barnes (2007) offers several common-sense suggestions for distinguishing “useful attributes from those of doubtful utility,” including, among other characteristics, their clear and useful meanings in a geologic and/or geophysical context as opposed to mathematical terms. With Barnes’ distinctions in mind, you can visualize a “utility spectrum” for seismic attributes (Figure 1), in which using an attribute or combination of attributes (e.g., by way of principal component analysis) to identify and correlate features of interest proceeds with varying degrees of attention to the true physical meaning(s) of the attribute(s). At one end of this spectrum is the mentality that “I’ll use this attribute for correlation when it helps me find what I’m looking for, regardless of what it means physically.” At the other end is “I’ll only use attributes whose physical meaning I fully understand.” In practice, there is no right or wrong mindset or approach — only



05/12/11 2:06 PM


“What does it physically mean?”

“Does it help me correlate?”

Figure 1. Utility spectrum for a seismic attribute. Depending on experience and the project at hand, you analyze attributes with varying degrees of attention to their physical meaning.

personal preference. Your position on this spectrum for any project in which attributes are important will depend on preference (that is, experience), seismic data quality, and availability of calibration information, such as well control. Seismic attributes are used to assist interpretation at all scales, ranging from analyzing regional depositional systems to mapping fine details of structure, stratigraphy, and rock properties.* They are also used to illustrate data quality (see Chapter 9). Two of the most commonly used seismic attributes are amplitude and coherence.

Amplitude The reference or baseline value for the amplitude of reflection seismic data is zero; so amplitudes are positive or negative in accordance with agreed polarity and display conventions, as discussed in Chapter 2. In terms of the simple two-layer model shown in Figure 1 of Chapter 2, the magnitude and algebraic sign (positive or negative) of the amplitude of a reflection from a single, isolated acoustic-impedance (AI) boundary is directly proportional to the magnitude and algebraic sign of the reflection coefficient (RC) at that boundary; the convolutional model (Figure 9 of Chapter 2) extends this relationship to the general case of the composite seismic response to a reflection coefficient series. In Figure 2, amplitude A is the departure of the waveform from the baseline value, as shown by the red arrows. The time separation between the apex of a peak (or trough) and the apex of its adjacent trough (or peak) is often called delta time or delta T (DT), and the absolute value of the difference in amplitudes measured at the same two points is called delta amplitude or delta A (DA). Notice that Figure 2 refers to data that are

*Excellent references for seismic attribute analysis are Seismic Attributes for Prospect Identification and Reservoir Characterization by Chopra and Marfurt (2007) and Interpretation of 3D Seismic Data by Brown (2011).



05/12/11 2:06 PM

Chapter 3: Seismic Attributes 23 Figure 2. Seismic amplitude (red arrows) and the quantities DT and DA (blue arrows) that are based on picking adjacent peak and trough reflections. This display is for time-processed data.

0 +

–

∆A ∆T

processed in the time domain, and DT is equal to one-half of the period of the displayed waveform. Before you proceed with any quantitative analysis of seismic amplitudes, you must ensure that amplitudes were handled carefully and consistently through all stages of data acquisition and processing. In many cases, this involves some detective work on your part, but your effort is well worth the trouble if you are to gain any meaningful and reliable information or value from amplitude analysis. You need to review all processing steps involving gain recovery and preservation of relative amplitudes as well as the quality of stacking (to be discussed in Chapter 4), giving particular attention to processes such as automatic gain control (AGC) that affect prestack or poststack amplitude balancing. Many interpreters prefer to apply AGC to their data when the primary objective is structural interpretation, but they would not use the same data for amplitude or attribute studies. At the same time, the data used for those types of studies must have sufficient dynamic range to include the maximum processed amplitude values and not have restricted this range by excluding or clipping values (a workstation-related issue).



05/12/11 2:06 PM


Any quantitative use of seismic amplitudes must be based on careful picking of reflections so that amplitudes are consistently and accurately measured or extracted from the seismic data. You should automatically track these reflections whenever possible (data quality permitting) so that they are consistently picked as peaks (maxima) or troughs (minima) if they are to be used for amplitude analysis. You can autotrack — track or pick reflections using computer-based processes (see Chapter 7) — or you can use an event that has not been autotracked as a reference horizon to construct a gate or window of sufficient size to contain the individual reflection or reflection package whose amplitude you want to measure. These gates do not need to be symmetrically arranged around a reference horizon and can be defined by two separately tracked horizons. Seismic amplitudes are manifestations of geology because they are the response to AI contrasts that are themselves measures of rock and fluid properties. Changes in amplitudes therefore reflect changes in geology, and every seismic line or volume exhibits a range of seismic amplitudes that can be correlated to trends in rock and fluid properties and ultimately to lithology and pore-fluid type. Amplitudes at the extremes of this range are anomalous in the sense that they are out of the ordinary or are departures from an established trend and, as such, can be of particular exploration interest. For example, in Cenozoic basins with predominantly clastic fill, anomalously high amplitudes known as bright spots have proven to be attractive exploration targets, although not guaranteed or completely risk-free. Bright spots reflect the reduction of the acoustic impedance of a reservoir sand caused by the presence of hydrocarbons in the pore-filling fluid in comparison to the acoustic impedance of the same reservoir filled with brine. Validation of seismic amplitude anomalies as direct hydrocarbon indicators (DHIs) is a very important element of successful exploration and development programs in many areas of the world. In general, you observe reflections as having anomalous amplitude when you visually inspect data (see Figure 3). This is a very qualitative measure, and amplitudes identified in this way are said to be above background, meaning that there is an ambient or background level of reflectivity associated with the overall impedance trends (i.e., the geology) of the area under investigation. Figure 4 shows the results of one quick and easy technique for highlighting amplitude anomalies by simply clipping or blanking all amplitudes below a certain threshold or background level. You should reference amplitude values to a statistically based background level and calibrate them to well control if you intend to use them for quantitative purposes such as calculating reservoir thickness and estimating reserves.



05/12/11 2:06 PM

Chapter 3: Seismic Attributes 25

t

Figure 3. A 2D time-migrated line with an anomalous high-amplitude reflection, indicated by the black arrow. Red events are negative amplitudes (troughs) and blue events are positive amplitudes (peaks). The trough-over-peak signature of the anomaly suggests that it can be interpreted as a thin (below tuning) hydrocarbonbearing reservoir. Note similar trough-over-peak anomalies to the right of the highlighted event (courtesy PGS).

Most if not all modern workstation systems and interpretation packages have standard routines for extracting amplitudes of individual reflections as well as of user-specified intervals of interest. In the latter case, the attribute is usually a statistical measure of the data within an interval (such as average absolute amplitude, maximum positive amplitude, or root mean square [rms] amplitude). As indicated by Barnes (2007), amplitude-related attributes such as these often provide duplicate measurements. You should test different attributes before choosing the one that works best — gives the most stable and clearly defined results, obviously involving interpretive judgment and experience. As Barnes goes on to say, “If you can’t tell which one works best, then it doesn’t matter which one you choose.” Not only does Barnes imply that you should take the time to test different attributes on your data, but he also requires that you have some idea of what the different attributes mean physically. The extraction of amplitude-related attributes over user-specified intervals can be particularly helpful when integrated with the results of sequence



05/12/11 2:06 PM


t

Figure 4. Same line as in Figure 3 but displayed with amplitudes less than approximately half of the maximum absolute amplitude value for the entire line clipped to highlight anomalously high and low amplitudes. Clipping was done visually by dynamic modification of the display color table. The clipping value can be thought of as a threshold between background and anomalous amplitude values (courtesy PGS).

stratigraphic or seismic facies analysis to aid in identifying and interpreting depositional systems, especially valuable in regional studies. You should carefully select and then pick the horizons that bound intervals of interest so that these intervals are neither too broad (effectively mixing or masking several different amplitude signatures) nor too narrow (excluding data that are needed for accurate characterization of an interval). Again, appropriate definition of intervals involves interpretive judgment, and any integrated analysis using seismic facies and attributes should include well control when calibrating results. The quantities DT and DA shown in Figure 2 are used in tuning or timeamplitude (time-amp) analysis (see Chapter 6, especially the wedge model shown in Figure 2 of that chapter) to study thin beds. In general terms, you should recognize that meaningful results from tuning analysis depend on critical factors in data processing and interpretation. For processing, these factors include true relative amplitude recovery and preservation, knowledge of the seismic wavelet and the signal-to-noise ratio (S/N) of the data, and the availability of carefully edited well data for calibrating the seismic



05/12/11 2:06 PM


response. For interpretation, the seismic data being analyzed must have the objective peak-and-trough reflections consistently and accurately picked, which requires autotracking and thorough quality control before proceeding with tuning calculations. Measurement of DT also provides a quick means for estimating the dominant frequency within a window of seismic data. By definition, dominant frequency is the predominant frequency determined by measuring the time between successive peaks or troughs (the period T of a waveform, as shown in Figure 2 of Chapter 2) and taking the reciprocal (Sheriff, 2002). In practical terms, the dominant frequency within a window of data can be thought of as the frequency of the waveform that dominates your view of the data within that window. Remember that the dominant frequency tends to decrease with increasing reflection time owing to attenuation, so any estimate of dominant frequency is applicable only to a window of data and not to the full reflection record. Using a visual estimate of DT in milliseconds for well-defined, coherent reflections within a window of interest and recalling that DT as defined in Figure 2 is equal to one-half the period of the waveform, you can calculate the dominant frequency in hertz (Hz, cycles per second) as 1000/(2 × DT). In the example shown in Figure 5 for an estimated DT of 30 ms for the Dominant frequency = 1000/(2 × 30 ms) = 17 Hz

t

~30 ms

Figure 5. Example of estimation of dominant frequency based on observed peak-totrough time separation DT (courtesy PGS).



05/12/11 2:06 PM


high-amplitude peak/trough reflections within the window outlined in red, the dominant frequency is 1000/(2 × 30 ms) = 17 Hz. Note that this technique is implicitly meaningful only for time-domain data because depthprocessed data cannot be characterized by a measurement (DT) that can be made only in the time domain unless the depth data have been converted back or stretched to the time domain using the velocity model with which they were processed. For additional information on physical principles and techniques for analysis and interpretation of seismic amplitudes, refer to Seismic Amplitude Interpretation by Hilterman (2001).

Coherence Coherence is a seismic volume attribute; it is run only on 3D seismic data and measures the trace-to-trace similarity of the seismic waveform within a small analysis window. Coherence technology was originally developed by Amoco (Bahorich and Farmer, 1995) to enable more complete use of the abundance of information contained in a 3D seismic volume to complement standard interpretation techniques. Because important elements of geology such as faults and stratigraphic features (e.g., channel margins) are evident as discontinuities in seismic data, an attribute such as coherence can be very useful in identifying and visualizing these features (see Figures 6 and 7). Generating a coherence volume is an automated process that requires selecting values for several input parameters, most important of which is the size of the data-analysis window (in three dimensions). A very large window will include too much data and produce output with a pronounced structural overprint, whereas a window that is too small will include too little data and produce output that is more a manifestation of noise in the data rather than geologic content. Input-parameter values usually are chosen following a series of tests to determine which combination of values produces the most interpretable output. As with the results of many other analytical processes, the quality or interpretability of the output depends heavily on the noise content of the input; a very noisy data set will probably contain very little useful or reliable coherence information. Your experience comes into play in selecting input parameters for coherence processing and evaluating/interpreting coherence output; that experience also helps you properly gauge the value added to your interpretation through the use of coherence data. Many interpreters generate a coherence volume for their 3D data as one of the first steps in a new interpretation project. A cursory review of these coherence data can be very helpful in gaining initial impressions of the geology of the project area and building work flows and schedules for ensuing



05/12/11 2:06 PM

Chapter 3: Seismic Attributes 29 a)

b)

Figure 6. (a) Traditional 3D time slice; faults parallel to strike are difficult to see. (b) Coherence time slice; faults are clearly visible. From Bahorich and Farmer (1995).

interpretation tasks. Coherence displays such as that shown in Figure 7b also can be very useful for quality control of fault picks to ensure that faults are not miscorrelated in structurally complicated areas. Some interpreters prefer to pick faults primarily on coherence slices rather than on vertical sections when data quality permits. Coherence data usually are viewed on a horizontal slice through the coherence volume or along a suitably tracked horizon (the latter producing a coherence horizon slice). Alternatively, the original 3D volume can be flattened along a suitably picked horizon, and coherence can be generated from that volume to produce a coherence horizon slice. These slices can be particularly helpful for interpreting details of stratigraphy, but you must pick the input horizon very carefully — entirely by automatic tracking if possible — so that tracking artifacts are not passed through the coherence process and subsequently interpreted as real geology.



05/12/11 2:07 PM

30 First Steps in Seismic Interpretation a)

b)

Figure 7. Comparison of a conventional 3D horizontal slice through (a) a reflectivity volume and (b) a coherence volume generated from the parent reflectivity volume. In this example from a depth-processed 3D volume, faults that appear as dark lineations on the coherence slice are more clearly seen and accurately interpreted on the coherence data than on the reflectivity data (courtesy BP).

Inversion Although you can consider seismic inversion to be more a process than an attribute in the sense that amplitude and coherence are attributes, some discussion of inversion is warranted because interpretation of inversions most often involves analysis of attributes extracted from the inverted data. The place of seismic inversion within the sequence of acquisitionprocessing-interpretation of reflection seismic data is best visualized using Figure 8, a slightly modified version of Figure 1 from Chapter 1. In simplest terms, the inversion process involves calculating AI data from reflectivity data; as such, seismic inversion can be considered the first step in interpreting ideally processed seismic data. As shown in Figure 8, conventional reflectivity data, the ideal seismic response, provide information about the boundaries between subsurface layers, whereas seismic inversions (AI) measure the properties of the layers themselves. If you approximate an RC series as the derivative of an AI function, then the inversion of “optimally processed” reflection seismic data effectively is integrating those data to produce AI data. The following equations show that this approximation is based on the assumption that the difference between



05/12/11 2:07 PM


Lithology

AI

Ideal seismic response

RC – +

+

No depth scale implied

No time scale implied

Inversion Figure 8. Seismic inversion, creating AI data from an “ideal seismic response,” is an inverse process that can be considered an interpretive process (refer to Figure 1 of Chapter 1).

adjacent samples in an AI series is incrementally small, that is, AI varies slowly and smoothly with depth. From the equation in Chapter 2, RC( j ) =

( AI ( AI

( j +1) ( j +1)

− AI( j ) ) + AI( j ) )

.

The first equation below contains the key assumption that allows the RC series to be approximated as the derivative of an AI function: ∆AI(t ) , 2AI(t ) AI(t ) ∫ RC(t ) ~ ln 2 + C1 , RC ( t ) e∫ ~ C AI(t ), RC(t ) ~

2

where C is a constant.



05/12/11 2:07 PM


Inversion is done in the time domain, so depth-processed data must be converted to time before inversion processing and then converted back to depth after inversion if desired. Note that scaling of the inverted data is implicit in the integration; this scaling is determined by calibration to well control or to regional impedance trend curves. Of course, any reflection seismic data can be inverted, but the quality of the inversion depends on calibration and the quality of the input reflectivity data. Lindseth (1979), who refers to seismic inversion as generating a synthetic sonic log from a processed seismic trace, captures the uncertainty inherent in seismic inversion: The inversion of seismic reflection data is much more demanding (and revealing) of data acquisition and processing quality than are conventional displays. While inferior data, as in many seismic operations, does not totally impede the execution of any process, the quality of output will be degraded, eventually reaching a point where any benefits from the procedure are doubtful. In the context of Lindseth’s comment on the dependence of seismic inversion on the quality of data acquisition and processing, following is a list of conditions you should check about the processing of input reflectivity data before proceeding with inversion: • • • • • •

Amplitudes are true relative amplitudes. Amplitude variation with offset (AVO) effects are accounted for. Data are zero phase. Seismic wavelet is invariant (at least over the window of interest). Bandwidth is maximized. All multiple reflections have been removed (see Chapter 7).

All of these conditions are rarely if ever met, so the quality of a seismic inversion will always need to be assessed carefully in terms of its correlation to well data and the accuracy of its representation of real geology. Seismic inversions are correlated in much the same way as are conventional reflectivity data (see Chapters 7 and 8). Although a primary objective of correlating both types of data is to define the boundaries of the intervals of interest, you pick zero crossings on inverted data to identify these boundaries and then examine attributes of the defined layers to study their internal properties. Contrast this with picking troughs, peaks, and, occasionally, zero crossings on zero-phase reflectivity data, where the picked horizons also define layers of interest but extracted attributes



05/12/11 2:07 PM

Chapter 3: Seismic Attributes 33 a)

b)

Figure 9. Comparison of inverted and conventional reflectivity data. (a) Inverted data on which the top and base of a low-impedance interval are picked as zero crossings (black = relatively high impedance, yellow/brown = relatively low impedance). (b) Corresponding reflectivity data onto which the blue horizons from (a) have been directly transferred (black = peak, yellow/brown = trough). On this display, the blue horizons coincide approximately with a trough and a peak that correspond to the top and base, respectively, of the same low-impedance interval; these correlations are confirmed by well control not shown on these images (courtesy BP).



05/12/11 2:07 PM


provide information only about layer boundaries and not internal layer properties. Figure 9 compares inverted data and the reflectivity data from which the inversions were generated; the blue horizons were autotracked as zero crossings on the inverted data and transferred directly to the corresponding reflectivity data. The good correlation of the zero-crossing picks on the inverted data to the troughs and peaks on the reflectivity data confirms that, in terms of wavelet phase, the inversion process can be thought of as applying a phase rotation of –90° to the reflectivity data (see Figure 5 of Chapter 2), which is consistent with the concept that acoustic impedance can be calculated by integrating a reflection coefficient series. Note also in the leftcenter of Figure 9a that the upper blue horizon is not defined (cannot be automatically tracked) where there is no zero crossing between two closely spaced low-impedance layers. In Figure 9b, however, the upper blue horizon can be tracked automatically as a trough through the same area to a point of termination with the lower blue horizon (peak). Seismic attributes play an increasingly important role in interpretation as you move along the value stream from exploration through appraisal and development to production. Often, attributes are a deciding factor in selecting well locations and designing well trajectories, whether in an exploration or a production setting. Although you can interpret attributes solely on the basis of uncalibrated observations, with well control for calibration you can more accurately and confidently explain the geologic meaning and significance of the attributes of your data.



05/12/11 2:07 PM

Chapter 4 Velocity Reflection seismic data measure the traveltime of a seismic pulse from its source to subsurface reflectors and back to a receiver or array of receivers. A complete description of geology requires that any interpretation derived from these data, recorded in the time domain, be placed accurately in the 3D depth domain, and you use velocity information to accomplish this transformation from time to depth. The importance of time-to-depth conversion would be evident in the look of consternation on a driller’s face if you told him that the exploration target in a proposed well was at 5.3-s two-way time, with no mention of the depth to which this reflection time corresponds. Velocity is defined as distance traveled per unit time. It is a vector quantity, that is, it has magnitude and direction. The scalar quantity associated with velocity is speed. Although we customarily talk about the velocity of propagation of compressional (P-) and shear (S-) waves through different materials, in a strict sense we are describing only the speed of wave propagation through these materials because usually we do not specify the direction of measurement. In other words, we simplify our statements by assuming the materials are isotropic. Many materials exhibit velocity anisotropy, which means that the velocity of P- and S-waves in those materials depends on the direction in which it is measured (in the direction of wave propagation). The concept of velocity anisotropy is very important for accurate seismic imaging and time-depth conversion and is addressed later in this chapter. Three different types of velocity are commonly used in the oil and gas industry. The first, interval velocity Vint, is the distance traveled per unit time, where the distance traveled is the thickness of a single well-defined layer or stratum: Vinti = 2

∆zi . ∆ti



05/12/11 7:46 PM


Here, Vinti is the interval velocity of the ith layer, Dzi is the actual thickness of the ith layer, and Dti is the two-way-time thickness of the ith layer. The reciprocal of interval velocity is called slowness. Average velocity Vavg is the distance traveled per unit time, where the distance traveled is the total thickness of many layers or strata measured from the top of the uppermost layer to the base of the lowermost layer: Σ 2 ∆zi . Σ∆ti ΣVinti ∆ti

Vavg = =

Σ∆ti

Root-mean-square (rms) velocity Vrms is a statistical quantity calculated from Vint and Dt: 2 Vrms =

ΣVint2 i ∆ti Σ∆ti

.

These definitions are generalized for an arbitrary number of subsurface layers; Figure 1 illustrates them for simple two-and three-layer models. Three sources of velocity data are used in the oil and gas industry: sonic logs, well-velocity surveys, and seismically derived velocities.

Sonic logs A borehole sonic tool continuously emits a high-frequency signal from a transmitter and records that signal at a receiver; the transmitter and receiver are located within the same wireline tool. The travel path of the signal from transmitter to receiver is typically 3–10 ft (1–3 m) long, and the tool is engineered to compensate for its tilt within the borehole and for variations in borehole size. The traveltime from the transmitter to the receiver is the interval transit time (ITT), also known as slowness, measured in microseconds per foot. With correction by a factor of 10–6 to convert microseconds to seconds, interval velocity (measured over the very small interval between the transmitter and receiver in the sonic tool) is equal to the reciprocal of interval transit time: Interval velocity =

1 Interval transit time



05/12/11 7:46 PM

Chapter 4: Velocity 37 a)

Datum Z1

t is thickness in two-way time

t1

Vint1 = 2 Z1 / t1

t2

Vint2 = 2 Z2 / t2

Z Z2

Vavg =

2 Z/ t/

2

2

Vrms =

t = (2 Z1 + 2 Z2)

Vint

t = (Vint1 2

Z1

Z2

Z

Z3

Vavg = 2

Vrms =

2 Z/ 2

Vint

t is thickness in two-way time

t1

Vint1 = 2 Z1 / t1

t2

Vint2 = 2 Z2 / t2

t3

Vint3 = 2 Z3 / t3

t = (Vint1 2

t2) / ( t1 + t2)

2

t = (2 Z1 + 2 Z2 + 2 Z3) t/

t1 + t2 )

t1 + Vint2

Datum

b)

/(

2

t1 + Vint2

/(

t1 + t2 + t3) 2

t2 + Vint3

t3) / ( t1 + t2 + t3)

Figure 1. Schematic of the definitions of average, interval, and root-mean-square (rms) velocities for (a) two-layer and (b) three-layer models.

For example, a measured interval transit time of 80 ms/ft is equal to an interval velocity of 12,500 ft/s (3800 m/s): 1 1 ft = = 12, 500 ft/s −6 (80 µs/ft × 10 s/µs) 0.00008 s



05/12/11 7:46 PM


Errors in measuring true rock velocities with a sonic log can occur for several reasons. First, when formation velocity is less than the mud (drilling-fluid) velocity no refraction occurs at the borehole wall, and the signal does not take a standard path to the receiver. Second, areas of washed-out borehole allow the signal to travel through the mud rather than the formation, so correlation between slow velocity and large borehole size indicates that the measured velocity probably is too slow. This commonly occurs below a casing shoe or in salt. Finally, fracturing and invasion of porous formations can cause the measured velocity to be different from that of the undisturbed rock. This occurs particularly in gas-bearing formations where the mud filtrate tends to invade the rock and remove the low-velocity effect of the gas.

Well-velocity surveys The basic field setup for a well-velocity survey is illustrated in Figure 2. In such a survey, the seismic source — an air gun for a marine survey and most commonly vibroseis for a land survey — is placed as near to the wellbore as possible. In a marine environment, an air gun is hung over the side of the rig or drillship into the sea. On land, if an air gun is used instead of vibroseis, it is put in a water- or mud-filled pit adjacent to the rig. The survey illustrated in Well

Datum

Tstat Source

Tobs

D

Vavg =

D Tvert + Tstat

Tvert

Receiver (geophone)

Figure 2. Schematic of the field setup for a conventional well-velocity or checkshot survey. In the figure, Tobs is recorded traveltime from source to receiver, Tvert is recorded traveltime converted to vertical traveltime, Tstat is vertical traveltime from datum to source depth, and D is depth from datum to receiver.



05/12/11 7:46 PM

Chapter 4: Velocity 39

Figure 2 uses a single downhole geophone. However, for increased operating efficiency and cost savings, most velocity surveys now use an array of evenly spaced geophones rather than a single geophone, thereby enabling recording of arrival time at many depth points for a single shot. The survey begins by lowering the geophone to the total depth (TD) of the well and clamping it to the borehole wall. The source is fired, and the arrival time of energy that travels directly from the source through the earth to the geophone (the first arrival) is recorded; the arrival time of the source impulse at a receiver placed very close (but not too close) to the source also is recorded to establish a time datum for the survey. The source is fired several times at each geophone station, and the records for each station are summed with bad records edited out, to produce a final output record for each level. The survey proceeds by raising the geophone up the borehole, stopping at predetermined levels to clamp the geophone and fire the source. The levels at which the geophone is positioned can be spaced regularly to achieve a desired vertical sampling of velocities or set at specific geologic boundaries or formation tops. The survey is finished when the geophone reaches the level at the top of the survey program or when the recorded data fail to meet acceptable signal-to-noise (S/N) criteria. This type of well-velocity survey, known as a check-shot survey, provides direct measurements of average velocity, from which you can calculate interval velocity, as shown in Figure 3. Datum

Z1 = depth of station 1 Z2 = depth of station 2

T2 Z1

Z = station spacing = (Z1 – Z2)

Z2 T1 = one-way arrival time at station 1 T1 Z

T2 = one-way arrival time at station 2 Vavg to station 1 = Z1 / T1 Vavg to station 2 = Z2 / T2

Vint = (Z1 – Z2) / (T1 – T2) =

Z / (T1 – T2)

Figure 3. How to calculate interval velocities from data collected in a well-velocity survey.



05/12/11 7:46 PM


There are two main differences between a check-shot survey and a vertical seismic profile (VSP). First, the listening time for a VSP is longer than that for a check-shot survey because a VSP records not only the direct arrival of energy from source to receiver but also energy reflected from impedance boundaries below the geophone position in the well (see Figure 4). Second, the station spacing for geophones in a VSP is always regular and fairly short (50–75 ft; 15–20 m) to achieve the sampling required to avoid aliasing the desired signal (see Chapter 6). The VSP records downgoing (direct) and upcoming (reflected) energy, so it contains the same information as a checkshot survey with the added advantage of denser sampling of the subsurface velocity field. The source and receiver geometries in a well-velocity survey vary according to wellbore geometry and the objectives of the survey (see Figures 5 and 6). The source can be placed very near (essentially at) the well or some distance away and can be fixed throughout the survey or moving with respect to the well or the wellbore trajectory. The receivers also can be fixed with a moving source, as in the case of a walkaway VSP, or moving, as in the case of a walkover VSP when the receiver is located vertically beneath the source. In itself, VSP processing is a complicated subject that is not discussed in this text; however, Figure 7 provides a good idea of what raw VSP data look like before processing is done. Needless to say, well-velocity Well

Datum

Tstat Source

Tobs

D

Vavg =

D Tvert + Tstat

Tvert

Receiver (geophone) Reflector

Figure 4. Same field setup as shown in Figure 2 but including upcoming arrivals as recorded in a VSP. The green arrows indicate downgoing and upcoming raypaths.



05/12/11 7:46 PM

Chapter 4: Velocity 41 Wellhead

Offset

Source Receiver Figure 5. Field geometries for well-velocity surveys with a fixed source. Walkover

Walkaway

Source Receiver

Figure 6. Field geometries for well-velocity surveys with a moving source.

surveys should be designed carefully in conjunction with the results of forward seismic modeling.

Seismically derived velocities Although sonic logs as well as check-shot surveys and VSPs directly measure P-wave velocity, they operate at very different geologic scales and use considerably different source and receiver instrumentation. They



05/12/11 7:46 PM


Time

Depth

Figure 7. A raw VSP record (depth of receiver on the x-axis, traveltime to receiver on the y-axis). Downgoing arrivals dip from right to left; upcoming arrivals dip from left to right (courtesy BP).

are direct in the sense that they measure the elapsed time of propagation of a compressional pulse from an energy source to its arrival at a receiver a known distance from the source. When using a near-wellbore source, a check-shot survey or VSP typically measures only the vertical component of propagation velocity and does not account for the possibility (probability) of velocity anisotropy. In contrast, the velocity derived from analyzing reflection seismic data provides an indirect measure of P-wave velocity that does include the effects of anisotropy. Seismically derived velocities actually are calculated using known acquisition geometry and observed reflection arrival times. The validity and utility of these velocities depend heavily on the accuracy of velocity-analysis techniques and the degree to which certain critical assumptions underlying the velocity calculations are met. The concept of multiplicity of subsurface (or multifold) coverage, in which reflections from a single subsurface point are recorded using multiple source-receiver geometries, establishes the basis for conventional seismic velocity analysis. As conceived, multifold acquisition was developed to more effectively enhance coherent signal and eliminate unwanted random and coherent noise, which could be achieved by adding, or stacking, the multifold data. Central to the stacking process is estimating the optimal



05/12/11 7:46 PM


stacking velocity, determined in the velocity-analysis step of a normal seismic processing sequence. This velocity analysis is based on the assumptions that the propagating media are horizontal layers that are homogeneous (uniform in physical properties throughout) and isotropic (having the same physical properties regardless of the direction of measurement). Given these conditions, the recorded traveltimes, source-receiver distances, and stacking velocity are related by the so-called normal-moveout (NMO) equation. Unfortunately, earth scientists know that these assumed conditions are rarely, if ever, met in the real earth. Figure 8 illustrates a source and multiple-receiver array positioned over a single horizontal reflector at depth. In this diagram, the distance X between the source and any receiver in the array is referred to as the sourcereceiver offset, which is zero when the source and receiver are coincident. The reflector in Figure 8 is the boundary between two horizontal layers that, respectively, have constant P-wave velocities V1 and V2 and densities r1 and r2 (review Figure 1 in Chapter 2). For the geometry shown in Figure 8, a reflection from the boundary between the layers arrives at successively later times for receivers with greater offset because a signal travels successively greater distances through the upper layer for larger offsets. Figure 9 shows that as the entire source-receiver array is moved, a given point on the reflector is effectively sampled multiple times (called fold of coverage), each time by energy traveling along a different path. Figure 9 also illustrates how the recorded energy from these multiple paths for the same reflecting point, referred to as a common depth point (CDP) or, more accurately, as a common midpoint (CMP), can be collected or gathered into a)

b) Offset X

Source

Offset X

Receivers

V1, ρ 1 Reflecting points Depth Z

V2, ρ 2 Reflection time T

Figure 8. Seismic acquisition for (a) a simple two-layered earth model (depth domain) and (b) the corresponding shot record (time domain). This model forms the basis for the common-depth-point (CDP) method of acquisition.



05/12/11 7:46 PM


Shot #1

A

B

3 2 1

C

A B C

V1, ρ 1 Shot #2

A

B

C

V2, ρ 2

Common depth point (CDP) Offset X A B

V1, ρ 1 Shot #3

A

B

V2, ρ 2

T0

V1, ρ 1

Reflection time T

C

C

V2, ρ 2

CDP gather

Figure 9. Building on the model shown in Figure 8, the sequence of three different source-receiver configurations shown on the left illustrates how moving a sourcereceiver array allows you to sample a given reflecting point in the subsurface (the CDP, marked by the red dot) multiple times, each with a different incident angle and raypath. As shown on the right side, collecting or gathering the traces corresponding to these different raypaths for the same subsurface reflecting point enables formation of a CDP gather, on which seismic velocity is analyzed. NMO is the increase in reflection time for a given reflecting point that occurs as the source-to-receiver distance increases.

a single group known as a CDP or CMP gather. For a given CDP gather, the increase in reflection traveltime with increasing offset, referred to as normal moveout (NMO), is described by the well-known NMO equation: T 2 = T02 +

X2 V2

where T is the two-way traveltime to a receiver at offset X and T0 is the twoway traveltime for X = 0. All of the offsets X for the geometry in Figure 9 are known, and T for each offset is recorded. It is possible to correct each T in a CDP gather to T0 using the NMO equation and the appropriate value for V. Figure 10 illustrates how accurate moveout corrects the recorded arrival time T of a



05/12/11 7:46 PM


Offset X

Offset X

T0

T0

Reflection time T

Reflection time T

Figure 10. NMO correction, by which each of the traces in a CDP gather is corrected so it represents the trace that would have been recorded if the source and receiver were coincident (offset X = 0). The second-order NMO equation, which is an approximation, describes this behavior in terms of reflection time, offset, and velocity (NMO or stacking velocity). The velocity term in the NMO equation, although having units of velocity, can be most accurately described as the variable which, when solved for, best flattens the event on the CDP gather.

reflection at each receiver to T0, flattening the corrected reflection across the gather. Figure 10 also illustrates the power of stacking moveout-corrected traces in a gather to produce a composite trace in which signal is enhanced and unwanted noise is greatly reduced. Conventional seismic velocity analysis solves for the unknown quantity V, commonly referred to as the stacking velocity Vstk or NMO velocity VNMO, by applying moveout corrections to all traces in a gather with a suite of velocities and then measuring the accuracy of the corrections according to how well reflections are flattened across the moved-out gather. This measure is valid only when assuming normal hyperbolic moveout, which is in itself an approximation. Notice that in the NMO equation, reflection time T is a hyperbolic function of offset X. Accurate, good-quality velocity data are those which, in making the move-out correction, result in exact flattening of primary reflections in a corrected gather, assuming hyperbolic moveout. Stacking velocities can be identified, or picked, in several ways, one of which uses a contoured semblance plot or velocity spectrum on which the stacking velocity that results in the best flattening of a reflection corresponds to the maximum value for the semblance or trace-to-trace similarity of that reflection across the NMO-corrected gather. When picking velocities on a velocity spectrum, a data processing geophysicist also looks at a real-time display of the NMO-corrected gather at the analysis location, on which he considers how well the selected velocity flattens the reflection at the two-way time of the velocity pick. He adjusts each pick until satisfied



05/12/11 7:46 PM


that it produces optimal event flattening, and for each analysis location he creates a series of two-way time/stacking-velocity pairs that will be used in stacking. His assessment of flattening is very interpretive and depends heavily on his visual acuity, the noise content of the data, and the smoothness of the velocity field, such that the assumption of hyperbolic move-out is valid. In accordance with the NMO equation and in view of the interpretive nature of NMO velocity analysis, the accuracy of velocity picks generally decreases with increasing reflection time (depth) because the amount of NMO decreases with increasing depth. You can see this effect on velocity spectra as smearing or stretching of semblance clusters along the velocity axis with increasing reflection time. Figure 11 illustrates how a velocity spectrum is produced. The NMO equation is applied sample by sample to the entire time range of an uncorrected gather (from T = 0 to 5.9 s two-way time, or TWT) over a velocity domain of V = 4800 to 12,000 ft/s (1460–3660 m/s). For a reflection at a given time, there is a moveout velocity for which the measure of semblance at that reflection time is maximum, that is, for which the trace-totrace alignment of the reflection character at that time is most consistent across the entire offset range of the gather. The (V,T) pair corresponding to that point is marked, or picked, on the velocity spectrum, and similar maxima are picked over the full time range of the gather to produce the best time-velocity function for moving out the entire gather. These velocity picks are the white dots connected by white line segments shown in Figure 11b. Figure 12 shows how moveout with velocities that are too slow or too fast results in a moveout correction that is too large (overcorrected) or too small (undercorrected). In Figure 12a, the moveout correction is too large because the velocity is too slow; the applied velocity function is shifted to the left of the best function toward the slower velocities, and the overcorrected events can be viewed as smiling — think of s for “slow and smiling.” In Figure 12b, the moveout correction is too small because the velocity is too fast; here, the applied velocity function is shifted to the right of the best function toward the faster velocities, and the undercorrected events can be viewed as “frowning” — think of f for “fast and frowning.” The uncorrected gather, moveout-corrected gather, and stacked trace in Figure 13 summarize the velocity analysis/NMO correction/stacking sequence. NMO correction is very important because you very often see only stacked seismic traces and are unable to directly verify the accuracy of preceding velocity analysis and moveout correction. Figures 14 and 15 show that even relatively small errors in moveout velocities can affect the appearance of final stacked traces, which ultimately affects how you



05/12/11 7:46 PM

Chapter 4: Velocity 47 a)

Velocity spectrum

CDP gather

T 2 = T02 + X 2 V2

b)

Figure 11. Velocity spectrum on the left for (a) the raw (no moveout correction) and (b) moveout-corrected CDP gather on the right. Highest semblance values on the spectrum correspond to the dark blue, purple, and red areas. Stacking velocity picks in (b) are marked as white dots connected by white line segments.



