48 Oilfield Review

There was a time when exploration con-sisted simply of following surface signs suchas seeps, creek beds and salt domes, anddrilling where the party boss poked his stick;but those days are gone. Petroleum seismol-ogy has revolutionized the search for hydro-carbons and brought about a period ofremarkable discoveries. Seismic explorationhas expanded dramatically with the tandemadvent of 3D seismic technology and pow-erful computer capabilities. Exploration hasbeen a race at sea and on land for evergreater efficiency. The current limits of con-temporary technology were apparentlyreached last year, however. In marine seis-mic acquisition, eight streamers formed750-meter [2460-ft] wide swaths and10,000-meter [32,800-ft] offsets, and in landseismic acquisition, very large-channelcapacity, 24-bit recording systems andthree-component sensors were employed.

Computing power has also grown expo-nentially, with massively parallel computerssuch as the CM-5 (576 nodes and over 74 gigaflops of power) cutting processingtime to a tenth of what it was just ten years ago. Sophisticated software systemsare now integrating the process flow soseamlessly that engineering, petrophysical,geological and geophysical data can bemerged at the explorationist’s workstation tofacilitate interpretation.

Despite these tremendous technologicalstrides, easy discoveries have become athing of the past. All of the world’s moreobvious reservoirs have been found, andmost have been tapped and put into produc-tion. Those in accessible areas that havebeen the easiest to identify, the hydrocarbonstrapped in structural faults and anticlines thatmanifest themselves so readily in 2D and 3Dseismic data, are almost all in production, asare the stratigraphically entrapped reservoirsthat revealed themselves as bright spots inotherwise lackluster seismic data.

Today, fully 40% of the oil found in maturehydrocarbon provinces is being found instratigraphic traps.1 In fact, most of theremaining oil and gas in these areas probablylie in such traps, making future explorationdeliberate searches for these elusive geo-logical units. Worldwide, except for thoseprincipally structural accumulations in Russiaand the Persian Gulf, some 60% of knowngiant oil and gas fields (500 million bbl [79 million m3] oil and 3.5 Tcf [4.7 billionm3] of gas) are stratigraphic. These includeVenezuela’s Bolivar Coastal with 30 billionbbl [4.9 billion m3], the 6 billion-bbl [1 bil-lion-m3] East Texas field, and Mexico’s PozaRica field, with 2 billion bbl [0.3 billion m3].Furthermore, as exploration presses fartherinto the world’s deepwater provinces, theprospect is frequently a stratigraphic trap,too—the Shell Mars field, with 700 millionbbl [111 million m3], for example, and Ram-Powell with 250 million bbl [38 million m3].

Exploring for Stratigraphic Traps

Jack CaldwellAbu ChowdhuryPeter van BemmelHouston, Texas, USA

Folke EngelmarkLars SonnelandStavanger, Norway

Norman S. NeidellZydeco Energy, Inc.Houston, Texas

Discovery of new resources in mature provinces and deep offshore

demands new technology to find the most elusive of petroleum

reservoirs, stratigraphic traps, which may well hold the vast majority

of the new millennium’s yet-to-be-discovered hydrocarbons.

For help in preparation of this article, thanks to ColinHulme and Dick Ireson, Geco-Prakla, Gatwick, England;Bobbie Ireland and Rutger Gras, GeoQuest, Houston,Texas, USA; Greg Leriger, Geco-Prakla, Houston, Texas;Evelyn Medvin, Coherence Technology Company, Hous-ton, Texas; and Roy Nurmi, Schlumberger Wireline &Testing, Houston, Texas. GeoViz, MultiWave Array, Nessie and SCT (Seismic Clas-sifier Toolbox) are marks of Schlumberger. CoherenceCube is a mark of Coherence Technology Company.SUMIC is a mark of Statoil (Den Norske Stats Oljeselskap).CM-5 is a mark of Thinking Machines Corporation.

Winter 1997 49

Regardless of the method of exploration,today’s undiscovered stratigraphic traps aredifficult to find and are frequently encoun-tered purely by accident while drilling step-out wells to further delineate a field. Thatnotwithstanding, with ever greater success,the technological advances in the geo-sciences that helped reveal so many ofyesterday’s hydrocarbons are now being putto the task of identifying these less dis-cernible reserves. These lie in isolated depo-sitional units beneath seismic data’s hillsand valleys, the pinchouts, unconformities,reefs and barrier bars collectively known asstratigraphic traps (above).

Structural Versus Stratigraphic TrapsHydrocarbons migrate upward from theirsource through porous subterranean stratauntil their route is blocked by a layer ofimpermeable rock, and they accumulatebeneath the sealing body, structure or trap.Geologists divide these traps into two types,structural and stratigraphic. Structural trapsare formed by tectonic forces after the sedi-mentary rocks have been deposited. Theygenerally fall into two categories: anticlines,where the rock has been folded or bentupward, and faults, where movement alonga joint or fracture has driven an imperme-able layer above a permeable layer.

Stratigraphic traps, on the other hand, aremost often formed at the time the sedimentsare deposited. They fall into three cate-gories: pinchouts, most common in stream

environments where a channel through aflood plain has been filled with permeablesand that was then surrounded by less per-meable clays or silts when the channelmoved; unconformities, where a permeablereservoir rock has been truncated and cov-ered by an impermeable layer following anondepositional period or a time of erosion;and carbonate reefs, fossilized coral con-structions and associated deposits that arosefrom ancient ocean shelves and were over-lain by layers of both permeable and imper-meable rock (see “Types of StratigraphicTraps,” page 51).

