3-D Seismic Interpretation

'....,*E*WX..*-*

r,.r

" r.{

3-DSeismic Interpretation

3-D seismicdatahavebecomethe key tool usedin the oil andgasindustryto understandthe subsurface. In addition to providing excellent structual images,the densesampling of a 3-D survey can sometimes makeit possibleto mapresewoirqualityandthe distributionofoil andgas.The aim of this book is to help geophysicists andgeologistsnew to the techniqueto interpret3-D datawhile avoidingcommonpitfalls. Topicscoveredinclude basicstructuralinterpretationand map-making;the useof 3-D visualisation methods;interpretation of seismicamplitudes,includingtheirrelationto rock andfluid propertiesiandthe generationanduseof AVo andacousticimpedancedatasets. Also includedis the increasinglyimportant neld of timelapse seismic mapping, which allows the interpreter to ftace the movement of fluids within the reservoirduringproduction.The discussionof the acquisitionandprocessingof 3-D seismicdatais rntendedto promote an understandingof important data quality issues.Extensive mathematicshas been avoided,but enoughdetailis includedon the effectsofchangingrock andfluid propertiesto allow readers to maketheir own calculations. The authorsofJ-D SeismicInterpretationareprofessionalgeophysicists with manyyears,experience in the oil industry. They are still actively inter?rcting 3-D seismicdataand arc thercfore able to summarise the currentbest practice.The book will be indispensable for geoscientists leamingto use 3-D seismic data, particularly $aduate studentsof geophysicsand petroleum geology, and new enftants into the oil andgasindusty. Mike Bacon was awarded a Ph.D. in geophysicsfrom the University of Cambridge before becoming a Principal Scientific Offrcer at the Institute of Geological Sciencesin Edinburgh (now the British Geological Suwey). After working as a lecturer in the Geology Depanment of the University of Accra, Ghana, he took a position with shell uK where he worked for 19 years asa seismicinterpreter ard asteam leaderin seismic special studies. Dr Bacon is a co-authorof Introduction to seismic Interpretation by Mce'llrin et al. (1979).'d is a memberof the editorialboardof the petroleumindusty magazine FirstBreak.He is a Fellow of the GeologicalSocietyand a memberof the EAGE (EurcpeanAssociationof Geoscientists andEngineers). Robsimm is a geophysicistwith 16 years' experiencein the oil and gasindustryand a specialistin the rock physicsinterpretationof seismicdatain both explorationand production.After gainingan M.sc. and Ph.D. in marinegeologyat UniversityCollegeLondon,the early part of his careerwas spentwith Bdtoil plc and rricentrol plc as a seismicinter?rcter.He subsequently took a positiqnat Enteryriseoil and progressedfrom North Seaexplomtionto productionand equity determination,prior to becoming a'' rntemalconsultantto assetteamsand management. since 1999Dr Simm hasprcvidedindependent consultancyand taining servicesto numerousindependentand multi-national oil companiesthroush his companyRock PhysicsAssociatesLtd. TerryRedshawgaineda Ph.D. in numericalanalysisfrom the unive$ity of wales beforebecominga GeophysicalResearcher with WestemGeophysical.Since1985he hasbeenemployedby Bp in a variety of roles.Thesehaveincludedresearchinto imagingand inve$ion algorithms,as well as leadinga team supplyingBP's worldwide assetswith supportin the areasof seismicmodelling,rock properties,AVO and seismicinversion.Dr Redshawworks.atpresentin Bp,s ExplorationExcellenceteam,which helps operatingunits to carry out the tecbnical wo* neededto evaluateoil prospectsand decidewhether to drill them or not.

ffirrio, 'wi, um.. llhn' "*

3-DSeismic Interpretation M.Bacon R,Simm T,Redshaw

CAMBRIDGE

UNTVERSITY PRESS

PUBLISTTED BY TIIE PRESSSYNDICATE OF TIIE UNIVERSITY OF CAMBRIDCE

The Pitt Building, TrumpingtonStreet,Canbridge,Unitei Kingdom The EdinburghBuilditrg, Canbridge CB2 2RU, uK 40 west 20rhSreeL New York, l.[Y 1001l-4211,USA 477 WilliamstownRoad,Port Melboume,VIC 3207,Ausfialia Ruiz de A1arc6n13,28014Madrid, Spain Dock House,The Waterftont,CapeTown 8001,SouthAfrica http://ww\t cambridge.org @M. Bacon,R. Simm andT. Redshaw2003 This book is i$ cop].right.Subjectto statutoryexception and to the provisionsof rclevantcollectlvelicensingagreements, no reproductionof any part may takeplacewithout the written perrussionof CarnbridgeUnivenity hess. First published2003 Pdntedin drc Unit€d Kitrgdomat the Unilersity hess, Cambddge Typefaces'fimesl0,5ll4 pt. ard HelveticaNeue

Slsten IATEX2E [rB]

A cataloguerecordfor this book is awilable Irom theBritish Library Library of CongressCataloguingin Publicationdata Bacon,M. (Michael), 194G 3-D seismicinterpretation/ by M. Bacon,R. Simm, T. Redshaw. p. cm. Includesbibliographicalreferclces andindex. ISBN 0 521 79203? (hadback) 1. Seismicreflectronmethod, 2. Seismicprospecting. 3. Petroleum- Geology. 4. Natual gas Geology. I. Titlei Three-Dseismicinterpretation. II. Sirffn, R. (Robert),1959T. (Terence), 1957 IV Tide. m. Redshaw, QE539.824 2003 62/.1 592-4c21 20O3Mr2Or ISBN 0 521 792037 hardback

Gontents

Preface

page rx

-

1

Introduction

1.1

Seismicdata Migration of seismic data Data density Usesof seismicdata Roadmap Conventions:seismicdisplay.units Unit conversions References

2 3 7 9 13 14 15 16

2

3-Dseismic dataacquisition andprocessing

17

2.1 2.2 2.3 2.4 2.5 2.5.1 2.5.2 2.5.3 2.5.4 2.5.5 2.5.6

Marine 3-D dataacquisition Marine shearwave acquisition 3-D land acquisition Other types of seismic survey 3-D dataprocessing Reformat,designature, resamplingandgain adjustment Deconvolution Removingmultiples Binning Stacking and migration Post-migrationprocessing References

1.2 1.3 r.4 1.5 1.6 1.7

I

18 26 30 34 35 35 39 39 43 46 53 .)l

vi -

Contents

-

3

interpretation Structural

Well ties 3.1 3.1.1 The syntheticseismogram 3.1.2 TheVSP Workstationinterpretation 3.2 3.2.1 Displaycapabilities 3.2.2 Manualhorizonpicking 3.2.3 Autotrackers 3.2.4 A ributes 3.2.5 Viewingdatain 3-D Depthconversion 3.3 methods 3.3.1 Principlesof vertical-sfretch 3-3.2 Use of well velocityinformation 3.3.3 Use of seismicvelocities 3.3.4 I-ateralshifts References

57 57 58 66 71 72 77 8l 84 88 89 89 94 96 98 100

-

4

interpretation Geological

102

4.1 4.2 4.3

Seismicresolution Seismicsffatigraphy Interpretation tools Someexamples Faults References

ro2

5

seismic amplitudes Interpreting

120

5.1 5.2 5.3 5.4 5.5 5.5.1

Basicrock properties Offsetreflectivity Interpretingamplitudes AVO analysis Rock physicsfor seismicmodelling Fluid effects

120 171 tz5 130 139 140

4.4 4.5

106 109 113 117 118

vii -

Contents

5.5.1.1 Calculatingfluid parameters 5.5.1.2 Calculatingmdtrix parameters 5.5.1.3 Invasioneffects 5.5.2 P-wavevelocityandporosity 5.5.3 P-wavevelocityandclay content 5.5.4 P-wavevelocity anddensity 5.5.5 Shearvelocity 5.5.6 Dry rock moduli 5.6 Assessingsignificance References

143 144 1/t<

146 r46 146 148 150 151 153

--

6

lnversion

6.1 6.2 6.2.L 6.2.2 6.3 6.3.1 6.3.2 6.4

Principles Procedures SAIL logs Extendingthe bandwidth Benefitsof inversion Inferringreservoirquality Stochasticinversion AVO effects References

159 164 164 166 170 111

7

3-Dseismic datavisualisation

172

Reference

179

8

Time-lapse seismic

180

8.1 8.2 8.3 8.4 8.5

Rock physics Seismicmeasurements Seismicrepeatability Seismicprocessing Examples References

183 184 186 187 188 191

155 155

r51 r57

-

viii -

Contents

issues 1 Workstation Appendix A1.l A1.2 A1.3

Hardware Sottware Datamanagement Reference

r93 193 194 194 195

-

2 Glossary Appendix

196 209

x

Preface

-l

The combinationofthe inter?reter'singenuitywith evenmorecomputer theserespects. in the future. powerwill surelyleadto furtherdevelopments We haveincludeda numberof examplesof seismicdisplaysto illustratethe various interpretationtechniques,and to give the readera feeling for the typical quality of modem seismicdata.We are grateful to the fbllowing for permissionto reprofor 1igs.2.2,2.8,2.16,2.23duceproprietaryor copyrightmatedal:BP Exploration andStatoilfor fig. 5.l 2; 2.24.2.27. 2.30.234-2.37,8.3 and8.7-8.8; ChevronTexaco for figs.3.1,3.3,3.5-3.6,3.8-3.13,3.17-3.18, andProduction ShellUK Exploration (BP Explo3.20-3.24,4.4, 4.6, 5.6,6.2-6.8and 6.10;the WytchFarmpartnership (UK) Ltd, ONEPM McGee Resources plc, Kerr ration OperatingCo Ltd, PremierOil Ltd andTalismanNorth SeaLtd) for figs. 7.1-7.6; the GeologicalSocietyof London for 1ig.1.6(b);the McGraw-Hill Companiesfor fig. 5.3; the and Dr R. Demyttenaere andEngineers(EAGE) andDr J. Hendrickson EuropeanAssociationof Geoscientists for fig. 5.16; the EAGE and Dr P. Hatchell for figs. 8.4-8.5; the EAGE and Dr J. Stammeijerfor fig. 8.6; the Society of ExplorationGeophysicists(SEG) for G. H F forfig. 1.6(a),theSEGandProfessor fig.4.1,theSEGandDrS. M. Greenlee Gardnerfor fig. 5.1,theSEGandDr H .Zengfor fig.4.7,IheSEGandDrW. Wescottfor fig. 4.8, andthe SEGandDr L. J. Woodfor fig. 4.9.Figures3.1,3.3 and3.24werecreatedusingLandmarkGraphicssoftware,fig. 4.6 usingStratimagicsoftware(Paradigm Geophysical),flg. 5.15(b)using Hampson-Russellsoftwareand fig. 6.3 usingJason software. Geosystems The text is intendedas an aid in developingunderstandingof the techniquesof 3-D interpretation.We have not been able to include all the possiblelimitationson applicabilityand accuracyof the methodsdescribed.Careis neededin applyingthenl in the real world. If in doubt,the adviceof an experiencedgeophysicistor geologist shouldalwaysbe sought.

I

1 Introduction

If you want to find oil andgasaccumulations, or producethemefficientlyoncefound, then you needto understandsubsurfacegeology.At its simplest,this meansmapping subsurfacestructureto find structureswhereoil and gasmay be trapped,or mapping faults that may be barriersto oil flow in a producingfield. It would be good to have a map of the quality of the reservoiraswell (e.g.its thicknessandporosity),partly to estimatethe volumeof oil that may be presentin a giventrap, andpartly to plan how bestio get the oil or gasout of the $ound. It would be betterstill to seewhereoil and gasare actuallypresentin the subsurface, reducingthe risk of drilling an unsuccessful explorationwell, or evenfollowing the way that oil flows throughthe reservoirduring productionto makesurewe don't leaveany moreof it thanwe canhelp behindin the ground.Ideally,we would like to get all thisinformationcheaply,which in the offshore casemeansusingasfew boreholesaspossible. Onetraditionalway of understanding the subsurfaceis from geologicalmappingat the surface.In many areas,however,structureand stratigraphyat depthsof thousands of feet cannotbe extrapolatedfrom geologicalobservalionat the surface.Geological knowledgethendependson boreholes.They will give very detailedinformationat the pointson the mapwheretheyaredrilled.Interpolatingbetweenthesecontrolpoints,or extrapolatingawayfrom them into undrilledareas,is wheregeophysicalmethodscan be mosthelpful. Although someusehasbeenmadeof gravity andmagneticobservations, which respondto changesin rock densityandmagnetisation respectively, it is theseismicmethod thatis by far themostwidely usedgeophysicaltechniquefor subsurface mapping.The basicideais very simple.Low-frequencysoundwavesaregeneratedat the surfaceby a high-energysource(for examplea smallexplosivecharge).They traveldown though the ea.rth,and are reflectedback from the tops and basesof layers of rock where there is a changein rock properties.The reflectedsoundtravelsback to the surfaceand is recordedby receiversresemblingmicrophones.The time taken for the soundto travel from the sourcedown to the reflecting interfaceandback io the surfacetells us aboutthe depth of the reflector, and the strengthof the reflected signal tells us about the change of rock propertiesacrossthe interface.This is similarto the way a ship'sechosounder can tell us the depth of water and whether the seabedis soft mud or hard rock.

2 -

lntroduction Initially, seismicdata were acqulqd along straightlines (2-D seismic);shooting a number of lines acrossan area gave us the data neededto make a map' Again, the processis analogousto making a bathymetric map from echo soundingsalong a numberof ship ffacks. More recently, it hasbeenrealisedthat there are big advantages to obtainingvery closelyspaceddata,for exampleasa numberofparallel shaightlines very closetogether.Insteadofhaving to interPolatebetweensparse2-D lines,theresult is very detailed information about the subsurfacein a 3-D cube ('{ and } directions horizontally on the surface,z direction vertically downwardsbut in reflection time, not distanceunits).This is what is known as3-D seismic. This book is an introductionto the ways that 3-D seismiccan be usedto improve our understandingof the subsurface.There are severalexcellent texts that review the principlesandpracticeof the seismicmethodin general(e'g. Sheritr& Geldart,1995)' on the distinctivefeaturesof 3-D seismic,and aspects Our inlentionis to concentrate that are no different from the corresponding2-D caseare therefore sketchedin lightly' The aim of this first chapteris to outlinewhy 3-D seismicdataaretechnicallysuperior to 2-D data.However,3-D seismicdata are expensiveto acquire,so we look at the balancebetweenbetterseismicquality and the cost of achievingit in differentcases' The chaptercontinueswith a roadmapof the technical material in the rest of the book, and concludes with notes on some important details of the conventions in use for displayingseismicandrelateddata. A complementaryview of 3-D seismicinteryretation,with excellentexamplesof colour displays,is providedby Brown (1999).

-I

1.1

data Seismic with a would be a I -D point measurement -The simplestpossibleseismicmeasurement single source(often referredto asa slzal,from the dayswhen explosivechargeswerethe most usualsoufces)andreceiver,both locatedin the sameplace.The resultscouldbe displayedasa seismictrace,which is just a graphof the signalamplitudeagainsttraveltime, conventionallydisplayedwith the time axis pointing verticallydownwards'Reflectorswould be visibleastraceexcursionsabovethe ambientnoiselevel Much more wi I sourcesandreceiverspositionedalonga straightline usefulis a 2-D measurement, on the sudace.It would be possibleto achievethis by repeatingour 1-D measurement at a seriesof locationsalong the line. tn practice,many receiversregularly spacedalong the line areused,all recordingthe signalfrom a seriesof sourcepoints.In this case,we can exffact all the tracesthat havethe samemidpoint of the source-receivetoffset. This is a commonmidpointgather(CMP) Thetraceswithin sucha CMP gathercanbe added together(stacke if the increaseof travel-time with offset is first correctedfor (normal moveout(NMO) correction).The detailsof this processarediscussedin chapter2'

4

lntloduction

I

D p t h

T

i I i

timesection. raysandresulting of normal-incidence Fig.1.1 Sketch path of the incident ray to the reflector, so the angle of incidence at the reflecting horizon must be 90o.Not only arereflection points not directly below the surfacepoint whereverthis horizon is dipping, but for some surfacelocations there may be several different reflections from the hodzon, and for other surfacelocations there may be no reflections received at all. The display produced by plotting seismic tracesvertically belowthesurfacepointswill, assketchedin thelowerhalfoffig. 1.1'behardto interpret in any detailedsense.This problemis solvedby a processingstepcalledmigration, which repositionsreflectors to their correct location in space.There are various ways of carrying this out in practice, but the basis of one method (Kirchhoff summation) rs illustratedin fig. 1.2.This showsa point scattererin a mediumofuniform velocity;this 'cat's eye' that reflects any incident ray directly back reflector is to be thought of as a along the path by which it arrived. If a seismic line is shot above such a reflector, it appearson the resulting sectionasa hyperbolic event.This suggestsa migration method as follows. To find the amplitude at a point A in the migrated section, the hyperbola correspondingto a point scattererat A is superimposedon the section' Wherever it crossesa trace, the amplitude value is noted. The sum of these amplitudes gives the amplitude at A in the migrated section. Of course,not all the amplitude values in the summationtruly relate to the scaltererat A; however,if thele are enoughtraces'energy received fiom other scattererswill tend to cancel out, whereasenergy truly radiated from A will add up in phasealong the curve. (A more completediscussionshowsthat various correctionsmust be applied before the summation,as explahed, for example, in Schneider,1978.)

c -

Migralion of seismic data Position D p I h

T

t i

Fig. 1.2 Sketchofrays reflectedfrom a point scattercrand resulting time section.

The snagwith such a procedureis that it repositionsdata only within the seismic section.If data were acquiredalong a seismicline in the dip direction,this should work fairly well; if, howevel we acquiredata along a line in the strike direction,it will not give correctresults.If we havea 2.5-D sffucture,i.e. a 3-D structurein which the dip sectionis the sameat all points along the structure,then on the strike section all reflectorswill be horizontal,and the migrationprocesswill not repositionthem at all. After migration,dip and strike sectionswill thereforenot tie at their intersection (fig. 1.3(a)).This makesinterpretationof a close grid of 2-D lines over a complex structurevery difficult to cany out, especiallysincein the real world the local dip and strikedirectionswill changeacrossthe structure. In general,someof the reflectionson any seismicline will come from subsurface pointsthat do not lie directly below the line, andmigratingreflectionsas thoughthey do belongin theverticalplanebelowthe line will givemisleadingresults.For example, fig. 1.3(b)showsa sketchmap of a seismicline shot obliquely acrossa slope.The reflectionpoints are locatedoffline by an amountthat varieswith the local dip, but is typically 250m. If we seesomefeatureon this line that is importantto preciseplacing of an explorationwell (for examplea smallfault or an amplitudeanomaly),we haveto bearin mind that the featureis in reality some250 m awayfrom the seismicline that showsit. Of course,in sucha simplecaseit would be fairly easyto allow for these shiftsby interpretinga grid of 2-D lines.If, however,the strxctureis complex,perhaps with many small fault blockseachwith a differentdip on the targetlevel,it becomes almostimpossibleto map the strxcturefrom sucha grid. Migration of a 3-D survey,on the other hand, gatherstogetherenergy in 3-D; Kirchhoff summationis acrossthe surfaceof a hyperboloidratherthan along a hyperbola(fig. 1.4).Migrationof a tracein a 3-D surveygatherstogetherall the reflected energythat belongsto it, from all other tracesrecordedover the whole (;, y) plane.This

l,

-

lntoduclion (a)

Strikesection

DiDsection --------tl-

(b)

Fig,1.3 (a)Fora 2.5-Dsfucture,dip andstrikelinesdonottie aftermi$ation;(b)mapviewof aredepthin feet(ft)). reflectionpointsfor a 2-Dline(contours means that events are conectly positioned in the 3-D volume (provided that the migration processhas been canied out with an accuratealgoritbm and choice of parameters,as discussedfurther in chapter2). This is an enormousadvancefor mapping of complex areas;insteadof a grid of lines that do not tie with one another,we have a

-

Datadensity

/z-\ LI e

//

?;tt:;ilr,o*oo)\

3 D migration (sumacrosshyperboloid) Fig.1.4 Kirchhoffmigrarionin 2-D and3-D. volume oftrace data,fromwhich sectionscan be chosenfor display in any orientationwe want. Furthermore, focussing of the data is also improved. For unmigrated data, the limiting horizontal resolution can be taken as the size of the Fresnel zone, an areasurrounding any point on the reflector from which reflected energy arrives at the receiver more or less in phase and thus contributing to the signal at that reflection point. The radius / of this zone is given approximately by " ^ h J

-

---.

z, where .1"is the dominant wavelengthof the seismic disturbanceand ft is the depth of the reflectorbelow the source-receiver point (seee.g. McQuillin et al., 1984).This canamountto severalhundredmetresin a typicalcase.MigrationcollapsestheFresnel zones;2-D migrationcollapsesthezoneonly alongtheline direction,but 3-D migration collapsesit in both inline andcrosslinedirections,to a valueapproachingI/2, which may amounito a few tensof metres.This will improvethedetailthatcanbe seenin the seismicimage,thoughvariousfactorswill still limit theresolutionthatcanbe achieved in practice(seesection4.1). -

1.3

Datadensity When 3-D seismicfirst becameavailable,it resultedin an immediateincreasein the accuracyofsubsurfacestructuremaps.This waspartlybecause of theimprovedimaging

,

Ia

8

lntroduction

I

Amplitudeon a singleline 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 -

rO

6')

(f)

l'-

-

c\t

t

O

(\t

O

r

(

At

i

)

c)

F

-

(r)

-

s

x

mapview Amplitude:

Fig. 1.5 Top: graph of amplitude versusposition along a single line. Bottom: map view of amplitude variation across many similar parallel lines.

9 --

Usesof seismic data discussedin the last section,but also becauseof the sheerdensity of informalion available.Mappingcomplexstructuresfrom a grid of 2-D datais a subjectiveprocess; the interpreterhasto makedecisionsabouthow to join up featuresseenon lines that might be a kilometreor more apart.This meansthat establishingthe fault pattemin a complicatedareawill be time-consuming,and the resultingmaps will often have significantuncertainties. 3-D data,with their densegrid of traces,allow featuressuch as 1-aults or stratigraphicterminationsto be followed and mappedwith much greater (seesection3.2.2). assurance More recently,it has beenrealisedthat the densityof coverageallows us to make moreuseof seismicattributes.This will be discussed in detailin chapter5, but a typical examplemight be that we measurethe amplitudeof a seismicreflectionat the top of a reseruoir,which increaseswhen hydrocarbonsare present.Such an effect is often quitesubtle,becausethe amplitudechangemay be smallandalmostlost in thenoisein the data.Consistentchangesacrossa 3-D datasetstandout from the norsemuch more clearlythan changesalonga 2-D hne. Figure 1.5 showsa syntheticexampleillusfratingthe powerof seeingdensedatain map view.At the top is a graphof amplitudealonga singleline; the left_handhalfhas a meanvalueof0.1l andthedght-handhalf of0.l2, anduniformly disnibutedrandom noisewith amplitudeA 0.01 hasbeenadded.Workingfrom this graphalone,it would be ha.rdto be certain that there is a higher averageamplitude over the right_handpart, or to saywherethe changeoccurs.The lower part of fig. 1.5 showsa contourmap of the amplitudesof40 suchlines,eachwith the amplitudestepin the sameplacebut a differentpattem of randomnoise;the lines run from bottom to top of the area.It is immediatelyobviousthatthereis a stepchangein averageamplitudeandthatit occurs halfway up the area.As we shall seein chapter5, correlationof amplitudeanomalies with structurecanbe a powerfultestfor hydrocarbonpresence;this syntheticexample showswhy interpretationofamplitudeanomaliesis muchmoresolidlyfoundedon 3_D datathanon a srid of 2-D data.

1.4 Uses ofseismic data Seismicdataare usedboth in explorationfor oil and gasand in the productionphase. The type and quality of data gathered are determined by the balance between the cost of the seismicand the beneflt to be gainedfrom it. The generalpattem is as follows. (1) Early exploration.At this stage,knowledgewill probablybe very sketchy, with little or no well information.The presenceof a sedimentarybasinmay be inferred from outcropgeology,or indirectlyfrom geophysicalmethodssuchasgravityand magneticsthat distinguish sedimentaryrocks from metamorphic basementon the basis of their density or magneticsusceptibiliry(see e.g. Telford et al., 1976:)-

't0 -

lntoduction At this stageevena small number of 2-D seismicprofiles acrossthe basin, perhaps tens of kilomeffes apart, will be very helpful in defining the general thickness of sedimentsand the overall structural style. (2) Perhapsafter some initial wel1shave been drilled in the basin with encouragrng results, exploration moves on to a more detailed study, where the aim is to deflne and drill valid traps. More seismic data are neededat this stage,although the amount dependson the complexity of the structues' Simple anticlines may be adequately defined from a small number of 2-D profiles, but imaging of complex fault architectureswill often be too poor on 2-D data for confident interpretation' If wells are fairly cheapand seismic data are expensiveto acquire (as is often the caseon land) it may be best to drill on the basisof a grid of 2-D lines' If weils are vety expensivecompared with seismic acquisition (the typical marine case), then it will alreadybe worthwhile at this stageto use 3-D seismicto make sure that wells are corectly located within the defined traps. This might' for example, be a matter of drilling on the upthrown side of a fault, or in the correct location on a salt flank to intersect the pinchout of a prospectivehorizon An example where 3-D seismiccompletelychangedthe structuralmap ofa fie1dis shownin fig' 1 6(a) (redrawnafterGreenleeet a1.,1994).This is the AlabasterField, locatedon a salt flank in the Gulf of Mexico. The first exploration well was drilled on the basis of the 2-D map ard was abandonedas a dry ho1e,encounteringsalt at the anticipat€d pay horizon. The 3-D survey shows that this well was drilled just updip of the pinchout of the main pay interval. This is a casewhere seismic amplitudes are indicative of hydrocarbon presenceand are much easier to map out on 3-D seismic. (3) After a discovery has been made, the next step is to understandhow big it is' This is the key to deciding whether developmentwill be profltable. At this stage' appraisalwe11sare neededto verify hydrocarbonpresenceand investigatereservoir quality acrossthe accumulation.Detailed seismicmapping may reducethe number of appraisal wells needed, which will have an important impact on the overall economics of the small developmentstypical of a mature hydrocmbon provrnce' The next stepwiil be to plan the development.An exampleof the impact of 3-D on developmentplaming is shown in fig. 1.6(b) (redrawn after Demyttenaere€r 4/ ' 1993).This showspart of the CormorantField of the UK North Sea,where oil is trapped in Middle Juassic sandstonesin four separatewesterly dipphg fault blocks. The left-hand side of the figure showsthe initial map of one of thesefault blocks based on 2-D seismic data; the absenceof intemal structural complexity led to a developmentconceptbasedon a row of crestal oil producerssupportedby downflank water injectors. The right-hand side of the figure showsthe map of the fault block basedon 3-D seismic; the compartmentalisationof the fault block led to a revised developmentplan with the aim of placing producer-injector pairs in eachmajor compartment.

11 -

Uses of seismic data (a)

3-Dseismic

2-Dseismic

1975 (2-Dseismic)

(b)

O Producer I tnlector

(3-Dseismic)

---

E r o s i o ne d g e

Fig. 1.6 Changesto maps owing to 3 D seismic: (a) AlabasterField, culf of Mexico (redrawn after Creenlee?t ll/., 1994,with permissionof the authorsand the SEG)I O) CormorantField, UK North Sea (redrawnafter Demyttenaereet a/. (1993) with permissionof the authorsand the Geological Society of London).

(4) During field life, additionalproducersandinjectorsmay be neededto optimiseoil recovery.An accuratestructural map will certainly be needed,and any information that can be gleanedfrom seismicon lateralva.riationof resewoirquality will be highly welcome.A furthercontributionfrom seismicis possible:we cansometimes

12 -

lntroduction seehow the distribution of hydrocarbonsin the reservoir evolvesduring production by repeatingthe seismicsurveyafter someproduction hastakenplace (4-D seismic, discussedin chapter8). This will showwhere'for example,oil is not being swept towardsthe producerwells, perhapsbecausefaults folm a bmrier to flow; additionai wells canthenbe targetedon thesepocketsofbypassedoil. In this application,3-D seismicis essentialbecausethe better focussingand denserdata are neededto look for subtle clues to reservoir quality and hydrocarbonpresence. The decisionon whetheror when to shoot3-D seismicis essentiallyan economlc one.Doesthe valueof the subsutfaceinformationobtainedjustify the cost?This issue has beendiscussedby Aylor (1995),who collateddata on 115 3-D surveys'At the time, the averagecost for a proprietary marine survey was US $4'2 million, and for a land surveyit was US $1.2 million. The average3-D developmentsurveyresulted in the identification of six previously unknown high-quality drilling locations. It also goodfrom badlocations:before3-D theaverageprobabilityofsuccess(POS) separated of a well was 51Vo,wheteasafter 3-D the locationsfell into two groups,with 70% of locationshavinganincreasedPOSof757o,andtheremaining3070oflocationshavinga muchreducedPOSof only 17%. 3' D seismicwasalsovery effectiveat targetingsweet spotsin the reservoh:initial productionratesper well averaged565 banels per day (b/d) without 3-D and 2574 b/d with it. Using this information togetherwith information on direct 3-D suwey costs(for acquisition,processingand interpretation), andthe indirect costs due to the delay in developmentwhile the 3-D survey was being acquired and worked,it wascalculatedthatthe average3-D surveyaddedUS $14 2 million in value, most of which camefrom the addition of previously unrecogniseddrilling locationsand the higher initial production rates.Resultsfrom this limited databasethus indicate the would be instructive, positivevalueof using3-D seismic.studiesof a largerdatabase butunfortunatelyindustry-wideinformationon 3-D seismiccostsandbenefitsis elusive (Nestvold& Jack,1995). However,the oil hdustry asa whole is convincedof the valueof 3-D survey,ascan be seenfrom the growthof 3-D seismicacquisitionworldwide.Accordingto a suraey in FirstBreak,19,M7 8 (2001)),onshoreannual by IHS EnergyGroup(summarised 3-D acquisitionincreasedfrom 11000 sq km in 1991to 30000 sq km in 2000' while annualoffshore3-D acquisitionrose from 15000 sq km to 290000 sq km Over the sameperiod,2-D acquisitionfell from 260000linekm to I 10000km onshore,andfrom 1300000 km to 840000km offshore.The striking increasein offshore3-D coverage is no doubt due to the efficiency of marine acquisition andresulting low cost per square kilometre. It may also reflect the high fixed costsof marine survey' Almost all modem marineseismicis shotby specialistcontractors,who needto keeptheir boatsworking continuously; this resultsin a mix of commercial arrangements,including surveysshot exclusively for one oil company,surveysshot for a group of companies,suweys shot at the contractor'srisk with the intention of selling the final processeddataon the open market, and various hybrids betweenthem.

13 -

RoadmaD

1.5

Road map Chapter2 is devotedto explaininghow 3-D seismicdataare acquiredandprocessed. The interpreterneedsto haveat least an outline knowledgeof thesetopics,for two reasons.One of them is the needto understandwhat the limitationsof the data are. Often, the interpreteris strugglingto get as much informationas possibleout of a seismicdataset,andhasto decidehow far his conclusionsarerobust,or whetherthere is a charcethathe is beingmisledby noise.The otherreasonis thatthe interyreterwill picture,what be asked,whenhis besteffortsstill leavehim unsureaboutthe subsurface canbe doneto improvethedata.He will thensometimesfind himselfin a dialoguewith acquisitionandprocessingexperts,andneedto speaktheir language.Chapter2 aimsat equippinghim to do this. Althoughmost spaceis givento the specificissuesthat arise for 3-D, the methodsthat are no different from the 2-D casehave been sketchedin to give a reasonablycompleteaccount. Chapters3 and4 describethe basicinteryretationprocess.The distinctionbetween structural and geological interpretation is an artificial one, in the sensethat both are going on simultaneously as the interpreter works through his data. However, many interpretersspendmuch of their time making structuralmapsor planning well trajectories. Therefore, the basic mechanicsof workstation interpretation are coveredat some lengthin chapter3. Chapter4 considerssomeof the ways that 3-D seismiccan lead to enhancedgeologicalunderstanding, and what someof the problemsare,especially becauseof the limited resolutionof seismicdata. The availabilityof densegridsof datahasrevolutionisedour ability to makeuseful inferences from measuringseismicattributes,suchasthedetailedstudyofamplitudesof individualseismicloops.This topic is thereforecoveredin detailin chapter5. Inversion of seismicdatato acousticimpedanceis coveredin chapter6; this is an old ideathat hasbecomemuchmoreusefulwith the availabilityof high-qualitydense3-D datasets. It conveftsthe standardseismicsection,which emphasises the layerboundarieswhere reflectionsoccur,into a form that emphasises the propertiesof individuallayers.This can then be a stafting-pointfor discussionswith other disciplines,for examplethe reservoirengineer, An areaof rapid progressat presentis the useof morepowerfulcomputerworkstations to give the interpreter a better appreciationof the 3-D nature of the subsurface, viewing 3-D bodiesdirectly ratherthanjust seeing2-D sectionsthroughthem.This is explainedin chapter7. Thereis increasinginterestin usingrepeatedsurveysoverproducingfieldsto follow changesin seismicresponseproducedby changesin porefill (e.9.developmentof a gascap);this is anotherold ideathathasbecomemorefeasiblewith the availabilityof high-quality3-D surveys.Suchsurveys,usuallycalled 'time-lapse'or '4-D' seismic,

14 -

lntroduction are discussedin chapter8. Appendix 1 containsa brief summaryof the hardwareand workstationsin practice,andfinally softwareissuesinvolvedin managinginterpretation Appendix2 contairsa glossaryoftechnicalterms.This is not intendedto be exhaustive; a definitivedictionaryof geophysicaltermshasbeencompiledby Sheriff( 1991).

-

1.6

seismic display, units Gonventions: Thereare two topicsto mentionherethat may causeconfusionto the unwaryreader: displaypolarity andsystemsof units.Display polarity is the more importantof these; arguingaboutpolaritywasteslargeamountsof interpretertime.The problemis this: if downwards,whenwe makea wiggle we haveaninterfaceat which impedanceincreases displayofa seismictracewith thetime axisvertical,doessuchaninterfacegiveriseto a deflectionto theleft (a trough)or to theright (apeak)?Classicallyon papersections,the peakswereshadedto producea displayin which the peakswereblack andthe troughs appearedwhite, anda similarconventionis oftenusedfor workstationdisplaysso that peaksareblack or blue andtroughs arered or white. In principle, polarity is fixed at the Many time of recordingthe data,and preservedthroughoutthe processingsequence. so that a singleinterfaceis represented modemdatasetsaretransformedto zero-phase by a singleloop with somelower-amplitudewiggles on either side of it. Processors 'SEG normal' or 'SEG reverse'.This often describethe polarity of the final data as (SEG), refersto a conventionpromulgatedby the SocietyofExplorationGeophysicists accordingto which SEGnormalwould correspondto an increasein impedancedownwardsbehg represented by a peak.The reverseconvention(SEGreverse)is commonly employedin somehydrocarbonprovinces,e.g.the UK North Sea.Unfortunately,it is quitepossiblefor mistakesin acquisitionor processingto resultin final displayswith poladty oppositeto the processors'statedconvention.The interpreterneedsto check for himself what the polarity of a given datasetreally is. It might be thoughtthat this can easilybe doneby comparingthe datawith well synthetics(section3.1).However, many seismicsectionsshowlong interaalsof wigglesofabout the sameamplitudeand frequency,andoversuchanintervalitmay beeasyto establishplausibletiesusingeither polarityconvention,if bulk shiftsareallowed;suchshiftsare almostalwayspresentin A goodcheckis to Rndan isolated realdataowing to variouslimitationsin processing. interfacewith a largeand sharpimpedancechangeacrossit, which will give rise to a shongandisolatedseismicloop; inspectionof this will revealthe polarity of the data. In the marinecase,it is temptingto usethe seabedfor this purpose,but careis needed becausethe true seabedmay have been removed by the application of a fface mute during processing.In view of the potentialfor confusion,it is good practiceto state 'increaseof polarityexplicitly in discussions of seismicdata,with somephrasesuchas: impedancedownwardsgivesrise to a red (or blue,or black,etc.) loop'. This is a cumbersomeconvention,but is al leastclear.In somecases,whereconversionto zero-phase

15 -

Unitconversions hasnot beencarriedout or hasbeenunsuccessful, a singleisolatedimpedanceinterface may giverise to a complicatedreflectionsignal,with severalloopsof roughlythe same amplitude.In this case,polarity is not a meaningfulidea,anda sketchof the response ofan isolatedinterfaceshouldaccompanyseismicdisplays.A moredetaileddiscussion of theseissueshasbeengivenby Simm & White (2002). There is no uniform conventionin the industryregardingunits of distance.Both feet/inchesandkilometres/metres/centimetres units are commonlyemployed,and are often freely mixed (e.g.horizontaldistancesin metresandvertical distarcesin feei). This is easyto copewith usingtheconversions in section1.7.ln this book,both systems areuseddependingon the sourceof dataunderdiscussion.In the real world, units are almostinvariablyannotatedon displays,so confusionshouldbe minimal. Much more confusionis generated by inadequatelydocumenteddisplaysof well data;depthsmay (measured depth,the distancealonghole from a fixed referencepoint, be asmeasured e.g.the derrick floor of the drilling rig), or relativeto a geographicaldatum(usually sealevelfor marinedata),or mayhavebeencorrectedfor well deviationto givevertical depths(again,relativeto derrickfloor, sea-level,etc.).Closeinspectionof displaysof well logs is often neededto establishwhat the depthreferenceactuallyis. (A similar problemariseswith onshoreseismicdata,wherezerotime will usuallycorrespondto a datumplane at someparticularelevation,which may not howeverbe documented in a way that is easilyretrievable.)Anotherpossiblesourceof confusionin horizontal positioningof well and seismicdata arisesfrom the use of differentmap projection systems.Many differentsystemsarein use,andevenwithin a givenprojectionsystem thereare differentpossiblechoicesfor projectionparameters.This can easily cause problemsin relatingwells to seismicsurveydata.Sincenew well locationsare usually chosenfrom a seismicsurveygrid, at the intersectionof a pa.rticularinline andcrossline, it is obviouslycritical to be ableto translatethis intersectioninto a point on the ground where the rig will actually be placed.One of the problemsof 3-D seismicis that workstationvolume; interpretationis oftencarriedout on a moreor lessself-contained by differentpeopleat suchvolumesoften havea complicatedhistory of reprocessing differenttimes, and it may not be easyto check whetherthe real-worldcoordinates assignedto this volume are correct.If thereis any doubt at all aboutthe coordinate systemsbeingused,the sewicesof a specialisedsurveyorareneeded.