05/12/11 7:46 PM


Velocity spectrum

T 2 = T02 + X 2 V2

CDP gather

Overcorrected

b)

Undercorrected

Figure 12. Example of a CDP gather (right) for which the applied moveout correction is (a) too large or (b) too small. The gather is overcorrected in (a) and undercorrected in (b) because the moveout velocities are too slow in (a) and too fast in (b), respectively, as shown in the spectrum (left).



05/12/11 7:46 PM


Raw CDP gather

NMO-corrected CDP gather

45-fold CDP stack

Figure 13. (left) A raw CDP gather, (center) the same gather with best moveout correction applied, and (right) the single output trace generated by stacking the moveout-corrected and muted gather.

interpret them. Note that the volume of data processed is reduced in stacking by a factor equal to the fold of the data; with reference to Figure 13, the CDP gather composed of 45 traces stacks to a single output trace, which is approximately a 98% reduction in the number of traces handled from prestack to poststack. All of the traces in the NMO-corrected gather shown in Figure 13 were summed to create the single stacked trace on the right side of the figure; however, any user-specified number or combination of traces from the gather can be summed in the stacking process. Partial stacking, which does not use all of the traces in a corrected gather, is usually done to enhance imaging by stacking only those traces for which signal strength is higher over a given range of offsets or incidence angles than for other ranges or the entire gather. The change in signal strength can accurately reflect real subsurface conditions such as illumination, amplitude variation with offset (AVO), or amplitude variation with angle (AVA); but it can also represent inaccurate velocity estimation — events are not flattened across the full range of offsets or angles. Figure 16a is an example of partial stacking of the near and far halves of the corrected gather shown in Figure 13, and Figure 16b is an enlargement,



05/12/11 7:46 PM

50 First Steps in Seismic Interpretation – 3%

– 2%

– 1%

Best

+ 1%

+ 2%

+ 3%

Figure 14. Groups of stacked traces (each group represents a single stacked trace that has been duplicated 10 times for visual effect) corresponding to the best corrected gather from Figure 13 and the same gather corrected with a range of different velocity functions (1%–3% greater or less than the best function). Note that the stacked responses change as the applied moveout-velocity functions change.

showing in detail the differences in reflection character between the two partial stacks. You can quickly visualize from Figure 16a how correlations of partial stacks depend critically on the reflectivity of the different ranges of offsets or angles included in the stacks. As an interpreter, it is your responsibility to always inspect corrected gathers before partial stacking to ensure that data have been NMO corrected as accurately as possible and that no undesired illumination or AVO/AVA effects carry through into the stacks. For the simple case shown in Figures 8 and 9, the NMO velocity VNMO is equal to the interval velocity Vint of the single subsurface layer through which the seismic energy passes. For the case of a single dipping subsurface layer, VNMO is equal to Vint divided by the cosine of the angle of dip of the layer. For the case of multiple horizontal subsurface layers, VNMO is approximately equal to rms velocity Vrms. For the case of multiple parallel, dipping



05/12/11 7:46 PM

Chapter 4: Velocity 51 – 3%

– 2%

– 1%

Best

+ 1%

+ 2%

+ 3%

Figure 15. Displays of 21 adjacent CDPs (the center CDP corresponds to the gather shown in Figure 13) with the same range of moveout-velocity functions as in Figure 14. The stacked responses change as the applied moveout-velocity functions change.

subsurface layers, VNMO is approximately equal to Vrms divided by the cosine of the angle of dip of the layers. If there are many dipping nonparallel layers in the subsurface, then accurate estimation of VNMO is a ray-tracing problem and not a simple or straightforward calculation. Note that Vrms is a good approximation for VNMO (or stacking velocity Vstk) only under particular conditions. Generally, these terms are not synonymous. Interval velocity can be calculated from VNMO using the well-known Dix equation (Dix, 1955): for i horizontal layers and small offsets (Vrms ~ VNMO), (VNMOi )2 Ti − (VNMOi −1 )2 Ti−1  (Vinti ) = (Ti − Ti−1 ) 2

It is very important to note that the Dix equation is valid only for horizontal layers and small offsets (a small offset is equal to or less than the depth of



05/12/11 7:46 PM

52 First Steps in Seismic Interpretation a) Raw CDP gather

b)

Near half (inside)

NMO-corrected CDP gather

Far half (outside)

Near half (inside) CDP gather

Near stack

Far half (outside) CDP gather

Near stack

Far stack

Far stack

Figure 16. (a) Raw and NMO-corrected CDP gather from Figure 13 with near- and far-half CDP gathers and their associated near-trace and far-trace partial stacks. The near-stack and far-stack displays represent the single near-half and far-half stacked traces, respectively, which have been duplicated 10 times for visual effect. The red box outlines the area enlarged in (b). (b) Enlargement from (a) showing the difference in reflection character between the near-trace and far-trace partial stacks for the NMO-corrected gather. Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:46 PM


the layers to which the Dix equation is being applied; see Dix [1955]). This equation will probably continue to be misused to calculate interval velocities in situations that do not satisfy these conditions. Having calculated Vint from VNMO, quantities such as average velocity and interval transit time can be calculated easily. Figure 17 shows Dix interval velocities superimposed on a velocity spectrum from which stacking velocity picks for the Dix calculations were taken. The sensitivity of interval velocity calculated with the Dix equation to the precision of stacking velocity picks is best illustrated by a numerical example. Stacking velocity picks from NMO velocity analysis are for the two reflections (peaks) marked by red arrows on Figure 18. For pick 1 (upper peak), VNMO1 is 5557 ft/s (1694 m/s) and T1 is 2.994 s. For pick 2, VNMO2 is 5737 ft/s (1749 m/s) and T2 is 3.210 s. The conditions required for valid application of the Dix equation — horizontal layers and small offsets at the level of the target reflections — are reasonably well met in this example. The Dix interval velocity calculated for these picks is 1

 ((5737 )2 ( 3.210 ) − (5557 )2 (2.994 ))  2 Vint =   = 7816 ft/s (2382 m/s) . ( 3.210 − 2.994 )   Velocity spectrum

CDP gather

Interval velocity

Figure 17. Velocity spectrum from Figure 11b, showing Dix interval velocities calculated from stacking velocity picks. Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:46 PM


2.0 s

Pick 1

3.0 s

Pick 2

4.0 s

5.0 s

Figure 18. Time-migrated seismic line for which stacking velocity picks for the two reflections marked by red arrows are used to illustrate the sensitivity of the Dix interval velocity calculation.

Now increase the stacking velocity value for pick 2 by 3% to 5909 ft/s (1801 m/s) and recalculate the interval velocity 1

 ((5909 )2 ( 3.210 ) − (5557 )2 (2.994 ))  2 Vint =   = 9532 ft/s (2905 m/s) . ( 3.210 − 2.994 )   In this example, an increase of 3% in a stacking velocity pick results in a corresponding increase of 22% (7816 to 9532 ft/s; 2382 to 2905 m/s) in the calculated interval velocity. Figure 22 shows that a change in one stacking velocity pick causes two corresponding changes in interval velocity because the altered pick is used in two interval velocity calculations, one for the interval above the level of the revised pick and one for the interval below. The sensitivity of Dix interval velocities to stacking velocity picks strongly suggests that you should review the results of stacking velocity analyses, usually in the form of NMO-corrected gathers, before using calculated interval velocities for interpretive purposes such as lithology prediction, pore-pressure modeling, or time-depth conversion. Using Dix interval



05/12/11 7:46 PM


velocities without this quality assurance can lead to serious technical errors and consequent expense. Keep in mind these limitations when calculating interval velocities from stacking velocities: • Layer geometry — dip, nonparallel layers, and layer curvature can invalidate the results of Dix calculations. • Layer properties — the Dix calculation assumes layers are homogeneous and isotropic. • Precision of data analysis — picking reflection times and stacking velocities is interpretive. Velocity (ft/s) 0.0

0

2000

4000

6000

8000

10000

12000

14000

1.0

Two-way time (s)

2.0

3.0

4.0

5.0

6.0 Stacking velocity

Interval velocity

7.0

Figure 19. Plot of the stacking velocity function (blue) and calculated Dix interval velocities (red) for the CDP location in Figure 18. The annotations in black on these functions represent the change in one stacking velocity pick and the corresponding changes in two calculated interval velocities.



05/12/11 7:46 PM


Remember that stacking velocity is the velocity that optimizes the effectiveness of stack; it results in the best enhancement of signal and reduction of noise by the stacking process. We interpret stacking velocities using a variety of analytical tools and displays and assess the quality of the stacks produced. Because of the inherent subjectivity of these interpretations, correct or objectively determined stacking velocities as such probably do not exist. Only under certain conditions and having made several important assumptions does stacking velocity approximate true propagation velocity. These conditions and assumptions (horizontal, homogeneous, and isotropic layers and small offsets) must be recognized and understood before any interpretation or application of seismically derived velocities can proceed correctly and meaningfully.

Velocity anisotropy Anisotropy is the variation in a physical property depending on the direction in which it is measured. Velocity anisotropy is important when working with seismic data because the strata through which a seismic pulse travels very often exhibit velocity anisotropy. In other words, the pulse does not move at the same speed in all directions of propagation. The velocity anisotropy commonly encountered in working with reflection seismic data is called polar anisotropy, also known as vertical transverse isotropy (VTI), a condition associated with layered strata in which the axis of symmetry is vertical (perpendicular to layer boundaries; the layers themselves are transversely isotropic). If the axis of symmetry is tilted, then the velocity anisotropy is called tilted transverse isotropy (TTI). Five independent Thomsen parameters are used to characterize polar anisotropy; the two most often mentioned in discussing this type of anisotropy (Sheriff, 2002) are d (delta) and h (eta). The parameter d is associated with weak anisotropy, most critical for describing polar anisotropy and frequently associated with the short-offset moveout correction applied to vertical velocity: VNMO = aII × (1 + d), where aII is the P-wave velocity parallel to the axis of symmetry for anisotropy. This velocity is equivalent to that measured in a well-velocity survey. The parameter h captures the deviation of long-offset P-wave moveout from what it would have been in an isotropic medium. It is calculated from d and



05/12/11 7:46 PM


e (epsilon), the latter of which is another of the five independent Thomsen parameters for polar anisotropy:

η=

ε −δ 1 + 2δ

Correction for velocity anisotropy must be done whenever NMO velocities are used in depthing processes such as time-to-depth conversion or depth migration. For time-to-depth conversion, recall that the vertical propagation velocities measured in well-velocity surveys are slower than NMO velocities, as related by d, so any time-to-depth conversion performed using uncorrected NMO velocities will result in calculated depths that are too large. Similarly, isotropic prestack depth migration (PSDM) using a sediment-velocity model based on NMO velocities will result in lateral and vertical mispositioning of migrated reflections; the vertical mispositioning often is of greater magnitude than the lateral mispositioning, so a so-called Z-to-D vertical correction (where Z is migrated depth and D is true vertical depth) based on calibration to well control must be applied to achieve accurate depthing. If anisotropy information from well data is available, then depth migration can be done anisotropically (APSDM) so that the anisotropy condition is accounted for during migration. Figure 20 illustrates the value of APSDM for producing a depth volume that exactly ties well control. Of course, the problem of how to extend anisotropy parameterization away from well control remains.

Time-depth conversion Because seismic data are acquired in the time domain T, they must be imaged in or converted to depth as measured in real earth space — in the domain of depth measurements made in wells — to describe subsurface geology accurately. You must never forget that a time-processed image can be, and very often is, distorted because of the complexity of the velocity field through which the seismic energy passes and that accurate seismic imaging and depth conversion are necessary to interpret and map seismic data correctly. Velocity data provide the essential link between the time and depth domains that enables you to (1) provide spatially accurate maps of the subsurface, (2) predict well depths, and (3) calculate bulk rock volumes and hydrocarbon reserves. As mentioned in the preceding section, depths measured on isotropic depth-migrated data (Z-depths) need to be corrected to D. Depths on anisotropic depth-imaged data do not require this correction, at least at the



05/12/11 7:46 PM


b)

Figure 20. (a) A 3D isotropic PSDM showing a mis-tie between the yellow seismic horizon at 8687 m (28,500 ft) and the correlative formation top at 8077 m (26,500 ft) true vertical well depth (dashed blue line). (b) A 3D anisotropic PSDM of data in (a), showing exact well tie to the interpreted yellow seismic horizon at 8077 m (26,500 ft) true vertical well depth (courtesy TGS).

calibration points for an anisotropic velocity model as shown in the example in Figure 20. A fair amount of depth imaging is done with isotropic models, owing largely to lack of anisotropy data for parameterizing velocity models, so Z-to-D conversion is necessary to accurately correlate and map these data.



05/12/11 7:46 PM


Techniques for vertical conversion from T to D and from Z to D fall into five general categories, depending on the type and quantity of available velocity data. The following techniques apply to vertical conversion only with no lateral repositioning of points in the subsurface: • • • •

Single velocity function (sonic log, check-shot survey, VSP, trend curve) Multiple velocity functions (any combination of single functions) Layered velocity model Continuous velocity model (e.g., stacking or migration velocity volume) • Calibrated velocity model (with single or multiple T-to-D or Z-to-D functions) The two critical factors in all of these techniques are (1) the quality and distribution of available velocity control and (2) the accuracy of interpolation, vertically and laterally, between these control points. Interpolation must be mathematically correct and, perhaps more importantly, must be geologically reasonable in the sense that trends introduced into the interpolated data are consistent with known geology or sound geologic models. Obviously in cases for which more than one model fits the available velocity control, the final depth conversion must be accompanied by an estimate of uncertainty that reflects multiple possible outcomes. Vertical conversion with velocity functions uses one or more time-depth functions (Figure 21) to calculate depths to individual seismic horizons or to any points within the range of the function(s). This is commonly done in areas where velocity varies smoothly vertically and laterally and is not defined discretely by formation boundaries. For Z-to-D conversion, the procedure can involve two steps, the first converting from Z to T followed by converting from T to D. Alternatively, these two steps can be combined into a one-step process converting directly from Z to D. In the latter case, there is no actual velocity involved in the conversion; rather, the conversion factor is dimensionless, based on the ratio between the Z-to-T velocity from the two-step conversion and the T-to-D velocity from that same process (this conversion accounts for d anisotropy). Vertical conversion with a layered velocity model, such as the one shown in Figure 22, uses interval velocities defined for each interval in the model to calculate thicknesses of those intervals, which are then summed appropriately to determine the depth to each layer boundary or interpolated to estimate depths to points within intervals. The layers in these models are most often defined by interpreted seismic horizons or surfaces calculated from those horizons. The interval velocity for each layer in the model can



05/12/11 7:46 PM


2000

2000

4000

4000

6000

6000

8000

8000

10000

10000

12000

12000

14000

14000

16000

16000

0 00

00

10

00

90

00

80

0

00

2.5

70

2.0

00

1.5

60

1.0

00

0.5

Average velocity (ft/s)

50

0.0 0

D (ft)

b)

One-way time (s)

40

a)

Figure 21. (a) A time-depth function from a conventional check-shot survey and (b) an average velocity-depth function calculated from the original time-depth data. Vint1 Vint2 Vint3

Horizon 1 Horizon 2 Horizon 3 Horizon 4

Vint4 Time

Vint5 Vint6

Horizon 5 Horizon 6 Horizon 7

Vint7

Figure 22. A vertical slice through a layered interval-velocity model. The full model would be a 3D interval-velocity volume. Interval velocities are assigned to layers whose boundaries are defined by or calculated from interpreted seismic horizons.

be assigned as a constant velocity or as a velocity function applied only within that layer. This is commonly done in areas where velocity corresponds closely to geologic formations or specific structural elements, e.g., layered depth conversion involving salt most often assigns a constant interval velocity to a mapped salt body.



05/12/11 7:46 PM


Figure 23. A 3D PSDM seismic line with its associated depth-migration velocity model shown as a colored overlay (interval velocity increases from blue to red). The irregularly shaped high-velocity feature in this image is a salt body (courtesy PGS).

Vertical conversion with a continuous velocity model, such as a stacking velocity volume or as an interval velocity–depth model used for depth migration, differs from using discrete velocity functions only in the sense that interpolation between control points is an implicit part of creating a continuous model. Figure 23 is an example of a depth-migration velocity model superimposed on a PSDM seismic line. Such a continuous model is built primarily as an integral step in the depth-migration process, and it is an extension of its original purpose to use the model for Z-to-D conversion. In fact, because continuous velocity models are often based on seismically derived velocities, they should not be used for vertical conversion if they have not been calibrated to well control or, in the absence of well control, to trend curves or some other source of information that accounts for velocity anisotropy. In summary, you will most frequently use four basic types of velocity: • Interval velocity — distance traveled per unit time, where the distance traveled is the thickness of a single well-defined layer or stratum. • Average velocity — distance traveled per unit time, where the distance traveled is the total thickness of many layers or strata measured from the top of the uppermost layer to the base of the lowermost layer; is calculated directly from the known depth of a receiver and the oneway traveltime to that receiver as recorded in a well-velocity survey.



05/12/11 7:46 PM


• Stacking velocity — also known as normal moveout (NMO) velocity, a quantity having the units of velocity that is estimated during stacking velocity analysis by flattening reflections on an NMO-corrected gather via application of the NMO equation. • Root-mean-square (rms) velocity — a statistically defined velocity which, under the constraints of horizontal layers and small offsets, is approximately equal to stacking velocity; when these constraints are satisfied, this approximation allows you to calculate interval velocities from stacking velocities using the Dix equation.



05/12/11 7:46 PM

Chapter 5 Migration Because of the physics of wave propagation, the 3D nature of subsurface geology, and the techniques with which reflection seismic data are acquired and processed, seismic data must be migrated, or repositioned, to place reflections in their true subsurface positions. This migration, which addresses the image-fidelity element of data quality, is done routinely as part of a data-processing sequence for all 2D or 3D data before or after stacking in the time or the depth domain, depending on the complexity of the geologic structure and the subsurface velocity field.* As an interpreter, you are concerned with migration because in every interpretation you describe the size and position of the elements of geology that you see in your data, and you can’t accurately do this without taking migration into account, either in data processing or as part of your interpretation work flow. The need for migration is illustrated in Figure 1, which is a 2D model of a single dipping reflector with constant P-wave velocity above the reflector and seismic source and receiver coincident at point SR. By convention, the recorded two-way traveltime t to the dipping reflector is plotted on a vertical trace at point SR, even though the true normal-incidence reflecting point on the reflector is not located vertically below point SR. The dashed red curve in Figure 1, which is an arc of a circle with radius equal to t, represents all possible positions for the reflector for a given value of t. This curve is known as a wavefront, a locus of equal traveltimes through a propagating medium for an impulse occurring at t0 = 0 (see Figure 2). The migration *A paper by Gray et al. (2001) contains an excellent historical perspective of migration as well as practical treatment (with a minimum of high-level mathematics) of migration problems and solutions, and a paper by Etgen et al. (2009) provides a comprehensive overview of the current state and future direction of depth imaging in exploration geophysics.



05/12/11 2:12 PM

64 First Steps in Seismic Interpretation X u

SR

m

t

t

Migrated

Unmigrated

sin m = tan u

Figure 1. Schematic of migration of a dipping interface in a 2D constant-velocity model, where u is unmigrated dip angle, m is migrated dip angle, t is two-way traveltime, and X is horizontal distance. Impulse at t0

t1 t2 t3 t3 > t2 > t1

Figure 2. Schematic (2D view) of the definitions of wavefronts and rays. A wavefront is a locus of equal traveltimes for a pulse propagating through an elastic medium. The shape of a wavefront depends on the velocity distribution in the transmitting medium; in this example, the wavefronts are circular because the propagation velocity is constant and isotropic. A ray (red arrow) is a line (or curve) everywhere perpendicular to wavefronts that represents the travel path of a pulse from t0 to traveltime t.

operation moves the reflecting point from its position vertically below point SR along this curve to its true subsurface position, that is, to its migrated position at the point of normal incidence reflection. Notice in the very simple example of Figure 1 that the relationship between the dip angles of the unmigrated and migrated reflectors is derived directly from trigonometry of the triangles formed by the origin of the diagram, point SR, and the reflecting points on the unmigrated and migrated reflectors.



05/12/11 2:12 PM

Chapter 5: Migration 65

Datum t

Unmigrated Migrated

Figure 3. Migration of a horizon (dashed black curve) interpreted on a 2D unmigrated seismic section. The final migrated horizon (solid red curve) connects points of tangency to the arcs (wavefronts) constructed from the source positions.

A horizon interpreted on an unmigrated 2D seismic section can be migrated by constructing arcs for many source points along the section and then drawing a smooth curve connecting points of tangency to these arcs, as shown in Figure 3. Think of this process as generating a wavefront for the observed traveltime to the unmigrated reflection (horizon) at each individual source point and then defining the migrated horizon as the surface tangent to all of these wavefronts. The shapes of the arcs (wavefronts) depend on the velocity distribution in the section above the dipping reflector. In the simplest case of constant velocity, the arcs are circular; but they become more complicated when the velocity distribution varies vertically or laterally (or both). Hence, repositioning reflections — migration — is a velocitydependent process. Before the advent of 3D data or computerized migration as an essential step in a standard data-processing sequence, migration of 2D seismic data was addressed in several ways. One primary way was to interpret horizons and faults on unmigrated data. The horizon maps constructed from the interpreted lines were then migrated using an appropriate velocity function or distribution. This technique, called map migration, was done separately for each interpreted horizon, requiring consistency of velocity trends, vertically and laterally, from one horizon to the next to produce geometrically correct and geologically reasonable maps. Another way to migrate data was to interpret horizons and faults on unmigrated data; then the interpreted lines were migrated individually using an appropriate velocity function or distribution (as illustrated in Figure 3). Because this migration could be done only in the plane of an individual 2D



05/12/11 2:12 PM


line, migration could not properly account for the effects of a 3D structure — for reflections from points located out of the vertical plane of the line. These out-of-plane reflections (see Figures 4 and 5) are called sideswipe. Only in the relatively uncommon case of a 2D line being true dip to actual subsurface structure can 2D migration be trusted to be accurate; even at that, its results depend on the accuracy of the velocity used for migration. The failure of 2D migration, whether manual or computerized, to handle 3D subsurface structure accurately is the source of the mis-tie problem present in virtually all 2D interpretation projects; 3D imaging is required to address this issue properly. An example of the power of migration to more accurately define true subsurface geology is shown in Figure 6. Figure 6a is a 2D unmigrated seismic line on which you see what is commonly referred to as a bow tie,

t

Figure 4. The antiformal feature within the red circle on this 2D time-migrated display is an example of the out-of-plane effect known as sideswipe on 2D seismic data. Apparent structural discordance such as this is an obvious positive indication of sideswipe; the antiform and the dipping reflections that dominate the bottom half of the display cannot coexist as reasonable subsurface geometries. Even the dipping reflections on this display will be mispositioned (mismigrated) if the 2D line is not a true dip line. Several fault-plane reflections also can be seen in this image (courtesy BP).



05/12/11 2:12 PM

Chapter 5: Migration 67 a)

t

b) t

Figure 5. (a) Image of a 2D time-migrated seismic line, showing sideswipe (crosscutting antiformal reflections within the yellow circle). The red arrow marks the intersection of this line with an orthogonal 2D time-migrated line. (b) Image of a 2D time-migrated seismic line orthogonal to the line shown in (a). The red arrow marks the intersection of the two lines. There is no sideswipe on this line, and the dipping salt body to the left of the line is the source of the sideswipe reflections observed on the orthogonal line shown in (a). The distance between the intersection of the two lines and the edge of the salt body on this line is approximately 8000 ft (2450 m) (courtesy WesternGeco).



05/12/11 2:12 PM


b)

t

t

Figure 6. (a) Image of a 2D unmigrated line exhibiting a classic bow-tie reflection configuration. (b) Prestack time migration (PSTM) of the line shown in (a). The crossing reflections in the center of the unmigrated image are resolved to reveal a relatively simple syncline. Note also that small faults, especially on the left side of the image, are more sharply defined. Focusing of reflections in general is improved (courtesy PGS).

so named for the pattern of crossing reflections in the center of the image. This reflection configuration as it appears cannot in all likelihood represent real geology, so migration is needed to resolve the actual structure. Figure 6b is the migrated version of this 2D line; the bow-tie reflections have been



05/12/11 2:12 PM


repositioned to their true locations to reveal a relatively simple syncline. This type of structure is called a buried focus syncline because the center of curvature of the syncline is below the recording surface of the seismic data (see Figure B-11 in Sheriff [2002] for illustrations of the raypath geometries and a synthetic record section for this structure). Comparison of Figure 6a and 6b shows that in addition to resolving the true structure of the syncline correctly, migration more clearly defines small faults, especially to the left of the syncline, and generally focuses reflections more sharply. The smoothness and clarity of the migrated image in Figure 6b suggests that the orientation of this line is very nearly perpendicular to the axis of the syncline, that is, the line is a dip line. Keep in mind, though, that this is still 2D migration, no matter how striking the results, and that 3D migrated data would be needed for optimum imaging accuracy. Reflection seismic data are migrated in the time or the depth domain, depending on the complexity of the subsurface structure and the subsurface velocity field (see Figure 7). As a result of progress in computer power and sophistication of migration algorithms as well as in response to the advance of exploration into more challenging subsurface settings, migration is now done routinely on prestack data, although there are still many areas in which poststack imaging in time or depth can provide acceptable results.

PoSTM

PrSTM

Poststack time migration

Prestack time migration

Simple velocities Simple structure

Simple velocities Complex structure

PoSDM

PrSDM

Poststack depth migration

Prestack depth migration

Complex velocities Simple structure

Complex velocities Complex structure

Figure 7. Different migration types for prestack and poststack time and depth domains. Most seismic imaging is now done on prestack data, so the acronyms for poststack time (PoSTM) and poststack depth (PoSDM) migration are no longer commonly used. The shortened acronyms PSTM for prestack time migration and PSDM for prestack depth migration are now widely accepted.



05/12/11 2:12 PM


Reflected S Incident P

Reflected P 1

VP1,

1

VP2,

2

q1´

2

Transmitted P

Refraction (P) sin sin

1

=

2

V1 V2

Transmitted S

Reflection (P) q1 = q1´

Figure 8. Snell’s law for reflection and refraction of P-wave energy at and across an acoustic impedance (AI) boundary. The critical angle of incidence θc is the angle at which θ2 = 90° (sin θ2 = 1), that is, sin θ1 = V1/V2, and no energy is transmitted across the AI boundary into the deeper layer. Raypaths for reflected and transmitted shear (S-wave) energy are shown by the dashed arrows.

Depth imaging is needed primarily for areas of large lateral velocity contrasts in the subsurface where ordinary time-domain imaging fails because it does not account for refraction of seismic energy (defined by Snell’s law; see Figure 8) at the boundaries across which these contrasts occur. Because depth imaging includes the effects of refraction in calculating travel paths by way of traveltimes through an interval-velocity model, its results more accurately describe the true positions of subsurface reflectors. However, you must be aware that migration output in depth does not guarantee one-to-one correspondence with true geology. Depth imaging can fail when the depthmigration velocity model is inaccurate, either in defining the geometries of anomalously high- or low-velocity bodies or in assigning specific velocity values, gradients, or anisotropy parameters in the velocity model. Figure 9 clearly illustrates the differences between time and depth migration, demonstrating that accurate description of geology requires depth imaging in areas where there are large lateral velocity contrasts, in this example caused by salt bodies. To produce an accurate image of subsurface features of interest, seismic data must first be acquired in such a way that energy reflected from those features is recorded at the surface. The term illumination is defined as



05/12/11 2:12 PM


Two-way time below sea level (ms)

a)

Depth (m below sea level)

b)

Figure 9. (a) A 2D PSTM seismic line, approximately 75 km (47 mi) long, from offshore Brazil. (b) A 2D PSDM image of the line in (a).The differences between the two images are striking. The PSDM image is a more accurate representation of subsurface geology, certainly leading to a very reasonable explanation for the location of the exploration well (annotated in green) on the right side of the line (courtesy PGS).

the placement of seismic sources and receivers so that seismic energy will fall on desired portions of reflectors and be recorded for processing. Obviously, you cannot migrate reflections to their true subsurface positions if the energy reflected from those positions was never recorded. Often you will find yourself correlating horizons through poorly imaged zones that were only partially illuminated or not illuminated at all, so that you are effectively conducting a model-guided interpretation, connecting illuminated and properly migrated patches of the subsurface together in a geologically reasonable way. This is to be expected, especially in frontier exploration or in areas with severe imaging problems such as subsalt plays, and you must be sure to risk your interpretation of these areas accordingly. If illumination



05/12/11 2:13 PM


Figure 10. Matrix of migration algorithms in modern depth-migration methods, illustrating the range of migration algorithms that can be used to address different subsurface imaging problems. In general, greater structural and/or velocity complexity in the subsurface requires algorithms from the upper-right quadrant of the matrix, which involve increased time and cost in their applications (Figure 1 by Biondi in Herron [2009]).

modeling was not done as part of acquisition design for your data, then it is good practice to do this modeling using first-pass interpretation results to identify the areas in which your correlations are probably less reliable and to provide input for additional data acquisition. Seismic migration has become more important as exploration targets are being sought in increasingly challenging and complex settings. There are many different migration approaches and algorithms, some better suited to specific imaging problems than others, all having their own strengths/ limitations and corresponding cost implications (for example, see Figure 10). As an interpreter, you will often contribute to decisions involving which migration algorithms to use for a given problem, so you will need to develop at least a basic understanding of how the different algorithms work. This is part of building experience, and it requires you to work closely and communicate effectively with processing geophysicists. You will frequently be called on to assess the quality of migration output. For all of the mathematical and computational complexity of migration,



05/12/11 2:13 PM


your assessment will often consist exclusively of visual and very nonquantitative determination of improved S/N and reflection continuity — ultimately, whether the output appears to look more reasonable geologically within the context of expectation or realization of some geologic concept or model. At best, these will be subjective assessments, and you will make them with greater confidence as you gain experience.



