4

Rekindling Interest in Seismic While Drilling

S E I S M I C S

Seismic while drilling promises elegant solutions to some limitations of conventional borehole seismics.

But so far, it has failed to grab a significant slice of the market. Recent advances may create new opportu-

nities for the technique.

Richard MeehanDouglas MillerCambridge, England

Jakob HaldorsenMasahiro KamataBill UnderhillHannover, Germany

For help in preparation of this article, thanks to AndrewKurkjian, Schlumberger Wireline Services, Houston,Texas, USA.In this article, TOMEX is a mark of Western Atlas Interna-tional, Inc.1. The principles, history and use of borehole seismics

are covered by:Hardage BA: “Vertical Seismic Profiling. Part A: Prin-ciples,” Handbook of Geophysical Exploration, Sec-tion 1. Seismic Exploration vol.14A, Helbig K and Tre-itel S (eds): London, England: Geophysical Press,1983.Oristaglio ML: “A Guide to Current Uses of VerticalSeismic Profiles,” Geophysics 50 (December 1985):2473-2479.

2. Babour K, Joli F, Landgren K and Piazza J-L: “Ele-ments of Borehole Seismics Illustrated With ThreeCase Studies,” The Technical Review 35, no. 2 (April1987): 6-17.

3. Oristaglio M, Beylkin G and Miller D: “The General-ized Radon Transform: A Breakthrough In SeismicMigration,” The Technical Review 35, no. 3 (July1987): 20-27.

nThe concept: to image the subsurface using normal bit noise and surface geophone.

The aim is to turn conventional boreholeseismics on its head. Instead of locating seis-mic sources on the surface and conveyingreceivers downhole, seismic while drillinguses bit vibrations as a downhole source andsurface geophones to measure signals (above).

This inversion promises timely seismicinformation without interrupting drilling andwithout deploying any downhole hardware.Further, the service may be employed inenvironmentally sensitive areas where sur-face seismic sources are disruptive—for

example, in jungle roads built specially toaccommodate sources. Despite theseadvantages, a widely-used seismic-while-drilling service has so far proved elusive.

This article outlines the principles of seis-mic while drilling and its potential uses,looks at developments to date, and exam-ines how recent research initiatives arerekindling interest in the technique.

Oilfield Review

nConventionalborehole seismics.A basic verticalseismic profile(VSP) surveyemploys a staticseismic sourcewhile the geophoneis moved to differ-ent levels in thewell (top). To createa lateral image ofthe formation,walkaway VSP usesa fixed geophonelocation and a sur-face source that ismoved along a linethat extends fromthe well (bottom).

Source

Reflector

Downgoing multiple

Direct wave

Geophone position Time1

2

3

4

5Reflected primaryReflected upgoing multiple

Subsurface reflector

S1 S2 S3 S4 S5 S6

Source positions

Well

Geophone

Fault

S1 S2 S3 S4 S5

WellHyperbolic direct arrival

Standard VSP

Walkway VSP

S6

The PrinciplesWell location is usually selected using sur-face seismic images. Once drilling is under-way, it is useful to know the bit’s positionrelative to the seismic section. However, thisinformation is not easily available becausethe vertical axis of the seismic section ismeasured not in distance but in “two-waytime”—the time the seismic waves take totravel through the earth, bounce off a sub-surface reflector and return to surface.

To relate the position of the bit to the seis-mic section, it is necessary to convert thevertical axis from time to depth. This con-version requires knowledge of the velocity

January 1993

of seismic waves through the formation.Velocity varies significantly with rock typeand usually has to be measured, rather thanmodeled, using a combination of sonic logsand borehole seismics of a well after it hasbeen drilled.

The theory of borehole seismics has beenknown for many decades.1 At its simplest, ageophone deployed on wireline records thetime that seismic waves take to travel from asurface source to a receiver at known depthin the well. These times are doubled to tiein with two-way time on the surface seismicsection. This simple service is known as a“checkshot” survey.

But there are subtleties that add signifi-cantly to the usefulness of borehole seis-

mics. Good quality data, sampled finely andin sufficient depth, enable a vertical reflec-tion image or vertical seismic profile (VSP)to be created. In a basic VSP survey, theseismic source is static and the geophone ismoved to different levels in the well. Theimage may be displayed either in time, tomatch the surface seismic section, or indepth, to match wireline logs.2

Alternatively, the geophone location maybe fixed and the surface source movedalong a line that “walks away” from the rig.Walkaway VSP produces an image of thesubsurface with lateral coverage that is typi-cally between half and a quarter the welldepth (left). In deviated wells, various com-binations of VSP and walkaway VSP may beemployed to provide the required images.3

Today, borehole seismics delivers a rangeof high-resolution images. However, like allwireline-delivered services, drilling muststop and the drillstring must be removedprior to running the survey. Therefore, bore-hole seismics is typically carried out duringopenhole logging, usually just before casingis run. The results certainly offer usefulinformation, but this may be too late. Thewell may already be in the wrongplace—for example, on the wrong side of afault subsequently revealed by walkawayVSP—and a costly sidetrack may beneeded. Furthermore, it may be expensiveor impossible to locate sufficient surfacesources to create a satisfactory walkawayVSP image.

In seismic while drilling, compressionalwaves emitted by the active bit radiate bothdirectly to surface and downward from thebit, reflecting off formation boundaries. By

5

using surface geophones to detect thissound, the inverse of checkshot, VSP andwalkaway VSP surveys may be obtained.

These techniques offer several advantagesover conventional borehole seismics:drilling need not stop and, because themeasurements are made continuously, theinformation allows well trajectory decisionsto be made before it is too late. Further,using the bit as a source may make it practi-cal to perform large-scale borehole seismicjobs where surface sources are impracti-cal—for example in towns or environmen-tally sensitive areas.

However, seismic while drilling presentssignificant technical challenges. The signalemitted by conventional seismic sources iswell controlled—either an impulsive explo-sion or a sweep from a vibrator of knownsignature—making the time between itsemission and detection relatively easy todetermine. On the other hand, the bit’s sig-nal is essentially continuous and uncontrol-lable. A geophone on surface records con-tinuous seismic radiation as it is transmittedthrough the ground.

In addition, the environment around adrilling rig is very noisy. The comparativelylow-level energy of the drillbit seismic signalis often completely submerged in noise.Onshore, the geophone traces include sev-eral noise components. Some noise that cor-relates with the seismic signal is caused bybit vibrations travelling up the drillstring andthe fluid-filled annulus and then “rolling”along the air-ground interface to the geo-phones—this is called correlated ground roll.Uncorrelated ground roll comes from thevibrations of surface equipment like the mudpumps and engines. Random noise is causedby events like a passing truck or train.

The challenge is to recognize the unknownand variable signature of the bit, to improvethe signal-to-noise ratio and to convert acontinuous emission to one in which dis-crete seismic events may be recognized.

The HistoryThe idea of using drill bit vibrations as anenergy source for subsurface investigationdates back to the 1930s. The original con-cept was proposed for cable drilling, which

6

made processing more straightforwardbecause it applies force to the formationthrough discrete impacts creating a pulsedsignal. The technique failed to mature asrotary drilling became the norm.

Three decades later, rollercone bit vibra-tions were analyzed, initially to assess theireffects on the drillstring and the rig. But thiswork also showed that the vibrations pos-sess sufficient energy and bandwidth to beused as a seismic source.4 Then, in 1985,Société Nationale Elf Aquitaine, Paris,France, patented “instantaneous acousticlogging in a wellbore” using a drill bit assource. The technique overcomes the con-tinuous nature of the source by relating thesignal measured by surface geophones to asignal measured on the drillstring using acrosscorrelation technique.5

Geophones near the rig measure the seis-mic signal and an accelerometer at the topof the drillpipe measures bit vibrationstransmitted through the drillstring. Theaccelerometer receives a continuous signalsimilar to that reaching the geophones. Butthe speed of the signal through the drillstringis different from that through rock, so thetwo traces will be time shifted. By usingcrosscorrelation, the value of this shift maybe estimated.

If the time shift and the velocity of the sig-nal through the drillstring are known, thetravel time of the signal in rock may be cal-culated. But components in the drill-string—drill collars, stabilizers and tooljoints for example—affect propagation ofthe signal, distorting the signature and mak-ing drillstring velocity difficult to calculate.

One proposal by Elf exploits differences indrillstring travel time before and immedi-ately after a new joint of pipe has been con-nected to the string. A crosscorrelationbetween the accelerometer and geophonesignals is made before the connection.Another crosscorrelation is made after theconnection but before any hole has beendrilled. These two crosscorrelations are thencrosscorrelated with each other, giving atime shift attributed to the extra joint ofdrillpipe. By successively combining these,the drillstring velocity may be calculated.

Since the initial Elf patent was filed, a seis-mic-while-drilling-service has been devel-oped and offered under license as theTOMEX survey by Western Geophysical.Time-to-depth conversions have been car-ried out and images comparable to VSP sur-veys have been produced.6 Despite theseresults, seismics while drilling has so farfailed to capture a significant share of thepotential market.

New SystemsGiven the promising potential of seismicwhile drilling, Schlumberger researchershave developed two prototype real-time sys-tems that mirror conventional techniques ofborehole seismics and may promote morewidespread use of the technology.

First, this article looks at an experimentalcontinuous checkshot system designed to bepermanently on site delivering time-to-depth conversions. Then, it examines animaging-while-drilling system being devel-oped to deliver an image comparable towalkaway VSP. This system uses a largearray of geophones and is designed to be onsite for only a few days.

Continuous checkshot: A key to a viablecontinuous checkshot service is cost-effec-tiveness. If large quantities of equipment andmany people are permanently required onsite, costs are generally high. No matter howgood the product, it will be too expensive.

With this in mind, GECO-PRAKLA Engi-neering, Hannover, Germany, is testing afully automated system that keeps the num-ber of data-gathering channels to a mini-mum. Currently undergoing research trialsin the field, the system also employs newprocessing techniques to ensure that perfor-mance and real-time product delivery arenot sacrificed when hardware is simplified.

The system employs 12 or fewer geo-phones, deployed radially starting about200 m [656 ft] from the rig. There are twoaccelerometers on the swivel—the lowestpart of the drilling equipment not rotating.Data are transferred via cable to the rig-based recording unit. Reducing the numberof geophones improves system economicsand makes automation feasible, but presentsa signal processing challenge.

Uncorrelated noise has traditionally beenreduced by summing the responses from alarge number of independent receivers—anirreversible process. With sufficient traces,the in-phase signal is satisfactorily reinforcedand the out-of-phase noise attenuated—thebigger the array, the better the data. Corre-lated ground roll is more troublesome. But,by deploying sufficient geophones in arrayswith carefully chosen directions, correlatednoise may be attenuated.7

However, the effectiveness of this direc-tional attenuation technique is limited, par-ticularly when there are few geophones. Tofacilitate more sophisticated noise removal,the continuous checkshot system preservesindividual geophone data, rather thaninstantly summing them (next page).

The first task in this method is to removeground roll from the individual geophone

Oilfield Review

4. Deily FH, Dareing DW, Paff GH, Ortloff JE and LynnRD: “Downhole measurements of Drill String Forcesand Motions,” Transactions of the American Societyof Mechanical Engineering for Industry (1968): 217-225.Lutz J, Raynaud M, Gstalder S, Quichaud C, Raynal Jand Muckelroy JA: “Instantaneous Logging Based on aDynamic Theory of Drilling,” Journal of PetroleumTechnology 24 (June 1972): 750-758.

7January 1993

5. Staron P, Arens G and Gros P: “Method for Instanta-neous Acoustic Logging Within a Wellbore,” Interna-tional Patent Application under the Patent Coopera-tion Treaty No. WO 85/05696 (May 20, 1985).

6. Rector JW: “Utilization of Drill Bit Vibrations as aDownhole Seismic Source,” PhD dissertation submit-ted to Stanford University Department of Geophysics,September 1990.Rector JW and Marion BP: “MWD VSP and Check-shot Surveys Using the Drill Bit as a Downhole EnergySource,” paper OTC 6024, presented at the 21stAnnual Offshore Technology Conference, Houston,Texas, USA, May 1-4, 1989.Rector JW and Marion BP: “The Use of Drill-BitEnergy as a Downhole Seismic Source,” Geophysics56 (May 1991): 628-634.

7. Hardage BA: Crosswell Seismology and Reverse VSP,vol. 1. London, England: Geophysical Press Limited(1992): 79-101.

Processing

Crosscorrelation

Deconvolved signal—reflections removed

Geophones

Accelerometer signal

Geophone signals

Filtered signal

Adaptive filter

Time shift

Time of signal in formation

Time of signal through drillstring

Drillstring image

Computation

Bit

nDrillstring image.Drillstring vibra-tions measured atsurface are used tocalculate thechanges inimpedance thatthe vibrationsencounter as theytravel from the bot-tom of the drill-string to the top. By deciding whatpart of the imagecorresponds to thebit, it is then possi-ble to work out thedrillstring traveltime.

bit

with the average of the two traces measuredby the accelerometers on the drillstring. Thiscrosscorrelation also gives the time shiftbetween accelerometer and geophone sig-nals—the difference in signal velocitythrough the drillstring and formation. Thetime shift is then used to calculate the signalvelocity through the formation.

Determining formation velocity alsorequires knowing the drillstring travel time.As already noted, the many components inthe string complicate calculation of thistravel time and a number of methods havebeen proposed—like Elf’s double crosscor-relation process.

The continuous checkshot system uses anew technique called drillstring imaging,also devised at SCR, to model changes in theacoustic impedance of the drillstring, giving abetter understanding of the velocity of thesignal through the drillstring (next page,top). The time shift and the drillstring traveltime are then used to compute the forma-tion travel time.

Imaging while drilling: The continuouscheckshot system employs a relatively smallnumber of geophones to gather datathroughout drilling. It is not designed todeliver complex images like those producedby walkaway VSP. This requires a largedata-gathering exercise employed for a shorttime and different processing techniques.Researchers at both SCR and GECO-PRAKLA Engineering are constructing com-plex images using long data recordsobtained by large arrays of surface geo-phone at a range of radial offsets. The scaleof data gathering rivals that used for con-ventional 2D surface seismics, employingarrays of 40 to 220 stations with up to 48geophones per station.

The imaging technique is based onresearch started at Schlumberger-DollResearch, Ridgefield, Connecticut, USA.

traces. Signal processing experts at Schlum-berger Cambridge Research (SCR), Cam-bridge, England, have developed adaptivenoise filters to do this. Adaptive filters havebeen used for many purposes since the1970s—current examples include telephonechannel equalization to prevent echoes onlong-distance telephone lines and antinoisesystems to reduce environmentally undesir-able sounds.8

For seismic while drilling, the filterexploits differences in the moveout of com-ponents within the traces. Ground roll

approaches the geophones from the sideand exhibits moveout across the array.However, the wavefront of the seismic sig-nal approaches the array from below, haszero moveout and is in phase (next page,bottom). By using these differences in move-out to distinguish between different parts ofthe trace, the adaptive filter effectively atten-uates the ground roll, while allowing theseismic signal to pass (see “Cutting Noisewith an Adaptive Filter,” below).

Random noise may be removed by cross-correlating the individual geophone traces

Adaptivefilter

d

Filtered signal

Geophone 2 signalGeophone 1 signal

Ground roll

8. Widrow B, Glover JR, McCool JM, Kaunitz J,Williams CS, Hearn RH, Zeidler JR, Dong E andGoodlin RC: “Adaptive Noise Cancelling: Principlesand Applications,” Proceedings of the Institution ofElectrical and Electronic Engineering 63 no. 12(December 1975): 1692-1716.

9. Desler JF, Farmer PA and Haldorsen J: “Method ofand Array for Vertical Seismic Profiling,” EuropeanPatent No. 0 294 158 A2 (May 31, 1988).Miller D, Haldorsen J and Kostov C: “Methods forDeconvolution of Unknown Source Signatures fromUnknown Waveform Data,” U.S. Patent No.4,922,362 (May 1, 1990).

10. Oristaglio M, Beylkin G and Miller D, reference 3.Miller D, Oristaglio M and Beylkin G: “A New Slanton Seismic Imaging: Migration and Integral Geome-try,” Geophysics 52 (July 1987): 943-964.

Cutting Noise with an Adaptive Filter

nAn adaptive filter at work.

To illustrate how an adaptive filter removes

unwanted noise, consider an array of two geo-

phones (above). The seismic signal from the drill

bit arrives at both geophones virtually simultane-

ously. But the wave front of the ground roll arrives

at the geophone closest to the rig before it arrives

at the second geophone. Therefore, the trace from

the first geophone contains an advanced version

of the ground roll with respect to the trace from

the second geophone. The size of this advance

depends on the speed of the wavefront and the

separation of the geophones.

To remove the ground roll component from the

second trace, the output from the first geophone

is delayed by a short time d—where d is less

than the advance. Now the trace from the first

geophone contains a delayed version of the seis-

mic signal compared to the second geophone, but

still has an advanced version of the ground roll.

The time-delayed first trace is then weighted

by an adaptive filter and subtracted from the sec-

ond trace. Because the filter input contains an

advanced version of the ground roll, it can suc-

cessfully remove ground roll from the second

geophone. But because it contains only a delayed

version of the seismic signal, the signal in the

second geophone’s trace cannot be removed. The

filter weights are adapted so that the total power

of the output signal is minimized. When the mini-

mum power has been achieved, as much of the

ground roll as possible has been removed.

A similar technique may be used even if the

seismic signals do not arrive simultaneously, so

long as the seismic signal has a different move-

out across the array compared to the ground roll.

And for the continuous checkshot system, the fil-

ter is extended to cover all the geophones and not

just two.

8 Oilfield Review

Ground roll Seismic signal

Geophones

9January 1993

nContinuous checkshot data gathering and processing. Geophonesignals are filtered and crosscorrelated with the deconvolvedaccelerometer signal. This yields the time shift—the difference intravel time taken by the signal through the formation and its timeup the drillstring. The accelerometer signal is also used to createa drillstring image. The time shift and the image may then be usedto compute the time the signal takes to travel through the formation.

nHow geometryaffects moveout.Ground rollapproaches thegeophones from theside and exhibitsmoveout across thearray—the wavefront reaches thegeophone nearestto the rig first, thentravels to the nextgeophone and soforth. The wave-front of the seismicsignal approachesthe array frombelow. Because thedistance betweenthe nearest and far-thest geophone issmall—typicallyabout 20 m [66 ft]—relative to thedepth of the well,the seismic signalarrives at each geo-phone virtuallysimultaneously.

Tim

e

Surfaceequipment

Pow

er

Drillpipe

Heavyweight drillpipe

Bottomholeassembly

Change in impedance

Frequency

The processing capitalizes on the relativeabundance of geophone data and tracks thewavefront as it travels through the forma-tion, potentially estimating the bit signal andthe earth’s response without usingaccelerometer data.9 However, by employ-ing the accelerometer input, data may becompressed, making it feasible to store themassive volume of information collectedover three or four days.

Each trace contains a common bit signa-ture, and noise that varies from trace totrace. With time-delay curves, stacking anddeconvolution filtering, new signals are cre-ated that represent what the traces wouldhave looked like if the source had been anoiseless pulse—the earth impulseresponse. This converted form is thenmigrated to create an image (see “Spikingthe Traces,” page 10).10

The feasibility of imaging while drillinghas been demonstrated by a number ofresearch experiments. For example, inMulmshorn, Germany, in 1991, imageswere created using 240 groups of 24 geo-

phones that compare very favorably withwalkaway VSP images of the same interval(above).11

A rigorous way of assessing the resolutionof seismic images is to analyze the distribu-tion of the signal across the frequency range0 to 100 hertz. Four types of seismic imageresulting from the Mulmshorn experimenthave been compared. The spectrum from aconventional VSP-derived image wasalmost uniformly distributed—the ideal. Thespectrum for walkaway VSP was almost asgood. Both surface and seismic-while-drilling surveys show some deficiencies atthe higher frequencies, although the imagesare still acceptable (right).

A subsequent test, at Voelkersen, Ger-many, further revealed the scope of thetechnique. In March 1992, an experiment

Starting with signals created by a noisy continu-

ous source, the objective is to create the earth

impulse response—the traces that would result if

the source had been a noiseless pulse. This

involves a number of steps (next page, left).1

Estimation of source signal—In contrast to the

continuous checkshot system where the seismic

signal arrives virtually simultaneously at a short

array of geophones, the imaging system uses an

array spanning a large distance—kilometers—so

the signal does not arrive at each geophone at

the same time. Each trace measurement contains

a replica of the source signal that must be time

shifted to make them appear to arrive at surface

simultaneously. These shifted traces may then be

stacked to create a single trace, which represents

an estimate of the source signal.

Finding the time delays assumes that the bit

signal travels to surface as an approximate spher-

ical wave and is received by the array as an

approximate hyperbola. Best-fit analysis may be

used to find the hyperbola that fits the data. How-

ever, if geophone data are correlated with

accelerometer data, the signal-to-noise ratio is so

improved that the time-delay curve is much more

obvious, making best-fit analysis redundant.

Deconvolution and filtering—Once the traces

have been shifted and stacked, a process called

deconvolution is carried out. This removes the

source signature and reduces each signal’s first

arrival to a spike. Additional copies of the source

signature are present in the data because of

reflections in the earth. A single filter is applied

to all the traces, so these additional signals are

also converted into spikes, with amplitude and

time delay relative to the first arrivals. The spikes

provide the basis for the seismic-while-drilling

image.

The deconvolution filter is similar to a conven-

tional noise-reducing Wiener filter and is data-

adaptive—its nature changes according to the

data.2 Wiener deconvolution requires an estimate

10 Oilfield Review

nWalkaway VSP image (left) and a seismic-while-drilling image(right) obtained using similar processing, from a research experi-ment carried out in Germany in 1991—the gamma ray log isshown for reference. The strong reflector at about 4700 m is theZechstein salt formation. The reservoir is in the thin sands visiblein the gamma ray log at about 5000 m. Hints of the sands may beseen in both images. However, the reflector directly above theZechstein is better revealed by the seismic-while-drilling imagealthough this may be partially a function of the different sourcearray employed for the wireline survey.

nComparing the spectra of four seismictechniques used on the same well.