Pinchout and unconformity traps are oftenfound in sand-shale beds in basins thathave experienced considerable tectonicactivity, where there are unconformities andoverlaps as well as marine, coastal and flu-vial facies. Reefs, however, are frequently

found either on stable shelves beside troughscontaining fine clastic sediment or in evap-orite basins. Since reef traps exhibit ananticlinal structural expression, their delin-eation with traditional seismic methods isordinarily accomplished with relative suc-cess, unlike pinchouts and unconformities,whose physical subtleties have evadedexplorationists until recently. For this rea-son, we are focusing mainly on nonreefstratigraphic traps. We examine why theyhave been so difficult to identify, and thendescribe the seismic acquisition and pro-cessing techniques that have been devel-oped to bring these features to light.

Why Are They Difficult to Find?Stratigraphic traps were first identified inPennsylvania in 1880, but exploring for themremained a mystery until the mid-1930s,when seismology began to be applied in thepetroleum industry.2 Explorationists find struc-tural traps far easier to identify than strati-graphic traps in both 2D and 3D seismic data

■■3D image of a stratigraphic trap insouth Texas. This three-dimensionalGeoViz visualization highlights ameandering channel system. Datashown include 2D seismic lines(multlcolor vertical sections) andwell logs (green and tan curves).Gas-saturated sands (red) andwater saturation (blue) are clearlyseen. A productive well was accurately positioned on the basisof this interpretation.

2. For background on early exploration for stratigraphictraps: Halbouty MT: “The Time is Now for All Explo-rationists to Purposefully Search for the Subtle Trap,”and other contributions to AAPG Memoir 32: TheDeliberate Search for the Subtle Trap. Tulsa, Okla-homa, USA: American Association of Petroleum Geol-ogists (1982): 1-10.

Erosional channel base

Channel infill sequences

Transgressive channel top

50 Oilfield Review

because structural traps are seen as highlydipping reflections and discontinuities in oth-erwise smooth reflections. Thus for years,acquisition and processing techniques havebeen tailored to accentuate these features,allowing interpreters to concentrate theirefforts on faults, anticlines and reef formationsrather than their more subtle stratigraphiccounterparts. Other methods, such as detect-ing bright spots (reflections with anomalouslyhigh amplitudes), are used with some successon both structural and stratigraphic traps. Cer-tainly, many bright spots do illuminate strati-graphic traps, which make these traps so eas-ily identified with today’s techniques thatexperts say few await discovery (right). But,until recently, the search for the more elusiveunconformities and pinchouts has beenthwarted by limitations in seismic data,namely low resolution, noise and a multitudeof technological barriers—the need to expandbandwidths, attenuate multiples (false reflec-tions), reduce differences of scale betweenseismic data and well logs, bring core and logmeasurements into harmony, and improvesynthetic seismograms.

Stratigraphic traps are generally visuallysubseismic, so thin or so conformable totheir surrounding geometry that their sub-tleties are nearly invisible in traditional seis-mic data. The detail that can indicate astratigraphic trap in seismic data may wellbe just a small part of a single seismic trace,perhaps only a small bump on an otherwisesmooth wiggle or a change of curvature orslightly different wave shape that wasn’t pre-sent earlier.

Detecting these subtleties requires data ofthe highest possible quality. For these pur-poses, high quality means large bandwidthor a wide range of frequencies to resolvesmall features, and low noise. Yet somestratigraphic traps generate the very noise inthe seismic data that makes them hard tosee. For example, some stratigraphic trapsare associated with unconformity surfacesor surfaces of major velocity contrast.Reflections off these surfaces can reverber-ate or reflect multiple times, giving rise tothe type of seismic noise called multiples.The presence of these multiples in data canconceal the trap and make it difficult todefine its exact depth. Frequently, however,there is no strong reflection associated witha stratigraphic trap. Instead, the trap is asso-ciated with a gradual change in lithologyinstead of an abrupt contrast. In these cases,a more typical seismic signature would besmall changes in the character of a reflec-tion from one trace to another (right).

■■Bright spot in the seismic record. A bright spot anomaly is a high-amplitude seismicevent. On the left panel, the bright spot shows up in the center as a black reflection ofpositive amplitudes. On the right panel, the same bright spot shows up as a seismicattribute, here in pink and red.

■■A channel-formed stratigraphic trap, the Flounder formation, in Gippsland basin, off-shore Australia with diagnostic features indicated. Note the sigmoid pattern, the truncatedseismic events (arrow) indicative of an unconformity, and the amplitude changes fromone trace to another (between the chevrons).

Winter 1997 51

Stratigraphic traps were first recognized in 1880 by

Carll, but not so named until 1936 by Levorsen,

who defined them simply as traps “in which a vari-

ation in the stratigraphy is the chief confining ele-

ment in the reservoir which traps the oil.” He

noted that, “the dominant trap-forming element is

a wedging or pinching-out of the sand or porous

reservoir rock, a lateral gradation from sand to

shale or limestone; an uplift, truncation, and over-

lap, or similar variation in the stratigraphic

sequence,” to differentiate them further from

structural traps.1 Structural traps, conversely, are

usually formed by tectonic forces after sedimen-

tary rocks are deposited and include anticlines or

folds and faults (right).Stratigraphic traps include pinchouts, in which a

lens of permeable sand is surrounded by less per-

meable silts and clays. These form in both land

and submarine stream environments. Here sand

is deposited in the stream channel and in coastal

settings when beach sediments are covered by

impermeable clays that are laid down if sea level

rises. Unconformities represent a gap in the geo-

logical record due to erosion, nondeposition, or

both (right). The reservoir seal may be created by

alteration of the exposed portion of the reservoir

rock itself, or by deposition of a later impermeable

layer. Reefs, either mound or shelf-margin carbon-

ate units, form by growth of coral and deposition of

calcite precipitated from seawater.