05/12/11 2:13 PM



05/12/11 2:13 PM

Chapter 6 Resolution By definition, resolution is the ability to separate two features that are close together (Sheriff, 2002). Resolution applies to seismic data and to products derived from interpreting seismic data (e.g., maps) in the temporal and spatial domains. We often speak about the resolving power of seismic data and what can be done to increase it because, in doing so, we’ll be able to interpret finer details of subsurface geology. The concept of resolving power of seismic data follows from the sampling theorem, also known as the Nyquist theorem, which formally states that band-limited functions can be reconstructed from equispaced data if there are two or more samples per cycle for the highest frequency present (Sheriff, 2002). On the basis of this theorem, you can describe commonly used measures of temporal and spatial resolving power of seismic data. The sampling of seismic data is specified by a sample rate, such as 2 or 4 ms, and sampling frequency is defined as the inverse of the sample rate, which is 500 Hz for a sample rate of 2 ms and 250 Hz for a sample rate of 4 ms. The Nyquist frequency is defined as half the sampling frequency: For a sample rate of 2 ms, the Nyquist frequency is 250 Hz; and for a sample rate of 4 ms, the Nyquist frequency is 125 Hz. According to the sampling theorem, when there are fewer than two samples per cycle of a given signal, a signal at one frequency yields the same values as those for another frequency, and the one signal can be mistaken for the other. This frequency ambiguity, or aliasing, is illustrated in Figure 1, in which a 200Hz sine wave is aliased, or misread, as a 50-Hz sine wave when sampled at 4 ms. When acquiring seismic data, you can prevent frequency aliasing by using an antialias filter during recording to attenuate frequencies above the Nyquist frequency.



05/12/11 2:13 PM


4 ms

200 Hz

50 Hz

Sample rate = ∆t

Nyquist fN = 1 / 2 ∆t = 1 / 2 (4 ms) = 125 Hz

Figure 1. A 200-Hz sine wave that aliases as a 50-Hz sine wave when sampled at 4 ms. The Nyquist frequency in this case is 125 Hz, so a 125-Hz antialias filter used during recording would attenuate the 200-Hz signal.

The temporal resolving power of seismic data is usually described by the tuning thickness of the data, which is based on the fundamental equation that relates velocity V, dominant frequency f, and wavelength l: V = fl . The dominant frequency in this equation can be estimated easily from the time separation between adjacent peak and trough reflections on a seismic section (see Figure 5 in Chapter 3); in general, the dominant frequency changes vertically and laterally on a seismic section. Given estimates of the dominant frequency of the data and the vertical propagation velocity in the vicinity of the features to be resolved, you can calculate the wavelength of the seismic signal from the preceding equation, from which the Rayleigh limit of vertical resolution is derived as l/4. Because propagation velocity and the dominant frequency of the seismic signal change vertically and laterally throughout the subsurface, it follows that temporal resolving power will vary across a given area of investigation. Tuning thickness usually is visualized with the aid of a diagram known as a wedge or tuning model (Figure 2). The purpose of such a model is to illustrate the seismic response to the wedge and determine the thickness



05/12/11 2:13 PM

Chapter 6: Resolution 77 Distance (ft)

Depth (ft)

VP = 7000 ft/s

VP = 6000 ft/s

V=fλ Tuning thickness = 50 ft

TWT (ms)

No density contrast

30-Hz Ricker wavelet

Figure 2. Wedge model using a 30-Hz Ricker wavelet and P-wave velocities VP of 6000 and 7000 ft/s (1800 and 2100 m/s) for the wedge (in blue) and encasing medium, respectively. Using the formula V = fλ, the tuning thickness for this model is calculated to be 50 ft (15 m), which corresponds to the point on the model (red line) at which the trough-peak amplitude response is greatest.

for which the amplitude response is maximum, that is, for which constructive interference of the individual responses from the top and base of the wedge is maximum. The point at which this composite amplitude response is maximum is the tuning thickness for the model, with given input wavelet, wedge geometry, and layer impedances. Notice that above the tuning thickness, the seismic responses from the top and base of the wedge are separate and distinct (the bed thickness is “resolved” by the time separation between these individual responses). Below the tuning thickness, the waveform of the composite response does not change, but its amplitude decreases as the bed thickness decreases. These observations suggest that with good data quality (based on seismic processing from which wavelet phase and true relative amplitudes can be reliably determined), careful horizon interpretation, and available well data for calibration, you can use seismic data to estimate layer thicknesses, a study commonly referred to as tuning or time-amplitude (time-amp) analysis (see Chapter 3).



05/12/11 2:13 PM


Z = depth Z

Z + /4

Reflector

First Fresnel zone

Figure 3. Schematic of the geometry of the first Fresnel zone.

The spatial resolving power of seismic data is usually described in terms of the Fresnel zone, defined as the portion of a reflector from which reflected energy can reach a detector within one-half wavelength of the first reflected energy (Sheriff, 2002). Figure 3 shows the geometry of the first Fresnel zone, which is the smallest and innermost of a succession of higher-order annular Fresnel zones. The equation 1

V  TWT  2 Fr =   2 f  defines the radius of the first Fresnel zone Fr in terms of the two-way traveltime (TWT) to a reflector, the average propagation velocity V to that reflector, and the dominant frequency f of the seismic signal impinging on the reflector. This formula implies that the size of the first Fresnel zone almost always increases with depth (corresponding to increasing propagation velocity and two-way time) and decreasing dominant frequency of signal (owing to attenuation). Fresnel zones are measured with respect to unmigrated seismic data. Seismic migration collapses these zones; however, 2D migration collapses the zones only in the direction of shooting of the 2D line. For 3D data, a full 3D migration collapses the first Fresnel zone to a circle with a diameter of l/2 (radius = l/4), where l is the dominant wavelength of the seismic signal. Spatial sampling is an important consideration when designing 3D seismic surveys. The size of the unit of area into which a 3D survey is subdivided, called a 3D bin, ideally should be sufficient in terms of the Nyquist theorem, to properly sample the dip of the steepest reflector and/or the area of the smallest feature of interest within the survey. Figure 4, which represents a 2.5D model of the subsurface (the third dimension in the strike direction



05/12/11 2:13 PM

Chapter 6: Resolution 79 Bin at surface

4

q = True dip q

Maximum bin spacing = Vavg/(4f sinq)

Figure 4. Schematic of the maximum 3D bin spacing required to image the maximum dip of target reflectors in terms of average velocity Vavg to the targets, maximum dip q of the targets, and dominant or peak frequency f of the seismic signal.

is perpendicular to the plane of the page), illustrates how the required bin size is related to the average velocity Vavg to the target reflectors, the maximum dip q of those reflectors, and the dominant frequency f of the seismic signal. Notice that the Rayleigh resolution limit (the tuning thickness) also appears in this relationship. It is important to realize that 3D survey design, in addition to addressing technical requirements such as maximum dip and minimum area to be imaged, must also take into account economic considerations that can balance or even outweigh technical factors. Seismic trace displays can exhibit aliasing related to spatial sampling, as shown in Figures 5 and 6. Figure 5 shows four arrays, each consisting of four identical variable-area wiggle traces. Traces in Figure 5a are aligned such that the zero crossing marked in red (the red horizon) is correlated horizontally from trace to trace. On the succeeding arrays (Figure 5b–5d), each trace within the array is shifted downward by a constant amount from the trace on its left, with the amount of shift increasing from array to array. The dip of the red horizon increases in direct proportion to the amount of trace-to-trace shift in each array and in the direction of the shift (from left to right). In Figure 5c and 5d, the dashed blue horizon that dips from right to



05/12/11 2:13 PM


b)

c)

d)

Figure 5. Aliasing in an array of four identical traces. The red horizon is the correct trace-to-trace correlation; with increasing vertical shift of adjacent traces, another possible correlation, marked by the dashed blue horizon, appears. a)

b)

Figure 6. Aliasing caused by deleting every other trace from the original four-trace array in Figure 5. The red horizon is the correct trace-to-trace correlation; the dashed blue horizon is an aliased correlation.

left is marked as a possible correlation of the same zero crossing; note that as the dip of the red horizon increases, the dip of the dashed blue horizon decreases, and vice versa. This correlation ambiguity is a manifestation of aliasing, which in this example is related to the trace spacing and the magnitude of the dip (the trace-to-trace shift) of the red horizon. Figure 6 illustrates aliasing behavior by changing the interval between traces (effectively, the trace-to-trace sample rate) rather than by trace-totrace vertical shift, as done in Figure 5. The array of traces in Figure 6a is identical to Figure 5b. In Figure 6b, every other trace has been dropped, effectively doubling the trace interval and halving the trace sample rate (the



05/12/11 2:13 PM

Chapter 6: Resolution 81

left-to-right-dipping red horizon is in the same position on both arrays). In Figure 6b, the undersampled array, the dashed blue horizon dipping from right to left, is shown as a possible correlation of the zero crossing from trace to trace; this aliasing is related to the change in spatial sampling rate. As stated, the Nyquist theorem applies to reconstruction of band-limited functions from equispaced data, and these can be functions of time or distance. Referring to band-limited functions of distance, the wavenumber k of a waveform is defined as the number of wave cycles per unit distance, which is the inverse of wavelength l. These parameters are the spatial equivalents of the frequency f and period T of a time-domain waveform (as shown in Figure 2 of Chapter 2). The spatial sample rate Dx for a 2D seismic line is the common-depth-point (CDP) interval, and for a 3D survey this rate is the bin size. For a given Dx, the sampling wavenumber is 1/(Dx) and the corresponding Nyquist wavelength and Nyquist wavenumber are 2Dx and 1/ (2Dx), respectively. When working with a grid of 2D data, you need to think about the size of a feature that can be resolved in terms of the line spacing of your grid. To illustrate this concern, consider a simple 2D anticline that is X units wide and thus has a wavenumber of 1/X. Again for simplicity, assume that you are working with a grid of 2D lines that is Y units by Y units square, and that the axis of the anticline is parallel to one of the directions of the lines in the grid. The Nyquist wavenumber for this grid of lines is 1/(2Y). The Nyquist theorem states that a feature with a wavenumber larger than the Nyquist wavenumber will be aliased, that is, the anticline will be aliased if 1/X > 1/ (2Y). Using a numerical example, if the anticline is 2 km wide, it will be aliased by a grid with a line spacing greater than 1 km; said another way, you need a grid spacing of 1 × 1 km or less to map this anticline accurately. Geologic features come in all shapes and sizes, and you need to understand resolution as one of the fundamental elements of seismic data quality to represent geology properly in an integrated interpretation. Each seismic data set has a characteristic resolving power. Good practice dictates that you should investigate each of your data sets carefully to be fully aware of temporal and spatial limits of resolution, especially in the context of your interpretation objectives.



05/12/11 2:13 PM



05/12/11 2:13 PM

Chapter 7 Correlation Concepts The most important concept to keep in mind when correlating reflection seismic data is that the reflection seismic method provides indirect timedomain measurements of the 3D space of subsurface geology. The character and position of the reflections you correlate, which are the seismic responses to impedance contrasts across real geologic boundaries, depend on the geometry and properties of the subsurface velocity field — critical factors in stacking quality and in the accuracy of positioning your data (migration), controlling what you see and where you see it on a record section. The fundamentals of acquisition and processing of reflection seismic data and the nature of the earth imply that there will always be some distortion in every seismic image, so you should never consider any seismic section — no matter how carefully processed — to be a true geologic cross section. In other words, all seismic sections require interpretation. As you interpret, you must remember that there is no such thing as a noise-free seismic record, that all seismic data have limits to their temporal and spatial resolving powers, that virtually every 2D seismic line suffers from its inability to image a 3D earth accurately, and that even the most sophisticated 3D PSDM volume will not be an exact replica of the subsurface. Again, these elements of data quality affect the interpretability of reflection seismic data.

First look When taking your first look at a seismic section, you should see the whole section and not focus or concentrate on any particular portion of the image. Scan the section from top to bottom and from side to side, taking in as much of the image as you can to form initial concepts about the geologic setting and overall quality of your data. In the workstation



05/12/11 2:14 PM


environment, use the computer’s capability to modify display size and scale to change the magnification (zoom and unzoom) and/or the aspect ratio (stretch or compress) to facilitate viewing the data. You might also try different color tables to see if any features in the data are more clearly visible (to your eyes, at least) using a particular color scheme. It’s a good idea to inspect all of your data using movie or animation functionality on your workstation before beginning your actual interpretation. This allows you to more fully assess data quality and provides an additional check that you have all of your data. Beyond knowing the acquisition and processing history of your data, at first you should assume as little as necessary about your project area; you’ll have plenty of time and opportunity throughout your interpretation to focus effort and carefully integrate observations and correlations into a consistent story. You are not well served by running off with premature speculation or unwarranted assumptions early in your project. Your initial assessment of data quality — noise content, resolving power, and fidelity of imaging — will affect three important interpretive decisions: 1) Are the data of sufficient quality to deal with the interpretive issues at hand, or is there need for reprocessing or additional acquisition? 2) What fraction of the data can be correlated using automatic versus manual picking (in the workstation environment)? 3) What is the best way to record variations in data quality across the project area? Answers to these questions determine how and with what confidence horizons and faults (if any) can be correlated; the geologic setting and complexity of the project area, together with any prescribed objectives for the interpretation, determine how many horizons (and faults) need to be correlated to describe the geology accurately and to meet business objectives.

Horizons versus faults The essence of correlation is recognizing patterns in seismic data, followed by associating these patterns with known analogs or modeled representations of real geology. At its most basic level, seismic interpretation involves correlating two primary types of geologic surfaces: horizons and faults. A horizon is the surface separating two different rock layers (Sheriff, 2002), which gives rise to a seismic reflection according to the acoustic



05/12/11 2:14 PM

Chapter 7: Correlation Concepts 85

impedance contrast between the two layers (recall Figure 1 in Chapter 1). According to Vail et al. (1977), two types of physical surfaces are present in sediments at the time of deposition: stratal surfaces and unconformities. Each of these can cause seismic reflections if there is a sufficient impedance contrast across it. Stratal surfaces have chronostratigraphic implications. Many are geologic-time surfaces because they are former depositional bedding surfaces that were synchronous over their areas of occurrence; unconformities have chronostratigraphic significance because, by definition, all of the rocks below the unconformity are older than those above the unconformity. A fault is a fracture or fracture zone along which there has been displacement of the two sides relative to one another parallel to the fracture. Sheriff (2002) defines a fault as a displacement of rocks along a shear surface. The fundamental difference between correlating seismic horizons versus faults is that the former is based on recognizing and tracking continuous or predictably changing patterns of reflections, whereas the latter is based on recognizing discontinuities or offsets of patterns that are otherwise continuous or predictably changing (see Figure 1). Of course, faults themselves often can be tracked as predictable patterns of discontinuities,

t

Figure 1. Image of a seismic line, illustrating the difference between correlating seismic horizons as continuous or predictably changing patterns of seismic reflections versus faults as discontinuities or offsets of patterns that are otherwise continuous or predictably changing (courtesy WesternGeco).



05/12/11 2:14 PM


depending on the structural setting and quality of imaging. Occasionally, reflections from fault planes are imaged clearly (see Figure 4 in Chapter 5) and can be used to assist fault picking. In a strictly geometric sense, you can think of correlating horizons and faults as marking the boundaries that define common dip families (packages of reflections with internally consistent character and orientation), after which you explain the geologic nature of the boundaries and then reconstruct the geologic history of the interpreted data. You have two primary concerns in correlating faults; both depend on the quality of your data — specifically, noise content and imaging fidelity: 1) Tracking fault surfaces on individual lines (or on 3D horizontal slices) and from line to line (2D) or across a volume (3D). 2) Accurately correlating horizons across faults. Addressing the first concern, you should naturally pick faults on images that most clearly show the discontinuities, offsets, or reflection terminations that are your evidence for faulting. Whether 2D or 3D, a seismic line that is true dip to a fault, assuming that migration and other processes are done satisfactorily, should afford the clearest and sharpest view of the reflection terminations that you would pick as the fault (the block diagram in Figure 2 illustrates the definitions of strike and dip). In working with 2D data, you are limited to the orientations of lines in your grid of data, which often are not true dip to the faults of interest; so when correlating faults, you will have deal with the problems inherent in 2D imaging (see Chapter 8). In the 3D world, however, if neither the inline nor the crossline direction is in true dip orientation to faults, then you can create arbitrary lines that are in true dip orientation to faults, on which you can very accurately pick the faults, data quality permitting. Figures 3 and 4 are an inline and crossline, respectively, from a depthmigrated 3D survey on which a normal fault has been picked. In this example, the fault in question is more clearly imaged and easily picked on the line that is more nearly true dip to the fault (in this case, the inline) rather than on the orthogonal line (the crossline), on which there is no obvious termination evidence for it. Here, you would pick the fault first on the inline, and then you would use that pick as a tie or reference point to find evidence for the fault on the crossline. Figures 5 and 6 illustrate the difference between dip and strike views of normal and reverse (thrust) faults, respectively. As you can envision, termination evidence for a fault will be hard to see on lines that are parallel or subparallel to the fault trend, no matter whether the fault is a normal or a reverse fault. Note also that regardless of the apparent dip of



05/12/11 2:14 PM

Chapter 7: Correlation Concepts 87 Bedding surface

Horizontal surface

a c

b

Line ab = strike Angle c = dip Figure 2. Diagram defining strike and dip. Strike is the direction of the line formed by the intersection of an inclined plane with an imaginary horizontal plane (line ab). Dip is the angle of the inclined surface below the imaginary plane measured perpendicular to strike, or the true dip (angle c). Apparent dip is the angle of any line on the inclined surface that is not perpendicular to strike. Apparent dip is always less than true dip and is equal to zero along the line of strike.

a fault, interpretation of that fault as normal or reverse depends on how you correlate reflections across it. If you are working with 3D data and decide to correlate faults on horizontal slices, you have the option of picking faults on reflectivity or coherence data. As mentioned in Chapter 3, coherence data are intended to highlight discontinuities such as faults, and these data can provide excellent images on which you can pick faults with good confidence. The examples illustrated in Figures 6 and 7 of Chapter 3 show that fault picking can be done with considerably more confidence and accuracy on coherence data than on the reflectivity data from which the coherence data were generated. However, this result depends heavily on the quality of the parent reflectivity data. Not all discontinuities that you observe on seismic data are necessarily faults; there are real geologic features such as unconformities that mark discontinuities between reflection packages, and there are imaging



05/12/11 2:14 PM

88 First Steps in Seismic Interpretation a) z

b) z

Figure 3. (a) Uninterpreted version of an inline from a depth-migrated 3D survey (displayed with no vertical exaggeration). The vertical black line marks the intersection of this image with the 3D crossline shown in Figure 4. (b) Interpreted version of the same 3D inline with a normal fault shown by the solid yellow line. The red dot marks the intersection of the fault on the crossline image in Figure 4 (courtesy PGS).



05/12/11 2:14 PM

Chapter 7: Correlation Concepts 89 a) z

b)

z

?

Figure 4. (a) Uninterpreted version of a crossline from a depth-migrated 3D survey (displayed with no vertical exaggeration). The vertical black line marks the intersection of this image with the 3D inline shown in Figure 3, and the red dot marks the intersection of this image with the normal fault picked on Figure 3. (b) Interpreted version of the same 3D crossline with a normal fault shown by the dashed yellow line. The control point provided by the fault as picked on the inline guides the picking of the same fault on the crossline, on which the termination evidence is much less distinct (courtesy PGS).



05/12/11 2:14 PM

90 First Steps in Seismic Interpretation Dip line to fault

Strike line to fault up down

Dip line to fault

Map view

Strike line to fault

Seismic horizon

Vertical views

Figure 5. True dip and strike views of a listric normal fault.

Dip line to fault


Map view

Dip line to fault


Seismic horizon

Vertical views

Figure 6. True dip and strike views of a reverse (thrust) fault.



05/12/11 2:14 PM


artifacts that appear as discontinuities and can easily be mistaken for faults. You might say that all faults are discontinuities in reflection patterns, but not all discontinuities in reflection patterns are faults. Herron (2000b) discusses an example of a discontinuity observed on depth-migrated data that might have been interpreted as a fault but actually was an imaging artifact. This discontinuity was observed on a first pass of prestack depth migration (Figure 7) that used salt-body geometry interpreted on poststack depth-migrated data. The discontinuity was “healed” on a second pass of prestack depth migration (PSDM) that used revised salt-body geometry based on interpretation of the first pass of PSDM (Figure 8). In addition to serving as a warning that depth-migration artifacts can masquerade as faults, this example emphasizes that depth migration is an iterative process in which multiple passes of imaging (time and funding permitting) are often required to achieve accurate results.

z

Figure 7. First iteration of PSDM using the base of salt, shown as the light blue horizon, from interpretation of a poststack depth-migrated data set. The discontinuity in question is the trend of disrupted reflections, marked by the yellow arrows. The base of salt picked on this image is shown as the green horizon (courtesy WesternGeco).



05/12/11 2:14 PM


z

Figure 8. Second iteration of PSDM. The discontinuity seen in Figure 7 is no longer apparent. This second iteration uses the updated base of salt, shown as the green horizon (courtesy WesternGeco).

An unconformity is an important geologic boundary that has correlation aspects of both horizons and faults. Although you typically correlate the seismic response to such a boundary as a horizon — as a coherent or predictably changing reflection — it actually involves tracking reflection terminations or discontinuities between groups of reflections that are genetically related and distinctly different in character from other groups of reflections. These are the so-called seismic sequences above and below the unconformity. At times, you will have to pick an unconformity manually because of its discontinuous and variable character caused by lateral variation in impedance contrast between the post- and preunconformity sections, but in other settings you will be able to automatically track an unconformity quite easily because of its laterally consistent signature (as in Figure 9). Recognizing and accurately tracking unconformities and “their correlative conformities” (Mitchum, 1977) is of central importance in seismic stratigraphic interpretation.



05/12/11 2:14 PM

Chapter 7: Correlation Concepts 93 a)

t

b)

t

Figure 9. (a) Uninterpreted 2D time-migrated line, showing a well-imaged angular unconformity. (b) Enlargement of a portion of (a), showing detail of reflection terminations that mark the position of the unconformity (courtesy PGS).

In addition to correlating/picking horizons, faults, and unconformities as the fundamental elements of an interpretation project, at times you will be called on to identify and map the distribution of distinctive patterns of reflections. Always observe and note reflection patterns and configurations, even when you do not intend to map them. This activity is the part of seismic stratigraphic interpretation known as seismic facies analysis, where the term analysis includes not only identifying and mapping reflection patterns and their bounding surfaces but also calibrating the patterns to well control. Mitchum et al. (1977) define seismic facies as a group of seismic reflections



05/12/11 2:14 PM


whose parameters (configuration, amplitude, continuity, frequency, and interval velocity) differ from adjacent groups. Sheriff (2002) defines them as the character of a group of reflections involving the general amplitude, abundance, continuity, and configuration of reflections. From these definitions, it is clear that accurate and reliable identification of seismic facies depends heavily on seismic data quality and requires pattern-recognition skills. Although you can analyze seismic facies on any good-quality seismic data, 3D data greatly facilitate this activity because the power of 3D imaging allows you to view facies from different angles and perspectives, thus enabling more confident interpretation. The standard classical reference for principles and practices of seismic stratigraphic interpretation is Seismic Stratigraphy — Applications to Hydrocarbon Exploration (Payton, 1977). Other sources of information on this topic are Bertram and Milton (1996), Hardage et al. (2011), and Hart (2011).

Multiple reflections Your fundamental skills in pattern recognition, knowledge of geology, and ability to assess data quality affect your ability to identify and correlate valid and geologically meaningful patterns of reflections. These are the primary reflection responses that represent true subsurface geology, whether they are horizons, faults, or unconformities. At the same time, this knowledge and experience gained over time enables you to correctly identify and avoid correlating spurious features such as multiples, imaging artifacts, and noise. These nonprimary reflections should be clearly marked on interpreted lines so as not to be included in the interpretation of primary events. Figure 10 shows that in contrast to a primary reflection, one that represents seismic energy which has been reflected only once, a multiple reflection is energy that been reflected more than once. Figure 11 illustrates three types of multiples commonly encountered when interpreting reflection seismic data (see Figure M-13 in Sheriff [2002] for a more complete summary of multiple types). A multiple reflection is often described by its period, which is the difference between the traveltime of the multiple and the traveltime of the primary reflection from the deepest reflecting point on the multiple’s travel path (see Sheriff [2002] for corresponding definitions of long- and short-path multiples). When describing multiples, the terms longperiod and short-period are relative, in the sense that they depend on the depths to and distances between the primary and the multiple-generating reflecting boundaries and also on propagation velocities. For example, in Figure 11, the traveltime of the double multiple is longer than that of the



05/12/11 2:14 PM

Chapter 7: Correlation Concepts 95 Black = Downgoing

Red = Upcoming

Depth Primary

Multiple

Figure 10. Schematic of primary and multiple reflections. The primary reflection is seismic energy that has been reflected only once; a multiple reflection is seismic energy that has been reflected more than once. Double

Ghost

Peg-leg

Depth

Figure 11. Three common multiples.

peg-leg multiple, but you can’t absolutely characterize either of them as long period or short period unless you know something about depth and velocity. You must always be aware that residual multiple energy may remain in data, even though demultiple processing has been applied. There are no perfect demultiple techniques. When trying to decide whether a particular reflection might be a multiple, you should be able to identify the primary



05/12/11 2:14 PM


reflection from the interface that served as a multiple-generating boundary before concluding that the reflection in question is in fact a multiple. You can often recognize multiples because they cut across primary reflections. Multiples usually are generated at the surface or at interfaces in the shallow subsurface where the geology is relatively simple and the arrival times of the multiples are close to or coincide with those of primary reflections from deeper and usually more highly structured reflectors — hence, the crosscutting reflections. These multiples also have relatively high amplitudes because they are generated at interfaces across which there are relatively high impedance contrasts (such as the seafloor in marine data). For these reasons, a long-period multiple such as the seafloor double multiple is particularly obvious as a crosscutting event on marine seismic data, especially in deepwater and ultra-deepwater settings. By the same reasoning, short-period multiples such as peg-leg or intrabed multiples are not easy to identify and separate from primary reflections; they interfere with the primaries and hinder accurate interpretation of the primary reflection signatures. These multiples are especially difficult to remove in processing. The positions of multiples can be predicted simply on time-processed data using the reflection times to primary horizons that serve as the multiple-generating boundaries. These positions are calculated by summing the traveltimes for intervals between primary horizons, as shown in Figure 12; the real data example for these calculations is shown in Figure 13. Notice in these figures that the dips (in time) of the multiples are exaggerated in comparison to the dips of the primary horizons, making the multiples easier to identify. This technique for predicting multiple positions cannot be used on depth-processed data because the multiples are placed in depth according to their traveltimes through the velocity model used for the depth imaging (see Figure 14).

Manual versus automatic tracking In the modern workstation environment, you correlate or pick horizons manually or automatically, choosing between these two modes of picking on the basis of reflection continuity and ultimately on data quality (Herron, 2000a). Manual picking or tracking on a workstation is exactly what it sounds like — the computerized equivalent of what used to be done by hand with a colored pencil on a paper section, literally connecting the dots to create a curve that follows the trend of the reflection you are correlating. In contrast, automatic picking, or autotracking, is done on data for which a computer algorithm can accurately and, obviously, more rapidly reproduce what was formerly done by hand, reflection continuity



05/12/11 2:14 PM


a)

T1

b) T2

c)

d) T2

T2

T1 T2

Figure 12. Method for calculating positions of multiples on time-processed data. (a) Blue horizon DT1 is the two-way reflection time to the seafloor, (b) red horizon DT2 is the two-way reflection time to the top of salt, (c) dashed purple horizon is the seafloor/top-of-salt peg-leg multiple (DT2 + DT1), and (d) dashed red horizon is the top-of-salt double multiple (DT2 + DT2). t

Top of salt

Seafloor/top-of-salt peg-leg multiple Top-of-salt double multiple

Figure 13. A 2D time-migrated seismic line, showing seafloor/top-of-salt pegleg multiple and top-of-salt double multiple. These data are the basis for the calculations illustrated in Figure 12. The seafloor double multiple appears faintly below the top-of-salt reflection on the left side of the display (courtesy BP).



05/12/11 2:14 PM

98 First Steps in Seismic Interpretation 0 ft P

M z

35,000 ft

Figure 14. Multiple reflections on 3D depth-migrated data. These data are the product of a 3D PSDM sediment flood on which no demultiple processing was applied. Primary reflections identified as the top and base of salt are marked with dashed blue lines. Note that the positions of multiples cannot be calculated without using information from the PSDM velocity model. For example, you cannot calculate the position of the seafloor double multiple M (green arrow) by doubling the distance to the primary seafloor event P (red arrow); rather, you must calculate the distance traveled through the PSDM velocity model for the two-way reflection time of the double multiple. The depth of the double multiple is more than twice the depth of the seafloor because the sediment velocity is greater than seawater velocity and increases with depth (courtesy BP).

and data quality permitting. As the interpreter, you set the parameters that control the operation of an autotracking algorithm, again based on your assessment of data quality. As few as one or two or as many as seven or eight parameters, depending on the sophistication of the algorithm, are usually set to default values that you can modify. You choose final tracking



05/12/11 2:14 PM


parameters based on testing ranges of values on representative samples of your data. In addition to selecting parameters for autotracking, you must also decide on the number and locations of seed points for autotracking, which are the starting points for correlations to be made using the chosen tracking parameters. No one set of tracking parameters is necessarily optimal for a given horizon in a given project area. It is most common for these parameters to vary across any area, especially large areas — a consequence of naturally occurring spatial variations in geology as well as nongeologic variations in data quality. In every case, the final tracked horizon must represent your view accurately, regardless of how it was picked. To ensure that this is so, when using an autotracking algorithm, you must devote time to qualitychecking your results to be sure the tracker has placed picks in the same places that you would have. An equally critical aspect of autotracking is to remember that you are responsible for understanding how an autotracker works before using it. Most if not all fault picking is done manually, although some workstation systems have automatic fault tracking algorithms whose effectiveness is directly proportional to the clarity of evidence for faulting and, again, ultimately depends on data quality. Figure 15 illustrates the observations that you make in deciding whether to pick a horizon manually or automatically. Figure 15a shows an uninterpreted 3D depth-migrated line on which the most prominent reflection is a peak rising from the lower-left corner of the image and extending all the way across the section. Regional knowledge indicates that this event marks the top of a salt sheet. Figure 15b shows the results of autotracking two horizons that you decided could be autotracked on the basis of event continuity and overall data quality. Both horizons were seeded with single control points positioned on the left side of the image (exact locations of the seed points are not shown). A comparison of Figure 15a and 15b shows that the upper horizon terminates at the point where it no longer has the continuity or character that it does at its seed point or over the extent where it was successfully autotracked. This termination suggests that the autotracking parameters for the horizon on this line are acceptable, i.e., the autotracker stopped where you would have stopped. Also, there is a short segment of picks where the autotracker made its pick above the overall smooth trend of the horizon; this relatively small error, or mis-pick, is a result of the autotracker trying to honor a relatively high-amplitude reflection slightly above the smooth trend of the horizon. It could be corrected by adding more seed points or manually tracking in the immediate vicinity of the mis-pick.



05/12/11 2:14 PM

100 First Steps in Seismic Interpretation a) z

b) z

c) z

Figure 15. (a) Uninterpreted 3D depth-migrated line. (b) Partially interpreted version of (a), showing two autotracked horizons, the deeper of which marks the top of a salt sheet. (c) Partially interpreted version of (a), showing a manually tracked horizon (light blue dashed line) that marks the base of a channel. The light blue horizon cannot be autotracked because in this example the seismic response to the base-of-channel surface does not have a distinctive or laterally continuous character (courtesy WesternGeco). Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 2:14 PM


Further inspection of the tracking results on Figure 15b indicates that the shallow horizon terminates at a left-to-right-dipping boundary above the top-of-salt horizon that separates two intervals with distinctly different internal reflection configurations. This boundary, which becomes less well defined from left to right across the image, can be drawn with a concave-upward shape (the dashed light blue horizon in Figure 15c) and is interpreted as the base of a channel, the axis of which would be roughly perpendicular to the plane of the image. Even though this boundary can be observed clearly on the image, the horizon picked to represent it must be tracked manually because there is no consistent reflection response suitable for autotracking. In other words, the fundamental nature of geology and the seismic response to geology can and do conspire to prevent you from autotracking every surface of interest, no matter how good your data quality might be. You should always inspect your data at the level of the horizon to be autotracked to assess data quality, geologic complexity, and suitability for autotracking because these factors determine not only your choices for tracking parameters but also where and how many seed points you will need to pick. Figure 15 illustrates that for relatively simple geology and good data quality, you need very few seed points for good tracking (in this case, only one per horizon). But in general, more highly variable data quality and greater complexity mean that you need proportionally more seed points for accurate tracking. In this context, you should be careful not to waste time picking more seed points than the autotracker needs to operate efficiently and accurately. Your goal is to strike the right balance between time spent on picking seed points and adjusting tracking parameters and the number of tracking iterations required to achieve acceptable results. As with many other aspects of seismic interpretation, you develop this balance with experience. Autotracking errors are most evident on vertical seismic sections. An example of an autotracking error (not all are this obvious) is shown in Figure 16, on which the red horizon (picked as a trough) was tracked from right to left. The automatic pick is accurate as far to the left as the yellow arrow. Beyond this point, it is seriously wrong, having jumped irregularly from cycle to cycle and eventually following a migration artifact to the far left against the base-of-salt reflection. The pick is not where you would have put it. The red horizon must be tracked manually to the left of the yellow arrow on this image in order to produce a geologically reasonable representation of the structure below salt. Figure 17 is an example of an autotracking error caused by failure to constrain tracking properly with seed points and fault picks. The red horizon



05/12/11 2:14 PM


z

Base of salt

Direction of tracking

Figure 16. An autotracking error. The automatic pick for the red horizon is obviously wrong to the left of the yellow arrow. This horizon should be tracked manually below the salt on this image (courtesy BP).

in Figure 17a, seeded as a trough, has been tracked incorrectly across a fault (not shown on this figure) that cuts the horizon at the point marked by the yellow arrow. The picked horizon is not offset by the fault; you would say that the horizon as picked does not honor the fault. This horizon was tracked using a single seed point, and the results were the same whether the seed was placed to the right or the left of the point of intersection with the fault. Note that although the correlation across the fault is incorrect, you cannot tell the exact nature of the error from the display alone without knowing the location of the seed point; it is either to the left of the fault, in which case the correlation across the fault is one cycle too high, or to the right of the fault, in which case the correlation across the fault is one cycle too low. This type of miscorrelation commonly occurs where reflections align across a fault in such a way that the autotracking algorithm fails to recognize the discontinuity. The fault is effectively invisible at the level of the individual event being tracked but is clearly evident upon inspecting a broader window of data. You can solve this problem by picking the fault before autotracking and by specifying seed points in the upthrown and downthrown fault blocks. The correct correlation across the fault for a seed location to the left of the fault is shown as the dashed red horizon in the downthrown fault block in Figure 17b; note that several other faults on this line probably



05/12/11 2:14 PM

Chapter 7: Correlation Concepts 103 a) t

b) t

Figure 17. (a) An autotracking error caused by failing to pick a fault before tracking and not specifying seed points on both sides of the fault. The yellow arrow shows the point at which a down-to-the-right extensional fault intersects the red horizon. (b) The actual fault pick. Note the down-to-the-right extensional fault marked in yellow, which actually offsets the red horizon. The dashed red horizon to the right of the fault is the correct correlation of the red horizon, given an initial seed point to the left of the fault (courtesy WesternGeco).

would need to be picked before autotracking to prevent similar miscorrelations. Most if not all interpretation applications can automatically stop or block autotracking at specified boundaries such as previously picked



05/12/11 2:14 PM


faults, meaning that you should pick faults before attempting to autotrack horizons in faulted areas. You do most but not all quality control of autotracking results using a map view of the autotracked horizon, although you make corrections to the input grid or seed horizon on vertical sections. Following are some of the observations you can make on the map view of an autotracked horizon that indicate probable errors in autotracking: 1) Abrupt changes in display colors which, when highlighted by dynamically adjusting or flexing the color table, suggest that the autotracked horizon has jumped cycles. 2) Holes or voids in the horizon where no automatic pick was made. These should be investigated on vertical sections through the areas in question to help you decide whether they represent real geology, such as fault heaves, or poor tracking, in which case you may need to adjust tracking parameters. 3) Sharp or angular features in the horizon that do not look geologic. 4) Features that follow the 3D inline or crossline directions. With reference to Figure 17, consider that you would not see the miscorrelation of the red horizon on this line as a discontinuity in map view because the horizon was smoothly but erroneously tracked through the fault. This tracking error, so clearly evident on a vertical section, would probably appear as an abrupt change or discontinuity in areas of the map where the fault displacement is larger or smaller and where reflections are no longer fortuitously aligned across the fault. Autotracking errors are corrected by editing existing picks, adding seed points, and/or updating tracking parameters. You run the sequence of seed picking–autotracking–quality control, updating iteratively until you are satisfied that the autotracked horizon is a sufficiently accurate representation of the surface as you would pick it, in the extreme, on every line in your data set. A final interpretation most often is a composite of different regions in which the horizon was manually or automatically tracked as dictated by data quality and geologic complexity, and you are responsible for keeping track of (no pun intended) which picking mode was used for which portions of a horizon. The importance of quality control in autotracking results cannot be overstated: There may be nothing more damaging to your project results, and to your reputation, than to have to apologize for autotracking errors that surface long after you completed your interpretation.