-

1.7 Unitconversions Conversionsare statedto four significantfigureswherenot exact. Length: 1 inch : 2.540cm 1 foot : 12 inches: 30.48cm I metre: 100cm: 3.281ft

16 -

lntroduction Density: 1 g/cm3 : 1000kglm3 : 62.43lblft3 Volume: l litre : 1000cm3: 0.03531ft3 1banel (bbl): 0.1590m3 1m3 : 6.290bbls

-

References Aylor, W. K. (1995).Businessperformanceand value ofexploitation 3-D seismic.Tre leading Edge, 14,'79'7,80t. Brcwn, A. R. (1999). Interpretation of Tbee-dimensional Seismic Data (5th edn). American Associationof PetroleumGeologists,Tulsa. Demyttenaere,R. R. A., Sluijk, A. H. & Bentley, M. R. (1993). A fundamentalreappraisalof the structureofthe CormorantField andits impacton lield developmentstrategy.lnl.PetroleumGeolog! ofNorthwest Europe: Prcceetlingsof the 4th Confercnce,l.R.Parker (ed.),pp l151 7, Geological Society,London. Greenlee,S. M., Gaskins, G. M. & Johnson,M. G. (1994). 3 D seismic benefitsfrom exploration through development:an Exxon perspective.The Leading Edge, 13.1304. McQuillin, R., Bacon, M. & Barclay,W. (1984).An Inttuduction to SeismicIntel pt"tation (2nd edn). Graham & Trotman Ltd, London. Nestvold,E. O. & Jack, l. (1995).Looking aheadin marine and land geophysics.Iie Leatling Edge,

14.ro6t 7. Schneider,w. A. (1978).Integral formulation for migration in two dimensionsand threedimensions Geophysics,43, 49 16. Sheriff, R. E. (1991). Encyclopedic Dictionary oJ Exploration Geophysict (3fd edn). Society ol Exploration Geophysicists,Tulsa. Sheriff, R. E. & Geldart, L. P (1995). E plorution Seismology. Cambridge University Press, Cambridge,UK. Simm, R. & white, R. (2002). Phase,polarity and the inierpreter'swavelet.Fit.tr B/ea,t,20,21'7-81. Telford, W M., Geldart, L. P, Sheriff, R. E. & Keys, D. A. (1976). Applierl Geophysics.Cambidge University Press,Cambridge,UK.

I

2

andprocessing 3-Dseismic dataacquisition

The aim of seismicdataacquisitionand processingis to deliverproductsthat mimic cross-sections throughthe eafth.In order to do this, the corect amountand typesof data must be acquired,and processingappliedto removeunwantedenergy(suchas multiples),and to placethe requiredeventsin the corect location.At the sametime, a balanceneedsto be struckbetweencost andtimelinessof data,while attainingalso the importantobjectivesof safeoperationsanddoing no harmto the environment. It is not the aim of this chapterto give a full accountof seismicacquisitionand processing;rather we aim'to concentrateon those aspectsthat are specificto 3-D operationsor arerecentinnovations.For thosewho requiremoreinfbrmationon energy sources,instrumentation, receiversand generalacquisitiontheorythereare a number of detailedreferencessuchas Stone(1994) and Evans(1997).In additionthereare goodintroductionsin someof the moregeneraltextssuchasSheriff& Geldart(1995) or McQuilliner a/. (1984). The vast bulk of seismicdata currentlyacquiredis 3-D, owing to the tremendous in tems of interpretabilitydiscussedin chapterL Todayit is unusualfor advantages the major oil companiesto drill explorationwells prior to a 3-D surveybeing shot, processedandinterpreted.Surveysrangein sizefrom a few tensof squarekilometres to severalthousandsquarekilometresfor explorationpurposesin for field development ftontier basins.Land 3-Ds arelesscommonthanmarine,partly becauseoftheir higher cost,but alsobecauseland well costsarerelativelylow. In order to achievethe aims outlined above,surveysneedto be plannedto cover adequatelythe areaof interest,tating into accountdata repositioningdue to migration. To achievethis, the actualrecordeddata must cover an areathat is largerthan the target areaby a migration aperture(fig. 2.1 and later in this chapter).In addition, the tracespacingneedsto be small enoughin all directionsto avoid dataaliasing. Ideally, subsurfacecoverageshouldbe uniform with a consistencybetweenthe contributionfrom different offsets (the distancebetweenthe sourceand receiver) and azimuths(the direction betweenthe sourceand the receiver).Budget, access, water curents or timing issuesmay mean one or more of these guidelineswill haveto be sacrificedand a balancestruckbetweenoperationalexpediencyand data oualitv.

18 -

3-Dseismic dataacquisition andprocessing

Targetsurveyarea

Actualfullfoldsurveyareaallowinglor migration aperture

Limitof shotsandreceivers to givefullfoldarea Fig. 2.1

Relationship between target area and acquisition area.

Modem powerful computersenable data processingto begin in the field or on the boat shortly after the acquisition has sta.rted,leading to rapid delivery of products evenfor the largestsurveys.This is importantnot just for the financialimplications of improvedtumaroundtime, but also becausedecisionson dataquality, suchas the effectsof bad weatherandincreasedswell noise,canbe madeby examinationof real dataquality.Only themosttime-consuming processes suchaspre-stack3-D migration needdedicatedprocessingcentreswith largepowerfulcomputers. -a

2.1

Marine 3-Ddataacquisition In general,3-D marinedataacquisitionis simplerandfasterthanlandacquisitionsince in all but the most heavily developedoffshore areasthere are few obstacles,leading to routine and rapid data gathering.In standardmarine acquisition,a purpose-built boat (fig. 2.2) is usedto tow one or more energysourcesand one or more cables containing (pressuresensitive)receiversto record tle reflections from the underlying rocks. At present,the source is nearly always an array of air guns tuned to give an energypulseof shortduration with directivity characteristicst}lat concentratethe energy vertically downwards. In the past, other sourcessuch as water and steam guns were used,andthesemay be encountered on older 3-D surveys.Evans(1997)givesa good descriptionof the workings of an air gun; briefly, the expansionand collapseof the air bubble in the water acts as an acoustic source that sends sound waves through the water and into the rock layers below the seabed.At changesin the rock acoustic

19 -

Marine 3-Ddataacouisition

Fig. 2.2 Examplesof modern marine 3-D seismic vessels.

impedance,part of the soundwaveis reflectedbackto the surfacewhereit is captured by the receiversandtransmittedto the boatfor fuflher processingor writing to tapefor storage(fig.2.3). In the early days (mid-1980s)of 3-D data acquisitionthe boatswere not powerful enoughto tow more than one cable and one set of guns, so the 3-D acquisition geometrywasjust a seriesof closelyspaced2-D lines (fig. 2.4).Owing to the high operatingexpenseofthis design,the surveystendedto be rathersmallandusedonly over developingfields.To reducetime wastedduring operationsseveralnovel geometries were tried including circle and spiral shooting(ng. 2.5). Thesehad varying degrees of successbut certainlyincreasedthe difficulty of dataprocessing.They do not lend themselves to modemmulti-cablesurveys.The benefitof the much clearersubsurface pictureobtainedfrom 3-D data led to a growing demandfor such surveysand rapid advancesin technology.Today,speciallydesignedseismicvesselsare powerful and enoughto tow multiple cablesand deploy two or more gun arraysthat sophisticated lines to be collectedfor eachpass arefired altemately.This allowsmultiple subsurface of the boat,significantlyincreasingthe efficiencyandreducingdataacquisitioncosts. Figure2.6 showsa typical mid- 1990slayoutof four cablesandtwo gun arrays.giving

20 -

3-Dseismic dataacquisition andprocessing

Apprcrimalely 6 kilomerres belweensourceandiudhesheeivef

cable withre@lverelemenls

ExampteRetlection Po nls betweena s nqlesholand E@ive6

Fig, 2.3 Basicsof marine acqLlisition.The boat travelsthrough the water and every few metres fires the sourcewhich emits a soundwave into the water.This travelsthrough the water and into the rock layers.At changesin the acousticpropeniesof the rock, as generally occur whereverthe lithology changes,part ot' the sound wave is reffectedback. The reflectiorl tmvels up to the surlace where it is capturedby rcceiverswithin a long cable towed behind the boat-The receiverstransmit the recordedsignal bdck to the boat where it is storedon tape and may be partly processed.

Fig. 2.4 Basic 3-D acquisition.After shooting a line, the boat Lurnswith a relatively large radius befbre shooting a line jn the oppositedirection. The boat then tums agaii and shootsa line adjacent to the original line. This is repeatedseveraltines until finally the line AB is shot. The full survey may contain severalrcpetitionsof this basic design.

eight subsurfacelines tbr eachpassof the boat. The guns fire altemately;eachgun generates asmanysubsurface linesastherearecablesbeingtowed.The gun separation is suchthat the lines recordedby one gun are interleavedwith the lines shot by the secondgun. Usually the separationbetweenlines is between25 and 37.5m. It is diflicult to reducethis lurther with a singlepassof the boatwithoutan unacceptable risk of cablesbecomingentangled, butoccasionally surveysareshotwith 12.5m line spacing

21 -

Maline3-Ddataacouisition

SpiralShooting. used aroundcircularly symmelricbodies suchas salldomes

Overlapping CircleShooting. Thiswasan attemptto replaceconventonalshaight lineshooting, removing lhe needto stopshoolingwhile the boatturned

Fig. 2.5 Early non-standard3-D marine acquisitiontechniquesdesignedfor eflicient operations.

Fig.2.6 Thebasicsof multi source,multi-streamer acquisition. As theboatsailsalong,theblue andthered gunanayfire alternately. Whentheredgunsf,re,thefour red subsurface linesarc acquired. Whenthebluegunsfire,thebluesubsudace linesarerecorded. The conliguration shown giveseightsubsurface linesper sail-lineof theboat,25 m apaftfrom eachother.

and modern steerablecables may make this more common in the future. The separation between traces recorded along a line is controlled by the receiver spacing and is usually between 6.25 and 12.5 m. As with the original marine survey design (flg. 2.4), a number of lines are shot together with the boat travelling in one direction, followed by a similar section with the boat travelling in the opposite direction. Occasionally, the

22 -

3-Dseismic dataacquisition andprocessing

Fig. 2.7 Generationof multi fold coverageduring seismic surveys.As the surveyproceeds, different combinationsof shot and receiverrecord the samereflection point. The difference betweenthe arrival times of the reflection from shot rcceiverpairs with increasingseparation allows subsurlacevelocitiesto be estimated.The addition of different tmces with the same refiection point improvesthe signal to noise ratio of the final section.

areasofthe surveywheretheboatchangedshootingdirectionmay leadto stripesin the data.This is causedby the long time intervalbetweentheseadjacentlines being shot, so that theremay be changesin tide, wave,curents or evenwater temperature(and hencewatervelocity)betweenthe two lines.Otherpatternsmay appearin the data,for instanceowing to failureof part of the sourcegun a[ay. Suchstripingmay be difficu]t to removb e y l a t e rd a l ap r o c e : s i n g . As with 2-D operations,the samesubsurfacelocationis recordedmany times by traceshavingincreasingseparationbetweenthe sourceandreceiverposition(fig.2.7). During processing,this multiplicity of datais usedto increasethe signalto noiseratio of tbe final stack,to pick subsurfacevelocitiesand to discriminatebetweendifferent recordedevents(suchasprimary andmultiplereflections). Modern boatsare capableof towing as many as 12 cableseachtypically between 4 and 8 km in length,though it is unusualto see more than l0 actually deployed. This largeincreaseof datagatheredwith eachpassof the boat meansthat,despitethe high costofbuilding andmaintainingthesevessels,their operationalefficiencypermits large explorationsurveys,coveringthousandsof squarekilometres,to be routinely acquiredin a matterof weeks.At the time of writing, boatsacquireup to 40 km2 of datadaily at costsof aroundUS $5000per squarekilometre.As a resultof the high initial cost of marinevesselsand the needto havenear-continuous operationsto be financially effective,almost all marine 3-D seismicboatsare owned by specialised senicecompanies andnot by energlcompanies. In orderto be ableto tow thecablesthroughthewateroffto the sideofthe boat,large devicescalled paravanes are usedto positionthe headof the outermostcablesto the

23 -

Marine3-Ddataacquisition

Fig. 2.8 A paravaneusedto control the position of the head of the receivercable.

requiredlocations.Figure2.8 showstheimmensesizeofthesedevices.Towingseveral closelyspacedmulti-kilomeftelong cablesrequiresthe boat to maintaincontrol over their positioningat all times.Having cablescrossover and wrap aroundeachotheris an expensiveand dangeroussituationand is to be avoidedif at all possible.To keep control,the boat needsto be continuouslymoving throughthe water at a speedof at least3 knots,andnormaloperationsarecarriedout at a speedofaround4 knots.It also

24 -

andprocesslng dataacquisition 3-Dseismic clear means that the seismic vessel requires a large turning area and needs to remaln of other shipping in the area. Often. smaller boats known as picket vessels are used cluring the seismic survey to warn other boats to keep their distance while opemtions the ale underway. The actual length of the receiver cables depends on the depth to a target and also on the expected velocity profile in the eanh A deeper target requlres good rule longer cable in order to pick velocities accurately and temove multiples A main of thumb is that the cable length should be at least as tong as the depth to thc The target level, although longer cables may be required under special circumstances to warn reflector a radar contains ends of the cables are attached to a tail buoy that and shipping of its presenceand to enablethe position of the cablesto be monitored recorded. At the end ofthe required surfaceline the boat needsasmuch as 2 h to turn safely r eady to shoot the next swathe of lines This is non-productive tirne and so data are usually acquired with the longest axis of the suwey area as the shooting direction for the sake of efficient operations.Other considerationssuch as dominart dip of the subsuface and currentstrengthanddirectionarealsotakenillto accountwhen designingmlrine seislntc (fi9 2 9) as operations. Data are usually shot in the dominant subsurfice dip dilection this is usually the direction with the smallestdistancebetweenrecordedtraces.Thete may be circumstances such as those illustrated in fig 2 10 where strike or another shooting direction is preferred. For some targets, such as those under in'egular salt be bodiesexhibiting largevelocity contrastwith the surroundingsediments'theremly jn no single optimum acquisition direction; surveysmay have to be shot severaldifl'erent more clirections to achieve good results. Currently this is rale, but is likely to become strong common as companlesexpenment with multi-direction shooting ln an areawith

._.-=.--:-:-----

stnKeurreqlon

relative Fig. 2.9 Dip and strike difeclions Over the yearsthere have bcen scveraldebateson the veloclty on hinges n1eritsoi shooting in the dip and strike dircclions. The arglrmentgenerally are not complexity.Vclocitiesare conrplexand aflectedby dip in the dip direction whereasthcy pnctlce' ln complexity this affectedby dip jn the strike direction. However.processingcan unt-avel the most is usually of shooting the direction inostacquisitionis in the dip direction.largelybecaLlse Jcn.elt .rn pl
25 -

Marine3-Ddataacquisition

Fig. 2.10 Strike shootingover a salt dome. Dip shooting (in the plane of the sectron)would see complex ray pathsdue to the presenceofthe salt, while strike shooting (in and out of the page) would be largely unaffectedby the salt.

currents,it maybe difficult for theboatto travelin or againstthedirectionof thecurent andmaintaingoodcontroloverthe cablesandthe requiredshotpoint spacing,leaving little choiceon shootingdirection. Modem seismicboatsare designedto tow large numbersand lengthsof receiver cablesthroughthe water while remainingacousticallyquiet so as to avoid unwanted noiseon the recordeddata.Specialexpansionjoints are usedin the cablesto ensure thatthe cablesthemselves do not transmitnoisealongtheir length.The streamers have neutralbuoyancyin the water so that they are easyto tow at a fixed depth (usually between3 and I I m dependingon the seismicresolutionrequiredand the seastate) belowthe sea-surface. This avoidsnear-surface noiseproblemscausedby the actionof waves.In thepastcableswerefilled with oil-basedfluidsto givethemneutralbuoyancy, but solid streamersare becomingmore common as they are both environmentally friendlierandacousticallyquieter.Specialwingeddevicescalled,birds,areattachedto thecableto keepit at the requireddepth.Theseareremotelycontrolledby an on-board computerand permit changesin receiverdepth(for instancein responseto changing weatherconditions)during operationswithout any needto bring the cablesback on board. To be able to shootand processdata accuratelyfrom a 3-D surueythe boatshave sophisticatednavigationsystems.During operationsthe boat and cablesare subject to currentsand winds, which will often causethe cablesto be pushedawayfrom the simple line-astemconfiguration,a situationknown as ,feathering,(fig. 2.11). This meansthat the positionsof shotsand receiversmust be continuouslyrecorded,and this infomation needsto be storedtogetherwith the tracedata.Todaythe entireworld has 24 h satellitecoveragethat allows the use of GpS satellitepositioningof the vesselto an accuracywithin a metre.The streamershavea seriesof compasses and transducersalong their length.The transducerssendout signalswhosetravel times

26 -

3-Dseismicdataacquisition andprocessing

ActualCableLocalion

PlannedSurvey Lrnes

+ ++ t t l t t l t

t

l

Directionoi Current Fig. 2.11 Cable featheringdue to action ofcuirent. between each other and fixed receivers on the boat give a wealth of information their position relative to each other and the boat. This information

on

is used by poweful

on-boardcomputersto provideaccuratemeasurements of absolutepositionsol every shot and receiverthroughoutthe survey.This information is later mergedwith the seismictracedata during processing.On board,the navigationis also usedto check that the suryeyhas actuallybeen shot within the requirementsdelinedby the client. In many surveysit is necessary to re-shootsomelines becausethe odginal lines were not positionedcorrectly due to wind/wave/cunenteffects.This re-shooting,known as infill, is an extra expenseand every effort is made to keep it to a minimum by accuratesteeringof the boat.On-boardcomputersystemspredictthe headingrequired to allow the boat to steeralongthe requiredtrajectorytaking into accountthe curent conditions,and ensurethat the shotsare automaticallyfired at the corect locations. Despite all this, the effectsof curents and winds usually mean that some infill is required. In areaswherethereis alreadyinfrastructure(platfom andproductionfacilities)in placeit is generallynot possibleto shootthe entire surueywith the standardmarine In such configuration.A safedistancemustbe keptbetweentheboatandanyobstacles. wherea sourceboat on a caseit is commonto usetechniquessuchas undershooting, onesideof theobstacleshootsinto a receiverboaton theotherside(seefig. 2.12).Care must be takenduring interyretationthat any differencesin the data in the undershot areaarenot a resultof the differentacquisitionparameters. -

2.2

Marine shearwaveacquisition Most marine acquisition to date has been concemed with compressional (P) wave reflections from a P-wave source, since fluids (including sea-water) do not perrnit the

27 -

Marine shearwaveacquisition

Lineof Midpoints

e

X

sourceBoat

Fig. 2.12 Principles of undershooting.Boats needto keep a safe distancefrom any obstaclesand therefore cannot acquire data close to rigs, etc. In order to record rcflections fiom undemeath such obstacles,undershootingis used.This is achievedusing two boats,a sourceboat and a receiverboat. The sourceboat and receiverboat stay on either side ofthe obstaclewith the sourceboai firing shots to be recorded by the second boat. The reflection point lies halfway between source and receiver, giving lines of reflecteddata from undemeaththe obstacle.lt is still not possibleto acquire some of the shoft offset data, but the result is much better than having a no-data zone within the survey. Since the sourceboat is not towing a large cable it is able to approachcloser to the obshuction.

transmissionof shear(S) waves.Recently(Garctta(1999)andTatham& McCormack (1991)) there has been an interestin the use of marine shearwaves,becausethey have advantagesin areaswhere gas clouds obscure deeperreflectivity and also becausethey may producestrongerreflectionsin areasof low P-waveacousticconffasts.Arother proposeduse is in areaswith very high velocity contrasts,suchas in imaging throughsalt or basaltlayers;curently there are very few documentedsuccesses.In such cases,the S-wavevelocity in the high-velocityzone can be close to the P-wavevelocity in the surroundingsediments,giving a simpler ray-pathfor the convertedray andthepossibilityof increasingthe angularcoverageat depth(fig. 2.13). Marine shearwave explorationrequiresconversionof the P-waveinput to a shear wave in the subsurface,normally at the targethorizon, togetherwith ocean-bottom receiversto allow the reflectedshearwaveto be recorded(fig.2.1q. Standardmarine acquisitioncontainssomedatathat haveconvertedfrom P to S and back to P again. Howeveqit is low in amplitudedue to the doubleconversion,and the similarity in termsof moveoutto multiplesmeansthat its identificationandseparationis extremely difflcult. The ocean-bottom receivercontainsfour phones,onepressure-sensitive hydrophone as in normal marine seismicrecordingand three mutually perpendicularvelocitysensitivegeophones.This is often referredto asfour-cotrlponentrecording,or 4C for short.A variety of differentrecordingconfigurationshavebeen developedover

{

I

{

I

1 I i

2A -

andprocessing 3-Dseismic dataacquisilion

-

P_wave

-

s_wave

Fig.2.13 The simple geomefy of convertedmodesthrough high-velocity layers.The diagram shows the potential benefit of a double converted ray-path. The high-velocity layer leads to a large amount of ray-bending for the P-wave ray-path and a lack of angular coverage at depth. The ray-paththat containsthe P to S conversionthrough the high-velocity layer is much simpler and gives larger angular coverage at ths target. In practice, it is very difficult to separateout such double conversionsbecauseof theil low amplitudesand similarity in velocity to multiples

-

P_wave

-

S-wave

Fig.2.14 Conventionalma ne shearwave geometry.Gas-saturatedsedimentsare very disruptive to the passageof P-wave energy. The P-wave reflection will therefore be weak and highly scattered On the other hand, the reflected S-waves are not affected by the presence of gas and travel through undisfurbed. The shear wave data are recorded at the seabed to avoid energy loss at a second conversionback to P-wave.

29 -

Marine shearwaveacquisition the years.Multiple deploymentsof single phonesthat sink to or are plantedin the seabedand later recoveredis one technique.More popularare cablesthat are laid on or draggedacrossthe seabedduringacquisition.A recentinnovationis cablesthat are permanently buried over a producing field and are hooked up to a recording vessel whenevera surveyis required.Although the initial surveyusing this last technique is very expensive,any repeatsurveyis relativelycheapsincethe boat is not required to lift and drag the cables.In operationsusingcables,two vesselsare neededthough eachcan be substantiallysmallerand lesspowerful than a standardmarine3-D boat. Oneboatis attachedto umbilicalsfrom the cablesandis therecordingboat,remaining stationarywhile a secondboat tows the sourceandsailsaroundthe surveyareashooting a densepattemof shots.In this way the acquisitionbecomesmore similar to land operations,whereit is alsocommonto shoota seriesof shotsfrom differentlocations into staticreceivercables.In all casesthe separation betweenlinesofreceiversis much greaterthan with standardmarine techniques,fold being built by a densecoverageof shots.Figure 2.15 showstwo examplesof marinefour-componentshootinggeometries. The sourceusedis generallya standardmarine source,althoughoccasionally with some tuning of the gun array to give a signature that is more uniform over a largerrangeof anglesthan the more directionalsourceusedin standard3-D seismic surveys. Both acquisition cosls and processingtime are greater for convertedwave data, so it is seldomusedin areaswith a good P-waveimage.Indeed,the amplitudeversus offsetinformationin conventionalP-wavedatacontainsinformationon the shearwave contrasts(seechapter5). However,the techniquehasprovedusefulin areaswherethe P-waveimageis obscuredby gasclouds.

StaggeredOrthogonalShooting Geometry 4 Cables:450 m separation 26 Shotlines:450m separation 25 m shotdistance

ParallelGeometry 4 Cables:450 m separation 21 Shotlines:flipjlop25 m shotdistance 75 m separation

7.35km


4.Skm

3.0km

30 -

3-Dseismic dataacquisition andprocessing

Fig.2,16 Land acquisitionexamples.

2.3

3-Dlandacquisition Therearea numberof majordifferencesbetweenmarineandlandacquisition.In thelatter,the shotsandreceiversaredecoupled,the surveyareahasnon-uniformtopography, and obstaclesand hazardsare more common.Land operationsoccur in a variety of areas,from bakingdesertsto fiozen wastes,andfrom junglesto cities(fig. 2.16).Each surveyneedsto be designedspecificallyfor the terain and region coveredso as to ensuregood data quality and optimisedturnaroundtime, while being non-intrusive to the environmentand ensudngthe safetyof the seismicoperation'screw and local inhabitants. The favouredland sourceis the vibrator since it is the most efficient to operate, haslittle environmentalimpact,and hasbettercontrol over the sourcecharacteristics altematives.The vibratoris a largetxck that contains than the main explosive-based a mechanicaldevicefor shakingthe groundthrougha controlledset of frequencies. Full detailsare givenin the acquisitionreferencesincludedat the end of this chapter

3l I

3-Dlandacquisition

Outerlinelofrlaidatflop

E E

E

6 receiverlines 6.64km lenoth 400 m betweenadjacentreciiverlines 165receiverstalionsat 40 metress€paration(perllne) afotal of 996 active stalions) 2 vibratorlines.101VPs Deiline 996 slationsfixedand recordedduring.acquisition ofthe 2 vtDrabrrnos Rol| moveupto nextpairofvibratorlinesis 1000mefes Flop:outerlinelefrlaid(5 lineffop) Requires:101VPs per squarekiiometr€ 75 recelverstationsper square kilomeke

.

l

Fig.2,17 Land acquisition. BasicX design. t:

Several vibrator trucks are often used together to increasethe energy of the source, and cycled severaltimes at the same location so as to be able to sum the results of different shots and increasethe signal strength relative to noise in the records. After shooting is completedat one shot location (often called vibrator point or VP for short), the trucks drive to the next location. In areasof easyaccessand few obstaclesthis is a relatively efficient processand it is possibleto shoot asmany as 800 \?s per day. This is, however,still far below the efficiency achievedin marine work. Sourceand receiver positions are surveyed(generally by Global Positioning System,GPS) and are marked in advanceby flags so as to allow rapid movementto the next VP. Once a section has been shot, the cable is picked up by the field crew and addedfurther along the line ready for later shooting, so asto allow continuousoperationsduring daylight hours. Many different layoutsof shotsandreceiversareavailablefor land operationsin areas without obstructions,and the skill of the surveydesigneris to ensurethe most effrcient arrangementthat provides the required data quality. Figtres 2.17-2.19 show how the

1 I

It

1 I I

t t ,

I

32 -

3-Dseismicdataacquisition andprocessing

Fig.2.18 Land acquisihon.Completion of a swatheof data by repetitionof the basic X design.

Fig.2.19 Land acquisition.Completion of the suNey by repetition of many swathes.

33 -

3-Dlandacouisition total surveyis generallybuilt up from continuousrepetitionof the basic acquisition patternor fbotprint.In fig. 2.17 we seesix receiverlines with shotlines forming an X pattern.As the shotsmoveposition,differentsectionsof the geophonelines are made live by the recordinginstrumentsin orderto maintainthe samerelationshipsbetween sourcesandreceiversasin fig. 2.17.Figure2.18 showspartof a full line swathewith a full shotpatternin place.The whole arrangement is thenmovedfor the next setof six receiverlines (fig. 2.19).In this casethereis an overlapof oneline betweenone setof six receiverlines andthe next set.Specialsoftwareis availableto the designerof land seismicsurveysto allow determinationofthe fold, azimuthalcoverageanddistribution of source-receiverdistanceat any location.Balancesneedto be struck betweenthe amountof equipmentavailableand the needto continuallypick up and move cables andreceivers.Usually in land acquisitionit is not economicto get the simpleuniform geometryof marineoperationsand often thereis a small overprintof the acquisition designin the final processed data. In areasthat are difficult for vehicleaccess(e.g.mountains,forests),an explosive sourceis used.Usuallythe areacoveredper day is not asgreatasachievedby vibrator trucks.Holesneedlo be ddlled into the earthto ensurethe explosivechargesare well coupledto bedrockandnot fired in the shallowerweatheredlayers,which would cause excessive amountsofnoise in thedata.This reducesthe acquisitionratesto somewhere between50 and 100 shotpointsper day.The lack of vehiclesmay meanthe majority of movementof equipmenthas to be by helicopter,an important issue for safety. Anotherissueis the amountof time that may haveto be spentcuttinglines in lbrested areas. Whatevertype of land sourceis used,the entire surveyneedsto be correctedfor the arival time changesdue to both topographyand variationsin the thicknessof the near-surface layer (fig. 2.20).The nearsurfaceis generallyheavilyweatheredand usuallyhasalteredacousticpropeftiescomparedwith the lessaffecteddeeperlayers. This generallyresultsin a much slowervelocity,but in areasof permafrostthe near surfacemay be substantiallyfaster.Usually a separatecrew is responsiblefor drilling andmeasuringupholetimes(thetime requiredto travelfrom the bottomof the drilled hole to the surface)throughoutthe area.Theseare then usedto determinethe depth to the baseof the weatheredlayer, and the velocitiesin the weatheredand bedrock layersfrom which the staticcorectionscan be derived.Sheriff& Geldart(1995)give an introductionto thesecorrections.It is impossibleand uneconomicto drill holes at every shot and geophonelocation,so first anival times from the field recordsare alsousedto estimatestaticcorections.In somecircumstances, specialreftactionprofile crews are usedto determinethe near-surface proflle. velocity When calculating the staticsto be applied,it is importantto ensureconsistencyacrossthe area.The largeredundancyin 3-D datameansthat thereis often conflictinginformationabout the staticsrequiredat any one location and specialsoftwareis usedto generatethe corlectlons.

94

andplocessing dataacquisition 3-Dseismic

-l

For land acquisitionthe surfacetopography Fig. 2.20 Weatheringlayer correctionsfor land data andthicknessoftheweatheleozonemustbecorectedfotsinceotherwisetheunderlyingstlucture layer changes uphole infolmation together would have an overprint of elevationand weathering th€ near sudace Once the velocity of the with refraction recordsis used to build a model of bedrockandweatheringlayer'ndthedepthtothebaseoftheweathefinglayefhavebeen shot and receiver statics The statrc at any determined, the correction is simply the sum of the - EilV*whete E"is the surfaceelevation'E* isthe to"utioni,ssl.pfv (f. - E*)lv*+(E, y' and yb elevationofthe seismicdatum' and elevationof the baseof the weatheredlayet Eris the are the velocitiesin the weatheredand bedrock layers

--

survey typesof seismic 2.4 Other andland surveysmustterminatein Marine surveysrequirea mrmmumdepthof water the transitionzoneandsurveys areasclosefo the shore.The zonein betweenis termed speciallydesignedswampbuggies involvea varietyof specialisedequipmentsuchas tbeseareasis moretrme consumrng with shallowdraughts.Althoughthe operationin tides,they canbe consideredasa andhas someextraconsfialntssuchasthe timing of mixture of madneandland operations' generallyover a producingfield' 4-D seismicsurreys are repeatseismicsurveys' to determinedifferencesin seismic that are speciallyshotat incrementaltime periods referredto in the title is time' the ..rponr"iu" to production.The fourth dimension of two horizontalco-ordinarcsand other three dimensionsbeing the standardones fluidwithin thereservoirrnaychange one verticalspatialco-ordinateBoth pressureand propertiesof the rock' Even the during production andboth of theseaffect the acoustic dueto compartmentalisation of a changein part of the reservoirmay indicate absen-ce marine'with one to four component faulting, etc. These4-D sun/eysmay be land or marine conventionaltowed-sueamer receivers,althoughthe majorityto d;G havebeen

35 -

3-Ddalaprocessing surveys.Jack(1997)givesan excellentintroductionto the subject,which is considered in moredetailin chapter8.

-

2.5

3-Ddataprocessing As with the sectionon dataacquisition,we do not aim to giveherea completeoverview of seismicdataprocessing. Rather,we will againconcentrate on thedifferencesbetween 2-D and 3-D. Yilmaz ( 1987)gives a detailedbut very readableaccountcoveringall aspectsof dataprocessing. In mostrespects3-D dataprocessingis very similarto 2-D dataprocessing.Indeed, any processthattreatstracesindividually,suchasstatisticaldeconvolution, gainrecovery,tracemuting andfrequencyfiltering,is appliedin an identicalmannerin 3-D and in 2-D. Someof the otherprocesses are appliedin a 2-D sense,alonglinesratherthan acrossthe 3-D datavolume. Thereare threemain processes that are differentfor 3-D data:binning, spatialfiltering, and any processthat repositionsdata such as migrationand dip moveout.In additionseveralotherapplications,althoughgenerallyrun on selectedlines from the 3-D volume,may benefitfrom the pdor applicationof one of the above.An example of such a processis velocity analysis,which is normally run on a regulargrid of 2-D lines but will improvein quality after 3-D migration.An imporlantaspectof 3-D dataprocessingis to ensurespatialconsistencyof parameters suchasthe velocity field. Figure 2.21 showsa typical processingsequencefor 3-D data.We will briefly run throughall the steps,but thosethat are significantlydifferentfrom 2-D processingare highlightedin bold andwill be discussedin moredetail. Modern commercialprocessingsystemscontainhundredsof algorithms.Thereare ofien severalwaysto tackleanygivenproblemandthe processinganalystmay haveto try severaldifferenttechniquesandparameters beforebeingsatisfiedthatanypafticular stepis optimised.It is not possiblein this book to evenintroduceall of the techniques but the readeris refenedto Yilmaz (1987)and Hattonet al. (1986)for more detailed descriptions.

2.5.1 Reformat, designature, resampling andgainadiustment The first steptn frg.2.21 is a refomatting operation.This just takesthe datacoming from the receiversand puts them into trace order.Normally the data are written to tapeat this stagein one of the designatedindustryfbrmatsso that the raw recordsare retainedto form the basisof possiblelater reprocessing. If startingfrom field tapes, reformattingincludesconverlingthe datafrom standardindustryfomat into whatever fbrmat the processingsystemuses.The secondstep,designature,takesthe wavelet

36 -

andprocessinq 3-Dseismic dataacqulsition l. Reformat 2. Designature 3. Resamplefrom 2 to,l ms 4. Low cut filter (5/12 minimum phasefilter) 5. Removebad traces 6. Merge navigationwith seismicheaders 7. Spatial resamplingftom I 2.5 to 25 m groups.Normal moveout (NMO) conection, K-litter, tracedrop, inverseNMO 8. Sphericaldivergencegain conections 9. Deconvolutionbetbre stack 10. Shot interpolation to double fold in CMP gathers I I . Radon demultiple on interpolatedgathers 12. High-frequencynoise removal 13. Drop of interpolatedtraces 14. Flex binning increasing from 37.5 m on near to 50 m on far offsets 15. Sort to common offset 16. Dip moveout (including approximateNMO) halving the number ofoffset planeson output 17. Pre-stack time migration using constant velocity 1600 m/s 18. InverseNMO 19. Re-pick velocities (0.5 km grid) 20. NMO Processinghereattercontinuing on three volumes: 21 . Stack to generate three volumes: near offset stack, far offset stack and full offset stack 22. 3-D constantvelocity inversetime migntion 23. Bulk static (gun and cable correction) 24. K notch filter to remove pattem causedby the acquisition 25. FXY deconvolution 26. FXY interpolation to 12.5 m x 12.5 m bin grid 27. Pre-migrationdata conditioning (e.g. amplitude balance,edge tapers,etc.) 28. One pass steep dip 3-D time migration (using time and spatially varying velocity field) 29. Zero-phaseconversionby matching to wells 30. Spectralequalisation 31. Bandpasslilter 32. Residualscaling Fig. 2.21 Typical 3-D processingsequence.

that was createdby the sourceand convertsit to a more compactform. For instance, air gunsoutputa signalwith a main peak,followed by a smallersecondarypeakdue to the re-expansionof the air bubble.Sucha sourcesignalis undesirablesinceevery reflectionwould be followed by a smallerrepetitionof itself. Designatureremoves the secondpeak,giving the inputwaveleta morecompactfom. A decisionneedsto be A zero-phase madeat this pointwhetherthe outputshouldbe zero-or minimum-phase. waveletis one thatis symmetricalaboutits centre,while a minimum-phasewaveletis onethatstartsat time zeroandhasasmuchenergynearthe startasphysicallypossible (fig.2.22).Therearearguments on both sides.It is certainlydesirableto usezero-phase

37 -

3-Ddataprocessing

Fig. 2.22 Comparisonof zero-phaseand minimum-phasewavelets.Also shown are zero-phase waveletsthat have had their phasesrotatedby 45' and 90'; the latter developsan antisymmetic 1blm. The desiredoutput fiom seismicdata processingis usually a seismic sectionthat represents the eanh reflectivity convolved with a zero-phasewavelet.becausesuch a wavelethas the greatest resolution tbr any given bandwidth. Seismic sourcescannotbe zero-phasesince that would imply output before time zero. Most air guns give signaturesthat are close to minimum-phase.This has the maximum amount offronFloading of the waveletpossiblefor any given amplitude spectrum. During processingthe data are conveftedto zero-phase.

suchasvelocityanalysiswill benefitfromsuchdata.However, datasincelaterprocesses higherfrequenciesin the waveletas it passesthroughit, and this the earthattenuates altersthe phaseof the wavelet.Even if the waveletgoing into the groundwere to be optionis zero-phase, it would not remainso at depth.To combatthis, if the zero-phase chosen,thenoneneedsto accountfor thephasedistortionwith depthby alsoapplying a deabsorptionfilter. If the minimum-phaseoption is chosenthen the applicationof the deabsorption filter is usuallyneglected,sinceabsorptionitself is a minimum-phase process.Thus,the waveletremainsminimum-phase(but sinceit is not a constantphase theentiresectionlater wavelet,its changewith depthmakesit impossibleto zero-phase on). Figure2.23 showsa waveletbeforeandafterdesignature. Marine dataare normallyrecordedwith a 2 ms samplinginterval.This is sufficient to recordfrequenciesup to 250 Hz, much higherthanthe frequenciesthat areactually recordedfrom theearth,pafiicularlyat thedeepertargetlevelswhereit is uncommonto higherthan30-50 Hz. Thedataareusuallyresampledto 4 ms,which recordfrequencies is sufficientforfrequenciesup to 125Hz. An anti-aliasfilter is appliedto ensurethatany This resamhigherfrequenciesin the nearsurfacedo not aliason to lowerfrequencies. pling reducesthevolumeof databy 5070andspeedsup all of thelaterprocessingstages

38 -

3-Dseismic dataacquisition andpJocessing

.IAV-

"''.*,

r

I

\-

Fiq. 2.23 Example of a wavelet before (top) and after (bottom) designature.The sourcefrom the air gun array still containsremnantbubble oscillations.A filter is applied to remove theseperiodic eventsfrom the rccordedseismic.The sharpeningof the wavelet at this stagegives increased resolution in the seismic data.In this case.no attemDthas been made to convert the waveletto zero-phase.

from thedata.TheseareoftenheavilycontamStep4 removesthelowestfrequencies inatedby noise(e.9.swell noisein themarinecase),which may swampanyunderlying signalandis bestremoved.Step5 is an editingstep;anyffacesthatappearexcessively noisyowingto poorcouplingor equipmentfailureareremoved.Step6 is theimportant processof assigningthe correctpositioninginformationto the tracesasreferredto in the acquisitionsection,so lhat actualoffsetsandlocationscanbe determined. Just as step 3 was associatedwith resamplingthe data in time, then step 7 is an identicalprocessin space.Typicallythereceivergroupintervalwithin a marinestreamer is 12.5m. This givesa CMP intervalof 6.25 m, which in all but the steepestdip areas givesan over-sampling in the inline direction.The dataaremerelyresampledto mimic

39

-I

3-Ddataprocessing 25 m receivergroup intervals,with every other trace being discarded.As with the resamplingin time,an anti-aliasfilter is appliedbeforethis step.Theresamplingfurther reducesdatavolumes,by a secondfactorof two. Step8 appliesa time-varyinggain to the datato boostup the amplitudesof the later arrivals comparedto the earlier ones.As the wavefront from the sourcetravels deeper into the earth,it coversa larger areaandalsosuffersamplitudedecaydueto transmission lossesandattenuation.Sphericaldivergencecorrectionis appliedto removethe lossin amplitudedue to the wavefrontexpandingwith depth.This expansionmeansthat the sane energyin the wavefront is spreadover an increasingareaasthe distancetravelled by thewaveincreases, andhencetheamplitudeof thewaveis less.Onecanseetlle same (in effect a 2-D sense)on wavescausedby droppingpebblesinto water.Near where the stonewas droppedin, the perimeter of the circular wavefront is quite small and its amplitude is large,but asthe wavefront expandsits amplitude decreases.In addition to a sphericaldivergencecorrection it is common to apply an additional exponentialgain functionwith time to accountfor the transmissionandattenuationlosscs.