Pinchout

Fold Fault

ReefUnconformity

Structural

Stratigraphic

Deposition Uplift and tilting

Uplift

Erosion

Deposition

Angular unconformity

■■The development of an unconformity. An unconformity is typically created through a process of deposition,uplift and titling, erosion and redeposition.

■■Types of Traps. Structural traps are generally classified as either anticlines or folds, or faults. Stratigraphictraps are generally categorized as pinchouts, unconformities and reefs.

Types of Stratigraphic Traps

1. Both citations by Gordon R: “Stratigraphic TrapClassification,” in Payton CE (ed): AAPG Memoir 26,Seismic Stratigraphy—Applications to HydrocarbonExploration. Tulsa, Oklahoma USA: American Associationof Petroleum Geologists (1997): 14-27.

52 Oilfield Review

Fortunately, many reservoirs formed bystratigraphic traps also have a structuralcomponent that makes them easier to dis-cover. They are found to have stratigraphicelements many years later, when cumulativeproduction volumes surpass original esti-mates.3 In the North Sea, Gulf of Mexicoand Campos basin of Brazil, giant strati-graphic traps are only recently coming tolight after decades of exploitation of smallerpredominately structural traps, suggesting abright future in exploration for stratigraphictraps in nominally mature provinces.

Because they are so hard to see, explora-tion for stratigraphic traps requires knowingwhere to look. It is fundamental to under-stand the structural setting and important,even if difficult, to recognize the deposi-tional setting of a prospect to know where toanticipate the occurrence of stratigraphictraps.4 The application of seismic stratigraphyto existing seismic data and well logs aids in

this understanding by providing neededinformation on the direction of sedimentaryflow, whether the sea level was rising orfalling, and other depositional conditions.5

Designing for DiscoveryExplorers don’t select a drillsite based onintuition and the whim of the party boss anymore. Nor are today’s seismic shoots simpleexercises in geometry; considerable geologi-cal and geophysical input goes into theplanning of a modern survey.6

To reduce the risk of an expensive dryhole, as much a priori knowledge of thelocation as possible is factored into thedevelopment of a stratigraphic model fromwhich to design a new, comprehensive 3Dseismic acquisition program. This prelimi-nary model indicates not only the targetdepth for the survey, but when a strati-graphic trap is the target, provides anapproximation of how large the trap is, sothat the final survey can render the desired

subsurface coverage. All known geologicallayers are included in the model, as are theirvelocities and densities. With ray tracing or,in the case of a targeted stratigraphic trap,applying the full-wave equation, a seismicsurvey can be simulated to optimize themeasurements and fine-tune the surveydesign (left).

The technology for both onshore and off-shore exploration for stratigraphic traps hasbeen in existence for 20 years, beginningwith sequence or seismic stratigraphy, then3D acquisition and its higher resolution ofthe details in seismic data. Offshore, how-ever, advances in technology have recentlybeen changing the way acquisition is beingperformed in deep water and matureprovinces. There, to alleviate the two mosttroublesome problems associated withexploring for stratigraphic traps, low-resolu-tion data and related multiples, the survey isconducted in one of two ways. Towed multi-streamer surface acquisition with longstreamers (from 3 to 12 streamers, 5000 m[16,400 ft] to 10,000 m long) can yieldwider bandwidths for higher resolution andobtain offsets long enough to attenuate mul-tiples (next page, top). And, more recently,four-component (4C) ocean-bottom cable(OBC) acquisition has become feasible,which not only achieves greater resolutionin the resultant seismic data, but providesmore reliable information on the lithologyand porosity of the target than can beobtained by towed streamers.7

Shear EnlightenmentAlthough the resolution and signal qualityachieved by towed acquisition haveimproved the ability to image stratigraphictraps, 3D seismic exploration is even fur-ther enhanced by employing the new tech-nology of four-component seabed systems.Geco-Prakla took the concept of a seabedsystem with external hydrophones andgeophones, originated by Statoil as SUMIC,and developed its new seabed system as anentirely self-contained cable with internalhydrophones and geophones, the Nessie4C MultiWave Array system (next page, farright). The four components comprise threeorthogonally mounted geophones and onehydrophone mounted within the ocean-bottom cables that are laid in direct con-tact with the seafloor.

This form of deployment allows measure-ment of rock properties that towed or verti-cal hydrophones cannot measure, sincehydrophones record only compressional (P)waves. The vast majority of surveys, includ-ing those recorded on land, rely entirelyupon P waves to obtain a seismic image, butappropriately deployed geophones, onshore

■■Synthetics for a survey design in the Gulf of Mexico. A synthetic seismogram shows thecorrelation between log and seismic data and can indicate how well a stratigraphic tar-get will be mapped. Here the most important tracks are tracks 3 and 4, representing thesynthetic and the seismic data (repeated for clarity). The impedance log in track 2 iscreated from the velocity log in track 1 and the density log in track 8. Track 6 indicatesthe semblance, or accuracy, of the match between the synthetic and seismic data, withgreen representing high semblance. Track 5 is the recorded seismic line and includes thewell trajectory (vertical red line)—which missed both red bright spots. SP and porositylogs are shown in tracks 7 and 8, respectively.

Winter 1997 53

or offshore, also record shear (S) waves,which react differently to the properties ofthe rock they penetrate (see “Full-WaveSpectrum: P and S Waves,” next page).