05/12/11 2:14 PM


Artifacts and interpretation pitfalls Despite efforts to produce noise-free reflection seismic data — the ideal seismic response of Figure 1 of Chapter 1 with maximum interpretability — there always will be some unwanted information in your data that will complicate or obstruct your interpretation. So you will need to learn how to recognize seismic artifacts, the undesirable effects of seismic acquisition and processing. Artifacts can be quite obvious or extremely subtle. In any case or degree of severity, they are hidden or deceiving pitfalls or difficulties that, if not correctly identified and properly handled, can completely invalidate an interpretation. These pitfalls will test your correlation skills sharply, and there is no doubt that, as an interpreter, you are ultimately responsible for recognizing seismic artifacts and pitfalls and dealing with them accordingly. Your ability to do so will depend on your knowledge of data acquisition and processing and your level of experience. Interpreters have been aware of seismic artifacts and pitfalls since the first day that seismic data were correlated from one station to another. Many of those artifacts might have been expected or anticipated through understanding the seismic method and its attendant limitations and ambiguities, while others probably were quite surprising, becoming evident only after drilled wells revealed what neither the seismic data nor the most thorough and insightful interpreters could ever have seen. We most often learn by our own mistakes, and we don’t really learn anything by guessing correctly. But because we usually are not eager or able to publicize errors, we find that opportunities to learn from the mistakes of others — from their close encounters with interpretation pitfalls — are not that numerous. One of the earliest publications that directly addresses seismic artifacts and pitfalls and the lessons to be learned from them is Tucker and Yorston’s (1973) classic Pitfalls in Seismic Interpretation, followed by the companion volume, Pitfalls Revisited (Tucker, 1982). The former is one of the earliest (if not the first) publications to draw attention to the different types of pitfalls into which the well-intentioned but unsuspecting interpreter can stumble. Tucker and Yorston describe and tabulate examples of three categories of pitfalls: velocity, geometry, and recording (acquisition) and processing. In the follow-up publication, Tucker adds two categories to the original list: stratigraphic mapping and general, the latter comprised of pitfalls that are more behavioral than technical in nature. Although both books include seismic images that are now dated, their treatment of pitfalls is still valuable for its candor and historical perspective. After reading the texts, you can easily imagine that even with the technological advances of the decades that have



05/12/11 2:14 PM


passed since they were written, you are not yet and may never be completely free of interpretation pitfalls. Perhaps the greatest danger or pitfall that you may face is correlating overly optimistically in low signal-to-noise (S/N) regions and then failing to properly communicate your lack of confidence and uncertainty (remember Dix’s threshold of impossibility in Chapter 1). You should persistently query your data and be aggressive in asking questions such as the following: • • • •

How likely is it that this reflection is not primary signal? What else might this be? Is there a geologically reasonable explanation for this? Am I being too optimistic, or should I be more skeptical?

You should avoid special pleading to support an interpretation, however optimistic, that is difficult to explain by sound geologic reasoning. By no means is optimism bad; vast quantities of oil and gas have been discovered by explorers who were not afraid to be optimistic. However, optimism should be tempered by the realities of physics and geology. This is the unmistakably human factor inherent in every seismic interpretation. Following are some of the more common interpretation pitfalls and related seismic artifacts. Interpreting coherent noise such as multiples or migration swings as primary signal. — Picking the seafloor/top-of-salt peg-leg multiple as the base of salt (Figure 13) is an excellent illustration of erroneously interpreting a nonprimary reflection as a primary event. Your first step in investigating whether a reflection might be a multiple is observing the similarity of its geometry to another event that you know or strongly suspect is a primary event. You should investigate which primary event on your data is the multiple-generating interface for the event that you think might be a multiple, what the travel path is for the multiple, and what reflectors are involved. Although there are not always definitive answers to these questions, they can show you the way to proceed with your interpretation. Often, it is helpful to sketch a simple diagram illustrating a possible travel path for a suspected multiple. In your analysis of the multiple, be sure to account for any polarity reversals at reflection points along its travel path. For example, the seafloor/top-of-salt peg-leg multiple in Figure 13 should be opposite the polarity of the top-of-salt and seafloor events because the travel path of the multiple includes a reflection of upcoming energy at the sea surface. This polarity reversal of the multiple can be seen clearly on the left side of the figure but less so in the center and right side, where the



05/12/11 2:14 PM


reflection response appears to be more complex. Notice that the polarity for this image appears to be opposite that of the SEG positive standard display convention (Figure 7 of Chapter 2), in that the seafloor reflection is a symmetric trough with well-developed side-lobe peaks. In many instances, coherent noise such as residual multiples is obvious because it cuts across primary (or what are interpreted to be primary) reflections, as shown in Figure 18. At the same time, the observation of two crosscutting reflections or sets of reflections in itself does not mean that one must be an artifact. Figure 19 shows an example of the reflection from the base of a gas hydrate stability zone (GHSZ) that cuts across more steeply dipping primary reflections; such an event, which is often mistakenly interpreted as a multiple reflection, is commonly called a bottom-simulating reflection (BSR) (see the May 2006 issue of The Leading Edge for an excellent collection of papers on the geology and geophysics of gas hydrates). Similarly, the flat spot shown in Figure 20 clearly cuts across primary dipping reflections, and there is no doubt that it is a primary reflection from a real hydrocarbon/ water contact. But you can also see in this figure a set of steeply dipping migration “tails” that equally clearly are imaging artifacts cutting across more gently dipping primary reflections. Especially in areas that require

t 2.0

4.0

6.0

Figure 18. A 2D time-migrated seismic line on which multiple reflections cut across primary dipping reflections. Multiple reflections such as the seafloor double multiple and peg-leg multiples from reflectors in the shallow subsurface are most clearly evident in the left and center portions of the line. The positions of these multiples can be calculated easily using the method illustrated in Figure 12 (courtesy BP).



05/12/11 2:14 PM


z

Figure 19. A 3D depth-migrated line showing the cross-cutting primary reflection (red arrow) from the base of the GHSZ in the deepwater Gulf of Mexico. No simple or geologically reasonable travel path could give rise to this event as a multiple reflection. Its depth below seafloor and polarity (a trough on an SEG positive standard polarity display) are consistent with it being interpreted as the base of the GHSZ (courtesy PGS). t

Figure 20. The 2D time-migrated image shown in Figure 10 of Chapter 2 on which the reflection from a flat-lying hydrocarbon-water contact cuts across primary dipping events. Crosscutting coherent noise, probably a migration artifact, is highlighted by the dashed red circle (courtesy PGS).



05/12/11 2:14 PM


complex imaging (PSDM) you will find that correlating dipping reflections as primary events versus migration artifacts is ambiguous and interpretive. Interpreting noise as discontinuous signal. — Interpreting data with very poor S/N can involve correlating random or discontinuous noise segments that accidentally or coincidentally align or are aliased along an interesting or favorable trend. This is tantamount to “seeing what you want to see” because in doing so you allow yourself to fit a tracked horizon to a prospective geometry. These correlations reflect the natural tendency toward optimism in most interpreters, which is not bad as long as it is balanced with well-founded geologic thinking (for example, comparing an interpretation to an established analog) and supported by other geophysical analyses such as seismic modeling. An excess of optimism can lead to correlations that are “almost sure to be misleading and therefore very expensive” (Dix, 1952), while a dearth of optimism can cause you to miss good exploration opportunities. Hence the need for balance. Interpreters at all levels of experience are susceptible to the tendency to interpret noise as discontinuous signal when they feel compelled or are instructed to “pick something, no matter how poor the data are.” This tendency has aspects of a condition or experience known as apophenia, a term coined by German psychiatrist Klaus Conrad that refers to seeing patterns or connections in random or meaningless data. You are well advised to heed the advice offered by Dix (1952): There is, of course, nothing to be gained by straining the imagination to such an extent that reflections are seen where none exist. Such a ridiculous procedure should be avoided. Though it is ridiculous it is quite natural, especially in those areas where a certain zone of weak reflections is known to be of major importance. Interpreting false time structures as real depth structures. — This pitfall is a result of lateral changes in velocity, which in time-domain imaging is commonly manifested as velocity pull-up, whereby a high-velocity body such as shallow salt in a normally compacting sedimentary basin creates false structure below it. Alternatively, an anomalously low-velocity body or interval causes a time sag that depresses the structure below it, as frequently observed in areas with shallow gas accumulations. Often the key to identifying these spurious, velocity-related features is the observation of coincidence, a particular feature that just happens to line up with another feature, the false image underlying the causative high- or low-velocity body. Figure 21, in which deep synclines underlie shallower anticlines one for one, very



05/12/11 2:14 PM


t

Figure 21. A 2D time-migrated line on which synclines directly underlie anticlines. The vertical (and coincidental?) alignment of these structures suggests that the synclines may be distorted images caused by anomalously low velocities in the overlying section, perhaps associated with gas-charged or overpressured sediments. Careful layer-based depth conversion or depth imaging would be needed to accurately resolve the true structure (courtesy PGS and the Ministry of Commerce, Industry and Tourism of the Republic of Cyprus).

strongly suggests that the image is probably distorted, in this case owing to anomalously low-velocity material, perhaps gas-charged or overpressured sediments, in the interval above the synclines; there is no simple or straightforward explanation for these synclines as primary structural features. Similarly, observing the coincidence that the “fault” shown in Figure 7 originates at the tip of the salt body on the left side of the image should cause you to take a second look at your salt-body geometry and depth-migration velocity model. This fault geometry is not geologically impossible, but it isn’t all that common either. The coincidence gives reason to investigate the depth imaging further, and the results of the second pass of depth imaging shown in Figure 8 essentially confirm that the “fault” on the initial pass of PSDM is an artifact.



05/12/11 2:14 PM


A corollary to interpreting false time structures as real depth structures is interpreting false structures in seismic depth as real structures in true depth. This pitfall occurs in the depth-imaging domain and is caused by errors in a PSDM velocity model that are not large enough to significantly degrade the focusing of reflections but are large enough to position those reflections inaccurately. As mentioned in Chapter 4, isotropic depth-imaged data require a Z-to-D correction to account for anisotropy, but anisotropic depth-imaged data may need similar (usually smaller) correction, especially in areas between control points for calibrating the velocity model if the anisotropy parameterization of the model is not spatially accurate. You should carefully investigate the effects of perturbing velocity and anisotropy values in a PSDM velocity model on the resulting depth images, especially when the exploration targets have low relief. An example of this type of pitfall is shown in Figure 22, a line from a 3D isotropic depth-imaged volume on which very gently dipping reflections

z

Figure 22. A 3D isotropic depth-migrated line on which an apparent dip reversal in the subsalt section coincides with the edge of a salt sheet and an overlying seafloor scarp. The question is whether the dip reversal is real: Could it be a depthimaging artifact caused by a slightly inaccurate migration-velocity model? The vertical exaggeration of the image is approximately 2.5 to emphasize the relatively low dips in the subsalt section (courtesy PGS).



05/12/11 2:14 PM


on the left side of the image appear to be folded as they extend to the right below a shallow salt sheet. The location of the “anticline” in the center of the image coincides with significant changes in water depth and thickness of overlying salt, so you should question whether the apparent folding is an artifact of the migration-velocity model, and, ultimately, whether the subsalt section is deformed or tilted at all. In areas with significant lateral velocity variations, depth imaging is now routinely done to improve image quality and reduce distortion; however, even in areas where velocities are supposed to be well behaved, that is, have little or no lateral variation and depth imaging is not thought to be necessary, there can still be significant image distortion, especially when dealing with subtle or gently tilted or warped (long-wavelength) structures. In these areas, you should be very careful with time-to-depth conversion and be sure to characterize the uncertainty in conversion. In a similar way, when working with land data, you should always check that statics corrections, especially long-wavelength corrections, have been properly calculated and applied. These corrections are notorious for erroneously creating or destroying structures. When interpreting on partial stacks, failing to recognize that incorrect moveout or migration-velocity error has created spurious high-coherence events or destroyed real signal. — In working with partial stacks, you should always check prestack data to be sure the data have been moved out correctly, particularly when amplitude variation with offset (AVO) or amplitude variation with angle of incidence (AVA) effects may be present. In cases of complex imaging, examine depth gathers to see that primary events are flattened properly and, in terms of target illumination, that the appropriate angle ranges are being stacked. Variable amplitude response (or lack of response) can easily arise if moveout is not applied correctly. As Figure 16 in Chapter 4 shows, even with moveout based on the best picked velocities, the appearance of stacked data can vary markedly, depending on the range of offsets included in the stack. Assuming that wavelet phase is known and then misinterpreting observed amplitudes. — Although determining or estimating wavelet phase does not involve seismic artifacts in the strictest sense, it is nonetheless an interpretation pitfall that warrants mention, particularly as it applies to working with seismic amplitudes and attributes in general. As discussed in Chapter 2, knowledge of wavelet phase is critical for correct geologic interpretation of individual waveforms at the fine scale of thin beds and closely spaced reflectors. Assumption or plain ignorance of wavelet phase



05/12/11 2:14 PM


can result in serious misinterpretation of lithology and/or pore-fluid type from seismic attributes. As an example of this type of pitfall, consider the following story. A well-meaning geophysicist interpreted a trough-over-peak reflection on what were thought to be zero-phase data displayed with the SEG positive standard convention. He interpreted this reflection as a prospective gas sand, mapped it in a trapping configuration, and proposed an exploration well. The target reflection turned out to be the seismic response of an acoustically soft (low-impedance) shale overlying an acoustically hard (high-impedance) shale; there was no gas reservoir. This particular shale-over-shale configuration has a trough-over-peak signature on data that are phase-shifted by –90° and displayed with the SEG positive standard convention, so it can easily be misinterpreted as a gas sand. Postdrill analysis revealed that the original seismic data were in fact not zero phase, and many people were justifiably upset, thinking the expense of the well could have been saved if only the phase of the seismic data had been known. Recognizing artifacts requires that you maintain a healthy skepticism about the information content of your data and a keen sense of knowing what to throw away and what to keep. Remember that the acquisition and processing histories of your data often contain information that can help you decide whether certain reflections in your data might be artifacts. When in doubt, conduct additional studies such as seismic modeling to investigate suspected artifacts, and do not hesitate to seek the advice and counsel of more experienced interpreters. You should always keep an eye out for seismic artifacts, especially in the more challenging exploration areas where complex imaging is becoming commonplace and artifacts are likely to occur. Again, your skill in recognizing artifacts develops over the course of your career and is fostered by making and taking the time to look at as much data as you can.



05/12/11 2:14 PM



05/12/11 2:14 PM

Chapter 8 Correlation Procedures This chapter describes the basic correlation procedures used in a typical interpretation project, beginning with how to start an interpretation and then discussing fundamentals of the two main correlation techniques (loop tying and jump correlation). The section on correlations in depth-migration projects provides helpful information and guidance for handling aspects of interpretation that are specific to working with depth-migrated data, especially for building velocity models. The discussion of visualization emphasizes the importance of this procedure for validating correlations and communicating interpretation results. The chapter closes with a summary of individual processes and an example of a generic interpretation workflow.

Getting started After checking to be sure that all data (seismic, well, cultural, potential field, etc.) for your project are in hand, take time to review the data before beginning correlations. You often make many important observations with a minimum of bias at this early stage of a project. Here are the steps you should follow: 1) Look at the horizontal and vertical extents of the data to gain a sense of their dimensions and the scale of your project. 2) Become familiar with the orientation of the data (strike and dip directions and their areal consistency) with respect to the project base map and in terms of absolute compass directions. 3) View the data on a display scale that shows the entire extent of a line, and inspect the shallow section for (a) near-surface features such as channels (with associated velocity anomalies that can distort underlying time structure) or amplitude anomalies (possible drilling hazards), 115 Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:49 PM


(b) on land data, time structures that correspond to applied statics, and (c) velocity anomalies that create false time structures (e.g., pull-up under shallow salt). 4) Experiment displaying your data in workstation-based projects with different color tables or schemes to see if any particular one favors or complements your pattern-recognition skills or more effectively enhances features of interest. Color tables are not “one size fits all” for all interpreters and all data sets. The use of color in seismic interpretation is very personal; some interpreters prefer to use a grayscale color scheme for structural interpretation (see Figure 22 in Chapter 7), whereas others use multicolored schemes (see Figures 3 and 4 in Chapter 3) when working with seismic attributes. Still others use a rainbow scheme (Figure 23 in Chapter 4) when corendering a migration-velocity model with trace data. You should not hesitate to create a custom color table if doing so helps you interpret your data and communicate your results. Brown (2011) contains many useful illustrations, comments, and suggestions regarding color in seismic interpretation. 5) Look at the processing sequence applied to the data (often this is not readily available on a workstation project) and review in particular the type of migration used to see if that migration was appropriate for the structure and velocity complexity of the project area. You select horizons for interpretation and mapping early in a project; some horizons may be more geologically significant or more obvious than others, and in some projects there is neither time nor need for picking the entire set of common dip families that make up an individual seismic line or a whole data set. These choices, which often involve correlation with other control such as well information by way of well ties, are determined by the exploration objectives for the project, the geologic complexity of the study area, and the quality of the available seismic data. On occasion, you modify the number of horizons and faults that you interpret during a project as you learn more about the study area. Within the limits of project objectives and available time, there is no point picking horizons (boundaries) and faults that do not contribute meaningfully to building the geologic history of an area. Although there might be many interesting features and surfaces visible in your data, attempting to identify and track all of them can compromise the objectives of an interpretation, destroy interpretive focus, and waste time and resources. Correlating reflection seismic data to well information (logs, cuttings, cores, biostratigraphic and geochemical data) is a well tie (correlate = tie). It establishes the fundamental link between indirect geophysical measurements



05/12/11 7:49 PM

Chapter 8: Correlation Procedures 117

and real geology. Well ties very often involve using a model-based seismic response known as a synthetic seismogram that is created by convolving a seismic wavelet, preferably one extracted from the seismic data to which the synthetic will be correlated, with a reflection-coefficient (RC) series generated from acoustic impedance (AI) data calculated from sonic and density logs collected during a borehole evaluation program. Well ties also are made using reflectivity data from vertical seismic profiles (VSPs), as discussed in Chapter 4. In either case, you correlate the well-based seismic response to the surface seismic data based on the character of individual reflections and distinctive reflection packages or intervals, as shown in Figure 1. There is always some uncertainty in making the best correlation for a well tie, and there are two primary sources of this uncertainty: (1) quality of the reflection seismic data to which the well is being tied and (2) assumptions made in generating a synthetic seismogram (or in processing a VSP). Although details for creating a synthetic seismogram or processing a VSP are not included in this text, the following list includes some of the factors to consider when you find that a well-to-seismic tie does not work: • Well or seismic data (or both) are not positioned accurately. • Seismic data do not image the actual well location. • Seismic data processing is deficient (e.g., data are excessively noisy or contain residual multiples). • Time-depth relationship is incorrect. • Logs used to generate synthetic seismogram were not edited properly. • Synthetic seismogram used an incorrect or inappropriate wavelet. Each horizon that you decide to correlate, whether established by a well tie or recognized on your seismic data as a geologically significant boundary such as a major sequence boundary or an unconformity, should illustrate a particular element of the geology of your project area. You should carefully record your starting point for correlating a horizon in the same way that a field geologist identifies a type locality for a geologic formation. Often you can divide your project area into subareas separated by one of two types of barriers: geology (faults, intrusions, etc.) or control (gap in data, difference in data quality). Within these subareas may be several seismic lines that illustrate a certain geologic element, so choose one of these lines per subarea as a type line that you can use later as a display line when presenting your completed interpretation. (Such a line often is one that you can correlate across barriers between subareas with greatest confidence.) It is good practice to correlate several horizons at once during the course of an interpretation, although you should carry no more horizons than you



05/12/11 7:49 PM

118 First Steps in Seismic Interpretation Well location Synthetic seismogram

a) t

Well location Synthetic seismogram

b) t

Figure 1. (a) A 3D time-migrated line with well location identified and the synthetic seismogram generated from sonic and density logs for the well. This seismogram uses a wavelet extracted from the 3D data and is calibrated with a velocity survey acquired in the well. (b) The well tie is established by translating the synthetic seismogram to the well location on the line and then shifting it up or down to achieve the best visual match to the seismic data (courtesy BP). Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:49 PM


can comfortably and accurately handle (a matter of personal preference and interpretive style). Correlating several horizons at once facilitates interpretation of faults and generally gives you a broader view of the geology contained in your data. At the same time, you should not overly restrict your view by focusing on discrete horizons but should correlate intervals (with characteristic seismic signatures) as well as horizons. Phantom horizons that mimic the form of an adjacent (overlying or underlying) horizon are commonly used to carry interpretation through poor data areas where primary interpretation by character correlation is not possible (see Pennington et al., 2004). Such a horizon often is constructed by connecting discontinuous dip segments within poor data zones. Phantom horizons can be very useful for producing relatively coarse form maps in poor data areas when a rough idea of structural trends is needed; in fact, maps of phantom horizons usually are the only ones that can be produced in such areas. In most instances, phantom horizons should be assigned greater risk of accuracy and not be assumed to have primary stratigraphic significance. Be mindful of the advice offered by Nettleton (1940): Under favorable conditions where there are clear reflections from certain continuous beds, such as a continuous limestone under or within a thick shale section, the seismograph map may show accurately the depth to the stratum (limestone) giving a certain reflected event. Then the map may be almost as accurate as could be made by an actual contact by drilling each detector spread. More usually, the construction of the map is not so simple, and the results are more uncertain. More frequently than not, any particular reflection is not continuous over a wide area, and closer control with dip determinations or phantom horizons must be used to tie a picture together. In very difficult areas, such as portions of the Gulf Coast, there are not enough reliable reflections to make even a reasonably continuous phantom horizon. However, it is usually possible to get at least an approximate idea of the dip within one or two zones of depth at many or most of the detector spreads. The dips so indicated, of course, are the components along the line of the spread … . With fair dip control it may be possible to construct an approximate contour map to give a fairly good general picture, but it cannot be taken as representative of structure at any particular depth. Having divided a project area into subareas, your interpretation usually proceeds from the simplest (least complicated) to the most difficult subarea. Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:49 PM


In effect, you first quickly interpret the easiest areas, effectively isolating problem areas. In these problem areas, you will spend the most time and make critical interpretive decisions, following the so-called 80–20 rule: 80% of the work (the easiest) is done during the first 20% of project time, and the remaining 20% of the work (the hardest) consumes the remaining 80% of project time. Often, important control points such as wells delimit these areas, and you expend much interpretive effort to ensure that horizons can be correlated with confidence away from these points and into other areas. As shown in Figure 2, you use two basic procedures to correlate reflection seismic data: loop tying and jump correlation, as discussed in the following sections.

Loop tying Loop tying correlates horizons and faults in a geologically reasonable fashion from line to line across a grid of lines; remember that a fault is a geologic surface which you must tie consistently from line to line in the same way that you tie a horizon. Loop tying begins at a starting point (X on Figure 2) on a given line and proceeds by trace-to-trace correlation along that line to its intersection with another line, effectively spawning another starting point for correlation along the intersecting line. This procedure is repeated along lines in the grid until correlation returns to the original

Jump correlation

Area of interest

x

Loop tying

Seismic grid

Figure 2. Schematic of loop tying and jump correlation.



05/12/11 7:49 PM


s tarting point on the first line. The closed track of seismic lines along which correlations are made is called a loop. You should tie loops in a regular, disciplined manner, moving the interpretation successively from completed loops to adjacent untied loops and leaving no intervening lines uncorrelated. If possible, you should expand correlation away from the starting point(s) that serves as a fixed reference; when you discover correlation errors, you should trace them back to the starting point(s) to identify the correlations that need to be changed. With 3D data, it is good practice to begin tying loops on a relatively coarse grid and fill in as needed; the grid does not need to be uniform across an entire project area, but it should not vary greatly because large variations can affect the accuracy of subsequent operations such as autotracking, interpolation, or gridding. Whether correlating horizons or faults, you should pick on a grid that is sufficiently closely spaced for accurate definition (resolution, according to the Nyquist theorem) of all features of interest. Figure 3 is an example of a loop tie consisting of five line segments from a 3D survey. Although a loop normally is defined by four line segments as shown in Figure 2, line segment A is repeated on the far right side of this display to establish visual continuity of the data and picked horizons around

A

B

C

D

A t

Figure 3. A loop tie from a 3D survey. Correlating the shallow (yellow) horizon is straightforward, but correlating the deep (red) horizon requires a geologic explanation for correlating across a discontinuity in segment C (courtesy BP).



05/12/11 7:49 PM


the entire loop. On this display, the tracking of the yellow horizon is simple (as Dix says, “almost anyone can understand such a correlation” [1952]), but there is a correlation problem to be solved for the red horizon. Consider the starting point for this horizon to be the peak (black) shown in segment D. Correlating to the right into segment A is straightforward (it could easily be autotracked), but to the left the horizon does not track smoothly across segment C. There is a discontinuity about one-fourth of the way from right to left across segment C, at which point you must decide whether to correlate to a shallower or deeper peak before continuing the red horizon on to segment B. This discontinuity must have a reasonable explanation in terms of data processing or geology, and, given the overall appearance of the data, a geologic explanation such as a small fault or a depositional/stratigraphic difference is more likely. Continuing across segment C toward segment B, if the red horizon is correlated to the lower peak, it can be connected to the pick made from segment D to segment A (right side of the display). Alternatively, if the peak is correlated to the shallower peak in segment C and then carried on toward segment B, there will be a correlation error or mis-tie between segments B and A that is not as simply explained by geology as is making the correlation to the deeper peak in segment C. The loop tie illustrated in Figure 3 is procedurally straightforward, in the sense that you only need to know the exact positions of the intersections of the lines that form the loop; referring to Figure 2, these are the corner points of the loop. This example is for 3D data, so reflections will tie at the line intersections, that is, the reflections will directly line up with (correlate to) each other because the data have been 3D migrated. But what happens if the loop is formed from a grid of 2D migrated lines? Recall from Chapter 5 that migration can be done in two or three dimensions; more importantly, migration of any given 2D line cannot fully account for 3D effects unless that line is a true dip line. But all of the lines that form a loop of 2D lines cannot be true dip lines unless the geology is perfectly flat (in which case dip = 0 and there is no true dip direction); so in tying loops of 2D migrated lines, an interpreter will almost always encounter a mis-tie. In this case, a mis-tie is the misalignment of reflections at a line intersection caused by incomplete migration of one or both of the lines. Figures 4 and 5 are images of two of the four 2D time-migrated lines that form a loop as shown in Figure 2. The intersection of these images is marked by the vertical yellow line on each image. Note that the reflection geometry and patterns in Figure 4 (line A) show a major down-to-the-right extensional (listric) fault, with steep and variable dips in both fault blocks and well-defined left-to-right convergence of reflections in the right-hand fault block. In contrast, the reflections in Figure 5 (line B) are horizontal to



05/12/11 7:49 PM

Chapter 8: Correlation Procedures 123 Line A t

Figure 4. A 2D time-migrated dip line. The vertical yellow line marks the intersection with the line shown in Figure 5. Although the true dip direction cannot be determined from this line alone, the relatively steep dip on this line and the gentler dip on the orthogonal line in Figure 5 indicate that this line is more dip than strike oriented (courtesy WesternGeco).

relatively gently dipping at depth, with no obvious faults visible. These lines are perpendicular to each other, so line A must be nearer to true dip than line B. The lines will not tie at their intersection because at least one of them (probably line B and possibly both) is not fully and correctly migrated. Figure 6 illustrates this mis-tie, which becomes greater with increasing reflection time because of the increasing dip with depth on line A. Figure 7a illustrates that lines A and B can be tied by reflection character, but not necessarily in terms of absolute two-way reflection time, by simulating migration of line B. This is done by rotating line B updip into line A until there is good character match. Recalling Chapter 5, consider that line A was acquired and processed in the plane of Figure 1 of Chapter 5 and so is more accurately migrated, whereas line B was acquired and processed in the plane perpendicular to the figure and is less accurately migrated. Rotating line B moves reflections along the path of the wavefront drawn, as shown in Figure 1 of Chapter 5, to the point at which reflections on line A have been accurately repositioned by migrating that line. Lines A and B are tied when reflections are smoothly aligned and good character match is established.



05/12/11 7:49 PM

124 First Steps in Seismic Interpretation Line B t

Figure 5. A 2D time-migrated strike line. The vertical yellow line marks the intersection with the line shown in Figure 4. Although the true strike direction cannot be determined from this line alone, the relatively gentle dip on this line and the steeper dip on the orthogonal line in Figure 4 indicate that this line is more strike than dip oriented (courtesy WesternGeco).