Spiking the Traces

11. Haldorsen JBU, Miller DE, Walsh JJ and Zoch H-J:“Multichannel Approach to Signature Estimation andDeconvolution for Drill Bit Imaging,” paper BG5.5,presented at the 62nd Annual International Meetingand Exposition of the Society of Exploration Geo-physicists, New Orleans, Louisana, USA, October25-29, 1992.

3500500 m–500 m

Walkaway VSPD

epth

, mDrill bit seismic

4500

5500

500 m –500 m

Spectral distribution

100

0

Freq

uenc

y, h

ertz

Surface seismic

WalkawayVSP

Drill bitseismic

VSP

Gamma ray log

nBefore and after deconvolution. The left image shows the equivalent of fivehours of data. Some 302, one-minute bit-noise records have been correlated,truncated and stacked—representing a compression of more than four ordersof magnitude of the drill bit waveform. These data have then been treatedusing the deconvolution filter (right). This removes noise incoherent with thesource signature, leaving a sharp direct arrival together with additional sharparrivals from subsurface reflectors and coherent seismic sources.

of the noise. In this case, because the filter is

optimized to work uniformly on all traces, noise

estimation is carried out implicitly by considering

the ratio of coherent energy—the signal that is

aligned across the array—and the total energy at

each frequency.

The filter enhances everything coherent with

the signal—like reflections—and suppresses

incoherent noise, while preserving relative arrival

times (below). Built into the filter is a natural

noise suppressor, attenuating the energy at fre-

quencies where the signal-to-noise ratio is poor.

Once the traces have been spiked, the filter

automatically reverses the shift to create the

earth impulse response.

Refining the result—The earth impulse response

may be refined using an improved time-delay

curve obtained by picking new break times on the

deconvolved data. Stacking and deconvolution may

then be repeated using the new time-delay curve.

Processing—These traces may then be processed

as normal walkaway VSP data using conventional

migration.

11January 1993

Offset, m

Time shifting to align direct arrivals

Applying the deconvolution filter

0

1

2

3

Offset, m

0

Tim

e, s

ec

200

4

Shifting

0 200-200

Tim

e, s

ec

0

1

2

3

Tim

e, s

ec

0

1

2

3

-200

nThe deconvolution process.

1. Haldorsen J, Walsh J, Miller D and Zoch H-J: “Optimal ArrayFocusing in Deconvolution for VSP,” to be presented at the 55thAnnual Meeting and Technical Exhibition of the European Associ-ation of Exploration Geophysicists, Stavanger, Norway, June 7-11, 1993

2. Wiener N: Extrapolation, Interpolation and Smoothing of Station-ary Time Series. New York, New York, USA: John Wiley & Sons,1949.

0

–1000 m WellWest East

1000 m

1

2

Stack of 302 one-minute fieldcorrelated records

Deconvolved data from fivehours of recording

Tim

e, s

ec

–1000 m WellWest East

1000 m

was carried out using 7 kilometers (km) [4miles] of surface geophones in a suburbanarea (left ). Data were collected over fourdays, during which a rollercone bit drilled89 m [292 ft] of formation to a depth of3075 m [10,090 ft]. A preliminary imagewas produced in the field after only 25 m[82 ft] of drilling and within a couple ofhours of acquiring the data.

The target for this well was about 5 km [3miles] deep. Sand layers below the Zech-stein salt are typically porous and gas-filled.However, wells completed close to faultshave proved unproductive because of lowpermeability. The available 2D surface-seis-mic data had been interpreted showing agraben below the Zechstein, but the imagewas not clear. The goal of the experimentwas to obtain a better image of the targetzone to clarify the structure and guidedrilling to potentially productive locations.

The geophones were laid out in three sur-face lines. There were 140 channels, eachconsisting of 24 vertical geophones (nextpage). Information from two accelerometers,mounted on the rig swivel, was used tocompress the sampled data by crosscorrela-tion. Before processing, data were stackedfor each 1-m [3-ft] interval that wasdrilled—except for 11 m [36 ft] of drillingwhere seismic-while-drilling data could notbe collected. The 78 stacked records,deconvolved and corrected for time delay,were then imaged using standard VSPmigration. The images show good agree-ment with the surface seismic sections, butalso increased detail (below). This detailallowed the driller to plan the well locationwith greater precision.

So far, the feasibility of both continuouscheckshot and imaging while drilling hasbeen demonstrated, but only in a few envi-

12 Oilfield Review

nSeismic-while-drilling image superimposed on the conventional surface seismic

nGeophone locations used for the Voelk-ersen experiment. If conventional boreholeseismics had been used, seismic sources,rather than geophones would have beendeployed in this suburban environment.

1.5

–1500 m 1500 m

2.0

2.5

3.0

Tim

e, s

ec

WellWest East

ronments. To date, the systems are limited toonshore applications in wells with low devi-ation being drilled through hard formationswith rollercone bits, without topdrives andwithout mud motors.

Some other environments and applica-tions are not expected to cause too manyproblems. Offshore use will necessitaterethinking of hardware requirements anddifferent processing, but these are notexpected to be insurmountable. And, whiletopdrives have not been used yet, no majorproblems are anticipated.

However, other cases are more daunting.Polycrystalline diamond compact (PDC) bitsdo not produce the same signal as roller-cone bits. Horizontal wells will need a dif-ferent arrangement of sensors and new pro-cessing techniques. And operating with mudmotors is an unknown quantity.

January 1993

N

As researchers gather field experience, itwill be possible to better define the limits ofthe technology. This may then allowdeployment of systems that minimize poten-tially costly missed targets and subsequentsidetracks, and identify possible hazardsahead of the bit. For the explorationist, thesesystems offer the chance to refine the finaltarget while the well is being drilled andbefore it is too late. —CF

nHow the Voelk-ersen geophoneswere laid out. Twoparallel 3-kilometer(km) [1.9-mile]strike lines werejoined by a 1-km[0.6-mile] dip line.The dip line ranradially from therig from about 600m [1970 ft]. At 50-m[164-ft] intervalsalong these lines,there were 140channels, eachconsisting of 24 ver-tical geophonesplaced every 2 m[6.5 ft].

13

Well site

500 m

200m

Pointe Noire

Gabon

Congo

Cabinda(Angola)

River K

ouilo

u

100m

1000m

2000m

Nigel PlattPeter PhilipWoking, England

Stewart WalterARK Geophysics LimitedMilton Keynes, England

Going For The Play:Structural Interpretation in Offshore Congo

nThe coastline ofthe Congo, show-ing survey lines(red) of the deep-water nonpropri-etary survey runby GECO-PRAKLAin 1990 and linesfrom a previoussurvey (blue) runtwo years earlier.Current offshorefields are near thecoast in waterdepths less than100 m [330 ft]. Linedirection for thedeepwater surveywas chosen mostlyperpendicular tothe coast—that is,parallel to theregional dip, thedirection mostlikely to giveinsight into struc-ture and plays.

S E I S M I C S

Petroleum exploitation begins

with geologists and geophysicists

exploring unknown territory with

few wells for guidance and no visi-

ble sign of oil and gas. Their task

is to uncover the best prospects

and decide whether oil companies

should commit to millions of

exploration dollars. A key tool is

structural interpretation of seismic

data. A survey shot in deep waters

off the Congo mainland shows how

geologists apply this technique.

Producing fields

Zaire500 km

14

For help in preparation of this article, thanks to JohnBooth, Amerada Hess, Libreville, Gabon, andDominique Pajot and Phil Trayner, GECO-PRAKLA, Sta-vanger, Norway.

Orthogonal linesparallel to regionalstrike completedseismic coverage.

Given the difficulty of finding oil and gas, itis strange that the spirit of exploration geol-ogy is captured by the word play. Noconnection here with frivolous wanderingsthrough exotic lands by geological super-sleuths. Play in the exploration contextdescribes a geological configuration thatfavors the accumulation of hydrocarbons.Plays and their associated prospects are

what exploration geologists spend most oftheir professional lives looking for.

Until oil or gas is discovered, plays existmostly in the mind. Focusing on one oranother of the earth’s basins, explorationistssearch for that seemingly impossible con-catenation—source rock, migration path,reservoir and seal.1 Each vital ingredientmust be present, both at the correct physicallocation and at the right time.

Oilfield Review

1. For more on petroleum plays:North FK: Petroleum Geology. Boston, Massachusetts,USA: Unwin Hyman, 1985.Allen PA and Allen JR: Basin Analysis: Principles andApplications. Oxford, England: Blackwell ScientificPublications, 1990.

nSatellite gravity map of West African coast and mid-Atlantic,showing dense oceanic crust (red) to the left and less dense fill ofsedimentary basins (blue) along the coast. The southwest-north-east lineations are transform faults accommodating the differen-tial spread of the mid-Atlantic and African plates. The map wasderived from satellite measurements of mean ocean height, aparameter dependent on the cumulative mass distribution anddensity of the underlying rock. Dense rock such as basementattracts water from the surroundings and creates a bulge in theocean, increasing ocean height. Less dense rocks such as sedi-ments give rise to a depression in the ocean surface. (Courtesy ofARK Geophysics Limited.)

Source rocks rich enough in organicmaterial must have been buried and heatedto sufficiently high temperature and longenough to form petroleum. Petroleummigrates upward, so there must have been aconduit to guide it. And a porous, perme-able reservoir rock capped by impermeablerocks is required to receive and trap thefluid. Finally, the ensuing geological evolu-tion—commonly several tens or even hun-dreds of million years—must have left thereservoir and seal intact.

As explorationists focus on a sedimentarybasin, this juxtaposition of geological coin-cidence occupies the mind. Every scrap ofevidence is used to refine the notion of howthe basin might have evolved and whetherthe structural and sedimentary history mightfavor a play. Outcrop geology, satelliteimagery, magnetic and gravity surveys, andespecially seismic data contribute to theinterpretation process. This article showshow various plays in offshore Congo crystal-lized in the minds of GECO-PRAKLA geolo-gists as they reviewed a nonproprietary seis-mic survey shot in West African waters.

Geologists first became interested in theCongo in 1928, attracted by tar seepagesnear Pointe Noire on the coast and knowl-edge that the subsurface contained thickdeposits of salt—salt intrusions provide aclassic trapping mechanism for hydrocarbonaccumulation (previous page). The salts hadbeen identified in numerous boreholesdrilled for potash mining exploration.

Oil was discovered at Pointe Indienne, 20kilometers [12 miles] north of Pointe Noire,in 1959. But further onshore explorationproved fruitless—some oil shows, but noth-ing of commercial value. In the late 1960s,an Elf/AGIP partnership discovered oil off-shore. Further successes yielded five off-shore fields and oil production from theCongo now exceeds 115,000 barrels perday. Recoverable reserves are currently esti-mated at more than a billion barrels.

The GECO-PRAKLA nonproprietary sur-vey of 1990, which forms the basis of theinterpretation described in this article, cov-ered 15,000 square kilometers [5800 squaremiles]. The survey ventured from shallowwater at the edge of the continental shelf,where oil plays were known to exist, todepths up to 2000 m [6560 ft]. This isdeeper than the current limit of commercialexploitation, but the deepwater data werenevertheless crucial for elucidating basinstructure and migration pathways from deepin the basin. The survey lines were designedto connect with previous surveys conductedcloser to the coast and intersect any explo-

January 1993

ration wells where precise lithological datawould be available.

In preparation for designing and theninterpreting the survey, the GECO-PRAKLAinterpreters amassed and absorbed an enor-mous variety of data, including satellite sur-veys showing the gross underlying structureextending from the African continent to themid-Atlantic, and data from wells in thearea (above). Gravity and magnetic surveyshelped determine regional fault patterns andthus the optimal directions for seismic sur-veying. Estimates of geothermal gradientsthrough geological time were made avail-able for evaluating the thermal history oforganic-rich formations and the consequentprobability of hydrocarbon generation.Finally, the interpreters had to bone up on

West African geological history, brieflydescribed here:

The offshore basins of the Congo formpart of the West African Salt Basin, a largecollection of basins stretching 2000 km[1250 miles] from south Cameroon toAngola. The salt was deposited duringAptian times about 120 million years agowhen the rifting of South America from

15

Pre

-rift

Early extension

Cratonic rifting

Rift subsidence: lacustrine and fluvio-deltaic deposition

Major rifting

Lacustrine marl deposition and clastic infill

Ear

ly r

iftLa

te r

ift

Western African Shield

GuianaShield

ParanaibaBasin

GuaporeShield

SanFrancisco

Basin

ParanaBasin

NorthNigerianShield

ChailluMassif

CongoBasin

KalahariBasin

Afro-BrazilSag

Basins

LakeChad

R. A

maz

on

Pacific

R. C

ongo

SW NE

120 to 150 million years

Structural history of offshore Congo illustrating basin evolution during

the last 150 million years, with three paleogeographic maps—rift in the

Late Jurassic, transition in the Aptian and drift in the Late Cretaceous—

and sequence of southwest-northeast transects from ancient to present

(transect position marked red on maps).

Structural History of West Africa

From “Hydrocarbon Potential of the West

African Salt Basin, Regional Exploration

Report, Volume III: Congo.” Pinner, England,

1987. Courtesy of Fairway Exploration Ltd.

West Africa gradually evolved into a full-fledged drift. Later salt movement wouldcreate anticlinal folds capable of trappinghydrocarbons.

The story, however, begins thirty millionyears earlier during the Late Jurassic (see“Structural History of West Africa,” above).At this time, extensional faulting and subsi-dence took place in the part of the Gond-wana supercontinent that would eventuallybecome both the east coast of South Amer-ica and the west coast of Africa. Furtherstretching or extension in the Early Creta-ceous led to the formation of a large-scale

16

rift along the future western African andeastern Brazilian margins. A modern paral-lel is the Great Rift Valley extending fromthe Red Sea to the Zambesi River.

Initially, the rift and its basins were abovesea level and isolated from the ocean. Largelakes formed in which sandstone and shalesaccumulated. Some shales were depositedin oxygen-deficient water, allowing preser-vation of organic matter. These formationsare the source for the billions of barrels ofoil found in the West African basins.

By Aptian times, continued subsidenceand a rise in global sea level permittedincursion by the sea. At first, this was inter-mittent, with the sea alternately enteringand receding from the basins. This createdideal conditions of repeated evaporationand marine flooding to create thick depositsof halite, the Aptian salts.

Later, during the mid- to Late Cretaceous,the area was definitively submerged andcontinental breakup of Gondwana led to aseparation, or drift, of South America fromAfrica.2 Whereas the basins had previouslybeen linked on one continental plate, nowthey were separated by a widening tract ofocean, as the Atlantic opened through injec-tion of new oceanic crust at the mid-oceanridge. Sedimentation was now marine, withthick deposits of limestone, sandstone andshale. Further subsidence took place in theLate Tertiary and was probably associatedwith faulting related to the collision of theEurasian and African plates.

This big picture the interpreters learnedfrom extensive research into the region’s

Oilfield Review

Westward tilt and subsidence

Drif

tS

ubsi

denc

e

WalvisRidge

Rio GrandeRise

Subsidence and open marine deposition

SW NE

110 million years to present

2. Terms for geological time units have both physicaland temporal meanings. Early or Late Cretaceous, forexample, refer to early or late in the Cretaceousperiod. But Lower and Upper Cretaceous refer to thelower or upper part of the Cretaceous rock sequence,also called system.

3. For a review on seismic data processing:Boreham D, Kingston J, Shaw P and van Zeelst J: “3D

Tran

sitio

n

WalvisRidge

Benue

Trou

gh

Microplate

Erosion and peneplanation

Evaporite phase: restricted marine transgression

SW NE

120 to 110 million years

geological history. Another card in theirhand was knowing which formations hadproduced oil shows during drilling in thearea, which of these had produced com-mercially, and which formation was themost likely source for the shows—the mostimportant was the Marnes Noires, or blackmarls, continental deposits formed underlacustrine conditions late in the rift phaseand prior to the invasion of sea water.

The majority of commercial reservoirs inoffshore Congo occurs in sandy anddolomitic rock deposited during the earlydrift phase in the mid-Cretaceous, withstructural traps created by movement of the

January 1993

Marine Seismic Data Processing,” Oilfield Review 3,no. 1 (January 1991): 41-55.

transition-phase Aptian salts. Noncommer-cial oil shows have been found aplenty inrift and transition rocks of the Lower Creta-ceous, but no oil has been tested in UpperCretaceous or Tertiary formations.

The interpreters knew the source of oil,therefore, and in general how it most likelymigrated and became trapped (next page).What they did not know—and what theysought during the interpretation—was infor-mation on the three-dimensional structureof the deepwater Congo sediments— specif-ically the location, general distribution andsize of likely hydrocarbon prospects. Theirimmediate goal was to identify target areassuitable for detailed mapping with moreclosely spaced surveys.

We rejoin our interpreters as they inspectthe survey’s processed and migrated sec-

tions and begin the critical task of identify-ing formation tops—the lengthy data pro-cessing was previously accomplished at aGECO-PRAKLA data processing center.3

Increasingly today, the interpreter’s maintool is the workstation. Workstations offermany advantages over traditional eyeballingalong paper sections. They offer versatile

17

Phase

Present day shelf and costal plain

Sub

side

nce

Late

drif

tE

arly

drif

tE

arly

rift

Late

rift

Pre

–rift

NESWAgeEpochPeriod

Maastrichtian

Campanian

SantonianConiacian

Miocene

Oligocene

Eocene

Paleocene

TuronianCenomanian

Albian

Aptian

Barremian

Hauterivian

Valanginian

Berriasian

Pal

eoge

neN

eoge

ne

Tert

iary

Time

Quat. PleistocenePliocene

Olig

ocen

e er

osio

n

Cre

tace

ous

Ear

lyLa

te

Jurassic

Permian

P∈

Oil testedPossible play

Trans.

140

130

120

110

100

90

80

70

60

50

40

30

20

10

Ma

Marnes Noires

nStratigraphic summary of the Congo sedimentary basin along a southwest-northeasttransect. Older rift sediments are clearly separated from younger drift sediments by theAptian sandstone and salt formations. The key hydrocarbon source rocks are LowerCretaceous Marnes Noires below the Aptian. Proven oil shows are indicated by blackdots, possible plays by open circles.

18 Oilfield Review

data manipulation, real-time processing ofattributes—computed parameters that helpinterpretation—and faster map generation.Paper sections may still used for two-dimen-sional (2D) surveys with exceptionally longsections separated by great distances.4

The first task always is to review all thedata, checking that intersecting linestie—meaning that the same reflectorsappear at the point of intersection at thesame two-way time on both sections. This isperformed by comparing reflector patternsat the point of intersection. Scanning thesections, the interpreter will also note majorstructural features such as listric faulting andsalt bodies, and begin the highly skilled taskof constructing a picture of the subsurface,all the time drawing and updating conclu-sions on a working map. Perhaps the inter-preter’s most cherished skill is this ability tovisualize in three dimensions.

The basis of structural seismic interpreta-tion is the loop method, in which a seismicreflector representing a geological horizon ismapped around a series of intersecting sec-tions and then back to the original section(next page). If the reflector checks againstitself, then all is well. But if the reflectorarrives back higher or lower than its originalposition, then an error has crept in and everystep around the loop must be checked. Clos-ing the loop is easier said than done. Tracingthe continuity of a reflector can be trickyacross a fault and sometimes impossible ifthe formation pinches out laterally or hasbeen eroded to form an unconformity. Prob-lems may also occur in areas of steep dip,where 2D migration on each section fails toimage complex three-dimensional structurecorrectly, producing a slight mistie.

Once all major reflectors have been suc-cessfully mapped around one loop, the inter-preter then moves to an adjacent loop andbegins again. Mapping reflectors is completeonly after all loops have been analyzed andfound to tie with each other. The startingpoint for this exercise is always where theinterpreter feels most confident, generallywhere geological control is available fromlog data. Well logs are measured versusdepth, so they must be rescaled versus two-way seismic time using check-shot data.

Typically, the interpreter plots the majorgeological units from the logs on the seismicsection at the well location to obtain thebest correlation. Lithology changesobserved on the logs assist the correlation

nPrinciple of the loop method for seismicstructural interpretation—area shown isdetail from northwest corner of surveymap. Within a given loop, a geologicalhorizon is mapped on one section until itmeets an intersecting section. It is thenmapped on the new section until it meetsanother section. Once around the loop,the horizon must map onto itself. If it doesnot, every step around the loop must bere-examined for error. The interpreter thenmoves to an adjacent loop. Interpretationconfidence builds as more and moreloops check out for internal consistency.

19January 1993

4. For seismic interpretation:McQuillin R, Bacon M and Barclay W: An Introduc-tion to Seismic Interpretation. London, England: Gra-ham and Trotman Limited, 1979.

5. This follows the SEG “normal polarity” convention. Inthe “reverse polarity” convention, slow formation overfast formation produces a black peak.

6. Bell DG, Selnes H, Bjoroy M, Grogan P, Kileyni T andTrayner P: “Better Prospect Evaluation with OrganicGeochemistry, Biostratigraphy and Seismics,” OilfieldReview 2, no. 1 (January 1990): 24-42.

process. For example, a slow formationsuch as mudstone overlying a fast formationsuch as a tight limestone typically correlateswith a white band or trough on the seismicsection.5 Conversely, a fast formation over-lying a slow formation generally correlateswith a black band or peak. Once the bestcorrelation is found, the interpreter colorsthe section at the well location according tothe geological units on the log.

Our first example section from the survey(next page, top) intersects two wells. A seaof black and white traces, it awaits the inter-preter’s color coding. Interpretation tech-niques vary, but the result is little different.