Because stratigraphic traps are not neces-sarily associated with structural events andare not domal in shape, they are often invis-ible in P-wave data, but are equally as oftenreadily identified in combination P and Sdata. This is because shear waves havecomplementary information to that of the P waves, which allows more complete char-acterization of the elastic properties of therock and fluid and allows identification ofthe subtle changes in lithology that comewith stratigraphic traps.

Because shear waves do not travel throughwater, the cable must be in direct contactwith the seabed to accomplish the 4C shoot,a deployment that requires precise position-ing of the cables as they are laid from theback of the acquisition vessels. Once a seis-mic traverse has been shot, the cable iseither dragged or taken in and redeployed ina new location; additional lines of seismicdata are acquired until the entire area hasbeen covered.

The information obtained in a four-compo-nent survey complements that from P wavesin eight ways. First, because S waves are rel-atively unaffected by pore fluids, includinggas, they can be used to obtain structuraland stratigraphic information in areas wherethe presence of such fluids precludes coher-ent images from P waves only. Targets belowgas chimneys and gas clouds are notori-ously difficult to image with P waves, and S waves have been highly successful in this application.

■■Seafloor cable. The Geco-Prakla Nessie4C cable system is reeled out for deploy-ment from surface vessels.

■■A wide array acquisition. Back deck of the seismic acquisition vessel GecoTopaz conducting a survey with a wide array of long streamers.

3. Pettingill HS: “A Historical Look at Worldwide Tur-bidite Production: Future Opportunities in Strati-graphic Traps,” presented at the 1997 AAPG AnnualConvention, Dallas, Texas, USA, April 6-9, 1997,abstract A92.

4. Brown LF and Fisher WL: “Seismic-Stratigraphic Inter-pretation, Depositional Systems,” in Payton CE (ed):AAPG Memoir 26, Seismic Stratigraphy —Applicationsto Hydrocarbon Exploration. Tulsa, Oklahoma, USA:American Association of Petroleum Geologists (1977):213-248.

5. Mulholland JW: “Sequence Stratigraphy: Basic Ele-ments, Concepts, and Terminology,” The Leading Edge17 (January 1998): 37-40.

6. Ashton CP, Bacon B, Mann A, Moldoveanu N,Déplanté C, Ireson D, Redekop G and Sinclair T: “3DSeismic Survey Design,” Oilfield Review6, no. 2(April 1994): 19-32.

7. For a discussion of the system: George D: “Multicom-ponent Seismic Acquisition Complements Conven-tional Towed Streamer,” Offshore Magazine 57, no. 10(October 1997): 70-72.

54 Oilfield Review

Second, S waves yield independent infor-mation about rock properties, allowingmore complete prediction of both fluid typeand rock lithology. When only P waves arerecorded in a seismic section, it is frequentlydifficult to discern whether a detected directhydrocarbon indicator event is due to thepresence of hydrocarbons or is simply dueto lithologic changes. Shear-wave datalessen this difficulty considerably whenacquired simultaneously with P-wave data.In areas where P-wave data produce ampli-tude anomalies and the S-wave data do not,the presence of hydrocarbons is likely. If theanomaly is observed in both S-wave and P-wave data, it is most likely either a diage-netic or lithological phenomenon. The ratioof P-wave to S-wave velocities, VP/VS, isoften used to predict lithology, and the P-wave amplitude to the S-wave amplituderatio may turn out to help predict fluid-saturation differences.

Third, in deepwater acquisitions, the triple-sensor nature of the geophone data providesa unique opportunity for demultiple proc-essing. Shear waves contribute anothervelocity function to use in distinguishingbetween multiples. Since shear waves pro-duce multiples themselves that are similar tothose produced by P waves, this helps theinterpreter sort the principal reflections fromthe multiples to image more accurately.

Fourth, by acquiring both P and S waves,explorationists expect to illuminate shadowzones beneath high-velocity salt structures.Because the raypaths bend at a boundary oftwo different velocities, and salt bodiesoften have irregular boundaries, shadowzones, areas with no reflections, arecreated. By using both P and S waves, someof the shadow zones of one are illuminatedby the other. Therefore, better structural and stratigraphic images can be achievedbeneath salt.

Fifth, variations in seismic velocities, bothP-wave and S-wave, may help identify dif-ferent lithologies. The use of interval travel-time ratios, Ts/Tp—related to but easier tomeasure than interval Vp/Vs—over a rela-tively small time window in the seismicdata, may indicate lateral or verticalchanges in lithology and pore fluid type.Further, the use of these ratios, in conjunc-tion with seismic facies interpreted fromreflection patterns, provides a relativelypowerful way to begin inferring lithologicinformation at a scale more detailed thanmost seismic velocity models.

Conventional seismic surveys use compressional

or pressure (P) waves to penetrate the earth and

sea and reflect back data on the strata and struc-

tures these energy waves encounter. When these

waves become disturbances propagating through

the body of a medium, they either remain com-

pressional waves (P waves) or are converted into

shear waves (S waves).

Compressional waves occur when a liquid, solid

or gas is sharply compressed. The compression

sets off small particle vibrations in the same direc-

tion that the compressional waves are traveling.

Shear waves, on the other hand, are waves of

shearing action and occur only in solids. In shear

waves, the small rock particle motion is perpen-

dicular to the direction of wave propagation. They

may be generated by a seismic source in contact

with a rock formation or by the non-normal incidence

of P waves on rock. The generation of S waves

from a reflected P wave is called mode conver-

sion. It is slight for small incident angles and

becomes more pronounced as the offset increases.1

The velocity at which these waves travel is con-

trolled by the mechanical properties of the rock, its

density and elastic dynamic constants. In fluid-

saturated rocks, these properties depend on the

amount and type of fluid present, the composition

of the rock grains and the degree of intergrain

cementation, formation pressure and temperature.