Figure 7b shows that variable amounts of rotation of line B are needed to tie different reflections or reflection packages, in this case greater rotation being required to tie more steeply dipping reflections at depth. The different amounts of rotation reflect the reality of lateral and vertical velocity variations; consider that the “rigid” rotations in Figure 7 effectively are constantvelocity migrations along circular wavefronts. As a practical note, keep in mind that in addition to rotation of line B, some translation of the line may also be necessary for establishing a good character tie, based on data quality at or near the tie point or on lateral variations in seismic response caused by real changes in geology. The procedure described above achieves line ties by reflection character but does not necessarily determine the true position of the actual line tie in X-Y (horizontally) or Z (vertically in time or depth). In the simplest but not the most common case, the established tie point is correct if one of the two intersecting lines can conclusively be shown to be a true dip line; even this



05/12/11 7:49 PM

Chapter 8: Correlation Procedures 125 Line A

Line B t

Figure 6. The left portion of line A (Figure 4) juxtaposed against the right portion of line B (Figure 5). The mis-tie at the intersection of the two lines becomes more obvious with increasing reflection time and increasing dip on line A. Red arrows point to reflections that are obviously misaligned (courtesy WesternGeco).

assumes that the migration of the true dip line has been done correctly with an accurate migration-velocity field. Figures 8–10 illustrate two special cases and one general case of tying two orthogonal 2D time-migrated lines. For simplicity, these cases exclude 2D depth-migrated lines to avoid consideration of migration errors arising from differences in depth-migration velocity models; however, even for perfectly accurate velocity models, 2D depth-migrated lines can mis-tie for the same reason that 2D time-migrated lines do. The critical factor is not the domain in which the migration is done but whether the migration is twoor three-dimensional. The reflecting surface in Figures 8–10 is an inclined plane dipping from the top of the figures to the bottom; effectively, the true dip direction of this plane is parallel to the vertical axis of the figure (the top-to-bottom direction of the drawing), and the strike direction of this plane is parallel to horizontal axis of the figure (the left-to-right direction of the drawing). Structure contours are shown as horizontal dashed lines on each figure. The position of the schematic cross section shown in Figure 1 of Chapter 5 is marked as the true dip line, and the migrated line segment in



05/12/11 7:49 PM

126 First Steps in Seismic Interpretation Line A a) t

Line A b)

t

Figure 7. (a) Rotation of line B (strike line) updip into line A (dip line) simulates migration of line B in the direction of line A and enables matching of reflection character to tie the two lines. In this display, the character match is good in the shallow section but not in the deeper section (as marked by the red arrow), so line B needs more rotation into line A to tie deeper reflections. (b) Further rotation of line B (strike line) updip into line A (dip line) establishes good character match for deeper reflections (courtesy WesternGeco). Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:49 PM


that figure is the intersection of the inclined plane with the vertical plane. The recording datum of the 2D lines is a horizontal plane (the plane of the page), and the positions of the lines are plotted on this plane as heavy black lines, projected vertically downward to the inclined plane on which reflecting points are located. Figure 8 is the simple case in which one of the two 2D time-migrated lines is a true dip line. Migration of the dip line moves reflecting points on this line to their true subsurface positions (in the direction of the blue arrow on the figure), but there is no repositioning of reflecting points on the orthogonal line (the strike line). Even though the strike line is imaging updip of its surface location, reflections from the horizontal line of reflecting points on the inclined plane form a horizontal event for which there is no repositioning by migration in the vertical plane of the line. In this case, the mis-tie is at the mapped position of the intersection point of the two lines. The reflection from the inclined plane on the true dip line is correctly placed, having been migrated to the intersection point from a position downdip of the intersection; whereas the reflection from the inclined plane on the strike line, the reflecting point for which is updip of the line intersection, is incorrectly placed at the line intersection. You can see that at the point of intersection of the two lines as plotted on the horizontal recording plane, the

Surface location of 2D seismic line (true dip line)

Reflecting points on dipping plane Surface location of 2D seismic line (strike line)

Migration in vertical plane of 2D line

Updip Map view of contours on dipping plane Downdip

Figure 8. Schematic of a dipping plane and the mis-tie at the intersection of two orthogonal 2D time-migrated seismic lines, one of which is a true dip line. Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 7:49 PM


two-way reflection time to the inclined plane on the strike line is less than the two-way reflection time to the inclined plane on the true dip line, and the magnitude of this difference is the amount of the mis-tie. Thus, in a region where 2D time migration is acceptable — where there is simple or complex structure and simple velocity as described in Chapter 5 — the deeper of the two reflection times to a given horizon measured at the intersection of two 2D time-migrated lines is the more accurate (less incorrect) measure of the true two-way traveltime to that horizon. Figure 9 is more complicated but is not the general case for the inclined reflector and two orthogonal 2D time-migrated lines. In this case, each line is oriented 45° to the true dip direction, and both lines are apparent dip lines — neither can be recognized conclusively as more strike or dip than the other. Migrating each line (in the direction of the blue arrow) in its own respective plane is incomplete, yet reflecting points and the reflections from them are mispositioned on both lines by the same amount (both blue arrows terminate at the same map contour), giving the erroneous view that the lines do tie. This observation supports the previous statement that a character tie

Reflecting points on dipping plane

Reflecting points on dipping plane

Surface location of 2D seismic line


Updip Migration in vertical plane of 2D line

Map view of contours on dipping plane Downdip

True dip line

Figure 9. Schematic of a dipping plane and the mis-tie at the intersection of two orthogonal 2D time-migrated seismic lines, both of which have the same apparent dip.



05/12/11 7:49 PM


alone does not guarantee that the true subsurface location of a tie point has been determined accurately. Figure 10 is the general case for the inclined reflector and two orthogonal 2D time-migrated lines; strictly speaking, the most general case for tying intersecting 2D time-migrated lines would have an arbitrary angle of intersection between the two lines. Here, both lines are oblique to the true dip direction, and the one whose azimuth is nearer to the true dip direction is considered more the dip line and the other more the strike line. Both are incompletely migrated, the dip line less so than the strike line. A mis-tie is associated with these migrations, equal to the difference between the values of the contours at which the blue arrows terminate. Figure 10 also illustrates that the deeper of the two reflection times to a given horizon measured at the intersection of two 2D time-migrated lines is the more accurate (although still incorrect) value for the true two-way traveltime to that horizon. Although Figures 8–10 represent an almost ridiculously uncomplicated model of the subsurface, they illustrate several important concepts and useful

Reflecting points on dipping plane Reflecting points on dipping plane

Updip Migration in vertical plane of 2D line

Map view of contours on dipping plane Downdip


True dip line


Figure 10. Schematic of a dipping plane and the mis-tie at the intersection of two orthogonal 2D time-migrated seismic lines, one of which has greater apparent dip than the other.



05/12/11 7:49 PM


techniques when correlating 2D reflection seismic data. In practice, a grid of 2D seismic lines usually is laid out in an orthogonal pattern aligned with the predominant strike and dip directions. These directions are not always known at the outset of exploration in a given area; on land, they may be determined from outcrops or based on trends of surface features, and in a marine setting they may be as simple as perpendicular and parallel to the coastline. As exploration matures, the 2D program design is refined as these directions and variations become better known, although many programs are infill to existing surveys with no special attention given to dip and strike. When you correlate lines in a 2D grid, the most common procedure is to track horizons on the lines that are more nearly in the dip direction and have greater imaging accuracy and to use the strike lines to correlate horizons from dip line to dip line by character. This approach may help you visualize elements of geology that are defined implicitly in dip section, but it does not remedy the mis-tie problem that you must address in mapping. Of course, this procedure is not applicable in 3D work, although the trend and magnitude of dip in the subsurface can figure into 3D survey designs, as mentioned in Chapter 6. Because migration is a velocity-dependent process, the mis-ties you observe on 2D time- or depth-migrated data may be caused not only by geometric effects — the fundamental limitation of 2D imaging of a 3D subsurface — but also by differences between the migration velocities used to migrate individual 2D lines. These velocities almost always are determined on a line-by-line basis and do not necessarily tie at line intersections. You can easily envision a situation similar to Figure 9, in which you can tie two lines because a geometric mis-tie is effectively cancelled by differences in the migration velocities used for the lines. You can just as easily imagine an instance where two 2D depth-migrated lines imaging horizontal strata (e.g., undeformed abyssal plain sediments) do not tie because the velocity model for one of the lines is faster (its rate of increase of velocity with depth is greater) than the model for the other line. It is important to realize that not all mis-ties are caused by 2D imaging; they can also be the result of differences in acquisition and/or processing between adjacent or overlapping surveys and, as such, can be problems in both 2D and 3D projects. Mis-ties of this kind often are corrected by simple static or bulk shifting of one survey to another, which requires that one survey be established as a reference to which the other is corrected. This procedure can account for relatively small mis-ties in terms of vertical position of reflections but cannot do so for changes in reflection character caused by processing differences, which usually appear as differences in phase and/or amplitude, or for mispositioning caused by having imaged the surveys with different migration velocities.



05/12/11 7:49 PM


In addition to migration of seismic lines, a process known as map migration can help you accurately position subsurface reflecting points. This process follows correlation of 2D unmigrated seismic lines, and it uses the same geometric principle relating unmigrated and migrated surfaces as illustrated in Figure 1 of Chapter 5. When using map migration, you correlate horizons on 2D lines by loop tying and/or jump correlation and then create a contoured surface of the two-way reflection time to a given horizon. Usually such a surface is purposefully smooth, but accurate positioning of faults is problematic, owing to the lack of precision in picking faults on unmigrated data. You migrate your mapped surface by repositioning contours using an average velocity distribution for that surface and the true dip direction and dip rate measured directly from the contoured map. This process can be done relatively easily by hand using velocity information for the horizon to be migrated (from seismically derived velocities or well velocity surveys; see Chapter 4), but difficulties arise in areas with complex structure, such as steep dips, closely spaced faults, intersecting faults, or several different fault trends. In the modern interpretation environment, map migration is done with computerized algorithms and rarely by hand. Chun and Jacewitz (1984) provide equations that are easily adapted to map migration computations. The omnipresence of 3D data has nearly eliminated map migration from interpretation but not from the modeling tool kit. Loop tying applies to correlation of horizons and faults, but there is an additional concern when correlating faults on a grid of 2D data. These correlations can be problematic if the grid is widely spaced with respect to the distance between individual faults, a manifestation of spatial sampling (see Chapter 6). You can accidentally miscorrelate, or alias, fault picks if you make your correlations in map view only and have not carefully tied the faults from line to line. Figure 11a shows a map view of several fault picks marked on three parallel seismic lines, but the picks have not been correlated from line to line. In this example, the faults are normal faults and have trapezoids or tents annotated on the downthrown blocks; for simplicity, the faults are shown schematically as single fault cuts. When you first look at this display, you should ask at least three questions: 1) What is the interpreter’s confidence in these fault picks? 2) How accurate are the positions of these fault cuts? 3) Has the interpreter been able to tie the faults from line to line on any intersecting lines (not shown in Figure 11)?



05/12/11 7:49 PM


b)

c)

Normal fault (tent on downthrown block)

Figure 11. Correlation ambiguity for fault cuts in map view. (a) Fault cuts marked on three parallel seismic (dashed) lines. The faults are normal faults, with tents annotated on the downthrown blocks; they have not been correlated from line to line. (b,c) Two different correlations for the fault cuts shown in (a). The correct fault correlations can be determined by loop tying the fault surfaces using intersecting lines (not shown) or by correlating from line to line based on similarity of fault shape, position, and displacement of interpreted seismic horizons.

Obviously, the third question is the most important for making the correct fault correlations. But if the answer to that question is “no,” then correlations based only on the marked positions of the fault cuts are ambiguous. Figure 11b and 11c shows two possible correlations. The correct correlations should be made by loop tying the faults using intersecting lines; if no intersecting or crosslines are available, then the correlations should be made, albeit with less confidence, by correlation based on similarity of fault shape, position, and displacement of interpreted seismic horizons. In addition to tying the faults, you should ensure that your correlations are consistent with the senses and amounts of displacement of your correlated horizons. Important sources of information that you might use to guide or support your correlations are well data that establish fault positions and displacements based on missing or repeat section and geologic studies that



05/12/11 7:49 PM


describe the regional trends and/or structural style of your project area. In a 3D interpretation, you can check your fault correlations by displaying fault picks on horizontal reflectivity or coherence slices, such as shown in Figures 6 and 7 in Chapter 3. Note in these examples that fault trends are particularly well defined on the coherence data, again as enabled by good data quality. You can also validate your correlations of faults and horizons by viewing arbitrary lines parallel to faults, effectively within individual fault blocks, to be sure that you haven’t aliased any fault picks or missed any cross faults. Even though you will spend a lot of time at the helm of powerful workstations and will very often be totally immersed in 3D data, you should not lose sight of the real world of geology, in which reflecting boundaries are 3D curved surfaces frequently interrupted by sharp discontinuities (faults). The impedance contrasts across these surfaces vary laterally and widely in scale, and the seismic responses to these surfaces overlap and interfere. None of these complications is addressed in the discussions of Figures 8–10. Coupling your awareness of this reality with the ever-present noise in seismic data and the assumptions, approximations, and uncertainty implicit in data acquisition and processing/imaging, you should be very careful when working with 2D data and very appreciative of the immense value of 3D data.

Jump correlation In its most general sense, jump correlation involves correlating an area or segment of data having a particular or distinctive reflection character to a noncontiguous or nonoverlapping area or segment of data. This type of correlation, which is based on similarity of reflection character, position, and geologic setting, puts the greatest demand on your pattern-recognition skills because your correlation is not visually continuous. As shown in Figure 2, you use jump correlation to tie outlying lines that do not intersect the main control grid of available data; this is historically associated with interpreting grids of 2D seismic data, but on a large scale it can be thought of as correlating from any data set to another nonintersecting 2D or 3D data set (for instance, from one 3D volume to another nonoverlapping 3D volume). In the modern interpretation environment of workstations and 3D data, the most common use for jump correlation is to pick horizons across faults, in which the noncontiguous or nonoverlapping areas are separated by faults. You most often make jump correlations on a workstation using a tool called a correlation polygon, with which you outline a limited area of data in one region (e.g., in a fault block) that you will translate or jump across a discontinuity or fault to make a visually continuous correlation to reflections in another region (such as in an adjacent fault block; see Figure 12). In



05/12/11 7:49 PM


drawing a correlation polygon, it is important to include enough but not too much data in the polygon for you to correlate an interval of reflections and not just an individual event; in this way, you take full advantage of pattern- recognition skills and avoid tunnel vision in focusing too narrowly on a single reflection. In some workstation applications, an entire fault block, defined by a previously picked fault, can be translated along that fault to investigate possible correlations into the adjacent fault block. This can mitigate, but not necessarily eliminate, the tendency toward tunnel vision. One helpful check on the validity of jump correlations across faults is flattening the interpreted seismic section on a correlated horizon (in Figure 12c, you would flatten on the red horizon). You can do this by hand on a paper copy of the interpreted line by cutting it along the traces of picked faults and then sliding the block adjacent to a given fault up and down along a)

t

Figure 12.b)(a) A 2D time-processed seismic line. The objective is to correlate t the red horizon to the right into the fault block between the two picked faults (in yellow). The red box, often referred to as a correlation polygon, outlines the segment of data that will be translated and/or rotated to establish the best character correlation with data in the adjacent fault block. (b) The correlation polygon from (a) has been translated into the adjacent fault block and rotated counterclockwise to achieve good character correlation (visual alignment of reflections) at the level of the red horizon. (c) Final correlation of the red horizon based on jump correlation into the fault block between the two picked faults. Correlation along this line into other fault blocks would proceed in similar fashion by selecting and then translating/rotating other correlation polygons (courtesy BP).


c)


t

05/12/11 7:49 PM


b)

c)

t

t

Figure 12. (Continued)

the fault to find the best visual correlation. Most workstation systems now provide flattening on a user-specified horizon; some let you select a fault block defined by existing fault picks and then translate it to make your correlation — a simulation in the electronic world of the manual procedure described above. In this jump correlation or any related flattening process, you are not doing a rigorous structural restoration of the line you’re correlating; you are only using a technique to facilitate character correlation.



05/12/11 7:49 PM


Proper structural restorations require accurate seismic correlations as input and involve considerations such as decompaction and balancing to preserve length or volume. They are usually done by specialists using software applications designed for this type of work. Figure 13 illustrates the importance of viewing the full seismic section when making jump correlations because the character of a relatively large interval, vertically or laterally, often provides the best evidence for making the most likely jump correlation. Again, flattening this line on the red or the blue horizon would help validate the accuracy of the correlations. Figure 14 shows how using geologic insight can help you make the correct jump correlation when alternative correlations are possible. The red and green faults on Figure 14a are picked with good confidence (other faults that might also be picked are not shown). At first glance, the pronounced highamplitude reflection highlighted by the yellow arrow appears to be downthrown to the right across the green fault, making it a reverse fault. A more inclusive top-to-bottom view of this line reveals individual reflections and reflection packages that correlate across the green fault as shown in Figure 14b, making it a normal fault. The bulk of correlation evidence, in combination with knowledge of the regional structural setting of the area, indicates that the green fault is in fact a normal fault. The false correlation suggested by the bright reflection highlighted on Figure 14a can be readily explained and confirmed by additional correlation and mapping. The anomalously high amplitude of the bright reflection to the left of the green fault is the reflection response to a thin hydrocarbon-charged sand. To the right of the green fault this sand is either brine-filled or absent, thus having no correlative high-amplitude response. The anomalously high amplitude of the bright reflection to the right of the green fault, which appears to be downthrown from the fault block to the left, also is the reflection response to a thin, hydrocarbon-charged sand but not from the same sand as that corresponding to the bright event in the adjacent fault block. This stratigraphically older sand is hydrocarbon charged to the right of the green fault but is either brine-filled or absent to the left of the green fault. The preceding example shows that often you can resolve conflicting correlation evidence by incorporating sound geologic reasoning and knowledge of regional geologic trends in addition to taking full views of the sections being correlated, that is, not being distracted by or focusing too closely on individual reflections. Many times the question is asked whether faults or horizons should be picked first in a given interpretation project. There is no right answer to this question. In some projects, faults are very clearly imaged and can be picked first, providing a framework for subsequent picking of horizons (recall the



05/12/11 7:49 PM

Chapter 8: Correlation Procedures 137 a)

b)

c)

z

z

z

Figure 13. (a) Restricted portion of a depth-migrated seismic line on which there appears to be no clear-cut evidence for major faulting. (b) Same line as in (a) but displayed with greater vertical extent of data. Now there is very good evidence for a major fault downthrown to the right in the left-center portion of the image; other smaller faults are also visible. The lack of good evidence for this fault in the shallow part of the image is the result of fortuitous alignment of reflections masking the fault; when more data are displayed, the discontinuities that define the fault are readily apparent. (c) Partial interpretation of (b) with a major downto-the-right fault picked as shown. A correlation polygon would help establish correlations for the red and blue horizons, and either of these could subsequently be used for flattening to validate the correlations (courtesy PGS).



05/12/11 7:49 PM

138 First Steps in Seismic Interpretation a) t

b) t

Figure 14. (a) Time-migrated line with two picked faults marked in red and green. The yellow arrow points to a high-amplitude reflection that appears to be downthrown to the right across the green fault, a reverse fault. (b) Three different reflections marked in yellow above and below the bright event highlighted in (a) provide good evidence that the green fault is a normal fault and not a reverse fault (courtesy BP).



05/12/11 7:50 PM


discussion of horizon autotracking in Chapter 7). In other projects, horizons need to be picked first to guide fault picking. In all cases, the only definitive statement to be made is that all final horizon and fault correlations must be geophysically consistent and geologically reasonable, and the order of and manner in which these correlations are done according to the work flows used are matters of interpretive judgment based largely on experience. No matter how you proceed with an interpretation, you should track the progress of correlation across your study area on an index map, on which you mark tied line intersections and fault cuts (this is done automatically on a workstation). You select one of the following end-member procedures to use in correlating and mapping: 1) Interpret all lines in the area; then grid and contour the entire area at once. 2) Interpret only a few lines in a relatively small area, grid and contour that area to establish trends, and then proceed by steps in this fashion across the entire area. Whether using one of these approaches or some combination of the two (often the case), you might consider tying loops on a relatively coarse or loose grid and then adding interpretation to or infilling that grid — essentially the process used in seeding autotracking in a 3D seismic interpretation. This approach affords you a view of the geology of the area fairly early in your project, which can guide additional detailed interpretation or, if needed, provide a quick view of the project. Remember that interpretation on a coarse grid is, by itself, no excuse for not interpreting all available lines in a project. As stated before, it is good practice to mark fault cuts for each interpreted horizon to assist fault correlations, especially in structurally complex and/or poor-data-quality areas; on a workstation, you can do this automatically, usually with an option that you turn on or off. Comparing fault cuts on different horizons helps to identify miscorrelations and to ensure that the fault patterns on all interpreted horizons are reasonable and consistent with one another. During the course of an interpretation, it is very important for you to make notes directly on seismic lines and maps to mark features of interest and questionable areas; unfortunately, you cannot do this very easily on a workstation. Notes focus attention on areas that are critical elements of the interpretation or that need further work. In many instances, these notes should become part of the permanent record for your interpretation because they can have important implications for project and data management.



05/12/11 7:50 PM


Correlations in depth-migration projects Although detailed treatments of depth-migration principles and practices are not included in this text, it is important to discuss the interpretive input required in a typical depth-migration processing sequence. The following generic depth-migration processing sequence for marine 3D data used for subsalt exploration outlines the four main steps of this sequence. 1) Water-flood migration to image the seafloor. In this step, the entire data volume is migrated with the velocity of seawater. 2) Sediment-flood migration to image the top of the salt. Having picked the seafloor on the water-flood migration from step 1, the migrationvelocity model is updated by replacing the seawater velocities below the seafloor horizon with a sediment-velocity field. The data volume is migrated a second time. 3) Salt-flood migration to image the base of the salt. Having picked the top of salt on the sediment-flood migration from step 2, the sediment-velocity model is updated by replacing the sediment velocities below the top of salt horizon with salt velocity. The data volume is migrated a third time. 4) Final migration to image the subsalt section. Having picked the base of salt on the salt-flood migration from step 3, the migration-velocity model is updated by replacing the salt velocity below the base-of-salt horizon with sediment velocities. The data volume is migrated a final time. Interpretive input is required after each of these steps. Tracking the seafloor on the water-flood migration is usually straightforward and is especially important in areas with significant seafloor relief. Tracking the top of salt on the sediment-flood migration is less straightforward and can be very difficult where this surface is rugose or steeply dipping. Tracking of the base of salt on the salt-flood migration is often very difficult and is sometimes done only in model-guided fashion. The full integrated interpretation is performed on the final migration, where the preceding efforts bear fruit. In settings with multiple overlying or irregularly shaped salt bodies such that along any vertical line through the 3D volume there would be more than one top-of-salt penetration (and, necessarily, more than one base-of-salt penetration), steps 2 and 3 are repeated to image successively deeper salt bodies properly. As you can easily envision, the accuracy of correlations at each step in this migration sequence is critical to the accuracy of following steps. Said another way, inaccurate correlation of the top of salt will necessarily preclude accurate imaging of the base of salt.



05/12/11 7:50 PM


Correlating the top- and base-of-salt reflections, or the horizons that bound any bodies with anomalously high or low velocity, in a depth-migration project must be done in such a way that the bodies are completely closed spatially before proceeding with the final migration. In a workstation system that has multivalued horizon capability [for any X-Y (bin) position there can be multiple Z-values], a salt body can be completely defined with a single horizon, even though the top- and base-of-salt events are picked separately at succeeding stages in the normal processing/model-building sequence. For example, the top-of-salt horizon picked on the sediment flood could be updated to include picks for the base-of-salt horizon on the salt-flood migration. Good practice would include copying the original top-of-salt horizon for archival purposes before updating it to include base-of-salt picks. Unfortunately, for the time being, most workstation systems are single valued, meaning that for any X-Y position, there can be only one Z-value. This functionality constrains the interpretation technique that you can use to pick the top- and base-of-salt horizons to ensure that the final picks completely close a salt body. Figure 15 illustrates a technique for picking top- and base-of-salt horizons that helps you construct closed salt bodies. Figure 15a shows the outline of a salt body on which the top-of-salt horizon (Top salt 1) actually extends beyond the right-hand edge of the body. This horizon is intentionally projected as shown and must be done so manually to ensure it will intersect and overlap a similarly projected base-of-salt horizon (Base salt 1). In this example, the base-of-salt horizon and a deeper top of salt (Top salt 2) are picked in similar fashion to ensure they overlap and hence close the body. This second deeper top of salt would be named and picked as a separate horizon. When all of the top- and base-of-salt horizons for all salt bodies in a project have been picked, the overlap areas are trimmed away (using workstation functionality); then the salt bodies are completely defined as shown in Figure 15b. You should exercise care in picking the edges of salt bodies to ensure they are smooth and fully closed; even small errors in picking at these edges can have detrimental effects on the final migration. Note that accurate salt-thickness maps can be calculated easily from these carefully picked top- and base-of-salt horizons. As mentioned, if there are complex or multiple overlying salt bodies in the project area, you may have to repeat the sediment- and salt-flood migrations in your depth-migration processing sequence to accurately image these features and, ultimately, your exploration or development targets. In conjunction with processing geophysicists and project managers, you decide whether to do this based on the extent and degree of



05/12/11 7:50 PM


z

Top salt 1

Trim overlap

Base salt 1

Trim overlap

Top salt 2 b)

z

Figure 15. (a) Outline of an arbitrarily shaped salt body; two top-of-salt horizons (light blue) and one base-of-salt horizon (red) are shown. The top- and base-ofsalt horizons are intentionally picked to overlap as shown to ensure the salt body is completely closed, and the overlap areas (dashed rectangles) are trimmed away prior to final depth migration. (b) Final salt-body outline; if using a workstation system with multivalued horizon functionality, this shape would be picked as a single horizon.



05/12/11 7:50 PM


salt complexity, the imaging accuracy required for acceptable definition of your targets, and the time and funding available for additional processing and interpretation. Figure 16 is an example of a final depth migration on which multiple top- and base-of-salt horizons for a relatively shallow salt body were picked using the technique illustrated in Figure 15. In actual practice, you would pick each horizon after the migration step specifically intended to image it. Ideally, you would need two separate applications of the sediment-flood–salt-flood sequence to most accurately image the salt overhang on the right-hand side of the body. In this example, the extent of the overhang is relatively limited, and a single application of sediment flood followed by salt flood produces acceptable imaging as shown. However, this is not always the case; your decision whether to proceed with extra imaging steps will be project specific and not always simple or straightforward.

z

Figure 16. A 3D PSDM seismic line showing multiple top-of-salt (dashed blue lines) and base-of-salt (dashed red lines) horizons defining a salt overhang. These horizons are drawn with overlaps according to the technique shown in Figure 15 (courtesy PGS).



05/12/11 7:50 PM


On a depth-migration project in which you pick top and base of salt, you may occasionally observe a salt suture, a reflection that marks the boundary between two salt bodies which have merged or coalesced. The acoustic impedance contrast that gives rise to such a reflection can be a measure of the difference in rock properties between the adjoined salt bodies or between salt and sediments entrained in the suture zone. Accordingly, there can be portions of the suture zone across which there is no impedance contrast, e.g., clean salt against clean salt, so no reflection will be visible; the suture is not detected, but geologically it is still there. In these areas, you pick the suture, if you choose to do so at all, in model-guided fashion based on the shapes of the individual (unsutured) salt bodies or the trend of the suture where it is visible. Figure 17a is a schematic of the correlations of three events you would pick to define two merged salt bodies: top of salt, base of salt, and salt suture. Figure 17b shows the top- and base-of-salt events that you pick for building a migration-velocity model for depth imaging; note that the salt suture shown in Figure 17a is not needed for imaging unless you intend to assign different velocity values to the two salt bodies. Figure 17c and 17d shows how you can define the two salt bodies separately using the salt suture and appropriate portions of the top- and base-of-salt horizons that you originally picked; in particular, note that the salt suture actually is part of the top-of-salt surface for the body on the left and at the same time is part of the base-of-salt surface for the body on the right. You construct these “geologic” top- and base-of-salt horizons by merging the suture with edited copies of the top- and base-of-salt horizons picked for imaging. As you can imagine, editing and merging horizons in this way requires careful data management, to say nothing of the demands on data management caused by picking salt (especially base of salt) on multiple seismic volumes processed with different migration algorithms, each having its own distinct imaging advantage. The essential work of picking top and base of salt is unfortunately regarded by some as a bottleneck in the depth-migration processing sequence. For reasons such as poor image quality and the need for model-guided picking, these correlations can often be “so difficult as to be impossible” (Dix, 1952) and thus cannot be done quickly and easily in every instance. It may be that the places in which the correlations are the most important are where they take the most time. Follow your own standards of quality in this work within the greater context of project schedules and availability of computer resources, and always strive to maintain a stable balance between these frequently opposing concerns. Never forget item 3 in the list of characteristics of an interpreter (Herron, 2003): You can tolerate the criticism that you are never working fast enough.



05/12/11 7:50 PM


a) z

c)

Correlations

b)

Top of salt Base of salt Salt suture

d)

Figure 17. (a) Schematic of top-of-salt (blue lines), base-of-salt (red lines), and salt-suture (green line) horizons picked for two merged salt bodies. (b) Top- and base-of-salt horizons picked for two merged salt bodies, the horizons used for velocity model building and depth imaging. The salt suture is not needed for imaging unless the two salt bodies are assigned different interval velocities. (c) Top- and base-of-salt horizons used to geologically define the individual salt body on the left in (a) and (b). The horizon originally picked as the salt suture is now part of the top of this salt body. (d) Top- and base-of-salt horizons used to geologically define the individual salt body on the right in (a) and (b). The horizon originally picked as the salt suture is now part of the base of this salt body.

Visualization Visualization is seeing with the mind’s eye, and exploration for oil and gas has always required that geoscientists visualize subsurface geology. To do this in practice in the world of 2D seismic interpretation has always been difficult, and to communicate results and make recommendations to nongeoscientists has been even more difficult. You might say that the more experienced and talented interpreters have a native ability to visualize the subsurface clearly and to render their interpretations with well-drawn maps, cross sections, and block diagrams as well to fully use the power



05/12/11 7:50 PM


of technical language to communicate within and outside their discipline. The emergence of 3D seismic techniques and accompanying advances in computer capabilities has created space for you to develop your skills for interpreting seismic data and visualizing your results. There may never have been, nor should there have been, a sharp line between interpretation and visualization, and this is especially true now with the preponderance of 3D data. Whatever distinction exists between the two may continue to fade as technology moves ahead and 3D data are truly interpreted in three dimensions — time will tell. In the meantime, the need to visualize geology in seismic data and especially to use visualization to effectively communicate interpretation results will continue to be critical for exploration success. Visualization of seismic data and interpretations has two primary objectives: 1) To manipulate seismic data in order to gain perspective and increase the amount of reliable geologic information to be derived from them. 2) To construct accurate 3D representations of geology derived from seismic images. To address the first of these objectives, you can increase the vertical exaggeration of a display by reducing its trace spacing to enhance dipping reflections, angular relationships, and faults. In the world of paper sections, a display with compressed trace spacing is affectionately and appropriately called a squash plot (Figure 18), a deliverable specifically requested from a geophysical contractor. If you do not have the time or funds to generate these displays, which involves some effort in data processing, you can quickly and easily achieve the same visual effect by viewing your data from a foreshortened perspective. This involves looking at a seismic section along rather than across your line of sight, which causes the size of a feature’s dimensions on the section to appear relatively shorter, thereby enhancing angular relationships (see Figure 19). Many jokes about sleeping on the job originated in situations when an interpreter using this technique was observed laying his head down on a desk or work table to get a foreshortened view of his data. You can now take the same perspective view of your 2D or 3D data in workstation applications that allow you to rotate an image or volume on a screen. You just change the angle between your line of sight and the image by moving the image; whereas with a paper section, you fix the image and move your head to change that angle. It is similarly easy to change the trace spacing or to vertically stretch traces on a workstation display to create a



05/12/11 7:50 PM

Chapter 8: Correlation Procedures 147 Original display

Trace spacing = x

Squash plot

Trace spacing = x/4

Figure 18. A squash plot of a seismic line is generated by plotting the line on the left with a trace spacing which is one-fourth that used for the original display.

a)

b)

Figure 19. (a) Original display of the line shown in Figure 18. (b) A foreshortened display of the line shown in (a) that is created by rotating the original display in the direction indicated by the red arrow.



05/12/11 7:50 PM


b)

Figure 20. (a) A squash plot and (b) a foreshortened display of the line shown as the original display in Figures 18 and 19a.

real-time squash plot, although this is not foreshortening in a true sense (Figure 20). A helpful technique for visualizing geology across a grid of 2D seismic lines involves laying out dip lines in sequence, on a worktable or fastened to a wall, and aligning them along a common reference such as the tie point to a particular strike line. This allows you to scan the sequence of lines back and forth to check for consistency of correlations and to see how the geology changes along strike. You can also take a foreshortened view of the sequence of lines. Do not wait until the end of a project to use this technique as a final visualization or quality-control step, but apply it whenever it can help you make correlations. The movement of your view from line to line in sequence is much the same as the animation of images on a workstation, the difference being stationary images and a moving observer versus moving images and a stationary observer. The use of motion in interpretation, whether of the interpreter or the images, is fundamental to our attempts to interpret and visualize in three dimensions. In trying to represent the results of any seismic interpretation, you are faced with the singular problem of rendering 3D features on 2D displays.



05/12/11 7:50 PM


No doubt the best way to describe geology is to construct a true 3D scale model, but this is a very complicated and impractical solution, especially in view of the power and flexibility of available visualization applications. In a workstation setting, an important concern is to ensure that your interpretation and visualization software are compatible. If they are incompatible, you have to reformat your data and interpretation products so they can be loaded into a visualization application. Ideally, both are part of the same system, but this is not always the case, depending on your company’s information technology (IT) architecture. Maps, the most common products of an interpretation, are 2D renderings of 3D surfaces that you create routinely by hand for projects involving paper sections and by computer gridding and contouring applications for workstation-based projects. Other 2D products are fence and block diagrams, which require good graphics skills to depict 3D geology accurately, especially to represent perspective (e.g., foreshortening) properly. Manually drawing these illustrations is a challenging task; many workstation systems have functionality specifically designed to create illustrations as part of your interpretation, and in many instances you can capture electronic images from your workstation screen. But without exception, you must ensure that all 2D representations of your interpretation, no matter how you make them, are accurate and comprehensible to those viewing them.

Interpretation processes and work flows In a general sense, a work flow is an outline of the steps in a procedure for doing something. Because there is a certain amount of granularity in any outline, that is, what one person considers an individual step may to another person be the consolidation or merging of several steps, it follows that the details of work flows for a given interpretation project probably vary from one interpreter to another. Table 1 summarizes individual interpretation steps or processes, and Table 2 is an example of a generic interpretation work flow. Interpretation work flows such as Table 2 should address all of the interpretation processes to be undertaken in a given project and thus must be designed to meet the specific objectives and requirements of that project. Given the wide variety of interpretation projects in our business, ranging from exploration through appraisal and development to production, it follows that there is no single



05/12/11 7:50 PM


work flow that can or should be applied routinely to every project. Rather, you should build your own library of work flows based on experience gained from different projects over the course of your career and from sharing experience and exchanging effective practices with other interpreters. As a seismic interpreter, you will devote much of your time to performing the nuts-and-bolts correlation procedures described in this chapter. The summary of interpretation processes and the generic interpretation work flow presented in Tables 1 and 2 should be particularly helpful as you gain experience in the art of correlating seismic data, and you are encouraged to develop your own work flows and interpretive style. In seismic interpretation as in the pursuit of any craft, you bring individual talents and insights to your work and should naturally strive to hone your skills and synthesize best practices throughout your career.