Some interpreters identify all faults on thesection (red markings) and then track reflec-tor continuity. Others may concentrate onreflectors and reflector terminations—onlapsand truncations— concentrating on thestratigraphy and marking in only those faultsthat bear on the work at hand.

In this case, the interpreter marks in themain faults and then concentrates on reflec-tor continuity between the two wells. Later,as the picture between the wells clarifies,deeper faults are marked in and reflectorcontinuity mapped away from the wells,southwestward—left on the section—toreach a line intersection.

To the layperson, some of these markingsmay appear hard to follow—for example,the continuity across deep fault blocksbetween the wells (next page, bottom). Theinterpreter, however, brings to bear varioustricks of the trade and an abundance ofexperience in outcrop and subsurface geol-ogy. The layering that varies so dramaticallyon either side of the deep faults is surmisedfrom the rather consistent pattern of nearby,parallel, high-amplitude reflections. Moreinsight into fault interpretation can comefrom laboratory experiments that simulatemillions of years of basin evolution—see“Analog Modeling of Basin Evolution,” page24.6

At each line intersection, formation topsare transferred to the new section (page21)—in this case running in a northwest-southeast direction—and the interpretationprocess moves toward another line intersec-tion. And so on around the loop. Then theinterpreter moves to an adjacent loop, andthe process is repeated. Once all loops havebeen interpreted, the sections can be col-ored to highlight the geology (pages 22 and23, top).

In both example sections, the thin Aptiansand (orange) forms a boundary betweenthe rift sediments below and the drift sedi-ments above. Overlying the sandstone, thesalt (purple) intrudes into the overlying sedi-ments, creating structural traps for oil gener-ated in the Marnes Noires below (brown).

The first section that lies parallel to thedirection of geological dip clearly showsfault blocks in the deep rift sediments andlarge-scale listric faults in the shallower driftsequences. A listric fault has a pronouncedcurved slip face. In this example, the sedi-ments to the left (southwest) of the faulthave been displaced downward and rotatedclockwise. The second section that parallelsthe coast appears simpler, but is no less

(continued on page 22)

nTop: a southwest-northeast section along the regional dip line,tying with two wells 13.5 km [8.4 miles] apart. First step was cor-relating the seismic section to data from the two wells. Next,major faulting and formation tops were interpreted between thewells.

nBottom: the same section with tentative interpretationextended toward intersection with orthogonal strike line,marked by red arrow.

20 Oilfield Review

NE

SW

0 5km

nThe strike section, oriented northwest-southeast. Following correlation at the point ofintersection with the previous dip section (red arrow), interpretation is continued towardanother intersection to the left (blue arrow). The sedimentary layering is more uniformthan in the previous section because the orientation is parallel to the coast, perpendic-ular to direction of greatest geological dip.

21January 1993

NE

SW

0 5km

22 Oilfield Review

nCompleted interpretation of the previously illustrated sections. The next step in the interpretation process is to digitize two-waytimes to the formation tops and create maps for each formation top.

7. For a review of magnetic and gravity techniques:Dobrin MB and Savit CH: Introduction to Geophysi-cal Prospecting, 4th ed. New York, New York, USA:McGraw-Hill Book Company, 1988.

NW SWSE

important. Here, it is vital to determine thattraps apparent on the previous section showfour-way closure—that is, they truly form atrap for hydrocarbons. On this example, thecomplex structure around the well at the farright is probably due to listric faulting seenon the previous section.

The interpretation seems clean and fin-ished, yet questions often remain. Forma-tion tops may be uncertain, particularly inthe deeper section beyond the range of wellcontrol and where seismic data lose theirresolution. The exact shape of the all-important salt intrusions may be subject todifferent interpretations. A solution forresolving these cases lies with magneticand gravity data, acquired concurrentlywith the seismic data.7

The technique for both data types is tocreate a 2D model of the subsurface fromthe seismic interpretation, compute themagnetic and gravity responses at the sur-face and then compare with actual data.The position of formation tops and faults inthe model and perhaps also the subsurface’svarying magnetic susceptibility and densityare then adjusted to ensure the best possiblematch. Adjustments to the model geometrymust be carefully cross-checked with theseismic data. Initial estimates for rock sus-ceptibility and density come from laboratorymeasurements and logs.

The usual strategy is to work upward(next page, bottom). Interpreters start withthe magnetic data since metamorphic andigneous basement rocks commonly domi-nate the magnetic response—practically allsedimentary rock is nonmagnetic. Excep-tions are volcanics that are interbedded withsedimentary formations—a crucial point inthe North Sea, for example, where volcanicsand volcaniclastic sediments can be interca-lated in reservoir horizons. In West Africa,

however, shallow volcanics are not impor-tant, so the magnetic data can be usedexclusively to fine-tune the structure anddepth of the crystalline basement.

Interpreters next turn to the gravity data,which are influenced by the entire geologi-cal section, and use them to constrain inter-pretations of the deep sedimentary structure.For the Congo survey, gravity firmed up thestructural interpretation of the rift-stage sedi-ments—that is, the Upper Jurassic andLower Cretaceous deposits, including theMarnes Noires, resting on the basement butbeneath the Aptian salt. Gravity modelingwas also critical in defining the geometryand distribution of the salt structures. Butthe original seismic interpretation wasassumed to provide a reliable picture of theyounger formations, which for the Congosurvey meant all formations above the salt.

Obtaining a consistent interpretation onall sections and fine-tuning that with mag-netic and gravity data took the GECO-

(continued on page 26)

23January 1993

nMagnetic and gravity modeling in two dimensions for fine-tuning the seismic interpretation. Magnetic modeling is first usedto constrain the interpretation of the deep igneous and metamorphic basement structure. Then gravity modeling is used torefine interpretations of the deep rift sediments and salt intrusions. (Courtesy of ARK Geophysics Limited.)

0

2000

6000

-30

30

-10

10

.013

.01 .009 .01

.008

4000

10 20 30Distance, km

Dep

th, m

Mag

netic

Fie

ld, n

T

Magnetic

SW NE

2000

4000

6000

0

2.7

2.5

2.72.7

2.55

2.152.57 2.572.42.6

2.35

2.5

2.72

2.152.57

-10

0

10

10 20 30Distance, km

Dep

th, m

Gra

vity

Fie

ld, m

Gal

Gravity

NESW

NE

Ken McClay

Royal Holloway, University of LondonEgham, England

nSchematic of a deforma-tion apparatus used tosimulate basin evolution.

Plastic detachment

Rigid footwall

10 c

m

Deformable hanging wall

Analog Modeling of Basin Evolution

Searching for plays depends on a deep under-

standing of how sedimentary basins form, then

tying that understanding to the available data,

which is usually seismic. While there is no sub-

stitute for first-hand geologic experience, experi-

mental tools have been developed that shed light

on basin evolution and complement the geolo-

gist’s field-based knowledge. One such tool can

simulate the complete life of a basin in less than

an hour. Developed during the last ten years at

Royal Holloway, University of London, England,

the analog basin modeling apparatus described

here was initially sponsored by BP. Funding now

comes from a consortium of six oil companies.1

The simulation takes place inside a long box

comprising two ends—one fixed at the left, repre-

senting the upthrown or footwall side of the fault,

and one movable in the horizontal direction, rep-

resenting the upthrown or hanging-wall

side—and two glass sides to allow observation

and recording of what happens inside (above).2 A

membrane runs along the bottom of the box and

is attached to the hanging wall moving with it.

With the hanging wall positioned about midway

along the length of the box, the experimenter fills

the box with horizontal layers of alternately col-

ored sand particles simulating layers of sedimen-

tary rock. Mica can be substituted to model

anisotropic formations, and clay can be substi-

tuted to achieve a different composition. Typi-

cally, the sand layers are 30 centimeters (cm) [12

in.] in length and10 cm [4 in.] in depth, equiva-

lent to a sedimentary section between 1 and 10

km [0.6 to 6 miles] thick.

1. ARCO British Limited, BRASOIL U.K. Ltd., BP Exploration,Conoco (U.K.) Limited, Mobil North Sea Limited and Sun OilBritain Limited.

2. McClay KR: “Extensional Fault Systems in Sedimentary Basins: AReview of Analogue Model Studies,” Marine and PetroleumGeology 7, no. 3 (August 1990): 206-233.

24

The hanging wall is then slowly moved right-

ward in simulation of extending the basin. The

sand particles subside and fault, as in real life,

and the experimenter records the events through

the glass side with time-lapse photography. As

the sand surface subsides, the experimenter

refills the depression with sand of a different

color, mimicking the accumulation of so-called

synrift sediments that are deposited during exten-

sion. In less than an hour, the complete history of

the basin is on film. At any stage, the action may

be stopped and frozen by impregnating the sand

box with epoxy and allowing it to set. Cuts of any

orientation can then be made through the basin to

examine structure and fault pattern deformation.

To simulate different geological conditions, the

footwall can be constructed in a variety of

shapes: vertical to observe plane extension, dif-

ferent molds in the shape of curved or listric fault

surfaces to observe lower angle fault detach-

ment. The membrane may be plastic, or rubber,

or a combination of both. Plastic conveys the

hanging wall’s lateral movement to all the sand,

while rubber spreads the extension

uniformly—these configurations mimic differing

structural styles in basin development. And the

bottom of the box can be inclined to simulate lat-

eral variations in basin thickness.

The sequence of snapshots shows the develop-

ment of a rift basin formed by stretching sedi-

ments over a listric fault and basement inclined

downward at 10° (next page, left). First, a depres-

sion develops against the listric fault and is filled

with synrift sediments. Next, a V-shaped fault pat-

tern termed a crestal collapse zone appears, fol-

lowed by another farther to the right. Each of

these is bounded to the left by an accompanying

synthetic fault, meaning a fault with the same

displacement direction, and to the right by an

antithetic fault, meaning a fault with the opposite

displacement direction. The synthetic faults

rotate as extension proceeds becoming quite

curved or listric, while the antithetic fault

remains straight. Between the two crestal col-

lapse zones, a high-standing zone termed a horst

remains relatively unrotated.

As extension continues, faulting in the left col-

lapse zone becomes more complex and begins to

affect the synrift sediments. Because of their

shorter history, though, the synrift sediments are

less faulted than the rift sediments, a clue to the

interpreter trying to distinguish the synrift from

rift. Meanwhile at depth, the layers rotate dramat-

ically yet seem completely unfaulted. Since the

seismic record at depth is often noisy and hard to

interpret, this analog result may provide the inter-

Oilfield Review

nExtension of sedimentsup to 100% over a listricfault and basementinclined downward at 10°.Blue, white and black lay-ers are sand particles, sim-ulating rift sediments,sieved into the box beforeany extension. Red, whiteand black layers simulatingsynrift sediments aresieved in as extensioncauses a depression nearthe listric fault.

nSection from a GECO-PRAKLA nonproprietarysurvey shot in the Gulf ofMexico showing extremerotation above a listricfault—compare with thesimulation. The right of thesection is bounded by asalt intrusion.

January 1993

km0 5

preter with likely scenarios for fault geometries.

The validity of these experiments may be appreci-

ated by noting the similarity of the fully extended

model to an actual seismic section shot over listric

faulting in the Gulf of Mexico (above).

Over 400 analog simulations have been run at

Royal Holloway, University of London with a vari-

ety of geometries and sedimentary cover. Simula-

tions have been run in which a basin is first

extended, then compressed, resulting in uplift

and fault reversal, a process called basin inver-

sion. Other experiments have simulated strike-

slip faulting, in which a sedimentary basin is

sheared by opposing horizontal movements. Salt

intrusions have also been modeled—these exper-

iments help understand structural trapping.

25

26

nTwo examples of two-way time maps obtained from the structural seismic interpretation. Left: two-way time to the base of theAptian, showing the geological structureat the base of the Aptian salt.

Right: a two-way time thickness map,between the seabed and theMiocene—darker color at the bottom indi-cates greater thickness. Thickness maps,obtained by subtracting two-way times toconsecutive layers, provide an indicationof differential subsidence and its controlon sedimentation. These factors help pre-dict potential reservoir distribution, geom-etry and thickness.

Oil Window

0

2000

4000

6000

Plio./ Pleist.

Miocene

Paleocene

Cenomanian

Albian

TransitionLate RiftEarly Rift

Pre–rift

Dep

th, m

Time, millions of years before present140 120 100 80 60 40 20 0

Oil window

Ocean

N 10 0 km

PRAKLA interpreters most of three months.The next task, which took one month more,was to map two-way times to formation topsand assess plays. This would formalize theworking maps they had been constantlyupdating while interpreting the sections.

Mapping 2D seismic events is halfmechanical and half art and experience.The mechanical part, comprising digitizingthe location of formation tops and interpo-lating between sections, is greatly facilitatedby the workstation. But resolving ambigui-ties in the large areas between the sparselines of a 2D survey—for example decidingif a fault seen on one section is the same as,or different from, that seen on another—requires considerable geological expertise.Thus, interpreters generally take the com-puter-generated maps as a starting point andthen manually adjust the results to achieve ageologically valid interpretation of the struc-ture that is consistent with all available data.

Mapping not only elucidates structure butalso permits the creation of time-thicknessmaps, by subtracting the two-way times ofconsecutive geological tops (left). Knowingthe seismic velocities of the formations, it isthen possible to compute actual thickness.Thickness is one parameter that determinespotential reservoir size, a key factor in evalu-ating pay. Thickness is also important forreconstructing the sediments’ burial history.This describes how quickly the sedimentswere deposited and how each formation’stemperature evolved throughout geologicaltime. It then becomes straightforward tocheck whether formations deemed likely toform hydrocarbons, such as the MarnesNoires, actually experienced the requisitetemperature conditions (below, left).

Oilfield Review

nBurial histories for offshore Congo forma-tions based on one well’s lithologic col-umn. The oil window, indicating suitableconditions of temperature and time for oilgeneration, covers the rift sediments thatinclude the Marnes Noires, the most likelyoil source for the area. Temperature isderived from depth using a variable pale-otemperature gradient. The ocean (top)can be seen to enter the study area about130 million years ago, regress and thenre-enter permanently.

Salt-induced structures, Albian to Neogene

Salt-induced structures, Albian to Maastrichtian

Salt-induced structures, Albian to Turonian

Salt diapirs with crestal anticline or flank structure

Base Aptian structure

Rift section structure

Migrating Miocene channels

Tertiary graben collapse zone

Limit of surface outcrop

N

Pre-Aptian

Post-Aptian

Base Miocene UnconformitySW NE

Post-Aptian

Pre-Aptian

10 0 kmnSummary oflikely plays in partof the Congo off-shore survey, color-coded according togeological age.Schematics show-ing source, migra-tory path andreservoir trap forreservoirs in thepre- and post-Aptian sediments.Brown formationwith dots is theMarnes Noiressource rock.

At this point, all the factors necessary topinpoint plays are at hand. The burial his-tory confirms the potential of organic-richhorizons as source. Detailed comprehen-sion and mapping of basin structure revealslikely migratory paths and trapping mecha-nisms. The thickness maps indicate the dis-tribution and geometry of sediment bodiesand assist in recognizing commerciallyinteresting reservoirs. Deciding the locationof plays now demands of the interpreter ajuggling of these factors and the picking oflocations and depths where all indicationsappear simultaneously favorable. The resultis a play map that oil companies can use todecide on concession bidding, or perhaps todesign further seismic surveys with closerline spacings or even 3D surveys. The sameinterpretation concepts apply throughoutthe exploration phase and ultimately permitthe selection of favorable sites for explo-ration drill holes (above).

These first steps in the search for new dis-coveries occur years, sometimes even adecade or more, before hydrocarbons comeon stream and repay the exploration invest-ment. Just as an infant learns to walk—onestep at a time until confidence builds—sogeologists feel their way into frontier areas.They tread carefully at first, but then rapidlyadvance to the best areas as more informa-tion becomes available. In the search forplays, the geologist is king. —HE

27January 1993

28

For help in preparation of this article, thanks to AJBerkhout, Delft Technical University, The Netherlands;Phil Christie and Doug Miller, Schlumberger CambridgeResearch, Cambridge, England; Cengiz Esmersoy andLisa Stewart, Schlumberger-Doll Research, Ridgefield,Connecticut, USA; Kjell Erik Fagertun, Jan Farestveit,Gary Hutton, Ottar Sandvin and Alan Strudley, GECO-PRAKLA, Stavanger, Norway; Niels Kinneging, GECO-PRAKLA, Delft, The Netherlands; Phil Kitchenside, DizMacKewn and Richard Woods, GECO-PRAKLA, Orping-ton, England; Roland Marschall, GECO-PRAKLA, Han-nover, Germany; Terence Redshaw, BP-Statoil R&DAlliance, Trondheim, Norway.

Paul FarmerStavanger, Norway

Sam GrayDan WhitmoreAmoco Production CompanyTulsa, Oklahoma, USA

Graham HodgkissOrpington, UK

Andy PieprzakHouston, Texas, USA

Davis RatcliffAmoco Production CompanyHouston, Texas, USA

David WhitcombeBP ExplorationHouston, Texas, USA

Structural Imaging: Toward a Sharper Subsurface View

S E I S M I C S

You’re trying to fix the location of a star using a telescope with optics

that distort the image. Sometimes you can’t see the star, but when you

can, it might look fuzzy or appear to be the wrong size or in the wrong

place. Structural imaging is like fitting the telescope with lenses that

bring the image into crisp focus at the right location.

Ask a geophysicist for a simple definition ofstructural imaging and you might get ananalogy like this, echoing the parallelbetween optics and acoustics. But in directterms, structural imaging boils down to this:the branch of seismology in which pro-cessed seismic data undergo additionalpasses to create a large-scale picture of thesubsurface. Its goal is to provide a map tolocate traps and plan a drilling strategy foroptimal drainage.1

Structural imaging lays the foundation forother seismic techniques that investigateprogressively smaller features (pages 30-31). After structural imaging, seismic stratig-raphy jumps to the next level of detail,characterizing the arrangement of layerswithin rock formations (see “SequenceStratigraphy—A Global Theory for LocalSuccess,” page 51). Next, lithostratigraphicinversion attempts to describe lithology ofindividual rock layers and evaluate proper-ties and distribution of pore fluids, throughanalysis of variation of seismic signal

1. For a general review of the literature:Fagin SW: Seismic Modeling of Geologic Structures:Applications to Exploration Problems: Tulsa, Okla-homa, USA: Society of Exploration Geophysicists,1991.French WS: “Practical Seismic Imaging,” The LeadingEdge 9 (August 1990): 13-20.Johnson JD: “Structural Imaging in the Real World,”The Leading Edge 11 (January 1992): 32-36.Redshaw T and Lasseter L: Seismic MigrationExplained! and Seismic Migration Defined! Houston,Texas, USA: BP Complex Structures Group, 1991.

amplitude with spacing between source andreceiver, called offset (see “HydrocarbonDetection With AVO,” page 42). The qualityof these finer-scale techniques rests largelyon the quality of structural imaging.

Today, structural imaging is advancing ontwo fronts. One is improving image qualityof conventional structures—their positionand shape become known with greateraccuracy. The other is the ability to imageareas of more complex structure associatedwith large, rapid changes in velocity. Exam-ples are low-velocity layers, structure belowsalt or gas, or multiply folded and faultedformations. Understanding of these difficultsettings promises to better quantify reservesin established fields and help define newprospects that eluded more conventionalapproaches.

In all but the simplest geologic settings,imaging with seismic energy has three fun-damental problems: the image starts outblurry, has the wrong shape and is in thewrong place (next page). These problems are

(continued on page 31)

Oilfield Review

Sheriff RE: A First Course in Geophysical Explorationand Interpretation. Boston, Massachusetts, USA: Inter-national Human Resources Development Corpora-tion, 1978.Yilmaz Ö: Seismic Data Processing. Tulsa, Oklahoma,USA: Society of Exploration Geophysicists, 1991.For a review of processing:Boreham D, Kingston J, Shaw P and van Zeelst J: “3DMarine Seismic Data Processing,” Oilfield Review 3,no. 1 (January 1991): 41-55.

January 1993

Raypath

330 350 370

330 350 370

800

1000

1.6

2.0

1.6

2.0

Dep

th, m

Tim

e, s

ecTi

me,

sec

Distance, mTheory

Stack section

After migration

330 350 370

End of reflector foundby following raypath

back to surface

Real end ofreflector

Smearing of eventaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaReflected rays

Zero-offsetstack section

Migrated section

Geophone

Bowtie

Event plottedvertically belowwhere reflectionis recorded

Tim

eD

epth

Tim

e

nThree common curses of unmi-grated seismic data. Events thatend abruptly (left), such as a fault,may not be sharply imaged bystacking because energy at the ter-mination of the event is diffracted,smearing the image over manytraces (inset above). Migration col-lapses these diffractions to sharplyrender the reflector. aaaaaaaaaaaaDepthDepth

Dep

th

Reflection recordedbetween

these points

Acquisition geometry

Dipping reflector in constant

velocity earth

Recorded reflector misplaced

Actual reflector

Correctly positioned

migrated reflector

Midpoint

As shot-receiverseparation increases,reflection point movesupdip

After stacking

After migration

A syncline may appear as a “bowtie”on the stacked section (above). Migra-tion moves the events so the sectionlooks like a syncline. Dipping reflec-tors (below, left) present several prob-lems: the reflector is misplaced later-ally and vertically, it appears longerthan the actual reflector and dips lesssteeply. The steeper the dip, and morecomplex the velocity field, the moredifficult the solution to these problems.

The inset (above) shows that reflec-tion points move updip because dif-ferent offsets do not image the samepoint. Stacking them will thereforemove the reflection point updip. Thisis corrected with dip moveout(DMO), a simple type of migration—sometimes considered to be “premi-gration”—that transforms data soeach common midpoint gather con-tains reflections from the samedepth point. (After Redshaw and Lasseter,reference 1. Inset after Sheriff, reference 1.)