Soft, loosely consolidated rocks are generally less

rigid and more compressible than hard, tightly

consolidated rocks. As a result, P and S waves

travel slower in soft rock than in hard. Extremely

unconsolidated rocks support only weak shear-

wave propagation.

P waves and S waves propagate through rock

and reflect differently at interfaces (next page,top). A P wave reflecting as a P wave, called a P-P

reflection, is symmetric about the point of reflec-

tion at the common midpoint (CMP) halfway

between the source and receiver. A P-S reflection

is asymmetric at the point of reflection, called the

common conversion point (CCP), which is different

from the CMP. The difference in reflection points

must be taken into account through processing.

The combination of P- and S-wave data rather

than P-wave alone can yield previously unavail-

able information about fluids in the pore spaces

and improve the potential for identifying the

prospect lithology. Because gas in a formation

causes the compressional velocity to slow but has

little effect on shear waves, the combination of

compressional and shear measurements helps in

identifying gas-related amplitude anomalies. In

addition, S-wave data are also able, in some

cases, to image structures that P waves cannot

adequately portray, such as reservoirs with gas

clouds in porous rocks above the reservoir (nextpage, bottom).2

Until recently, P and S waves could be acquired

simultaneously only onshore, but with the advent

of ocean-bottom cable (OBC), both can be

obtained. Although the marine seismic source con-

tinues to generate only P waves, once those waves

have reflected off deep strata, they may be con-

verted and propagate upward as S waves. The new

4C seabed cable system is in contact with the

ocean floor and can record both waveforms.

Full-Wave Spectrum: P and S Waves

1. Sheriff RE: Encyclopedic Dictionary of ExplorationGeophysics, SEG Geophysical Reference Series Number 1.Tulsa, Oklahoma, USA: Society of ExplorationGeophysicists, 1973.

2. George D: “Shear Data Acquisition Permits Correlation ofSeismic, Log Data,” Offshore Magazine 55, no. 3 (March1995): 37-38.

Winter 1997 55

■■P and S waves. When P waves (red) reflect as Pwaves, they do so symmetrically about a commonmidpoint (CMP). When P waves convert upon reflec-tion to an S wave (black), they do so asymmetrically,about a common conversion point (CCP).

■■Value of shear waves. Shear waves are useful forimaging beneath gas clouds, as they travel throughthis low-velocity zone relatively undisturbed. The lat-eral displacement of the CCP due to P-S reflectionmust be taken into account in processing.

Receiver1.4 kmBoat heading

Shot Receiver Shot

CMP CCP

P wave S waveP waveS wave

Gas

P-S

P-P

5000

7000

3000

4000

■■Imaging through gas with shear waves. The lower panel shows an image created by migrating P-P CMP-stacked seismic data. The center of the image is weak and disrupted by a gas cloud obscuring the crest of thestructure. The top panel shows migrated P-P CCP-stacked data. The reflectors are clearly imaged all the wayacross the panel.

56 Oilfield Review

Sixth, experience with a few situations inthe North Sea has shown that some reser-voirs have low P-wave reflectivity while hav-ing relatively high P-to-S mode-conversioncapabilities. Therefore, these reservoirs,which can not be seen at all on P-wave data,become visible using the mode-convertedshear waves. An extension of this idea leadsto the conclusion also that by acquiring bothP and S waves, seismic information and logand core data can be correlated more con-vincingly, and perhaps explains why correla-tion has been difficult in some areas.

Seventh, a stationary acquisition system likethe Nessie 4C MultiWave Array permits true3D acquisition, meaning complete offset andazimuth distributions within the data. Towed3D surveys, while providing coverage of a3D volume, do so with a series of essentially2D traverses, as the source is in line with thereceiver cable. By acquiring P and S wavespropagating in all azimuths, velocityanisotropy (the variation of a property withdirection) of P and S waves may be deter-mined. Velocity anisotropy can be especiallypronounced in S waves, and depending onthe type and amount, can be used to help

Oblique Complex oblique

Sigmoid Complex sigmoid-oblique

Oblique (parallel) Shingled

Truncation

Sequenceboundary Apparent

truncation

Downlap

Downlap surface

Toplap

Sequenceboundary

Onlap Onlap

■■Depositional and termination pat-terns. Types of depositional and ter-mination patterns sought in seismicdata indicative of stratigraphic traps,including discontinuities. [After Visher, reference 9, main text.]

Sigmoidofflap

Shelfedge Oblique

offlap

Aggradationalofflap

Shelfedge

Channel-overbankcomplex

Drape

MoundSlope front fill

Apparenttruncation

Reconciliation between static and dynamic models

Construction of property maps

Construction between attributes and petrophysical properties

Construction of attribute maps

Seismic stratigraphic interpretation

Relationship between seismic response and petrophysical properties

3D cube inversion

3D cube reprocessing:Enhanced frequency content

3D cube reprocessing:Improved vertical resolution

Data migration

Raw data

■■Methodology Used by Geoscientists.

Winter 1997 57

discriminate rock types, detect source rocksand identify principal fracture directions. Forexample, shales are often highly anisotropic,displaying transverse isotropy wherein thevertical velocity is different from the horizon-tal velocity. Sometimes this can be observedin 4C seismic data, and could indicate ashale, suggesting the existence of what is acommon sealing formation for many strati-graphic traps. Velocity anisotropy is also animportant consideration when processingsurface seismic data for stratigraphic interpre-tation, as small errors in velocity can impactthe resolution of the final seismic image.