Table 1. Summary of interpretation processes. Ensure you have all of your data (includes data review) Establish well ties/select starting points for correlation Correlate horizons and faults Measure times/depths on sections* Post values on (grid) map* Contour map** Perform quality control (QC) Verify transfer of information from sections to map* Ensure map is geologically reasonable *Done automatically on a workstation **May be done automatically on a workstation



05/12/11 7:50 PM


Table 2. Generic interpretation work flow. Establish objectives of interpretation Make sure you have all data (and verify positioning of data) Review acquisition and processing histories of data if available Review data (use animation/visualization applications) Make initial assessment of data quality Determine structural style Estimate strike/dip directions Estimate phase and polarity of data if possible If multiple data sets, establish reference data set Establish starting points for correlation Make well ties Identify framework horizons and sequence boundaries If multiple data sets, confirm or reselect reference data set Correlate horizons and faults Set up nomenclature system for data management Determine size of correlation grid(s) Determine need for manual versus autotracking Determine need for seed (automatic) tracking versus interpolation of grid(s) Review quality (QC) of all results



05/12/11 7:50 PM



05/12/11 7:50 PM

Chapter 9 Data Quality and Management The importance of data quality and management continues to increase as we address ever more demanding interpretation challenges and move irreversibly farther into the workstation environment. In the glare of technological developments enabled by massive growth in computing power, we must not lose sight of the critical needs to assess data quality as an essential element of every interpretation project and to manage the burgeoning volumes of data and interpretation products that threaten to overwhelm us. More than you may realize, these two concerns can affect the quality, timeliness, and ultimately the business value of your interpretation.

Data quality A discussion of seismic data quality necessarily begins by defining exactly what is meant by “quality.” In its most general sense, quality is the degree to which something fulfills its intended purpose. All measures of seismic data quality are inherently subjective, so it is important to know why a particular data set was acquired and processed the way it was so as to set the proper context for assessing its quality. For example, you wouldn’t grade data from a conventional 3D survey purposely acquired and processed for deep exploration as poor quality because they aren’t suitable for evaluationing shallow drilling hazards. Similarly, you wouldn’t consider data from a 2D high-resolution shallow hazards survey as poor quality because they are useless for deep exploration (compare Figures 1 and 2). Given the purpose for a data set, you evaluate quality based on specific characteristics according to the degree to which the data set suits its purpose. Assessing seismic data quality is one of the most important aspects of your job as a seismic interpreter. It is an expectation that you satisfy and a requirement that you meet in every interpretation project. Your ability to 153 Downloaded 17 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/


05/12/11 2:16 PM


t

1.5 s

2.5 s

Figure 1. Line from a conventional 3D time-migrated data set that was purposely acquired and processed for use in deep exploration (below 4.0 s). Compare this line to the 2D high-resolution line in Figure 2 (courtesy WesternGeco).

describe and effectively communicate your evaluation of data quality develops over time as you gain experience, expand your knowledge of seismic data acquisition and processing, and broaden your exposure to different elements of geology in a wide range of settings and environments. At times, assessing data quality can seem to be no more than a beauty contest. In a comparison of several lines from the same area or of the same line with different processing sequences applied, the one that is most pleasing to the eye or looks most geologic or has the highest signal-to-noise ratio (S/N — perhaps the most common quantitative measure of data quality) or is visually dominated by smooth and coherent reflections and sharply defined faults is judged to have the highest quality. Although this view may often be correct, it should be taken only after you have reviewed the acquisition and processing history of the data and placed the seismic image into the context of the geologic setting in which the data reside. After all, beauty can be only “skin deep,” and data processing has been known to turn the occasional sow’s ear into a silk purse. Your view of data quality takes shape and



05/12/11 2:16 PM

Chapter 9: Data Quality and Management 155

t

1.5 s

2.5 s

Figure 2. Line from a 2D high-resolution (time-migrated) shallow hazards survey that was purposely acquired and processed to optimize resolution of the shallow section. Compare this line to the conventional 3D time-migrated line shown in Figure 1; both lines are located along the same surface track (courtesy BP).

evolves throughout a project as you look at all of your data. So take care not to assess and communicate data quality until you have had the opportunity to view all of the available data. Similar to the perils of the one-line interpretation, assessing data quality based on the view of only one or a small number of lines from a larger grid or survey can be seriously misleading. The three primary elements of seismic data quality are detection, resolution, and image fidelity (Schoenberger, 2004, personal communication). All of the time and effort expended in acquiring and processing seismic data in one way or another addresses one or more of these elements, with the intent of providing you, the interpreter, with as near to an ideal representation of the true reflection seismic response to subsurface geology as possible (refer to the “ideal” seismic response in Figure 1 of Chapter 1). Detection is determination by the seismic method that a feature is present, or gives rise to a seismic reflection that is recognizable as signal above the level of ambient or background noise. The power of a data set to detect subsurface features of interest is often described in terms of its S/N. Resolution is the ability to resolve by the seismic method two features that are close together



05/12/11 2:16 PM


(refer to Chapter 6). The so-called resolvable limit, then, for discrete seismic reflectors is the minimum separation for which you can tell that more than one interface is involved in the observed seismic response (Sheriff, 2002). Finally, image fidelity is the degree to which the processed seismic response accurately depicts the true subsurface positions and magnitudes of seismic reflections. It consists of correct focusing and accurate positioning of reflections. Usually you assess overall data quality using a range of subjective terms such as “excellent,” “good,” “fair,” “poor,” and “unusable” (avoid colloquialisms). You should be mindful of and prepared to discuss and illustrate which of the three elements of data quality dominates in your assessment. It is good practice in any interpretation project to map areal variations in data quality, remembering that quality very often will vary with time/depth as well. In other words, a single data-quality map for a particular project will not necessarily represent variations in quality and, ultimately, in confidence of picking from horizon to horizon. A common product of mapping seismic data quality is shown in Figure 3, in which a green-yellow-red or “traffic light” color sequence is used to represent quality grades of good-fair-poor. As with any geologic or geophysical map, it is essential to include a legend on a data quality map to ensure that color and grade conventions are communicated clearly. You can construct maps such as Figure 3 by hand or in a workstation by picking a horizon that reflects different grades of data quality. The workstation

Good Fair Poor

Figure 3. A “traffic light” data-quality map for an interpretation project area. The legend shows how the colors correspond to the different grades of data quality.



05/12/11 2:16 PM


approach involves manually tracking a horizon composed of quality picks whose values or ranges of values correspond to different grades of data quality (see Figure 4); the horizon could represent overall data quality or the quality or confidence associated with picking specific horizons of interest. This approach requires additional picking effort and concentration on continuous assessment of data quality — well worth the time in terms of value added to overall interpretation results. Data-quality maps can be used as stand-alone maps or as overlays to other maps created during an interpretation project. Do not consider any interpretation to be complete until it includes an assessment of data quality in the form of a finished data-quality map(s) and accompanying descriptive text in a final project report. Your conclusions about data quality are a critical element in characterizing uncertainty and risk for your interpretation as well as providing a basis for recommendations to acquire or reprocess data.

Quality of blue pick

Color table

Time (depth)

Map view

Figure 4. Constructing a data-quality map from quality picks associated with an interpreted horizon. The quality picks are made as a single horizon whose times (depths) or time (depth) ranges uniquely correspond to specific quality grades. The color table for the resulting map is designed to portray the different grades in a traffic light scheme according to the time (depth) of the quality picks.



05/12/11 2:16 PM


As mentioned, grading the quality of a seismic data set is undoubtedly subjective and depends heavily on your experience and your knowledge of seismic acquisition and processing. At the same time and in the same way that you improve correlation skills by looking at as much data as possible, you also develop the ability to evaluate data quality and effectively communicate the results of a quality evaluation. Simply stated, it is impossible for anyone who actively seeks to enhance interpretive skills and judgment to look at too much data.

Data management Data management involves procedures and practices to help you keep track of the large quantities of data and interpretation products common in the oil and gas business. In the modern interpretation setting, data management is primarily an electronic concern. In the preworkstation days, data management involved activities such as printing and storing maps and paper sections; losing, misplacing, or physically damaging maps and paper sections; sharing maps and paper sections with other interpreters; and nuisances such as running out of a particular shade of colored pencil or other critical resource. In retrospect, there were recognizable boundaries in those days — sets of tangible constraints that limited the amount of data and information a company or individual could amass. In contrast is the modern workstation environment, in which there appear to be few boundaries, especially in view of the increasing power and efficiency of computers and the seemingly unlimited amount of available electronic storage space. The quantity of data volumes and derived interpretation products continues to expand unchecked. It is more important than ever for you to manage data, that is, to bear responsibility and accountability for what to keep, where to keep it, how to find it later, and, perhaps most importantly, what to throw away. The increasing number of depth-migration projects has placed special demands on the need for effective data management. These projects involve multiple stages of interpretation on several different migration volumes (sediment floods, salt floods, etc.), and in the most complex settings there may be migrations with different algorithms in a particular stage of the project. For example, as illustrated in Figure 10 of Chapter 5, there are a number of different algorithms, each with its own technical and cost benefits, that you might use to address a difficult base-of-salt imaging problem in a subsalt interpretation project. In such a case, your final base-of-salt surface could be an amalgamation of horizons picked on several different data volumes into a single surface you would then use for depth migration. If you



05/12/11 2:16 PM


would ever need to refine or update this surface, you would have to know which part of it had been picked on which version of migration, which requires that you properly manage your horizons and migration volumes. In such a project, it will not always be clear which migrated volume affords the most accurate base-of-salt image, so your experience often will dictate which image is most nearly correct (or least incorrect) or most geologically reasonable in the sense that it is analogous to known geology. Recommended practices for good data management are independent of specific workstation systems. Most commercial workstation systems include basic data-management functionality that varies only in specifics from system to system. The most important practice might well be selfrestraint in creating and maintaining only those data volumes and products that contribute measurably to a project and that serve as parent entities from which you can recreate other volumes and products as necessary; this, of course, depends on the frequency with which the child entities are used and the amount of time it takes to recreate them. The practice also involves communication with other interpreters on the same project to avoid needless duplication of data and products.*

Nomenclature systems Effective management of an interpretation project requires that, from the outset, you establish and maintain nomenclature systems for data sets and interpretation products. These systems must be intuitive, rigorous in the sense that they must contain core information without fail, and flexible or expandable to accommodate information that is unique to the type of data or product being managed. A well-designed system standardizes nomenclature for any type of data or product that can be categorized: trace data files, arbitrary lines extracted from 3D volumes, horizons, faults, mapping files, wells, and culture files (such as lease outlines and political boundaries). Within a system, you should subdivide any particular category in whatever way facilitates information storage or retrieval. It goes without saying that you should thoroughly document all nomenclature conventions and link this documentation to its project, even if there are corporate standards for nomenclature, in the same way and for the same reason that every map has a scale bar, north arrow (or some azimuth reference), and legend. *See Herron (2001) for a more detailed discussion of seismic data management, with particular reference to the pitfalls involved in failure to manage multiple data sets effectively and the inability to conduct an integrated interpretation based on multiple data sets.



05/12/11 2:16 PM


A hypothetical nomenclature system for interpreted horizons in a workstation project might read as follows: Each horizon name must have a minimum number of core elements, each delineated by an underscore (_). These elements must appear in the following order in each unique horizon name: • Project identifier • Numerical designation for the approximate geologic age of the horizon (this number increases with the age of the interpreted horizon and enables the horizon list to be displayed in order of horizon age) • Color assigned to the horizon • Biostratigraphic age of the horizon • Name of the trace data file on which the horizon was interpreted (for individual horizons only, does not apply to merged horizons) • Initials of the interpreter who picked the horizon Following this convention, the components for a horizon named X2005_0600_green_A_pl20_ewfa04_DH would be defined as in Table 1. In our hypothetical system, a process such as interpolation, filtering, or attribute extraction applied to an original (parent) horizon would be specified in the horizon name after the designation of the biostratigraphic age or the trace data file but before the interpreter’s initials. In our example, the parent horizon smoothed with a 5 × 5 equally weighted spatial filter would be X2005_0600_green_A_pl20_ewfa04_sm5x5eq_DH. Although it is common practice to name trace data files carefully in 2D and 3D workstation projects, many interpreters do not recognize the Table 1. Components in a hypothetical nomenclature system for an interpreted horizon named X2005_0600_green_A_pl20_ewfa04_DH. Term

Definition or meaning

X2005

Exploration project that began in the year 2005

0600

Numerical designation for the horizon

green

Color assigned to the interpreted horizon

A

Named lithologic unit (in this case, the designated “A” sand)

p120

Biostratigraphic age of the horizon (in this case, lower Pliocene)

ewfa04

Name of trace data file on which the horizon was interpreted (in this case, the final gained version of a PSTM data volume with a sample rate of 4 ms)

DH

Interpreter’s initials



05/12/11 2:16 PM


importance of using consistent and intuitive nomenclature for arbitrary lines extracted from 3D volumes. You should assign names to these lines in such a way that another interpreter can easily grasp the purpose for which the line was created. For example, if an arbitrary line is a well-tie line, then somewhere in the name for that line should be the name(s) of the well(s) to which the line is tied. At the same time, you should specify the orientation of an arbitrary line in X-Y-space whenever possible. You can easily do this by naming the line according to its compass direction. For instance, if an arbitrary line is extracted in the northeast to southwest direction, then its name should include the characters N45E or S45W to indicate its azimuth. Compass directions are preferred over the standard abbreviations NE or SW because often the line direction cannot be described by these abbreviations, e.g., N60E or S30W. The operating philosophy in nomenclature schemes is to establish and maintain order and consistency in your data management — to eliminate obfuscation, if you will. The importance of systems such as this cannot be overstated. The story is told of an interpreter who arrived at a new posting and asked which of the many interpretations in his new project was the best or latest or least incorrect. No one could answer this question, and the previous interpreter had left no decipherable records (a problem in itself). So the interpreter did what almost everyone else would do: Judging that it would take longer to evaluate all of the existing interpretations than to do his own, he proceeded with his own interpretation. The unfortunate result was to further entangle the project with more picked horizons and faults, many of which were probably duplications of existing work. This was not lack of data management; rather, it was data mismanagement. Eventually, data mismanagement has a negative business impact that may not become evident until many years and interpreters later. The excuses that “It doesn’t matter because some day this project will have no value” or “Don’t worry, we can get by” are unacceptable because poor data management is habit forming, and its effects can be latent, harmful, and unpredictable.



05/12/11 2:16 PM



05/12/11 2:16 PM

Chapter 10 Other Considerations In preceding chapters, we considered interpretation from the complementary standpoints of seismic fundamentals and data quality, leading naturally to discussions of correlation concepts and procedures. This chapter addresses additional topics and issues you probably will encounter in the normal course of work as a seismic interpreter. Ranging from technical fields such as 4D seismic and seismic modeling to philosophical concepts such as interpretive judgment and the interpretation paradox, the following sections are intended to raise your awareness of these topics as building blocks in the foundation of your career. Also included in this chapter are several sections on professional development, the substance of which is based solely on the experience of the author.

Gridding and contouring In the modern workstation environment, gridding and contouring interpreted 2D or 3D seismic data are accomplished routinely and relatively quickly as automated processes. This is in sharp contrast to the previous generation of interpretation of exclusively 2D seismic data on paper sections, in which gridding, the measurement of two-way reflection times to picked horizons at specific user-defined points, and contouring consumed a considerably greater fraction of an interpreter’s time and were dramatically more subjective and dependent on the interpreter’s individual skills and experience. Gridding of interpreted seismic data, even 3D data which are inherently “gridded” by the nature of their acquisition and processing, is now done primarily for two reasons: to facilitate automated contouring and to manipulate picked horizons (e.g., calculate the thickness of the interval between two horizons).



05/12/11 2:16 PM


Gridding programs vary in their degrees of simplicity and user friendliness, but all have the common functionality of allowing you to specify the spatial frequency or grid spacing with which to sample your data in preparation for contouring or horizon-based computations (bearing in mind, of course, the fundamentals of spatial sampling according to the Nyquist theorem as discussed in Chapter 6). Most often you use these programs in iterative fashion, testing different combinations of gridding and contouring parameters and then visually assessing output in the form of contoured maps until achieving acceptable results. It is also important to ensure that the input grids for any calculation involving gridded horizons were generated using common grid parameters. In addition to processing speed, gridding programs have the overwhelming advantage of precision; the same input data and gridding parameters always yield the same results. In the historical interpretation setting of 2D paper sections, so-called gridding parameters corresponded to selection of the interval between points along a line at which you measured reflection times. You defined this interval based on the dimensions of your 2D seismic grid and the dominant wavelength of features of interest as well as on data quality. As a natural consequence of machine-based picking of horizons and faults, the speed and precision with which you now measure or simply record reflection times or depths are vastly improved in the workstation environment—in comparison to the historical interpretation setting, where you had to measure times manually using an analog device such as a ruler or timing strip, for which your visual acuity and/or patience and attention span often factored into your precision. In the historical setting, contouring a seismic horizon was very time consuming and involved considerable artistic input and skill on the part of the interpreter. It was the part of the interpretation work flow into which you blended style, based on experience, as well as sound knowledge and understanding of geologic principles. To this day, the inability to easily (if at all) introduce subjective elements of geology, i.e., trends or styles of contouring, into an interpretation is one of the most serious limitations of machine-based gridding and contouring. The freedom to incorporate style through contouring can be critical to an interpretation in the sense that contouring should be appropriate for the geologic setting of interpretation, effectively using the “empty space” between data points while honoring those points within their associated uncertainties. Tucker’s (1988) classic paper, which is very good reading even in our workstation-dominated time, refers to seismic contouring as a “unique skill,” the practice of which is becoming less common in the interpretation community and is often nostalgically referred to as a “lost art.” His Figures 4–6 illustrate how the same input data can be contoured three different ways, each conveying a distinct message about the geology they represent.



05/12/11 2:16 PM

Chapter 10: Other Considerations 165

In spite of the power and flexibility to grid and contour data on a workstation, it is still good advice for you to know how to manually time and record reflection times and to contour a seismic horizon map because this experience affords perspective on the strengths and weaknesses of automated processes and enables you to conduct an effective interpretation outside of the workstation environment, if needed. Remember that gridding and contouring, whether manual or machine-based, are essential steps in every interpretation work flow. But also remember that the greatest value in any interpretation lies in the accuracy of your correlations and the geologic thinking incorporated into your work. Be wary of slick maps and presentations, which are easy to create using automated gridding and contouring but can mask or smooth through errors and are not necessarily based on accurate, thorough, and consistent correlations.

4D seismic In 4D seismic techniques, the fourth dimension, time, is added to the three spatial dimensions of conventional 3D seismic data. A more accurate description of the fundamental concept of 4D seismic is time-lapse seismic, referring to seismic data that are reacquired or repeated at a later time at the same location as an initial line or survey. In a rigorous sense, you can say that reshooting a 2D seismic line at a later time produces a 3D line, but you must be careful to specify what the three dimensions are. The purpose of time-lapse seismic is to accurately measure true changes in subsurface conditions, such as reservoir pressure and pore-fluid content, which can be detected by the reflection seismic method. Successful timelapse surveying requires that the repeat data be acquired and processed in a manner as similar as possible to the initial data to ensure that the time-lapse signal is not confused with or obscured by differences in acquisition and/or processing. This process requires exacting control over acquisition (source and receiver equipment, recording fidelity, and positioning accuracy) and processing (sequence and algorithms, very close QC) of the original and the time-lapse data. The initial survey in a time-lapse project is called the baseline survey because it is the data set to which all subsequent work is referenced. The repeat survey, of which there can be several, is called the monitor survey because it is used to measure or monitor changes in the subsurface that have occurred since the baseline survey was acquired. These changes can be seen by careful comparison of the position and character of reflections and associated seismic attributes between the baseline and monitor surveys.



05/12/11 2:16 PM


Interpretation of 4D surveys and analysis of differences between baseline and monitor data are customarily done in one of two ways. The first way is to interpret the baseline and monitor surveys separately, followed by differencing interpreted horizons. This approach allows you to study time shifts and attribute differences between baseline and monitor data. The time shifts are manifestations of geomechanical changes, such as stretching overburden and compaction of reservoirs, and can be up to 8–10 ms (e.g., from producing Pliocene-Pleistocene turbidite reservoirs at subsea depths of 10,000–15,000 ft [3048–4572 m] in the Gulf of Mexico; see Ebaid et al. [2008]). The amplitude/attribute differences are the result of impedance changes related to variations in pore pressure as well as fluid type and saturation within producing reservoirs. The amplitude/attribute changes can be quantified and analyzed with stacked or prestack seismic data in the reflectivity or acoustic/elastic-impedance domains. The second way is to interpret a difference survey, created by subtracting the baseline survey from the monitor survey, done after any static differences or time shifts between the surveys have been purposefully removed in data processing. This difference survey necessarily focuses on amplitude/ attribute changes and significantly reduces the amount of time devoted to interpretation because only one survey, rather than two, is interpreted. However, the method places an additional requirement on the fidelity of data processing for accurate statics corrections. A difference survey clearly is not useful for studying 4D time-shift effects. The choice of approach is up to the interpreter, obviously in coordination with processing geophysicists. In creating a difference survey, differencing by subtracting baseline data from monitor data, rather than subtracting monitor data from baseline data, is by convention only. In the simplest terms of normal incidence or stacked-trace response, the measurable change in acoustic impedance (AI) caused by reservoir production is positive; the AI increases as the reservoir “hardens” when reservoir pressure decreases or water saturation increases. The change in AI is negative, that is, it decreases as the reservoir “softens” when the reservoir pressure increases, as in response to water injection or when free gas enters the reservoir pore fluid by exsolution (referred to as gas breakout). Note that an increase in reservoir pressure, which by itself softens a reservoir, can harden that same reservoir by dissolving gas in the pore fluid. So there can be opposite impedance effects to resolve when analyzing the impedance changes in a reservoir undergoing artificial pressure support. Analysis of impedance changes can be done in the prestack domain, a topic beyond the scope of this text. In any case, quantitative interpretation of time-lapse seismic should proceed only with careful calibration to



05/12/11 2:16 PM


well control from the producing field over which the time-lapse data were acquired.*

Seismic modeling Although details of seismic modeling are beyond the scope of this text, several comments of a philosophical nature are in order. Modeling is a topic for which approximations play a critical role. Models should be built with a complexity neither greater than nor less than that required by the detail of investigation at hand. (As noted by Albert Einstein, “The model should be as simple as possible, but no simpler” [Hawking, 2010].) For example, there is no point expending the time and energy to construct a 3D model to study a structural problem that could be adequately addressed with a 2D model. In other words, a 2D approximation can be quite satisfactory for some problems. Similarly, one should not expect a 2D modeling study to shed meaningful light on a complex 3D problem. You are responsible for determining when and how approximations are to be applied in modeling studies. Forward seismic modeling is done to investigate the expected seismic response from a given subsurface model. A forward model can help guide or evaluate the acquisition and/or processing of actual seismic data from a setting that the model ideally represents, or it can be used to help decide among alternative interpretations by comparing the results of modeling to those possible interpretations. Forward modeling is the opposite of inverse seismic modeling, in which actual seismic data are used to determine a model that could have given rise to the observed seismic response. As opposed to forward modeling, inverse modeling is nonunique. The validity and utility of any modeling results depend critically on the model input; remember, garbage in, garbage out. As mentioned in Chapter 1, geoscientists should not allow their work to be driven by models; to do so is to put the objectivity essential to all interpretations seriously at risk. In the words of Stephen Jay Gould (1995), “Models in science are judged as useful or detrimental, not as true or false.”

Interpretive judgment In any field of conscious endeavor, increasing opportunities to exercise judgment come with experience, and correlating seismic data is no exception. Experience is having been put to the test, which involves having A very good reference for 4D seismic principles and applications is Calvert’s Insights and Methods for 4D Reservoir Monitoring and Characterization (2005).

*



05/12/11 2:16 PM


learned by doing, from both success and failure (but probably more lasting and meaningfully from failure; see Petroski [1992, 2008]), and having developed confidence in one’s abilities and acceptance of the fallibility of individuals and groups. You will regularly be called upon to use experiencebased judgment in choosing approaches to solve interpretation problems and decide about seismic correlations — not the correlations that “almost anyone can understand,” but those that are “so difficult as to be impossible” yet must be attempted. On any given day, you might make hundreds of correlation decisions, some much more challenging than others, giving few or none of them a second thought. But as you go about your work, you contribute, often subconsciously, small bits of knowledge from each of your projects to your ever-broadening and deepening bank of experience, building an interconnected network of decisions and results that is stimulated and triggered into action by new problems. This network enables and fosters creative thinking and serves as a source for synthesizing innovative approaches to problem-solving, without which neither the science nor the art of seismic interpretation can advance and flourish. Figure 1 is an example of using judgment to solve a correlation problem. Figure 1a is an image taken from a 3D isotropic PSDM volume on which a shallow salt sheet is clearly seen. In spite of the good to very good quality of this line, there is at least one major correlation problem: how to correlate events in the extrasalt section on the right through a disrupted zone below the edge of the salt sheet into the subsalt section on the left. Immediately, you should wonder if the disruptions or discontinuities are faults or possibly depth-migration artifacts. Is there more to the appearance and location of these features than mere coincidence, and what in your store of geologic and geophysical knowledge and experience will help you decide? Observing the steepness of the edge of the salt sheet and the associated overlying seafloor scarp, both of which necessitate depth migration for accurate imaging, you reason that any errors in picking the seafloor, top of salt or base of salt, and/ or assignment of seawater, sediment, and salt velocities might have caused imaging errors that are manifested as the disruptions. But there is neither time nor money available to revise the depth-migration velocity model and remigrate the data; you must proceed with an interpretation with the data in hand. So you make a jump correlation into the subsalt section and try to correlate the disruptions as faults through the 3D volume. In doing this, you find that the “faults” form a pattern that can be interpreted as geologically reasonable or inferred to be spurious because it coincides closely with the trend of the salt edge and does not fit with known fault trends elsewhere in the study area. If, for the sake of argument, you conclude that the disruptions



05/12/11 2:16 PM

Chapter 10: Other Considerations 169 a) z

b) z

Figure 1. (a) A 3D PSDM line on which reflection continuity is disrupted below the steeply dipping edge of a salt sheet. (b) The dashed blue line illustrates simplified correlation of an arbitrarily chosen reflection from extrasalt to subsalt through the disrupted zone below the edge of the salt sheet. This correlation is based on your judgment that the disrupted zone is a manifestation of inaccurate depth imaging and that the true reflection configuration is best represented by the smooth correlation as shown (courtesy PGS).



05/12/11 2:16 PM


are indeed migration artifacts, then you must think about how to carry your correlations through the disrupted zone and how to map them. You know that there is greater correlation uncertainty in the disrupted zone than elsewhere on the line, even if you are confident in your jump correlation into the subsalt section. You decide to make a smooth correlation through the zone as shown in Figure 1b, realizing that you might be missing some details of real structure that are obscured by what you have interpreted as imaging artifacts. At the same time, you consider the possibility that there are other, more subtle problems in the image, such as the sharp flexure in the extrasalt section immediately to the right of the edge of the salt sheet. You will have to address this when you create a final true depth map. You are aware that you have not conclusively solved this correlation problem; rather, you have developed a solution that is geophysically consistent and geologically reasonable in view of the available data and in light of your experience. You are fully cognizant that more data, perhaps a remigrated depth volume, some well control, or more mapping of regional fault trends, might cause you to revise your correlations and maps.

Curiosity and interpretive thinking The very first item on Herron’s (2003) neither exhaustive nor exhausting list of characteristics of an interpreter is that an interpreter is naturally curious about the earth. This is much the same as Dix’s (1952) exhortation for geophysicists to look at seismic records “with the assumption that every event has a significance that we can find and that is worth finding.” This persistent inquisitiveness translates into a manner of thinking for the practicing interpreter — not a bias or rigidity in approaching problems, but a continual querying of data in seeking to understand the full meaning of observed seismic responses. Following is an example of one line of inquiry, not the “right” one or the only one, that you might follow in trying to explain what you see on the image in Figure 2. You recognize immediately that you must find out how these data were processed in order to establish context for what you see. In other words, are you looking at time- or depth-processed data, and is this a final migration or the output from an intermediate stage in the processing sequence? The most obvious feature on the line is a relatively flat-lying, lozenge-shaped body in the center of the image that has three candidate reflections (labeled 1, 2, and 3) defining its base. There are essentially no continuous reflections below the deepest of the three possible base-of-body events, and the overall strength of reflections below the body is very low. You ask a few questions about regional geology and learn that there is active salt tectonism where the data were acquired. From this, you infer that the lozenge-shaped feature is



05/12/11 2:16 PM


z

1 2 3

Figure 2. An example line extracted from a 3D survey for which you might be asked to make and then explain your observations. For the sake of argument, you begin with no knowledge of the processing that has been done to the data (courtesy WesternGeco).

a slice through a salt body of unknown shape. You know that time-domain imaging is likely to fail in this area owing to the large lateral velocity contrast between the relatively shallow salt and surrounding sediments, so you suspect that you might be looking at a depth-processed image, perhaps the output of a sediment-flood migration, because the top of salt is well imaged but three possible base-of-salt reflections are visible. Only one of these three reflections, all of which originate at the left-hand terminus of the body, can be the primary base-of-salt reflection. Processing geophysicists confirm that you are indeed looking at a line from a 3D sediment-flood (depth) migration. Although the upcoming saltflood migration will reveal the position of the true base of salt, you are nevertheless curious about the origins of the three events you see: Which is the primary base of salt reflection, and what are the other two? You reason that the shallowest of the three possible base-of-salt reflections, event 1, is probably the true base of salt because any nonprimary reflection would have a greater traveltime than the primary reflection and so would occur deeper on the depth-migrated image. You realize that if you have a well through this salt body very near to or on this line, you could use the sediment-velocity model



05/12/11 2:17 PM


and a reasonable value for salt velocity to estimate the depth to the base of the salt on this image. But you check your database and find no such well. You proceed, assuming that the deeper candidate reflections, events 2 and 3, are not primary reflections unless one or both of them is an intrasalt reflection. You propose that either or both them are nonprimary reflections, perhaps multiples. But multiples of what? If they are multiples, then they were not removed by demultiple processing, but you don’t know what, if any, demultiple processing has been done. You can’t predict positions of multiples as you might if you were working with time-processed data (see Chapter 7), so you must estimate these positions in depths for traveltimes through the sediment-velocity model along possible nonprimary travel paths. You suspect that the polarity of these events might help determine which is which because the polarity of a seafloor/top-of-salt or a baseof-salt/top-of-salt/base-of-salt peg-leg multiple would be opposite that of the primary top-of-salt event. The character of the seafloor and top-of-salt reflections suggests that the data are not exactly zero phase, but their character is distinctive enough that you can distinguish events of opposite polarity. You eliminate peg-leg multiples involving the top or base of salt and an extra reflection from the seafloor or any suprasalt reflecting surface. Such a multiple would roughly mimic the top- or base-of-salt event rather than fan downward from the edge of the salt body as you have observed. A peg-leg multiple involving the water column would occur much deeper on the image than the observed base-of-body reflections do. Your initial hypothesis is that event 3 might be an intrasalt multiple reflection: a base-of-salt/top-of-salt/base-of-salt peg-leg multiple. Figure 2 shows that the polarity of this reflection is opposite that of the top of salt reflection, as it should be. Proceeding cautiously from this tentative identification of event 3 as a peg-leg multiple, you now try to explain the nature of event 2, which has the poorest continuity and weakest amplitude of the three events. You reason that this reflection must involve a travel path within the salt body because its origin coincides with the edge of the salt body like the proposed peg-leg multiple below it. You formulate three possible explanations for event 2: 1) It is a primary reflection from the base of a layer of unknown lithology, the top of which is apparently conformable with the true base of the salt marked by event 1. This layer thickens away from the edge of the salt body, at least in the plane of this section, and you don’t have a simple geologic explanation for this configuration. 2) It is the primary base-of-salt reflection, and event 1 is a primary reflection from an intrasalt interface (you could make a similar argument for



05/12/11 2:17 PM


event 3 being the primary base-of-salt reflection and events 1 and 2 being primary reflections from intrasalt interfaces). 3) It is a base-of-salt reflection, but one leg of travel through the salt involves propagation of energy as a shear wave (S-wave) converted from incident compressional (P-wave) energy. By observation alone, you can’t tell at which interface (top or base of salt) the mode conversion takes place. The third explanation for event 2 is quite plausible. Although you don’t usually think about S-waves when working with marine seismic data, you know they are a real physical phenomenon. Using an estimated S-wave velocity for salt, you find that the calculated position of a reflection involving converted wave energy is fairly close to the observed position of event 2. Unable to think of a simple geologic explanation for this event as a P-wave reflection, you conclude that it involves mode conversion and propose that a modeling study be undertaken (time and funding permitting, of course) to understand the physics of its origin. At the same time, you revisit your previous thinking and wonder if event 3, which you originally thought might be a peg-leg (P-wave) multiple, might also be a converted wave reflection, in this case having converted from a downgoing P-wave to an S-wave at the top of salt, reflected from the base of salt as an S-wave, and then converted back to an upcoming P-wave at the top of salt. This explanation also is consistent with the observed polarity of event 3. Because you don’t have access to prestack data (depth gathers) that might help you identify these events based on move-out correction, you prudently decide to reserve final judgment until your modeling study is completed. Even if the modeling is never done, for the time being you leave your questions unanswered but not forgotten. In speaking with other interpreters, you learn that this triad of baseof-salt reflections has been observed in other depth-imaging projects and is common (you might recall having seen a similar set of three reflections in Figure 14 of Chapter 7). On the salt-flood migration for your data (not shown in this text), you observe a single, sharply focused base-of-salt event, and the other candidate base-of-salt reflections have disappeared as you expect. On the final depth migration (also not shown), you see that reflections extend smoothly from the extrasalt area on the left in Figure 2 into the subsalt portion of the image. One of your fellow geoscientists also refers you to a published case study (Ogilvie and Purnell, 1996) in which modeling results and real 3D data confirm the occurrence and significance of mode-converted waves in subsalt exploration. The preceding discussion is a single example of the chain of interconnected observations, inferences, hypotheses, and tentative conclusions that



05/12/11 2:17 PM


often forms during the course of an interpretation. This chain is pinned by hard data (e.g., a control point for the base of salt in a well) and is as strong as the validity of your observations and sound application of physical and geologic principles. In this example, the salt-flood and final 3D migrations validate identification of event 1 on the sediment-flood migration as the true base-of-salt reflection, but you need the results of further studies such as forward seismic modeling before you can conclude which of your tentative interpretations of events 2 and 3 is correct.