2.5.2 Deconvolution Step 9, deconvolution,is a processthat sharpensthe waveletand removesany short period reverberations. The theory is explainedin a numberof texts; see,for exam_ ple, Robinsonand Treitel (1980).A digital operatoris designedfor eachtracethat is convolvedwith thetraceto removeunwantedringiness.The operatoris designedauto_ maticallybasedon characteristics of the tracesard a few simpleparameterssupplied by theprocessinganalyst.Theseincludethe operatorgap.The ideais thatthe operator will not changethe waveletfrom time zeroto theend-timeofthe gap,but try to remove periodicityat timesbeyondtheendof thegap.Deconvolutionmaybe predictiveor spik_ ing, with the differencebetweenthe two beingthe lengthof the operatorgap.A shorr, or no gap,givesmaximalwaveletcompression(hencethe namespiking)while a large gap (32 ms or more) attemptsto removeperiodicitycausedby shortperiodmultiples whoseperiod is longer than the gap. In practice,both forms actually perform a mixture ofwaveletcompression andde-reverberation. Careshouldbetakenwith spikingdecon_ volution sincetheoreticallyit assumesminimum-phaseinput data.The outputof data thathaveundergonetoo severea spikingdeconvolutionprocessis relativelyvariablein phase.Becauseofthis, it is becomingrareto applyspikingdeconvolutionbeforestack. Consistencyof waveletshapeandamplitudebecomesincreasinglyimportantas more attemptsare madeto infer subsurfaceinformation from the amplitudesof the reflection even$.

2.5,3 Removingmultiptes One of the consequences of dual sourceshootingis that the numberof tracesin a comrnon midpoint (CMP) gatheris half that of single sourceshooting.Each source

40 -

andprocessing dataacquisition 3-Dseismic

Multiple Energy

Fig. 2.24 CMP gatherat 100 and 50 m trace spacing.Comparisonof CMP gather at original trace spacingand after shot record interpolation(seefi8. 2.25) to interpolatetracesso that the new gather has half the trace spacingof the original. The high-energymultiple packetis badly aliasedin the original gather,making removal ofthese multiples very difncult. After shot record interpolation the multiples are better sampledand thdre is little aliasedenergy

contributesto dataon differentsubsurfacelines (fig. 2.6). The time intervalbetween shotsis unchangedfrom the singlesourcecase,becausea sourcecannotbe successive flred until the listeningperiodfor the previousshothasended.When the dual sources arefired alternately,therearethereforehalf asmany shotscontributingto everyCMP. This may lead to dataaliasingof any steeplydipping multipleswithin a CMP gather. For example,a commongeometryis to have25 m receivergroup inte als within a cableandtwo sourcesfired altemately,with a shotevery25 m. This meansthat each sourceis fired only every 50 m. The fold in a CMP gatherwill be only one-quarter of the number of tracesin a shot gather,and the offset incrementin a CMP gather will be four timesthat in a shotgathet.Fig'.re 7.24 showsan exampleCMP gatherat 50 and 100m tracespacing.To reducethe possibilityof aliasing,additionalshotsare methods createdby interpolationbetweenthe actualshots.Generally,frequency-space are employedsincetheseusethe lower unaliasedfrequenciesto generateestimatesof The procedureis explainedby the srackingdiagramin fig 2.25. thehigherfrequencies. The shot recordeddata are sortedinto receivergathers;eachreceivergatheris then eachexistingtrace.Thesetracesarethenused interyolatedto addonenewtracebetween CMPSUsuallyshotinterpolation togetherwith theoriginaldatato form better-sampled has beenperformedthe additional and when this is requiredfor multiple suppression (invented)ffacesare discarded.

41 -

3-Ddataprocessing

Interpolation occurswithin the receiver gamer Shot posttron along theline

a

* a a rFa

t

a a

a

t

r

a

a

r

a

a

a a a a ! a

!

a a a a a a a a a

* . .

a

a

a

a

r

a

a a * a

a

a

t

a a

* a a a a a r

a

a

a Traceswith a Common Midpoint Receiverpositionalong the line

Fig. 2.25 Stacking diagram to illustrate shot rccord interpolation.The grey receiversmark the position of the original recorded data. The data are sorted into receiver gathers in which the trace spacingis halfwhat it is in CMP gathers.Interyolation is used to crcatethe additional tracesshown in red. Once sortedinto CMP gathersthe number of traceswithin each gatherhas doubled and the trace separationhas halved.

Fig.2.26 Primary andrnultiple reflections.The red andpurpleeventsare primaries;theyhavea singlereflectionalongtheray-path.Thegreeneventhasmultiplereflections andin thiscaseis the first-ordermultipleof thepurpleevent.Thetimingofthe greeneventmaybe simila.rto underlying primaries,andif it is not removedit mayobscurethedeeperreflectivity.

The next stepin fig. 2.21 is theprocessto remove multiples from the data.Figure 2.26 explains the differences between primary events and multiples. Here again there are a large number of choices of technique. Methods are based either on velocity moveout differences, or on prediction based on the timing and geometry of the cause of

42 -

andplocessing dataacquisition 3-Dseismic the multiple. In the fomer classare the most commonlyusedtechniques,Radonand FK demultiple.In both casesthe data are normal moveout(NMO) correctedto approximatelyflatten either the multiple or primary events.(The conceptof NMO is explainedin Appendix 2 and the equationsfor it are given in section 3.3.3) The data are then transformed to Radon or FK space and the unwanted events are removedby rejecting(or passingandsubtracting)the multipleevents.The methodswork and becausemultiples spendlongertravellingin the (generallyslower)near-surface hencehavea moveoutvelocity that is slowerthan the primary eventsarriving at the sametime (ng.2.27). Multiple eliminationin areasof shongwaterbottom reflection

?Br$

, la{a I

*ffi

w & 1

914 I

tlgB

Stll I

Fig, 2.27 Gather showing primary and multiple events,illustrating moveout dift-erencesbetween the two. In this examplethe pdmary NMO velocity has beenusedto flatten the primary events The multiples, in this casefrom a deep seabed,are dipping from top left to bottom right. This example showsthe removal of the multiple by the Radon transform technique,where eventswith moveout on the farthest offset sreater than 100 ms and less than 1500 ms were temoved.

43 -

3-Ddataprocessing coefficientsor with large velocity contrasts(such as sub-saltplays in the Gulf of Mexico) is an areaof activedevelopmentwithin the geophysicalindustry.Recently introducedwaveequationmethodsarebasedon the predictionof the multiplefrom the primarylocation.Thesetechniquesshowconsiderable promisebut theyare expensive comparedwith moveouttechniquesanddo not easilylend themselvesto true 3-D implementationowing to acquisitionlimitations.Currentlythey are appliedto datain a 2-D senseandfail if thereis a strongcross-dipcomponentin the multiple-generating surface.

2.5.4 Binning Databinning is the processwhereeverytraceis assignedto midpointlocations(i.e. a locationhalfwaybetweenshotandreceiver).To achievethis the suryeyareais coveretl by a largenumberof rectangularbins as shownin fig. 2.28.Tracesare then assigned a CommonMidpoint (CMP) location at the centreof the bin in which they fall. In rnarineacquisition,if the streamerswere perfectly straightbehind the boat, all the bins would containa regularsamplingof offsets,which would be highly desirablefor seismicprocessing. However,because offeatheringthisis not thecase.In pafticular,the

Planned CDPlines

OriOinaf Uinbasedon regutar geometry !

Fig. 2.28 Data binning. The size of the bins is determinedby the original acquisthon..l.he bins are centredon the plannedlocationswith width equal to the CMp (also known rs CDp) sDacineand length equal ro the line spacing.

44 -r

andprocessing dataacquisition 3-Dseismic

f

oin orioinar

ffiein

trace "rp^n"iot C Near-offset

aFar-otfsettrace

Fig. 2.29 Bin expansion. Flexi-binning is used so as to ensurc that each bin has a regular distribution of ffaces with thg futl range of offsets. Thrce examples ale illustrated here. Bin A is expanded for a near offset trace with relatively small expansion. Bin B is seeking a trace with larger offset. The bin expansion is greater and has thrown up three possibilities. The trace closest to the bin cente is the one that gets borrowed. Bin c already has a far offset trace and no bin expansion is required. The red trace will end up being used by both bin A and bin D After binning, the traces are treated as though they came from the centre of the bin though it is normal to retain and use the absolute offset information for later processing

furthest offsets may be non-uniformly distributed with somebins having two or more traceswith a similar offset while othershavenone. To make the distribution of offsets moreuniform a processcalled flexi-binning is used Herethe bins are extendedby fixed amounts(generally) in the crossline direction, perhapsby as much as 507oeither side (fig.2.29).lf no ttacewithin a certain offset rangeis found within the original bin, then a trace within the extendedbin is usedasthough it falls in the original bin. If more than one trace within an offset rangeis found, then it is usual for the trace closestto the bin centre to be used. The bin expansionis normally small for the near offsets, since the cable position is better controlled close to the boat, and grows with increasing offset' Note thattheprocessasdescribedjustcopiesa tracefrom onelocationto anotherThis copying meansthat there may be small static shifts between traces within a binned CDP, since one trace may come from tlte extreme top edge of a flexed bin and the next offset may come from the bottom edge.Figure 2.30 shows a binned gather with

45 E

3-Ddataprocessing

Fig.2.30 Exampleof a binnedgather

somecleardifferencesbetweenadjacenttraces.This may substantiallyreducethe final ftequencycontentofthe stackeddata.Somecontractorsoffer full interyolationoftraces onto bin centresor full pre-stackmigrationproceduresthat can conectly handlethe real tracepositioningratherthanassumingtracesareregularlysampledat bin cenftes, both of which providefor a mathematicallybettersolution.Owing to the cost this is not routinelyappliedat present,but with theever-reducing costof computerpowerand the desirefor higher-frequency datafor improvedinterpretationit is becomingmore common.

46 -

andplocessing 3-Dseismic dataacquisition

Fig, 2,31 Seismic migration. During seismic data recording the reflectjon from point A will be recordedby a coincident sourceand receiverat location B. It will be plotted on a stackedsectionas a point A* vertically below B with distancegiven by length AB. Migration movesthe point back to its subsurlacereflection point and plots it vertically below point C, thus giving a sectionthat looks more like a cross-sectionthrough the earth.

andmigraiion 2.5.5 Stacking thedatafor seismicmigration. areto prepa.re The next stepsin theprocessingsequence Migration is one of the key stepsin seismicdata processing- it is the step which attemptsto move the recordeddata so that eventslie in their correctspatiallocation ratherthan their recordedlocation (flgs. 2.31 and2.32). The mathematicsof how to perform this processis well definedby the wave equationand has beenextensively researched over the pastthreedecades.There are a numberof texts,suchas Stolt & written Benson(1986),Berkhout(1982,1984)andBancroft(1997,1998),specifically are more there aboutseismicmigrationthat describethe mathematicsin detail. Once a large numberof optionsrangingfrom migratingall the pre-stackdata to stacking datain a CMP followed by post-stackmigration.Thereis alsothe issueof whetherto usetime or depthmigrationand also the type of algorithm(Kirchhoff, implicit finite difference,explicit finite difference,FK, phaseshift, etc.).In recentyearsthe choice hasbecomeevenwider with the ability of somealgorithmsto incotporatethe effects of velocityanisotropy.To a largeextentthe detailsof the algorithmareunimportantto the interpreter.What mattersarethe accuracyandcost.Thesearedeterminedby issues suchasthe largestdip thatcanbe properlymigrated,the frequencycontentof the final migrationand the time requiredto perform the operation.The choiceof whetherto migratedatabeforeor after stackingis largely dependenton the velocity regimeand the subsurfacedips presentin the data.Large dips may meanthat shallower(slower) eventsarrive at the sametime as deeperevents,giving rise to the two eventsneeding to stackat the sametime but with differentvelocities(fig. 2.33).Suchan occurrenceis known as a stackingconflict andthe solutionis to migratethe databeforestack The

47 -

3-Ddataprocessing

DepthModel

distance

distance

Fig. 2.32 Seismic migration moves events fiom their recorded position to their true subsurface position. This diagam shows the effect of migration on a syncline. The rays are reflected at right angles to the surface. It can be seen that the blue rays from the right hand of the syncline are sometimes recorded to the left of the green rays reflected from the left side of the syncline. This gives a complicated bow tie pattem in the cofiesponding time section. The effect of migration is to move the events back to their true subsurface location.

earliest forms of migration were basedon an approximation to the wave equation and ignored ray-bending at velocity boundaries.Today such techniquesare known as time migration methods.In areasof lateral velocity contraststhe straight ray approximation of timemigrationcanseriouslymispositionevents.The solutiontothisproblemis depth migration, which correctly handlesvelocity variation. Of course,thereis the additional complication of deriving the velocity model neededto drive depth imaging and this is anotheractive areaof researchand development.Most methodsrequire al iterative approach.An assumedvelocity-depth model is used, the data are migrated pre-stack using this model, and the images acrossthe migrated CMP gather are compared.If the model velocity is too high then the further offsets will have the event positioned too deepcomparedwith the near offsets, and converselyif the velocity is too low the eventswill be imaged shalloweron the far offsetsthan on the nears.Only if the velocity

'::

:t ,"

4A -

3-Dseismic dataacquisition andprocessing

E Fig, 2.33 Example of eventscausinga stackingconflict. Rellectionsfrom both B and C will be recordedat the point A with approximatelythe sametravel time. Since eventC is higher in the sectionit requiresa slower velocity to stack the eventthan the deeperrellection B. This createsa stackingvelocity conflict that is solved by partially migrating eventC away from eventB. Provided the migration velocity is slower than the true earth velocity the partial migration will move the dipping event to somewherebetweenits unmigratedlocation (B) and its true location (C).

modelis nearlycorrectwill theeventsappearflat.Figure2.34showsan iterationfrom a pre-stackdepthmigrationwherevelocityanalysisis beingperlbrmed.Eventswill only be flat if theyhavebeenlully migratedin a 3-D senseandit is expensiveto repeatthis for the entiresection.Usually the velocity iterationstepis perlbrmedonly on a series of targetlines using an algorithrn(suchas one of the vadantsof KirchhoftJthat can outputonto a discretegrid without havingto migratethe entiredataset.Sincelateral velocity vadationsalsogive rise to stackingproblemsmostdepthmigrationbenefitis gainedfiom working pre-stack. In the eadydaysof 3-D data,migrationwasperformedastwo runsof2-D migration. First the inlines were2-D migrated,and thenthe datawould be sortedinto crosslines and migratedagain.This was known as two-passmigrationandis relativelycommon on older surveys.It was usedbecausecomputerpower and memory were not large enoughto allow thefull 3-D datasetto be migratedin onepass.The two-passtechnique is theoreticallycolTectonly for a constantvelocity earth,and once computerscould hold sufficientdatato allow full 3-D migrationit actuallybecamecheaperto perform the migrationin onepassratherthanthe two becauseit avoidedthe sortfiom inline to crosslineorientation.Todayit is extremelyrarefor 3-D migrationto be performedas two 2-D migratlons. The migrationapproachoutlinedrn frg. 2.21 is a hybrid pre-stacktime migmtion comesfrom an areawith approachandcoversstepsl5 to 28. The processingsequence relativelysteepdips but without largelateralvelocity variations.Therefore,pre-stack

49 -r

3-Ddataprocessing

Fig, 2,34 Examination of CMP gathersduring an iteration of pre-stackdepth migration. pre-stack depth migration is requiredin arcasof complex lateral velocity variation. In such areasthe standard normal moveout equation is insufncient to align events owing to the different paths that the waves take through the subsurface.In order to apply pre-stackdepth migration one needsto supply the corect velocity-depth model. This is generally not known and so is built in an iterative fashion during the pre-stackmigration. The correct velocity model is the one where the migratedgathers are flat acrossall offsets.Here we seegathersdudng a pre-stackmigration iteration. Some of the eventsare flat, indicating that the correct model has beendetemined. Othersare still dipping and the velocity model needsupdating to flatten them also.

time migration is the preferredsolution.Sincefull pre-stacktime migration is still relativelyexpensivea cheaper,fasterapproximationhasbeenused.As computerpower continuesto increase,theseapproximationsare being replacedby the full pre-stack migration solution.Steps 15 and 16 are an applicationof Dip Moveout (DMO), so calledbecauseitremovestheeffectofdip on stackingvelocitiesandtracepositions.As part of the DMO process,onealsoneedsto apply an approximateNMO corection. At this stageit is usualto havepickedvelocitiesapproximatelyon a sparsegrid, sayevery I km in all directions.This is generallysufficientsincedetailedvelocity analysisfor final stacking(maximumanalysissepararionof 0.5 km) is performedafterDMO and constantvelocily migrationin step 19. Thereis an excellentarticle on the benelitsof DMO by Deregowski(1986)andmoredetailon algorithmsin thenotesby Hale ( 1991). One of the benefitsof DMO is that it takesa constantoffset sectionand transforms

I

( '' :l

..

50 -

3-Dseismicdataacq[isitionandp]ocessing

the datato a zero-offsetsection,thus allowingconventionalpost-stackmigrationto be applied.This is how the pre-stackmigrationprocessin step17 is applied ln this step' we arenot attemptingto geta final migration;insteadwe aretrying to moveeventswith potentialstackingconflictsapafisufflcientlysothatwe candeterminea uniquestacking velocity.This allowsus to use a computationallyvery fast constantvelocity forrn of migration- in this exampleperformedwith a velocityof 1600nr-/s.It is importantthat this stepdoesnot usetoo high a velocity sincethen eventscan disappearcompletely if they are steeplydipping.Generallyone usesa velocity that is closeto the slowest seenin the section(1480 m/s for water velocity). Step 18 removesthe approximate corection for NMO appliedduringthe DMO process(step16) andthis allowsfor the detailedvelocity analysisto be performedand appliedin steps19 and 20 The data are usually savedto tape after DMO and the removalof the initial NMO correction to allow any later re-processingto begin from this stage.It is commonpracticeto performthe final velocity analysisat a spatialinteryalof 250-500m thoughthis too is The spacingneededdependson reducingin the desireto retainthe higherfrequencies. the variabilityin the velocityfield; largevariationsrequiremorefrequentanalysisthan moregradualchanges.For really detailedwork, the adventofbetter automaticvelocity picking algorithmsmeansthat it is possibleto pick velocitiesfor everyCMP location within the targetarea. After theflnal velocityanalysisandmoveoutco ectionthedataarestackedStacking togethertracesthatcontainthe samereflectioninformationboth improvesthe signalto (random)noisecontent(by thesquareroot of thenumberoftracesstacked)andreduces any residualcohefentnoisesuchas multipleswhich stackat velocitiesdiffelent flom the primary events.During stacking,mutes(zeroingthe datawithin specifiedzones) areappliedto thedatato ensurethatNMO stuetchis not a problemandthatanyresidual thestackedsection(Rg 2 35)' multiplesleft on thenear-offsettracesdo not contaminate Theremay be someamplitudevariationwith offset (AVO) eff'ectsin the data,which can be usedas hydrocarbonindicators,so this informationis retainedby stackingthe data over selectedoffset ranges.Chapter5 containsmore detailson the physicsand useof AVO data.An offsetto anglerelationshipis usedthatdependson the subsurface velocitiesto definemute patternsto producetwo or more anglestacksas well as the from thispoint onwardsneedsto be applied stackoverall usableoffsets.All processing to eachof the indiridualstackedsections The result of the stackingafter step 17 is a numberof partially migratedsections requiringthe removalof the constantvelocity migration.This is doneby step22, an inversemigrationor demigrationusingthe sameconstantvelocityasusedfor the initial mlgranon. Step23 addsa staticcorection basedon the depthof the sourceandthe receiverso Step24 removessomeof the effectsof the pattem thatthe datumis now at sea-surface. left in the data from the acquisitionvariability.SincedifferentCMPs will containa different combinationof tracesin a regularpattem this may show itself in the final

51 -

3-Ddataprocessinq

Fig. 2.35 Example of mutesapplied during stacking.Here we seetwo mutes applied to a set of gathers.The red mute is applied to stop eventsbeing distortedby the shetch causedby the normal moveout processat large offsetsand shallow times. The green inner trace mute has been applied to removeremnant multiples on the near tracesthat have not been removedby other demultiple processessuch as Radon.

stackedsection.For instance,CMP I may containt r a c e s1 , 5 , 9 . . . ,C M P 2 w i l l h a v e o f f s e t s2 , 6 , 1 0... , C M P 3 o f f s e t s3 , 7 , 1 1. . . , C M P 4 4 , 8 , 1 2 . . . ,C M P 5 w i l l b e b a c k to 1,5,9.. . . This four-tracepattemmay be visible in the final sectionparticularlyin the near surfacewhere the fold of data (numberof tracesstackedtogether)is low owing to the stretchmute.Any regularlyrepeatingpattemwill showitself asa strong componentin the inline spatialFourier transformand can be removedby a notch filter. Step25 is a 3-D noiseremovalprocess.The applicationofnoise suppression is much moresuccessfulwith 3-D datathan2-D owing to theextradimension,datavolumeand spatialconsistency for thealgorithmto work with. Usuallyrelativelysmalloperatorsare used,working on maybefive tracesin both directions.This allowsthe filtersto adaptto relativelysuddenchangesin reflectordip while retainingsufficientdatato distinguish

52 -

dataacquisition andprocessing 3-0seismic signal from random noise. Step 26 is very similar to step 10, the difference being that we a.renow perfonning the operationpost stack in order to make the crosslinespacing the CMP spacingalong a line is the sameas the inline spacing.As alreadydiscussed, generally much finer than the spacingbetweenlines owing to cost and the possibility of cablesbecoming entangled.Again methodsthat attempt to interpolate frequencies beyond the standardaliasing criterion are employed since no benefit would be gained otherwise. Both steps25 and 26 together with step27 arc used to condition the data for the final migration processof step 28. Within step27 we are attempting to remove any suddenamplitude variations belweentracessince theseare not consistentwith the behaviour of wavesand would causeartefactsduring the migration process. Step28 is the final, high-fidelity migration processusing the spatialvarying velocity field from step20. It is usual not to use the raw velocitiesas originally picked, but ratherto usevelocitiesthat havebeensmoothedboth spatiallyandin time. Oftentrials with various percentagesof the picked velocity field are performed and the results are analysedfor evidenceof over (too high velocities)or under (too low velocities) migration.A compositevelocity field is usedfor the final migrationbasedon these tests.The resultof 3-D migrationis oftenvery impressivecomparedto 2-D migration. The reasonis simple:3-D migrationplacesthe energycorrecdyin a three-dimensional sense,whereas2-D migrationcan only do this in the direction the data were shot; any complexity perpendicularto the plane of the 2-D sectionremainsunresolved. Figure2.36 showsan exampleof seismicdatabeforeandaftermigration.The effectof the migration is to narrow the width of the large anticlinal feature and steepenthe dips of the sides.

Fig. 2.35 Exanple of seismic data before and after migration. Data on the left are unmigrated, on the right they are migrated. Notice how the migration process has reduced the apparent size of the structure. Note also that whereas the picture on the left has reflections that crcss each othel, those on the right are geologically consistent.

53 -

3-Ddaiaprocessing

2.5,6 Post-migration processing After migrationthe dataare more correctlypositionedalthoughthey may still not be perfectlylocatedowing to efiors in the velocityfield, useof time migrationratherthan depth migration, anisotropy,etc. However,at this stagethe data shouldlook like a cross-section throughthe earthwith reflectionscorrespondingto changesin acoustic propertiesand unwantedeventssuchas multiplesand noiseremoved.The final step is to convertthe datato zero-phase which centresthe peakof the seismicwaveleton the impedancecontrast.Even when an attempthas beenmadeto keepthe datazero_ phasethroughoutthe processingsequence!it is likely that thereis still considerable phaseuncertaintydue to the effectsof attenuation.Zero-phasingis best pedormed with well data.An operatoris determinedby least-squares matchingthe seismicdata to the reflectivitygeneratedfrom the well data (White (19g0) and Walden& Whire (1984)).Sincethereis likely to be someeffor in the positioningof the seismicdata after the migration,this matchingis usuallydonefor a numberof tracessurrounding the well location.The fit betweenthe seismictraceandthe well log generated synthetic ls comparedand the trace locationthat gives the best match is chosenas the trace for estimationof the wavel'et.The processis explainedin more detail in chapfer3 (section3.1.1).Figure2.37 showsthe waveletextractionobtainedfrom one of the commerciallyavailableinterpretationpackages.Oncea wavelethasbeenextractedan operatorcan be designedto convertit, and hencethe data,to zero-phase.Note that suchan operatoris really only applicableto the time-gateusedfor the extraction.It is extremelydiflicult to conlidentlyzero-phase dataover largetime windows. The final threestepsin theprocessingsequence outlinedin fig. 2.21areall concerned with fine tuning the dataprior to loadingto the interyretationworkstationfor detailed analysis.Step30 appliesan equalisationon a traceby tracebasisto ensurethe spectral contentofeachtraceis broadlysimilar,while step3l appliestime-varyingbandpass fil_ tersto reducethe higherfrequencieswith time to eliminatethosethataremosuynolse due to their attenuationon passingthlough the rocks. Finally, step 32 appliestime_ varying trace scalingto ensurea balanced-lookingsectionwith time. One approach is to apply AutomaticGain Control (AGC). This appliesa time-varyinggain to each traceindividually,with the gain calculatedso as to keepthe averageabsoluteampli tudeconstantwithin a window that slidesdown the trace.A shorttime-windowAGC (say200 ms) is highly undesirableif any usewill be madeof amplitudeinformation subsequently, becauseit tendsto destroylateralamplitudechangesthatmay be impor_ tant.A long gateAGC (say 1000ms or more)is usuallyacceptable, however,because the gain is not muchinfluencedby the amplitudeof any singlereflectorAll thesefinal stepsareimportant,because theycanaid or inhibit theinteryretabilityofthe datasetthat is deliveredto the interyretationworkstation.Removetoo many high-frequencydata and subtledetail may be missing;leavetoo much noiseand automaticbatchtracking of horizonsmay be compronrised.

54 -

3-Dseismic dataacquisition andprocessing

al

:,: :,:

: i : r : :r: : ! :r:

:,:

-f f 1t 0 2A 30

L r r t r t t t t l t l i , 0 1 02 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0 1 1 0 1 2 0 1 3 0

Fig.2.37 Example of \ravelet estimationby matching data to wells. The waveletmay be estimated by deriving a matching operatorbetweenthe seismicdata and the well log predictedreflectivity. To do this a small area of tmces cenhed on the well location is examined and the match bet\aeen each trace and the well log is calculated.The best match traceis determinedand the waveletis estimated togetherwith the syntheticthat is predictedusing the wavelet and Iog derivedreflectivity. Several QC pammetersare also given, the most impo ant being the signal to noise estimate.The bottom left-hand display showsthe signal to noise plot for tracesaround a well; high signal to noise is shown in blue. In this casewe have a very good match with a signal to noise ratio of over 4 that is obtainedfrom a tlace very close to the actual well location. The display on the right shows the syntheticin white overlain on the recordedseismicdata.

With luck, hard work and appropriatetestingandparameterselectionthe result of the completeprocessingsequenceshouldbe a sectionthat resemblesa cross-section of the reflectivity throughthe earth with optimisedseismicresolutionand signal to noiseratio. In practiceit often takesseveralpassesthroughthe processingsequence belbrethis goalis satisfactorilyachieved.Eachpassthroughthedatauseslessonsleamt from the previousprocessingroute so that the next iterationmay spendmore time on velocity analysisor multiple attenuationor use a differentmigrationstrategyThere is also a different requirementfrom the data as we move from explorationthrough

55 r-l

Belerences appraisalanddevelopment. Onceexplorationtargetshavebeenidentifiedtheremay be a wish for a detailedgeophysicalanalysisof theprospectivelevels.Techniques suchas AVO analysis,seismicinversionandamplitudeanalysis(seelaterchapters) all impose specialrequirements on thequalityofthe seismicdatathatmaynot be metby theinitial processing for explorat ion screening. In someareaswith severeproblems(e.g. sub_basalt or beneathcomplexsalt bod_ ies), where the primary signalis stronglyattenuatedor scattered, it may be difficult or impossibleto generateinter?retablesectionswith currenttechnology. The desire for better seismicsectionsremainsstrong and there is active research in most as_ pects of seismicdata acquisitionand processing.However, the industry shoulclbe proud of the improvementsmade in seismictechnologyover the past two decades. Dataquality hasimprovedtremendouslyandwith it our ability to seehydrocarbonsat depth.

E

References Bancroft, J. C. (1997). A practic.ll Lln.lerst,tn.lingof pre_ and poststackMilrations, Volume l (Po,jtstack).Society ofExploration Geophysicists, CourseNotes SeriesNo. ? B4ncroft, J. C. (f998). A pructical lLndefitanding oJ pre_ anl poststack Mryratlons, Votume2 (Pr"ertdc,t).Society of Exploration Geophysicists,Course Notes SeriesNo. 9. Berkhout, A. J. (1982).SeismicMigratktn. Imaging ofAcoustic Eneryy by Waw l,ield Extrapolation. A. TheoreticalAspects(2nd edn). Elsevier Scientific publishins. Amsterdam_ Berkhout, A. J. ( f984). SeismicMigration. lmaging o.f Acoustic Energvby WaveField Extrapolation. B. Pructical Aspects.Elsevier Scientific publishing. Amsterdam. D e r e g o u s kSi .. M . I l q 8 6 r .W h a li , D M O : F i r r rB n a L . 4 t J | / l y t . p p j. 24. Evans,B. J. (1997).A Handbook For SeismicData AcquisitionIn Explofarlor?. SocietyofExploration Geophysicists,GeophysicalMonograph SeriesNo. 7. Garotta,R. ( 1999).Srear Wares.ln,mAcquisition to Interpret(tion Society ofExploration ceophysi_ cists, DistinguishedInstructor Series,No. 3. Hale'D' (1991) Dip Moveout processing.society ofExploration Geophysicists, course Notes series No. 4. Hatton, L.. Worthington,M. H. and Makin, J. (19g6). SeismicData processing: Theoryan l practice. Blackwell Scientific. London. Iack,l. (1997). nne-l,apse Seismicin Resenoir Management.Society ofExploratlon ceophysicists, 1998 DistinguishedInstructor Short Course. McQuillin, R., Bacon, M. & Barclay,W (19g4).An Introduction to Seismic Inrerpretatio, (2nd edn). Graham & Trotman Ltd. London. Robinson,E. A. and Treitel, S. (1980). Geoplzys ical SignalA allsis.ptentice Hall, EnglewoodCliffs, New Jersey. Sheriff.R. E. & celdafi,L.p. (1995\. Erytorution Seismology(2nd edn).Cambridge Universirypress, Cambridge,UK. Stolt, R. H. and Benson,A. K. (1986). Sersnic Migrution: Theory and pftrctice. ceophysicalpress. Stone,D. J. (1l994).D esigningSeismicSut-velsin Tv)oand ThreeDin erriors. SocletyolExDloration Geophysicists.GeophysicalReferenceSeriesNo. 5.

56 -

andprocossing +D seismicdataacquisilion Ththam,R. H. & McCormacKM. D. (1991).MulticomponentSeismologyin PetroleumExploration, Societyof ExplorationGeophysicists,InvestigatioNin GeophysicsSeries,Vol. 6. Walden,A, T. & White, R. E. (1984).On erors offlt andaccuracyitr matchitrgsy ietic seismograms andseismictaces. GeophtsicalProspecrtng,32, 8'7l-9I. White, R. E. (1980). Paftial coherencematching of synthetic seismogramswith seismic haces. GeophysicaI Pmspecti ng- 28, 333-58. Yilmaz, O. (1987).SeisnicDota Prccessing.Societyof ExplomtionGeophysicists,Iavestigationsin GeophysicsSeries,Vol. 1.

I

3 Structural interpretation

This chapteris ntainly aboutthe most lundamental interpretatron activity:making maps of horizons.Historically,it was the need for bettcr maps or complex structuraltraps that was a key drivcr in the early adoptionof 3_D seismic.Usually,however.it is not enough just to map the top of the reservoir. To understandhow structures were tormed and when, it is usually nccessaryto rnap a rangc of marker horizonsaboveand helow the targct.AIso, dcpth conversionwill. in most cases.requirc thc mapping of several horizonsabovc the targetlevcl. This chapterbeginsby consideringhow stratigraphic horizonscncounteredin wels can be tied to particularreflectionson a seismicsurvey; this rs an rssuetor all seismic Interprctation'and is often easieron a 3-D survey bccauseof the more nearrycorrecl p.sitioning ofsubsurfacefcatures.Having decidedwhat to map, the interpreteris taced with the dauntingproblemol working with a huge numberol.sersmrctraces.all of which should ideally be taken into account.Happily, thc power of computer workstations has increasedfasterthan the quantity of traces waiting to be Interpreted;the chapter continueswilh an explanationof how thc tracking ol.rellectin-qhorizons through a 3-D volurrrecan be partly automated.This is not a matteronly of mapping the horizon itsclf: in an area of even moderatestructurtl complexity, ir is rhe mapping of fault systcmsthat will consurremuch of thc interpretation ctlbrt. Semi_automated methods can help here roo. For compieteness, the chapterconcludcswrth a bricf discussionof how lo converta reflection_timehorizon map into a depthntap;the rssuesrnvolvedare lhc sameas thosewith 2_D scismic.but the grealer densityof tlata in 3_D surveysmay m a k eI h e t a s l ,e r \ i e r i n p r a c t i ( e . -

3,1

Wellties Onc ol the first srepsin interpretinga seismic datasctis to establishthe relationship betweenseismicrcflectionsand stratigraphy.For structuralmapping, it may be suth cientto cstablishapproximaterelationships(e.g.,reflection X is nearBasecrelaceous,). althoughfor ntorcdetailedwork on altributes,as describedin chapter-5.rt ls usuallynec essaryto be more preciseand establishexactlyhow (tbr example)the top or a rcservolr

58 -

Structural interDletation is expressedon the seismic section.Although some information can be obtainedby relating reflectors to outcrop geology, by tar the best source of stratigraphic infbrma tion. whereverit is available,is well control.Often wells will have sonic(i e. formation velocity) and formation density logs, at leastover the intervalsof commercilllinterest: lrom these it is possibleto constructa syntheticseisnrogtumshowing the expected seismicresponsefbr comparisonwith the real seismicdata. In addition, some wells will have Veitical Seisn it Profil e 1V!P) data, obtained by shooting a surfacc seismic sourceinto a downhole geophone,which has the potentialto give a more precisetie betweenwell and seismicdata. In this sectionwe shall discussthe use of both these types of data.

seismogram 3.1.1 Thesynthetic The basic idea is very simple. To a first approximationwe can calculatethc expected seismic responseof the rock sequenceencounteredin the well by rreating it us rl one-dimensionalproblem. That is, we calculatethe effect as though the intedacesin the subsudace are horizontal and the ray-paths are vertical, so that rays are normally incident on the intefiaces.This is usually a reasonablefirst approximation'but means that we areignoringthe way that seismicresponsevaricswith angleof incidence.which will be discussedin chapter5. ln somecases,we may haveto usethe short-offsettraces, rathcrthanthe full stack,for comparisonwith the calculatedwell response.to make sure that the approximationis valid. lf wc think of the subsurlaceasa numberof laycrs.each with its own acoustic impedance A, then the reflection coeflicient at the /?th interface for P-waves at zero-offset is given by the folmula R , ,- ( A , * r

4,,)/(A,a + 4,,).

where A, and A,,+l are the acousticimpedanceaboveand below the interface:acoustic impedanceis the productofdensity and seismicP-wave(sonic)velocity.The derivation of this lbrmula can be tbund in Sheriff & Geldart (1995), tbr example:it is a simple particularcasc.valid tbr normal incidence.of the Zoeppritz relations'which describe how the reflectioncoefticientvariesas a function of incidenceangle. Both density and sonic valuesare routincly logged as a function of depth in boreradiationfiom adownhole holes.Densityis inferrcdtiom the intensityof back-scattered gamma-ray source: the amount of back-scatter is proportional to the electron density in the rock which is in tum proportionalto thc bulk density.Sonic velocity is determined from thc travel-timeof a pulseof high-fiequency(e.g.20 kHz) soundbetweena downhole source and downhole receivers;the sound travels as a retiacted arrival in the boreholewall. Becauseof the methodsemployed,the valuesobtainedfor velocity and density are those in the fbrmation closc to the boreholewall, i.e. within a few tens of centimetres of the borehole. This may or may not be representativeof the fbrmation as a whole. lt is possiblethat the well has drilled an anomalousfeature.e.g a calcarcous

59 -

Wellties

concretion.at a pafticular level, which woultl give log rerdings quite different liom those typical of the fbrmation in general.Also, in permeablelbnnations the zone ad_ jacent to the boreholewall will be invarledby the drilling fluiti, and this may alter the log responsefiom the original formation values.[n some cases.seisnricvelocity may vary significantlywith liequency.a phenomenonusuallycalledr/lspcrrrrzr.ln rocks of low porosityand pemeability, the sonic log measurer:l at 20 kHz may not be a reliable guide to velocitiesat seismicfrequenciesof 20 Hz. Multiplying the sonic and density logs togetherwill give us an acousticimpedance log. A typical display from a well synthericsoftwarepackageis shown in tig. 3.1. On the lcfi-hand side there is a scalemarked in borh depth anclreflection(two-way) time (TWT): how we tind the correspondence betweendepth and TWT is explainedbelow. The valuesat the top of the scalc show that a reflectiontime of zero corresponds to

l0ll111l r5 20 25 J0 35

5 - 1 0 - 1 0 - 500

1DSYN 3-5-70-900

Fig.3.1 Synlhelicscisrnogrant gencrdlion. Tracksshowtime/depthscale.sonicand dcnsrlylog. calculaledimpedance. and synrhcticlbr rwo \\,avclels (5 l0_,:1050 Hz and 3 5 70 90 Hr. both 7erc pnase).