29

30 Oilfield Review

Seismic Data Processing

ACQUISITION

PreprocessingQuantify survey geometry, select CMP intervals for velocity analysis

(2 to 5 km)

Optimization of Processing ParametersGain removal and amplitude correction, statics correction, mute test,deconvolution, noise suppression, multiple attenuation, matching filter

Conventional Processing(NMO, DMO correction, etc.) (CMP stack, etc.)

Prestack time migration Poststack time migration

Interpretation of important velocity interfaces

Sei

smic

com

pan

y an

d o

il co

mp

any

pro

cess

ing

grou

ps

Inte

rpre

tatio

nd

epar

tmen

t

Macro modelVelocity verification: check ofwavefronts using layer stripping insubset of data.

Structural ImagingProvides large-scale arrangement of velocity layers

Seismic Stratigraphic InterpretationDisplays section in depth and time; provides detail in each velocity layer& qualitative evaluation of depositional environment

Lithostratigraphic Inversion and AVOProvides elastic parameters, porosity, density, fluid type

PrestackPoststack

DEPTH MIGRATION

Sei

smic

pro

cess

ing

team

and

oi

l com

pan

y q

ualit

yco

ntro

l/int

erp

reta

tion

team

s

Sei

smic

pro

cess

ing

and

oil

com

pan

y in

terp

reta

tion

team

s

Incr

easi

ng d

etai

l

Conversion of time maps to depth mapsusing “apparent velocities” (depth stretch)

Image-ray depth conversion

T

I

M

E

D

E

P

T

H

2. For a classic work on time migration:Hubral P: “Time Migration—Some Ray TheoreticalAspects,” Geophysical Prospecting 25 (1977): 738-745.

caused mainly by refraction—bending ofrays as they pass through rocks of differentvelocity2 —and diffraction of seismic energyas it passes through rocks of different veloci-ties, shapes and thicknesses. To make theimage interpretable, seismic energy must befocused into a sharp, correctly shapedimage, and the image moved to the correctlateral and vertical position. The sharper theimage of a structure and truer its shape andposition, the more accurately the structurecan be evaluated and drilled. The method ofsharpening, shaping and relocating imagesis loosely termed migration, often consid-ered synonymous with structural imaging.

Mathematically, migration is performedby various solutions to the wave equationthat describe the passage of sound throughrock. Called migration types, these numer-ous solutions or algorithms often take thename of their authors (Gazdag, Stolt) or thetype of solution (finite difference, integral).Migration types may be thought of as a fam-ily of tools, each with shortcomings andadvantages. Choice of the optimal type isnot always obvious and relies on the experi-ence of the seismic practitioner.

January 1993

nSteps in seismicdata processing andinterpretation. Struc-tural imaging is aprecursor to othermethods of seismicinvestigation.

Velocity m

Types of migration are applied in abroader category of migration called class:poststack or prestack, two-dimensional orthree-dimensional (2D or 3D), time ordepth. These classes form eight possiblecombinations. The trend in imaging is frompoststack to prestack, 2D to 3D and time todepth migration. This trend is best under-stood by examining the strengths and weak-nesses of these classes.

Poststack migration, still the most com-mon form, assumes the section built ofstacked traces is equivalent to a zero-offsetsection, meaning each trace is made as ifthe source and receiver are coincident (see“Seismic Basics,” next page). Chief advan-tages of poststack migration derive fromstacking: compression of data, removal ofmultiples and other noise, and fast, inexpen-sive processing. Poststack migration holdsup even in fairly strong lateral velocity vari-ation, but at some level of velocity variation,stacking breaks down and prestack process-ing is required. A limitation of stacked datais its removal of true amplitude information.

Prestack imaging is done on unstackedtraces, taking 60 to 120 times longer than

Velocity Model Verification

Yes

odel satisfactory

No

Compare depth migration with inputvelocity model, using full data set

Image-ray depth conversion

Initial macro model

Depth migration

Match?

poststack imaging, but with the potential toretain amplitude variation with offset (AVO)and phase changes useful for later analysis(page 33, top). Prestack time migration ispreferred when two or more events occur atthe same time but with different stackingvelocities (page 33, bottom). Prestack depthmigration is advantageous when velocitiesin the overburden or the target are complex,but requires large computer resources andremains rare. However, as massively paral-lel computers become more widely avail-able, migration technology will shift towardprestack depth migration.

In 2D migration, energy reflected only inthe plane of the section is correctly imaged,whereas 3D migration uses energy fromboth in and out of the plane of the section.In general, 3D migration will have higherresolution because it can move energy fromoutside the plane back to its correct position.The cost of higher resolution, however, is

(continued on page 34)

31

Adjust macro model 2D prestack migration: revise commonimage point so traces are flat3D poststack migration: revise model tofocus energy on boundary, converging theimage section and the model.

32 Oilfield Review

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaOffset

Shot #

Two-

way

tim

e

1 2 3

Offset

Shot #

Two-

way

tim

e

1 2 3

Offset

Shot #1 2 3

CMP gather

Stackingvelocity

NMO correctedCMP gather

StackedCMP

+ =

+ + =

1

2

3

CMP

Shot #

nBasic seismic acquisition geometry and processingfor a flat reflector and uniform velocity. For simplicity,only three receivers are shown, but there may be hun-dreds of receivers. The green ray always bounces offthe same point, which is always halfway betweensource and receiver. The green rays are said to have acommon midpoint (CMP), although they have differ-ent distances—or offsets—between source andreceiver. Traces sorted to combine reflections from thesame midpoint are called CMP gathers. Reflectionscome later in time for farther offsets. This time delayis called normal moveout (NMO) and is corrected byusing stacking velocity. Finally, these traces arestacked into a single trace. Stacked traces from hun-dreds or thousands of CMP gathers are used to createa CMP section. (After Redshaw and Lasseter, reference 1.)

nPrinciple of stackingvelocity calculation.Zero-offset velocity isdetermined as if thesource and receiverwere at the same point.aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaOffset x Zero offset

1. Sometimes erroneously called root-mean-square (RMS)velocity, because it is the root-mean-square average ofvelocities in a model of parallel horizontal layers. Stackingvelocity approaches RMS velocity as offset approacheszero only if layers and reflectors are parallel and layers areisotropic. Seismics often has many names for the samething and confusion may arise because different mean-ings for the same words are not clearly communicated.

The basic principle in seismic surveying is that

sound is reflected at an interface of materials that

have sufficiently different acoustic properties.

Reflectivity is determined by density and acoustic

velocity, which vary greatly with lithology and less

so with pore fluid composition. Seismic reflec-

tions, therefore, are used mainly to map the shape

and position of formation boundaries and some-

times to identify fluid type.

Sound energy is generated at sea by an explo-

sive or a burst of air bubbles from an air gun, and

on land usually by an explosive charge or a vibra-

tor. Sound reflected back to the surface is detected

and converted into an electrical impulse and dis-

played as a seismic trace. This is a curve in which

the vertical axis is two-way travel time (from sur-

face to reflector and back to surface) and the hori-

zontal axis represents distance from the source.

The position of reflectors in time is represented as

wiggles in the trace, with wiggle amplitude repre-

senting energy. Once the rock velocity is known,

time measurements can be converted to depth.

Each trace is made from one recording of a

source-receiver pair in one position. For efficiency,

traces from one source are recorded at several

receivers simultaneously. The source and

receivers are moved a certain distance along the

survey line and another recording made. For imag-

ing, traces are usually sorted into a collection

called a common midpoint gather, in which all

traces have a common midpoint. Each gather is

often summed into a stack, a single trace averag-

ing wiggles of the entire gather. Stacking com-

presses data into a more manageable volume and

increases signal-to-noise ratio. Stacked traces from

many gathers are grouped into a stacked section—

Seismic Basics

a 2D slice of the earth in which the vertical axis is

time. Each trace is located midway between the

source and receiver, called the midpoint.

The goal is to make the stacked section appear

similar to the geologic cross section. Take the sim-

ple case of one source and three receivers over a

flat reflector (top). The distance between source

and receiver—called offset—increases with

receiver number. As offset increases, so does the

time for sound to travel through the earth to the

receiver. Wiggles at increasing offsets therefore

appear later in time, a phenomenon called move-

out. The wiggles form a hyperbola, making a flat

reflector appear to dip and curve. This is called

hyperbolic moveout. How can the wiggles be

realigned so that when they are stacked they indi-

33January 1993

nHow more than one event can arrive at the same point on thesurface from different points in the subsurface. Two reflectionsarrive at the same time from the same horizon (left). Standardnormal moveout (NMO) corrections will have stacking velocityconflicts because stacking velocity varies with dip. The solutionis dip moveout (DMO) correction (see pages 30-31). At right, tworeflections arrive at the same time from different horizons. Nei-ther DMO nor NMO will work. Prestack time migration is thesolution. (After Redshaw and Lasseter, reference 1.)

nComparison of 2Dpoststack andprestack depthmigration, with themacro model ofvelocity layers, forthe Ouachita for-mation (Texas/Oklahoma, USA).The prestack depthmigration providesa more easily inter-pretable image inthe middle of theinterval.

Tim

e, s

ecD

epth

, ft

Macro model

Poststack depth migration

Prestack

0.05

0.35

0.65

1.0

0

5000

15000

10000

Distance, ft

Dep

th, f

t

0 30000

5000

15000

1000 2000

10000

1650

0

1350

0

1430

0

1600012000

14000

1600018500

17000

16000

17000

18000

19000ft/sec

Two reflections arrive at the sametime from the same horizon

Two reflections arrive at the sametime from different horizons

DMO required Prestack time migration required

cate the true shape of the reflector? They are lined up

using stacking velocity.1 This can be calculated as:

Distance from zero offset to offset X

[(Travel time at offset X)2 - (travel time at zero offset)2]1/2

Applying the Pythagorean theorem to the result

provides the (essentially horizontal) velocity of

rock between zero offset and offset X (previous

page, bottom).

Stacking velocity is the beginning rather than

the end of the quest for velocity, ultimately needed

to convert time measurements to depth. As a first

approximation of the velocity of rock above the

reflector, stacking velocity allows reflections to be

moved near their true positions. This approxima-

tion begins to fail as the earth grows more com-

plex—many reflectors, dipping beds, faults,

strongly folded or tilted reflectors or lateral or ver-

tical variation in velocity between reflectors. Nev-

ertheless, stacking velocity provides a foundation

for more sophisticated methods.

34 Oilfield Review

nVisualizing a North Sea salt diapir struc-ture with 2D time migration, depth migra-tion and depth migration output in time.Colors denote the interval velocity fielddetermined prior to depth migration. Thevelocity model permits a more certainmigration by including lateral variations.In the poststack time-migrated section,reflectors are poorly defined, on both theflanks and base of the diapir. Definitionimproves in the prestack depth migration,performed with the velocity field shown.Finally, for comparison between theprestack depth and time migrations, thedepth migration output is converted backto time, revealing a marked improvementin definition of the diapir flanks and base.

8600 8700 8800 8900 9000 91000

1

2

3

4

0

1

2

3

4

0

1

2

3

Tim

e, s

ecD

epth

, km

Tim

e, s

ec

Shot points (25 meters between points)

4900 3200 1500Velocity, meters/sec

Original poststack time migration

Prestack depth migration

Prestack migrated depth-to-time conversion

0 meters 500

greater acquisition costs with longer pro-cessing time—what takes days in 2D mighttake weeks in 3D.

The choice of 2D or 3D migration is firstdetermined by acquisition geometry. Dataacquired with a 2D scheme—a singleacquisition line with shot and receivers in aline—can only be 2D migrated, but 3D datacan undergo 2D or 3D migration. An inex-pensive, fast approximation to 3D migrationis 2D migration in orthogonal directions,called two-pass 3D migration. It is strictlycorrect only for a constant-velocity earth,but errors are small if the vertical velocitygradient and dip angle are small.

Time and depth migration differ on sev-eral levels. In simplest terms, time migrationlocates reflectors in two-way traveltime—from the surface to the reflector andback as measured along the image ray—whereas depth migration locates reflectorsin depth. A migrated seismic section with atime axis, however, is not necessarily a timemigration. A depth migration may be con-verted to time. This is sometimes done tocompare velocity modeling for depth migra-tion with velocity assumptions used for timemigration (right).

The significant difference between timeand depth migration is the detail with whichthey view the behavior of sound in theearth. To time migration, the earth is simplein both structure and velocity; to depthmigration, it may be complex. Time migra-tion assumes negligible lateral variation invelocity and therefore hyperbolic moveout.Using only stacking velocity, or someapproximation, time migration can makesharp and correctly shaped and positionedimages—if structure and velocity are gener-ally simple. Time migration can handlesome complexity of structure, but only lim-ited variation in velocity.

When velocity and structure becomeobviously complex, rays are bent, produc-ing nonhyperbolic arrival times and distort-ing the results of time migration (next page,above). Reflectors become blurry, move to

nA general scheme for selection of migration class.(From Redshaw and Lasseter, reference 1.)

Simple velocities + simplestructure = poststack time migration

Simple velocities + complexstructure = prestack time migration

Complex velocities + simplestructure = poststack depth migration

Complex velocities + complexstructure = prestack depth migration

Incr

easi

ng v

eloc

ity

35January 1993

nA salt flankconundrum in theBarents Sea. Thepoststack timemigration(stretched todepth with raytracing) showstwo weaknessescommon to thetechnique: breakupof near-verticalreflectors and,marked by thelower oval, “pushdown” of reflec-tors due to thelarge lateralvelocity contrastbetween rock(1800 to 3000m/sec) and salt(4000 m/sec). Agood interpreterwill spot this fea-ture as a result ofthe reflectorbeing under thelip of the salt. Inthis case, thenext step waspoststack timemigration. Thebroken reflectoron the upper rightflank is holdingtogether a littlebetter. Below, thepush-down effectis reduced, butthe reflector isstill somewhatbroken. Finally, aprestack depthmigration clearlyimages reflectorsflanking the salt.

3. Berkhout AJ: “SEG Annual Meeting ’92: A SuccessfulMeeting Characterised by Progress and Uncertainty,”Journal of Seismic Exploration 1 (November 1992):311-314. Berkhout notes that 50 of 400 papers con-cerned migration, most of them depth migration.

4. May BT and Covey JD, “Structural Inversion of SaltDome Flanks,” Geophysics 48 (August 1983): 1039-1050.

Dep

th, k

mD

epth

, km

Dep

th, k

m

Prestack depth migration0

1

2

3

4

0

1

2

3

4

0

1

2

3

4

Poststack time migration

Poststack time migration

Breakup ofnear-verticalreflectors

“Push-down”effect

Near verticalreflectorsimaged better

Push-downeffect lesspronounced

Accurateimagingof near-verticalreflectors

Push-downeffect resolved

the wrong place or become too long.Accounting for this ray bending requireswhat is called a macro model—a model ofvelocities between reflectors. This is neededmainly to eliminate lateral positioning errorcaused by refraction, but also to sharpen theimage. Construction, revision and verifica-tion of this macro model are the goals ofdepth migration, and are the main contribu-tors to its difficulty and cost. Depth migra-tion is also more sensitive to errors in veloc-ity than time migration.

Although time migration handles onlysimple problems exactly, it remains thedominant technique in exploration, andusually lays the foundation for the depthmigration macro model. In the productionsetting, some operators prefer to stretch timemigration to the limit before jumping todepth migration (see “Misconceptions inStructural Imaging,” page 39). Still, depthmigration has become a valuable toolbecause it is the only one that can handlethe most difficult problem in imaging—strong, rapidly varying lateral changes invelocity. Depth migration also remains thefocus of most imaging research.3

When is depth migration needed? Or, toput it another way: When is the velocityfield complex enough to mislead timemigration? For many operators, a stepbefore depth migration is a processing stepto convert the time-migrated section todepth using image-ray depth conversion.This procedure uses an image ray, which isshot downward perpendicular to the surfaceand is bent by an amount predicted bySnell’s law applied to the velocity model.The ray passes through the correct lateralposition of an event, which in the timemigration would appear vertically below thestarting point of the ray. If the image raystrikes the reflector a considerable lateraldistance from the starting point, velocityvariation may be interpreted to be sufficientto warrant depth migration.

In many areas, the decision to use depthmigration usually arises when imaging thenear-vertical flanks of salt domes (right). Inthis setting, BP tests the macro model withray tracing.4 If the depth or lateral positionof the image is not displaced far enough toaffect well placement, then BP does notbother with depth migration. To preserve

36 Oilfield Review

5. Sugrue MJ and Rockliff DC: “Depth Migration ofDiapirs from the Central Graben of the North Sea,”Technical Programme and Abstracts of Papers, 53rdEAEG Meeting and Technical Exhibition, Florence,Italy (May 26-30, 1991): 100-101.

6. Marschall R: “Macro Model Optimization and Depth-Migration,” Expanded Abstracts, 2nd InternationalCongress of the Brazilian Geophysical Society, Sal-vador-Bahia (October 27–November 1, 1991): 944-947.

nInterval velocities(top) correspondingto the time slice(bottom) developedusing GECO-PRAKLA’s VELMOD3D ray-tracing andvelocity inversionprogram. The topfigure shows thetopography of thehorizon containingthree salt domes.Color variation rep-resents computedinterval velocities.

U K North SeaGlobal Depth Horizon and Velocity FieldLayer 4

Column

Row

Dep

th, k

mIn

terv

al v

eloc

itym

/sec

0

2

4

300

500

700

900

2000

1500

1000

500

2200

2500

7. Common image point gathers go by several othernames, among them coherency panels. See VersteegRJ: “Analyse du Problème de la Détermination duModèle de Vitesse pour l’Imagerie Seismique,” PhDthesis, University of Paris VII, 1991.

8. Stork C: “Making Depth Migration Work in ComplexStructures,” Expanded Abstracts, 62nd SEG AnnualInternational Meeting and Exposition, New Orleans,Louisiana, USA (October 25-29, 1992): 939-943.

steep-dip events, BP is careful not to finishprocessing with a low-cut filter (5 to 40 Hz).In removing noise, the filter may inadver-tently erase low-frequency, steep-dip events.5

Selection of the appropriate type and classof migration is only half the story of imag-ing, however, and probably the less impor-tant half. The main concern in imaging isthe engine that drives the depth migrationalgorithm—the velocity macro model.

The Macro Model“Imaging practitioners often state that wereally have a velocity problem, not a migra-tion problem.”

Jeff JohnsonManager of GeoscienceAmoco Production Company

The macro model is a numerical descriptionof the subsurface on the scale of hundredsof meters. It contains either two-way traveltime or depth to the main reflectors and thevelocities and densities between them. Itdescribes the acoustic propagation charac-teristics of the subsurface and is used fordepth migration to include ray bending. Inother words, the macro model functions asthe air traffic controller for the migrationalgorithm. It tells the algorithm how far tomove the reflector—up a little, down, to theleft or right, or hold that line. Inadequateknowledge of velocity results in the imagebeing under- or overmigrated, misplaced orblurred. Even the most advanced migrationalgorithm will fail to focus the image ifdirected by a flawed macro model.

Techniques of macro model building arecontroversial, proprietary and fast evolving.It remains the weak link in depth migration,so attention today focuses on increasinglysophisticated ways to model, update andverify velocity for depth migration.6 The costand computation time of these techniquesincrease with their capability to handlelarge, rapid changes in velocity.

A starting point is stacking velocity,obtained for conventional time processing.Stacking velocity is calculated from the dif-ference in arrival time at different offsets,assuming the layers over the reflector have aconstant velocity. Stacking velocity valuesare often inaccurate because they are anaverage over a large volume of rock, whichoften has nonuniform velocity. Stackingvelocities may be constrained with datafrom sonic logs or previous seismic data.Still, stacking velocities may provide thebest, first macro model.

A first-guess model of the earth is usuallydeveloped by picking main reflectors from a

Common image pointgathers lie on curve

Common image pointgathers lie on straight line

nComparison of common midpoint (CMP)and common image point (CIP) gathers.A CMP gather is made prestack andbefore migration. If events are nonhyper-bolic, stacking will not work and prestackprocessing is required. A CIP gather isalso made prestack, but after migration.The degree of curvature in the gatherindicates error in the velocity model.

nMacro modelupdate using cur-vature analysis ona prestack depthmigration. Velocitypicks in the sec-tion, V1 and V2, donot result in a flatevent, shown atleft. The degree ofcurvature in thegather is used toupdate the veloc-ity. After updating,the upper reflectoris flattened slightlyand the lowerreflector broughthigher in the sec-tion. Events in thecommon imagepoint gathers nowlie on a flat line,indicating that theupdate achieves acorrect macromodel.

CMP

CIP

Common midpoint gather

Common image point gather

poststack time migration. Times are assignedto each reflector and velocities to intervalsbetween reflectors (previous page). A first-order approximation is the computation ofconstant velocity values between reflectors.Now, workstations readily permit estimationof velocity gradients vertically betweenreflectors and sometimes horizontally alongthe event. The macro model may then beused to make a synthetic seismogram basedon part of the model, which is iterativelymodified until the synthetic matches themeasured seismic data.

From here, the main challenge is to find afast, accurate way to update and verify themodel, usually starting with a time-migrated section. Three approaches areused—curvature analysis, layer strippingand focusing analysis.

Curvature analysis can be done severalways. One approach uses so-called com-mon image point (CIP) gathers, which canbe made by selecting prestack data, depthmigrating it, then creating a section madefrom offsets that image the same subsurfacepoints7 (above, right). If the same event onall the traces lies on a flat line, then eachmigration has imaged the event at the samelocation and the macro model is correct. Ifthe same event on all the traces is curved,there is an inconsistency in the migration,indicating that a velocity variation needs tobe accounted for (right). The degree of cur-vature provides a measure of the velocityerror. However, formulas for updating veloc-ities are normally based on curvature of CIPgathers, which assume a horizontally lay-ered earth, or at best, slightly dipping layers.The method therefore breaks down in com-plex subsurface structures. Nevertheless,curvature analysis at least gives an indica-tion of the error, and a satisfactory result canusually be obtained in no more than four orfive iterations.