An eighth benefit of acquisition with theNessie 4C MultiWave Array is it allows forthe calibration of AVO (amplitude variationwith offset) analysis derived from streamersurvey data. An AVO effect occurs when thereflection coefficient at an interface changesas a function of distance between sourceand receiver. When P-wave energy strikes aparticular interface, some of it will be trans-mitted as P waves and some will bereflected as P waves, while some of theenergy will reflect as S waves and be trans-mitted as S waves. Some lithology-fluidcombinations generate dramatic AVOeffects, and the observed AVO signatures, oranomalies, can be diagnostic of hydrocar-bons. But these effects are not seen in con-ventionally processed seismic data, becausethe processing step of stacking averagesamplitudes from traces at different offsets.

If an AVO effect is suspected, the seismicdata can be processed to preserve and high-light rather than average amplitude varia-tions. The data can also be visualized in 3Dcubes in the same way stacked data are visu-alized. But in this case there would be multi-ple cubes: one for near-offset data, one formedium offsets and one for far offsets. A mul-titude of 3D cubes can be produced, limitedonly by how much offset information is cap-tured in the analysis. Geco-Prakla scientistshave created as many as 23 different offsetranges for a survey (previous page, top).

Once an AVO anomaly has been detected,it needs to be interpreted to identify therocks and fluids that created it. This is doneby generating models and comparingobserved AVO effects to the modeled ones.Most models contain information from P waves alone, and the shear-wave veloci-ties required for the model are extrapolatedfrom distant well logs or inferred fromempirical transforms relating P to S veloci-ties. Shear-wave velocities extracted fromNessie 4C MultiWave Array data providecrucial input for construction of the modelsand allow for more reliable interpretation oflithology and fluids in the trap.

ProcessingGeco-Prakla researchers have developedthe SCT Seismic Classifier Toolbox softwaresystem for seismic stratigraphic mapping orinversion specifically for the identificationof stratigraphic traps. The SCT methodol-ogy is based on work done by Peter Vailwhen he was at Exxon.8 An extension ofthe Vail technique, which was developed

using 2D seismic data, the process beginsby identifying depositional environmentsand their sequence boundaries in the 3Dseismic data volumes and, within thesequence boundaries, identifying certainstratigraphic and geometric patterns thatshould be sought (previous page, bottom).9

Using advanced image processing algo-rithms that allow the geometry of interbedreflections between the sequence bound-aries to be enhanced, primitives—computertemplates—were developed to map thesepatterns automatically.

The goal is to determine the lithologiccomposition of the stratigraphic objects. Theobjects are defined by their boundaries,which may be subtle and difficult to delin-eate. Boundaries are sometimes identifiedimplicitly by the way some of the seismicreflectors terminate against them. This isespecially the case with submarine fan sys-tems and their associated channels. For thisreason, the Geco-Prakla SCT methodincludes reflection termination recognitionsystems (below).

■■Terminations. Both top and base of a submarine fan system in the Gulf of Mexicohave been interpreted. The customized color bar at the lower right of this seismic dis-play emphasizes the high amplitudes of the gas-bearing fan sand at the center of thescreen. The gamma ray log curve (yellow line) is interpreted to present a submarinefan system. Termination of the fan system is represented by the edges of the blackbright spot within the red and blue layers. The sand thickness decreases below seismicresolution near the edges.

8. Mitchum RM and Vail PR: “Seismic Stratigraphy andGlobal Changes of Sea Level,” in Payton CE (ed):AAPG Memoir 26, Seismic Stratigraphy—Applicationsto Hydrocarbon Exploration. Tulsa, Oklahoma, USA:American Association of Petroleum Geologists (1977):49-213.

9. Visher GS: Exploration Stratigraphy.Tulsa, Oklahoma,USA: PennWell Books, 1990.

58 Oilfield Review

The patterns produced by SCT processingof the 3D volume are then passed throughclassification, a type of generic inversionprocess wherein the desired stratigraphicpattern defines a set of attributes that aresought in the data. This produces either acatalog of patterns that the interpreter canuse, or these patterns can go directly intothe actual 3D volume following patternenhancement.

At this point, the system is trained by theinterpreter, who can simply point at a partic-ular pattern and instruct the SCT system torecognize comparable patterns in the seis-mic dataset. The toolbox runs all of the seis-mic volume through the enhancement stepand classifies based on the patterns thathave been presented as the training data,either from the catalog or the identified pat-terns in the dataset. This produces a classvolume, or voxel set, in the cube (above).Each voxel will have one value for each ofthe classes or stratigraphic patterns thatwere defined for the SCT process or it willhave a value that indicates there is doubt inthe pattern that has been identified.

At the end of the session, a set of voxelvolumes is produced from the kinematic, orgeometric, part of the seismic data, indicat-ing the density of the bedding between the

sequence boundaries and how the beddingterminates within the retaining system. Atthis point, the dynamic information isadded, such as the reflection strengths of thevarious geometric patterns that were identi-fied. These help discriminate between thetypes of lithology that may be lying betweeneach of the stratigraphic objects.

Even if the 3D seismic data being pro-cessed through the set are solely acoustic orP-wave information from conventionaltowed streamers, there is some dynamicinformation that can be used to identifystratigraphic traps, including the same set ofprimitives. In addition, AVO methods canprovide a fair understanding of lithologies.This is an implicit way of getting shear-waveinformation without the added cost of multi-component acquisition.

Available well logs are now added to theprocess. To put the well and seismic data onthe same scale, the log is used to construct asynthetic seismogram by convolving reflec-tion coefficients from the sonic and densitylog with a band-limited seismic wavelet.This process achieves the same low resolu-tion as that of the selected wavelet.