The interpretation paradox An interpreter is naturally curious about the earth — a statement not made in jest. When coupled with your problem-solving skills, this very personal characteristic eventually leads you to deal with the interpretation paradox: When interpreting seismic data, you think that more control will enable you to solve your interpretation problems; however, more control, although helping to solve old problems, tends to uncover new ones. This paradox often manifests itself in statements such as “If only this line were a little longer” or “If only we had extended this 3D survey a little farther” or “If only I had a little more time to work on this” or “If only we had drilled the well a little deeper.” The common phrase in these statements is “a little,” representing that quantum of information or time you wish you had. Unfortunately, because you are curious and because the earth has so many secrets awaiting discovery at scales large and small (and of course because seismic response is never ideal), you will always be able to find several interpretation problems worthy of your attention in virtually every project. But temporal and financial resources are finite; and since you can’t afford to lose sight of your project objectives, whether in business or academia, you can’t pursue every interesting and potentially rewarding line of investigation that presents itself. Your experience will guide you in choosing which problems to address and which to defer. You will learn how to avoid unwarranted “science projects” (disparagingly called by some) and how to argue effectively for and commit to worthwhile studies — and how to decide between the two.

Approximations Approximations are essential to good geologic thinking. Consider the staggering number of approximations involved in attempting to understand geologic time in terms of a human lifespan and how to consistently



05/12/11 2:17 PM


visualize geologic processes (such as seafloor spreading and plate convergence, deposition of deep marine sediments, diagenesis, hydrocarbon migration) that operate at vastly different rates and scales. As discussed and implied throughout this text, indirect measurements of the subsurface of the earth by the reflection seismic method necessarily involve approximations in physics, mathematics, and geology. For example, Chapter 3 describes several important approximations in the basic theory for inverting reflectivity data. To be successful as a geoscientist, you must be creative within a framework of approximations built by science and your own knowledge of your craft. This is expected of you, especially in exploration, but is more easily said than done. In his lecture titled “The Uncreative Scientist” given in 1967 at the University of Chicago, physicist Richard Feynman attempted to link the concepts of creativity and approximations: In particular, there are one or two skills that I see in the creative physics students — but I think it applies more generally — that are missing in the noncreative ones. And that skill is how to deal with approximations. This is very difficult to teach because it’s an art. Noncreative students can’t learn it — it goes back to the business of them wanting something exact when what they really have to deal with is approximations. You must approximate if you’re doing physics because what you’re doing is to take one view of a nearly hopelessly complicated situation. Nothing is simple. The world is enormously complicated … . But the fact to appreciate is that you do approximations all the time … . I’m just trying to show you that in the simplest situations — in every situation — approximation is necessary in order to think. Feynman’s words echo the admonition offered by C. H. Dix (1984) in an interview for The Leading Edge: “After all, the earth is a nonhomogeneous mess and no matter how hard you try not to smooth it out or idealize it in your processing, you have to do some of that. So eventually you’re going to have to use your instinct.” Feynman’s “approximate” and Dix’s “smooth it out” say much the same thing, and you are well advised to heed the words of these scientists as you proceed with your work. Although not an approximation in the strictest sense, the concept of significant figures is frequently ignored or abused in computations based on the results of seismic interpretation, such as time-to-depth conversion, estimation of hydrocarbons in place, and calculation of recoverable reserves. Hence, it is important to remember that accuracy is freedom from error



05/12/11 2:17 PM

176 First Steps in Seismic Interpretation Reference value Probability density

Accuracy

Precision

Value

Figure 3. Representation of the concepts of precision and accuracy (with permission from Wikipedia).

(closeness to true value) and precision is repeatability (see Figure 3). A measurement system or calculation is valid if it is accurate and precise. By definition, the significant figures of a number are those digits that carry meaning contributing to its precision, including all digits except for (1) zeros serving as placeholders to indicate the scale of the number and (2) spurious digits introduced, for example, by calculations carried out to greater accuracy than that of the original data or measurements reported to a greater precision than the equipment supports. Always be careful to honor item 2 to avoid communicating false accuracy in your calculations. At the same time, you should always specify the ranges of uncertainty associated with your measurements and calculations.

Uncertainty and risk You should always take time to characterize the uncertainty associated with the results of a seismic interpretation because the risk of any venture based in part or in whole on those results will depend on your assessment of data quality and measure of uncertainty. For these reasons, the difference between uncertainty and risk must be considered and clarified as much as possible. With reference to Figure 3 and according to Sheriff (1992), uncertainty has to do with the precision with which a measurement is known and does not necessarily imply anything about accuracy. This description of uncertainty as a quantitative measure is complete; but because seismic interpretation is very much a subjective and qualitative activity, in characterizing uncertainty you must allow for nonquantitative aspects that do in fact carry implications about accuracy. For instance, because seismic response is most often if not always nonunique, it is possible that a very



05/12/11 2:17 PM


reasonable interpretation can represent several significantly different geologic models, any one of which by itself can be confidently and precisely determined in quantitative terms. So the uncertainty of any seismic interpretation includes the precision of measurements made and calculations based on its results as well as the degree to which it fits a range of geologically reasonable models. Implicit in this view is the notion that you should place interpretive results within an established quantitative range of uncertainty and, in doing so, be sure you clearly understand and communicate the meaning and limits of that range. Perhaps the most common measure of uncertainty is specifying an error bar or plus-or-minus description, usually expressed as the variation about a mean value (“The predicted depth to the formation is 20,000 ft [6096 m] plus or minus 2%” or “The predicted depth to the formation is 20,000 ft [6096 m] plus or minus 400 ft [122 m]”). In any real measure of an interpretive result, you are expected to know all of the factors that control the magnitude of the uncertainty and how you assigned their values. Your estimation of uncertainty doesn’t need to be symmetric about a mean or most likely value because the individual factors contributing to uncertainty, especially those of a geologic nature, are not necessarily symmetrically distributed in the earth. Intimately associated with uncertainty is the concept of risk, which involves assessing the state of uncertainty, or lack of certainty, regarding the occurrence of a given outcome within a range of possible outcomes for a condition or activity. Risk is most easily expressed qualitatively by degrees within a range using subjective terms such as “low,” “medium,” or “high.” The danger is obviously that these words don’t have the same meaning to all who will base further computations or decisions on them (you might say that their meanings are not precise). In probabilistic terms, you say that certain outcomes have a greater or lesser chance of occurring. In the petroleum industry, this translates into the ever-optimistic determination of probability of success, which communicates the likelihood that the results of your integrated interpretation, if acted upon, have a specified chance of finding oil and/or gas within the ranges of the location and the quantity you have determined. Remember always to speak of ranges rather than single-point values because you have uncertainties associated with the proposed location and quantity. Difficulties arise in attempting to assign quantitative measures to qualitative states of risk, and risking systems most often are established using agreed-upon numerical values or ranges of values associated with what are inescapably qualitative assessments. These qualitative judgments can be tempered based on a study of a large number of known outcomes and the factors contributing to them.



05/12/11 2:17 PM


Every measurement or calculation from a seismic interpretation has its associated uncertainty that must be communicated as part of the interpretation. Keep in mind that risk assessment is a qualitative process and must include consideration of seismic data quality and interpretation uncertainty; it is heavily influenced by experience and is meaningfully communicated only within a system of conventions that objectively assigns numerical values for risk to ranges of possible outcomes. This book does not address how risk systems are developed and applied, largely because they are most often built as proprietary systems by individual companies. It is sufficient to say that all risking systems include at least three primary risk factors: source, reservoir, and trap. Each of these can be subdivided into a number of interdependent subfactors; for example, “reservoir” could be subdivided into “reservoir presence” and “reservoir effectiveness.” Any one or more of what one company might consider to be only one component of a primary factor could in itself be thought of as a primary factor by another company. The interdependencies among factors and subfactors, however defined, would be determined within each risking system.

The workstation environment Undoubtedly, most seismic interpretation will be done in a workstation environment for the foreseeable future, and the power and sophistication of data processing and interpretation applications will continue to evolve. The history of oil and gas exploration and concomitant advances in science and technology are such that you always will be engaged primarily in the business of describing geology. What will continue to change are the tools you use and the nature of the environment in which you use them. Among all of the changes that have taken place, the move from working with paper sections to electronic images on workstations has had the most far-reaching consequences, especially by replacing many tasks formerly done by hand with vastly more efficient and reliable automated processes that have enabled interpreters to reap the considerable rewards of increased productivity. But can it also be said that some situations remain in which interpretation in the old paper section or analog domain is preferable to working in the digital domain of the 21st century? In seeking to answer this question, it is fitting to review the strengths and weaknesses of the modern workstation environment (see Table 1). The last item under workstation weaknesses is perhaps the most important of all. You are a craftsman, and the quality of your work is measured by your judgment and skill in using your tools. The workstation is the most powerful and versatile tool you have, but it is not the only tool in your kit.



05/12/11 2:17 PM


Table 1. Strengths and weaknesses of the modern workstation environment. Strengths

Weaknesses

• Organization of large volumes of data (especially 3D data)

• Misapplication or abuse of poorly understood technologies

• Integration of different types of data

• Reliability of hardware systems and software

• Mechanization of processes formerly done by hand (e.g., event picking, timing, posting, contouring) • Speed, power, and precision of individual operations • Flexibility of display, real-time display modification • Prompt documentation of interpretation work (e.g., electronic screen captures)

• Seismic title blocks (acquisition and processing parameters) usually not displayed • Limitations on interpretive perspective and capacity to incorporate individual style into interpretations • Inability to represent subjective elements (uncertainty) in interpretations • Lack of appreciation of workstation limitations, and perception that the workstation is a panacea, not a tool

Your job requires that you continually improve your skills in using your employer’s workstation system of choice. To the degree possible, you should become “fluent” in working with as many different systems as you can — somewhat akin to learning several languages. All viable workstation systems include the same sets of tools needed to perform basic interpretation tasks. The differences among systems are for the most part manifestations of the uniqueness of each system’s fundamental architecture, which in turn reflects the approaches to interpretation of the geoscientists and software engineers who designed the system. You will hear the term functionality used to describe what you can do on a particular workstation. You must remember that as an interpreter you decide what you need to do in your interpretation project (that is, you develop your own work flows), so you should never become a slave to workstation functionality. Always be on the lookout for ways to increase and improve the functionality of the workstation system on which you work. “Will the computer replace the interpreter? In our opinion, this is unlikely.” —Etgen et al. (2009)

Ergonomics Ergonomics, the science of designing the workplace environment to fit the user, is especially important in the modern workstation interpretation



05/12/11 2:17 PM


setting. In the historical setting of interpreting paper sections, your ergonomic concerns might have included preventing back strain caused by leaning over a drafting table for too long, bending over to pick up a pencil that you had dropped, or keeping from cutting your finger on the sharp edge of a seismic section. Now these concerns involve conditions such as maintaining proper sitting height, posture, and positions of your hands and forearms; having your workstation monitor(s) set at the appropriate height; and ensuring good lighting. Lighting has always been important because you can easily stress your eyes whether working on paper sections or at a workstation; in fact, this may be an even more serious matter in the workstation environment because the effect of looking at a computer screen for hours at a time not only can be tiring but also hypnotic. Many new interpreters become so engrossed in their work that they are genuinely surprised when told that they have been sitting at their workstations, intently correlating horizons and faults, for two or three hours without a break. So in addition to having a well-designed ergonomic setup for your workspace, you should also discipline yourself to take periodic breaks for your mental as well as physical health, both of which contribute to your effectiveness as an interpreter. Most companies provide ergonomics consultation regularly, usually in conjunction with an office move, or by request; you should take advantage of this service. Ergonomics is not just corporate-speak. It is a personal matter that you should take seriously.

Presentations Many professionals consider presentations to be the bane of their existence. Like taxes, they are necessary and unavoidable, and you can’t just hire someone to do them for you. The following suggestions for effective presentations are born of the experience of inadequate preparation, untimely scheduling, failed projector bulbs, nervousness, poorly composed slides (especially the dreaded word slides), restless or inattentive audiences, hidden agendas, and many other alternately frustrating and comical episodes too numerous to recount. The objective of every presentation, independent of its content, should be unambiguous communication of information. In any presentation, you should do the following: • Introduce yourself; some in the audience may not know who you are. • Clearly state the title and objective(s) of your presentation. • Tell your audience what map scale(s) you are using, and give a reference direction (the proverbial “north arrow”) before being asked.



05/12/11 2:17 PM


• Describe the phase and polarity (with reference to an accepted polarity convention) of all seismic data shown in the presentation before being asked. • Be able to discuss the acquisition and processing history of your data. • Be prepared to discuss the control used to make every map you show. • Describe every map symbol you are using, and include map legends as needed. • Avoid garish or kaleidoscopic color tables for displayed lines and maps. • Compose graphics/slides to be as simple as possible. • Maximize the “data-ink ratio” of your slides: Use every “drop of ink” on the slides to convey measured quantities and information, and minimize words, lines, arrows, and other redundant or extraneous annotation (see Tufte, 2001). • Never use a long word when a short one will do (very good advice from George Orwell [1946]). • Don’t be afraid to say “I don’t know” when you don’t know something. • Always acknowledge the work of others when you use it. • Look at your audience, not at the floor, ceiling, walls, or projection screen. • As trite as it sounds, “Tell them what you’re going to tell them, tell them, and then tell them what you told them,” and don’t do so in a condescending manner. • Build flexibility into a presentation, even a short one, so you can adjust your delivery as you continually assess the attention span and interest level of the audience. In summary, give a presentation in exactly the same way that you would want to see and hear the same presentation from someone else. An effective presentation should be long enough to cover all of your essential points yet short enough to maintain the attention of your audience. No doubt there are aspects of marketing in the presentation of interpretation results, but thorough, accurate, and consistent technical work, communicated as such, effectively sells itself, whether in getting you into good business deals or keeping you out of bad ones.

Career development The progression of formal and on-the-job training or learning that is best suited for developing a career as a seismic interpreter has been discussed many times by interpreters young and old and by supervisors, functional



05/12/11 2:17 PM


managers, and training managers. This issue can be broken down into three questions: 1) Are there ideal academic qualifications or degree programs for a career as an interpreter? 2) Should you begin your career in an interpretation assignment, that is, on a project for which you actually begin correlating seismic data, or should you begin in a data acquisition or data processing assignment? 3) Where in the exploration-through-production value stream is the best place for you to begin your career? This text does not propose a “best” learning and development sequence for new interpreters or comment on how companies achieve their training and career development objectives — why the structure and content of inhouse training programs are what they are. Instead, we discuss the questions above in terms of the strengths and shortcomings of alternative answers to each. This is especially appropriate because each interpreter brings unique talents and personality to the job, so there is no one-size-fits-all course of learning and development. The best training sequence should be tailored to individuals within a broader and more generalized framework. The first question might be rephrased as, “Must you have a degree or degrees in geology to be a good interpreter, whatever ‘good interpreter’ means?” The embedded question here is about the level of formal education required, as is the case for some companies, or the level shown by historical performance to be best for the job. The short answer to this question is “no.” It might be said that if you have talent in visualization and problem solving, regardless of your specialization in formal education, then you have a good foundation for interpretation work and can learn what you need to learn about physics and/or geology on the job. This of course does not consider personality traits that might favorably incline you to the demands of interpretation work (see Herron, 2003). At the same time, it would be foolish not to recognize that maximum exposure to geology in the classroom and in the field, as early and often as possible, is of great benefit to a career in interpretation — highly recommended, but not absolutely required. Labeling a person as an interpreter rather than as a geologist or a geoscientist should never prevent a person from going on field trips that provide first-hand experience with the geology expected to be seen in seismic data. Also, the titular formality of geologist versus geophysicist should be abandoned; a person who works with seismic data other than for highly specialized purposes should be a geoscientist. The title “interpreter,” unless applied in a very informal sense, should probably be discarded.



05/12/11 2:17 PM


The second question involves two schools of thought, one of which is that if you’re hired to be a seismic interpreter, then you should begin correlating seismic data as soon as possible, given a certain amount of basic training beyond your formal education, so you can begin to build handson interpretation experience early in your career. In this scenario, you are expected to acquire working knowledge of data acquisition and processing on the job, with augmentation by way of formal industry-sponsored or inhouse instruction. The opposing school is that, as a new hire, you should intentionally learn about data acquisition and processing at the outset, preferably in an assignment to an acquisition or processing group, because you should not begin to correlate seismic data until you have sufficient experience with the processes and techniques that control data quality. There certainly are merits to both views, and the decision on which approach to take may be based not so much on what’s best for you but, unfortunately, what an individual company has determined is best and possible in terms of budgets, human resources, and business demands. In a perfect world, the concerns of both schools might be satisfied if interpretation projects were organized and run so that data acquisition, processing, and interpretation were truly integrated by a team of geoscientists with the requisite experience in each discipline. On this team, you as a novice interpreter (geoscientist) could learn about each discipline through master/ apprentice relationships with more experienced team members, and this learning could be supplemented by timely formal classes. In any case, you have to stay active in your first few assignments, building and maintaining good working relationships with senior technical staff to ensure that you are gaining breadth and depth of experience as well as receiving the formal training you need. Your development objectives and a plan to accomplish them should be established and agreed upon according to company procedures and guidelines. Most companies have training or experiencebroadening programs, but the ultimate responsibility for career development is yours. The third question is related to the function of the business unit or group to which you are initially assigned, ranging from exploration (including regional studies) through appraisal and development to production. In a very general sense, we think of geoscience work as becoming more detailed and fully integrated with technical disciplines such as reservoir engineering as we move from exploration to production, an obvious result of the increasing quantity of and dependence on well data farther along in the value stream. Without a doubt, one of the most important aspects of working with seismic data is the correlation of well data to seismic data, the well-to-seismic tie that links geology and seismic response.



05/12/11 2:17 PM


This link always includes some uncertainty; but in dealing with this uncertainty, you gain valuable experience and develop greater understanding of the strengths and limitations of seismic and well data. Some interpreters tie wells to seismic data, assuming that the greater amount of uncertainty resides in the seismic data and that the well data are always (or almost always) correct. Others tend to be willing to recognize the limitations of well data and spread uncertainty more evenly between the well and seismic data. Well-to-seismic ties are always data dependent and very often are quite subjective, hence the importance of experience and appreciation of uncertainty in working with seismic and well data at the same time, which is more likely to happen with greater frequency in a production rather than an exploration setting. A classic well-to-seismic conundrum that severely tests interpreters is the case of tying two wells to a single seismic line. Imagine that you make each tie independently and with acceptable confidence, but you cannot correlate reflections between the tie points with the same degree of confidence in any geophysically consistent or geologically reasonable way. You must make some adjustments, within the limits of uncertainty of the well data (e.g., the well-log data have not been edited properly) or the seismic data (e.g., the imaging is inaccurate). But how to decide? You must produce “most likely” correlations no matter what and will have to choose the appropriate weights to assign to the well and the seismic data, based on experience and knowledge of acquisition and processing of both types of data. If it is true that some geoscientists are too quick to assume that well data are ground truth and that all problems with well-to-seismic ties originate with the seismic data, then it is important to develop an in-depth understanding of the consistency of seismic data acquisition and processing, especially 3D data, as well as to recognize the limitations of these data. It may be (again, this is not always the case) that this understanding is more likely to develop in a setting where there is less well control, as toward the exploration end of the value stream. Here you can concentrate primarily on building correlation skills, confidence, and a greater understanding of data limitations before you are required to extend or modify your correlations to make well ties. Perhaps this is analogous to learning rules, so that with advancing skills you know when and how to stretch those rules.

Advanced interpretation As with most disciplines or courses of study, seismic interpretation makes increasing demands on its students and practitioners as their proficiency grows. It is safe to say there is general agreement on the fundamentals



05/12/11 2:17 PM


of seismic interpretation, but there probably is no consensus on what constitutes advanced interpretation. You might characterize this lack of consensus by asking the question, “Does the ‘advanced’ in advanced interpretation mean an increase in the complexity of the interpretation problems to be solved or an increase in the sophistication of tools and techniques brought to bear on interpretation problems?” The two parts of this question are not mutually exclusive. As is the case for constructive discussion of any well-posed question, there is merit in an answer that incorporates the strengths of arguments that support either endmember position. Notice that both parts of the question include “increase,” which implies that advanced interpretation, however defined, involves an increase in something — what you might know as experience. So you can rephrase the initial question as, “Do you advance in experience by working on interpretation problems with an ever-increasing degree of complexity and difficulty or by developing proficiency with an ever-increasing number of sophisticated high-tech tools?” This might seem to be a word game now, having substituted “ever-increasing” for “increase” but not really having changed the core question. Rephrasing again, but not in an “either-or” format, the question becomes, “How do you gain meaningful interpretation experience?” There is a relatively simple answer to this question: You gain experience by looking at and working with as much data as you can. The more varied the geologic settings of your project areas, the more complex your problems; the more sophisticated your tools, the better — as long as you look at your data with Dix’s (1952) advice in mind, that “every event has a significance that we can find and that is worth finding.” As your experience grows, you may find that “advanced” interpretation skills are not necessarily introduced in a classroom but rather are learned in team settings or at professional society meetings where you share your experience with other geoscientists and benefit from honest discussions of not only successes but also failures.

Time spent and value added As with any of the technical tasks performed in exploring for and producing hydrocarbons, seismic interpretation has value that is a function of the time invested in it. Of particular interest and concern is the form this function can take. Figure 4 is a representation of the 80–20 rule mentioned in Chapter 8. It illustrates that the value of an interpretation, in terms of observations, inferences, correlations made, and information gained, increases sharply early in an interpretation project, i.e., very much is learned with very little time invested. This proportionality changes as time passes,



05/12/11 2:17 PM


with increasingly smaller increments of value added for increasingly greater amounts of time consumed (diminishing returns). In effect, most of an interpretation, usually the easiest or most straightforward parts, is done quickly, with the bulk of the time spent on problem areas. The actual value of an interpretation hypothetically reaches a maximum after a certain time (point A in Figure 4). Continuing with the interpretation beyond this point can result in a decrease in the overall value of the project, if for no other reason than the wasting of time. This is similar to “gambler’s ruin” — using the commutative law of time and money in throwing good money (or time) after bad. Before this critical point of diminishing returns is reached, you should undertake additional tasks or pursue different lines of investigation that can add value more quickly (upper curve in Figure 4). For example, in a 2D interpretation project, such tasks might include reprocessing, acquisition of additional 2D data, or recommendation to acquire 3D data. In any case, always strive to maximize the value gained from your efforts over time, realizing that “obviously it is foolish to go ahead and correlate where no correlation is possible” (Dix, 1952). A confounding element in the time-value relationship for seismic interpretation, whether 2D or 3D, is that the value of a particular interpretive task

Addition to initial project or new project

Value

A

Initial project

Time

Figure 4. In a typical interpretation project, the tendency is to add incrementally less value as time passes; beyond a certain point, marked A on the figure, you can actually begin to erode value. You can prevent this from happening by adding to your interpretation project (e.g., acquire new data or reprocess existing data and carefully avoid “busy work”) or by moving to a new project.



05/12/11 2:17 PM


may not always be apparent at the time the task is done. The general form of the curves in Figure 4 assumes that the value of a task can be measured accurately at any time in its duration. If there is a lag between completion of a task and realization of its full value, then the time-value relationship represented in Figure 4 can look very different. This means you may have to defend work whose value is not readily apparent at the time you are doing it. In most cases, you successfully advance arguments for doing such work only by illustrating its value from experience. And what is meant by experience? As said before, experience is having been put to the test. A more lighthearted definition is that experience is the acquired ability to recognize a mistake when you make it again. No interpreter has ever passed every test faced or answered every question posed without having erred at one time or another. So, contrary to what many might think, an experienced interpreter does not know all the answers; rather, he or she knows when and how to ask questions that can be answered with confidence.



05/12/11 2:17 PM



05/12/11 2:17 PM

References Bahorich, M., and S. Farmer, 1995, 3D seismic discontinuity for faults and stratigraphic features: The coherence cube: The Leading Edge, 14, 1053–1058. Barnes, A. E., 2007, Redundant and useless seismic attributes: Geophysics, 72, no. 3, P33–P38. Bertram, G. T., and N. J. Milton, 1996, Seismic stratigraphy, in D. Emery and K. J. Meyers, eds., Sequence stratigraphy: Blackwell Scientific, 45–60. Brown, A. R., 1996, Seismic attributes and their classification: The Leading Edge, 15, 1090. Brown, A. R., 2011, Interpretation of three-dimensional seismic data, 7th ed.: AAPG Memoir 42 and SEG Investigations in Geophysics No. 9. Calvert, R., 2005, Insights and methods for 4D reservoir monitoring and characterization: SEG/EAGE Distinguished Instructor Series No. 8. Chopra, S., and K. J. Marfurt, 2007, Seismic attributes for prospect identification and reservoir characterization: SEG. Chun, J. H., and C. A. Jacewitz, 1981, Fundamentals of frequency domain migration: Geophysics, 46, 717–733. Dix, C. H., 1952, Seismic prospecting for oil: Harper & Brothers. ———, 1955, Seismic velocities from surface measurements: Geophysics, 20, 68–86. ———, 1984, Interview with TLE editor Robert Dean Clark: The Leading Edge, 3, 14–17. Ebaid, H., A. Tura, M. Nasser, P. Hatchell, F. Smit, N. Payne, D. Herron, D. Stanley, J. Kaldy, and C. Barousse, 2008, First dual-vessel high-repeat GoM 4D survey shows development options at Holstein field: The Leading Edge, 27, 1622– 1625. Etgen, J., S. H. Gray, and Y. Zhang, 2009, An overview of depth imaging in exploration geophysics: Geophysics, 74, no. 6, WCA5–WCA17. Feynman, R. P., 1985, Surely you’re joking, Mr. Feynman!: Bantam Books. Gould, S. J., 1993, Eight little piggies: W. W. Norton & Co. ———, 1995, Dinosaur in a haystack: Harmony Books. Gray, S. H., J. Etgen, J. Dellinger, and D. Whitmore, 2001, Seismic migration problems and solutions: Geophysics, 66, 1622–1640.


SEG-SEISMIC-11-0601-REF.indd 189

05/12/11 2:18 PM

190 First Steps in Seismic Interpretation Hardage, B. A., M. V. DeAngelo, P. E. Murray, and D. Sava, 2011, Multicomponent seismic technology: SEG Geophysical Reference Series No. 18. Hart, B. S., 2011, An introduction to seismic interpretation: AAPG Discovery Series No. 16, CD-ROM. Hawking, S., and L. Mlodinow, 2010, The grand design: Bantam Books. Herron, D. A., 2000a, Horizon autopicking: The Leading Edge, 19, 491–492. ———, 2000b, Pitfalls in seismic interpretation: Depth migration artifacts: The Leading Edge, 19, 1016–1017. ———, 2001, Problems with too much data: The Leading Edge, 20, 1124–1126. ———, 2003, Characteristics of an interpreter: The Leading Edge, 22, 49. ———, 2009, Interpreting depth-imaged data: Case studies, examples and pitfalls from the interpreters’ perspective: The Leading Edge, 28, 364–367. Hilterman, F. J., 2001, Seismic amplitude interpretation: SEG/EAGE Distinguished Instructor Series No. 4. Lindseth, R. O., 1979, Synthetic sonic logs — A process for stratigraphic interpretation: Geophysics, 44, 3–26. Mitchum, R. M., Jr., 1977, Glossary of terms used in seismic stratigraphy, in C. E. Payton, ed., Seismic stratigraphy — Applications to hydrocarbon exploration: AAPG, 205–212. Mitchum, R. M., Jr., P. R. Vail, and J. B. Sangree, 1977, Stratigraphic interpretation of seismic reflection patterns in depositional sequences, in C. E. Payton, ed., Seismic stratigraphy — Applications to hydrocarbon exploration: AAPG, 117–133. Nettleton, L. L., 1940, Geophysical prospecting for oil: McGraw-Hill Book Company. Ogilvie, J. S., and G. W. Purnell, 1996, Effects of salt-related mode conversions on subsalt prospecting: Geophysics, 61, 331–348. Orwell, G., 1946, Politics and the English language: Harcourt. Payton, C. E., ed., 1977, Seismic stratigraphy — Applications to hydrocarbon exploration: AAPG. Pennington, W. D., A. Minaeva, and S. Len, 2004, Uses and abuse of “phantom horizons”: The Leading Edge, 23, 454–456. Petroski, H., 1992, To engineer is human: The role of failure in successful design: Vintage Books. ———, 2008, Success through failure: The paradox of design: Princeton University Press. Sheriff, R. E., 2002, Encyclopedic dictionary of exploration geophysics, 4th ed.: SEG Geophysical Reference Series No. 1. Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geophysics, 50, 609–614. Tucker, P. M., 1982, Pitfalls revisited: SEG Geophysical Monograph Series No. 3. ———, 1988, Seismic contouring: a unique skill: Geophysics, 53, 741–749.



05/12/11 2:18 PM

References 191 Tucker, P. M. and H. J. Yorston, 1973, Pitfalls in seismic interpretation: SEG Geophysical Monograph Series No. 2. Tufte, E. R., 2001, The visual display of quantitative information: Graphics Press LLC. Vail, P. R., R. G. Todd, and J. B. Sangree, 1977, Seismic stratigraphy and global changes of sea level, part 5 — Chronostratigraphic significance of seismic reflections, in C. E. Payton, ed., Seismic stratigraphy: Applications to hydrocarbon exploration: AAPG, 99–116. Wikipedia, Accuracy and precision, accessed 15 June 2011, http://en.wikipedia.org/ wiki/accuracy_and_precision. Yilmaz, O., 2001, Seismic data analysis: Processing, inversion, and interpretation of seismic data: SEG Investigations in Geophysics No. 10.