An alternative that is attractive mainlybecause it is free from dip limitation isadjustment to the macro model based onray tracing, determining the arrival time ofrays through the macro model and adjustingvelocity boundaries accordingly. The mainlimitation to this method is rapid variation inlateral velocity. Ray tracing works for bothtime and depth domains, whereas curvatureanalysis on image point gathers is limited tothe depth domain.8

Because curvature analyses may fail insettings of complex structure or large veloc-ity contrast, the time-migrated image maynot clearly show boundaries needed todevelop the macro model. An alternativewith 2D data is layer stripping with prestack

37January 1993

nMacro model building by layer stripping. Based on stacking velocities, thevelocity field in the upper section is chosen as 1500 m/sec, and migration per-formed. This gives the flat reflector at the top of the section. The velocity fieldbelow this reflector is then chosen as 2000 m/sec and it is stacked, revealing thesecond, more jagged boundary. The process is repeated to find the third boundary.

nPanels explaining focusing analysis, which indicate when the velocity model is cor-rect (left set) and when velocity is overestimated. (After Faye and Jeannot, reference 10.)

Offset

Real depthand focusing depth

0

Depth

Tim

e

Offset

Real depth

0

Depth

Tim

eFocusing depth

Migration depth

3000 m/sec

2000 m/sec

1500 m/sec

Third boundary

1500 m/sec

2000 m/sec

Second boundary

0

1

2

3

Dep

th, k

m

Iteration for boundary 2 Iteration for boundary 3

data (right ). In this procedure, the interpreterstarts with a picked interval velocity (oftenderived from stacking velocities) and amajor reflector near the top of the section.The section just below this reflector is depthmigrated with a sequence of trial intervalvelocities, one of which clarifies the nextvelocity boundary below. Then anothersequence of interval velocities is pickedfrom this boundary down and this newinterval is migrated. The process is repeated,picking interval velocity and migrating alayer at a time until the entire section iscomplete. Velocities can be verified usingmoveout analysis. The layer strippingapproach proved successful for Conoco inthe Marmousi Workshop, a migration con-test sponsored by Institut Français du Pét-role, in which eight oil companiesattempted to produce the best image possi-ble for a synthetic example.9

Focusing analysis attempts to determineerrors in the macro model for prestackmigration, usually 2D.10 The wave equationis used to show what a gather would looklike if the source and receiver were movedto the depth of the reflector. Under this con-dition, the macro model is correct if diffrac-tions collapse to a point—that is, if theyfocus. But if the model is wrong, the focusmay fall above or below the depth of thesource and receiver to the so-called focusdepth (right ). For a flat reflector, the realdepth of the reflector lies about halfwaybetween the focus depth and the migrationdepth with the provisional velocity model.This real depth is then used to update themacro model.

Opinion on focusing analysis is divided.On the plus side, the method has no dip lim-itation; maximum dip is determined by themigration algorithm. On the minus side,results are difficult to interpret. Complex sub-surface structure and gross error in themacro model create spurious focus energythat may mislead interpreters.

State-of-the-art velocity model buildinguses layer stripping with 3D rather than 2Dprestack data. This can take 10 to 100 timeslonger to compute than with 2D prestackdata, and so has limited application. Itsadvantage is its superior capability in han-dling complex velocity fields. Becauseprestack data are so large—even a smallsurvey of 40 square miles [100 square kilo-meters] can have 10 million traces— veloci-

38

ties are obtained along only a few lines ofdifferent azimuth, and often on short sectionsof the lines.11 The practitioner usually limitsthe area of interest to the peak of a structuralhigh, which presents the challenge of highdips. But even if this area can be reliablyimaged, expanding the local result to thewhole area of interest is difficult. There is nogood method for this. Practitioners tend torevise the macro model by trial and error.Verification, however, is easier, using thesame discrepancy between common depthpoint gathers as in moveout analysis.

Which Migration Algorithm?To perform migration, the seismic practi-tioner has two choices: the class of migra-tion (meaning 2D or 3D, prestack or post-stack, time or depth) and the type ofmigration algorithm.

The choice is partly constrained by thedata acquisition scheme. Two-dimensionaldata can undergo only 2D migration. Sam-ple density and, in 3D surveys, proximity ofother lines, can affect accuracy of the macromodel and therefore the choice of algo-rithm. Some algorithms are more tolerant ofmacro model error than others.

Selection of a migration algorithm isalways a balancing act (next page). Influ-encing the decision are special problems ofthe structure being imaged (such as steepdip or complex overburden velocity), eco-nomics, computer resources, available time,availability of expertise and certainty of themacro model. The ideal migration algorithmhas the following characteristics:

Oilfield Review

January 1993

2D Seismic Data

3D Seismic Data

Prestack depth

Limited offset stack +poststack depth

Prestack time +poststack depth

Prestack time

DMO + poststack time

DMO + poststack depth

Poststack depth

DMO + poststack time

Poststack time

DMO + poststack time

Poststack time

No migration

Cost

Cost

Prestack depth

Cross-line time + inline depth

Prestack time +poststack depth

Prestack time

DMO + poststack time

Astronomical

HighPoststack depth

DMO + poststack depthPoststack depth

Poststack timeDMO + poststack timeCross-line time + inline depth

Poststack time

No migration

DMO + poststack time

Very high

High

High

Low

Low

Very low

Very high

Lower

High

Low

High

Low

Low

Very low

Cost

Cost

Cost

Cost

OverburdenVelocities

OverburdenVelocities

OverburdenVelocities

OverburdenVelocities

Com

plex

Simple

Com

plex

Simple

Complex

Simple

Complex

Simple

Complex

Simple

Complex

Simple

Cost

Cost

Kno

wle

dge

of d

ata

incr

ease

s

Sta

ge in

exp

lora

tion/

app

rais

al/d

evel

opm

ent

pro

cess

nBP’s guidelines for selecting migration strategies for 2D and 3D data, and (next page)comparison of three migration algorithms. (From Redshaw and Lasseter, reference 1.)

9. Wang SS, Baumel RT, Hanson DW, Bell DW, Boyd M,Cavanaugh TD, Cox VD, D’Onfro PS, Durrani JA andStandlee LA: “Common-Offset Depth Migration as aVelocity Analysis Tool,” in Versteeg R and Grau G(eds): The Marmousi Experience, in Proceedings ofthe 1990 EAEG Workshop on Practical Aspects ofSeismic Data Inversion, Copenhagen, Denmark (May30, 1990): 139-158.

10. Faye J-P and Jeannot J-P “Prestack Migration Veloci-ties from Focusing Depth Analysis,” ExpandedAbstracts, 56th SEG Annual International Meetingand Exposition, Houston, Texas, USA (November 2-6, 1986): 438-440.

Depth migration improves the accuracy of reflectors.True—sometimes. Time migration coupled with

image-ray depth conversion may give as good a

section as depth migration at a fraction of the cost.

Depth migration is generally superior only in the

presence of rapid changes in lateral velocity. In

general, depth migration costs about ten times

more than image-ray depth conversion.

Prestack depth migration generally gives infe-rior results compared to poststack depth migrationbecause prestack migration is more sensitive toerrors in the velocity model.

Prestack migration is more sensitive to errors in

the macro model and will give inferior results—if

the macro model is wrong. If the macro model is

correct, prestack depth migration can give superior

results.

Depth migration-based velocity analysis (hightech) is better than image ray-based velocity anal-ysis of a time migration (low tech).

This high-tech approach offers no significant

advantage where velocities or structures are simple.

The high-tech approach is advantageous when there

is complex structure, complex velocity, or both.

Velocity derived from well data should be con-sidered the true velocity.

True, in the absence of significant velocity

anisotropy. But where there is velocity anisotropy,

use of well velocity—which is often the vertical

component of velocity—may allow focusing the

image, but at the wrong depth—or moving the

image to the right depth, but out of focus. Today,

most commercial techniques achieve a compro-

mise between focusing and positioning an image.

In the future, depth migration will include esti-

mates or even measurements of velocity anisotropy.

Misconceptions inStructural Imaging

39

11. Routine 3D-prestack Kirchhoff imaging is done onlimited regions of interest, using small scalar com-puters. See Hutton GD and Berg KI: “3D LocalImaging Utilizing Paraxial Ray Traced Green’s Func-tions,” Expanded Abstracts, 62rd SEG Annual Inter-national Meeting and Exposition, New Orleans,Louisiana, USA (October 25-29, 1992): 1007-1009.

4

Comparison of Common Migration Algorithms

Type

Kirchhoff

Phase shift(Gazdag)

f-x(frequency-space)

• Nonrecursive Kirchhoffaccurate for steep dips

• Migrates selected timewindows

• Accurate to 90° dip ifvelocities vary only vertically

• May be extended to includelateral velocity variation

• Generates no noise artifacts

• Accurate for steep dips andvertical and lateral velocityvariations

• Can be fast for large, 3Dpoststack data

• More expensive than finite-difference methods

• May generate false events

• Expensive• Not widely available

• Cannot perform commonoffset migration

• Although more accuratethan Kirchhoff migration, toocomputationally intensivefor 3D prestack migration

Advantage Limitation

Kirchhoff Migration performs a weighted sum (an integral) along diffraction curves and positions the sum at the apex of the curvein the migrated section. It is based on Kirchhoff’s solution to the wave equation and requires ray tracing for computation of traveltimes, which may give inferior results in complex velocity fields.

Phase shift migration, also called Gazdag migration, uses Fourier transform to convert space and time axes to wavenumber andfrequency, respectively. It is nearly exact in the absence of lateral velocity variation.

f-x migration, also called omega-x and recursive Kirchhoff migration, is used for almost all 2D prestack and 3D poststack depthmigration. See reference 11, page 39.

•Handles any range of dips•Handles large velocity changes—more

than 30 to 40%, possibly up to 100% inany direction

•Given a correct velocity field, accuratelypositions images in all three coordinates

•Preserves amplitude and phase•Generates no artifacts•Is inexpensive and fast.

No algorithm can do all of these. Somecan do only one adequately (above). Forexample, f-x migration easily handles largechanges in velocity and provides excellent

0

image resolution. But it can image reflectorsdipping only to 80°. On the other hand,nonrecursive (integral) Kirchhoff migrationhandles dips to 90° and provides excellentresolution, but has problems with largevelocity changes.12

Production specialists use several algo-rithms. Amoco, for example, may use onealgorithm for the top of the salt, another forthe flanks, and a third for imaging the baseof the salt and below. In complex velocityfields, GECO-PRAKLA may employ severalalgorithms in layer stripping.

Imaging in the FuturePerhaps the most significant research inimaging today concerns introduction ofvelocity anisotropy into macro model build-ing. In some rocks, especially shales, veloc-ity measured horizontally may vary fromthat measured vertically by up to 50%. Thisanisotropy can profoundly affect velocitymeasurements made on core, by well logs,vertical seismic profile and estimates fromstacking. Of interest are the effects of twotypes of anisotropy: that associated withdepositional fabric, called layer induced,and that associated with postdepositionaldeformation, called azimuthal or fractureinduced. These effects have been recog-nized for at least 20 years, but have escapedroutine measurement.13 The hope is thatbetter characterization of these anisotropicvelocity fields will lead to better location ofstructures. Improved image resolution willbe a secondary effect.

In highly anisotropic formations, betterknowledge of anisotropy may also lead toimproved ties between seismic and logvelocities. In the late 1980s, BP discoveredthat in the North Sea closely tying seismicvelocity to well logs resulted in undermigra-tion by 7%. Velocity anisotropy was thecause—horizontal velocity, measured byseismic methods, was 7% greater than verti-cal velocity, which dominates the log read-ing.14 However, this discrepancy may alsobe caused by seismic surveys and well logssampling at different frequencies, and there-fore measuring different velocities. Until thissampling discrepancy and effects of aniso-tropy can be resolved, some operatorschoose not to constrain seismic velocity bywell velocity, but to use log velocity only asa guide.

Better accounting for anisotropy may alsoimprove an emerging method called turningray imaging.15 A turning ray is one thatpasses through horizontal, goes up to hit thereflector then down before returning to thesurface (next page). These rays are useful inimaging beneath and around salt overhangs.Current imaging processing assumes veloc-ity increases smoothly with depth and is

Oilfield Review

nWhat’s hiding under the salt? Use of turning ray migration to image the flanks of a salt wedge in the Gulf of Mexico. In the unmi-grated 2D stack (left), part of the steeply dipping energy, visible on the right, has traveled along turning raypaths and reflected froma salt overhang. The overhang is not clearly imaged. In the migrated section (right), the image produced by turning ray time migra-tion clearly shows the right side of the salt overhang.

Turning ray reflection

Constant velocity

Varying velocity

Saltdome

Unmigrated 2D-stack Turning ray time migration

isotropic. Consequently, turning ray pathsare assumed to be arcs of circles, which per-mits easy conversion from travel time todepth. But often velocity is anisotropic andthe arcs are not circles, making time-depthconversion more difficult and reflector loca-tions less certain. Better knowledge of aniso-tropy may help account for noncircular arcsand allow better positioning of reflectors.

The chief computational challenge is 3Dprestack depth migration, which requires a

January 1993

12. Integral and recursive approaches are both based onthe same wave equation that governs transmission ofenergy in the earth. Recursive migration propagatesthe recorded data back into the earth one depth-stepat a time, typically every 10 meters [33 ft]. This pro-cess gives data that would have resulted from havingsources and receivers located at the current depth.By comparison, the integral approach migrates dataat each image point directly via a weighted sum ofthe recorded data that reside on a travel-time curve.This curve is assumed to be hyperbolic in timemigration implementation, although not in depthmigration implementation.

13. Vander Stoep DM: “Velocity Anisotropy Measure-ments in Wells,” Geophysics 31 (October 1966):900-916.

gigaFLOP machine to handle data often onthe order of terabytes (one trillion bytes).16

When this becomes possible, the macromodel could be built with prestack 3D data,allowing accurate description of abrupt lat-eral variations in velocity.17

Structural imaging is both pushing andbeing pushed by advances in computer sci-ence. Increased computer power has beenlinked with the trend from 2D to 3D, fromtime migration to first-generation depth

14. Whitcombe DN and Collyer VA: “Evidence for theExistence of Anisotropy Beneath the Southern NorthSea,” presented at the 50th meeting of the EAEG,The Hague, The Netherlands, 1988.

15. Hale D, Hill NR and Stefani J: “Imaging Salt withTurning Seismic Waves,” Geophysics 57 (November1992): 1453-1462.Ratcliff DW, Gray SH and Whitmore ND: “SeismicImaging of Salt Structures in the Gulf of Mexico,”The Leading Edge 11 (April 1992): 15-31.

16. One gigaFLOP is 1 billion floating point operationsper second, a measure of computational speed.Highnam PT and Pieprzak A: “An Implementation ofa Fast Accurate 3-D Migration on a Massively Paral-lel Computer,” Expanded Abstracts, 61st SEGAnnual International Meeting and Exposition, Hous-ton, Texas, USA (November 10-14, 1991): 338-340.Kao JC: “A Practical Implementation of the Multi-processing 3D Kirchhoff Prestack Migration Schemeon the Cray Y-MP Systems,” Technical Programmeand Abstracts of Papers, 54th EAEG Meeting andTechnical Exhibition, Paris, France (June 1-5, 1992):284-285.

migration, from poststack to prestack.18 Thefuture will probably see closer linkage ofvelocity model estimation with migration,which will expand its role to include lithol-ogy and pore fluid identification. These pos-sibilities pivot on advancement of massivelyparallel computation, the tool that willallow seismics to see ever more clearly.

—JMK

41

Cabrera JJ, Perkins WT, Ratcliff DW and Lynn W:“3D Prestack Depth Migration on a Massively Paral-lel Computer: Implementation and Case History,”Technical Programme and Abstracts of Papers, 54thEAEG Meeting and Technical Exhibition, Paris,France (June 1-5, 1992): 268-269.

17. Landa E, Kosloff D, Keydar S, Koren Z and Reshef M:“A Method for Determination of Velocity and Depthfrom Seismic Reflection Data,” GeophysicalProspecting 36 (1988): 223-243.

18. Berkhout AJ: “Massively Parallel Processing in E&P:Geoscience or Computer Science?” Journal of Seis-mic Exploration 1 (1992): 119-120.Berkhout AJ: “Trends in the Seismic Industry,” Jour-nal of Seismic Exploration 1 (1992): 3-8.

4

Hydrocarbon Detection With AVO

Edward ChiburisECGeoHouston, Texas, USA

Charles FranckRoyal Oil & Gas CorporationCorpus Christi, Texas, USA

Scott LeaneyJakarta, Indonesia

Steve McHugoOrpington, England

Chuck Skidmore Amoco Production CompanyHouston, Texas, USA

S E I S M I C S

Imagine a geophysical technique with the volumetric coverage of surface seismic that could delineate zones

of gas, oil and water. In many ways, that summarizes the potential of interpreting seismic reflection ampli-

tude variation with offset, or AVO.

nBright spots—froma gas sand, andfrom a high veloc-ity basalt. Numbersat top are shotpoints. (Courtesy of BillOstrander and ChevronCorporation, reference 5.)

Tim

e, s

ec

0

0.5

1.0

1.5

2.0

0170 150 130 110 90

150 130 110 90 70A

A

Gas

In the late 1920s, the seismic reflectiontechnique became a key tool for the oilindustry, revealing shapes of subsurfacestructures and indicating drilling targets.This has developed into a multibillion dollarbusiness that is still primarily concernedwith structural interpretation. But advancesin data acquisition, processing and interpre-tation now make it possible to use seismictraces to reveal more than just reflectorshape and position. Changes in the charac-ter of seismic pulses returning from a reflec-tor can be interpreted to ascertain the depo-sitional history of a basin, the rock type in alayer, and even the nature of the pore fluid.This last refinement, pore fluid identifica-tion, is the ultimate goal of AVO analysis.

Early practical evidence that fluids couldbe seen by seismic waves came from “brightspots”—streaks of unexpectedly high ampli-tude on seismic sections—often found to

2 Oilfield Review

For help in preparation of this article, thanks to JackCaldwell, Barry Donaldson and Jim Hovland, GECO-PRAKLA, Houston, Texas, USA; Al Frisillo, Amoco Pro-duction Company, Tulsa Research Center, Tulsa, Okla-homa, USA; Michael Mikulich, Chevron Corporation,San Francisco, California, USA; Bill Murphy and AndyReischer, Schlumberger-Doll Research, Ridgefield, Con-necticut, USA; Bill Ostrander, Benecia, California, USA.* In this article, DSI (Dipole Shear Sonic Imager) andELAN (Elemental Log Analysis) are marks of Schlumberger.1. An intermediate step called normal moveout correc-

tion is omitted here for brevity, but described in“Structural Imaging: Toward a Sharper SubsurfaceView,” page 28.

Tim

e, s

ec

0.5

1.0

1.5

2.0

Basalt-dry

nSingle-layeracquisition geome-try, associated syn-thetic traces show-ing AVO effect ofgas sand, and amultilayer geome-try. Synthetics showamplitude increas-ing with offset(deflections to theleft becoming morenegative). Althoughseismic coordinatesare based on offset,theory relatingchanges in ampli-tude with materialproperties of thereflector is basedon angle (θ).Because the reflec-tion point is mid-way betweensource and receiver,and common to alltraces, it is calledthe common mid-point (CMP). A col-lection of traces isknown as a gather,in this case a CMPgather. In simplecases (top) with lay-ers of uniform den-sity and velocity, adirect relationshipexists between off-set and angle ofincidence. But mostof the time (bottom),variations in den-sity and velocitybend rays, requir-ing ray-trace mod-eling to relateangle to offset. aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaOffset 4

Offset 3

Offset 2

Offset 1

S4 S3 S2 S1 R1 R2 R3 R4

Offset 4

Offset 3

Offset 2

Offset 1

Shale

Gas sandCommonmidpoint (CMP)

S1

Gas sand

Shale 1

Shale 2

CMP

Synthetic traces—CMP gather

Amplitude increases with offset

R1

θ1

θ2

Single-layer geometry—Direct relationship between θ and offset

Multilayer geometry— Complex relationship between θ and offset

signify gas. Bright spots were recognized inthe early 1970s as potential hydrocarbonindicators, but drillers soon learned thathydrocarbons are not the only generators ofbright spots. High amplitudes from tight orhard rocks look the same as high amplitudesfrom hydrocarbons, once seismic traceshave been processed conventionally (previouspage). Only AVO analysis, which requiresspecial handling of the data, can distinguishlithology changes from fluid changes.

An analogy for the physics of AVO is theskipping of a stone across a pond. Everyoneknows that if a stone is dropped or throwninto water from directly above, it sinksinstantly. But skimmed nearly horizontally, itbounces off the surface of the water. Theamplitude of the bounce, which was zero atvertical incidence, increases with the angleof incidence.

Now replace the water with rubber andrepeat the process. This time the verticalbounce is high, and the high-angle bounceis low. The amplitude of the bouncedecreases with angle of incidence, a dramat-ically different behavior from the water case.

Analogous concepts applied to seismicsform the basis for inferring formation prop-erties—density and compressional andshear velocities—from seismic reflectionamplitude variation with angle of incidence.And because formation density and velocitydepend on the fluid saturating the forma-tion, reflection amplitude variation also per-mits identification of pore fluid.