60 -

StructuralinterDretation a depth ol l I I ft: this is becausein this offshorewell, the datum for seismictimes is sea-levelbut log depthsare measuredrelativeto a point on thc drilling rig (the kelly bushingin this case)which is I I I tt abovesea-level.Thereis scopefbr confusionhere. particularlyin the caseof deviatedwells where the depth scalemay representdistancc as measurcdalong hole or altemativelymay havebeencorrcctedto true verticaldepth. Using erroneousdatum valuesis a common causeofproblems with well synthetics.and they nced careful checking.The next threepanelsshow the sonicand density logs.and the acousticimpedancelog fbrmed by multiplying them together.ln routinc logging practicc.thc samplingintervalfbr sonic and densityvalueswill be half a tbot in depth: thus the acousticirnpedancclog will typically sho$,hne detail where thin interbeds of dill-crcntlithologies are present.From this log it is possibleto get an inmediale impressionof what causesthe principal scismic rcflcctions.They are ofien causedby sucldenmarkeclchangesin impedanceduc to nraior changesin lithology. though they may also result fiom small low-amplitudeimpedancechangesif they arc cyclic at the right frcqucncy10resonatewith the seismicpulse (Anstcy & O'Doherty.2002). 'lhc ncxt stcp is 10 convert thc acoustic impedancelog, calculatedfiom log data recordedas a function ol depth.into a log as a function of (two-way) travcl time. This is easy if we know the time {lepth (I Z) relation lbr thc wcll. which can in principlc bc obtainedby simply integratingthe sonic log. though in practicetwo problens arise. One of them is that effors (fi)r example.minor miscalibrationof the sonict(nl) tend to accumulatewhen the log is integratedovcr nranythousandsof t'eet.Another problemis that the sonickrg is harcllyeverrun in the shallowcstpart ol the hole. For thesereasons. it is usualto calibratethe I-Z curve by meansof somedircct observationsof travel-lime from a surfacesourceto ardownholcgcophone((h(k .\hols),e.g. at intcrvalsof 5(X)li along the entire borehole:the intcgratcdsonic is then adjustedto match thcsccontrol points. It is also possibleto adiust thc sonic log itself. and theu to usc this adjusted log to createthe impedancevaluesand the syntheticscismogram.This is often a bad idea: the piecewiscadiustmentof the sonic log tends to creatca step charge at each checkshot.and thus a spuriousrellection.Obviously.it is possiblcto clcate a smooth adiustnentto the sonic log that avoidsthis problem.but a sinlplerapproachis to adjust only thc I Z curve and not the sonic log. A reflectivitycurve is then calculatedfiorr the impedanceusing the tirrrnulagiven above.This reflectivity sequenccis convolvecl with the wavelet thought to bc prcscnt in the seismic data to gcncratethe synthetic seismogram.the expectedrcspoltscof the loggedinterval.shownon the right hand side of tig. 3. | . There may bc considcrableuncertaintyabout the correct wirveletto use. The amplitude spectrumof thc waveletcan be estimatedfrom thc seismicdata. bul in ordcr to dcscribethe waveletcompletelythe phasespectrunris also needed.This describes the relativeshitis of the wavclbmrs at cach fiequency (fig. 3.2). Two particulartypes o1'waveletarc otien used:the minimum-phaseand zcro-phasewavelet.A mininrumphasewavclct is a causalwavelet.i.e. it has no anrplitudebefbre a delinitc start time.

61 -

Wellties

WaveletPhase

ZeroPhasewaveletand selecledfrequency components 3 612

40

72Hz Phasedescribesthe relativetimingof the vanouslrequencycomponentsthat make up the seismicwavelet

Amplitude

60 FrequencyHz

20

60

FrequencyHz

Fig.3.2 'Ihc phascspecrruDr rclarcd lo rhc(imingrelarionship ol rhevrrioust.rsquenrr conlponenls ln thewavelet. \\Jlichin thisexamDle is TeroDhlse. Of ail the waveletsthat havethis propeny and havc any particularamplitudespectrum. the minimum-phasewavelctis thc one that has the most conccntrationor energyclosc to the start timc. This does not nccessarilymcan that the leading loop is thc lrrgesl: somemlnlmum-phascwavcletshavetheir greatestamplitudein the secondloop. Thesc wavelctsare importantbccausethe actualsourcesignaturefrom explosilesol lir -uuns is close to ninimum-phase.Howevcr,as noted in chaptcr2, thc waveletis oftcn con verted to zero-phasedurinc proccssing.This producesa wavelet like that shown in fig.3.2. synrmetricalabout the zero-time antl so with energv at negattvcttmes and not causal.This waveletis prcferredlbr interpretationbccausethe strongcentralpeak at trme zero is easyto relateto thc reflectorconcerned.[n practice.the processors at_ tempt to convertthe waveletto zero-phascis rarcly perlcct.and mixed_phlrse wavelets are common. The choicc of wavelet can ntake a considerablcdiff'erenceto the appcaranceol. synthetictraces(Neidell& PoggiagliolIni.Ig77).Thereareseveralpossibleapproachcs. A simple methodis to make syntheticswith theorcticalwaveletsof ditfbrentfrequencl, contenl.both zero- and minimum-phasc.and look fbr I good visual match to the real seismic.This may be adequarcto idcntifv rhc loopscorrespondingro key stratigraphic m a r k e r ss. o i s a u s e f u sl t a t o m a k i n g s t r u c t u r a l m a p s . A b t a l l y d i f f e r c n t a p p r o a c h i s t o start from a measuredsourcesignaturc.and to calculatethe way that it is modilietl by

62 -

interDletation Struclural passingthrough the earth: however.the etTectof the earth lilter on waveletphasemay bc hard to estimate.For detailedwork involving measurementof loop amplitudes(scc chapter5), it may be bestto estimatca wavelettiom the data.This is easilydoneby using an algorithn)thatwill calculatea wavclctthatgivesthe bestlit betweenthe syntheticand coefficicnt the real data.The goodnessof fil can bc evaluatedfrom the cross-correlation significance ofhigh correlation betweenthe synthcticand the real seismic.However.the dependson the lengthof the wavclctcomparedwith the analysiswindow. lf the wavelet is made long cnough.then a perfectmatch can always be obtained.but such wavelets are ofien implausible(e.g.having high-amplitudeoscillations)becausethey are trying to fit the noise in the real seisnricas well as signal.A good match would be one with a high correlationover a long gatc using a short wavelet.An empirical approachis to use a gate of 500 ms or more with a wavelet consistingol only 2-3 loops. but a nrore rigorousapproachhas been put fbrward by White ( 1980).who uscsa statistical methodto constrainwaveletlength.With this typc of approach.not only is the wavelet shape derived, but also the timing rclirtive to time zero. This is important becausc some minimum-phasewaveletscan look approxinlatelysymntctrical.and so roughly like a zero-phasewavelet,but the main loop is dclayed tiom zero time. If we want to measureamplitudeson seisnricdata.it is importantto measurethe right loop, e.g. the one correspondinglo lhe top ol l rcservoir. In fig. 3.1, a zero-phasewavelct of fiequency content5-50 Hz has been used.The displayrepaysclosc scrutiny.Firstly.it is a good ideato checkthe polarityofthe display. 'nomal' and As mentionedin chaptcr l. thereis otien confusionaboutwhat the words 'reverse' mean as applied to the polarity ol zero-phaseseismicdata. By boking at an isolatedintertacewith a sharpimpedancechangc.the polarity of thc syntheticcan be seendirectly.Thus in tig. 3.1 there is a sharpincrcasein impedanceat about 3020 ms which corrcspondsto a white trough (cleflectionto the leti) in the synthetic.Next, it is easy to see where high-amplitudereflectionsare to be cxpected:thesewill be the casicsteventsto pick on the 3-D datasetto fbrm a basisfor structuralmapping.A sharp impedancechangcgivesthe best response:the ramp-like impedancechangeat around 3200 ms causesonly a moderateevcnt.even though the total changein impcdanceis large.The fine structurewithin the impedancelog is not rcpresentedat all in thc s1'nthetic. Detailedcomparisonfbr a particulartarget interualwill show what hope there is of mapping,tbr cxample,the top and baseol a reservoirfrom the seismicdata.It is often useful to calculatesyntheticsfbr a rangeof high tiequency cutofl.s,to seewhat bandwidth would be neededto reveal significantdetail: if a moclerateincreasein resolution would improve the informationsignificantly.it is worth consideringadditional seismicprocessing.inversion(chapter6). or re shootingthc survey. The comparison of thc synthetic seismogram with traces extracted fiom the 3-D datlrsetaround the well location is shown in hg. 3.3. In this case,there is a good visu.rl match fbr the main events. but not tbr the weak events in the central part of the display.In sucha case.it may be usefulto calculatesyntheticseismogramsthat includc

63 -

Wellties

.-r-..&!

Fig,3.3 Sl nrherrrseisntngrant \ufrerpL'scd on seismicsectionat the well locatior.

& -

interDretation Struclural the etTectsof multiples.Idcally, scismic processingshould have attenuatedmultiplcs to low levels relative to the primaries.If a better fit to the real seismic is observed when multiples are added into the well synthetic,then the interpreterneedsto be on his guard:even if multiple energyis not obvious on a sectionview and doesnot cause on problemsfor structuralinterpretation,it may still corrupt amplitudemeasurements target refleclors.An advantageof having the densedata coverageof the 3-D survcy is that it is possibleto observehow correlationvaries as l function of trace location. ldeally one would hope to seea bull s eye patternof high correlationcentredaboutthc well location.thoughthe real world is not always so simple. Sometimesit is easy to obtain a good match betweenwell syntheticand seismic dataset:sometimesit is very difficult. Thcre aremany possiblereasonsfbr a mismatch.

Seismicsurveyproblems

. incorrectzero-phasing(or other defectiveprocessing) . rnultiples . incorrect spatial location duc to shot/receiver mispositioning or (more commonly) incorrect

Syntheticseisnrogram delects

migration velocities . defective logs . hydrocarbon effects . inadequatespatialsanpling

ln addition.thcrcarethe effectsof amplitudevariationwith incidenceangle:in principlc this meansthat syntheticseismogransshould be calculatedfor a rangeof anglesand addedtogetherto simulatethe slackedtracc of the real data (seechapter5). It is important to understandthe causesof an observcdpoor well to seismictic. If the problem lies with the seismicdataset.similar problcmswill ofien be presentin all wells within thc survcyarea.Incorrectspatiallocation.if the resultof mispositioningof sourcesand receivers,will probably mean that all ties will be improvedby applying a constantlateralshifi to the seismictracelocations.Ifthe problemis inconectmigration of lhe surlacc scismic. then the tie points will be shified updip or downdip by an amountrelatedto the steepness of the local dip: thus flat-lying overburdenmight reed little lateral shifi. whereasdeeperreflectorswith dips of 30 might need lateral shifts of hundredsof metres. Problems duc to the synthetic seismogram will be difl'erent from one well to another. Variousproblcmscanarisewith the wireline bg data.They may needsubstantialediting to removeintervalsof incorrectreadings.Usually.the petrophysicistis interestedmainly in obtaininggood quality logs over the rcservoirinterval,which he or she will use to evaluatcrcscrvoirquality and hydrocarbonsaturation:other intervals,especiallyshales, will not have receivedintensescrutiny for log quality at thc time of acquisition.It is quite common to find data gaps and noisy intervals,owin€:tbr exampleto cycle skips on thc sonic log where the automatictrnvcl-timemeasurementsystemtriggerson thc wrong part of the signalpulse.Defectivclog intervalscan bc editedby removingnoise

65 -

Wellties

spikesand by replacingerroneousdata,either by plausibleconstantvaruesor pernaps by valuesderived fiom anotherwcll. Over a particularintcNal. it may be possibleto calculatevalucsfor one lo-efiom othcr logs.e.g.densityfrom gantma ray antisonic:the requiredrelationshipscan bc establishedlrom a nearby well where all the logs are of good quality.whatever replacementmethodis used,careis needecl to avoiclintroducing artificial suddenjumps in the sonic or density curvesal top and bottom of the cdited interval.as they wr)uldgeneratespuriousreflcctionsin the syntheticseismogram.l_ogs over hydrocarbon-bearingrescrvoirsshould also be treatedwith great suspicion:il there is signiticantinvasionof drilling lluid inro the fbrmation. either or borh of the density and sonic logs may be recordingvaluesin a zone closeto thc boreholcwhere the hydrocarbonshave been partly swept away by the clrilling fluid. wnosepropernes are therefbrenot representative of the virgin fbrmation. It is possibleto cstinare these eflectsusingmethodsdescribedin chaptcr5. but the resultsareoficn unreliablebecause of uncertaintyaboutthe cxtent of the invasionand thus the magnilutleof the efl'ect.n log rcsponse.Finally. as notcd abovc.the logs samplethe subsurfaceonly within a f.ew centlmetresaround the borchole.whereassurfhccseismicdata responoro properUes that are averagedlaterallyover at leastseveraltensof metres.Thus. lbr example.a loc^ll calcareousconcretion.which happenetlto bc drilleclthrough by a wcll, could show a marked efftct on logs but have no seisnticexpressionbecauseof its limited lateral cxlent. Even when thc wireline log data are correctand reprcsentativc of the fbrmation.the approachdescribedabovemay nor resulrin a correctsyntheticcalculation.Implicitly. the method assumesthat we can trcat the propagationof the seismicwave rnrougn a I D earth model using ray theory.This is correct if thc seismic wavelengthis short comparedwith the layer thickness.lf thc wavelcngthis greaterthan aboutten trmesthe layerthickness(aswill ccrtainlybe the casefbr surfhcescismicresponsemodelledfiont closelysamplcdwirelinedata).thenit is more appropriateto approximatethe subsurtace layering as an eflectivemcdium (Marion et u!., 1994).The ef'lectivemedium velocity Vs is calculatedas follows. Supposewe have a stack of thin layers in each of which there are log measurements of P vclocity l./n.shearvekrcity l,', and dcnsity p. [n each layer we determinethe shearand bulk modulus 1p and K) from the equations

and

x - c(v,i i*) Over an intcrval (typically a quartcrof the seismicwavelength).we then calculalcthe arithmeticdensityaverageand the hamonic averageof lr and K. Thcseaveragevalues arc then usedto calculatea mean valueof l,n and V,, using the samerclationsbclween the clasticmoduli and velocitiesas befbre.This effectivemccliumcalculationis known

66 -

Structural inlenretation as Backusaveraging(Backus, 1962).The differencefrom the simple approachoutlined previouslydependson the velocity and density variationbetweenthe different layers. F'orshale-sandalternation.the ditlerenceis usually small: for shale-dolomitc,it can be quite large,perhapsreachingas much as 207r. It often happens that pq)r well ties cannot be definitely traced to a specilic cause in the seismicor the log data.As pointedout by Ziolkowski ct a1.( 1998),thcre is a fundamental problem:seismictracesare not in realitythe simple convolutionof a physically meaningfulwaveletwith the reflectioncoefhcientseries.ashasbeenassumedabove.In reality,the seismicdatawe collect are the responseof a layeredearthto a point source, includingintemalmultiplesand free-surfaceeffccts;conventionalprocessingcombiner theserecordsto producestackedsectionsand attemptsto removemultiples,diffractions, P S conversions, andso on. It is not suryrisingif we thenfind thatsometimesthe wavelet that gives the best fit bctwcen the stackedseismictrace and thc wcll syntheticvaries lrom one well to another.Somctimesit may not be possibleto havemuch conlidencein the well tie. even after carefulediting of the log data and considerationof all the other possiblecomplications.At this point. it is usefulto have a diilerent line of evidenceto help understandthe causeof discrepancies betweenthe well syntheticand the surface seismic,and this is where the VSP can help.

3.1.2 TheVSP Essentially,the techniqueis to record a surfacesource using downhole geophones; a general account has been given by. fbr example,Oristaglio (1985). The simplest geometryis the caseofa verticalwell with a seismicsourceat the wellhead(fig.3.,1(a). where the sourccis shown slightly separatedfrom the wellheadfor clarity).Recordings would be made ol the sourceby a geophoneat a seriesof downhole locations,e.g. at intervalsof50 ti overa verticaldistanceof4(X)0li to give recordsat 80levels.The record fbr any particular level will contain both upgoing and downgoing waves;the lbrmer are reflections from horizons below the geophone. and the latter are the direct arrivals, as sketchedin fig.3.,l(a). Both upgoing and downgoing arrivalswill be contaminated by multiples due to reflectionsboth above and bclow the geophone.but the upgoing waves that immediately follow the downgoing direct arrival have a useful property: they are liee fiom multiple energy,becauseany multiple bouncesin the ray-pathwould delay it and makc it arrive signilicantly later than the direct ray. AIso. the waveform of the direct arrival at a geophonetells us what the waveletis at that particulardepth. (The seismic wavelct changesslowly with depth owing to a progressiveloss of the high frequencies.)This measuredwavelet must also be the wavelet presentin the reflections that immediately fbllow the direct arrival. becausethe travel path through thc carth is nearly identical. Therefbre. if we can separatethe upgoing and downgoing arrivals,then the downgoingwaveheldtells us the waveletand allows us to calculatea tilter operator to convert it to zero-phase.Applying the same operator to the upgoing

67 -

Wellties

(a)

(b)

Fig, 3.4 (a) Schenatic geomelry of ray-pathsfrom surlacesourceto boreholegcophones.for rhc simplc caseof a venical well and small otTseI,shown exaggcratedherefor clarityi (b) schcmalic graph of ardval timc againstSeophonedepth.

ot E

Struclural interoretation wavelieldproduceszero-phase reflectiondata,with a more accuratecontrolofthe phase than can be achievedlbr surfaceseismicdata. Fo unately.it is quitc straightforwardto achievethe separationbetween up- and downwavcs.As shown schemirticallyin lig. 3.4(b). thcrc is a diffbrencein changeof arrival time with depth betweenthc up- and downgoing waves.When the geophoneis locatedat the depth of a particularreflector.then the direct travel-timeis the samc as the reflected time to the event. As the geophone is moved up away from the reflector, the direct travel-timedecreases and the reflectiontime increases.For the simple caseof a verticalwell with sourccat the wellhead.it is obviousthat the decreasein travel-time fbr the dircct arrival will be the same as the increascin travel-timefbr the reflection. ln practice.it is fairly straighttbrwardto measurcthc travel-timesof the direct arrivals, thoughit is sometimeshard to identify the exacttime at which thc tracebeginsto deflect owing to the presenceofnoise. lftraces are staticallyshiftedby subtractingthcsctimes. then the direct anivals will line up horizontallyacrossa tracc display: if the tracesare shittedby addingthc lirst anival times(doublingthe slopcol-thefirst arrivaltravel time curve)then the upwardtravcllingevcntswill be horizontal(fbr horizontalreflectors.or nearly so if they aredipping). Filteringthe first type ofdisplay to enhancelaterallycon tinuouseventswill resultin an estimateof the downgoingwavelleld,which can then be subtractedfiom the datato leaveonly the upgoingwavefield.Applying the secondtype of trace shifi to theseupwaveswill then give us a display on which seismicreflectors arc near-horizontaland can be enhancedby median filtering. which emphasiscsncarhorizontal lineups in the dataset.A filter operatorcan also be applied to convert the wavelet (as measuredin the downwaves)to zero-phase.It is then possibleto fbrm a t on-idor stut'ktraceby stackingtogelherthe pans ol the upwavedatasetirnnrediately tbllowing thc dircct arrival.This tracc shouldthen bc zcro-phaseand fiee of multiples, and thus ideally suitedfor comparisonwith well syntheticand surtaceseismic.Since the VSP averagesseismicresponseover a distanceof a few tens of metresaroundthe borehole,the problemsof formation invasionand very small-scalelithologicalchanges are not present.lt is therelbreusuallyhelplul; the main problem is the ratherlow signal to noiseratio. Figure 3.5 showsan exampledisplay of deconvolvedupwaves,aficr the traces have been shified to make the reflected events line up horizontally. The correspondingenhanced(median-liltered)upwavedisplay is shown in hg. 3.6. The median filter hasbroughtout someconsistenteventsthatarescarcelyvisible in fig. 3.5; however, reliability ol theseeventswould needcarclul thought in a practicalapplication. Another advantageof the VSP is the ability t0 give good resultsin deviatedwells. where syntheticseismogramsare ofien unreliable.perhapsbecauseanisotropymakes the sonic log readings(which measurevelocity along the borehole)diffcr lrom the vertical seismicvelocity in the fbrmation; thus the calculatedimpedancecontrastsare not thoseseenby a nearly vcftically travellingray.A usefulVSP techniqueis the walkabove geometry,where the surfncesourceis placed vetically above thc gcophoneat a seriesof levels in the deviatedhole. In this way. an image is producedof the zone

69 -

Wellties

Geophonedepth, ft

qt

E t

o ! 4.O

Fig. 3.5 Trace display lbr zero otl.setVSP, alier subrractionol down{oing wdves.tracc atlgnntent ol upgorngwares.and deconvolution.

70 -

interpretation Slructural Geophonedepth, ft

3.5

dt

E (!

' o '

F

4.O

5.U

Fig.3.6 Samedatr as hg. 3.5 atlcr applicatiotlof medianlilter to cnhancehorizontalalignmenls

71 -

Workstationinteroretation below the well bore, which can be compareddirectly with an arbitrary line selected trom the surfaceseismicdataset.Note, however,that this techniquedocs not work fbr a horizontalborehole;if there is no vcrtical separationbetweenlevels,it is not casy to scparatc the upgoing and downgoing wavelields. as there is no difl-erencebctween them in the time-depth pbt. Although the VSP is olien the best way to establishthe tie berweenthe surlace seismicand the well infbnnation, it does have one disadvantage comparcdto thc well synthetic.This is that it givesno real insightinto how reflectionsare causedby changes in vclocity and dcnsity valuesfiom one fbrmation to another.This knowledgeis vital, as we shall see in chapter5. if we want to understandthe likely causesof changesin seismicreflectionamplitudeor characterfiom one part of the surveyto another.which may allow us to predict Iithology,reservoirtluality or porefill.

-

3.2

Workstationinterpretation Having identitiedsomehorizonsthat arc significantfor understandingthe geology and prospcctivityof an area.the next task is to map thcm acrossthe survey.Before carrying out any detailedwork. it is useful to inspcct the volumc as a whole ro ger a general inprcssion of structuraland stratigraphicfeaturesof intercst.This can be done by using the volume visualisationlcchniquesdiscussedin chapter7. Increasingly,lnlcrpretatron is being carrietlout in this environment(Kidd. 1999).However,much dctailedwork is still carriedout by picking the two-way reflectiontime to varioushorizonson some or all of the traccsof the survey.Thesepicked horizonsare fundamentarto the arrribute measurel]tentwork discusscdin chapter5. In the earliestdays of 3-D survey,horizonswere picked using methodscarriedover tronr interpretationof grids of 2-D lines.Thesewere prescntedto the rnrerpreteras a stackof paperprints;he would mark up the horizonsof intereston a line througha well locationand then folJowthem along the line to intersectionswith other lines.where thc picks would be transferredto the crossinglines.By working round a loop ofintersecting lines,it would be possibleto get back to the startingpoint, where it could be checked that the interpretationwas consistentarounclthe loop. Interpretationwoultl proccedby fbllowing the horizonsround a seriesof such loops until they had beenpicked on all the lines (see McQuillin et al., 1984.fbr a more detailedexplanationwith examples). The analogousmethod fbr 3-D data was b make papcr displaysof all the inlines and all the crosslines,with the idea of interpretingthem in a similar way. However,it is easy to see that the effbrt needed for a full manual interpretation is prohibitive. Suppose we have a quite small rectangularsurvey consistingof 500 lines. cach with 1000 traces. Then there would be 500 inline sectionsand 1000 crosslinesections.The number of intersectionsto be checkedwould be 500 000. Assuming the data were easy to pick. it might be possibleto vcrify thc intersectionsat a rate of, say.2000 per working clay,

72 -

SlrucluralinlelDretation so picking the entire surveywould lake a year ol solid mechanicaleffort. with no time allowed for thinking about the meaningof the data.The practicetherefbregrew up of interpretingonly a propo ion of the data.say cvcry I Othinl ine and crossline.Often this was sutlicienttbr structuralmapping:thc benefitof having datacorrectlypositionedin spaceafier 3 D migration was sccurcd.However,the llnc detail that was prescntin the closely spaceddata was lost. The use ol-workstationsfbr 3 D interpretationwas theretbrewclcomed by interpreters. They of'tered several advantages: (i) the ability to vicw sectionsthrough the data in any orientation, (ii) automatic book keeping of rnanuallypicked horizons; picks made on one linc would automaticallybe transterredto other lines or to map views. (iii) seni-automatedhorizon picking. (iv) calculationof pick attributesthat can bc uscd to extractadditionalinfirnnation. (v) ability to seethe data volume in 3 D. not just as sections. To achieveall this requircsthe use of fairly powerful workstations.and Appentlix I dcscribessomeof the hardwareand datamanagementrequiremcnts.Eachof the topics on the abovelist will now be addressedin turn.

3.2.1 Displaycapabilities The 3-D seisrr.ric traccscan be thoughtofas a volumc of seismicamplitudevalues.In the cxamplediscussedin the previousscction.therewould be 500 000 tracesarrangcdon a rectangulargrid in map view,5(X)inlinesby 1000crosslines.The two-way time on thc veftical cxis might rangefrom 0 to 4000 ms, sampledat 4 ms, giving us 1(X)0samplcs on eachtrace.As shown in lig. 3.7, it is possibleto view a rangeof differentslicesfrom 'cube' this of data. There are the obvious inlines and crosslincs,but also llorirontal (time slices slices).antl vertical sectionsat any orientationthroughthe volumc. These 'arbitrary lincs' do not have to be straightithey might. fbr example.be constructcdlo .join up a number of wcll locations Thereare two possiblemodcsofpresentingseismicscctionson the scrccn:as wiggle 'variable intensity' displays.ln either casc,there are limitations imposed tracesor as by the screenrcsolution.This might, for instance.be 1024 by 1024 pixels. (A 2r.tc1 is the smallestindependcntlycontrollableelement of a screendisplay.Softwarc can specify the brightnessand colour of each pixel on the screenbut cannot achicvc any higher (-r, r') resolutionthan the pixel.) To get reasonabledynamic rangeon a wigglc trace display, the trace would need to extcnd over. say, l0 columns of pixcls. If the tracesdo not overlap.this would imply that only 100or so tracescould be displayedat any one time. Tracescan be allowedto overlapin orderto view morc of them,but evcn so a wiggle tracedisplay will be limited to only a fcw hundredtraces.This is suitable for detailed work (e.g. wcll ties or study of lateral changesin loop character),but makesit diflicult to obtainan overvicw of the data.lt is therefbreolien betterto work in

73 -r

Workstation intemretation

Data cube

Inlines

Crosslines

Timeslices

Arbitrary lines Fig.3.7 Dillerentwaysof slicinga datacube. variablcintensitymode.In this case,eachtraceis assignedone column of pixels.within which each pixel correspondsto a time sample;the pixels are colour-codedto show amplitudeof the particularsample.Choice of colour-codingis under the interpreter's control, but popularchoicesare grey-scale(medium-greytbr zero-amplitude.shading to black for large positive and white for large negative amplitudes) and redT/trlucor rcd/trlack dual polarity (white for zero-amplitude. shading to recl for high negative amplitudeand blue/blackfor high positiveamplitude).It is useful to experimenrwith differentcolour bars:grey-scaleofien bringsout subtleevents(e.g.reflectionsobliquc lt:tthe bedding.which may be noise or may carry genuineinlbrmation about intemal

78 -

Stluctural interoretation map in the correctlocation; fault intersectionsare marked in a similar way. The tault intersectionsare joincd up between lines to establishthe lault pattern,and horizon contours are constructed from the posted values. All of theseprocedureshaveequivalentsin 3-D workstationinterpretation.Horizon picks are markcd by digitising with a pointing device (usually a mouse)on a scrccn display of a section.This can bc done using displaysin any orientation,as explained in the previousparagraph.Once a horizon pick has been made on any paniculartrace of the 3 D datavolume, it is availablefbr displayon any other sectionthat includesthat trace.For cxample.it is ofien best to stan by picking along a scriesof compositelincs that link the availablewells together.It may then be bestto intelprcta f-ewkey dip lines acrossthe structure:when theselines are displayed,the picks alreadymadeon the wcll traversescan be displayedautomatically,to ensureconsistentpicks.A coarsegrid ol dip and strikelinesmight thenbe interpreted,which can laterbe inlllled asmuch asrequired to definethe featuresof interest.At each stage.the picks alreadymadc on intersecting linescan be displayedon the currcnt scction.Justas with paperdata,this is a powcrlul check on consistency.and it is quite usual for picks to be deletedand reworkedas the interpretationproceeds.(To make selectivedeletionpossible,it is impofiant to retain infblmation on exactly what co-ordinateswere used to constructparticularcomposite sectionsthrough the data, if you are using anything more complicatedthan simplc inlines and crosslines;all softwareallows you to storethis infbrmation,but to do so is not alwaysthe delirult.) with horizonsaredigitisedfiom the screcndisplay Faultplanesandtheir intersections in r similar way. It is much easierto work with taults on lines crossingthem approximately at right mgles than on lines crossingthem obliquely. wherc the fault plane crossesthe beddingat a shallow angle.This is of coursewell known to the interpreter of 2-D data,wherc a line that crossesa significantlault obliquely will have a smeared imageof the subsurfacewith substantialamountsof rellcctionenergycoming from t-ea turesout of thc plane of section.ln the cascof 3-D data,the rcflcctedenergyhas becn repositionedso that lhe verticalsectiondoesnot containout of-planereflections,if the migrationhasbeencarriedout correctly.Even so. it is difficult to recognisefault planes that do not make a high angle with the bcdding,when projectedon the line ol section. This is becausefaults are almost invariablyrecognisedfrom reflectorterminations,as reflectionsfrom the lault plane itself are rarc: the lineup ol terminationsis much easicr to seeon a dip sectionthan a strike section(figs. 3.12 and 3.13). On the other hand, lines parallelto a fault may be very uselul to investigatchow one fault intersectswith another,which may be crucial to the integrily of a tault-boundedstructuralclosure. While all this picking is going on, the sottwarecancontinuouslyupdatea map display showing the horizon pick. by colouring in the traces on a basemap according to the TWT to the reflector. Fault intersectionscan be marked by special symbols on this map display.This nakcs it easy for the interpreterto keep track of what lines have been interpretedand of the emerging structuralmap. Usually interpretationworkstations

80 -

Stluctural interpretation

Fig.3.14 Schematicrnapshowingclillicultie\of cslablishingfuull pallcrntiom a grid ol 2 D lincs up by FrulI svmholssho\\'laull cutsrisihlc on r scfi(jsol-ersl-wesllines:theycould bc.ioinecl ol thcm).leadingto quilc dillerenl eitherlhc solid lincsof lhe dashedlire\ (or somccorrtrination Iirull maps.

fault Reverse

fault Normal Sectionview

Horizonoverlap

Horizonculout

Map view

Fig.3.15 Schenlulic\ectionand map viewsol normal nd reverscfitults.

u

-

interDletation Struclural takc placeuniler closeusercontrol.so thal difficult areascanbe picked with continuous control of the quality of the result. The quality of an autotrackcdpick may bc improvedby pre-conditioningthe seismic filtcring' has datausing imageprocessingtechniques.One approach,structure-oriented been introducedby Hoecker & Fehmers(2002). The idea is to stabilisereflectionsin the presenceof noise. without srnoothingover faults. The processconsistsof thrce steps:analysisof the raw datato determinethe local orientationof the reflcctors'edgc detectionto find reflectiontemrinations,and smoothing of the data along the local orientationwithout filtering acrossthe edgesdetectedin the previousstep As well as rcrnovingnoise.it is also possibleto usesucha filter to rcmovegenuinebut small scale featuresof the tlata.suchas very small faultsor small-scalestratigraphicfeatures.This opens up the possibility of an iterrLtivcapproachto automatedinterprctation.In the tirst pass.all the llne detail is removcd.permitting a rapid autotrackingof the rrain horizons.anclperhapsautomaticfault tracking.This lirst rcsult can then be finc-tuned by repeatingthe processon a tlatasetwith less aggrcssivesmoothing.stabilisingthe autotrackingby using the resull of the first pass as a seed grid. The processcan be repeatedthrough severalc1'clesol iteration.until either the data havc been interpreted to the requiredlevelofdetall or the limit sct by the noisein the datasethasbcenreached.

3.2.4 Attributes A major advantageof workstationinterprctationis that measurcmentsof the seismic bop bcing pickedare simpleto calculateand storc.The most obviousis loop amplitude' but bop width. averageamplitudein a window below or abovethe horizon.and many others are commonly available.The ability to see these measurelrentsin mup vieu tiom tlenselysampleddata is a key step in getting infbnnation fiom the seismicdata about poretill (presenceand type of hydrocarbons)and reservoirqurlity tportr'it1. net/gross,etc.).The way in which this can be done is the topic ofchapter 5 Amplitude e g. channelsystems For mapscan alsobe the key to rccognising stratigraphicf'eatures' accuratework, it may be importantto know how the softwarecalculatesloop amplitudc. Some early autotrilcketswould simply use lhe largeslscismic amplitude seenat any of the (usually'1 ms) sampleswithin the krop: since there is unlikelv to be a sample Modern exactlydt the loop maximum. amplitudeswere systematicallyundercstimated. autotrackersfit a curve to the amplitudesat the sampleswithin the loop in order to eslimatethe true maximunl value. A diflcrent type of attribute is particularly important to structuralmrpping. lt is possibleto analyseboth the picked horizonsand the scismictrace datl thcmselvesto look fbr lateraldiscontinuitics:we shallconsiderin this sectionthosethat are relatedto recognisinglaults.but they can also be usedto aid geologicalinterpretationin general' The simplest approachinvolvcs calculation fbr a picked horizon of the local dip v a l u ea n d i t s a z i m u t h( D a l l e yc t a / . . l 9 l t 9 ) . F i g u r e 3 . l 9 ( a )i s a s k e t c h m a p o f a l a u l t e d

86 -

Structural interpretation fault segments.Detailedmappingofthe relay rampsmay be importantin understanding whether a tault will provide a seal to a prospectivestructure.Study of the way that displacementof a horizon variesalong a fault is needcdto assesssealingcapacity(sec chapter.l). and is also a uscful chcck on the correctnessof the interpretation:along a single fault. the displacementshould increascsmoothly fiom zero at the ends of the fault traceto a maximum in the ccntre. It is also possibleto calculatcthc local curvatureof the picked horizon (Roberts, 2001). Faultsappearas bipolar high-curvatureanomalies,with high valucsof opposite sign producedwhere the fault plane intersectsthe horizon On its up- and downthrown sides.lt may alsobe possibleto relatecurvatureto fractureintensity.e.g.over saltswells. The main problernin usingcurvaturc( which is a problemfor dip andazimuthcalculation also) is to dccide thc length scaleover which the attributeis calculatcd.distinguishing larger-scale structuralf'eatures trom small-scalefeaturesthatmight be sedimentological or might be noise (e.9. apparentreflectorrugosity resultingliom the autotrackedpick wanderingup and down within a broad loop of low signalto noiseratio). Another approachis to make an illumination display of the picked horizon. Thc softwarecalculateshow the surlacewould look if seenfrom above when illuminated by a light source from a particulardirection. Usually the source ('sun') direction is set to be near the horilontal ('low in the sky') so that subtle highs and lows in the surllce are picked out by thc contrastbetweenthe bright surfacesfacing the sun and the dark shadowswhere surlaccslacc away fiom it. The etl'ectis to enphasisethose topographiclineationsthat in map view trend at right angles to the sun direction. Therefbre,completeinterpretationrequiresa numberol illuminationdisplayswith the sun in ditlerent directions.Better still is to have interactivccontrol of the sun position. with real time updating of the screendisplay as lhc sun is moved around the map. In this way the usercan choosesun positionsto emphasiseparticularf'eaturesof interest. Thesemethodsdependon havingan accuratelypickedhorizonon a densegrid. Ifthe horizonof interestis not casyto autotrack.the interpreterwill haveto do a good deal of editingandre-pickingbelbrehe is ableto usethesetools.A difterentapproachis to try to recognisefaultsas discontinuitiesin thc seismictracecube.without necessarilyhaving any horizonspicked at all. The basic idea is to calculate.over a limited time-gate,a measureof the similarity of a seismictraceto its neighbours(Bahorich& Farmer,1995; an implementationis the subjectof an Amoco patent).The calculatedsimilarity value is postedin the seismicclatacube at the centreof the tracewindow usedto calculateit, and the processis repeatedfbr every tracein the seismicvolume and for every possible window start time; the result is thereforea completedata cube of similarity values. Faultsare revealedas planesof low similarityl they are bestseenin a horizontalsection through the cube.The advantageover a simple time slice through the reflectivitydata is that the faults will bc visible whatevertheir orientation:on reflectivitysliccs,faults are ofien difficult to seewherethey run parallclto the strikeofthe beddingso that there

87 -

Workslation inlerpretation

il '

:

l"\

..,

.

;*.*r -'

\;1

Fig.3.20 Coherency mapshowing linealions duero faullsisolidlineis section of fig.3.21. are no obviousdisplacements ofbedding lineaments.An exampleis shownin llg. 3.20. trom the Lower Tertiary of the UK North Sea. A subtlued colour_scale,such as thc grey-scaleused hcre, or a sepia scale,is otien best for picking out subtle lincations. There are a number of lineationsduc to small taults. The bold black line marks thc locationof the seismicsectionshown in fig. i.21. The small lault in the centreof the line is easily fbllowed acrossthe map view in fig. 3.20. Another exampleis shown in lig. 3.22, whcre a salt diapir piercesthe horizon in the ccntre of rhe map and radial faults can be seen.cspeciallyin the south-eastemquadrant.Since the tault planesare surlaceso1-lowcoherencewhich are distinct fiom the higher coherencevalueswithin the 3-D volume generally.they can be visualisedin 3-D from any perspcctiveusing the techniquesdiscussedin chapter7. They can also be aukrtrackedusing techniques similar to thoseusedfbr horiz.ns, though this has not yet becomestandardpracticein the santeway as horizon autotracking.The coherencecubc methodolo-qycan also bc usedto revealstratigraphicdetail in the 3-D volume. such as channcl/fansystems(scc chapter4). Care is ncededin interpretingall theseattributeswlrerethcre is signilicantcoherent noisepresentin the scismicdata.Intcrferenceof noiseevents(e.g.multiples)with real reflectors gives rise to discontinuities in reflectors that can be misinterpreted as faults

o,t

-

interpretation Structulal or using an algorithm that did not take proper account of overburdencomplexities' then there may be systematic lateral shitis of interpreted features(e.9. faults) fiom their true locations.These lateral shifts will be discussedin section 3.3.4: for the momcnt we assume that they are not a problem. Then all that is needed to conven the reflectors mapped in two-way time into depth is a knowledge of the seismic vclocity in the Sometimes,a quite detailedvelocity model will alreadyhave been built subsurf'ace. by the seismic processors for migration purposes, but as we shall see these are not necessarilythe hestvelocitiesto usc for depth conversion To developour ideas,it is useful to look at a real seismicsection.A display of the entiresectionfiom suface to targetlevelon a workstationis usuallytoo poor in quality to usetbr detailedpicking. owing to the limited venical resolution.bul is worth making to plan the strategyfor depth conversion(lig 3.23). In this example from the UK Central North Sea,the objective is at orjust above the orange horizon' which is the top of the Ekofisk Formation. Two horizons have been picked in the overburden. They are levelsat which there is substantialdiscontinuityin the curve of sonic vclocity against depthat a nearbywell (fig. 3.24).The yellow horizon is encountcredat a level of about 1300 ms. Above this level, ignoring noise.velocity is ncarly constant;at the horizon there is a slight decreasein velocity. and then velocity increasesfairly steadilywith depth to the top ofthe Sele Formation(greenmarkcr in 1ig 3 23)' at which point there is an increasein velocity and a more complicatedvelocity-depthtrend which is only roughly approximated by a linear increasewith depth This suggeststhat a three-layer model would be suitable.with constantvelocity in the top layer and velocity increasing with depth in the other two. A similar analysisneedsto be carriedout at an early stage of every interpretation, as it detennines which overburden layers need to be picked to carry out the depth conversion. Sometimes, as here, only a few surf'acesare needed to give a reasonableapproximation. At other times quite a large number of surfaces may be needed,if thereare largevelocityjumps at a numbcr of horizons:this might be the case if carbonatcs or evaporites are intercalated within a sand/shalesequence.To decidewhethera given layer is worth including in the model. it is easyto calculatethe error introducedat the well by treatingit as part ol an adjacentlayer.To assesswhether the error is important is harder, and depends on the detailed geometry of the structure being mapped;critical areasto look at will usuallybe the culminationlrnd the possible spillpoino t l ' a n y\ l r u ( l u r i l lc l o \ u r c . If the velocity within a layer is constant,it is obvious how to convertthe two-way time thicknessinto a thicknessin depth. tf there is a velocity gradientwith depth' we proceedas fbllows. Supposewe have a layer which extendstrom depth:r to depth :1. at which the two-way times arc 1t and t] respectively.and that the vclocity at any depth within the layer is given by

In such a tbrmula. u is ofien referred lo as an instQntaneoustelocit)-: it describes the actualseis|nicvelocity at a pafliculardepth(or travel time) and may be contrastedwith

94 -

Structural interorelation the fault. If the picked time horizon is smoothedout in this way. then vertical-strctch depthconversionwill of courseput a kink into the depthhorizon underthe fault plane. If this is a problem, then it may be an adequateapproach to remove the distorted Part of the horizon under the lault plane on the depth map, and replace it by extrapolating thc horizon dip as seenoutsidethe fault shadow.Smoothingor filtering the velocity field is is accuratc anotherpossibleway to removethe distortion.If neitherof theseapproaches enough (as might be thc case lor listric faults above a hydrocarbonaccumulationor prospect. where the area affected by the tault shadow would be large). then pre-stack depth migration is needed. A different approach to dealing with the eflect of lateral variation in overburden velocity has beendescribedby Armstrong ct ol. (2001). Look again at lig. 3.23; there the largestof which is at the right-handend ol' are obviouschannelsin the near-surface. thc sectionand extendsdown to ir TWT of about400 ms. There is somecvidenceon the sectionthat the infill of this channelhasa low seismicvelocity: reflectorsimmediately undemeathit are pusheddown. Similarly. reflectorsare pulled up by presumedhighvelocity inlill to the channel-likef'eaturesvisible at a TWT of 800-900 ms Armstrong er a1.proceedby measuringthe push-downor pull-up on a reflectorimmediatelybclow the anomalousf'eature.and usethis informationto simulatethe effect on seismiclines acquiretiacrossit. As discussedin more detailin section3.3.3,differentsource-rcceiver otTsetsare affected by the anomaly to differing extents. depending on the sum of the delays experiencedat the two ends of the path. Simulaiion of CMP stackingof the modelled data then predictsthe time distortion in the stackeddata. whose eflect can therefore be subtractedout of the horizon time map.