In a case study for Statoil in the Danishsector of the North Sea, for example, theNorwegian state oil company had a 2Dline and two wells in which Geco-Prakla

was able to identify two stratigraphicsequences with the SCT Seismic ClassifierToolbox system and map them in 3D. Oneof the wells showed hydrocarbons in thesands of a submarine channel system. TheSCT approach, using fluid indicators fromthe ratio of P- to S-wave velocities, Vp/Vs,measured in both the wells and inferredfrom the seismic data, made it possible todetermine the reservoir’s fluid distribution.

To discriminate between hydrocarbon-bearing lithology and non-hydrocarbon-bearing lithology, it is essential to calibratethe lithological effect of the Vp/Vs ratio ver-sus its fluid effect. If perturbations are pre-sent, they can be associated with fluid ratherthan lithology. The Danish example was a2D traverse, not a 3D multicomponent OBCdataset, which would have allowed furtheridentification of each layer of the stratigra-phy with its full elastic field observations.Nevertheless, it was processed through anew set of software tools Geco-Prakla callsmodel-based processing, which assuredconsistency between the 3D geometricmodel of the dataset and its P-wave velocitymodels, and confirmed the identification ofhydrocarbons in the stratigraphic trap inzones not tapped by the wells.

Offset

■■Voxel cube. Volume cell, or voxel interpretation of a 3D survey from the Gulf of Mexicowith GeoViz software. The voxel interpretation in red highlights the distribution of a par-ticular seismic attribute, or characteristic, that maps the extent of a bright spot.

Winter 1997 59

Visualization and Attribute AnalysisLooking for stratigraphic traps in seismicreflection data, the geophysicist searches for subtle variations within formationsrather than the obvious structures. Thisrequires analyzing the seismic attributes, orcharacteristics, of the seismic traces. Keyattributes of the seismic data reveal whetherthe trap is structural or stratigraphic andprovide additional characteristics that arehelpful in determining the precise nature ofthe lithology.

In mapping sand channels, for example,the termination attribute is analyzed todetermine the shape of the channel, wherea reflector terminates against a particularchannel. If there are channel complexes andhigh amplitudes are present, they will alsobe visible in the seismic data’s texture andcan be verified. The texture of the seismictraces, another attribute, the rough orsmooth appearance of the data, serves as anexcellent indicator of lithology. When thesediments are deposited under high-energyconditions, as are sands, the seismic datalook chaotic; if quiet sedimentation, such asclays, then the data appear smooth (right).

Of the many different seismic attributes,the majority of which have arguable valuein exploration, amplitude, the maximumdeparture of a wave from the average value,is the most important indicator of strati-graphic traps. This is because amplitudescan light up stratigraphic traps as brightspots, in which the amplitudes are dramati-cally greater or less than those of the adja-cent layers. Even more important is whenthere is an updip drop in amplitude,because it can indicate a potential reservoirunit. In addition, when a stratigraphic inter-val thins with two beds coming together ashappens in a pinchout, wavelets from thesetwo beds merge causing reinforcement, orbrightening of amplitude, and the locationof that brightening is where the trap lies.

A great deal of information about the rockproperties of the formation is available fromamplitude. If true amplitudes are obtained,reflection coefficients derived from the datacan be employed to compute acousticalimpedance. This, in turn, can be related to density, velocity and porosity. Amplitudeis also a major factor in estimating the netpay of the stratigraphic field, becauseamplitude changes with the amount ofhydrocarbons present.

One useful attribute in identifying strati-graphic traps in 3D seismic data is ampli-tude versus azimuth, an aspect of anisotropythat combines both travel time and ampli-tude information and can indicate appropri-ate hydrocarbon reservoirs by whether theunit has closure and whether its amplitudeis dim in the updip direction. Another use,in mapping fracture zones, from which pri-mary production often comes, is that it indi-cates areas of weakened amplitude, whichis caused by a poor coupling of S wavesacross fractures. Azimuthal coverage is typi-cally 360° in a 4C survey, thus assuring theacquisition of this key attribute.

Accurate velocity measurement is alsoessential to correctly interpreting structuresas stratigraphic traps, and particularly ingauging variations in the recorded velocities.Incorrect analysis can result from followingthe wrong reflecting wavelet. If the mediumis 4C 3D seismic data, the measurement ofshear information is much more reliable,permitting the use of the anisotropic effectand identification of both the P-wave layervelocity and the S-wave layer velocity, andadding to the discrimination power withrespect to different lithologies, pore fluidsand abnormal pressure regimes.

Although amplitude and velocity are inte-gral to an accurate interpretation of 3D seis-mic data, different attributes are importantin different aspects of the interpretation pro-cess. To determine which are more impor-tant, discrimination analysis is undertakenby compressing the seismic data into a set of attributes that might be relevant for theparticular problem that has been identified,in this case the discovery of stratigraphictraps. From these attributes, a set of attributetransformations is selected.10

In what is called attribute space in themodel, points are defined for each positionwith several attribute values. These can bethought of as clusters. Separations betweenthe clusters permit discrimination, becausethey show that there is no overlap betweenthe two classes of attributes. If, for example,there are only two attributes, the attributespace is two-dimensional and appears as acrossplot, with each of the axes in that cross-plot now an attribute parameter. If there aretwo separate clusters in that crossplot, then

■■Textural contrast in seismic data. In this example from theGulf of Mexico, the seismic horizon has been artificially illumi-nated to reveal the seismic texture. The rough sections corre-spond to high-energy areas of erosion at the level of themapped surface, and may be channel incisions. The smootherareas represent quieter regions, depositional settings where fansystems may have developed.