January 1993

Conventional treatment of seismic data,however, masks this fluid information. Theproblem lies with the way seismic traces aremanipulated in order to enhance reflectionvisibility. In a seismic survey, as changes aremade in the horizontal distance betweensource and receiver, called offset, the angleat which a seismic wave reflects at an inter-face also changes (above). Seismic traces—recordings of transmitted and reflectedsound—are sorted into pairs of source-receiver combinations that have differentoffsets but share a common reflection pointmidway between each source-receiver pair.This collection of traces is referred to as acommon midpoint (CMP) gather. In conven-tional seismic processing, in which the goalis to create a seismic section for structural or

stratigraphic interpretation, traces in agather are stacked—summed to produce asingle average trace.1

Stacking enhances signal at the expenseof noise, making reflections visible, andcompresses data volume. But it destroysinformation about amplitude variation withoffset. Consider two reflections in the sec-tion: one has amplitude increasing with off-set, such as in the case of the stone bounc-ing off the water, and the other hasamplitude decreasing with offset, similar tothe stone bouncing off rubber. Once thereflection traces are stacked, they may have

43

1. Green G: “On the Laws of Reflexion and Refraction of Light,” inTransactions of the Cambridge Philosophical Society 7 (1839):245. Reprinted in Mathematical Papers of the Late GeorgeGreen. London, England: Macmillan, 1871.

Kelvin WT: “Reflexion and Refraction of Light,” PhilosophicalMagazine, Fifth Series 26 (1888): 420-422.

2. In the 1620s, Snell observed that light refracts at different anglesthrough different materials. In 1657, Fermat postulated the cor-rect expression for its reflection and refraction.

Knott CG: “On the Reflexion and Refraction of Elastic Waves,With Seismological Applications,” Philosophical Magazine, FifthSeries 48 (1899): 64-97.

Zoeppritz K: “Über Erdbebenwellen, VIIB: Über Reflexion undDurchgang seismischer Wellen durch Unstetigkeitsflächen,”Nachrichten der Königlichen Gesellschaft der Wissenschften zuGöttingen, Mathmatisch-physikalische Klasse (1919): 57-84.

3. Macelwane JB and Sohon FW: Introduction to Theoretical Seis-mology. New York, New York, USA: John Wiley & Sons, Inc.(1936): 147-179.

4. Muskat and Meres published hand-calculated reflection coeffi-cients in 1940, assuming Poisson’s ratio was a constant 0.25across the interface. Koefoed, in 1955, and Bortfeld, in 1961,simplified the Zoeppritz expressions by assuming that changesin material properties across the interface are much smaller thanthe average value of the properties on each side and that theincident angle does not exceed 30°.

Muskat M and Meres MW: “Reflection and Transmission Coeffi-cients for Plane Waves in Elastic Media,” Geophysics 5 (1940):115-148.

Koefoed O: “On the Effect of Poisson’s Ratios of Rock Strata onthe Reflection Coefficients of Plane Waves,” Geophysical

identical amplitudes—they may even bebright spots—while their AVO signatures arecompletely different. AVO analysis can usu-ally distinguish fluid contrasts from lithologycontrasts, but it requires carefully processedgathers that have not been stacked.

A Little TheoryAttempts at practical application of AVObegan about 15 years ago, but the physicswas understood around the turn of the cen-tury (see “History,” below, right). The gen-eral expressions for the reflection of com-pressional and shear waves at a boundary asa function of the densities and velocities ofthe layers in contact at the boundary arecredited to Karl Zoeppritz.2 Zoeppritz foundthat amplitudes increase, decrease orremain constant with changing angle ofincidence, depending on the contrast indensity, compressional velocity, Vp , andshear velocity, Vs , across the boundary.

Conventional seismic surveys deal exclu-sively with the reflection of compressionalwaves. When a compressional seismicwave arrives vertically at a horizontal inter-face, the amplitude of the reflected wave isproportional to the amplitude of the incom-ing wave, according to the normal inci-dence reflection coefficient:3

in which ρ is density and 1 and 2 signify thetop and bottom layer, respectively. Whenthe seismic wave arrives obliquely, the situ-ation is more complicated. The compres-sional reflection coefficient is now a tortu-ous function of the angle of incidence, thedensities, and Vp and Vs of the two layers incontact. The simplest useful approximationsto the Zoeppritz theory comprise the normalincidence reflection coefficient writtenabove plus at least three other terms—func-tions of angle and contrasts in density andthe two velocities. Nevertheless, the depen-dence of reflectivity on density, Vp and Vsmakes it possible to deduce fluid and rocktype. Gas, oil and water have different den-sities and acoustic velocities. They insinuatethis difference on the rocks they saturate.

ρ2Vp2 - ρ1Vp1

ρ2Vp2 + ρ1Vp1 '

44

2. Zoeppritz K: “Über Erdbebenwellen, VIIB: ÜberReflexion und Durchgang seismischer Wellen durchUnstetigkeitsflächen,” Nachrichten der KöniglichenGesellschaft der Wissenschften zu Göttingen, Math-matisch-physikalische Klasse (1919): 57-84.

3. This is strictly true only if transmission effects, such asspherical spreading and attenuation, are neglected. Italso holds when the layers are not horizontal, as longas the angle between the wave and the interface is 90°.

Synthetic AVOs from LogsIn much the same way that a blindfoldedexpert can identify a wine and its vintage, oran X-ray diffraction lab technician can iden-tify mineral components in a rock sample,the key to using AVO for fluid identificationis comparison of real data with a stan-dard—in this case a synthetic seismogram (a“synthetic” for short). This is an artificialseismic trace manufactured by assumingthat some pulse travels through an earthmodel—rock layers of given thickness, den-sity and velocity—and returns to berecorded. The earth model that producedthe synthetic can be modified, sometimesrepeatedly, until the synthetic matches themeasured data, indicating the earth model isa reasonable approximation of the earth.

The densities and velocities of fluid-satu-rated rocks necessary for the creation ofsynthetic traces preferably come from logsor cores. Missing data can be estimated

using theoretical or empirical equations.The synthetic traces show the expectedAVO effect for each fluid type.

Take, for example, the AVO effect of gasin sandstone predicted from logs in a gasfield operated by Texas-based Royal Oil &Gas (next page, top). Here, acoustic veloci-ties were measured with the DSI DipoleShear Sonic Imager tool. The seismic eventof interest is the circled blue reflection cor-responding to the interface between anoverlying shale and the gas sand. The tracerecorded at zero offset—0° from vertical,directly above the reflecting point—beginswith a small negative amplitude (tracedeflects to the left). The amplitude becomesmore negative as offset increases. The AVOresponse to oil is the same (next page, mid-dle). But when hydrocarbons are replacedwith water, the AVO response changes (nextpage, bottom). Now polarity becomes posi-

Prospecting 3 (December 1955): 381-387.

Bortfeld R: “Approximation to the Reflection and TransmissionCoefficients of Plane Longitudinal and Transverse Waves,” Geo-physical Prospecting 9 (December 1961): 485-502.

5. Aki K and Richards PG: Quantitative Seismology, Theory andMethods Vol. 1. San Francisco, California, USA: WH Freemanand Co. (1980): 123-155.

Shuey RT: “A Simplification of the Zoeppritz Equations,” Geo-physics 50 (April 1985): 609-614.

Attempts at practical application of AVO began

about 15 years ago, but the physics draws on 19th

century advances in optics and electromagnetic

wave theory. In the 1800s, Green and Kelvin spec-

ulated about the similarity of the reflective

behavior of light and elastic waves.1 Using

Snell’s law, Knott in 1899, and Zoeppritz in 1919,

developed general expressions for the reflection

of compressional and shear waves at a boundary

as a function of the densities and velocities of the

layers in contact.2 Although Zoeppritz was not the

first to publish a solution, his name is associated

with the cumbersome set of formulas describing

the reflection and refraction of seismic waves at

an interface. In 1936, Macelwane and Sohon

recast the equations to gain insight into the

physics and facilitate calculations.3

Before computers were widespread, AVO

effects were incorporated into synthetic seismo-

grams and other calculations using approximated

Zoeppritz equations.4 Today, personal computers

can generate synthetics based on the full Zoep-

pritz formalism, but approximations are still used

to gain physical insight into the relative influence

of velocity and density changes on seismic ampli-

tudes, and in the attempt to back out lithology

and fluid type from AVO data.5

History

Oilfield Review

nSynthetic traces showing AVO effects ofgas, oil and water, and an ELAN Elemen-tal Log Analysis output. The negativepolarity event at the top of the gasbecomes more negative with increasingoffset, a signature of gas (top). In this idealcase, the synthetics are created from anearth model based on measured logs con-verted from depth to time. Velocities weremeasured using the DSI Dipole Shear SonicImager tool. In the less desirable but morecommon situation, all three log measure-ments are not available. Missing data arecreated using empirical relations. Mea-sured logs (middle) from an African oilsand at 0.97 sec, as identified by ELANcalculation, have an AVO effect similar togas, but in some cases, AVO can distin-guish the two. Water sands identified bythe ELAN log deeper in the section showno amplitude increase with offset. TheAVO signature of water is virtually alwaysdifferent from that of oil or gas, makingAVO a hydrocarbon indicator (bottom). Inthis example, water is substituted for gasin the sand of the top figure using empiri-cal and theoretical equations.

45January 1993

Tim

e, s

ec

1.0

1.1

Shale Boundwater Quartz

Calcite Oil Movedhydr

Tim

e, s

ecTi

me,

sec

1.0

1.1

1.5

1.6

Gas

Oil

Water

0

0

0 50001000

3280 6560

50001000

Offset (horizontal distancebetween source and receiver), ft

Vp Vs bρ

8 9 3 4 2 2.25kft/sec kft/sec g/cm3

Vp Vs bρ

8 9 3 4 2 2.25kft/sec kft/sec g/cm3

Offset, ft

Offset, ft

Matrix, %

Cal

cula

ted

Mea

sure

d

Cal

cula

ted

nThe measuredAVO gather fromRoyal Oil’s Texasgas well. Ampli-tude of the smallnegative (blue)reflection at the topof the gas becomesmore negative asoffset increases.The seismic datahave undergonetrue amplitude pro-cessing describedin “Processing forAVO Interpretation,”(page 48).

Offset, ft

Tim

e, s

ec

0 2200 4400 6600

1.3

1.03

0.8

0.55

0 1000 5000

1.0

1.1

Offset, ft

Angle, deg

Tim

e, s

ec

0 8 16 24 32

Am

plitu

de

0

-0.05

-0.1

-0.15

Angle, deg0 8 16 24 32

Am

plitu

de

0.15

0.1

0.05

0

Angle, deg0 8 16 24 32

Am

plitu

de

0

-0.05

-0.1

-0.15

Top gas sand

No fluid change

Bottom gas sand

nQuantitative Zoeppritz prediction of AVO effects for three reflections in the Texas gas field of Royal Oil & Gas. At top of gas sand (left),amplitude doubles, from slightly negative to very negative, as offset increases. At the gas-water contact near the bottom of the sand(lower right) amplitude increases 50%, from slightly positive to more positive. Most reflections that do not involve a fluid change (upperright) show negligible amplitude change.

46 Oilfield Review

Tim

e, s

ec

1.0

1.1

0 50001000

Offset, ft Gra

dien

t (G

)

Inte

rcep

t (P

)

Gx P

Gas

nSynthetic tracesfrom page 45, topand associatedAVO compositetraces—gradient,intercept and theirproduct (GxP). Wig-gles in the G trace,for example, haveamplitudes equal tothe best-fitting gra-dient computed forevery time level inthe gather. Reflec-tors emerge as neg-ative or positivewiggles, easily dis-tinguishable fromnoise, which staysaround zero. Posi-tive product indi-cates hydrocarbon,in this case gas.

tive (amplitude deflection to the right), andamplitude decreases with offset.4

The AVO effect at any interface can bequantified with the Zoeppritz formulas, andplotted as a curve (above). At the top of thegas sand in Royal Oil & Gas’ well, and formost gas sands, Zoeppritz calculations pre-dict an increasingly negative amplitude withoffset. In this case, the predicted negativeamplitude increases 100%. Also shown arethe Zoeppritz-predicted AVO effects for thegas-water contact deeper in the sand, andfor a nonfluid, lithologic contact higher inthe section.

These curves of amplitude versus angle ofincidence can be used to make quantitativecomparisons between synthetic predictionsand amplitudes from real data, once thedata have been processed for true ampli-tudes. This is made easier by plotting ampli-tude versus angle of incidence squared,which converts Zoeppritz curves to straightlines. AVO behavior can then be succinctlydescribed by the line’s gradient, G, and nor-mal incidence intercept, P.

For typical reservoir rocks, the reflectionat an interface between a water-bearinglayer and a hydrocarbon-bearing layer issuch that a negative polarity reflectionbecomes more negative—intercept and gra-dient both negative—or a positive polarityreflection becomes more positive—interceptand gradient both positive (above, right ).

ft0 5000

0.55

0.8

1.05

1.3

0.9

1.0

1.1

1.2

1.3

-2 0 2

Tim

e, s

ec

Tim

e, s

ec

Offset, ft0 2200 4400 6600

Am

plitu

de

Angle2

GxP Section

0 50 100 150 200CMP

nAVO data, gradient and interceptcalculation and product section for theTexas gas field. On the top, amplitudesare measured across data traces frompage 45, top. Amplitudes are fit to astraight line (middle)—by changingaxes in the original Zoeppritz predic-tion plot from amplitude vs. angle toamplitude vs. angle squared—with acorresponding gradient and intercept.Traces containing the values of thegradient, intercept, or the product ofthe two form a seismic section, in thiscase, a product section from the Texasgas field (bottom). The region betweenCMP 70 to 100 shows a positive AVOproduct (red), and corresponds to thegas-producing sand. A second well4400 ft to the left has confirmed thisgas extent.

nGathers from well locations in twobright spots seen on page 42, showingdifferent AVO signatures. At left is thegather from shot point 81 of top section.Negative amplitude becomes morenegative with offset in gas sand. On theright is a gather from shot point 127 ofbottom section. Amplitude decreaseswith offset, indicating no hydrocarbon. 47January 1993

Gasshot point 81

Basaltshot point 127

4. Velocities and density for this example were com-puted using a technique described in “Taking Advan-tage of Shear Waves,” Oilfield Review 4, no. 3, (July1992): 52-54. For further reading see Murphy W, Reischer A andHsu K: “Modulus Decomposition of Compressionaland Shear Velocities in Sand Bodies,” Geophysics 58,no. 2, (February 1993): 227-239.

5. Ostrander WJ: “Plane-Wave Reflection Coefficientsfor Gas Sands at Nonnormal Angles of Incidence,”Geophysics 49 (October 1984): 1637-1648.

The simplest indicator of hydrocarbons istherefore the product of gradient and inter-cept. A positive product most likely indi-cates oil or gas. A product trace for theRoyal Oil & Gas example clearly revealsgas. G traces, P traces or product traces canbe plotted next to each other to producesections, similar to stacked sections, forAVO interpretation.

Interpretation of Actual AVOsHow do real AVO gathers compare withsynthetics? The real gather (previous page,bottom) observed at the Texas gas well andcarefully processed by GECO-PRAKLA (see“Processing for AVO Interpretation,” nextpage) shows the same AVO signature as thesynthetic gather generated using log datafrom the well (page 45, top). Both gathersshow a small negative reflection at normalincidence that becomes more negative withoffset. This signals hydrocarbons, and sureenough, the well did produce gas. The gra-dient and intercept are both negative, andtheir product positive. A section composedof product traces from every gather in theseismic line shows a zone of positive prod-uct (right). A second well drilled in the zoneconfirmed the presence of gas.

Because the synthetic was built from logdata—density, compressional velocity andshear velocity—rather than estimated val-ues, it closely matches the observed gather.Often estimated is shear velocity, and thiscreates a common stumbling block to AVOmodeling. Dramatic AVO effects appear ingas sands where the shear velocity is oftentoo slow to be measured with conventionalsonic tools. Introduction of the DSI toolremoves this impediment.

Once the AVO signature of hydrocarbonsis known, seismic data can be examined forfluids. For example, what would AVO anal-ysis have revealed about the two brightspots—one from gas, the other frombasalt—described on pages 42-43? Theanswer was published in 1984 by BillOstrander, then at Chevron USA (below).5

48 Oilfield Review

6. Chiburis EF: “Studies of Amplitude Versus Offset inSaudi Arabia,” Expanded Abstracts , 57th SEG AnnualInternational Meeting and Exposition, New Orleans,Louisiana, USA (October 11-15, 1987): 614-616.

1. The GRT technique is described in Beylkin G, Miller D andOristaglio M: “The Generalized Radon Transform: A Break-through in Seismic Migration,” The Technical Review 35, no. 3(July 1987): 20-27.

Carefully processed gathers from the welllocations show the difference between theAVO signature of gas and that of high-veloc-ity basalt. Gas shows the now-familiarincrease of amplitude with offset, whilebasalt shows a decrease.

AVO effects may also be tracked across areservoir to delineate a fluid contact. A tech-nique developed by Ed Chiburis, while atSaudi Aramco, has had remarkable successdelineating Saudi Arabian oil reservoirs.6 In26 of 27 cases, the technique predicted thepresence or absence of oil, which was laterconfirmed by drilling (next page, bottom).

The technique identifies changes in AVObehavior along a seismic line, and associ-ates those changes with changes in fluidcomposition. Once a given fluid has beenidentified in a well, the AVO behavior of thegather at the well is defined as the standardto look for elsewhere in the section.

To overcome the lack of true amplitudeprocessing in most data, Chiburis developeda normalization technique using anotherreflection that shows consistent amplitudein the section as a reference. Peak ampli-tudes of the target reflection in each AVOgather are picked interactively on a worksta-tion and normalized trace by trace to thereference event. Use of a reference eventremoves or minimizes amplitude distortionassociated with flaws in acquisition andprocessing. The technique also circumventsthe need for synthetics. The measured AVOresponse at the well serves as the standard.The major limitation of this method is thatthe geology and stratigraphy must be well-known in order to associate changes inAVO with changes in fluid type.

Another clever AVO analysis techniquepracticed by Amoco is to display and com-pare seismic sections made up of partialstacks (next page, top). Here, AVO informa-tion—or fluid discrimination informa-tion—masked by a full stack, is retained in apartial stack of the far offsets. A partial stackis similar to a full stack, except that eachtrace is the average of traces in a smallrange of offsets rather than all offsets in theCMP gather.

Processing for AVO Interpretation

Any properly acquired seismic survey, new or old, can be processed for AVO analysis. The goal of

processing is to preserve reflected pulse shape and amplitude. Changes in pulse with offset can then

be interpreted in terms of lithology or fluid contrasts at the reflector. Data destined for stratigraphic

interpretation or lithostratigraphic inversion (see “Structural Imaging: Toward a Sharper Subsurface

View,” page 28) also benefit from true amplitude processing. Every seismic data set creates its own

processing problems, requiring a tailor-made processing sequence. Here is a typical AVO processing

sequence, one that works for the data sets described in this article.

Basic Steps

True Amplitude Recovery (TAR)—compensates

for amplitude loss caused by wavefront spreading

and low transmission quality (Q) of the rock

through which the seismic wavefront travels.

Frequency wave number (F-K) filtering—is

required to attenuate coherent noise generated

by near-surface or seabed features such as rigs,

buildings and seabed channels. Ground roll, or

surface waves, common in land data, cannot usu-

ally be removed with this method. Correctly

designed receiver arrays can solve this problem.

Generalized Radon Transform (GRT) demultiple—

reduces amplitude of multiples (interbed or water

column reverberations) relative to primary

energy. Conventional demultiple techniques do

not preserve true amplitudes, nor do they elimi-

nate all multiples. The GRT demultiple separates

seismic arrivals by differences in their apparent

velocities, then suppresses multiples by an

inverse transform of only part of the data.1

Deconvolution—creates a new trace with wiggles

that indicate the location (in time) and the

strength of each reflector. Surface-consistentdeconvolution reduces pulse shape distortion

because the filter is the same for each shot and

receiver location.

Surface-consistent scaling and residual statics—

correct amplitudes and arrival times of raypaths

distorted by near-surface anomalies, such as

those caused by the unconsolidated (“weath-

ered”) zone on land or a rough ocean bottom.

Velocity analysis and Normal Moveout (NMO)—create and apply the velocity model that aligns

wiggles from all offsets. In conventional seismic

processing, velocity analyses are made every 2 to

3 km [1.2 to 1.8 miles]. Because most AVO

anomalies are caused by velocity variation,

closely spaced velocity analyses are required

every 0.25 km [0.15 mile].

Fine-Tuning Steps

CMP-consistent statics—sometimes called non-

surface consistent statics or trim statics, forces

alignment of selected events that have not been

properly aligned by standard processing, on a

CMP-by-CMP basis. This is a compromise, and

considered unscientific by purists.

Dephasing—attempts to restore each trace to

zero phase so reflection events can be tracked

and their amplitude changes quantified.

Mixing and median filtering—attenuate random

noise and reinforce signal. In median filtering,

traces within a small offset range are summed or

stacked. Mixing does the same kind of summing

over a small range of CMPs.

Bandpass filtering—removes high- and low-fre-

quency noise from traces.

I R A QI R A N

S A U D I A R A B I A

A R A B I A N G U L F

KUWAIT

QATARRiyadh

Known reservoirs

AVO locations

Dhahran

km

miles

Line

1

Line 2

Line 3

Line 4

Line 5

km

miles

Known extentof reservoir

0 200

0 125

0 10

0 6.2

nSaudi reservoirs, showingdelineation of spatialextent of a hydrocarbondiscovery using Chiburis’relative event AVO tech-nique. In the inset, the yel-low area under the curvesshows where AVO indicateshydrocarbon. Based onAVO analysis and the localgeology, the interpretationof the reservoir extent isshaded in dark green.

nNow you see it, now you don’t—a far offset stack of 2D offshoredata reveals hydrocarbons masked in conventional full stack.Plotted in two-way time, brighter colors indicate higher reflectionamplitude. Red and yellow are positive polarity, green and bluenegative. Insets show offsets stacked to create each section. Bedsdipping to the right are the only structure visible in all but the far-offset case. A flat spot is circled (above, left), indicating a fluid-fluid contact cutting across the dipping layers. (Courtesy of Rod VanK-oughnet and Chuck Skidmore, Amoco Production Company.)