3.3.2 Useof wellvelocityinformation Over the intervalwhere sonic logs and checkshotdata have been acquired,wells will have high-quality velocity information.lf there is only one well, velocity valuescan be read fiom the log, or averagevelocitiescalculatedfor particular layers using the known two way time (TWT) and depth at the top and bottom of the layer As we saw above, usually there are small static shifts between well syntheticseismogramsand real trace datal their cffect can be removed from the depth map by using TWTs fbr the top and base of the layer taken from the picked seismic trace data rather than fiom the well sonic/checkshotinfbrmation,when calculatingaveragelayer velocities.This is a reasonableapproach il the static shifts do not vary much from one well to another: ifthey do, it would be betterto identity and removethe causeof the shifis.If a (un * t: ) model is being used.then valuesfbr r'0 and A can be found by litting a line throughthe sonic log valuesplotted as a function of dcpth. If there are severalwells, then the simplest possibleapproachis to averagethe valuesfor eachintervalacrossall the wells. Another approachis to makc naps ol thc velocitiesin cachinterval.or perhapsof ui andi values' Interpolationbetweenthe wells may be difficult, however.They rnay be few in number,

95 -

Depthconversion

and will usuallyhavebeendrilled nearthe crestsofanticlines:the interpreter,however, may need a reasonablyaccuratedepth conversionof the intcrveningsynclines,either to map the spill points of thc anticlinalstructuresor to assessmaturity of hydrocarbon source rocks. At first sight the (u0 + li:) model gives the required help, but this is not always the case.The simplest casc is wherc the {' factor reflectsthc eff'ectsof compaction.as unconsolidatedsedimentsbecome more deeply buried over time. In sucha casc.the value found fbr qy may vary little liom one well to another This might be the caseil allthe rocks are currentlyat thcir greatestdepthofburial. Howevcr.when rocks that once were decply buried are later found near the surtaceafter a period of uplift and erosion,the velocitiesusually remain close to what they werc at the timc of deepestburial. If there has been variation of the uplifi liom one well to another. then u0 values will also vary. Sinple interyolationof u11betwecn the wells is valid only if uplift valuescan be similarly interpolated.Another possiblecomplicationis that the I factor may represcnta changein velocity due to lithologicaleffects,fbr cxample a consistentcoarseningor fining upwardsof a clastic sequence;( may then bc quite similar trom one well to another,but give no clue aboutthe effect of depthon velocity. Ratherthan derivingI fion the sonic log, it may therefbrebe pref'erablc. where several wells arc availablc.to determinea compactiontrend by plotting the avcragevelocity in each fbnnation againstmidpoint depth.The gradientof this line (rr) is not the same thing as the I valuc fi)r instantaneous velocity unlessthe intervalis rhin (time thickness much less than l/t). For thick intervals,with the notation of the pre!ious section, wc would lind: (--r--r)

/-t+--r\

1,.-,,w:u',+n(

:

/

so that ::( I

r(tz

tt)/4\:

: t + l \ t + K z )/ 2 ) ( t 2- t t ) 1 2

and

. 1 : 11 ( 2 u e+ ( : r . X r r- / r ) I - r(t1 rr) However the velocity maps are calculated.it is usually a requirementthat the iinal depth map should match the formation tops in the wclls. This will always be the case if the velocity derivationmethodologyhonoursthe well data exactly.as can easily be done if mapsare being madeofthe velocity in eachlayer.Howevel if someor all layers aredepth-converted usingconstantparametervalues(e.g.the averagefor the velocity in the layer,acrossall the wells), then there will be discrepancies betweenthe depth map and the true well depths. If they are large, the method of depth conversion used needs to be revisited;if they are small. the usual practiceis to grid up the mistie valuesand apply them as a correctionacrossthe whole map. The gridding algorirhm needsto be chosenso that it will not produceunreasonable valucsoutsidethe areaof well control.

-

inter0letation Structural as might be the cascif the gradientsbetweenwells are extrapolatcdbeyondthem lt is not usually possiblcto understandexxctly what controlsthcseresidualdiscrcpanctes' so the scopefor intelligcnt contouringtaking accountof geologicaltrendsis lirnited: this is why the residualsshould be quitc small belore this stepis taken. Quite oiien. well data give us intbrmation at so few points that it would be uselul to bring some extra informationinto play to interpolatebetweenthem. To get velocity infbrmationacrossthe whole of the seismicsurvcy,it is naturalto turn to the velocity liclds derivedduring thc processingof the scismicdata.

velocities 3.3.3 Useof seismic During thc courseof scismicproccssing.a dcnselysampledvclocity licld is gencrated in order to stack and to miSratethe data. It is ofien assumedthat stackingvelocities are root-meansquarc(rms) averagevelocitiesfronl the surlacedowtl to the rcflector conccmed.It was shown by Taner& Koehler ( 1969)that lbr a retlectorat the baseof n unilbrm horizontallayers.the reflectiotltime 7. correspondingto a soLrrce-receiver distance-r is given by f,'

7;

-

| C r . t ' ' ( r . r nF

.

I,;.,

wherc the coefticicntsC arc functionsol'the thicknessesand velocitiesof thc ii layers and !',.,,,,is the rms vclocity along the zero-offsettrajcctorydcfinetlb,

I r,irr Tl Stacking vclocities l'.1 are usually calculatcdby mcthods that assumca hypcrbolic time {listancc relationship,i.e. that tit a relationshipof the firrnr

r , , : r ,+i Ir',r to the travel-timcversusofl.setdata.The stackingvelocitiesarc therelbreonly an approximation to the rms averagc velocity fiom the surface to the rcflector concerncd However,there arc other reasonswhy the velocitiesthat give the strongcststack lrm plitudesand the most sharplytbcussedrcflectionsare only looselyrelatedto the actual seismicvelocitiesin thc real earth.Al Chalabi (1994) has provideda useful summary ol them,The most seriouscllects on stackingvelocitiesare duc to statics.structureand anisotropY. Staticseffccts arise when the survey is shot ovcr a near-surfacevelocity anomaly (Al-Chalabi. 1979).The geomctryis shownin lig. 3.26for the caseof a tntdel involving a step acrosswhich a near-surfacedelay is gencrated.When thc CMP locationis at A. only thc outer traccs of the CMP gather experiencethe delay: the bcst-fit hyperbola

97 -r

DeDthconversion

Fig.3.26 Skelchof cftectol ne r surtitce staricrypedelayson stacking velocity( fierAl Chalabi. t979). acrossall the data will be stecperthan the hyperbolaefbllowcd by the inner or outer traceafiivals scparately.correspondingto a lower velocity.When the CMp location is at B. most ol the rays seea doubledelay.once on the shotside and once on the receiver side:the outertracesexperienceonly a singlcdelay.The best-lithyperbolaacrossall the tracesis thenflatter(higher-velocity)thanwould be found lbr the unperturbedtraces.As shownat the bottomofthe tigure,the resultis an approximatclyantisymmetricvariation of stackingvelocity.with a wavelengthequal to rhe sprexdlength. The cffect can be large.with oscillationsup to l5% of the averagcvelocityto the reflcctor.tf the velocity anomlly is at dcplh, ratherthan at the surlace.both the width of the stackingvelocity responseand its amplitude are rcduced.[n gencral.lateral variationsin the stacking velocityfield with a wavelengthlessthtn the sprcadlength(maximumsourceto receivcr distance)are nol to be trustcd: if a sutficicntlylarge number of stackingvelocity data are available.the spuriousefltcts can be largely removedby smoothingthe data. The effect of slructurearisesin severalways. Dip can have a signilicantetfbct on velocityestimates.Often dip-independentvelocitieswill be availableif DMO hasbeen applied during proccssing:if not. a correctioncan be made (Levin. | 971). However, to usethe seismicvclocitiesfbr dcpth conversionwe usuallywant to calculateinterval velocitiesin eachindividual layer.This is doneby meansof the Dix tbrmula(Dix, 195-5 ):

t',,,- (

(v;rh v:r,,) l.Tr

7,,)

) "

98 -

inle0retation Structural where l';n1is the interval velocity in a layer with rnrs velocitics 1",,and V6 to its top and base, and l, and I/, are the coresponding normal incidence times. The fbrmula does not take accountof ray bending efIects.and gives incorrectresultsfor dipping interfaces(withdipslargerthanaboutT'forcasesmodelledbyAl-Chalabi).Reflector c u r \ A l u r cl l s o b i u s e .v e l o c i t i e si .t n d i s n o t e a \ y l o ( o r r e c l l o r . Anisotropy arisesas an issuebecausethe velocitiesdetennincdtiom seismicprocessingare,broadly speaking.horizontalvelocitiesthroughthe ground:for dcpth conversion.we of course need to have vertical velocities.Many rocks' however.exhibit anisotropy.with horizontalvelocitieslarger than vcrtical oncs. This may be intrinsic to the rock or an effect of small-scalcinterbeddingof fastcr and slower lithologies lt is possibleto measureanisotropy(e.g. from long offset VSP data: Annstrong cl '1l ' are availableand it has to be inferredfiom the 1995).but often no direct measurcments comparisonof scismicand well data. In general,migration velocitiesare closer to true velocitiesin the ground than are stackingvelocities,becauseol the removalof structureefl'ectsand much of the statics effect; anisotropyremains a seriousfactor' howcver. and it is not possibleto use a migration velocity directly tbr accuratcconversiontiom time to depth.However,if we are looking tbr a way to interpolatevclocitiesbetweenwells. migration velocitiescan be useful. We can compare the actual well velocitiesin a particularfbrmation with the migration velociticsmeasuredat the well locations'and so estimatea correctlon factor; if anisotropydocs not vary much laterally within the fbrmation' then it should be possibleto use a singlecorrectionfactor for it acrosslhe entirc area. All thesecommentsapply as much to 2-D as to 3-D seismicdata.The main beneht of 3-D data is that the velocity field will have been densely sampledin space.It is likely to be of better quality than velocitiesderived from isolated2-D lines because of the opportunityto spot mistakesby plotting out sectionsthrough the velocity cube (e.g. horizontal slices),and becauseit is dcnsely sampledit can easily be smoothed to remove the etl'ectsof statics.However. it is still only really suitableas a way ot interpolatingbetweenwells.

shifts 3.3.4 Lateral Sometimesthe accuratelateral positioning of events in the seismic datasetis very important.An examplemight be the caseofplanning a well to drill into a lault block on the upthrown sidc of a major fault. Therc may be a need to drill as close to the tault as possible,perhapsto ensuremaximum drainageof a rcservoircompartmcnt,but it will be crucial to drill on the correctsideofthe fault and not accidcntallyon the downthrown side.which might be outsidethe hydrocarbonaccumulationaltogether. Accuratelateralpositioningdependsmainly on the quality of the scismiemigration process:for modern surveys.any uncertaintyin the surtacepositions of shots and receiversis negligible by comparison.It is imponant to realise that migration may

oo E

DeDthconversion

Fig.3.27 Zero-ollserre|ecrior ray-parhtbr a dippinglayer. shilt reflectorsby large distanceslaterally.Consider the very simple case shown in fig. 3.27. where a singledipping reflectoris overlainby constant-velocityoverburden. An identiliablepoint S on the rcflector (perhapsa small fault) will be imaged on the stack section below the point O where the zero-ofl.setray intersectsthe surface; OS is pcrpendicularto the reflector.The migration processhas b shift the image laterallyto thc true location below P, over a distance .1. Then d:

O S s i nd

and if the two way travel time fbr the zero-offiet ray is t. then . u t d : - srn9. Of course,we do not directly observethe true dip in depth. but ratherthe dip on the (unmigrated)time section.the rate of changeof / with ./, which is given by 2sind L]

Call thisquantiryq. Then t

nt

all)

2

2

r,1t

The error 6 in rl due to an error du in l is then given by

d:2rdu

qt 4

or d

2dr.'

i-; For example,supposethe overburdenvelocity is 3000 m/s. For an eventat 2 s two_way t i m e w i t h a d i p o f l - 5 " .t h e l a t e r a sl h i f t w o u l d b e 3 0 0 0 s i n1 5 . . o r 7 7 6 m . A 2 E e - r o ri n

100 -

Structuralinterpretation This is estimatingthe velocity I would result in a 47' error in this shifi' or about31 m s' the of4 a time 25" at not likely to causeproblems.However.for a reflectordipping at shifiwou|dbe2535manda4T"errorinthiswouldbel()lm.enr'ughlocuus!'serious conccm when planning a well. In such a case' careful investigationof the migration velocity is ncededto eslablishits likely accuracy' variation' Complicationsarise as soon as the overburdenshows signilicantvelocity in panicularly if there are rapid velocity changeslaterally The etlect of ray bending assume algorithms the overburdenthen has to be taken into account Time migration shlpe of hyperbolicmoveout'and accounttbr lateralvelocity variationby varying the does structure velocity if the the hyPerbolawith map location:this is satisfactoryonly of not vary laterally lcross a CMP gather.The tcchnically correct aPproachin the case rapid lateralvariationis pre-stackdepth migration This is' houc'ver'time-consumlnP andexpensive,becauseoftheeffortnecdedtobuildacorrect3-Dvclocityrr.rodcl;if time the model is incorrect.the migratcd image may be worse than that from a simple tor migration. Various methods have therefbrebeen suggcstedto upply corrections idea basic The rays image lateralshift to time-migrateddata.Onc methodis thc useof (Hubral, 1977) of the imagc ray is that it startsverlically dounwards from a point al the surface. anclpropagatesthrough the subsurfacerefiacting at all velocity boundarics unti|thetfaveltimeisusedup.Thecorrectedhorizonsarepositioneduttheendpoinl case of the image rays. Howcver, image rays will corrcct for ray-bendingonly in the wherc the target horizon has zero time-dip; in othcr casesthe lateral displacement image derivedby this mcthod will be inconect.becausethe overburdensampledby thc ray is different from that seen by the actual physical rays reflected from the dipping surface(Calvert,2002).

-

References in Al-Chalabi.M. (1979).Velocity detcrminaliontiom seismicrcllectiondata ln: I)(r(lol'n(tlls Publishers' Science Applied pp l-6S (ed. A A Fitch). / Gtopht;ittt[ I:.tplorLnionntethrtt]s BitrkinS. (1994).Seismicvelocities- a critique Fi:1rBrzol. 12.589 96' Elgr' 2l '1'1 5 l ' Anstey.N. A. & O'DoherIy. R. F. (2002) Cyc les.luyersand rcflecliotrs Th( LeQ(litt! dala /:i-tl Armstrong,P N. Chmela.W. & [-caney.w S (1995) AVO calibrationusin!!borchole B r c a l . 1 3 .l l 9 2 U . velocilyanonaly ettects Amrstrong.T.. McAleer.J & Connolly.P (2001)'Rcmovalol ovcrburden 99 49. 79 fbr depth conversion.Ceophrsicul Prorpettin,(. layering J Caoph:-t Backus.G. E. (1962).l-ong-waveelaslicanisotropyproducedb) horiTontal Rts..67.112'740 ftalurcs:the Bahorich.M. & Farmer.S- (1995) l D seismicdiscontinuilyfbr faultsand struligraphic 8' cube.Th. LcadingEdg(.14. 1053 coherence positioning llr'\/ Calvert. R. (2002). Image rays and the old mylh about corecting iime migraled Break.2ll.115 16.

101 -

References

Dallcv.R. M.. Cievers. E. C. A.. Slanlpfii.C. M.. Davies.D. J..c{statdi. C. N.. Ruijrenberg. p A. & Vcnneer.(;. J. O_( I9lt9). Dip and azimuthdisplaystbr lD seisnricinterprelalion_ f.it.rtBt.?uI. 7.86 95. Drvie\. R. K.. Crawtbrcl.M.. Dula.W. F..Cole.M. J. & Dorn.c. A. (1997).Oulcropinterprerrrion ol. scrsmlcscalenorntallaultsin southcmOrceon:descriptionof slrucluralstvlesand evaluation ol \ub\uriaceinterprer{rion merhods.l he Ltudittg Etl.qt.l6. I 115,13. Di\. C. tl. (1955).Seismicvebcitiesliom surlaccneasurenents.a;..r)lr.!r .!.20. 611g6. Dorn.G. A. ( l99lJ).Modern I [] seismicinrerprelarion. Ih. L.illing Ldg(.11. 1262 11. Hcsthanmer.J. & Fossen.II. { 1997).The inlluenccof seismicnoisein strucluralinlerpretalion of sersmfcitttributenlaps.FirstBrt&.1S.209 19. Iloeckcr.C. & Fchners.G. (2(X)2). fast stnlcturalinlerprerarion wilh struclureorientedhltcrjng 7fu, Lr n I i n t Etlgc. 21. 23813. Hubral.I'. (1977). Time migntion someray theoretical aspects. 6ro2fir.srrr/ p rosptrting.25.13g '15. Kidd. (j. D. (1999).Fundamenhlsof.l,D seismicvolunre!isualizarion..1fu Lcur)ing Ettgt.lg. 102 t2. Le! in. F K. ( I97l ). Apparenrvclocilytionr dippinsintertacerellecrions. 6roprr..\i(.r.36. 5 l0-16_ Nlarion.D.. Mukerji.T. & Mavko.C. ( 199,1). Scalcellecrson velocirydjspet]\ion: tiom ray ro ellectivc nrediumrheoriesin slrarifiedmedia.Gcr.r2ln.sics. 59. l6ll 19. McQuillin. R.. Bacon.M. & Bilrcht,.W (1984) An tntkx!tution nt Sci.,;nic Intarptl. /r)r. (;raham & Trotntan.London. Neidell.N. S. & Po-ugiagliolni. E. ( 1977). Srraligraphic _ {eophy\ical modellinganclinrcrprerlrion principlesand techniques. ln,l.1r,rr. J_arairrtpht AltpliLutiont. llttltt.urb.n f,.tpl.ruriott (cd. C. E. Payton).AAPC memoir26, pp. .llt9 4l6. Orisuslio. M. L. ( l9lt5).A guidero currenrusesof vcnici.ilseisnicprofiles.6roplr.r rrs.50.2.1739. Roberls.A. ( 200I ). Curvalure (fibutesandtheirapplicalionIo JD intefpreledhoflzons.lir-\r Br.r?1. t9. lt5 l(x). Shcrili. R. E. & Celdarr. I-. P (1995). F.tTtlottttionStisnto!r4t.CambritlgetJniversirvpre\s. Cambridgc.UK. 'Ianer. M. fl. & Kochle( F_( 1969).Velociryspeclra digilalcompulerderivationandapplicarions olvclocilyfuncrions.Cc(4)rr\i(.\.-14.lt59-ltL Whitc. R. E. ( 19110). Prrlial cohcrencemrlchinS()1\ynrhericseisnrogranr; u,irhsei\D1ic rraces.ai.r) t1h\'.\i&|Pt1)s|.din!.28. -l.ll 5lJ_ Ziolkou,ski.A. Underhill.J. R. & Johnslon.R. G. K. (l99ll). Wavelels.wcll ties.and the searchtbr subtlestraligraphic traps.6co2lrr.slr r. 6-3.297-3 13.

I

interpretation 4 Geological

All seismicinterpretation is of coursedirected toward geological underslandhg of the subsurface.In the previouschapter,the objective of the interpreterwasto makemapsof surfaces,mainly in order to delineatetrapsby mapping the top of a reservoir Howeverhow doeshe or sherecognisewhere the reservoirsare likely to be in an undrilled area? What reflectorsare most likely to be the top of a reservoirbody? If thereare some but whatis their well dataavailable,perhapsreservoirshavealreadybeenencountered, lateral extent likely to be? What lateral changesin reservoir quality are likely, and how shouldthey be relatedto changesin seismicappearance? Thesequestionsare of coursejust asrelevantfor 2-D seismicas for 3-D, but the densedataprovidedby 3-D seismic offers more scopefor defining the extemal geomefiy and intemal architecture of reservoirbodies.The detailedmap view derivedfrom 3-D seismicis often more instructivethan an individualsectioncan be. Before embarkingon a moredetaileddiscussion,it is important to understandthe limitationson achievableseismicresolution;this is discussed in section4.1.Theprinciples of seismicstratigraphyarebriefly explainedin section4.2,includingtherecognitionof seismicfacies.Sometools to allow the interpreterto look for the expressionof different sedimentaryfaciesare describedin section4.3, and someexamplesof the results presentedin section4.4. The structuralgeologistalso has of coursean input to make to 3-D interpretation. The needfor validationof fault pattemsis lessthan in the caseof 2-D surveys,where aliasingof fault pattemscanbe a major issue.However,understanding of fault systems may be critical to understanding whetherfaultswill form effectivelateralseals.These topicsarediscussedin section4.5. -

4.1

resolution Seismic Both vertical and horizontalresolutionof seismicdata are limited, and this imposes limits on what geologicallysigniflcantfeaturescan actuallybe recognisedon seismic data.Verticalresolutionis determinedby the seismicsourcesignaland the way it is filtered by the earth. For example, the signatureof a typical marine air-gun array has

102

103 -

Seismic lesolution

0

10

(ft) 40 20 Thickness

50

60

70

80

g0

(a)

(b) Fig. 4.1 (a) Wedgemodel for bandwidth 6 50 Hz; (b) model for calcularingthin bed response from the differenceof identical waveletsslightly displacedin time, redrawn after Widess (1973) with permissionof the SEG. frequencies in the range 8-150 Hz; the upper frequency limit will be reduced as the

seismicsignalpropagates throughthe earth,perhapsto about50 Hz at a TWT of 2 s. The effect of finite bandwidthcan be studiedusing a simplemodel (fig. 4.1(a)).This showsthe zero-offsetresponse to a wedgeof materialincreasingin thicknessfrom zero

104 -

interpretation Geological to 100ft in 1 ft increments.The reflectioncoemcientat the top of the wedgeis -0.15 andthatat thebaseis +0.185. Thematerialof thewedgeis thereforesignificantlysofier thanthe materialaboveor below it, andthe materialaboveit is slightly softerthanthat belowit. Thesevaluesarebasedon thosefor a wedgeofporousgas-filledsandencased in Teniary shalesin the UK CentralNorth Sea.Figure 4.1(a) showsthe calculated waveletof bandwidth6-60 Hz. The polarity of the seismicresponsefor a zero-phase display is that a black peak marks a transitiondownwardsto an acousticallysofter material.Wherethe sandis absent,at the left-handend,thereis a weak white trough dueto the impedancedifferencebetweenthe shalesaboveandbelow the sandlevel.At the righfhand end,the top of the sandis markedby a sffongblack loop and the base by a sfiongwhite loop. Thereare small-amplitudewigglesbetween,aboveandbelow thesereflectors,causedby minor oscillationsin the wavelet,but it would clearly be possibleto pick the strongloops at top and basesandaccurately,and measurethe TWT intewal betweenthemto determinesandthickness.As the sandbecomesthinner, betweenthe top andbaseloopsreachesa nearlyconstantvalue howevel the separation at a thicknessof about40 ft. The point at which this happensis oftencalledthe ti;nlng in sand remainsnearlyconstant,andfurtherdecrease Afterthis,theseparation d?ictness. the intelferencebetween is the res tof This thicknesscausestheamplitudeto decrease. reflectionsat the top andbaseof the sand;thereflectionsfrom the top andbaseoverlap and, being of oppositepolarity,partly cancelone anotherBelow 40 ft thickness'the events.It is very importantto takethis into top andbasesandarenot visibleasseparate accountwhen estimatingreservoirvolumesin thin sands;usingthe isopachbetween the volume. top andbaseseismicreflectorswill grosslyoverestimate A methodto calculatethicknessesfor thin sands,below the tuning thickness,was discussedby Widess(1973),using a simple model wherelhe reflectioncoefficients are the sameat the top andbaseof thebed.As shownin fig. 4.1(b),the resultingsignal is the sum of the reflectionsfrom the top and baseof the bed, which are of courseof oppositepolarity;it is thereforethe differencebetweentwo identicalwaveletsslightly displacedin time. When the bed is very thin, the characterof the reflectionis that of the time derivativeof the incidentwavelet.Widessshowedthat the characterof the compositereflectionis unchangingfor bedswhosethicknessis lessthan about)'/8' of to thepredominantperiod where). is thewavelengthin thebedmaterialcorresponding the wavelet.For bedsthinnerthanthis,reflectionamplitudeis givenby 4nAb/1, where A is the amplitudethatwould be obtainedfrom the top of a very thick bed(i.e. with no interferenceeffect),andb is thethicknessofthe bed.Thustheamplitudeis proportional to bedthicknessfor thesethin beds,andthis canbe usedto predictbedthicknessfrom seismicamplitudeif the dataare calibrated(e.g.to a well) and if we can assumethat all lateralamplitudechangeis causedby changesin thicknessand not by changesin impedanceof the thin layeror of the matedalaboveandbelow it. As rhebed becomes thinner,the amplitudewill eventuallydecreaseso far that it is invisible.The thickness wherethis will happenis not easyto predict,becauseit dependson the level of seismic

105 -

Seismic resolution

The WedgeModel

Wedge Model

a50 E ;40

. TuningCurve

,9

d E

E

(msl truethickness Fig.4.2

Schematicwedgemodel for tuning eflects.

noise,andthe extentof complicationsfrom the presenceof otheradjacentlayers.If it is, say,half the resolutionlimit, this impliestharthe standardseismicmethodwill see only thoselayerswhosethiclless is greaterthan say20 ft. A model of a tuned hydrocarbonsandresponseis shown in fig. 4.2. This shows the amplituderesponseand apparentthicknessof a sandbed encasedin shale,using a zero-phasewavelet,and increasingthe bed thicknessfrom zero throughthe tuning range.As expected,thereis a linear increasein amplitudewith true thicknesswhen the bed is thin, while the apparentthicknessremainsconstant.Thereis a maximum amplitudeproducedby constructiveinteference,wherethe precursorof the reflection from the baseof the sandis addedto the main lobe of the reflectionfrom the top of the sand.Beyondthis point, the top and baseof the sandare observableas separate reflectors,andtheamplitudefallsto thevalueexpectedfor anisolatedtop sandreflector. Figure4.3 showsan exampleon an actualseismicline. The amplitudeof the gassand reflectionis highest(brightyellow) on the flants of the structurewherethereis tuning betweenthe top of the gassandand the gas-watercontact,andtlecreases towardsthe crestof the structure(orange-red)wherethe gas column is greater;in this panicular. example,thecolumnis nevergreatenoughto resolvethegas-watercontactasa separate event.In map view,the resultwill be a doughnut-shaped amplitudeanomaly,with the highestamplitudesforming a ring aroundthe crestat the point wherethe tuningeffect producesthe highestamplitude.

106 -

Geologicalinterpretation 't

u dadsano

Fig.4.3 Seisnicsectionshowinga tunedgassandresponse. Hodzontal resolution of seismic data is also limited. Migration in theory collapses a diffraction hyperbolato a focus whose width will be abouthalfthe dominant wavelength (Claerbout, 1985) if data are available at all dips, or more pessimistically a width about equal to the dominant wavelength (Stolt & Benson, 1986) if the dips rhat can be incorporated into the migration are limited to a maximum of 30 45.. This means that the best lateral resolution we could hope for, with the 6 60 Hz wavelet, might be about 60 ft, or about l8 m. However, hodzontal resolution is severely degraded by even small enors in migration velocity; as pointed out by Lansley (2000), a 0.5oloerror in migration velocity can degrade horizontal resolution by a factor of more than 5. In practice, therefore, hodzontal resolution will often be in the range 50 100 m. This is something that should be bome in mind when making decisions on exactly where on a seismic survey a well should be located. If it is very close to a major fault there will be a risk of reaching the target on the wrong side of the fault. The horizontal resolution also limits the geological detail that we can hope to see on seismic sections,though it is sometimes possible to see more than these formulae suggest(Goulty, 1997).

-

4.2

Seismic stratigraphy To understand the distributionof a reservoirsandin the subsurface. we needto seeit as part of a depositionalsystem.Thereare eight primary clastic(sandandmud) depositional systems(Galloway,1998):alluvial fan, fluvial, delra,shore-zone,shelf, slope, aeolianandlacustrine.Overtime,thedepositionalsystemswithin abasinchange;abandonedsystemsareburiedandeventuallymaybecomereservoirrocks.Usinga combina, tion of seismicdataandwell control,it may be possibleto understand the depositional systemwell enoughto be ableto predictsanddistributionandquality in areasthathave not yet beendrilled. This is achievedin parl by recognitionof individualdepositional unitsfrom the seismicdata,andin part by placingthemwithin an overallcontext.The

1O7 -

SeismicstratiolaDhv

lattertaskis aidedby the conceptsof sequencestratigraphy,which usesunconformity surfacesto defineboundariesofpackagesofrocks thatareof similarageanddeposited within a relatedfamily of depositionalsystems. Galloway(1998)givesseveralexamplesof the way in which it may be possibleto infer sanddistributionfrom detailedseismicmapping.For example,chamel systems areoftenseenin submarinefan,fluvial, anddeltaicenvironments. Channelsscourtheir bedsandbanksduring periodsof high-volumeflow, and depositsedimentwithin and aroundthe channelduring periodsof lower flow. Channelsmay be straightto highly sinuousin plan view, broad to narrow, and shallow to deep.They may be largely erosional(depositinglittle sedimentbeyondtheir banks)or depositional,buildinglarge levees.In general,muddy systemstend to havenarrow,deepsinuouschannelswith prominentlevees;in sucha system,sandsare often narrowisolatedlenticularbodies. Sand-richsystems,on the otherhand,tend to havebroad,low-sinuositychannelsthat do not have welfdevelopedlevees.Another type of sanddepositionis the result of unconfinedfluid flow. This is mostobviouslyfoundin a marineshelfor aeoliansetting, but also occurswithin other environments,suchas crevassesplaysalong riyers and turbiditelobesin submarinefans. Within a depositionalsystem,sandyreservoirandmuddysealassociations showpredictablepattems.Forexample,in a fluvial systemthebestsandsarefoundaschannel-fill deposits(e.g.point bars).Crevassesplaysalongthe channelbanksmay containsands depositedin small branchingflood channelsthat arepoorly connectedto the sandsof the channelfill. Muddy depositsin abandonedchannelsmay segmentthe top of the channel-fillsandwith shaleplugs;leveeson oppositesidesof a channelmay not be in pressurecommunicationif the channelis mud-filled.Verticalas well as lateralfacies changesmay be predicted.For instance,in map view a deltaconsistsof the deltaplain with a networkof distributaries,the deltafront with beaches.tidal flats and channel mouths,andthe submarinedeltashoreface andmuddyprodelta.As the deltabuildsout acrossthe shelf,a corresponding verticalsuccession is formed:a basalmuddyprodelta faciesis overlainby deltafront sands,cappedby lenticulardistributarychannelfill units and sealedby mudstonesdepositedwhen deltalobesare abandonedand transgressed by the sea.The distinctivecontributionof 3-D seismicis that mappingof theseindividual units will be much more reliablethan can be achievedwith a 2-D grid, and so inferencesbasedon the shapeof the bodies(e.g.channelsinuosity)will be muchmore reliable. The overall depositionalsettingcan be elucidatedusing the conceptsof sequence stratigraphy. A usefulsummaryofcunent thinkingon thistopichasbeengivenby Read(1996). ing & Levell As originallypublished,therewasconsiderable emphasison cycles ofsea-levelchangeasthecauseofsequencedevelopment andthemaincontrolon stratigraphicfacies;chartswerepublishedpurportedlyshowingglobal sea-levelbehaviour overgeologicaltime (Yail et al., 1977).The conceptof a universallyvalid, global,sealevel curvehasbeenquesiionedby many authors(see,for example,Underhill, 1991),

108

Geologicalinterpretation who point out the importanceof local tectonicsto local relative sea-levelchange. Howeverthis may be, the generalconceptof the sequencestratigraphicmodel has proved useful in predictingthe lithological successionat a continentalshelf margin during a single cycle of relativesea-levelchange.Depositionalsystemsmay be describedin termsof systemstracts,containingcontemporaneous depositionalsystems thatpasslaterallyfrom fluvial to deltaicto deep-watersystems. Thesesystemstractsare often interpretedin termsof their positionin the sea-level cycle, consistingof a major sealevel fall, a lowstand,a sealevel rise and a highstand.Sea-levelfalls resultin the formationof unconformitiesthat form the sequence boundaries,andexhibit sub-aerialexposureanda downwardshift in coastalonlap.In the initial sealevel fall, thereis erosionof the coastalsystemand depositionis confined to basinfloor fans.During formationof the lowstandsystemstract, a lowstand wedgeof sedimentis depositedthal consistsof leveedchannelcomplexesof slope fansandshelf-edgedeltaiccomplexes.As sea-levelrises,a transgressive systemsffact is formed; depositionis reducedin the basin and transgressive systemsform on the shelf and the coastalplain. At the top of the systemthere is a maximum flooding surface,andthe highstandsystemstractis markedby systemsthat aggradeandeventually progradeseawardasaccommodation spacecreatedby the sea-levelrise decreases. Sequenceboundariescan be recognisedon seismicfrom the onlap patterns,but can be difficult to pick. Study of isolated2-D seismicsectionsmay miss significantfeaturesbecausethey ignorethe map view and do not seelateralchangesparallelto the coast. Each systemsffact presentsits own reservoirassociations. Thus, in the lowstand tract, most sedimentbypassesthe fluvial and delia plain environments;shelf-margin delta lobes will offer reservoirtargetsin the delta front, and sandbodiesassociated with disfributarychamels.In the transgressive tract,stratigraphictrapsmay be formed in srike-parallelsandbodiessuchas wave/tidereworkedshelf sandbars and barrier islands.The highstandsystemstract often containsfluvial channel-fillsandsencased in overbankmudstones. One of the key elementsof sequencestratigraphyas formulatedby Yail et al. in AAPG Memoir 26 wasthe assertionthat seismicreflectionsgenerallyfollow chronostratigraphicsurfaces.Although this seemsto be correctin many cases,it is not immediatelyobvioushow to relate this to the way that seismicreflectionsare caused by impedancecontrastsacrosslayer boundades,which is fundamentalto the type of detailedpredictionof reseruoirpropertiesdiscussedin chapter5. The importanceof resonance betweenthe seismicpulseandcyclic sea-levelchangehasbeenemphasised by Anstey & O'Doherty (2002). For typical sedimentationrates,a cyclic sea-level variationwith a period of 1-5 million yearswould give rise to cyclic sedimentation pattemswiih thicknessesin the range 15-300 m; for a typical seismicvelocity of 3000m/s, theywould havea TWT thicknessof 10-200ms,which is aboutthe sameas the periodrangefor typical seismicwaves(frequencies5-100 Hz). The effect of the

109 -

InterDretationtools cyclic changein sea-levelis to causea changein accommodation spacethat will cause depositionat anyparticularpoint to changefrom morelandwardto moreseaward,and back again.If thereis a systematicimpedancedifferencebetweenlandwarddeposits and seawarddeposits(causedperhapsby their being sand-proneand shale-prone, respectively),thentherewill be cyclic changesin impedance.During onehalf-cyclethe impedancewill generallybe increasing,soreflectioncoefficientsat individualthin beds will all be positive,andover the otherhalf-cyclethey will all be negative.Supetposed on thesecyclestheremay be quitelargeimpedancejumps (andresultingreflectioncoefficients)at depositionalhiatuses.Howevel thecyclic effectswill dominateif thereis resonance with the periodsfound in the seismicpulse,which is why seismicreflectors tendto be chronostratigraphic markerslinked to cyclic sea-levelchange.Incidentally, this is an additionalreasonwhy well syntheticsmay not matchactualseismicdata;the resonantreinforcementof the signalmay be quite sensitiveto detailsof the wavelet usedto createfte synthetic.If it doesnot havea smoothbroad-bandspectrum,then the resonanteffectsmay be suppressed at frequencieswherethe waveletspectrumis deficient.