05/12/11 2:18 PM



05/12/11 2:18 PM

Index A accuracy, as distinguished from precision, 175–176 (see also precision, as distinguished from accuracy) acoustic impedance (AI), 4, 5, 9, 22, 30, 31, 34, 117, 166 and density logs, 117 and properties of layers, 30 and reflection-coefficient (RC) series, 30, 31, 117 and sonic logs, 117 calculation from reflectioncoefficient series, 34 magnitude and algebraic sign, 22 measurable change, positive, 166 reflection coefficient, 22 acoustic-impedance (AI) boundary, 9, 10 seismic pulse incident to, 10 and structural interpretation, 23 acoustic-impedance (AI) function, and reflectioncoefficient (RC) series, 30, 31 AI (see acoustic impedance) air gun, 38 aliasing, 75, 76, 79, 80 amplitude, 22–28 above-background amplitude, 24 acoustic-impedance boundary, magnitude and algebraic sign, 22 acoustic-impedance boundary, reflection coefficient, 22

and convolutional model, 22 and dominant frequency, 27 and gain recovery, 23 and manifestation of geology, 24 and quantitative analysis, 23 and reflection-coefficient series, 22 and seismic facies analysis, 26 and sequence stratigraphy, 26 and stacking, 23 and time-amplitude (tuning) analysis, 26 autotracking, 24 average absolute amplitude, 25 balancing, 23 baseline value, 22 convolutional model, 22 ΔA, delta amplitude, 22 ΔT, delta time, 22 maximum positive amplitude, 25 quantitative use of, and picking of reflections, 24 root-mean-square (rms) amplitude, 25 amplitude anomalies, 24, 115 highlighting, 24 amplitude balancing, 23 amplitude spectra, 15 amplitude variation with angle of incidence (AVA), 49, 112 amplitude variation with offset (AVO), 32, 49, 112 amplitudes, misinterpreted, and misestimated wavelet phase, 112–113 angular unconformity, 93

anisotropic depth-imaged data, 57, 111 anisotropic prestack depth migration (APSDM), 57, 58 anisotropy, 56 (see also velocity anisotropy) polar, 56 (see also polar anisotropy) vertical transverse, 56 tilted transverse, 56 antialias filter, 75 anticlines, 109, 110 apophenia, 109 apparent-dip 2D migrated seismic lines, and time migration, 128 approximation and thinking, 175 approximations, 174–176 artifacts, and interpretation pitfalls, 105–113 categories, 105 coherent noise, misinterpreted as primary signal, 106–109 correlation, 106 detection of, 113 false time structures interpreted as real depth structures, 109–112 noise interpreted as discontinuous signal, 109 partial stacks and nonrecognition of high-coherence events, 112 seismic artifacts, recognition of, 105 wavelet phase misestimated and amplitudes misinterpreted, 112–113


SEG-SEISMIC-11-0601-IDX.indd 193

06/12/11 7:30 PM

194 First Steps in Seismic Interpretation attenuation, 11, 27 and distance, 11 and dominant frequency, 27 correction of, deterministic, 11 correction of, probabilistic, 11 attribute, seismic, 21 automatic gain control (AGC), 23 and structural interpretation, 23 automatic tracking, versus manual tracking (see tracking, manual versus automatic) autotracking, 24, 27, 29, 96– 104, 121, 139, 151 average absolute amplitude, 25 average velocity, 61 AVO/AVA effects, 50

B balancing and decompaction, 136 bandwidth, 13 baseline survey, 165 basement, 19, 20 beam-migration depth-migration method, 72 bias, 6 bin size, and spatial sample rate, Δx, 81 biostratigraphic data, 116 block diagrams, 149 bottom-simulating reflection (BSR), 107 boundaries, subsurface, reflections, use of, conditions, 19–20 and visual estimation of wavelet phase, 19–20 bow-tie reflection, 66, 68 Brazil, 71 bright spots, 24 buried focus syncline, 69

C calibrated velocity model, for vertical conversion of points, 59 causal wavelets, 15 CDP (see common depth point)

CDP gathers, 47, 48, 49 Cenozoic basins, 24 channel, 100, 101, 115 channel margins, 28 check-shot survey, 38, 39, 40, 60 compared with vertical seismic profile, 40 circular wavefronts, and constant-velocity migrations, 124 clipping values, 23 CMP (see common midpoint) coherence, 28–30 and channel margins, 28 and coherence slices, 29 and coherence volume, 28 and faults, 28, 29 and noise, 28 and 3D seismic data, 28 coherence data, and faults, 87 coherence horizon slice, 29 coherence slices, 133 coherence time slice, 29 coherence volume, 28–29, 30 and data-analysis window, 28 reflectivity volume, and faults, 30 coherent noise, 106, 107 crosscutting primary reflections, 107 interpreted as primary signal, 106 color table, 116 common-azimuth depth-migration method, 72 common depth point (CDP), 43, 44, 47, 48, 49 gather, 44, 47, 48, 49 method of acquisition, model of, 43 common-depth-point interval, and spatial sample rate, Δx, 81 common-depth profiling, 1 common midpoint (CMP), 43, 44 compaction of reservoirs, 166 composite seismic response and reflection coefficients, 18 composite seismic responses, accurate interpretation and source wavelets, 18

compressional-wave reflections, 9 constant-velocity migrations and circular wavefronts, 124 continuous velocity model, for vertical conversion of points, 59, 61 contouring (see gridding and contouring) convolution, 16 and simulation of propagation of seismic pulse, 16 of reflection-coefficient (RC) series, 16 convolutional model, 16–18, 22 composite response, and interference, 16–17 scaled wavelet, 16 seismic response to reflection coefficient, 16 cores, 116 corrected gathers, inspection, before partial stacking, 50 correlation, and depth-migration projects (see depthmigration projects) correlation, jump (see jump correlation) correlation, processes and techniques that control data quality, 183 correlation ambiguity, 132 correlation concepts, 83–113 artifacts and interpretation pitfalls, 105–113 coherent noise interpreted as primary signal, 106–109 false time structures interpreted as real depth structures, 109–112 first look, 83–84 horizons versus faults, 84–93 manual tracking versus automatic tracking, 96–104 multiple reflections, 94–96 noise interpreted as discontinuous signal, 109 partial stacks and nonrecognition of high-coherence events, 112



06/12/11 7:30 PM

Index 195 wavelet phase misdetermined and amplitudes misinterpreted, 112–113 correlation of seismic records, 4, 5 and most common intellectual difficulty, 5 and pattern recognition, 4 correlation polygon, 133, 134, 137 correlation procedures, 115–151 depth-migration projects, correlations, 140–145 interpretation processes and work flows, 149–150 jump correlation, 120, 133–139 loop tying, 120, 121–133 reflection seismic data, procedures, 120 visualization, 145–149 cosine wave, 12, 13 and sine wave, phase relationship, 13 as function of time, 12 shape of waveform, variables in, 12 crosscutting coherent noise, 108 cuttings, 116

D data, phase of, description, 15 data acquisition and processing, as related to seismic interpretation, 5 data management, 158–159 (see also data quality, and data management) data quality, and data management, 153–161 (see also data management; data quality, responsibility for assessment) and depth-migration projects, 158 and detection, 155 and image fidelity, 155–156 and resolution, 155–156 assessment of, 153 data quality, 153–158

data-quality maps, 156–157 nomenclature systems, 159– 161 power of data set, 155 primary elements of, 155 quantitative measurement of, 154 relative to purpose of data, 153–155 resolution, 155 “traffic-light” map, 156–157 variation, mapping of, 156 data quality, responsibility for assessment, 3 (see also data management) data-quality maps, 156–157 decompaction and balancing, 136 δ, Thomsen parameter, and short-offset moveout correction, vertical velocity, 56 ΔA (delta amplitude), 22, 23, 26 and time-amplitude analysis, 26 and tuning analysis, 26 ΔT (delta time), 22, 23, 26 and time-amplitude analysis, 26 and tuning analysis, 26 Δti (two-way-time thickness, ith layer), 36 Δx (spatial sample rate), 2D and 3D seismic lines, 81 relative to bin size, 81 relative to common-depthpoint interval, 81 relative to sampling wavenumber, 81 Δzi (thickness, ith layer), 36 demultiple processing, 95, 172 density logs, 117, 118 and reflection-coefficient (RC) series, 117 and synthetic seismogram, 118 depositional systems, 26 depth domain, 43 depth gathers, 112 depth imaging, 58, 70, 169 and isotropic models, 58 failure of, conditions, 70

inaccurate, manifestation of, 169 depth-imaging domain, pitfalls, 111 depth migration, 57, 70–71, 73, 91, 173 and correction for velocity anisotropy, 57 and time migration, differences, 70–71 artifacts, and faults, 91 methods, migration algorithms, 72 depth-migration artifacts, 91, 168 depth-migration processing sequence, marine 3D data, subsalt exploration, 140 final migration, and imaging subsalt section, 140 salt-flood migration, and imaging base of salt, 140 sediment-flood migration, and imaging top of salt, 140 waterflood migration, and imaging of seafloor, 140 depth-migration projects, 140–145, 158 and data management, 158 and subsalt exploration, 140 correlations in, 140–145 depth-migration processing sequences, steps in, 140 final migration, and imaging subsalt section, 140 salt-flood migration, and imaging base of salt, 140 sediment-flood migration, and imaging top of salt, 140 waterflood migration, and imaging of seafloor, 140 detection, 3, 155 diagenesis, 175 difference survey, 166 dip, 87 dip line, time-migrated, 123 direct hydrocarbon indicator (DHI), 24 display formats, reflection seismic data, 15, 16, 17 display lines, 117



06/12/11 7:30 PM

196 First Steps in Seismic Interpretation distortion, inevitability of, 83 Dix equation, 51, 53 conditions for valid application of, 53 Dix interval velocities, 54, 55 example, 55 sensitivity, to stackingvelocity picks, 54 Dix interval-velocity calculation, sensitivity, illustrated, 54 Dix’s correlation procedure, in modern seismic interpretation, 2 Dix’s threshold of impossibility, 106 domain of information, 4 dominant frequency, 27, 28, 76, 78, 79 and attenuation, 27 and depth-domain data, 28 and time-domain data, 28 and tuning thickness, 76 calculation of, 27, 28 double multiple, 94, 95, 96, 97, 98

E 80-20 rule, 120, 185 elastic medium and seismic pulse, 100 elastic waves, 9 ε (Thomsen parameter), 57 ergonomics, 179–180 η (Thomsen parameter), and deviation of long-offset P-wave moveout, 56 evidence, rules of, 2 extensional fault, 122, 123

F f (dominant frequency, seismic signal), 79, 81 false time structures, 109–112, 116 fault block, 133 fault correlations, 131–133 and coherence slices, 133 and horizontal reflectivity slices, 133

fault cuts, 139 comparing, to identify miscorrelations, 139 fault heaves, 104 fault picks, aliasing of, 131 faults, 28–30, 68, 69, 84–93, 94, 116, 117, 119, 120, 122, 123, 131, 133, 136, 138, 168 accurate positioning of, 131 and coherence data, 87 and depth-migration artifacts, 91, 168 and tents, 131 correlating, concerns in, 86 faults versus horizons, 84–93 interpretation of, 119 listric normal, 90 normal, 86, 131, 136, 138 reverse, 86, 90, 136, 138 tracking of, 85–86 2D and 3D data, compared, 86 fence diagrams, 149 first arrival, 39 flat spot, seismic, 19, 20 fold of coverage, 43 fold of data, volume of data, and reduction in stacking, 49 foreshortened perspective, for viewing data, 146, 147, 148, 149 formalities, titular, recommended abandonment of, 182 forward seismic modeling, 167, 174 4D seismic (time, and three spatial dimensions), 165–167 amplitude-attribute analysis of, differences, 166 baseline and monitor surveys, 165 compaction, 166 difference survey, 166 fluid saturation, variation of, 166 geomechanical changes, 166 “hardening” of reservoirs, 166 impedance changes, analysis of, 166

interpretation and analysis, 166 overburden, stretching of, 166 pore pressure, variation of, 166 purpose, 165 requirements, 165 “softening” of reservoirs, 166 4D time-shift effects, 166 Fourier analysis, 11, 13 and waveform phase, 13 Fresnel zone, 78 and spatial resolving power, seismic data, 78 geometry, 78 measurement, 78 frowning events, 46

G gain recovery, 23 gambler’s ruin, 186 gas breakout, 166 gas-charged sediments, 110 gas hydrate stability zone (GHSZ), 107, 108 Gulf of Mexico, 108 gates, and reflections, 24 geochemical data, 116 geology, value of, in geophysics, 182 ghost multiples, 95 gridding, 121 gridding and contouring, 163–165 gridding programs, and subjective elements of geology, 164 manual operations, merits of, 164–165 purposes, 163 ground truth, 184 Gulf Coast, 119 Gulf of Mexico, 108, 166

H “hardening” of reservoirs, 166 high-amplitude response and hydrocarbon-charged sand, 136



06/12/11 7:30 PM

Index 197 horizon, 10, 84–85 horizon autotracking, 139 horizon identification, wildcat exploration compared with production projects, 10 horizon names, 160–161 horizon time, 21 horizons, 84–93, 94, 136, 160 components, 160 picking, 93, 136 versus faults, 84–93 horizontal reflectivity slices, 133 hydrocarbon migration, 175 hydrocarbon/water contact, 19, 108 for estimation of wavelet phase, 19 hypotheses, working, 170–174

I illumination, 49, 70–73 image fidelity, 3, 156 imaging accuracy, 2D data compared with 3D data, 66, 69 imaging artifacts, 94, 170 impedance changes, analysis of, prestack domain, 166 implicit uncertainty, data acquisition and imaging, 133 impossibility, threshold of, 2 inclined reflector and two orthogonal 2D timemigrated lines, mis-tie, 129–130 interpretation, 1, 2, 184–185 advanced, 184–185 and seismic data quality, 2 interpretation paradox, 174 interpretation pitfalls (see artifacts, and interpretation pitfalls) interpretation processes and work flows, 149–150 generic interpretation, 149, 151 granularity in work flow, 149 processes, summary of, 150 interval transit time (ITT), 36, 37, 53

and interval velocity, 36, 37 and the term “slowness,” 36 interval velocity, 36, 37, 39, 51, 53, 55, 60, 61, 62 and interval transit time, 36, 37 calculated from stacking velocities, 55 calculated with Dix equation, 53, 62 calculation, 39, 51 model, layered, 60 intrusions, 117 inverse seismic modeling, 167 inversion, 30–34 and quality of data acquisition and processing, 32 and synthetic sonic log, 32 and time domain, 32 assumptions about processing input reflectivity data, 32 correlation of, compared with conventional reflectivity data, 32–33 relative to acquisition-pro cessing-interpretation, reflection seismic data, 30 inverted data, scaling of, 32 isotropic depth-imaged data and Z-to-D correction, 111

J jump correlation, 115, 131, 133–139, 168, 170 and correlation polygon, 133 and fault blocks, 133–134 and pattern recognition, 133 and tunnel vision, 134 common use of, 133 flattening, on correlated horizon, 134–135

K k (wavenumber), relative to wavelength, 81 Kirchhoff multiarrival depthmigration method, 72 Kirchhoff single-arrival depthmigration method, 72

L lag, 12 λ (wavelength), 76, 81 relative to wavenumber, k, 81 layered interval-velocity model, 60 layered velocity model, for vertical conversion of points, 59 lead, 12 limestone, 119 limitations, necessity of understanding, 2 line ties by reflection character, 122–126 lines, mis-tied, critical factor, 125 listric normal fault, 90, 122, 123 dip and strike views, 90 long-wavelength structures, 112 loop, 121 loop tying, 115, 120–133 and 3D data, 121–122 and faults, 120

M map migration, 65, 131 and subsurface reflecting points, positioning of, 131 maximum positive amplitude, 25 migration, 4, 63–73, 116, 170 algorithms, matrix of, for variations in complexity of subsurface geology, 72 and dip-field complexity, 72 and multipathing waves, 72 and velocity complexity, 72 artifacts, 170 for structure, simple or complex, 69 for velocities, simple or com plex, 69 map migration, 65 purpose, 63 steep/overturned reflectors, 72 time domain compared with depth domain, 69 types of, for prestack and poststack depth domains, 69 migration, of hydrocarbons, 175



06/12/11 7:30 PM

198 First Steps in Seismic Interpretation migration swings, interpreted as primary signal, 106 migration velocities, differences between, 130 migration velocity model, 116 mis-tie, 122, 123 2D data, 122 mis-tied lines, critical factor, 125 (see also time migration; time migration and apparent-dip 2D migrated seismic lines) mis-ties caused by differences in acquisition and processing, 130 mode-converted waves, subsalt exploration, 173 model-based interpretation, 6 monitor survey, 165 moveout-corrected CDP gathers, 47, 48, 49 moveout-corrected traces, stacking, and noise reduction, 45 moveout correction, as related to velocity, 46 moveout velocities, and errors in stacked traces, 46–47 multifold acquisition, 42 multifold coverage, and conventional seismic velocity analysis, 42 multiple reflections, 94–96, 97, 98 common multiples, 95 demultiple processing, 95 double multiple, 94, 95, 96, 97, 98 ghost, 95 long-period, 94 peg-leg, 95, 96, 97 period, 94 positions of, prediction, 96, 97 recognition of, 96 short-period, 94 multiple velocity functions, for vertical conversion of points, 59 multiples, 94, 106, 172 interpreted as primary signal, 106

multivalued horizon capability, of workstation system, 141

N NMO (see normal moveout) NMO-corrected gathers, 54 noise, 106, 109 coherent, interpreted as primary signal, 106 interpreted as discontinuous signal, 109 nomenclature systems, 159–161 attributes, 159 horizon names, core elements and purposes, 160–161 horizons, components, 160 importance, 159 requirements of, 159 noncausal wavelet, 13, 15 and causal wavelet, 15 normal hyperbolic moveout, 45 normal-incidence reflection, point of, 64 normal moveout (NMO), 43–45 correction, 45 equation, 43, 44 traces, stacking, and noise reduction, 45 normal-moveout (NMO) velocity, 45, 57 and correction for velocity anisotropy, 57 and stacking velocity, Vstk, 45 Nyquist frequency, 75, 76 Nyquist theorem, 75, 78, 81, 121, 164 Nyquist wavelength, and sampling wavenumber, 81 Nyquist wavenumber, and sampling wavenumber, 81

O observe-interpret-test cycle, 6 oil-water contacts, tilted, 19 onland data, 116 optimal stacking velocity, 42– 43 orthogonal 2D time-migrated lines, tying, 125, 127– 129

overburden, stretching, 166 overpressured sediments, 110

P partial stacking, purpose, 49 partial stacks, and nonrecognition of high-coherence events, 112 particle motion as function of time, and seismic waveform, 11 particle motions, positive and negative reflection processes, 11 pattern recognition, and correlation of seismic records, 4 peg-leg multiples, 95, 96, 97, 106, 107, 172, 173 phantom horizons, 119 phase, 10 (see also wavelet phase) phase lag, 12 phase rotation, 13, 34 and inversion of reflectivity data, 34 phase spectra, 15 pitfalls, interpretation (see artifacts, and interpretation pitfalls) plate convergence, 175 polar anisotropy, 56 (see also anisotropy, defined, polar) and reflection seismic data, 56 and vertical transverse isotropy (VTI), 56 characterization, and Thomsen parameters, 56 polarity, 10 polarity reversals, 106 pore pressure, variation, and impedance, 166 poststack depth migration (PoSDM), 69 poststack time migration (PoSTM), 69 power, of data set, 155 precision, as distinguished from accuracy, 175–176 (see also accuracy, as distinguished from precision)



06/12/11 7:30 PM

Index 199 presentations, 180–181 requirements, 180–181 prestack depth migration (PSDM), 57, 58, 61, 69, 91, 92, 98, 110, 111, 168, 169 and Z-to-D vertical correction, 57 isotropic, and migrated velocities, 57 line, 169 velocity model, 98, 111 volume, 168 prestack time migration (PSTM), 69 primary reflection, 94, 95 principal component analysis, 21 probability of success, 177 problem solving, 182 PSDM (see prestack depth migration) pull-up, and shallow salt, 116 pulse width, 17 P-wave conversion to S-wave, 173 P-wave energy, 173 P-wave reflections, 9

Q qualitative assessments, and quantitative expressions, 177 quality, 2–3 (see also seismic data quality) of data, determinants in, 4

R Rayleigh limit, vertical resolution, 76, 79 rays, 64 reflection and reflector, distinguished, 11 reflection coefficient (RC), 4, 5, 9, 16, 18 and composite seismic response, 18 defined in terms of AI, 9 responses, 18 reflection-coefficient (RC) series, 16–18, 22, 30–31, 34, 117

and acoustic-impedance data, 117 and four coefficients, 16 and properties of layers, 30 as derivative of AI function, assumptions, 30–31 as distinguished from reflectivity data, 30 convolved with zero-phase wavelet, 16, 18 derivative of AI function, assumptions, 31 reflection seismic data, 15, 16, 120 composite response to boundaries, 16 correlation, procedures, loop tying and jump correlation, 120 display formats, 15 variable density, 15 reflection times, manual recording, 165 reflections, multiple, 94–96 reflections, ties, 3D data, mi grated, 122 reflectivity data, correlation, compared with correlation, seismic inversions, 32, 33 reflectivity data, zero-phase, 32–33 reflectivity volume, coherence volume, and faults, 30 reflector and reflection, distinguished, 11 refraction, of seismic energy, 70 reservoir engineering, 183 resolution, 3, 75–81, 155–156 aliasing, 75, 76 and Nyquist theorem, 75, 78 antialias filter, 75 resolvable limit, 156 sampling frequency, 75 sampling, specified, 75 resolvable limit, for discrete seismic reflectors, 156 reverse fault, dip and strike views, 90 reverse time depth-migration method, 72 risk, and uncertainty, 176–178 and geologic models, 177

and risk, quantitative expression of, 177, 178 and symmetry of estimate, 177 qualitative assessments and quantitative expressions, 177 risking systems and risk factors, 178 uncertainty, and aspects free of implication about accuracy, 176 uncertainty, and implication about accuracy, 176 uncertainty, common measure of, 177 uncertainty, relative to precision of measurement, 176 risk assessment and seismic data quality, 178 rock velocities, error in measurement, sources, 38 root-mean-square (rms) amplitude, 25 root-mean-square (rms) velocity, 39, 62 rotation of image, 146

S salt, 19, 20, 60, 67, 70, 91, 97, 98, 101, 109, 110, 116, 140, 141–144, 145 and pull-up, 116 and salt-flood migration, 140 base-of-salt reflection, 101 closed bodies, top and base, picking, technique, 141–143, 144 correlation, top and base, accuracy, 140 sutures, 144, 145 top and base of, picking, and tracking, 140, 144 salt bodies, 141, 144, 145 merged, 144, 145 spatial closing and correlation, 141 salt-body geometry, 91 salt-flood migration, and imaging base of salt, 140, 141, 171, 173, 174



06/12/11 7:30 PM

200 First Steps in Seismic Interpretation salt floods, 158 salt overhang, 143 salt sheet, 99, 100, 111, 112, 168, 169 salt sutures, 144, 145 AI contrast, 144 picking of, 144 salt tectonism, 170 sampling frequency, seismic data, defined, 75 sampling wavenumber, 81 relative to Δx, 81 relative to Nyquist wavelength, 81 relative to Nyquist wavenumber, 81 seafloor, 96, 140 tracking, 140 seafloor double multiple, 107 seafloor scarp, 168 seafloor spreading, 175 search image, 3 sediment flood, 98, 141, 158 sediment-flood migration, 140, 171, 174 and imaging base of salt, 171, 174 and imaging top of salt, 140 sediment-flood–salt-flood se quence, 143 seed points, 99, 101, 102, 103 seed tracking, 151 SEG positive standard display convention, reflection seismic data, 16 seismically derived velocity, 41–56 seismic amplitudes, quantitative use of, and picking of reflections, 24 seismic attribute analysis, references, 22 seismic attributes, 21–34, 116 amplitude, 22–28 coherence, 28–30 horizon time, 21 inversion, 30–34 two-way traveltime, 21 use of, 22 utility spectrum, 21 seismic data, correlation with well data, 10 seismic data, sampling, 75

seismic data quality, 3, 155, 156 (see also data quality, and data management; quality, defined) detection, 3, 155 elements of, 3 image fidelity, 3, 156 resolution, 3, 155 seismic facies, defined, 93–94 seismic facies analysis, 26, 93 seismic flat spot, 19 seismic event, 10 seismic interpretation, subjectivity of, 3 seismic inversion, 30–34 and quality of data acquisition and processing, 32 and synthetic sonic log, 32 and time domain, 32 assumptions about processing input reflectivity data, 32 correlation of, compared with conventional reflectivity data, 32–33 relative to acquisitionprocessing-interpretation, reflection seismic data, 30 seismic migration, importance, 72–73 seismic modeling, 167 appropriate complexity, 167 forward modeling, 167 inverse modeling, 167 utility of, 167 seismic pulse, normal incidence at acoustic-impedance boundary, 9 seismic records, limitations of, 83 seismic response, 9–20 and standard impedance configuration, 15 described by way of convolutional model, 9 measurement of, 9 seismic sequences, and unconformities, 92 seismic stratigraphy, 94 seismic waveform and particle motion as function of time, 11 selectivity, 3 semblance plot, 45

sequence boundary, 117 sequence-stratigraphic analysis, 26 shale, 119 shear waves, 9, 173 short-offset moveout correction applied to vertical velocity, 56 shot-profile one-way depthmigration method, 72 sideswipe, 66, 67 signal processing, 13 signal-to-noise ratio (S/N), 26, 39, 154, 155 significant figures, 175, 176 sine wave, and cosine wave, phase relationship, 13 single-valued workstation systems, 141 single velocity function, for vertical conversion of points, 59 sinusoidal wave, a periodic function, single frequency, time domain, 12 slowness, 36 small offset, 51, 53 smiling events, 46 Snell’s law, 70 “softening” of reservoirs, 166 sonic logs, 36–38, 117, 118 and reflection-coefficient (RC) series, 117 and synthetic seismogram, 118 interval transit time, 36 rock velocities, error in, sources, 38 source-receiver offset, 43 source wavelet, and wavelet phase, 17 source wavelets, and differences between composite responses, 18 spatial sample rate (Δx), 2D and 3D seismic lines, 81 speed, a scalar quantity, 35 squash plot, 146, 147, 148 stacked traces, 46–47, 50 and errors in moveout velocities, 46–47 stacking, of data, 42, 43, 49 (see also stacking velocity)



06/12/11 7:30 PM

Index 201 and optimal stacking velocity, 42, 43 and volume of data, 49 partial, purpose, 49 stacking velocity (Vstk), 45, 51, 54, 55, 56, 62 (see also stacking, of data; Vstk) and Dix equation, 62 and NMO velocity, 45 and optimization of signal, 56 and true propagation velocity, 56 and root-mean-square velocity (Vrms), 45 identification of, 45 stacking-velocity volume, 61 standard impedance configuration and seismic response, 15 statics corrections, 112 strike, 88 strike and dip, 115 strike line, rotated into dip line, 124 strike line, time-migrated, 124 structural restorations, 136 structures, false, in seismic depth, interpreted as real, in true depth, 111 subsalt exploration, and modeconverted waves, 173 subsalt plays, 71 subsalt section, imaging, 140 subsalt targets, 3 syncline, 69, 109, 110 buried focus, 69 synthetic seismogram, 11, 46, 117, 118 and well tie, 11, 118 generated with invariant wavelet, 11 two-way time/stackingvelocity pairs, 46 synthetic sonic log, 32 S-wave conversion to P-wave, 173 S-waves, 9, 173

T T (period, time-domain waveform), 81 t (traveltime), 16

tents, 131 θ (maximum dip), 79 Thomsen parameters, 56 δ, and short-offset moveout correction, 56 η, and deviation of long-offset P-wave moveout, 56 3D bin, 78, 79 relation to average dip, 79 relation to average velocity, 79 3D data compared with 2D data, general, 133 3D features on 2D displays, rendering of, 149 3D scale model, 149 3D time-migrated line, 118 tie point, 124–125, 148 correct, conditions for, 124–125 to a strike line, 148 tilted plane-wave depth-migration method, 72 tilted transverse anisotropy (TTI), 56 time-amplitude analysis, 26, 77 time-depth conversion, 57–62 (see also vertical conversion, T to D and Z to D) time-depth function, from check-shot survey, 59, 60 time domain, 43 time-domain imaging, 171 time-domain waveform, 81 time-migrated dip line, 123 time-migrated strike line, 124 time migration, 70–71, 125, 127, 128 (see also mistied lines, critical factor) and depth migration, differences, 70–71 and true-dip lines, 125, 127, 128 2D, and mis-tied lines, 125, 127, 128 time migration and apparentdip 2D migrated seismic lines, 128 (see also mistied lines, critical factor) time sag, and shallow gas accumulations, 109 time spent and value added, correlation of, 185–187

diminishing returns, 186 80-20 rule, 185 gambler’s ruin, 186 hidden value, 186 time structures, false, interpreted as real depth structures, 109–112 time-to-depth conversion, and correction for velocity anisotropy, 57 titular formalities, recommended abandonment of, 182 Tobs (recorded traveltime, source to receiver), 38 trace spacing, 146 trace stretching, 146 tracking, manual versus automatic, 96–104, 151 (see also automatic tracking, versus manual tracking) errors, examples, 101–104 faults, picking of, 99, 101– 104 parameters for, 98–104 precautionary measures, 101 quality control, importance, 104 seed points, 99, 101, 102, 103 seed tracking, 151 vertical seismic sections, 101 tracking artifacts, 29 “traffic-light” data-quality map, 156–157 transformation of observations, reflection-time domain into depth domain, 6 trough-over-peak amplitude response, 19, 25 and hydrocarbon-bearing reservoirs, 25 trough-over-peak reflection, misinterpreted, 113 true-dip lines, 66, 122, 125, 127, 128 and time migration, 125, 127, 128 true propagation velocity, 56 true relative amplitude recovery and preservation, 26 Tstat (vertical traveltime, datum to source depth), 38



06/12/11 7:30 PM

202 First Steps in Seismic Interpretation tuning analysis, 26, 77 tuning model, 76–77 tuning thickness, 19, 76–77, 79 and resolution, 76 turbidite reservoirs, 166 Tvert (recorded traveltime, converted to vertical traveltime), 38 2D data compared with 3D data, general, 133 2D imaging of subsurface, fundamental limitation, 130 migration velocities, differences between, 130 2D migration, failure of, 66 2D seismic lines, orthogonal pattern aligned with predominant strike and dip, 130 tracking horizons on dip lines and strike lines, 130 2D time-migrated lines, 122, 123, 124 two-way time/stacking-velocity pairs, 46 two-way traveltime (TWT), 21, 46, 78 type line, 117 type locality, of geologic formation, 117

U uncertainty, 176–178 and aspects free of implication about accuracy, 176 and geologic models, 177 and implication about accuracy, 176 and risk, quantitative expression of, 177, 178 and symmetry of estimate, 177 common measure of, 177 qualitative assessments and quantitative expressions, 177 range of, 176 relative to precision of measurement, 176, 177 unconformities, 85, 87, 92, 93, 94, 117 and seismic sequences, 92

angular, 93 utility spectrum, 21

V V (average propagation velocity), 78 validity, of measurement or calculation, 176 Vavg (average velocity), 36, 37 variable amplitude response, 112 variable-density display format, reflection seismic data, 15 velocity, 35–62 and Dix equation, 62 and sonic logs, 36 and transformation from time to depth, 35 and well-velocity surveys, 36 anisotropy, 35 average velocity, schematic, 37, 61 basic types, 61, 62 error in, sources, 38 interval velocity, 35, 37, 39, 61 moveout, errors in, and stacked traces, 46–47 P-wave, and anisotropy, 42 root-mean-square (rms) velocity, 37, 39, 62 seismically derived, 41–56 sonic logs, 36–38 sources of data, 36 stacking, 52, 62 time-depth conversion, 57–62 vector quantity, and associated scalar quantity, 35 velocity anisotropy, 56–57 well-velocity survey, 38–41 velocity anisotropy, 42, 56–57 (see also anisotropy) and check-shot survey, 42 and time-to-depth conversion, 57 and vertical seismic profile, 42 correction for, conditions, 57 polar anisotropy, 56

velocity anomalies, and false time structures, 116 velocity picks, accuracy, as related to depth, 46 velocity pull-up, 109 velocity spectrum, 46 vertical conversion, T to D and Z to D, 59–60 (see also time-depth conversion) calibrated velocity model, 59 continuous velocity model, 59 layered velocity model, 59 multiple velocity functions, 59 single velocity function, 59 uncertainty, estimation of, 59 with layered velocity model and interval velocities, 59, 60 with velocity functions, and time-depth functions, 59–60 vertical exaggeration, seismic display, purpose of, 146 vertical resolution, Rayleigh limit, 76 vertical seismic profile (VSP), 40, 42, 59, 117 advantages over check-shot survey, 40 and velocity anisotropy, 42 and well ties, 117 and well-velocity surveys, 40 critical factors, 59 geophones, station spacing, 40 walkaway VSP, 40 walkover VSP, 40 vertical transverse isotropy (VTI), 56 and tilted transverse isotropy (TTI), 56 vibroseis, 38 Vint (interval velocity), 35, 36, 37, 50, 53 visualization, of subsurface geology, 145–149, 182 and motion of observer, 148–149



06/12/11 7:30 PM

Index 203 foreshortened perspective, 146, 147, 148, 149 objectives, 146 rotation of image, 146 trace spacing, 146 trace stretching, 146 VNMO (normal-moveout velocity), 45, 50, 51, 53 accurate estimation, and ray tracing, 51 and calculation of interval velocity by Dix equation, 51 relative to Vint, 50 relative to Vrms, 50, 51 Vrms (root-mean-square velocity), 36, 37, 51, 52 as approximation for VNMO, or stacking velocity Vstk, 51 root-mean-square velocity, 36, 37 volcanics, 19, 20 VSP (see vertical seismic profile) Vstk (stacking velocity), 45, 51 (see also stacking velocity)

W walkaway vertical seismic profile (VSP), 40 walkover vertical seismic profile (VSP), 40 waterflood migration, and imaging of seafloor, 140 water injection, effect on acoustic impedance, 166 waveform, 10 and frequency, amplitude, and phase characteristics, 10 mathematical description for, 10 wavefront, 63, 64, 65

wavelet phase, 17, 18, 19, 20, 112–113 (see also phase) and hydrocarbon/water contact, 19 and seafloor reflection, 19 and source wavelet, 17 critical value of, 112 estimation of, 17–18, 19, 20 impedance boundaries and seismic responses, 20 limitations in estimating, 20 misestimated, and amplitudes misinterpreted, 112–113 wavelets, 11–12, 13, 14, 15, 16, 17 and extraction from seismic data over windows, 11–12 and interpretation of geology, 17 causal, 16 finite, band-limited, 13, 14 higher-frequency components and preferential reduction in strength, 11 noncausal, 15 scaled, 16 zero-phase, seismic response, 15 wavenumber, k, 81 wedge model, 26, 76 well data, correlation to seismic data, importance, 183 well locations, selection, 34 well tie, 11, 116, 117, 118, 184 and density log, 118 and synthetic seismogram, 11, 117, 118 and vertical seismic profiles, 117 correlation for, uncertainty, 117 unsatisfactory, causes, 117 well trajectories, selection, 34

well-velocity surveys, 36, 38–41, 131 and check-shot surveys, 38, 39 and vertical seismic profiles, 40 field setup, 38 procedure, 39 source and receiver geometries, 40 with fixed source, field geometry, 41 with moving source, field geometry, 41 wiggle traces, 15 work flow and interpretation processes, 149–151 generic interpretation, 149, 151 granularity in work flow, 149 processes, summary of, 150 workstation systems, singlevalued and multivalued, 141, 142

Z Z-depths, and isotropic depthmigrated data, 57 compared with anisotropic depth-imaged data, 57 zero-phase wavelet, 13, 14, 15 phase rotation of, 14 seismic response, 15 time duration for given bandwidth, 13 Zoeppritz equations, 9 and nonnormal angles of incidence at AI boundary, 9 Z-to-D (migrated depth-to-true vertical depth) correction, 57, 111 isotropic depth-imaged data, 111



06/12/11 7:30 PM



06/12/11 7:30 PM

First Steps in Seismic Interpretation [D.A. Herron, 2011] @Geo Pedia .pdf

Recommend Documents