10. For a discussion of transformation in the context oftime-lapse or 4D seismic: Pedersen L, Ryan S and Son-neland L: “Seismic Snapshots for Reservoir Monitor-ing,” Oilfield Review8, no. 4 (Winter 1996): 32-43.

60 Oilfield Review

■■Integration of conventional seismic line with Coherence Cube (Created by CoherenceTechnology Co. using GeoViz software from GeoQuest). Zones of low coherence (black)are interpreted as discontinuities.

High coherence

Coherence CubeConventionalseismic data

2000

2500

3000

3500

4000

Low coherence

Lineation anomaly

Coherence processing

Channel

■■Multiple images of a subtle channel processed for coherence. What appears to be a barely perceptible, low-amplitude body in the black and white section of stacked data is revealed by Coherence processing (right) to be asubtle submarine channel, with its trend and areal extent (red arrow) clearly visible. (Images courtesy of Coherence Technology Co.)

High-Coherence Event

Low-Coherence Event

■■High- and low-coherence events. In a high-coherence event (top),waveforms are similar from trace to trace. In a low-coherence event(bottom), waveforms are disimilar. (Images courtesy of Coherence Technology Co.)

Stratigraphic features frequently are better imaged

by Coherence Cube processing, a recently per-

fected methodology that may be applied either

after or during the processing of 3D seismic data

and utilized during the interpretation to further

reveal the stratigraphic trap (below right). A non-

traditional procedure, Coherence Cube processing,

which was developed in the geophysical labs at

Amoco and licensed exclusively to Houston-based

Coherence Technology Company, processes 3D

seismic data not for imaging reflections, but for

imaging discontinuities by analyzing waveform

similarity (below). Traces that are similar to each

other are mapped with high-coherence coeffi-

cients, and when similarities end, discontinuities

may be inferred. As a consequence, when visual-

ized in a 3D volume or cube, coherence coeffi-

cients enhance the detection and understanding of

stratigraphic features (as well as faults) that are

often not visible in traditionally processed data.

Stratigraphic features are frequently difficult to

see in seismic data due to the low level or chaotic

nature of the seismic reflections they provide.

Coherence Cube processing brings stratigraphic

features into focus as it computes the variations in

the waveform regardless of the amplitude of the

Coherence Technology

Winter 1997 61

discrimination between those two cases withthat particular attribute set has been success-ful. If, however, the two clusters are overlap-ping, then there is no discrimination.

Geco-Prakla has patented a set of newattributes that allow the reconstruction of allthe information that exists in a given layer—for example, between the top and base ofthe reservoir. And it can be done in anorthogonal sense; for each new attribute thatis added to the classification system, addi-tional information is obtained, limited onlyby computer running time.

Once the attributes have been selected thatthe geoscientist feels are key to a valid inter-pretation, a number of identification method-ologies may be applied, including horizonslicing (above). Coherence Cube processingreveals the stratigraphic trap further (see“Coherence Technology,” previous page).11

Now, and in the years to come, thepetroleum industry is relying on discovery ofnew resources in two areas: matureprovinces and the deepwater offshore fron-tier. Both demand new technology to findthe elusive hydrocarbons that lie beneaththe earth and sea. Since stratigraphic trapsmay well hold the vast majority of the newmillennium’s yet-to-be-discovered hydrocar-bons, particularly when reefs are consideredamong them, “The time is now for all explo-rationists to purposefully search for the sub-tle trap,” in the words of Michel T. Halbouty.New acquisition technology now illumi-nates these stratigraphic traps as neverbefore, and remarkable advances in pro-cessing and interpretation software andmethodology reveal their attributes for anal-ysis, visualization and verification. Difficultto find, yes, but as the giant stratigraphictraps around the world testify, well worththe search. —DG

reflectors (previous page, top). Lateral definition of

these stratigraphic features can be seen best in the

horizontal or time domain. In areas of high dip or

where stratigraphic features transit different strati-

graphic horizons, flattening of key surfaces, hori-

zon slices, may be beneficial in obtaining a greater

understanding of the stratigraphy contained in the

dataset. However, interpretive bias can enter the

dataset when using horizon slices in tracing strati-

graphic features, since a geoscientist is required to

go through the difficult, time-consuming and sub-

jective process of picking the horizon.

The Coherence Cube technique increases the

probability of finding hydrocarbons by indicating

stratigraphic (and structural) traps that were not

visible with traditional procedures. Estimates of

3D dimensional seismic coherence are obtained

by calculating localized waveforms within the reg-

ular grid of a 3D seismic dataset. A sharp disconti-

nuity is produced by stratigraphic boundaries.

In areas such as the Gulf of Mexico, where high

seismic amplitudes frequently indicate hydrocar-

bon accumulations, their stratigraphic milieu is

more readily identified from coherence data

because they provide a different perspective in

combination with amplitude data.1

Offse

■■Horizon slice. GeoViz 3D workspace showing horizon and surface slicing, whichallows detection of pay sands adjacent to a horizon or surface. High relativeamplitudes are shown in red and yellow.

11. Risch DL, Chowdhury AN, Hannan AE and JamiesonGA: “How Modern Techniques Improve Seismic Inter-pretation,” World Oil215, Part I (April 1994): 85-94;and Part II (May 1994): 63-70.

1. George D: “Coherence Cube Reveals Stratigraphic FeaturesMore Readily than Traditional 3D Time Slice Plots,Offshore Magazine 56, no. 3 (March 1996): 22-23.