-

4,3

Interpletationtools Geologicalinteryretationof seismicdatamaybe simpleifthere is adequate well control, butin manycasestheinterpreterhasto makeinferences from theappearance ofobserved bodies.This mayincludeboththeirextemalform and,ifresolutionis goodenoughfor it to be visible,the geometryof intemalreflections.Somewaysoflooking for distinctive t-eatures are asfollows. (1) Verticalsections.Standarddisplaysas discussedin chapter3 may be adequateto show the geometryof individual bodies,particularlyif they are thick enoughto show distinctiveintemal reflections,such as the dipping foresetbedsof a delta front. The top and baseof the unit containingthe foresetscan be picked by the samemethodsas usedfor structuralinterpretation,and in somecasesit may be possibleto map a numberof verticallystackedor laterallyequivalentunits.A tool that may be useful is the instantaneous phasedisplay,which is derivedfrom the seismictraceasfollows (Bames,1998;Taneret a1.,1979).Supposewe definethe envelopeof the seismictraceat any particularTWT as the maximumvalue that the tracecanhavewhenmodifiedby applyinga singlephaserotationto the entire trace.Inprinciple,thiscouldbefoundby observinghow thetracechangeswhenthe phaseis rotatedthough the range0 to 360"; at any TWT, the maximumvaluethat the traceassumesduring the rotationis the envelopeor instantaneous amplitude, andthephaserotationthatgivesriseto thismaximumamplitudeis theinstantaneous phase(reversedin sign).Finding thesevaluesfor all times on the tracagivesus the envelopeand instantaneous phasetraces.The actualcalculationis in pracrice

lrit

1

: rl

",r r:l

,t, . , I I

,..i , i

,i;

'110 -

Geologicalinterpretation

computationally straightforward enoughfor many workstationsto havethe ability to createinstantaneous amplitudeand phasedisplaysfor any sectionin not much more than the time takento displaythe data.Essentially,the instantaneous phase displaylookslike a seismicdisplaywith a very shorlgateAGC applied;amplitude informationis suppressed, and peakscan be followed acrossthe sectionwith a constantphaseof 0", troughswith a constantphaseof 180', and zero-crossings with a phaseof 190'. Whendisplayedwith a suitablecolour-bar(i.e.onethatstarts andendswith the samecolour so that +180" hasthe samecolour as 180"),then phasesectionmakesit easierfor the interpreterto spotangular the instantaneous relationships(onlap,downlap,etc.) in low-amplitudepartsof the seismicsection. An exampleis shownin fig.4.4. Arrows highlightplaceswherethe instantaneous phasesectionshowsangularitiesmoreclearlythanthe standardreflectivitydata. (2) Hodzontal sections.Time slicescan revealmap-viewgeometry,suchas channel systems.However,if thereis strxcturaldip present,fte horizontalslice doesnot showdataref'erringto a singlestratigraphiclevel. (3) Horizonslices.By slicingthroughthedataparallelto a particulareventil is possible to seea map of amplitudechangesat a singlestratigraphiclevel.This is often the bestway to seechannelandfan features,which are recognisedby their geometry in map view The referencehorizonis usuallychosento be the strongestandmost continuousmarkerwithin thesequence, asthis canbe autotracked mosteasily.This is a goodway to look at the intemalgeometryofthin layersat or belowthe limit of seismicresolution;all the informationis encodedin the lateralamplitudevariation of thereflector.In sucha case,it maybewo hwhileto invertthedataby themethods discussedin chapter6, with the aim of increasingthe bandwidthandthus getting slightly moreinformationout of horizonslicesthroughthe invertedvolume.With thickerlayers,it canbe moreinformativeto look at amplitudes(e.g.Ims averageto avoidmutualcancellationof positiveandnegativevalues)within a window whose thickless is chosenso as to enclosethe layer of interest;reconnaissance of the featureusingverticalsectionswill showwhat window sizeto use.A refinementof this idea is to use 'stratal slices' (Zeng et a1.,200I). In this method,displaysare producedof a seismicattribute(e.g.amplitude)on a geologicaltime surface.This surfaceis createdby linearinterpolationbetweenpickedsurfacesthat arebelieved ' to be time-parallelreferenceevents,e.g.marinefloodingsurfaces;aswe sawearlier in the chapter,suchsequenceboundaryreflectionsare often strong,easilypicked andlaterallycontinuous. (4) Coherenceslices.Use of horizontal slicesthrough the coherencecube to map faults was explainedin chapter3. Horizon-parallelslicescan be usedto reveal map-viewinformationaboutintemal sfructureof a layer,in exactlythe sameway as for reflectivity.Subtleintemal discontinuitiescan be revealed.To understand featuresseenin map view,.they may needto be comparedwith their expression on vertical sections;standardreflectivitysectionsshouldbe usedfor this purpose

111 -

Interpretationtools

Reflectivity .i..

1

;

Instantaneous

,.}

Fig.4.4 Reflectivityandinstantaneous phasedisplaysof a seismicsection. as the method of calculating the coherencecube makes vertical sections through it hard to understand. (5) 3-D views. The types ofview discussedin chapter 7 can be very helpful in understanding a geological feature. Ifthe feature is marked by high-amplitude eventsthat can be voxel-picked, it is easy to obtain a rapid first impression of its 3-D shape. Where the seismic expression of the body is less obvious, it may be necessaryto

''

'l'l1 -

Interoretationtools

Reflectivity

.'-1 1_

:*-*g

s*,.L:*

"

i

tr\ F:\

Instantaneous ohase

)

Fig,4,4

I { c f i c c I i v i I y d D d i n s t n t a n e o l t sp h a s e d i r p l a l s o l a s c i s n l i c s c c t i o n .

as thc rncthotlof calculalingthe cohelelce cubc nrakesvertical sectionsthrou-ghit harcl1ounclerstand. (5) 3-D vJcu's.The tvpes ol'vicu, rliscussecl in chaptcr7 can be vely helptul in under' stanclinga reological feutLrrc. Il the leatuleis nafkcd by high-amplitudeeventsthat can bc voxel picked. it is easy to obtain a rlpid ljrst impfessioltof its 3 D shapc'. Whele the seismicexptessio| of the body is lessobvious. it nay be necessaryto

112 -

Geologicalinterpretation

tnpul Fig,4.5 Schematic (afiercumey,1997). neuralnetwork pick a top and basesurfaceof the body by the usualmanualmethods,and then displaythe seismicbetweenthesesurfacesasa semi-transparent 3-D body. (6) An entirelydifferentapproachis to use a classificationalgorithmto map areasof similar seismiccharacter.One approach10 this is to use neuralnet soliware.A generalintroductionto the principlesof neuralnetworksmay be found in Gurney (1997),for example.A simple exampleof such a nerwork is shownin fig. 4.5. Input dataarefed to nodesin the first layerof the network.Eacharrowedpathhas associated with it a weight,and the input valuesare multipliedby the weight correspondingto theparticularpaththattheytravel.On arrivalat a node,the weighted inputs are summed;if the sum exceedsa thresholdvalue, the node sendsout a high signalvalue(conventionally'1 ') to the next layerof nodes,or if the threshold value is not reachedthe node sendsout a zero signalvalue.The sameweighting procedureis carriedout alongthe pathsto the secondlayer,andthe nodesin that layersumthe inputsandoutputa 1 or a 0 dependingon whetherthe summedinput exceedsthe thresholdor not.This behaviourofthe nodesmimicsthatof neruecells (neurons)in biologicalbrains.The weightson the interconnecting pathsdetermine how the systembehaves.They can be determinedby a learningprocess,in which input dataarepresentedfor which the correctoutputis known.For example,if we wantedto predictwhethera cerlainlayer is sandor shalefrom seismicdata,then the input could be a set of seismicattdbutevalues(traceamplitudes,loop widths, etc.) at a well locationwhereit wasknown whetherthe layerwas sandor shale.If we hada numberof wells, somewith sandandsomewith shale.eachwith its own seismicattributevaluesfrom an adjacenttrace,it would be possibleto adjustthe neuralnetweightsiterativelysothatthe outputis a sand/shale flag whenthe seismic attributevaluesare suppliedasinput. An extensionof this ideais to predictvalues of reservoirpropertiessuchas porosity.One way of doing this is to classifytrace data accordingto their similarity to syntheticsproducedfrom wells with known

113 -

Someexamoles

rock properties(Trappe& Hellmich,2000).A differentway of traininga net is to askit to clusterinput data,without specifyingin advancewhat the clustersshould bel thenetlooksfor similaritiesin pattemofthe inputdataandgroupstheinputinto classesaccordingly.For example,theinputdatato a netmight be seismictracedata, over a window hungoff a referencepickedhorizonandlargeenoughto encompass severalloops.The netthenclustersthedata,looking at the shapeofthe tracesrather than their amplitude.A map is produced,colouredaccordingto the clusterthat a tracebelongsto. An exampleis shownin fig. 4.6. Herethe datahavebeensplit into l2 classes,with characteristic traceshapesasshownin the lower part ofthe figure. All theseffacesbegin at the maximumof a trough,becausethe window usedfor dataselectionwashungoff an autotracked trough.The map showsa prominentlineation(arrowed)to thenorth-eastof which the tracesarequitedifferentfrom those elsewherein the map.This lineationis inferredfrom well contlol to be the edgeof a majorTeniaryfan system;within the fan,to the south-westofthe lineation,there areadditionalvariationswhich arenot well understoodfor lack of well calibration.

4.4

Some examples In thissectionwe describeexamplesof someofthe techniques describedin theprevious section.Stratalslicingis demonstrated try fig. 4.7 (Zenget a/., 2001).This comesfrom the Miocene-Pliocenesectionof offshoreLouisiana.From well data,it is known that in this areasandsare acousticallysofterthan shale.The polarity conventionfor these slicesis that red : soft, so in generalwe expectred to correspondto sandandblue to shale.Slicesat differentlevelsshowvariousfeatures.In (a),we seemoderatelysinuous, channel-likefeatures.Basedon comparisonwith well penetrations, theseare thought to be fluvial channelsin a coastalplain environment,with fining-upwardchannelfill. A deeperstratalslice, (b), showsa very different channeltype, with low sinuosity. Well penetrationsshowblocky log pattems.Theseare incisedvalley fills that contain depositsof lowstandand transgressive systemstracts.A deeperslice still, (c), shows soft red amplitudeswith lobateto digitateplan-viewgeometry,gradinginto low- to variable-amplitude lobes.Wells in the channels(e.g. log l) show an upward-fining distributarychanneloverlying upward-coarsening, progradingdelta deposits.Delta front deposits(log 2) containthin interbeddedsandsand shales,while prodeltawells encountershales(1og3). The overallsystemis interpretedas a highstandshelfdelta. Another exampleis shownin fig. 4.8, this time from offshoreEgypt (Wescott& Boucher,2000).Theseare submarinedelta front channelcomplexes,formed during a late Miocene earliestPliocenetransgression, and are well imagedon horizonslices through a coherencyvolume. The deeperRosettachannelcomplex is well defined becauseit is incisedinto underlyinganhydrite;it is charactedsed by shary channel edgesandlow sinuosity.Thesechannelsareinterpretedto be sedimentbypassconduits,

114 -

Geologicalinterpretation

Fig. 4,6 Seismicfacies classillcationmap. basedon the traceshapesin the lower part of the figure. Image producedusing StratimagicrM soliware (paradigm Geophysical),which incorpomres SISMACEIM technologiesdevelopedby TotalFinaBlf.

115 -

Some examples

1 t a -

F

s: l{ r

(

- {

-

T

.>i . f t<1 ; - - = t : { 1 ; 1 l ( f {

L

J

al2

_

ts ;l

r3\

r . / 5^ - 4 6 ?

SP

rt

g

J K I

Res l

l

.o{ | 5

)

1 i J o { 2

<

{-: 1

l

l ; t 3 i (

SP

Res

t

t

€- :t1 {

5 \

Jol

+:] 2

J: + -

Infs(edsedim€ntEource Fault

-, - - LJpdip limitoltVF -', ^' DeDositional axis

Fig.4.7 Amplitude stratalslicesshowing (a) a Pliocenecoasralplain, (b) an Upper Miocene incised valley fill, and (c) an Upper Miocene highstandshelfdelta system.CH : channel, FP: fl66601u;n,tUO : incised valley fill, SH: shelf. Reproducedwirh permissionfrom Zeng et al. (2001\.

116

Geologicalinterpretation

-l'nuiI.!

!siv.' ddFriirs lill {h$rncl coofioed nhhin Ro\cxa Anhydrit€

Aluhi(m o{curr at nrcan&r trcfid folk'$ ing a steafEr gr.dfunt

Fig.4.8 Compositecoherencyimage showing the Rosettachannelcomplex (orange),with the image ofthe Abu Madi channelcomplex (56 ms abovethe Rosetta;yellow) superimposedon the upper right- Light yellow outlines levee/overbankdepositsassociatedwith this channelsysrem. Reproducedwith permissionfrom Wescott & Boucher (2000).

transportingsedimentacrossthe shelf into deeperwater.They filled with sediment as transgression continued,and the systemwas no longer confinedby the resistant evaporitecanyonwalls. Eventuallythe channelavulsedat a meanderbend,following a steepergradient,and formed the Abu Madi channelsystem,with higher-sinuosity channelsconfined within fan/overbankdeposits. An exampleof interpretationof downlapgeometriesis shownin fig. 4.9. This is fiom the SanJorgeBasin,Argentina(Woodet a1.,2000).This is a fluvial system;the successivesouth-to-northdownlaponto the M7 unconformityis atffibutedto lateral accretionof migratingpoint bars.The overlyingdepositsshowaccretionaryprocesses from north to south,with downlaponto an unconformityseparatingthis unit from the deeperpoint bars.The top of this intervalis a widespreadfloodingsurface,markedby alluvialdeposits.Mappingamplitudes,on both standardreflectivitydataandcoherency volumes,in slicesparallel to this flooding surface,allowed channelsystemsto be mappedandtheir evolutionfollowed over time.

117 -

Faults

'" tit *4 a ' 1 q

z

o; ,

L;-*'

E

fr;11,a1,,,ertrt1$

09

;r;;;="..,9G;

1 0

1 1

(b; :* ;=

0n

-."9-.:"

.,5** 01

to -

^,i;;;ir,,".r,

_

*

l l

1:4

T'a$ '0'

Fig.4.9 Southto nonrhseislnicsectionwith Sp and resisrivitywell logssuper.poseci: (a) interpreted and (b) uninterpreted. Notc incisedchannelsolrtholwell gl4. Reprocluced with pennissiontiom Wood .r.1/. (2000).

-

4.5

Faults As we saw in chapter3, lault systemscan be n.rapped in sontedetail fi.ol.t3-D seismic. althoughthereis room for confusionbetweensmall taults and seis|njcnoise.It is often necessaryto determinewhethera mappedtault will fbtm an adequatelateralseal to a mappeclundrilled structure,or to evaluatewhether a tault will be a signilicantbarrier

118

Geologicalinterpretation to fluid flow duringproductionfrom a reservoirIf we can assumethat faultsarcnot in themselveshydrocarbonseals,then hydrocarbonscan flow acrossthe fault wherever permeablelayersarejuxtaposedacrossit. This can be analysedby drawing sections alongthefault plane,showingthelayersintersectingthefault on boththeupthrownand the downthrownside(Allan, 1989).From thisjuxtapositiondiagram,the spillpointof a structurecanbe determinedasthe shallowestdepthat which hydrocarbonmigration acrossthe fault is possible.This procedureis not as simpleas it sounds.Many faults are not simplesinglediscontinuities,but are complexzonesconsistingof a seriesof interconnected faulr segments(Knipe et al.,l99g). Often, a large singlefault would be judged an effectivebarrierto fluid flow from the juxtapositiondiagrams,but the equivalentensembleof small-displacement faultsmight not be. Very carefulmapping of the faultsis thenrequired,usingamplitude,coherencyanddip maps,togetherwith reviewofvertical sections,to eslablishthe fault pattem.This is particularlydifficult to do lbr the smallfaultsthat gradeinto seismicnoise. A complicationis that the fault planeitself may be an effectivehydrocarbonseal, eventhoughpermeablestrataarejuxtaposed acrossit. Thiscanbetheresultof smearing of clay alongthe fault planeduringdisplacementalongthe fault. variousmethodsfor predictingthe presenceof clay smearhavebeensummarisedbv Foxfordet a/. (199g). Thesemay be basedon: (i) the percentage of shaleor mudstonelayersin the faultedsequence, (ii) thepercentage of shalein the sequence thatwasmovedpastanypoint on the fault surface, (iii) the along-faultdistancein rhe slip directionof a point on the fault surfacefrom a potentialshalesourcelayer,andthe thicknessof the layer. Foxford et al. usedthe secondof theseapproaches (the shalegougeratio, SGR), and tbund that an SGR of less than 207owas charactedsticof fault zonesthat did not containshalegougein thei' particularstudy.similar cutoff varueshavebeenfound in otherstudies.As the SGR increasesabovethis level, the fault planebecomesa more effectiveseal,ableto hold a longerhydrocarboncolumnovergeologicaltime or sustain a largerpressuredrop acrossit on a field productiontimescale.However,the thickness of the shaleygougecan be highly variableand unpredictable.It is thereforedifficult to useSGR in a quantitativeway to determinefault permeabilities(Manzocchiet al., 1999).compilation of datafrom existingfieldsis neededto reducetheseuncertainties (Hesthammer& Fossen,2000).

-

Beferences Allan, U. S. (1989). Model for hydrocarbon migration and entrapment.American Association ()t Peh'oleurnGeologistsBulletin, 73, 803-l l.

118 -

ceologicalinterpretation

to fluid flow du'ing productionfrom a reservoirIf we can assumethat faultsarenot in themselveshydrocarbonseals,then hydrocarbonscan flow acrossthe fault wherever permeablelayersarejuxtaposedacrossit. This can be analysedby drawing sections alongthefault plane,showingthelayersintersectingthe fault on boththeupthrownand the downthrownside (Allan, 1989).From thisjuxrapositiondiagram,the spillpointof a structurecan be determinedasthe shallowestdepthat which hydrocarbonmigration acrossthe fault is possible.This procedureis not as simpleas it sounds.Many f'aults are not simplesinglediscontinuities,but are complexzonesconsistingof a seriesof interconnected fault segments(Knipe cl at., 1998).Often, a large singlefault would be judged an effectivebarrier to fluid flow fiom the juxtapositiondiagrams,but the equivalentensembleof small-displacement faultsmight not be. Very carefulmapping of the faults is thenrequired,usingamplitude,coherencyanddip maps,togetherwith reviewof verticalsections,to establishthe fault pattern.This is pafiicularlydifflcult to do lbr the smallfaultsthat gradeinto seismicnoise. A complicationis that the fault plane itself may be an effectivehydrocarbonseal, eventhoughpermeablestrataarejuxtaposedacrossit. Thiscanbe theresultof smearing of clay alongthe fault praneduringdispracement alongthe faurt.variousmethodsibr predictingthe presenceofcray smearhavebeensummarisedbv Foxfordet er.l99g\. Thesemay be basedon: (i) the pelcentageof shaleor mudstonelayersin the faultedsequence, (ii) the percentage of shalein the sequence thatwasmovedpastany point on the fault surface, (iii) the along-faultdistancein the srip directionof a point on the fault surthcefrom a potentialshalesourcelayer,andthe thicknessof the layer. Foxford et a1.usedthe secondof theseapproaches (the shalegougeratio, SGR), and found that an SGR of ressthan 207owas characteristicof tault zonesthat did not containshalegougein their particularstudy.similar cutoff valueshavebeentbuntl in otherstudies.As the SGR increasesabovethis level, the fault planebecomesa morc effectiveseal,ableto hold a longerhydrocarboncolumnovergeologicaltime or sust.rin a largerpressuredrop acrossit on a field productiontimescale,However,the thickness of the shaleygougecan be highry variableand unpredictable. It is thereforedifficult to useSGR in a quantitativeway to detemine fault perneabilities(Manzocchict a/., 1999).compilation of datafrom existingfieldsis neecred to reducetheseuncertainties (Hesthan.rmer & Fossen,2000).

-

References Allan, U. S. (1989). Model for hydr.ocarbonmigrarion and entrapment_Anrri(.an AssociatiottoJ Pehol?un CeologistsBulIetin.73. 803-l l.

119 E

Beferences Anstey, N. A. & O'Doherty, R. F. (2002). Cycles, layers, and reflecrions.The Leading Edge,2l, 44-51. Bames,A. ( 1998).The complex seismic hace made simple. fr? L eading E(tge,17,4.j34. Claerbout,J. F. (1985).lmaging the Edfth's h,erior.. Blackwell Scienrific publications.Oxford. Foxford, K. A., Walsh, J. J., Watrerson,J., Garden, I. R., Guscotr, S. C. & Burley, S. D. (199g). Sffucture and content of the Moab Fault Zone, Utah, USA, and its implications for fault seal prediction.ln: Faulting, Fauh Sealing and Fluid Flov, in HJdtlcdrbon Reservoiru(eds.G. Jones. Q. Fisher & R. J. Knipe), Spec.Publ. Geol. Soc.Lond'147. Galloway,W. E. (1998).Clastic depositionalsystemsand sequences:applicationsto reservoirpredic_ tion, delineation,and chara.teizatlon. The Leading Edge,17,113_9. Goulty, N. R. (1997).Lateral resolution of2D seismicillustratedby a real data exafiple. FirstBreak, 15, 77-80 (and conection it First Break.l5. 331 2). Gumey, K. (1997).Ar Introduction to NeuralNenrorkr. UCL press,London. Hesthammer,J. & Fossen,H. (2000). Uncertaintiesassociatedwith fault sealinganalysis.per,?leum Geoscience.6.3'745. Knipe, R. J., Jones,c & Fisher, Q. J. (1998). Faulting, fault sealing and fluid flow in hydrocarbon reservoirs:an introduction.lN Faulting, Fault sealing and Fluid Flov, in Hydrocarbon Resenoir-s (eds.G. Jones,Q. Fisher& R. J. Knipe),Spec.publ. Geol. Soc.Lond.,I47. Lansley,M. (2000). 3-D seismic surveydesign: a solution.Filirl B t.eak,lg, 1624. Manzocchi,T., Ringrose,P S. & Underhill, J. R. (1999).Flow rhroughfault sysremsin high porosity sandstones.In: Sr/rrcauralGeologv in Reseryoir Characterizati.on(eds.M. p Coward, H. Johnson & T. S. Daltaban),Spec.Publ. Geol. Soc.Lond.,l27. Reading, H. G. & Levell, B. K. (1996). Controls on the sedimentaryrock reco:d. \n: Sedimentary Ewii'onments:Pracesses,Faciesand Strctigrdpl) (ed. H. C. Reading).Blackwell Science,Oxford. Stolt, R. H. & Benson,A. F (1986).S?is/rl. M/g,?tlon. Geophysicalpress. Taner, M. T., Koehler, F & Sherifl R. E. (1979). Complex seismic trace analysrs.GeophJsics,44, 1041-63. Trappe, H. & Hetlmich, C. (2000). Using neural nerworks to predict porosity thicknessfrom 3D \ei.mic dara.Fil \/ BrcaL,18, 111-84. Underhill, J. R. (1991). Controls on Late Jurassicseismic sequences,Inner Moray Firth, UK Nofth Sea:a critical test of a key segmentof Exxon,s original global cycle chart. Basl, Rer., 3, 79_9g. Vail, P R., Mitchum, R. M. & Thompson, S. (1977). Seismic srrarigraphyand global changesof sea level.Part3: relarivechangesof sealevel from coastalonlap.In:.SeistuicStratigraphy Applications to Hydrocarbon Exploration (ed. C. E. payton), Mem. Am. Ass.petrr.tl.Geol.,26,Tulsa. Wescott,W A. & Bouchef P J. (2000). Imaging submarinechannelsin the westem Nile Delta and mterpretingtheir paleohydrauliccharacterisricsfrom 3-D seismic.The LeaclingEdge,19,5g(191. Wjdess,M. B. (1973).How rhin is a rhin bed'tGeophysics,38, ltj 6_80. Wood,L. J.,Pecuch,D., Schulein,B. & Helton, M. (2000).Seismicattributeand sequencestratigraphjc rntegrationmethodsfor resolving reservoirgeometryin San JorgeB asin. Atgefiina. The Leading E d e ( , 1 9 , 9 5 26 2 . Zeng, H., Tucker, F. H. & Wood, L. J. (2001). Stratal slicing of Miocene_pliocene sedimentsin Vermilion Block 50 Tiger Shoal Area, oftshore Louisiana. Ir? aeading Edge,20,408 lg.

121 --

ofisetreflectivity Compressional waves(P-waves)differ from shear(S) wavesin thedirectronofparticle motionasthe wavepropagates throughthe rock. For p_wavesthe motionis parallelto the directionof travelof the wave,whereasfor S_waves it is perpendicular. The p_ and S-wavevelocitiesare relatedto differentrock properties.When p_wavespropagate througha rock,therearechangesin the volumeofindividual particles, whereasS_wave propagatloncausesbendingwithout changeof volume.Standard seismicsourcesemit P-wavesalmostentirely,so we usuallyseeS-wavesdirectly only whenp-waves have beenpartly convertedto S-waveson reflectionat an interface.Standard seismicnro_ cessingconcentrates on usingp-wavesto form an imageof the subsurface. Ho*eu.r. aswe shall see,the shearpropertiesof the rock areimportantto understanding AVO. Sometimesa quantitycalledpoisson,sratio (o) is usedinsteadofthe VplV" rutio.It is givenby , . . , 2

o:

0 5- ( ; )

-

-,

-_. '-(*)

Figures5.1 and 5.2 show typicarvaluesof theseparametersfor some commonrock tvDes.

5.2

0tfsetreflectivity We saw in section 3.1.1 how the reflection coefficient at an intedace depends on the acoustlc impedance contrast across it for the case of normal incidence. ln the real world, seismic data are always acquired with a finite separation between the source and receiver (usually termed the o.fsel). This means that reflection will be much more complicated, becausepart of the p_wave energy will be converted into a reflected and transmitted shear wave. The equations describing how the amplitudes of the reflected and transmitted P- and S-waves depend on the angle of incidence and the properties of the media above and below the interface were published by Zoeppritz (1919); the amplitudes depend on the contrast in poisson,s ratio across the inteface, as well as the acoustic impedance change. Figure 5.3 shows an example of how the amplitudes depend on incidence angle for a particular interface. For a plane inteface the relationship between the p_wave angles of incidence and transmissionis given by Snell's Law (fis.5.,1): sin €2

v t :

sin g1

V

If velocity increases with depth across the inteface, then there will be an incidence angle fbr which the transmission angle is 90.. This is the critical angle, at and beyond which there is no transmitted p wave, and therefore a high reflection amptitude.

1n -

Interprelingseismicamplitudes

1.8

Bulkdensity(g/cm3) 2.2 2.4 2.6 2.4 3.0

2.0

4.4

Dolomite

25 000 20 000

Limestone 4.2

Rocks:r./

1 50 0 0

7

Anhydrite

'6

12000

q)

Time-average (Sandstone) '6 I

E 6 g 3.e

Sandstone

8000 > 7000

^_i

6000

t.7l/0.25 t

I 3980 0.5

0.4

0.2

(g/cms) Logarithm ofbulkdensity Fig.5.1 Densityvscompressional velocity,redrawnwith permission fromGardnereral. (1974). The Zoeppritz equationsare rather complicated.It is easyenoughto write software to generatecurveslike those shownin fig. 5.3, but it is also helpful to haveapproximations that give moreinsightinto the undedyingrelationsbetweenreflectivifyandrock properties. Useful approximationsfor the PP reflection coefficient (i.e. both incident andreflectedwavesofP type,the most commonsituation)havebeengivenby Aki & Richards(1980)andby Shuey(1985).Approximately, R@) -- A+ B sinza + C sinz0 tan20 where

a : o.s/4& + 44\ p / \ V o v

/v \2/

^v

Ap\

B:o.s;e-r(;) (zlr+rl 'p .p/

P I

123 --

othet retlectivitv

PoissonRatioandLithology

0.4 (!

rr

0.3 NonhSea Chalk

3 o.zshigh

.4.

0.15 0.1

Al (km/s.g/cm3) Fig.5.2 Poisson's ratiovs acousticimpedance.

EnergyPartitioning at an Interface Reflected Compressional

Transmitted Compressional Critical angle

Describedby the Zoeppritz Equations

L^ L*

Angleof incjdence

Reflected Shear

Transmitted Shear

Critical angle

Yp12 800ftls Ys8000ft/s p 2.4glcm3 yp 2'1000 ft/s Ys12 000ftls p 2.65glcm3

Fig. 5.3 Energy partitioning at an interface,after Dobrin (1976) Inttuduction to GeoDh\sicar , Ptospecting, rcprod:ocedwith permission of the Mccraw_Hill ComDanies.

, . ,i1 'll lti' 'l

124 -

lnterprelingseismicamplitudes

Snell'slaw

Fig.5,4 Compressional wavereflectionandfansmissionangles. and A% c : 0_5------:. vp where R(0) is the reflection coefficient at the incidence angle 0 (strictly speaking,at t}le averageof the angle of incidence and the angle of transmissionas determinedby Snell's Law), Vnis the averageof the P-wavevelocity on the two sidesof the interface, V. the averageS-wavevelocity, and p the averagedensity, and the quantitiespreceded by A are the differencesin the relevant parameteracrossthe interface. Theseequationscan be usedto model the seisrnicresponsewhen the rock properties are klown. The simplest model is the single interface betweentwo layers of different properties, and is often already very instructive. However, it is often necessaryto understandthe seismicresponseof a thin layer, for example to study how the seismic expressionof a reservoir sandchangesasit approachespinchout. This canbe examined

125

--l

Intenreting amplitudes usrnga wedgemodel,a singlelayerwhosethickness is systematically variedfrom zero to the desired maximum. A further step toward realism is to construct a model using all the differentlayers within someinterval,as recognisedfrom wireline well logs. This is important for relating well data to real seismic response,but the interference effectsbetweena whole stackof interfacescan be hard to ;nderstandunlessthe main individual layers are modelled separatelyfirst. For snall incidenceangles,a fufiher simplification of the Zoeppntz equatronsis possible;the term in the aboveformulainvolving C canb" n"gt""t"d. This is certainly thecaseout to incidenceanglesof30" or so,andis often a reasolnable approximationout to 40-45', beyondwhich thedataarein anycase oftenmutedout from gathersbecause of NMo stretchand the presenceof direct a'ivals. Then we can write the reflection coefficientin termsofthe normalincidencereflectivity R0 andIheAV7 pradient.G: R(0):Ro16rinzB.

--

5.3

Interpretingamplitudes Sometimeswe caninterpretfluid fill from amplitudes on seismicdata.tsefbrewe can do so, we needto havereasonableconfidencein the validity of the amplitudesin the seismicdataset.As explainedin chapter2, modernp.o""rrirrg will try to avoid any stepsthatcauseamplitudeartefacts.Ideally,we would like to havesersmicdatawhere amplitudesareeverywhereproportionalto reflectivity. This is not achievable, but what canbe doneis to makesurethatlocal lateralvariation of amplitude(overa distanceof, say,a kilometre)on a particulargroupofreflectors proportional is to reflectivitychange. We can often assumethat the averageabsolutereflectivity over a long ttme window varieslittle, so a long-gateAGC can be appliedto the data.It is essential,though,that the gateis long enough(1000 ms or more) to avoid destroyingthe lateralvariations we are looking for; the gate shouldinclude many reflectors,* ,t u, tn. targetevent makesvery little contributionto the averageamplitude in the window Calibrationof amplitudeto reflectivityis possiblefrom a well tie, but thecalibrationrs valid only over a limited intervalvertically.In any case,it is a good idea to inspectthe entiresection from surfaceto the targeteventand below; if amplitude uno.uli", at targetlevel are seento be correlatedwith overlying or underlying changes(high or low amplitudes due to lithology or gaseffects,or overburdenfaulting, for examf,le),thenthey should be treatedwith suspicion.Sucha correlationmight havea genuinegeologicalcause, but carefulthoughtis neededto establishthatthe effectis not-anartefact.folowing the amplitudeanomalythroughthe seismicprocessing sequence from theraw gathersmay be helpful;this may revealan a-rtefact beingintroducei in a particularprocessingstep. To recognisehydrocarboneffects(Direct Hydrocarbon Indicato.s,DHIs) for what theyare,we needto know what to expect.The sketch in fig. 5.5 showswhatto look for

,'l

ltr li' ,ll ttl

126 -

Inlerpretingseismicamplitudes

----------------.>

€

ii

I

l I Shale I

I {

Brightspot

O lloas sand

-l FI

t

e'**"a

Polarity reversal

Dimspot

Red= hardoop(mpedancencrease) Blue= softloop(impedance decrease) Fig. 5.5 Schematicmodels fbr oil and brine sand response.

in threepossiblecases.This generalideais valid for both standardfull-offsetstacksand thenearandfar tracesub stacksthatwe shalldiscussin thenextsection,althoughwhich pictureappliesmaydependon whatoffsetrangewe look at.On theleft we havethecase wherea brine sandis softer(lower impedance)thanthe overlyingshale.The presence of oil or gas will make it softerstill. We thereforeseean increasein amplitudeover the crestof the structurewherethe hydrocarbonis present.This is the classic'bright spot'. Of course,an amplitudeincreasemight alsobe causedby a lithologicaleffect, for examplea decreasein sandimpedancedue to porosityincrease.A key testis that we would expectthe bright spot,if it is really theresultofhydrocarbons,to conformto structure:in map view,the amplitudechangeshouldfollow a TWT contour(or strictly speaking,a depthcontourafter time-depthconversion,thoughover a limited areathe TWT anddepthcontoursareoftensimilarin shape).We alsohopeto seea 'flat spot' at the hydrocarbon-water contact.This is alwaysa hard reflector(impedanceincrease). The flat spot shouldbe at the sameTWT as the amplitudechange.(If we haveboth oil andgaspresent,then we may seetwo flat spots,oneat the gas oil contactandone at the oil water contact,with matchingamplitudestepsin the top reservoirreflector, from very bright in the gas-bearing part to brightin the oil-bearingpart.)In themiddle, we seethe casewherethe brine sandis hard relativeto the overlying shale,but the hydrocarbonsandis soft.The top sandwill be a hardloop belowthe fluid contactand a sott loop aboveit, with a polarity changeat the contact.This caseis oftendifficult to particularlyif the structureis affectedby minor faulting.It is interpretwith confidence, often possibleto keepthe top sandpick on a hardloop acrossthe crestby interpreting a small fault nearthe contact.Inspectionof linesin differentazimuthsacrossthe crest

124 -

Interpretingseismicamplitudes beddingreflectorscancausethe flat eventto be brokenup into a seriesof segments that may individually appearto be slightly tilted, althoughthe ensembleremains flar). (3) Thereshouldbe apparentthickeningof the isochorein the interualabovethe top sandif the altemative'continuousstrongloop' interyretationis used. (4) Theflat eventat theOWC shouldrun horizontallyacrossinclinedbedding,resulting in apparentreflectorteminations belowit. (5) Crucially,the amplitudedimming, the flat spot extentand the apparentisochore changeshouldbe consistentin map view with eachotherandwith a mappedtrap, e.g. a dip closure.The amplitudechangeshouldfollow a structuralcontou if it is indeedcausedby a changein fluid type at the downdipedgeof a trap. This is where3-D seismiccan makea big contribution.Both the amplitudemap and the structuralmap are much moredetailedthan couldbe achievedusinga grid of 2-D d a L a\ o . t h i st e s ti s m u c hm o r er i g o r o u s . If all the testsare passedthen it is possibleto have a high degreeof confldence in the interpretationof fluid fill. It is quite usualfor the evidencenot to be so clearcut, however.In particular,a lailure of amplitudesto fit structuremay be causedby a stratigraphicelementof the trappingmechanism,or by complicationsdue to lateral changesin rock propefiies(e.g. porosity).The evidencethen needsto be weighed careiully togetherwith geologicalunderstanding. Is thereenoughwell control for us to be reasonablyconfldentof sandand shaleproperties?Is the seismicdata quality adequate? For example,flat eventsmay be multiplesof sub-horizontaleventshigher in the section.If so, they can certainly cut acrossbedding,as expectedfor a fluid contact,but will be likely to continueacrossthe top-sealas well as the reservoir.It is particula y suspiciousif the flat eventis one of a whole suite at differentTWTs, which pointsstronglytowardsit beinga multiple.It is oftenpossibleto seevery weak flattish eventson a seismicsectionif they are looked for hard enough,and they are often multiplesthat havebeenreducedin amplitudebut not quite eliminatedduring processing. In manycases,lateral amplitudechangesarerelatedto changesin porosityratherthan fluid fill. This is particularlytruefor well-consolidated sandsandcarbonates. Figure5.7 showshow theimpedanceof theChalkin theNorth Seais stronglyaffectedby porosity, but relativelylittle by fluid fi11.In thiscase,theTopChalkwill be an impedanceincrease (red trough with the usualNorth Seapolarity convention)if the porosityis lessthan (bluepeak)for higherporosities.At constantporosity,the 357o,changingto a decrease differencebetweenthe avemgeimpedancetrend (in blue) and the lower valueswith hydrocarbonfill (in dottedblack) is quite small;a similar impedancechangecould be causedby quite a small changein porosityat constantfluid fill. Conversely,the large impedancechangecausedby porosityvariationwithin the usuallyobservedrange(say 10-407o)is much greaterthan would be causedby any changein fluid fil1.If thereis enoughwell control to calibratethe relationship,it may be possibleto infer porosity

1N I

Intenroting amplitudes

tt o

CI

relatively smallchange in impedance r€latedto pore geometryandfluidfill

(, a

Firstordertrendof Al andPHI impedance of overlying Shale TopChalk= Trough

TopChalk= Peak

Porosity Fig.5.7 Chalkimpedance versusporosity(afterCampbell & Gravdal,1995). values from the Top Chalk amplitude. A similar approachcan be adoptedto deduce sand porosity from seismic. However, this method assumesthat the rock above the interface (e.g. a caprock shale) does not vary laterally. If there is a dsk that it does, then the inversion methodsdiscussedin chapter6 are a better way to estimatereservoir porostty. A completely different sort of DHI is the gas chimney. This occurs where gas has leaked from a deeperlevel into the overburden,typically along a fault plane, but the ovelburden is mainly shale with limited permeablezones (e.g. silts). The rcsult is a diffrrse cloud of gas-bearingmaterial, typically with low safrations. There may be a few high-amplitude gas sandreflections at the top or within the body of the cloud, but in geneml scatteringand absorptioncauseamplitudesto be much reducedbelow and within it, so that amplitude measurementsare usually meaningless.There is often an apparentsag in TWT of reflectors below the cloud, due to the velocity decreasein the gas-bearinglayers; this can cause great difficulty for accuratestructural mapping in depth.Shear-wavedata,which are almostunaffectedby the gas,may be the best way to image the horizons below the clorid. This is often important becausealthough the gas saturationswithin the chimney are too low to be of any economic value, the presence of the chimney points to the possiblepresenceof a leaking trap below it. TUningis a complicationfor the studyof amplitudes.As we sawin section4.1, am_ plitudes from a thin bed can be greateror smaller than the value expectedfor a single interface, dependingon the thicknessof the bed relative to the seismic wavelensth. It

r30 -

Interpreting seismic amplitudes is often possible to observethis effect on seismic sections,manifestedas an amplitude maximum at a particular apparentbed thickness. Sometimesit may be necessaryto model this effect and allow for it when interpreting amplitude variation. For bed thicknesslessthanthemaximumof thetuningcurve,it maybe possibleto mapbedthickness using the linear amplitude-thicknessrelarion, althoughthis will work only if thereis no lateral variation in acousticimpedanceof the bed andthe material encasingit. Accurate thicknesspredictionalsodependson usinga correcttuningcurve,which in turn depends on having ar accurateestimateof the wavelet presentin the data.

-

5.4

AVOanalysis AVO responsecanbe classifiedinto four categories(fig. 5.8) dependingon the values of R0 and G, the normal incidencereflectivity (sometimesreferred to as irtercept) and gradientvaluesdefinedin section5.2.Figure5.8 showstypical responses for different shale-sandinterfaces(i.e. typical top reservoirin a clastic sequence);a shale-sand interfaceusually exhibits a negativegradient,i.e. the reflectioncoefficientbecomes more negativewith increasingoffset.ClassI responseis characterised by an increase in impedancedownwards acrossthe interface, causing a decreasingamplitude with increasingincidenceangle.ClassII responsehas small normal incidenceamplitude

Classesof AVO Response

rmpeqance Contrast

+

Modeloriginally basedon shale/brine sandinterface

Lrl

o l

t J

o_

f_, I

AVOcurvescanbe generated from: seismicgathers models s i n g l ei n t e r f a c e syntheticgathers

Fig,5.8 TheAVO classes. ModifiedafterRutherford& Williams(1989),Ross& Kinman(1995) andCastagna & Swan(1997).