49January 1993

Far offset stack Near offset stack Full stack

1.5

1.9

Tim

e, S

ec

Where is AVO going?Some companies use AVO routinely in anattempt to reduce risk associated withpotential drilling locations. Others havetried the technique and found the process-ing too time-consuming or too difficult. Anincreasing number of practitioners is insist-ing on quantitative agreement between syn-thetic and observed data before they willuse the technique. Currently, most examplesof AVO interpretation are qualitative. In theRoyal Oil & Gas example (page 46, middleand bottom), the qualitative match betweenthe two is good, but quantitatively, the syn-thetics predict a 100% increase in ampli-tude with offset while the data show anincrease of more than 200%.

Eliminating the discrepancy betweenobserved and synthetic data is therefore afocus of AVO-related research, and toucheson five main topics—processing, syntheticmodeling, petrophysics, interpretation andinversion.

Researchers seek a true amplitude pro-cessing scheme to produce AVO data tracesthat can be compared quantitatively to com-puter-perfect synthetics. Conventional pro-cessing for structural imaging does not pre-serve amplitudes. Researchers are revisiting

50 Oilfield Review

basic processing steps such as deconvolu-tion, velocity analysis and migration with aview to AVO applications.

Current research in synthetic modelingaddresses a wide range of topics. Syntheticsare only as good as what goes into them.How should logs sampled every 6 inches[15 cm] be “averaged,” or blocked, to pro-duce layered earth models? Different block-ing techniques produce different synthetics.What is the effect of layer thickness on AVOsynthetics? The right combination of layerthickness and seismic wavelength gives riseto reverberations in the layer that alterreflected amplitude. Can seismic energy bemodeled as simple rays, or is it better to useseismic wave theory? In the examples pre-sented above, ray theory was enough. Butwhen angles become large and velocityvariations complex, more computer-inten-sive wave theory is necessary. How doesvelocity anisotropy affect AVO? As angle ofincidence increases, differences betweenhorizontal and vertical velocities cannot beignored in earth models.

In general, petrophysics is the linkbetween earth models and any seismicinterpretation, but it is particularly importantin AVO interpretation. Changes in porosity,mineralogy, cementation, stress, com-paction or other properties that modify thevelocity or density of the rock, can give riseto AVO signatures that mask fluid effects.Changes in fluid saturation, on the otherhand, may exhibit no change in AVO signa-ture. For example, in shallow or unconsoli-

dated sands, or overpressured zones, theAVO response is about the same for all satu-rations. Drilling will confirm the presence ofgas, but it might be just “fizz water.” Labora-tory and field measurements on reservoirrocks, and especially nonreservoir rocks,under in-situ conditions, are crucial to theconstruction of a reliable earth model.Improved understanding of rock propertiesat core, log and seismic scales will lead tomore unambiguous AVO interpretation.

Standard AVO interpretation fits reflectionamplitudes to straight line approximationsof Zoeppritz prediction curves. More refinedinterpretations quantify goodness-of-fit orother statistical analyses of the fit. Work isunder way to abandon the straight-lineapproximation and fit the real curve.

A great deal of research is devoted toinversion, the attempt to derive a likelyearth model starting with real data—theinverse of synthetic modeling. To date,results indicate that knowledge of the Vp /Vsratio is required for stable inversion. Some-times Vp can be estimated from seismicstacking velocities, but Vs cannot. Full inver-sion of AVO data for material propertiescontinues to intrigue researchers, but it hasyet to be proven feasible.

What is the future in AVO? One hot topicis three-dimensional (3D) AVO. Many oper-ators have already successfully interpreted3D seismic data sets for AVO by assemblingtwo-dimensional (2D) AVO sections inseries. Few have tried real 3D AVO, that is,considering source-receiver paths in differ-ent azimuths. This requires knowing veloc-ity anisotropy in the horizontal plane.

Time-lapse AVO is another topic thatshows promise. As a reservoir is produced,fluid contacts will move. Seismic surveysshot at different times can be analyzed forfluid changes using AVO techniques. Infor-mation about drained and undrained vol-umes can affect development and produc-tion plans.

AVO can be a powerful tool for hydrocar-bon detection. Although experts may com-prehend the theory behind AVO interpreta-tion, unwary practitioners make mistakes.Progress has been made in modeling, pro-cessing and interpretation, but improve-ments need further joint efforts from the seis-mic, logging and petrophysics communities.

—LS

Sequence Stratigraphy—A Global Theory for Local Success

S E I S M I C S

No exploration technique flawlessly locates a potential reservoir, but sequence stratigraphy may come close.

By understanding global changes in sea level, the local arrangement of sand, shale and carbonate layers can

be interpreted. This enhanced understanding of depositional mechanics steers explorationists toward

prospects missed by conventional interpretation.

1.0

2.0

3.0

Tim

e, s

ec

0 1milesJack NealRice UniversityHouston, Texas, USA

David RischHouston, Texas, USA

Peter VailRice UniversityHouston, Texas, USA

For help in preparation of this article, thanks to ScottBowman, Marco Polo Software, Houston, Texas, USA;Carlos Cramez and Bernard Duval, TOTAL Exploration,Paris, France; Andrew Hannan, GECO-PRAKLA, Hous-ton, Texas, USA; Ulrich Möller, GECO-PRAKLA, Han-nover, Germany; and John Sneider, Rice University,Houston, Texas, USA.

nA seismic section interpreted to show the sandy interval (yellow) predicted usingsequence stratigraphy. The heavy vertical line shows the well location.

January 1993

0 1miles

Conventional lithologic correlation mapsformation tops by interpreting well log dataalone. It looks at what is there without tak-ing into account how it got there. Sequencestratigraphy combines logs with fossil dataand seismic reflection patterns to explainboth the arrangement of rocks and thedepositional environment. Understandingthe relationships between rock layers, theirseismic expression and depositional envi-ronments allows more accurate prediction

of reservoirs, source rocks and seals, even ifnone of them intersect the well (above).

Sequence stratigraphy is used mainly inexploration to predict the rock compositionof a zone from seismic data plus distant,sparse well data. It also assists in the searchfor likely source rocks and seals. Expertsbelieve that as more people learn the tech-nique, it will become an exploitation toolfor constraining the shape, extent and conti-nuity of reservoirs.

51

Sequence Stratigraphy, Seismic Stratig-raphy—How Many Stratigraphies CanThere Be?Stratigraphy is the science of describing thevertical and lateral relationships of rocks.1These relationships may be based on rocktype, called lithostratigraphy, on age, as inchronostratigraphy, on fossil content,labeled biostratigraphy, or on magneticproperties, named magnetostratigraphy.

Stratigraphy in one form or another hasbeen around since the 1600s. In 1669,Nicholaus Steno, a Danish geologist work-ing in Italy, recognized that strata areformed as heavy particles settle out of afluid. He also recognized that some stratacontain remnants of other strata, and somust be younger. This conflicted with thewidely held view that all sediments weredeposited during the flood at the time ofNoah. Steno developed three principles thatform the basis of all stratigraphy—youngerlayers lie on top of older layers, layers areinitially horizontal, and layers continue untilthey run into a barrier. His work was neitherwidely publicized nor remembered and isoften credited to James Hutton (1726-1797)or Charles Lyell (1797-1875).

For 300 years after Steno, stratigraphersworked at unraveling the history of theearth, correlating fossils from one continentto another, assigning names, ages and even-tually physical mechanisms to the creationof rock layers. By 1850, most of the majorgeologic time units had been named. By1900, most layers had relative ages, androck types had been associated with certainpositions of the shoreline, which wasknown to move with time. Fine-grainedrocks such as siltstones and shales wereassociated with calm, deep water, andcoarse-grained, sandy rocks with energetic,shallow environments.

At the turn of the century, shoreline move-ment was attributed to tectonic activity—therising and falling of continents. This viewwas challenged in 1906, when EduardSuess hypothesized that changes in shore-line position were related to sea levelchanges, and occurred on a global scale; hecalled the phenomenon eustasy.2 However,Suess was not able to refute evidence pre-sented by opponents of his theory—in manylocations there were discrepancies betweenrock types found and types predicted by sealevel variation.

In 1961, Rhodes W. Fairbridge summa-rized the main mechanisms of sea levelchange: tectono-eustasy, controlled bydeformation of the ocean basin; sedimento-eustasy, controlled by addition of sediments

52

to basins, causing sea level rise; glacio-eustasy, controlled by climate, lowering sealevel during glaciation and raising it duringdeglaciation. He recognized that all thesecauses may be partially applicable, and arenot mutually incompatible. He believedthat while eustatic hypotheses apply world-wide, tectonic hypotheses do not and varyfrom region to region. Fairbridge summa-rized the perceived goal at the time: “Weneed therefore to keep all factors in mindand develop an integrated theory. Such anideal is not yet achievable and wouldinvolve studies of geophysics, geochemistry,stratigraphy, tectonics, and geomorphology,above sea level and below.”3

This brings us nearly to the present. In1977, Peter Vail at Exxon and several col-leagues published the first installments ofsuch an integrated theory.4 Vail developed anew kind of stratigraphy based on ideas pro-posed by L. L. Sloss—the grouping of layersinto unconformity-bound5 sequences basedon lithology—and by Harry E. Wheeler—the grouping of layers based on what hasbecome known as chronostratigraphy.6

Vail’s approach allowed interpreting uncon-formities based on tying together global sealevel change, local relative sea level changeand seismic reflection patterns. Thismethodology, named seismic stratigraphy,classifies layers between major unconformi-ties based on seismic reflection patterns,giving a seismically derived notion of lithol-ogy and depositional setting.

Subsequent seismic stratigraphic studiesin basins around the world produced a setof charts showing the global distribution ofmajor unconformities interpreted from seis-mic discontinuities for the past 250 millionyears.7 An understanding emerged thatthese unconformities were controlled by rel-ative changes in sea level, and that relativechanges in sea level could be recognized onwell logs and outcrops, with or without seis-mic sections. This led to the interdisci-plinary concept of sequence stratigraphy—alinkage of seismic, log, fossil and outcropdata at local, regional and global scales.The integrated theory sought by Fairbridgehad arrived.

This article focuses on the subset ofsequence stratigraphy that includes seismicdata, and so falls under the heading of seis-mic sequence stratigraphy. The techniquehas been shown to work in a variety of set-tings, in some better than others. Attentionhere is on an environment where it hasproved successful—sand and shale deposi-tion on continental margins.

Building BlocksThe concepts that govern sequence strati-graphic analysis are simple. A depositionalsequence comprises sediments depositedduring one cycle of sea level fluctuation—by Exxon convention, starting at low sealevel, going to high and returning to low.One cycle may last a few thousands to mil-lions of years and produce a variety of sedi-ments, such as beach sands, submarinechannel and levee deposits, chaotic flows orslumps and deep water shales. Sedimenttype may vary gradually or abruptly, or maybe uniform and widespread over the entirebasin. Each rock sequence produced by onecycle is bounded by an unconformity at thebottom and top.8 These sequence bound-aries are the main seismic reflections usedto identify each depositional sequence, andseparate younger from older layers every-where in the basin.

Composition and thickness of a rocksequence are controlled by space availablefor sediments on the shelf, the amount ofsediment available and climate. Spaceavailable on the shelf—which Vail calls“shelfal accommodation space”—is a func-tion of tectonic subsidence and uplift and ofglobal sea level rise and fall on the shelf. Forinstance, subsidence during rising sea levelwill produce a larger basin than uplift dur-ing rising sea level. The distribution of sedi-ment depends on shelfal accommodation,the shape of the basin margin—called depo-sitional profile—sedimentation rate and cli-mate. Climate depends on the amount ofheat received from the sun. Climate alsoinfluences sediment type, which tendstoward sand and shale in temperate zonesand allows the production of carbonates inthe tropics.

As an exploration tool, sequence stratigra-phy is used to locate reservoir sands. Indeep water basins with high sedimentationrates, sands are commonly first laid down assubmarine fans on the basin floor (nextpage, “A”) and later as deposits on the con-tinental slope or shelf (next page, “B”). Butas sea level starts slowly rising onto the con-tinental shelf, sands are deposited a greatlateral distance from earlier slope and basindeposits.9 Deposits during this time aredeltaic sediments that build into the basinand deep water shales (next page, “C”). Ifthe sediment supply cannot keep pace withrising sea level, the shoreline migrates land-ward and sands move progressively higherup the shelf (next page, “D”). Once sea

Oilfield Review

nSequence compo-nents in order ofdeposition, frombottom to top. Thesequence beginswhen sea level rela-tive to the basinfloor begins to fall.

The first deposits,sand-rich fans, arelaid down while sealevel is falling to itslowest point (A).

As sea level bot-toms out and beginsto rise (B), sandsand shale aredeposited in fans onthe continentalslope. Submarinechannels with lev-ees may meanderacross the fan.Slumps are common.

The continuingrise in sea level (C)allows wedges ofsediment to buildinto the basin, withsands near theshore, siltstones andshale basinward.

A rapid rise in sealevel (D) movessandiest sedimentslandward as beachesand sandbars.

Sea level thenrises at a lower rate(E), allowing sedi-ments to build basin-ward again. Sandysediments are usu-ally restricted to thenearshore margin.

Colors follow aconvention used byVail. In order ofdeposition, basinfloor fans are yel-low, slope fans arebrown and subma-rine slumps are pur-ple. River deltasthat build out dur-ing low relative sealevel are pink. Rocksassociated withhigher relative sealevel—and usuallyless likely to besandy—are greenand orange. Withinevery part of asequence, sand-prone zones areyellow.

E

DaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaBAaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaCaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa0

0

0

0

0

Submarinecanyon

Basinfloor fan

Channeland levee

SlumpFan lobe

Barrier barDelta

Maximumsea level

Reworked shore sands

Marsh

Minimumsea level

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaJanuary 1993

level reaches a maximum for this cycle,sands will build basinward as long as sedi-ment remains available (left, “E” ). Thesequence ends with a fall in relative sealevel, marked by a break in deposition. Thesequence repeats, however, as long as thereis sediment and another cycle of rise andfall in relative sea level that changes theshelfal accommodation space (see “ADetailed View of Sequence Stratigraphy,”next page).

(continued on page 56)

53

1. For a review:Schoch RM: Stratigraphy: Principles and Methods.New York, New York, USA: Van Nostrand Reinhold,1989.

2. Suess E: The Face of the Earth, vol 2. Oxford, England:Clarendon Press, 1906.

3. Fairbridge RW: “Eustatic Changes in Sea Level,” inAhrens LH, Press F, Rankama K and Runcorn SK (eds):Physics and Chemistry of the Earth, vol. 4. London,England: Pergamon Press Ltd. (1961): 99-185.

4. Vail PR and Mitchum RM: “Seismic Stratigraphy andGlobal Changes of Sea Level, Part 1: Overview,” inPayton CE (ed): AAPG Memoir 26 Seismic Stratigra-phy—Applications to Hydrocarbon Exploration(1977): 51-52.Mitchum RM, Vail PR and Thompson S: “SeismicStratigraphy and Global Changes of Sea Level, Part 2:The Depositional Sequence as a Basic Unit for Strati-graphic Analysis,” in Payton CE (ed): AAPG Memoir26 Seismic Stratigraphy—Applications to Hydrocar-bon Exploration (1977): 53-62.Vail PR, Mitchum RM and Thompson S: “SeismicStratigraphy and Global Changes of Sea Level, Part 3:Relative Changes of Sea Level from Coastal Onlap,” inPayton CE (ed): AAPG Memoir 26 Seismic Stratigra-phy—Applications to Hydrocarbon Exploration(1977): 63-82.Vail PR, Mitchum RM and Thompson S: “SeismicStratigraphy and Global Changes of Sea Level, Part 4:Global Cycles of Relative Changes of Sea Level,” inPayton CE (ed): AAPG Memoir 26 Seismic Stratigra-phy—Applications to Hydrocarbon Exploration(1977): 83-98.Vail P, Todd RG and Sangree JB: “Seismic Stratigraphyand Global Changes of Sea Level, part 5: Chronos-tratigraphic Significance of Seismic Reflections,” inPayton CE (ed): AAPG Memoir 26 Seismic Stratigra-phy—Applications to Hydrocarbon Exploration(1977): 99-116.Sangree JB and Widmier JM: “Interpretation of Depo-sitional Facies From Seismic Data,” Geophysics 44(February 1979): 131-160.

5. An unconformity is a surface separating younger fromolder layers, along which there is evidence of erosionor a significant break in deposition.

6. Sloss LL: “Sequences in the Cratonic Interior of NorthAmerica,” Geological Society of America Bulletin 74(1963): 93-113.Wheeler HE: “Time Stratigraphy,” American Associa-tion of Petroleum Geologists Bulletin 42, no. 5 (May1958): 1047-1063.

7. Haq BU, Hardenbol J and Vail PR: “Chronology ofFluctuating Sea Levels Since the Triassic,” Science 235(1987): 1156-1166.

8. In some cases the unconformity may correlate later-ally with a conformity. A conformity is a surface thatconforms to those above and below it, with no evi-dence of erosion or nondeposition.

9. In the case of the Gulf of Mexico, a 100-m rise in rel-ative sea level would cause a 150-km landward shiftof the shoreline:Matthews RK: Dynamic Stratigraphy. EnglewoodCliffs, New Jersey, USA: Prentice-Hall, Inc. (1984):394.

A Detailed View of Sequence Stratigraphy

The components of depositional sequences are

called systems tracts. Systems tracts are divided

into three groups according to relative sea level

at the time of deposition—lowstand at low rela-

tive sea level, transgressive as the shoreline

moves landward, and highstand at high relative

sea level. Systems tracts are depositional groups

that have a predictable stratigraphic order and

predictable shapes and contents. A close look at

systems tracts, their geometries and lithologies,

shows how sequence stratigraphy can be used to

foretell reservoir location and quality.

Each systems tract exhibits a characteristic log

response, seismic signature and paleontologic

fingerprint, and performs a predictable role in the

oil and gas play—reservoir rock, source rock or

seal. Gamma ray (GR) and spontaneous potential

(SP) logs are expected to read low in sands and

high in shales. Resistivity logs show the reverse,

reading high in hydrocarbon-filled sands and low

in shales.

Apparent layering interpreted on seismic sec-

tions—called stratal patterns—is determined by

tracing seismic reflections to their terminations.

The termination is categorized by its geometry

and associated with a depositional style. Fossils

are described by their abundance, diversity and

first or last occurrence, allowing dates to be deter-

mined based on correlation with global conditions.

Starting with the lower lowstand systems tract

at the bottom of a sequence, basin floor fans are

typically isolated massive mounds of well-sorted

grain flows or turbidite sands1 derived from allu-

vial valleys or nearshore sands (next page, “A”).

Log responses are blocky, with a sharp top and

bottom bracketing clean sand. Seismic reflec-

tions curve down and terminate on the underlying

sequence boundary—a feature called down-

lap—while the top may form a mound. The low-

stand facies makes an excellent reservoir, with

porosity often over 30% and permeability of sev-

54

eral darcies.2 It may be overlain by a thin clay-

rich layer that can act as a seal, but more often it

is overlain directly by the next depositional unit.

In these cases, the basin floor fan acts as a

hydrocarbon migration pathway. Fossil content is

minimal, since deposition rates are often very

high. Basin floor fans derive their hydrocarbon

from previous sequences.

In areas of high deposition rate, the major

component of the lower lowstand systems tract is

the slope fan complex. Slope fans can be exten-

sive and can exhibit several depositional styles,

depending on the vertical gradient of the slope

face and on the sediment source (next page, “B”).

The complex may include submarine channels

with levees, overbank deposits, slumps and

chaotic flows. Log responses commonly are cres-

cent shaped. A sharp base within the crescent

commonly indicates sand in a channel, with a

bell shape indicating fining upward as the chan-

nel is abandoned. On the other hand, channels

may fill with mud. On seismic sections, leveed

channels in the fan show a characteristic mound

with a slight depression in the top. Sand-filled

channels make excellent exploration targets, but

may be difficult to track.3 Sands flowing over

channel levees may be deposited as overbank

sheets and alternate with shales, creating sub-

parallel reflectors. Such sands can provide

stacked reservoirs with porosities of 10 to 30%,

but are usually very thin. Slumps from shelf edge

deltas create a chaotic or jumbled pattern—

“hummocky” in interpreter’s vernacular—easily

identifiable on seismic data. Hydrocarbon

sources for channel and overbank reservoirs are

deeper sequences. Seals are provided by a

widespread “condensed” section of shale, a thin

layer representing prolonged deposition at very

low rates that comes with the rise in sea level.

The sealing shale also contains abundant marine

fossils used for dating.

Part of the upper lowstand, the prograding

wedge complex derives its name from shallow-

ing-upward deltas that build basinward from the

shelf edge and pinch out landward at the preced-

ing shoreline (next page, “C”). Log response

shows more sand higher in the section and less

sand basinward, indicating a coarsening upward.

The seismic signature shows moderate- to high-

amplitude continuous reflectors that downlap

onto the basin floor. This depositional unit often

contains ample sand, especially near the sedi-

ment source. Updip seals are typically poor, how-

ever, and structural trapping is required for hydro-

carbon accumulation.

The transgressive systems tract represents

sedimentation during a rapid rise in sea level

(next page, “D”). The shoreline retreats landward,

depriving the basin of sediment. SP and gamma

ray logs show a fining upward. Retreat of the

shoreline gives rise to seismic patterns that

appear to truncate basinward. In practice, this

systems tract is commonly thin, and such pat-

terns are usually imperceptible on typical seis-

mic sections. Basal transgressive sands derived

from reworked lowstand sands can be excellent

reservoirs, except where shell fragments may

later cement the sands. Shoreface sands will fol-

low strike-oriented trends.

The top of the transgressive systems tract is

the limit of marine invasion, and is called the

maximum flooding surface. Widespread shale

deposition results in a condensed section. Abun-

dant fossils provide ages and well ties across the

seismic section. This clay-rich layer shows low

resistivity and high gamma ray readings. The

seismic pattern of this surface is downlap, which

becomes conformal—parallels adjacent reflec-

tors—basinward, and disappears above the shelf.