131 -

AVoanalysis

PoissonRatio,Al and Shale/Sand AVO Responses Deep

ttt I ctass = E c I!

0.3 --

10000 20000 30000 40000 Acoustjc lmpedance (ft/s,g/cm3) Fig.5,9 Schematic plotofPoisson's ratioagainst acoustic impedance. (positive or negative)but the AVO effect leadsto high negativeamplitudesat far offsets. Someauthorsdistinguisha classIIp, where the zero incidenceresponseis positive and therefore there is a polarity reversal at an intermediateoffset. class III responses have large negativeimpedancecontrastsand the negativegradient leadsto increasing amplitude with increasing angle. Class IV responseoccurs where a large negative amplitude decreasesslightly with offset. confusingly, a reflector is sometimesreferred to as exhibiting positive AVO if the amplitude, irrespective of sign, increaseswith offset. In a clasticsequence, the AVO classesare relatedto differencesin consolidationof sandsand shaleswith depth(fig. 5.9). The generalincreasein impedancewith depth (fig. 5.10) reflectsthe decreasein porositydue to compaction.ClassI responsesare characteristicof deepwell-consolidated sections,and classIII responses of relatively unconsolidatedshallowsediments.class IV caa occur in very unconsolidatedsands, shallowerthan about 1000m, or wheresoft sandsare found below a non-clastichard layer.The featurethatdistinguishes il from classIII is the very low gradient;in practice it is often hard to say in real seismicgatherswhethera low gadient is positive or negativebecauseof scalingissuesto be consideredshortly. A commonmethodofAVO analysisis theAVO crossplot.R0andG canbe calculated for eachloop on every CMP gatherin a seismicsurvey,by measuringthe amplitudeand calculatingthebestfit to a plot of R(0) againstsin2p. Theresultingpairsof Re,G values can be chartedon the crossplotand, aswe shall seeshortly, may give us information on fluid fill andlithology.Ofcourse,we do not directlyobservethe incidenceangleA, but

132 -

Interpretings€ismicamplitudes Acousticlmpedance -}

q)

o

Fig.5.10An example of impedance trendcurves, based oncregory(1977). insteadhaveto calculateit from the source-receiveroffset. A very rough approximation would be to assumethat the ray-path from surface to reflector is straight, but this invariably underestimatesthe angle of incidence becauseof refraction effects caused by velocity increasewith depth.Ray-tracing through the overburdenvelocity structure will give an accurateestimate,but an approximateformula for the incidence angle at offset X and zero-offsettravel time Z0is ,

/ V;,\/

X2

s i n ' 9 = [ " l t ; ;-; , ; - - ; = ) \ \ vms,/ \ /1 f (yrnsto)-u where V1n1 is the interval velocity at travel time 20, and V.-, is the rms velocity from the surfaceto Z6 (i.e. approximately,the stacking velocity). Measuring AVO gradient from real gathersis not entirely straightforward. Residual moveoutcan distort the measurement. Figure 5.11 showsftacesfrom a CDP gather, where correction for NMO has not been accurate,and as a result the reflectors are not flat. The true AVO gradientwould be found by measuringthe amplitude along the peak or trough of a reflection. If the amplitudesare taken insteadat a constanttime sample, asmaybe thecasewith softwaredesignedfor bulk processingofdata,thenthe gradient will be too high. One way round this is to use trace-to-tracecorrelation to follow the Ioop acrossthe gather (in effect, to autotrack it). This will not work in the caseof a classIIp response,becauseof the polarity reversal.A further problemin estimating AVO gradient is that there are bound to be scaling issuesbetweennear and far offsets. A sphericaldivergencecorrectionwill havebeenappliedduringprocessingof the data to correct for the systematicdecreaseof amplitude with offset due to the increasein length of the travel path. If this correction is inaccurate,then there will be a tendency for all reflectors to show the samechangein amplitude with offset, e.g. a systematic Thereareotherpossiblecausesof suchsystematiceffects:the seismicsource decrease. may emit a stronger signal in the vertical than in an oblique direction, and receiver

133 -

AVoanalvsis

ResidualMoveoutand AVO reflection 0.15 Y= 0.0823x+0.11 0.1

€ 0.0s =

sample y=

0 . 4 5 3 5 x+ 0 . 1 3 0 4 ^ ^

Fig,5.11 Tracestiom a CDPgalher(ofiiet increasing to rheleft) andmeasured amplitudes. sensitivitycan similarly favour the vefiical adval. This is not necessarilya problem if we are interested ir looking fbr lateral change in AVO response on a particular reflector,but it will confusecomparisonof the measuredresponsewith models based on well data. It may be possible to conect the error by applying a scaling lactor if we understandthe cause well enough. Alternatively. we can normalise the amplitude of a target reflector against that of another reflector (or group of reflectors) of known AVO responsc.One way to do this is to comparethe seismictracedata with well synthetics over a range of incidenceangles.It is possibleto calculatea well syntheticfor any angle of incidence by using the Zoepprttz equations to work out the reflection and transmlssioncoefflcients at every intertace, so long as we have a shearsonic log (or can prediclone using the methodsof section5.5.5).One of the benelltsof elasticinversion. described in chapter 6, is that it fbrces a careful study of such well ties to be made, to determine the wavelet amplitude at different offsets. Where there is no well control. it will be necessaryto rnake some assumption about how amplitudes should on average behave acrossthe offset range. This will depenclon whether we expect to have an even balance of class I and class III responses.or a majority of one or the other. If we expect an even balamce.then the averageamplitude over a series of rellectors in a long gate can be usedto scalethe amplitudeof the targetevent.In general,amplitudescalingis a major source ol uncertainty. It is much easier to use AVO qualitatively, as a tool to look for lateral variation in rellector properties (e.g. to recognise pay zones), thar to use it to rlake quantitative predictions, e.g. of reservoir porosity. Noise on the trace data tends to have quite a large effect on the gradient calculation. A more robust approach to poor data is to use partial stacks.The simplest method is to divide the tracesinto two sets.nears and fars, with equal numbers of tracesin each. The

T

134 -

amPlitudes seismic Interpreting

*:'

: ,'.'.--* SeismicGather

i , -,;.:4

Near Stack

Fig.5.12 Near and far stacksfor a classIII sand Polarity: red :

,

jt..

, , ' ' r| -

Far Stack soft (impedancedecrease

downwards).

nearandfar datacanthenbe stackedandmigrated.The resultingnearandtar volumes can be loadedto an interyretationworkstationand treatedin exactlythe sameway as the full offsetrangedata.Figure5.12 showsan examplelbr the caseof a classIII sand brighteningto the far offsets.The resultis that the far stackshowsa very with maLrked strongeventat the targetlevel.This approachhasseveraladvantages ( I ) The signalto noiseratio of the nearandfar stacksis high comparedwith that of a gather. (2) The horizon interpretationcan be copiedfrom the full to the far and near offset data;if residualmoveout(due to inaccurateNMO conection) is a problem,the to centrethe horizonson the appropriateloop canbe donesemtsmalladjustments automatically.It is thenpossibleto work with amplitudemapsof a givenreflector This removesmuch of the on nearsandfars asa way of extractingAVO response. in gradientestimation,thoughthe problem residualmoveoutproblemencountered will still causea degreeof mis-stackingand so affect the amplitudesof the near andfar displays.Sometimes,in the caseof a classIIp responseor wherecomplex ofreflectorsfrom nearsto fars' it interferencecausesbig changesin the appearance modelling reflectorson thetwo sub-stacks; maybe hardto recognisecorresponding the expectedresponsefrom a well datasetmay be helpful' (3) Scanningthroughthe appropriatesub-stackvolume (especiallythe far tracesfor classIII) is a quick way to look for anomalousamplitudeslt is much fasterthan scanningthroughgathers:thevolumeof datais less,of course'but alsoit is possible to get an immediateimpressionof how amplitudeanomaliesrelateto structure Being ableto view far andneartracesectionstogether(in adjacentwindowson the screen)is a powerfulaid to understanding.

135 -

AVoanalvsis

AngleMutesfrom ProcessedGather

Fig, 5.13 Selectionof daralrom a gatller fbr ct€ation of angle stacks.

(4) Pre-stackmigrationis becomingroutine,andis the bestapproachfor seriousAVO studyasit providessomeguarantee thatthetracesin a gatherall belongto thesamc subsurfacereflectionpoint. Where this has not beendone,migrationof the near ard far sub-stacksis a cheapway of positioningthe stackedamplitudedatain the right placein space. A possibleproblem with near and far offset stacksis that the sub-stacksr.epresenl different incidenceanglerangesat different two-way times. This can make it more diflicult than it need be to compareobserveddata with a modelledresponsefbr a particularincidenceangle.The solution is to produceangle stacks;theseare macle by stackingdatawithin an anglerangecalculatedby ray-tracingor the fbrmula given above.Figure5.13illustrates theprinciple. The presenceof dift'erentfluid fills (brine,oil or gas)affectsthe Rs and G valuesof a reflectorIn general,the progressionfrom brine to oil to gaswill move Rn and G for the reflectorat the top of the reservoirtowardsmorenegativevalues.Someschematic examplesare shownin fig. 5.14.SandA hassmallpositiveR0 andnegativeG for the brine and oil casesand negativeR0 fbr gas.The responsewill be classIl for bdne or oil, andclassIII lbr gas.The expectecl stackedreflectionfrom top sandis also shown (red : impedanceincreasedownwards: positivereflectioncoefficient).For brine

136 -

lniBrpretingseismicamplitudes

StackResponse Brine Oil Gas

G

StackResponse ..-----------^.; o.,n"s"nd 1 Brine Oil Gas StackResponse Brine oil cas sandA

{ > } .

:p

-

SandC

Be Fig,5.14Schematic sandresponses ontheRo-Gcrossplot.

sandthis is a red loop; the reflection coefficient decreaseswith increasing offset, but remainspositive so that the summedresponseover all offsets is positive. The oil sand starts off with a small positive reflection coefncient at zero-offset, and reversessign as it becomesmore negaiive with increasing offset. Summing over all offsets results in partial cancellation of the positive and negative amplitudes, leaving a weak blue loop. The gas casereflection coeff,cient starts off negativeand becomesmore so with increasingoffset,so the summedresponseis a strongblue loop. SandB hasnegative R0 with positiveG, i.e. a classIV AVO response. The stackedtop sandamplitudeis a soft loop that increasesin amplitudefrom brine to oil to gas,owing to the changein R0. Sand C has a positive R0 ard negativegradient, so exhibits classI behaviour.The top sandreflectoris a red loop thatdecreases in amplitudefrom brine to oil to gas,because of the decrease in Rs. A crossplotfor a top sard reflectoris shownin flg. 5.15(a).The brine and oil sand casesform separatehends.Within eachtrend, it is porosity (increasingto the left) that causesthe pointsto be strungout alongthe trendline. The oil trendis systematically displacedawayfrom the brine hend, towardsmorenegativevaluesof R6 and G . If there wasreasonable seismiccoverage(asis likely with 3-D data),thenit would be possible to fit a trend line to the brine sandand use distancefrom this trend as a predictor for pay sand(fig. 5.15(b)).In this example,the pointsarecolour-codedaccordingto TWI which allows us to seethat the points furthest toward the bottom left ($een ellipse) are slightly shallowerthan the points (red ellipse) nearerto the main trend. In this case,the data come from a drilled anticlinal structuretlat containsboth gas and oil. The points in the green ellipse are from the top of the gas reservoir, the points in the red ellipse from the top of the oil reservoir Projecting data onto the perpendicularto the trend line, asshown,is equivalentto making a weighted sum of intercept and gradient.From

137 -

AVoanalysis

(a)

0

I

0.2 0.4 0.6

G

t

I

a

a

rrtp-i E oil

0.8 1

1.2

brine

l.'78!.

i

a

-'1- a Fr' iI

I

1.4 1.6

0.1

0.2

0.3

0.5

R0 (b)

ApplyingRo+O.256G makes poinls all on line the zero 0.256=sinr(30.a.) 1,",

=5]= -' 1,; I

pointsawayfromthe line will be ve or +ve

Thiscan be viewedas a projectron

I

Projectionaxis

B6 Fig.5.15 (a) and (b) tsxrmplesol inlerccpt-llradient cros\Plolsfir-rop sdtd in brineand oil cases.

Shuey'scquation,we sce that this is ecluivalentto creatinga stack at some parliclrlar angleof incidence.An anglc stackfbr a small rangearoundthis valrLewould therefbrc bc the best way of looking ibr pay sand.Var-iousfluid factrn.'discrinin;rntshavebecl.) ploposedalong theselines. but in practiceit is better to makc a crossplotof thc real seisrric data alrd clroosethe best angle stack by fte ca]culationshown in iig. 5.l5(b). Note that in this exanrplc.the slope of lhe trend linc is cluitehigh at about 4. This meansthat an anglestackat around30 would be the bestfluid discr.irr.rinant. Modclling the expectcdtrend basedon well data.consideringtlte et'ttctsof varying porosity and

t

138 -

Interpreting seismic amplitudes

AVOMappingWeightedStackor Equivalent AngleMaps

ru ru ffit It il* il*

EI

!i!=l-

rl.l

:;-

t m EI EI EI EI

tiEr I t r ftII E l f t * r-I I1* eR-l ll' rol I 1 : rcl i---:;-----

L E I

I

aEJIft1

t:

rEql

li EH] E] T;

f- ::;;-.'----l

EGffiI fllxl TffiII K] EE] :6

tl-1

I

It-

_ 1 iIl]* : -r{tl Ii: B EE!I

rel EI EI EI

ffil EI

f

ffi---do

I:

TJ I il* It* []:

rql li l*

Fig.5.16 Amplitude maps at different anglesover the Auger Field in rhe culf of Mexico. The optimum coincidenceof structureand amplitude anomaly is at an angle of l8', which is therefore the optimum fluid stack.Reproducedwith pemission lrom Hendrickson(1999).

shalecontentin the sand,would give much lower values,in the range 1.5 to -3. The slopeof the trend observedin the real seismictracedatais stronglyaffectedby noise.The calculatedAVO gradientis very sensitiveto randomnoiseon the seismic traces,whereasthe interceptis fairly insensitive.Where signal to noiseratio is low (i.e. weak reflectors),the slopeon the crossplotwill be much higher than the wellbasedexpectation.A complementaryapproachto this crossplotanalysisis to makea seriesof amplitudemapsfor differentanglerangesand comparethe resultswith the expectedbehaviourfor fluid effects,e.g. amplitudechangecontbrmingto structural contours(Hendrickson, 1999;seealsofig. 5.16). A different way of visualisingthe interceptand gradientdata is to createa trace volumeof R6 x G values.This is usefulwherepay sandhas a classIII responsebut brine sandis classI. Pay sandwill then havea positiveRe x G valueand brine sand a negativeone.By usinga colourbar that accentuates the positivevalues,pay sandis easilyidentifiedin the R6* G volume(fig.5.17). The presenceof AVO effectscan be a problemfor well-to-seismicties if synthetic seismograms areconstructed for thezero-offsetcase.In principle,well syntheticsshould

139

Bockphysicsfor seismicmodelling

Intercept*Gradient

positive Fig.5.17 Intercept * gradient section. values aredisplayed in redandindicate paysand. be calculatedfor a rangeof incidenceanglesand stackedto give a tracethat can be comparedwith the real seismicdata.In practicethe matchfrom seismicto zero-offset syntheticis oftenreasonable for classI or classIII reflectors.It is with classIIp events thatthe worstproblemsadse,asthe stackedamplitudemay be very smallandperhaps oppositein sign to the zero-offsetresponse.A poor tie to the zero-offsetsyntheticis theninevitable. A complicationfor AVo anarysisis theeffectoftuning. BecausetheTwr difference betweentwo closelyspacedreflectorsin a gatherwill vary with incidenceangle(before NMo conection),it follows that tuning effectsw r vary acrossthe gatherand distort the AVO responsein waysthat are hardto recogniseexceptcloseto a well wherebed thicknesses areknown andtuningcan be modelled. -

5.5

Rockphysics forseismic modeiling To understandobservedamplitudeeffects,we often needto know how rock densities and seismicvelocities(both P and S) are affectedby fluid fill (brine,oil or gas),by porosity,by pressure,by clay content,and so on. This is a large subject.A detailed accountcanbe foundin Mavko er al. (1998),for example,andwe shallonly providea summaryhere.The methodswe shalldescribework bestfor medium_to high_porosity sandstones, andare applicableto carbonates only whenthe pore structureis r.elatively uniform with a poresizevery much smallerthanthe sonicwavelength.problemsarise in the caseof rockshaving low porosityandpermeability;velocitiesrecordedby the sonic log may be differentfrom thoseapplicableto surfaceseismicdata.becauseof dependence of seismicvelocityon frequency(dispersion)_

14O -

Interpretingseismicamplitudes

5.5.1 Fluideffects In general, oil or gas fill will reduce the P velocity significantly compared with the brine case, and for gas the effect is often fully developed at saturations of a few percent(fig.5.18(a)).With increasinggas saturationbeyondthis point, the lowering of density becomesimportanl and the seismic velocity sta.rtsto increase again. The density decreaseslinearly as gassaturationincreases.The combinedeffect on acoustic impedanceis illustratedin fi9. 5.18(b).The impedanceof gassandsdropssharplyfrom the brine casefor gassaturationsof a few percent,andthen decreasesalmost linearly as gassaturationincreases. Thus,low gassaturationsmay give reflectionsbright enough to be confusedwith commerciallysignificantaccumulations. The effectof oil is more linear over the entire saturationrange,with little effect at low oil saturationbut an often strong effect at high saturations.S velocity is only slightly affected by differences in fluid fill, via the effect on density; the S velocity is slightly higher for the oil and gas cases. The effect of fluid fill on P and S velocitiescan be calculatedusing Gassmann's ( 1951) equations.They are applicableat seismicfrequenciesto rocks with intergranular porosity and faidy uniform grain size, and describehow the bulk and shearmodulus o[ a rock arerelatedto rhe fluid fill. Thebulk modulusis a measureof resistance to changein volumeunderappliedstress, andthe shearmodulusis a measureofresistanceto changein shape.P andS velocities are relatedto the bulk modulusK, shearmoduluspr anddensityp by the equations: t' p l --

K++ p

and 1/-

Gassmann's equations assertthatthebulkmodulus(K"u,)of therocksaturated with a fluid of bulk modulusK6 is givenby _ _ j * , : K o , K o K^u - K*,

K^a - Kd

Q\K^a

K

'

where K.u is the bulk modulusof the matrix material,K6 is the bulk modulusof the dry rock frame,andd is the porosify.The analogousrelationfor the shearmodulusis givenby Gassmannas This meansthattheshearmodulusis thesameirrespective of fluid fill. Thisis intuitively reasonable, asall fluidshavezeroshearmodulusandareequallyunableto helpto resist changesin shapeof the rock urideran appliedshess.

141 -

Rockphysicstor seismicmodelling

Velocity vs Compressional Saturation

(a) 11600 11500 i1400

oil

1 13 0 0

vp

'/

1 12 0 0 11100 1 10 0 0 1 09 0 0

as

10800 10700

./

10600 10500 0.3

0.4

0.5

0.6

0.8

0.9

WaterSaturation

Saturationvs Acousticlmpedance

(0.)

8.000 7.800 7.600

AI

I

oi

7.404

I

Gas

7.200

-)-/

7.000 6.800 6.600 0

0.1

0.2

0.3

0.4

0.5

0.6

4.7

0.8

0.9

1

WaterSaturation Fig. 5.18 (a) Calculatedcurvesof P velocity (fts) versuswater saturationfor oil and gas casesin an example sandstone;(b) calculatedcurvesofP impedance(knts g/cmr) ve$us water saturation for oil and gas cases.In real rocks, a water saturationlessthan 0-15 is unlikely.

142

r-I

Inlrerpreting seismic amplitudes At f,rst sight theseequationsare not very useful,becauseto calcuratethe saturated moduli we need to know the moduli for the dry rock frame. It may be possibleto measuredry rock properties dfuectly in the laboratory or predict them from porosity and mineralogy using various theoretical models. Howevel laboratory measurements are not often available,and the theoretical predictions are quite sensitiveto the shape of thepores,which may not be known.This problemcanbe side_stepped ifall we want to do is to calculatethe moduli for someparticularfluid fiU (e.g. gasor oil) when we know the saturatedmoduli for someother fluid fill (e.g. brine). If we have a well with wireline log data, then the initial saturatedmoduli can be calculated liom the p and S sonicanddensitylogs.Sometimes,howevet thelog datawill be of doubtfulquality.A usefulcheckis possiblebecausethe Gassmanncalculationproceeds via calculationof the parametersfor the dry rock frame, andthesecan be compared with what is generally expectedfor the particularrock type.The workflow to do this is asfollows. (1) Calculatethe shearmodulusfrom the measured shearvelocityanddensity.If there is no shearvelocitylog, itcan bepredictedfrom otherlog daraby methodsdescribed below:

(2) Calculatethe saturatedbulk modulus from

r*,:vjo-

4lt'

3

(3) Derive the dry bulk modulus from ra. -

+r-d)-K."

K.ut

E--r-o (4) For QC purposes,calculatethe dry rock poisson ratio from 3Ka - 21.t 21,t1 6Ka The dry bulk modulusandpoisson'sratio can be compared with expectedvalues lbr particulartypesof rock, as explainedshortly.It may be necessaryto edit the datato makesurethat the valuesremainwithin a reasonable range. (5) When the valuesin step(4) are acceptable, calculatethe fluid moduti ibr the new case.Hereandin step(3) abovewe needto beableto estimate themoduliof mixtures offluids, i.e. hydrocarbons andbrine.The modulifor theindividualconstituents can be obtainedby methodsto be consideredbelow,andcombined usingthe equation

t s i

Kn

K-

l-,t* t\.h

whereSwis the watersaturation(decimalfraction),K6 is the bulk modulusof the hydrocarbonandK* is thatof brine.This formulashowswhy evena smallamount

143 -

Rockphysics forseismic modelling of gas(which will havea very low K5) causesa largedecreasein K6, resultingin a largedecrease in Vo. (6) Calculatepn andp6,the fluid andbulk densitiesfor the new case,using pn:p",S*+(1 - S*)ph and p b = p ^ ( 1 - Q )+ d p s . (7) DetermineK"u1using ,

K",t :

(a *

-

' 2

('-#J o , t 6

&-rr*-4;

K ,

(8) Derive V" using

(e)Derive Vofrom v -

K*r I 4tt13

We now addresssomeof the detailedissuesin thesecalculahuns. 5.5.1.1 Calculating fluid parameters Equationsto calculatefluid propertieshave been given by Batzle & Wang (1992). Properties are dependenton pressureand temperature.For brines, the salinity has a significanteffecton thebulkmodulus(high saltconcentration impartsgreaterstiffness). For oils, the propertiesdependon ApI gravity and gas_oilratio. If dataare available from analysisofoil samples,thenit may be possibleto usedirectlymeasuredvaluesof bulkmodulusanddensity.(Note,however,thatfor our calculationwe needtheadiabatic bulk modulus,and reservoirengineersusually measurean isothermalmodulus,so a correctionneedsto be applied.)For gas,the propertiesdependon the specificgravity, which reflectsthe concentrationof moleculesheavierthan methane.oil-case values are always intermediate betweenthe brine and gas cases,but exactly where they fall dependson the oil concerned:low ApI and GOR oils will be closeto the brine case, whereashigh API andGOR oils will be closeto the gascase.Figure5.19 showssome representatrve curvesfor North Seadata.Theinterplay of temperatureand pressure effectsmeansthat the shapesof thesecurvesare not intuitive,and for accuratework the fluid propefiiesneedto be calculatedfrom the Batzle& Wangequations.

144 -

Interpleting seismicamplitudes

FluidProperties (NorthSeaExample) Density(g/cm3) 0.5

1

Fluidmodulus(cPa) 1.5

0 500

500

1000

1000

1500

1500

c 2000

g 2ooo

fi zsoo

t sooo

o 3000

3500

3500

4000

4000

4500

4500

5000

5000

-

38"APtoil Gas

F i g . 5 . 1 9Typicalvaluesof dcnsityard ffuid modulusasa tunctionof depthin the North Sea.

5.5.1.2Calculatingmatrix parameters Elasticmodulianddensitics fol individualrninerals canbetakenfiom published values suchas thoseshownin fig. 5.20.Wherethereis more thanone mineralin a rock then a weightedaveragehas to be calculatedof the parameters. For densitythis is a straightforwardweightingby iiactional conposition.For Lhebulk modulus,thereare possibJe (Mavkoerrr1.. several approaches 1998).Thesimplestis theVoigt-ReussHill average.which calculatestheaverageofthe arithmeticandharmonicweightedaverage values.Thus,for a quartz-claymixturewe wouldca]culate a Voigtbulknrodulus: K":l'iq1'Vu1,*K.1V.1 and a Reussmodulus:

n** and then

"

K,+K' 2

where V is thc volume liaction of the constituentdenotedby the sufflx. For the case above, thc clay volume fraction would be taken as the shale volume fraction der-ived from petrophysicalanalysis.

145 -

Rockphysicslor seismicmodelling

Examplesof SolidMineralElasticParameters

Density n g/cm3, e astic moduli n GPa, veLocjtiesin km/s.

afierSimmons& Wang( 197I). Fig.5.20 Mineralproperties. 5.5.1.3 Invasion effects Unfortunately, the density and sonic logs recordedin the borehole may not record values truly representativeofthe formation. The pressurein the borehole is usually kept higher than the fluid pressurein the formation being drilled. This pressuredifferential forces dfllling fluid into permeable lbrmations, replacing the original fluid in an lnvaded z.one around the borehole. Many different types of drilling mud are in use. Water-based muds may contain water of any salinity from very low to nearly saturated. Forcing this water into an oil-bearingreservoirwill of coursechangeits density and seismic velocity.Similarly,if an oil-basedmud is used,it will changethe propertiesof a brineIillcd reservoir in the invaded zone. The extent of the invaded zone dependson several factors, including the pemeability of the formation and the length of time that the borehole is leti exposed to circulating fluids. The etfect on the sonic log also depends on the source-receiver spacing: the wider the spacing, the rnore likely it is that the tool will measurevaluesin the formation beyond the invadedzone. It is possibleto use resistivity logs to estimate the oil satulation in an jnvaded zone, and then use steps (5) and (6) of section5.5.1 to estimatethe fluid propertiesin the invadedzone, given the oil and water properties. Gassmann substitution can then be used to corect logs back to what they would be if the fomation fluid had not been disturbed. In practice' this may not be a reliable approach, given that the logs may or may not be reading in the invaded zone, depending on the distance that invasion has penetratedaway lrom the borehole. If different fluid zones of the same reservoir have been dlilled' perhaps with various mud types in different wells, then a useful check is possible. Stafting tiom

147 -

Rockphysicsfor seismicmodelling

SandstonePorosityvs CompressionalVelocity 6000 5500 yp 5000

.\o

Han et al. (1986)(40 MPa)dataset

4500 4000 OsebergOtz cemented

3500 3000 2500

OsebergClaycemented

NorthSea (Palaeocene) dataset

Troll20 N/pa Troll5 NIpa

GulfCoastexample

2000

0.2

0.1

0.4 Shallowunconsolidated exampre

Porosity

Fig, 5.21 Sandstonecompressionalvelocity vs porosity.Trends basedon Dvorkin & Nur (1995).

Effectof Clay on CompressionalVelocity 5000 4500 4000 th

3500 E o 3000

I I

a

2500 2000 1500 0.1

o.2

Porosity OutputfromXu-White(1995)model Fig. 5.22 Eff'ectof clay on compressionalvelocity (Xu White model).

147 -

Bockphysicsfo] seismicmodelling

SandstonePorosityvs CompressionalVelocity 6000 5500

yp 5ooo

.\t

Han et al. (1986)(40 MPa)dataset

4500 4000 OsebergQtz cemented

3500

OsebergClaycemented

3000

NorthSea

2500

(Palaeocene) dataset

Troll20 N/pa Troll5 lvlpa

GulfCoastexample

2000

0.1

0.2

0.4 Shallowunconsolidated example

Porosity

Fig.5.21 Sand\tonecompressionalvelocity vs porosity.Trends basedon Dvorkin & Nur (1995).

Effectof Clay on CompressionalVelocity 5000 4500 4000 >c 3500 - q. 3000

I

I

a

2500

f

2000 1500 0.'1

0.2

0.3

0.4

Porosity OutputfromXu-White(1995)model Fig. 5.22 Efiect of clay on compressionalvelocity (Xu-White model).

148

Interpretingseismicamptitudes

Density= sVor Lithology ss/snavg

a 1.741

snae

0.265

sandstone

0.261

limestone

0.225

dolomite

1.74

0.252

anhydrite

2.19

0.16

y is in kn/s,density Fig.5.23P velocity-densiry relarions. in g/cm3. havebeengivenby Castagna(1993),asshownin fig. 5.23.An improvedversionofthe limestonerelationhasbeengivenby Mavko et al. (1999): 386 p : 1.359Vp0 , wherethe densityis in g/cm3andthe velocityis in krr/s. where thereexist somewers with both sonicanddensitydatafor an intervalbeing studied,it is betterto usethemto constructa specificrelationshipratherthanusethese generalequations.

5.5.5 Shear velocity Ideally, shearvelocity should be determinedby direct measurement.However,reliable shearvelocity logging is a fairly recentintroduction,and many wells havep but not S wave logs. It is therefore useful to be able to predict S velocity from p velocity, for useasan input to fluid substitutionandAVO modelling.This is alsoa usefulcheckon measuredshearlogs,which aresometimesof doubtfulquality; shearlog processingis to someextenlinterpretive.Also, sometypesof sonictool arenot ableto recordshear velocitieslower thanthe compressional velocity in the drilling mud. Variousempiricalrelationshavebeenproposed.A useful methodis describedby Greenberg& Castagna(1992). % is predictedfrom Voandmineralfractionsusingfour Vo-% relations: sand % : 0.80416Vp- 0.85588 shale V" :0.769 69Vp- 0.86735 l i m e s t o n e % : - 0 . 0 5 5 0 8 V p+z t . 0 1 6 7 7 V -e 1 . 0 3 0 4 9 dolomite % : 0.5832l yp - 0.07775

't49 -

Rockphysicsfor seismicmodelling

VplVs 2500

2000

.'

| O-1 ..

._ vs

-

.,!

.'1.,' -d,

1500

. ' . '

1000

47

500

0 1500

%

:

-

.. 0.4

o-45

-

Castagnas sandline mudline Castagnas Quartzline shales OsebergOtzcement OsebergClayCemenl Trolt5 MPa Troll10MPa Troll20 lvlPa Linesof equalPoisson Ratioare alsoshown

: 0.49 :

2500

4000

V^ Fig. 5.24 Schematicplots of shearagainstcompressionalvelocity (r/s) using various empirical relationsfor a mnge oflithologies. Lines of constantPoisson'sratio are also shown.

wherethe velocitiesare in km,/s.The shearvelocity for the rock is calculatedas the averageof the weighted arithmetic and harmonic averages:

v*ir,: I

%,wi

il

and

ll

Yr",.:J-

l

il where the summationis acrossthe four possibleconstituents,each with a volume fractionw andshearvelocity %. Sometypical schematicplots of % againstyp are shownin fig. 5.24. The quartz line, V" : 0.8029Ve- 0.7509, was derived by one of the authors(R. W. S.) from the work of Murphy et al. (1993) and other data and predictshigher % than the Castagna(1993)relations.With increasingvelocity (in general,decreasingporosity), Poisson'sratio decreases.For Vo below about 2500 m/s, none of these relations is applicable;they tend to predict too high a shear velocity, perhapsby a wide margrn.

I i

150 -

lfierpreting seismic amplittdes

5.5.6 Dryrockmoduli As we sawabove,dry rock moduli arecalculatedasa stepin perfomring Gassmannfluid substitution. Comparing them with expectedvalues is a useful check on the accuracy of the input data. Dry rock moduli vary systematicallywith porosity. Murphy et al. (1993) deivedthe following empirical relations for clean sandsfrom laboratory measurements: ,(a : 38.18(1- 3.390+ 1.95A2) tt = 42.65(1- 3.484+2.196\, where @is the fractional porosity and the moduli are in Gpa. A differen! approach,though leading to a similar outcome, is the critical poros_ ity model (Nur er a/., i998). In this model, dry rock elastic moduli vary linearly from the mineral value at zero porosity to zero at the critical porosity, where the rock beginsto behaveas though it were a suspension(fig. 5.25). The value of the critical porosity depends on grain sorting and angularity, but for sands is usually in the range 3540Vo. The predictionscan be unreliableat high porosities,as the critical value is approached.Figures 5.26 and 5.27 show some examplesof plots of elasticmoduli againstporosityfor real rocks. There are variationsin the moduli of real sandstones due to the presenceof clay and microcracks.Many points plot well below the model predictions,which are only really applicableto consolidated sands.Dry rock Poisson'sratios are generallyin the range 0.1-0.25 for sandstones

(fie.s.28).

For carbonates,the picture is more complicated. Chalks are amenableto the same generalapproachassandstones,but in most carbonatesthe moduli are stronqlv affected by pore shape.Marion & Jizba(1997)give someexamples.

The CriticalPorosityModel

,*=r"l-fl

.^"=,ql_r) poros[y I

u

porosity 0c

Fig.5.25 Theconceptof cdticalporosiry.

151

Ass€sslngsignificanco

Porosityvs K6 45 40

-

Murphyof a/.(1993) Kd = 38.18(1 - 3.390+ 1.95O2')

t d * 25 20

NorthSeaTertiaryTrend

15 10

GultCoastexamDle Troll

5 0

o.2 Porosity

Shallowunconsolidated example

Fig,5.26 Plot of obseweddry bulk rnodulusagainstporosity for somereal sandstoneexarnDles. Brokenblack line = empiricallower boundfor thesedata.

Porosityvsp 45 40 30 'u 2 5

-

Murphyef at (1993) p = 42.65(1- 3.480 + 2.i9 O2l NorthSeaTertiaryTrend

20

Oseberg 10 c

unconsolidated 0.4 exampE

0

Porosity

GulfCoastexample

Fig,5.27 Plot of obsewedshearmodulusagainstporosity for somereal sandstoneexamDles. Brokenblack lne = empfuicallower boundfor thesedata,

5.6 Assessingsignificance In all but the simplest cases, assessingthe significance of an amplitude anomaly (or the lack of one where it might be expected) is difficult. When investigating an

152 -

amplitudes Interpreting seismic

Porosityvs Dry RockPoissonRatio 0.5 0.45 0.4 0.35

od

0.3

Ct

0.25 0.2 0.15 0.1 0.05

. : b

a

o o-'ot

a

'r

!rtfrr' I

o a

0

i

#

a t

a

a

0.3

0.1

0.4

Porosity Fig. 5.28 Dry rock Poissonratio versusporosity tbr some real sandstoneexamples.

amplitudechange,or a flat eventthat may be a DHI, the following checklistmay be useful. to showthe effectsof lithology/fluid fi11on seismicamplitude? Are thereanalogues Are thereenoughgoodquality well datato cary out modellingstudies? Are the well datafrom a closelyanalogouscase(e.g.sedimentaryenvironment)? Is well dataquality goodenough(e.g.correctionof logs for invasioneffects)? Is the amplitudechangepredictedfrom the modelbig enoughto be detectable? enoughscenarios,includingnon-paycases? Has modellingaddressed greaterthan the effectof likely porositychanges? in fluid effect of change Is the Is therea goodwell tie to establishseismicphaseandpolarity? Alternatively,canphaseandpolaritybe deducedfrom knownisolatedreflectors? Are horizonpicks uncertain? to preseryeamplitudes? Has seismicbeenprocessed In particular,hasshort-gate(< 1000ms) AGC beenavoided'l Doesthe seismichavegoodsignal/noiseandimaging? Are multiplesa problem? Do overburdencomplicationsaffectseismicpropagation? Are the pre-stackdatagoodenoughto useAVO methods? If thereis a flat spot,doesit havethe right polarity? reflectors? Is the flat spotdiscordantwith bedding-plane spoUdim spot? Is therea bright Is the degreeof brightening/dimmingasexpectedfor a fluid effect? Are severalfeaturesconcordantin indicatinghydrocarbons?

153 -

References Is there adequatedata coverageto map the features? Are the featuresconsistentfrom line to line? Do the featuresshow conformancewith depth structue? Are the effects consistentwith analogues? Fluid effects are sometimes more visible on seismic than Gassmannmodelling suggestswe havea right to expecti for this reason,we should always look for structureconforming amplitude changein the real seismicdata, evenwhere modelling saysthat the effects may be too small to see.It is sometimespossible to see the effect of hy_ drocarbonsindirectly, becausethey tend to preserveporosity during diagenesis.The biggest problem in understandingamplitude anomaliesis the need to bear in mind a rich enoughsetofpossible explanations.In particurar,it is hard to anticipateeffects due to lateral lithological changedifferent from any recordedin existing wells in the area; generationof a rangeof geological models,basedon analogues,is the key to exploring the possibilities.