This surface is usually a very continuous reflec-

tor. At the shelf edge, it can commonly be identi-

fied by changes in reflection patterns above and

below.

Oilfield Review

nComponents of sequences, their log responses, and predicted and observed seismic reflection patterns.

55January 1993

GRor SP ResistivityaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaABCDE Seismic

reflection patterns Seismic exampleDeposition

cross section

1. A turbidite is a rock deposited from sediment-laden watermoving swiftly down a subaqueous slope.

2. Sangree JB, Vail PR and Sneider RM: “Evolution of FaciesInterpretation of the Shelf-Slope: Application of the NewEustatic Framework to the Gulf of Mexico,” paper OTC5695, presented at the 20th Annual Offshore TechnologyConference Proceedings, Houston, Texas, USA, May 2-5,1988.

3. Sometimes channels can be seen in slices of 3D seismicvolumes, but sequence stratigraphic studies are not com-monly done on 3D data. An exception is the studydescribed on page 62.

Layers deposited during highest relative sea

level are known as the highstand systems tract

(above, “E”). Early highstand sediments are usu-

ally shaly. The late highstand complex, deposited

as the rise in sea level slows, contains silts and

sands. Some late highstand sediments are

deposited in the open air as fluvial deposits.

Gamma ray and SP responses show a gradual

decrease in gamma ray, indicating coarsening

upward associated with decreasing water depth.

Seismic reflections are characterized by sig-

moidal—S-shaped—stratal patterns, similar to

prograding wedge reflections. There may be

deltaic and shoreface sands at the top of the sec-

tion, but in general, this systems tract has poor

reservoir sands, and updip seals are uncommon.

Fossil abundances diminish as the marine envi-

ronment becomes restricted to the deeper parts

of the shelf.

Oilfield Review

Relative sea level

Age, Ma

Layer number

3 2 1 0

0

40

80

0 5 10 15 20 25 30 35

Distance, km

Dep

th, m

0 15

0

100

200

300

400

500

30

Expected gamma ray, SP

and resistivity responses

Lithologies indicated by logs

Gamma rayor SP Resistivity

Sand

Shale

5

10

15

20

25

30

35

56

nSequences simu-lated using ScottBowman’s tech-nique, showing allthe components onpage 53. Inputs areinitial basin shape,sedimentation rateand relative sealevel. Geologic timelines are numberedfrom 5 to 35, oldestto youngest. Thesequence begins at5. Expected logresponses are plot-ted at ten locations.The left curve is SPor gamma ray, theright curve resistivity.

Components of a sequence may berepeated or missing, depending on localconditions and the rate of sea level change,but the basic sequence structure is pre-dictable. Computer-generated models ofsequence cycles are used to show theeffects of sea level change, sediment supplyand depositional profile (below).10 Theseinputs can be varied to test their relativeimportance, or to produce stacks ofsequences in an attempt to match real data.

Searching for Sand—A Case StudySequence stratigraphy was applied in 1992in the East Breaks area, offshore Texas (nextpage, top). Data included a two-dimen-sional (2D) seismic line and logs from sevenwells. The seismic line was processed forstructural imaging (see “Structural Imaging:Toward a Sharper Subsurface View,” page28) and the structural interpretation usedfour other lines to view the basin as a whole(see “Going for the Play: Structural Interpre-tation in Offshore Congo,” page 14). In thiscase, the big picture shows a basin con-trolled by normal faulting to the north,

which was the direction of the sedimentsource. Layers dip and thicken to the south.

Initially, seismic data and logs were inter-preted independently to identify sequencesand their bounding unconformities. Log-derived boundaries were compared withthose from seismic data and the interpreta-tion refined iteratively. Detailed seismicinterpretation began with the most easilyinterpreted reflection patterns, and waspieced together—working upward, down-ward and back toward the wells—respect-ing the stratigraphy suggested by thesequence model.

Logs from wells on the seismic lines wereconverted from depth to time using thenearest check shot—here, 3 miles [4.8 km]away.11 Sands interpreted on spontaneouspotential, gamma ray and resistivity logswere associated with seismic reflections atthe well and tracked along the seismic sec-tion. Shales indicated by logs were noted forcorrelation with fossil data from cuttings.

Next was integration of biostratigraphy.12

Fossils from cuttings help identify and dateboundaries of each rock sequence. Fossil

diversity and abundance are measured ver-sus depth, which is converted to seismictravel time for easy comparison with theseismic section. Fossils of planktonic (float-ing) organisms are more widespread thanthose of benthic (bottom-dwelling) organ-isms and are therefore more useful in estab-lishing regional time correlations. However,in shallow-water environments, benthic fos-sils are used because nearshore conditionsmay be too variable for planktonic fossils.

Fossils are also indicators of relative sealevel. High fossil counts, or peaks, are asso-ciated with shales deposited during low sed-imentation. Such conditions occur in thebasin during time of high relative sea level,but also in deep water between fan depositsand outbuilding delta deposits (page 58,top). Two shale sections are expected withineach sequence, one at the top of the slopefan and the other at the maximum floodingsurface, associated with the furthest land-ward position of the shoreline. Biostratigra-phy also holds the key to paleo-bathymetry—a measure of topography ofthe ancient ocean floor—needed to interpret

57

Fossils

3.0

Dep

th, f

t

3000

4000

5000

6000

Tim

e, s

ec

1.0

2.0aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaA

B

C

D

E

SP

H. sellii

C. macintyrei

G. miocenica

D. tamalis

0 1miles

ForamNanno

nEast Breaks, offshore Texas, seismic line with sequence components interpreted in color, SP log and fossil abundance curves. Blockdiagrams from page 53 point to representative examples on the section. This seismic section has nine sequences. Sequence componentsshown are not necessarily from the same sequence, which is why D appears stratigraphically above E.

T E X A S L O U I S I A N A

East BreaksGreen Canyon

kmmiles0 100

0 161

G U L F O F M E X I C O

January 1993

10. Bowman SA and Vail PR: “Computer Simulation ofStratigraphy,” American Association of PetroleumGeologists Bulletin (1993): in preparation.

11. A check shot is a wireline survey that checks theseismic travel time from the surface to a chosendepth in a well. Depths are chosen from logs. A geo-phone is conveyed by wireline to the desired depthand a seismic source is set off at the surface. Thetravel time is recorded and doubled to compare withthe surface seismic travel time.

12. For an integrated biostratigraphic study:Bell DG, Selnes H, Bjorøy M, Grogan P, Kilenyi Tand Trayner P: “Better Prospect Evaluation withOrganic Geochemistry, Biostratigraphy and Seismics,”Oilfield Review 2, no. 1 (January 1990): 24-42.

13. Posamentier HW and James DP: “An Overview ofSequence Stratigraphic Concepts: Uses and Abuses,”in Posamentier HW, Summerhayes CP, Haq BU andAllen GP (eds): Stratigraphy and Facies Associationsin a Sequence Stratigraphic Framework, Interna-tional Association of Sedimentologists Special Publi-cation, 1992.

14. Vail P, Audemard F, Bowman SA, Eisner PN andPerez-Cruz C: “The Stratigraphic Signatures of Tec-tonics, Eustasy and Sedimentology—An Overview,”in Einsele G et al. (eds): Cycles and Events in Stratig-raphy. Berlin, Germany: Springer-Verlag (1991):617-659.

the depositional environment. Paleodepth isderived from benthic fossils with knowndepth habitats (next page top, right curve).Knowing water depth helps to interpret deepor shallow water rock types and expectedlayer thicknesses.

Once seismic, log and biostratigraphicdata are combined, a final, color-codedinterpreted section is made. Very highamplitude reflections may be highlightedwith hatching. These so-called bright spotsare analyzed for anomalies in amplitudevariation with offset associated with hydro-carbons (see “Hydrocarbon Detection WithAVO,” page 42). In the East Breaks exam-ple, the most promising prospect is a large,sandy basin floor fan. Shales interpretedabove and below could provide seal andsource rock, respectively.

Overcoming Limitations of SequenceStratigraphySequence stratigraphy has proven useful forpetroleum exploration, but it is commonlymisapplied.13 There is controversy overwhether the technique can be applied tocarbonate systems since it was designed toexplain sand-shale systems. Some expertsmaintain that sequence stratigraphy is easierin carbonates because carbonates areextremely sensitive to sea level change.14

There is unanimous agreement, however,that low sedimentation rates often pose spe-cial problems. When sedimentation rate ismoderate to high, layers within a sequenceare tens to hundreds of meters thick, com-fortably within the resolving power of a typ-ical seismic wave (next page, bottom). But

when sedimentation rate is low, severalsequences might fit within a seismic wave-length. Sequence stratigraphy cannot beconfidently applied here, but it has beendone countless times. A useful interpretationin thinly-bedded regions requires abandon-

Oilfield Review58

nTypical Gulf ofMexico fossil abun-dance peaks andpaleodepth curve.Fossil abundancecurves based onanalysis of cuttingsfor foraminifera—protozoa with cal-careous externalskeletons—andnannofossils, abroad category ofextremely small,usually algal fossils.Peaks indicate thepresence of shalesat the top of theslope fan complex(brown) and atmaximum sealevel (green). Theright curve indi-cates fossil habitatdepth, in whichdark blue is deepwater and lightblue is shallower.

nHigh (top) and low (bottom) sedimentation rates, shown on synthetic sequences. Thebasic shape of each layer follows the initial basin shape, but layer thickness varieswith sedimentation rate. Only thick layers can be resolved with seismic methods.

Plankton

Foramabundance

Nannofossilabundance

Paleodepth

Dep

th, f

t

G. miocenica

H. emaciatum

H. sellii

C. macintyrei

D. brouweri

D. brouweri “A”

D. pentaradiatus

D. tamalis

(?) R. pseudoumbilica

3000

5000

7000

9000

11000

13000

1 6 30ft x100

Distance, km

Dep

th, m

Dep

th, m

High sedimentation rate

Low sedimentation rate

0

600

0

0 30 60

0

600

30 60

ing small-scale features and concentratingon larger scale, longer term processes thatcontrol the generation of sequences.

With this in mind, Vail and coworkersproposed a hierarchy of stratigraphic cyclesbased on duration and amount of sea levelchange.14 Duval and Cramez at TOTALExploration worked with Vail to providesubsurface examples and to expand theapplication to hydrocarbon exploration.15

The hierarchy assigns frequencies to themechanisms of eustasy enumerated by Fair-bridge (page 52), viewed in light of platetectonics (next page, top). The first-ordercycle, which is the longest, tracks creationof new shorelines resulting from thebreakup of the continents. Although thisbreakup does not follow a cycle, it has hap-pened twice, with a duration of over 50 mil-lion years. The second-order cycle is land-ward and basinward oscillation of theshoreline that lasts 3 to 50 million years.This oscillation is produced by changes inthe rate of tectonic subsidence and uplift,caused by changes in rates of plate motion.Both first- and second-order cycles maycause changes in the volume of the oceanbasins resulting in long-term variations inglobal sea level. The third-order cycle is thesequence cycle, lasting 0.5 to 3 millionyears. Fourth- and higher order cycles maybe correlated with periodic climatic changes.

The following example, with its lowdeposition rate, approaches the limit ofinterpretation in terms of third-order cycles.It comes from the Outer Moray Firth basinin the UK sector of the North Sea, where theinitial basin shape, tectonic activity andvariation in the rate of deposition add atwist to the interpretation (pages 60-61).

Stratigraphic interpretation of the last 65million years of sediments in the OuterMoray Firth is more difficult than in the Gulfof Mexico because slower deposition in theCentral North Sea resulted in thinner units,many of which cannot be resolved by seis-mic waves. During this period, the OuterMoray Firth has 17 sequences totaling 5000feet [1524 meters] of sediments (pages 60-61, middle and bottom), compared to theGulf Coast, with 10 sequences totaling 9000ft [2750 m]. In the North Sea, however,depositional processes juxtaposed a varietyof lithologies, providing reliable calibrationpoints for accurate conversion of logs from

59

Distance

1 order Continental breakup

2nd order— Movement of shoreline

3rd order— Sequence cycle

4th+ order— Parasequence cycle (periodic)

> 5

0 M

a3-

50 M

a0.

5-3

Ma

0.01

-0.5

Ma

Basinward

Landward

Incr

easi

ng ti

me

Incr

easi

ng th

ickn

ess

nHierarchy ofcycles, decreasingdownward in theduration of sealevel change andin area of influence.At the top, the first-order cycle lasts atleast 50 millionyears and iscaused by majorchanges in the con-figuration of tec-tonic plates. Thesecond-order cycle,lasting 3 to 50 mil-lion years, is alsocontrolled by platemotion. It involvesmovement of theshoreline landwardand basinward, onthe scale of conti-nents. The whitearea shows thetime in which thereis no rock record,usually due to lackof depostition. Thethird-order cycle, of500,000 to 3 millionyears, is thesequence cycledescribed in themain text. It iscaused by long-term tectonic activ-ity and short-termglobal sea levelchanges. Fourth-and higher ordercycles, of 10,000 to500,000 years, areof shortest duration.They are driven bysea level changescaused by periodicclimatic variation.Cycles of this orderare called parase-quence cycles.

January 1993

15. Duval B, Cramez C and Vail P: “Types & Hierarchyof Stratigraphic Cycles,” presented at the Interna-tional Symposium on Mesozoic and CenozoicSequence Stratigraphy of European Basins, Dijon,France, May 18-20, 1992.

Tim

e, s

ec

0

1.0

2.0

A BW E

Synthetic Gamma ray

0 5km

nSynthetic seismo-grams and gammaray logs from twowells tying with theOuter Moray Firthseismic line. Syn-thetics, based onsonic and densitylogs, provide adepth-to-time corre-lation for integra-tion of log, paleoand seismic data.(Courtesy of Amoco UK.)

depth to time using synthetic seismograms(right, bottom). In the Gulf of Mexico, thisconversion is typically done with onlynearby checkshots; sands and shales com-monly show periodic alternation with depthat wavelengths that make comparisonsbetween seismic sections and synthetic seis-mograms nonunique.

Stratigraphic study is always preceded bystructural interpretation. In addition, a pale-ogeographic interpretation of the OuterMoray Firth shows that late in the Creta-ceous period—when the sequences understudy began to be deposited—a smoothbasin floor sloped gently from northwest tosoutheast. During a relative fall in sea level,sediments were deposited as slope fans.Their seismic expressions indicate lobeswith channels and some chaotic flows—large-scale slumps with jumbled seismiccharacter. As sea level rose, a wedge of out-building deltas was deposited. Sea levelmaximum is associated with a depositionalhiatus, shown only as a thin line. Depositssynchronous with this surface may be foundon what is now land in Europe, but in thebasin, sediments that correspond to periodsof high relative sea level are rare.

Why are elements of the classic Vailmodel missing from this sequences in thisbasin? One explanation is the competinginfluences of tectonic uplift and sea levelchange. As global sea level rose and fell,continual regional uplift kept the sea fromreaching levels high enough to allow forma-tion of units typical of high relative sealevel. Only once, at the top of the thirdsequence, does a thin layer of high relativesea level sediments appear (orange).Another interpretation is that thin, high rela-tive sea level sediments were deposited, buteroded, and so are not preserved in the sec-tion (pages 60-61, bottom).

Tim

e, s

ecTi

me,

sec

Raw seismic

Interpreted seismic

Geologic cross section

0

4

0

4

0

W

6

This section can also be interpreted interms of second-order cycles (page 62, top).The entire set of 17 depositional sequencescan be bracketed by five second-ordercycles, based on physical stratigraphy andbiostratigraphy. Major biostratigraphic gapsexist at the boundaries of each second-ordercycle, and the boundaries can be seen torepresent major changes in the depositionalstyle of basin fill.

Studies in 3DIf the volume of earth in a study area issmall enough, workstations can add a newdimension to sequence stratigraphy. In theGreen Canyon area of the Gulf of Mexico,interpreters concentrated on a fan depositedin a syncline on the continental slope 1 to 2million years ago. Regional sequencestratigraphy was established using 2D seis-mic data and paleontologic control from sixnearby wells. Zooming in on a subset of thisdata, interpreters assembled a series of 2Dpanels for 3D interpretation.

The top and base of the slope fan wereinterpreted over a six-block area (54 squaremiles [138 km2]). The thickest part of theslope fan coincides with the stacked chan-nels that carried shallow-water shelf anddelta sands into deep water, greater than200 m [656 ft] (page 62, middle). A series ofstacked channels, possibly filled with sand,is visible within the slope fan interval.

The goal of this interpretation is to identifyexploration targets. Although lithology of thechannel deposits is difficult to identify in thehorizon slice, geology predicts that the chan-nel will terminate in a sand-rich fan.16 Thechannel was tracked south, and a fan wasdiscovered in the next block of seismic data.

Dep

th, m

7000

16. A horizon is the surface separating two rock layers.In seismic data, a horizon shows up as a reflection.A reflection tracked in a 3D cube of seismic dataand displayed in plan view is called a horizon slice.Depths or times to the reflection are contoured orcolor-coded.

Oilfield Review0

E

0 10km

61January 1993

nComparison of seismic and log data from the Moray Firth, North Sea, including the original seismic line (top), the interpreted line (mid-dle) and the geologic cross section with gamma ray logs. Permian and Carboniferous basement (green) are followed by Jurassic toLower Cretaceous sediments (purple). Both show evidence of rifting during and after deposition. In the Upper Cretaceous, chalk (blue)filled the earlier rifts. The chalk was deposited in open marine conditions, with no land exposed nearby. On the left (west) edge, a Pale-ocene (Danian) chalk debris flow appears as a chaotic zone on top of the earlier chalk. Basin relief was minimal and sea level was highat the onset of the first sequence in the Tertiary. Low relative sea level deposits—slope fans and chaotic flows—are brown (color conven-tion is the same as in the Gulf of Mexico example). River deltas building out during low relative sea level are pink. Sea level maximaappear as thin green lines.

E

Cross section

Horizon slice

17. Posamentier HW, Allen GP and James DP: “HighResolution Sequence Stratigraphy—The East CouleeDelta, Alberta,” Journal of Sedimentary Petrology 62(1992): 310-317.

18. Jacquin T, Garcia J-P, Ponsot C, Thierry J and Vail PR:“Séquences de Dépôt et Cycles Régressif/Transgres-sifs en Domaine Marin Carbonaté: Exemple du Dog-ger du Bassin de Paris,” Contes Rendues del’Académie des Sciences de Paris 315 (1992): 353-362.

19. Galloway WE: “Genetic Stratigraphic Sequences inBasin Analysis I: Architecture and Genesis of Flood-ing-Surface Bounded Depositional Units,” AmericanAssociation of Petroleum Geologists Bulletin 73, no.2 (February 1989): 125-142.

20. Sloss LL: “Tectonics—The Primary Control onSequence Stratigraphy: A Countervailing View,” Dis-tinguished Lecture Tours Abstracts, American Asso-ciation of Petroleum Geologists Bulletin 74, no. 11(November 1990): 1774.

21. Hag BU: “Ten Commandments for Sequence Stratig-raphers,” presented at the International Symposiumon Mesozoic and Cenozoic Sequence Stratigraphyof European Basins, Dijon, France, May 18-20,1992.

T E X A S L O U I S I A N A

East BreaksGreen Canyon

kmmiles0 100

0 161

G U L F O F M E X I C O

W E

LowerPaleocene Upper

Paleocene

Lower OligoceneMiddle EoceneLower Eocene

nOuter Moray Firth data reinterpreted in terms of second-order cycles. Basinward movement of the shoreline is orange and landwardmovement is green. Colors correspond to second-order cycles (page 59, top).

nSeismic line and horizon slice from acube of 3D data in Green Canyon, offshoreLouisiana. The top (yellow line) and bottom(red) of a slope fan were interpreted usingsequence stratigraphy. A stack of subma-rine channels can be seen in the seismicline (oval). Normally, channels like thesewould be difficult to track, but in 3D thetask is simple. The 3D data can be flat-tened at the top of the fan, and sliced hori-zontally to reveal a map (bottom). Here, thechannel can be seen to meander fromnorth to south around an emerging saltdome. A slump off the flank of the salt (cir-cle) has fallen into the channel. This chan-nel was tracked to the next block south,where it terminated in a slope fan lobe,predicted to be sandy.

62

Salt

FrontiersSequence stratigraphy continues to evolve.One area of investigation is high-resolutionsequence stratigraphy, which is performedat a higher resolution than seismic wave-lengths, usually with log and outcrop stud-ies. ARCO, TOTAL and Esso scientists per-formed a very high-resolution sequencestratigraphy study on a roadside ditch,which served as an analog of an incised val-ley and delta system. The system measuredonly 50 cm [20 in.] from bottom to top butobeyed the same physical laws as systemshundreds of meters thick.17

Sequence stratigraphy is achieving somesuccess in areas where it was not designedto work. Studies in carbonates show thatalthough the depositional layering patternsare different from sand-shale systems, thetechnique has the power to explain and pre-dict sediment distribution and lithologiccontent.18 Sequence stratigraphy has alsobeen applied with success to nonmarinedeposits in continental basins and in marinebasins isolated from continental sediments.

Sequence stratigraphy, as proposed by theExxon group, is not without controversy.Some alternative schemes to explainsequences place more emphasis on sedi-ment supply19 or tectonic activity.20 What-ever their area of expertise, stratigraphersagree on the main problem in sequencestratigraphy—overextending the model to fitevery study area. Bilal Haq, while at theNational Science Foundation in the USA,compiled ten commandments for sequencestratigraphers21 and Henry Posamentier ofARCO Oil and Gas Company has writtenon the uses and abuses of the technique.13

They approach the subject from differentperspectives, but their message is thesame—where sequence stratigraphy works,use it; if it doesn’t work, the problem is withthe application, not the theory. —LS

Oilfield Review