--

References Aki, K. & Richards,P.G. (1980).QuantitativeSeiwtotogy.W.H.FreemanandCo. Batzle,M. L. & Wang,Z. (1992).Seismicpropertiesof porefluids. Geophysics,sj, 139640g. Campbell,S. J. & Gravdal,N. (1995).The predictionof high porosity chalksin the EastHod field. PetroleumGeoscience.l. 57J O. Castagna,J. P.(1993).Rock physics- the link betweenrcck propertiesandAVO response.In: Offset dependentrcflectivity - theoryandpEctice ofAVO analysis(eds.J. p. Castagna& M. M. Backus), In estigationsin Geopl,),.'rcr,E. Societyof ExplorationGeophysicists. Castagna, J. P & Swan,H. W. (1997).principlesof AVO crossplotting. The LeadingEdge, 16, 33742. Dobrin, M. B. (1976). Inttuduction to Geophtsicalpruspecting.Mccraw-Hill. Dvorkin,J. & Nur, A. ( 1995).Elasticityof highporositysandstones: theoryfor two North Seadatasets. Geophysics, 61, 1363J 0. Gardner,G. H. F., Gardner,L. W. & Gregory A. R. (1974). Formationvelocity and density: the diagnosticbasicsfor stratignphic traps.G€apiysics,39, 7?0-80. Gassmann, R ( 1951).Elasticwavesthrougha packingof spheres. G eophysics,16,673-g5. Greenberg,M. L. & Castagna,J. P (1992). Shearwave velocity estimationin porcus rocks: theo_ retical formulation,preliminary verificationandapplications.Gaopttysicalpmspecting,40, 195_ 209. Gregory A. R. (1977).Aspectsofrock physicsliom laboratoryandlog dataareimportantto seismic interpretation.In: Seismicstratigmphy- applicationsto hydrcca6ol explo'llliorr-MpG Memoir 26 (ed.C. E. Payron). Han, D.-H., Nur, A. & Moryan, D. (1986).Effectsof porosity and clay contenton wavevelocity in sandstones. c"opr.}Jr'cs,51, 2093-107. Hendrickson,J. ( 1999).Stacked.Geophysicalprospeaing,47, 663JO6. Marion, D., Nur, A., Yin, H. & Han, D. (1992). Compressionalvelocity andporosity of sand_clay mixtures.Geopiysics,ST,554-63.

il

ll

1g -

InteDrclingscismicamplitudes Marior! D. & lizba" D. (1997). Acoustic propertiesof carbonaterocks. In: Carbonateseismology (eds.I. Palaz& K. J. Matkurt), Geophys.Dev.Series,6. Societyof ExplorationGeophysicists. Mavko, G., Mukerji, T. & Dvorkin, J. (1998).TheHandbookof RockPrysicr. CambridgeUniversity Press. Murphy, W., Reischer,A, & Hsu, K, (1993). Modulus decompositionof compressionaland shear velocitiesin sandbodies.Geophrsics,5E,22719. Nur, A., Mavko, G., Dvorkin, J. & Galnudi, D. (1998).Critical porosity: a key to relating physical propertiesto porosityb, rocks.TheLeadingEdge,11, 35742. Ross,C. P & Kinman,D. L. ( 1995).Nonbright-spotAVO: two examples.Geophysics,6O, 1398-1408. Rutherford, S. R. & Willians, R. H. (1989). Amplitude versusoffset variations in gas sands. Geophysics,54,68O-8. Shuey,R. T. (1985).A simplificationof the Zoepgiz equations.Geophysics,5O,6W14. Simmons, G. & Wang, H. (1971). Single Crystal Elastic Cowtants and CalculatedAggregate Propeties.Michtgan Inst. of Tech,hess, Cambddge,Mass. WyIie, M. R. J., Gregory A. R. & Gardner,G. H, F. (1958).An experirneDtal investigationof factors affectingelasticwavevelocitiesin porousmedia.Geophysics,28,459-93. Xu, S.& White, R. E, ( 1995).A newvelocitymodelfor clay-sandmixitres. GeophysicalProspecting, 43.91-118. Tnewnt4 K. (1919').I.ll,errcflexion und durchgangseismischerWellendurchUnstetigkerlsflaschen. Uber ErdbebenwellenVIIB, Nachrichtender Koniglichen GesellschaJtder Wissenschalten zu Gottinge4 Math. Phys.,K'., 57-84 -

I

:1,

i;r ll

J

I

6 Inversion

The fundamentalideaof inversionto seismicimpedanceis very simple.A reflectivity seismicsectioncontainsreflectionsthat can be studiedby the methodsdiscussedin chapters3-5. Thesereflectionsshow wherethereare changesin acousticimpedance in the subsurface. Inversionis the processof constructingfrom this reflectivitydataset a sectionthat displaysthe acousticimpedancevariationin the subsurfacedirectly.As we shallsee,this oftenmakesit easierto interpretthe datain geologicalterms,because it fbcuses attention on layers and lateral variations within them, rather than on the propertiesof the interfacesbetweenlayersthat causethe seismicreflections.This is an idea that has beenknown for many years(see,for example,Lindseth, 1979)but has not been usedvery much until recently,probablybecausegood resultsrequire input reflectivitydataof excellentquality; the availabilityof modern3-D datasetshas triggeredat upsurgeof interestin the technique. This chapterbeginswith a summaryof the principles,then discussessomeprac_ tical processingworkflows with particular attention to the issuesthat are critical for the quality of the results,continueswith somepracticalexamplesto demonstratethe benentsof thetechnique,andconcludeswith a summaryof somespecialised advanced applications. -

6.1

Principles A simplemodelfor zero-offsetseismicresponseis illustratedin fig. 6.1.The subsurface is represented asa numberoflayers,eachwith its own acousticimpedanceA; the zerooffset reflection coefficient at the interfacebetweenthe /ith andthe (r + 1)th is given by R,1: (An+l _ A.)l@,t

+ A,)

or for a smallchangein impedance64, then R :6A/2A:

0.56(lnA).

The reflectioncoefficientseriesis convolvedwith the seismicwaveletto give the seismictrace.Inversionaims to startfrom the seismictrace,removethe effect of the

r55

l)

156 -

lnversion \,elocily O.Nlty

Al R.n.ction cefi

)

.

2

Lr \f

41 {

(

h i l , ' \

Fig.6,1 Principle oftheinversion method.

:

i I

wavelet to get back to the reflection coefficient series,and from them derive the layer impedances.It has to assumethal the starting seismic data are free from corelated noise(e.g.multiples).Also, the waveletpresentin the datahasto be estimated;many inversion methods derive the wavelet from a well tie and have to assumethat it does not changelaterally away from the well. There is also an amplitude calibration to be takeninto account real seismictracesare not directly outputas reflectioncoefficient values,but are scaledto give someconvenientbut arbitrary rms averageover a trace.At leastovera limitedtime-gate,theratio ofreflectioncoefficientto traceamplitudehasto be constantif inversionis going to work. Careis neededduring seismicprocessingto avoidstepsthatmight introduceartificialamplitudechangesverticallyor horizontally. However,locally variableeffectsin the overburden(e.g. shallowgas)can reducethe seismicenergypenetratingto deeperreflectorsand so reducethe reflectedsignal;left to itself, the inversion processwould try to interpret this as a decreasein impedance contrastacrossthe deeperinterfaces.To remove such aflefacts, a long-gate AGC may be applied,which scalesamplitudesso as to removelateralvariationwhen averaged over a TWT intervalof 1 s, for example. Anotherissueis that the seismictracescontaindata of limited bandwidth;the frequenciespresentdependon the rock propertiesandthe seismicacquisitiontechnique, but might be in the range from 5 to 50 Hz. This meansthat the low frequenciesin particular, which are critical for the estimation of absolute impedancevalues, cannot

157 -

Procedures be obtainedfrom the seismic data, but haveto be addedfrom elsewhere.usuallv from a model basedon well data and geological knowledge.

-

6.2

Procedures

6.2.1 SAIIlogs A very simple approach to inversion was described by Waters (1978). This is the Seismic Approximate knpedance Log (SAIL), which is derived as follows. Instead of the discretereflection coefncientsof section 6.1, we define a piecewisecontinuous function of reflection time I which is called the rcflectivify:

r(t) :

6R/6,: O.sd(nA)/dt #To

from which we seethat t

A(t\

|

:2 | r1t1dr tn-j:...............i:............... A(U, J

so that / t

r t

\

l

A ( i ): A ( o ) e xl zp I r u n t | . \ *

)

What we actuallyhaveis a seismicsignals(r). If we cal assumethat the waveletis zero-phaseandthe dataarenoise-free,then the signal s is a band-limited representation of r(r), related to it by an unknown scaling factor, so we can write / t

r r

\

l

A ( i 1: a 1 6 1 . * O l a / s { r ) d rl , \ * ) where a takes accountof this unknown scaling. Assuming that the exponentis small, as is likely in practice, the exponentialterm can be expandedasa seriesand truncated, to give the approximaterelation:

l

'

"

1

A(t): A(o\{ t +a / srrtdrJ | | " |

t

;

l

or

i s(r)dr.

( A ( t ) - A ( U \ / A ( 0 ): d , J

aO

15B ry

Inversion

ivhich rlkl,r's us to cliculalc lilctiolul changc in impcclirnccas a fllnction of timc by irltcgriilingfcflcctivity fionl r startiugtirue (r at rvhich we knor,,'lhe impctlanccli(0): l h c L l n k r o w ns c l l i n g c o n s l n n lp r c | c n l s u s 1 r - o r tnu n i n g t h i s i n k ) t r s o l u l cv i t l u c so 1 ' irnpetlirncT e .h c S A I L l f i , l c cr c s u l l su i l l c l c u l r b c l i n i t c d t o t h e s e i s m i cb ; i r c l w i t l t l t . IsoliLtc(l iDtcfl ccs \!hclc llrcrc is l lirrgciinpcclanccchange.suchls thc top ol a nrtssivc clilbonuleu ithin a clolniullntlyshalesequerce.u'ill shou rrpaslhe blnd-lirnitedvetsiolr thatthe seismic ol u rlep lurul ior. \i'hich ir a bipolal loop. ;\ linrilalion is thc assur)rplior) w i l v e l c li s z c r ' op h r r s ci :1 ' l L t r l l p o s s i b l e . t h i s n c c t l s t o b c c l l c c k L - d b v r n e aanwseolfl t i e . A larticularlv uselLrlfblmLrLitionol this rpplorch car be ilcrivcrl Lrythinkiug ol the fcflcctivitl tlilcc in thc Fouriel clomain.i.c. lrs lhe sLlmrraliol ol'a nutnbel of sitre rvuvcsol clillcrcnl lrcqucncic-s lrnclplllrscs.lrtc!{r'tlionol an inclividu l (onrlirnenl \inc uave ol iequelcv r')givesa siuewlve u'hosephasehasbeenshiltcd by 90 ard wlrose l L r n p l i l u d hc l s b c c n r r r u l t i p l i c d{ r 1 'l / o . I n l c g l a t i o no 1 ' t h c c n t i l e t f a c e f e q u i r c sl h n l this lfocess shoulclbe applied to ever',r, ilclividinl sine u'uvc conll:)onelt.Theletirte. llrc scisnrictrrce can bc intcgtatcd by applying u phuseshili ol 90 and r lligh-cut ' ctuve. f i l t e r t h l l i l a l l s o l ' l ' r s 1 / r ai l c r . o stsh e e n l i l e s e i s n r i cb a n d w i r l t hi.. c . a t 6 d B p e r o Nliirt inlcrl)rl]1alion solj\varcsLrilcscolrliiirllhe lirnctionllitl'to p|i)ccssscisnlicdiila in t h i s \ \ ' r L )s' .o S A l l - l l a c c sc u n b e e a s i l l ' g e n c f a l c $ ( li t h o u l l i l c n c c d l b r a n y a c l d i t i o n a l pnrccssingsoftu llc. Figule 6.2 is ar erurnpleof \\,hatcan be achievetlby this neans. Il is a scisrnicline ircrosslu iirel of the Central North Sel. offshole UK. The holizon pickcclin ycllorv

\-

- r'rf' ' r '

.t,'tl

l

tar

dFq. ,e\..-,r'-r!rt

\

*

'tiaL*|

*ti*tr

f

\ 1,. \

Fig.6.2

B l n r l l i n r i t ( - d i I l \ r - f s i o n . c s L r l t .p . o c l u c c ( lr t t h c w o f k s l i l l i o n L r s i n sl h e S A I I - n r L - I h o d o l o g y

't59 -

Procedures marksthetop ofa particularsffatigraphicinterval.The SAIL sectionis displayedwith a colour bar suchthat hard intervals are red and soft intervals are violet-purple; the same conventionwill be usedfor otherimpedancesectionsin this chapter.patchesof soft coloursbelowtheyellow markershowthelocationof softoil-filled sands.Theseareactuallychannelscrossingtheplaneof section,aswouldbeclearfrom a mapview;notethe smallbumpsin theyellowhorizonabovethe softpatches,which areformedbecause the sandychannel-fillcompactslessthantheshaleson eithersideasit is buried(dffirential compaction,which often resultsin sandyintervalsbeing markedby moundedtopography). Such a display can be usedin a qualitativeway, to map out the extentof a hydrocarbonaccumulation.It would be less straightforward to try to use the results to makea quantitativepredictionof, for example,reservoirporosity,becausethe traces showonly relativeimpedance.Making a plot of well impedancedata,with bandwidth limited to the seismicspectrum,is a usefulaid to understanding how the SAIL results relateto the real earth.They often tum out to be surprisinglyuseful for sucha simple technique,becausein many casesreservoirsare not thick enoughfor the missing low-frequencycomponentto be critical. A reflnement of this approachis to try to match the spectrumof the SAIL data to the spectrumof impedancein the real earth. If we computereflectivity from well log data,we usually find that the earthreflectivity spectrumis not flat over the seismic bandwidth,asis commonlyassumedin seismicprocessing;instead,it hasmoreenergy at higher frequencies.By analogywith the optical light spectrum,we can say that the earthreflectivity is not white (i.e. flat) but blue (i.e. higher amplitudeat higher frequencies).A first approximationis often that the well log reflectivityhas a slope of 3 dB per octaveacrossthe seismicspectrum.This can be allowedfor by high-cut filtering the 90' phase-rotated traceat 3 dB per octaveinsteadof 6 dB. A more exact approachis to determinethe actual reflectivity spectrumfrom a well log (Lancaster& Whitcombe,2000).

6.2.2 Extending thebandwidth To get more informationinto a seismicsectionthan is actuallypresentin the seismic ffaces, extra data have to be obtained from elsewhere;different algorithms differ in detail, but they all havethe following generalfeatures. (1) Begin with a modelthat describesthe subsurface; explicitly or implicitly, this will containa numberof layersof differentacousticimpedance. (2) Calculatethe seismicresponsefrom this model,using the waveleipresentin the seismicdataset. (3) Comparethe calculatedseismicresponsewith the real data. (4) Modify themodelsoasto reducethemisfit betweenthecalculatedandrealseismic, perhapsiteratively and perhapsincorporating constraintson the impedancevalues that may be assignedto particularlayers,on the complexityof the model,and on

)

160 -

Inversion the variationoflayer parameters tiom oneseismictraceto the next alonga seismic line. (5) Perhaps,addlow-fiequencyinformationobtainedfrom a modelbasedon geological data. The additionalinformationthathasbeentakeninto accountis thus: (a) thewavelet:removalofits effectsis equivalentto a deconvolution, extendingbandwidth at the high-frequencyend; (b) the model:the geologicalinput extendsbandwidthat the low-fiequencyend. To makeall thismoreconcrete,itis helpfulto work throughanactualprocessingflow. A commonapproachto building a subsurfacemodel is to split it up into macrolayers, probably severalhundredsof ms thick, boundedby the main semi-regionalseismic markers,and consistingof a singlebroadlithology.This makesit easierto construct a geologicalmodel to constrainimpedancevariationwithin a macrolayerInside the macrolayel the subsurfaceis represented by meansof a seriesof reflectivityspikes. Thesespikes,whenconvolvedwith the wavelet,shouldreproducetheobseruedseismic trace,and integrationof the reflectivitywill give the impedancevariationwithin the macrolayer'. To preventthe softwarefrom trying to reproduceall the noisepresentin the seismicsection,it is usualto imposea requirementthatthe reflectivityspikeseries is as simpleaspossible,eitherin the senseof usinga small total numberof spikesor usingspikesof small total absoluteamplitude;the algorithmwill tradeoff the mislit betweenreal andcalculatedseismicsectionagainstthe complexityof the spikemodel, usuallyunderuserconnol of theacceptable degreeof misfit.The resultis usuallycalled 'sparse a spike' representation of the subsurface. It is obviouslycritical to the successof this processthat we know the waveletac curately.This is usually obtainedfrom a well-tie siudy.As discussedin chapter3, well syntheticor VSP informationwill tell us how zero-phaseseismicought to look acrossthewell; comparisonwith therealdatatellsuswhatwaveletis present.Figure6.3 showsanexampledisplayfrom sucha study.To theleft, in red,is thecandidatewavelet, which in this caseis closeto beingsymmetricalzero-phase; to theright is a panelof six (identical)tracesshowingtheresultofconvolvingthis waveletwith the well reflectivity sequence derivedfrom log data.Theyshouldbe comparedwith thepanelof tracesin the middleofthe ligure,which arethe realseismictracesaroundthewell location.Various geologicallysignilicantmarkersare also shown.There is clearly a very good match betweenthe real and syntheticdata so far as the principalreflectionsare concemed, so it is possibleto haveconfidencethat the waveletshownis indeedthat presentin the data.Therearealso somedifferencesin detail,which will limit the accuracyof an inversionresult.Thesemay arisefrom impedectionsin the seismicprocessing, perhaps the presence of residualmultiplesor minor imagingproblems.With luck, the inversion processwill leavesomeof this low-energynoiseout of the invertedimage,becauseof the sparseness of the spikereflectivityseries.Anotherpossiblesourceof mismatchis AVO, which will be discussedlater in the chapter.

'161 -

Procedures

Seis

EMM

ssnt,llrHM

l\r9$lf, 1{(t I

1069 I

?3!t(i 10ffr70

r

l

l

/trn}A{.nrditor $6t_v.ll_ln nlrlt2t! 75S 7s? ?5t?6(r l06t 1(r{1t t06t7!

l

l

r

Fig.6.3 Wirvclet.rcai seismictt-tcesrnd svnthelictracesderivedusingwlvclel.

>

l

l

162 -

lnvelsion

In many implementationsthere are further pammetersto be adjustedto arrive at an optimal sparsespike inversionproduct.For example,it may be possibleto set constraintson the impedancevalueswithin eachof themacrolayers, basedon the well data andgeologicalknowledgeof lateralvariability.Theremay be a parameterthat setsthe sparseness of the series,determiningthe degreeto which the invertedproducttries to follow all the detailof the seismictraces.As discussedabove,thereis a waveletscaling parameterthat affectsthe integrationof reflectivityto yield impedance.This is often a crucialparameter;in effect,it determineshow muchofthe final invertedsectioncomes from the low-frequencymodelandhow much is derivedfrom the seismictracesthemselves.In principleit canbe determinedfrom a well tie, but sometimestherearequite large variations in scaling from one well to another,perhapsbecauseof variation in absorption.A usefulapproachis to look at theresultsofvariousplausiblechoiceson a trial pieceof seismicsection,comparingtheresultswith well dataandgeologicalknowledge of lateralvariabilityof differentlithologies.This is an interpretiveprocess,ard the solution will probably be appropriateonly over a limited rangehorizontally andvertically. Differentstratigraphiclevelsmay be bestimagedon differentinverteddatasets. Often,the mostdifficult part of the inversionprocessis ihe constructionof the lowfrequency model. This dependson information other than the seismic trace data, very often on well information. One approach is to make a model by interpolating well impedancevalueswithin the macrolayers(fig. 6.4).A low-frequencybandJimitedversionofthis modelis addedto thebandlimited productfrom theseismictraceinversion, using the spectrum of the exffacted wavelet to decide exactly what frequency range shouldbe takenfrom the model;this might typically be 0 5 Hz. Wherethereis good

Fig. 6,4

Low frequency model from interpolated well data.

16il -

Procedures well control, this processis straightforward; where there are few wells, there is obviously a dangerof imposing an incorrect low-frequency trend on the final result. This is particularly true ifthe wells aredrilled in updip locations andthe targetfor the inversion is downdip, with a greaterdepth of burial causing additional compaction and perhaps containingadditional stratigraphicintervalsnot sampledat all by the well conffol. It may be possibleto allow for compactioneffectsby putting in trendsofincreasing impedance with depth, but thick intervals with no well control are bound to causeproblems. A last resort if there is no well control is to createthe low-frequency model from seismic velocities (stacking or migration). The snagwith this, apart from the generalissuesof the unreliability of seismicvelocities as a measureof real rock propertiesdiscussedin chapter3, is that densityhasto be inferred from velocity in order to calculateimpedance; to do this, it is necessaryto know the lithology, which may be just the point at issue in an undrilled section. A section resulting from this process is shown in fig. 6.5. It is created from the samereflectivity data as were used to createfig. 6.2. The effect of the low-frequency information is apparentin the progressionfrom low impedance(purple) at the top of the seciion to high impedance(yelloVwhite) at the base.The high-impedanceprecursor (geen/red) to the target horizon (i.e. abovethe yellow marker) visible in fig. 6.2 is not a feature of fig. 6.5; probably it was an artefact produced by minor departtre of the wavelet from the zero-phaseassumedin the SAIL processing.Howevet the definition of the pay sandis not much different from that in fig. 6.2; in this particular casetlp simple SAIL approach is adequate,partly becausethe reflectivity data are close to zero-phaseand partly becausethe target layer is the right thicknessto havea response within the seismicbandwidth.

Fig. 6.5

Ful-bandwidth inye$ion oyer same section as fig. 6.2.

'165 -

Benelibotinversion

Fig.6,6 Section fromaninverted 3-Ddataset. planning development,for examplein choosinglocationsfor producersand waterinjectionwells. Anotherexampleis shownin fig. 6.8.In this casethe reservoiris the UpperJurassic Magnus sand of the UK Northern North Sea.The figure shows the localion of a calibration well and the picked top and base of the reservoir sequence.In this case the reservoirunit containsinterbeddedsandsard shales.The upperpart of the figure shows an inverted section from the 3-D volume. The sandsarc hard relative to the shales, and the application of an impedancecutoff allows separationof sandsfrom shales;in the lower part of the figure this has been applied to separatesands (shown in grey) from shales(in pink). The next stepwould be to estimateporosityin the sands;in this particular study a stochasticapproachwas used, and the results will be shown after explaining the underlying principles in the next section. Inversionis also a useful tool for the study of carbonatereservoirs.For example, SIory et al. (2000) were able to use an inverted datasetto map zones of high and low porosity within a producingreef. Becauseof its shallow depth (3800 ft), high seismicfrequenciescould be recorded,and a resolutionof 14 ft was achievedin the inverted dataset.Mapping of low-porosity, low-permeability tight zoneswas critical to ar understandingof reservoir behaviour durirg production; the tight zonestumed out

166 -

Invelsion EstimatedrelationshiDs between porosityandacousticimledanceof rhe Fulmar sandi

rggrrdxriontl -1

'*tiid F: b^lTry1r_;

omax - -5e-6' AIMP+63.0 @med= -5e-6*AIMP+61.5 Omin= -5e-6*AIMP+ 59.0

n

tu

207a

I

15E

'oT

IOVo I

5000000

7000000

l

a

5fiH'*',,

=s

9000000

3'i.;.. ^"'

000000

13000000

Acoustic Impedance Fig.6.7 Plotof acousticimpedance versusporosityin Fulmarsands(UK NorthSea),basedon severalwells. to be discontinuous, making them less effective as baffles to prevent early water influx into producing wells than had been hoped.

6.3.2 Stochasticinversion Seismicinformationcanbe incorporatedinto a reservoirmodelby usinggeostatistical methods;a useful introduction aothis topic is provided by Chamberset al. (2000). The problem in building a reservoir model is that there is usually only a small number of well penetrationswhere reliable data on reservoir propertiescan be obtained directly. To estimatethe propertiesbetweenthe wells,it is necessary to understandtheir spatial variability. As we move away from a well, the valuesmeasuredthere becomeless and less useful as a way of predicting properties,but we need to know how quickly this fall-off in predictivepower happens.Data can be analysedusing variograms,which are plots showing how the dissimilarity of a property increaseswith disaancefrom an observationpoint; they can be createdby comparing properties measuredin pairs of wells at different distancesapart. The range at which the dissimilarity reachesa constanthigh value tells us how far away from a well it is reasonableto extrapolatethe properties measuredthere. If this range is, say, severalkilometres, then interpolation

167 I

Benefibol inversion

SandDresence Fig.6.8 Inverteds€ctionandinfenedsand/shale sepamtion ill Magnussands,UK NorrhSea.

168 -

lnversion beiweenwells may be easyand reliable,whereasif it is only a few hundredmetres thentherewill be problems.Often,the rangeis differentin differentdirections,being for exzurplegreateralong strikeof a depositionalsystem.Knowing the variogram,a mathematicalmethodologycalledkriging allowsus to interpolatebetweenthe sparse well data in a way that honoursthe spatialvariability of the property.In addition, seismicinformationcanbe incorporatedby usinga methodologycalledcokriging.For example,impedancefrom an inverteddatasetmight be usedas an aid to predicting reservoirporosity.Cokrigingwould producea map that would honourthe 'hard' data atthewellsexactly,andwouldinfer valuesbetweenthewellsbasedon a combinationof kriged interpolationbetweenthem andthe 'soft' evidenceprovidedby the impedance data.However,the resultingmap is a smoothmap of most likely values;it doesnot capturethe true variability within the reservoir,becauseas we haveseenin chapter4 the spatialresolutionof a seismicdatasetis limited; inversionwill improveresolution vertically but not horizontally.Unfortunatelyfor the reservoirengineer,the flow of fluids in a real reservoirmay be stronglyinfluencedby the inhomogeneities that have been smoothedout of the interpolatedmodel. Stochasticinversionoffers a way to constructa suiteof models('realisations')of the subsurfacethat are compatiblewith the well andseismicevidenceandretainthe high spatialfrequencies. A way of doing this is outlinedby Haas& Dubrule (1994).Supposethat we have a reservoirvolumefor which we havea reflectivityseismicdataset,and a numberof well penetrations(fig. 6.9).At eachof four wells in the example,we havean acoustic impedancetracederivedfrom log data.We startby drawingat randoma seismictrace location,wherewe wish to estimatethe impedancetrace(shownin greenin fig.6.9). A candidateimpedancetraceis simulatedfrom the well data;in effect,impedancevalues are drawn at randomfrom a populationin sucha way that the statisticsof the spatial variabilityofacousticimpedancederivedfrom thewell informationarehonoured.This meansthatthecandidateftacewill containtheshort-wavelength verticalvariationthatis foundin the well logs.From this impedancetracethe reflectionresponseis calculated,

????? 1 ar.+

+

Fig, 6.9 Principle of stochasticinversion.

169 -

Eenelits of inversion using the waveletpresentin the seismicdata.The result is comparedwith the real seismictraceat the chosenlocation.If the fit is not satisfactory, ihe simulationof the impedancetraceis repeatedusing a differentrandomdraw from the population,and the comparisonwith the realdatarepeated.The simulationis repeateduntil eventually an impedancetraceis foundthatgivesa satisfactoryfit to the observedseismic.This is thentreatedasthetrueimpedanceat this location,andis mergedwith the well data.The entire estimationprocessis then repeatedat anotherrandomlydrawn tracelocation, 'well' usingtheaugmented datafrom thepreviousstepto estimateimpedanceat thenew location.Eventually,all the tracelocationswill haveestimatedimpedancetraces.The resultingdatasetis consisientwith the reflectivitydataand alsowith the detailedwell information.The solutionis not, of course,unique;differentfinal impedancevolumes will be foundif therandomorderin which tracesareselecledfor calculationis changed, andifdifferent randomdrawsaremadefrom theimpedancepopulationwhenestimating the traceat any particularpoint. This meansthat it is possibleto producea number of differentreservoirmodels,all of which honourthe seismicand well information; studyof the rangeof possibilitieswill quickly show which featuresof the impedance volumeare well constrained(andthereforefound in all models)and which arepoorly determined(and show high variability from one model to another).Eachimpedance volumecan be usedto estimate3-D porosityor lithology by the methodsconsidered in the previoussection. One simulationproducedby applying this techniqueto the exampleof fig. 6.g is shownin fig. 6.10.The scaleon the left showsporositywithin the sand;shaleunits

Fiq.6.10 Porosity simulation from stochasticinversion.

17O -

lnversion within the reservoir are coloured dark blue. The frequency content is obviously mu-h higher than the standardinversionshown in fig. 6.8. As noted above,this result L. consistentwith the well and seismic data, but there is no guaranteethat it is correcc we would needto examinea whole suite of realisationsto seewhich of the featuresare anl Becauseof this,the techniquetendsto be rathertime-consuming well constrained. expensive,andis generallyregardedas a ratherspecialisedtool.

-

6.4

AVOeffects So far it has been implicitly assumedthat the seismic data can be treated as thougt they were acquired at zero-offset. The discussionof AVO in chapter 5 showsthat in some casesthere are significant changesof amplitude with offset. In particular, with classIII AVO responsesit may well give better results to itvert a far-offset sub-stackThere is then a problem in that we need somehowto modify the acoustic impedance tog at the wells in order to have values that can be comparedwith far-stack inverted 1I amplitudes.It is not clearwhat we meanby acousticimpedanceat non-zero-offset. is possible to calculate the reflectivity at each interface and to integrate it to give an 'impedance'variation,but this is not a propertyof the rock in the sameway that true acousticimpedanceis; it is a functionof the incidenceangle. by Connolly(1999).He introducesthe idea A possibleapproachhasbeensuggested of an elasticimpedancewhich is a function of incidenceangle E(9), such that the reflection coefficient at an inbrface is given by Rn(e): (E"+1- E,\/(E,t

+ E")

in analogy with the equation connectingacousticimpedanceto zero-offsetreflectivity in section6.1.From this, usinga simplificationof the Zoeppritzequations,he obtains E(0): q'Pt o" where a, p and p are the P- and S-wavevelocities and density,respectively,andr, y, z are given by r: v:

( 1+ t a n z 0 ) 6/\.sln_u

z : (1 - 4Ksinz0\, whereK is an averagevalueof (p/al2 over the entirezoneof interest.This equation can be usedto calculate an elastic impedancecurve from well log data which can, in tum, be usedto constrainand calibrateinversionsof sub-stacksin just the sameway acousticimDedanceis usedfor zero-offsetdata.

171 --

Refurences A benefit of inverting angle stacks to elastic impedance is that a wavelet will be estimatedfrom well ties for each angle range independently.This takes care of the shift to lower frequenciesnormary seenon the far traces,but more importantly allows the amplitude of the near and far traces to be correctly matched. This removes one of the chief uncertaintiesin quantitative interpretation of AVO effects. As compared with AVo interpretation on reflectivity data,additional benefitsare the reduction of the effect of tuning andthe ability to distinguish lateral changein a reservoir from changes in the overburden,just as in acousticimpedanceinversion. Various approacheshavebeentaken to combining the elastic impedanceinformation from different anglerangesto yield rock parameters.possibilities include the estimation ofPoisson's ratio or ofcompressionarandshearelasticmoduri, which canbe diagnostic of different lithologies or fluid fill. However,the resultsare sensitiveto noisein the data, and careful comparisonof the predictedrock parameterswith wer data is an essential QC step.

--

References Buiting,J. J. M. & Bacon,M. (1999).Seismicinvelsionasa vehiclefor integationof geophysical, geologicalandpetrophysicalinformationfor reservohcharacterisation: someNorth seaexamples. b PetroleumGeologyof NorthwestEurope:proceedingsof the 5th Conferefice(eds.A. I. Flert & S. A. R. Boldy).ceologicatSocieryLondon. Chambers,R. L., Yarus,J. M. & Hird, K. B. (2000).petroleumgeostatisticsfor nongeostatisticians. TheLeading Edge,19,474-9. Connolly,P (1999).Elasticimpedance. TlreI eadingEdge,18,438_52. Haas,A. & Dubrule,o. (1994).Geostatisticalinversio''- a sequentialmethodof stochasticreservoir modellingconstrained by seismicdata.FiAr Break,12.5614. Lancaster,S. & lvhitcombe, D. (2000). Fast-track'.colowed" inversion.Expandedabstract, SEG AnnuaI M eeting, Calgary. Lindseth,R. O. (1979).Syntheticsoniclogs a processfor sfatigraphic jnterprctation.Geophysics, 44,3-26. Story C., Peng,P, Heubeck,C., Sullivar,C. & Lin, J. D. (2000).Liuhua l1_l Field,South China Sea:a shallowcarbonatercservoirdevelopedusing ultra-high resolution3-D seismic,inversion, andattribute-based reservoirmodelling.?, e LeadingEdge,19, g34 44. Waters,K. H. (1978).RefectionSeismology.JohnWiley, New york.

I

datavisualisation 7 3-Dseismic

Recently there has been a change in the way that interpreters view seismic data. The fiaditional method of working, as explained in chapter 3, has been to make different 2-D sectionsthrough the 3-D data volume, as inlines, crosslines,random tracks or time slices. The only way to view more than one section at a time was to open multiple windows and view each one in a separatedisplay. Today, largely thanks to relatively low-cost computer power and memory, it is possible to view entire datasetsso that the interpreter can quickly get a feel for tle actual 3-D nature of the trap. Indeed, several different data volumes can be viewed simultaneously to intenogate various attribute volumes at the sametime (fig.7.1). This has many applications.As shown in fig.7.1, reflectivity and coherencevolumes can be viewed simultaneously when interpreting a fault; this is a way to combine the lateral continuity information from coherency with the identification of the nature of a feature in the standard reflectivity section. Such technology cauralso be used to view different AVO volumes in the same display, or to examine reflectivity and acoustic impedance (inversion output) volumes at the same time. Different timelapse seismic volumes (see chapter g) can be displayed so that production-related changescan be more easily seen. Also, various types ofdata can be viewed together with the seismic traces.In fig.7.2 we see the top reservoir map viewed together with well trajectories and overlain with satellite imagery data of the surface geology. Figure 7.3 is a more traditional combina_ tion: the top reservoir horizon together with the 3-D seismic volume from which it was picked. The ability rapidly to scan through the entire seismic data volume together with the reservoir pick is a very quick way of understanding where the picks may require adjusting. For this type of work, display speedis critical. To check the consistency of a pick acrossa volume of seismic data, the inter?reter needs to keep in his head a picture of how the pick looked on the previous view while examining a new one. If it takes several seconds to redraw the display to show a new line, it is hard to maintain this mental picture; the ideal is to have the display change to follow cursor movement on the screenfast enough that the delay causedby the redrawing is not perceptible. One of the advantagesofhaving the entire data volume loaded in memory is that it is quick to sculpt out different poftions of the seismic data.There are many ways to do this. A simple techniqub is to use interpreted horizons to guide sculpting: for example, data 172

173 -

3-Dseismic datavisualisatiOn

Fig.7.1 Difitfent ddtasets c{n be viewedsintultaneously. Herc a volumeof slandardreflectivity scisnricis viewedtogctherwith a cohcrcncyvolume(in blue).Also shownis an interpreted laull Jrldllepickedrvilh rheuid of borhvolulnes.

Fig, 7.2 Sulface and subsLrrlace data vicwed togelher.In Ihis example.taken from thc Wyrch Fanr oiltield in louthernEngland.we can sccthe colour codedrcservoirmap displayedtogclhcrwilh sateLlile imagingof the surlitce.

t

174 -

3-Dseismic datavisualisation

Fig.7.3 Combination ofdatatypes. In thisexample, weseeseismic datatogether withthetop reservoir timemapandwelltrajectories. might be removedaboveandbelowa reservoirso asto leavea 3-D bodycorresponding to the reservoiralone.A more automatedmethodis sometimesreferredto as yo.re1 plclirg. (A voxel is the 3-D equivalentof a pixel, andis the smallestindependent unit out of which the 3-D volume is constructed,typically conespondingto a 4 ms time sampleand a 12.5m x 12.5m horizontaltracegrid.) An initial seedpoinl is chosen, and the softwarethen automaticallyseeksout all connectedvoxels with amplitudes less(or greater)thana giventhreshold(fig.7.4). This is mostoften appliedto acoustic impedanceor standardreflectivityvolumes,thoughany seismicattributethat can be calculatedatall pointsin a 3-D volumecouldbeused.Forexample,we might choosethe iniiial point at a high-amplitudevalue,say120,andthenlook fbr connectedpointswith amplitudesabovea thresholdof, say,100.The softwarestaftsby looking at the nearest neighboursof the initial point in all directions:up, down, and in the four orthogonal sidewaysdirections.Voxelswith amplitudeabovethe theshold areconnectedinto the body.The processthencontinues,with nearestneighboursofthesenewadditionsbeing examinedto seeif they are abovethe thresholdand combiningthem into the body if they are. The processcontinuesuntil a picked body is createdthat in all directions is surroundedby valuesbelow the theshold, or which in somedirectionsreachesthe edgeof the surveyvolume.If, for exarnple,a hydrocarbonreservoiris markedby bright amplitudes,this is a way of automaticallydeterminingthe connectedpay volume.In practice,trial anderroris neededin settingthethresholdvalue.Ifit is too demanding,the growthofthe pickedvolumewill be quickly arrested;ifit is too lax, thenthe connected

175 -

3-Dseismicdatavisualisation

Fig. 7.4 Automatic tracking of bright amplitudes.The 3 D volume can be usedto quickly display all data points connectedto a given seedpoint with ampliludes greaterfhan (or lessthan) a given threshold.This may give a very quick way of sculpting out the reservoiror estimatingvolumes when used with acousticimpedancedata.

Fig.7.5 Viewing data in transparencymode. The 3-D data volume can be viewed with part of the data renderedtransparcnt.This allows the interpreterto get a 3-D perspectiveon the entile reservoir interval,examining issuessuch as connectivity and rock volumes

-,.*-.---.---iia

I - F

-.*r{i!

'176 -

3-Dseismicdatavisualisation

Fig, 7.6

Viewing data fr-omany angle.Here wc view liom abovc a 3-l) dala volunte that has weakeramplitudesrendercdtransparent, giving a gootl impressionol reseNoirdistributiolt.

Fig.7.7 Specialglassesworn tbr 3-D viewing.Polarisedglassesarc worn thatallow lhe viewerto seethe imageas a 3-D volume.

177 E

3-Dseismicdatavisualisation

Fig,7.8 Exampleoffull immersivevisualisation. New rechnologyincludesfully immersive displays where the data are projectedonto each wall and the floors and ceilings of a specially designedroom. This allows the interpreterto walk through.thedata,examining details of the reservoirfrom all angles.

volume will increase to fill much of the 3-D dataset. If the picking is successful, then

a 3-D volumecorrespondingto the 'seismicpay' hasbeencreated.This may or may not correspondto the real pay volume.In particular,the limited resolutionof seismic meansthat seismicconnectivityis not the sameas connectivitybetweensandsin the realearth.Productionbehaviouror prospectvalidity may be stronglyaffectedby sands too thin to imageseismically. Figure 7.5 showsanotheruseful display.Part of the data is renderedtransparent, leaving only those amplitudesthat might correspondto reservoir.For example,all amplitudeslessthan 100 could be removedcompletely,and amplitudesin the range 100 128could be madeprogressivelymore opaque.This givesan efficientmeansof viewingandunderstanding the3-D geometryofa reservoirbody. This is readilygrasped ifthe semi-transparent cubecanberotatedin realtime,sothatit canbe inspectedfrom a varietyofanglesin quicksuccession. Settingup thetransparency functioncanbe tricky; the idea is to makethe voxelsof interestbright while removing(makingtransparent) the rest.Trial anderror are inevitable.but a studvof well imoedancesand calculated

174 -

3-Dseismicdatavisualisation

Fig.7,9 Cinema style seFup ofa visualisationcentre.Cinema-stylerooms with large screensand enough spacefor teamsto view data togetherhave made a large impact on sharing ideasacross disciplines.

reflectivitywill help in decidingsemiquantitativelywhat rangeof voxel amplitudes represents the 3-D body we are interestedin. An advantageof loadingthe entire3-D volumeis that onecanrapidly scanthrough the completedatasetandview it from any angle.Figure7.6 showsa displaywherethe dataareviewedfrom above;weakamplitudeshavebeenrenderedtransparent to make a map view of the reservoirdistribution.Techniquessuchas thesehavedramatically reducedthe time it takesto performa first-passinterpretationofnew explorationareas andhavealsoincreasedunderstanding of existingfields. An advancethat is still experimentalratherthan commonpracticeis the ability to projectthedataon to a flat screensothatit appearsto theviewerasthoughhe is looking at a real 3-D image.The technologyis very similarto that usedfor somefilms that are viewedin 3-D. The imageis displayedas a stereopair with rapid altemationbetween the lefl-eyeand right-eyepictures.The viewer wearsspecialglassesthat incorporate an electronicshutterso that the left sideis transparentwhen the left-eyepictureis on the screen,and the right sidefor the right-eyepicture.Thus the left eye seesonly the left-eyepicture, and the right eye the right one. In this way, the 3-D perspectiveis re created(fig. 7.7). An evenmoreexperimentaltechnologyis one wherethe dataare projectedon the wallsandfloor ofspeciallybuilt 'visionariums'.Thesedisplaythedata

179 -

Retuence in full 3-D so that the interpreterscan evenwalk through the datavolume (fig.7-8). This technology is in its infancy but may hold potential for revolutionising interprctation in the future (Stark et al., 2000). At present,however,it is rather expensiveto implement. So far we have discussedhow working and displaying 3-D volumes, instead of the traditional 2-D slices, has improved the efficiency of the interpreter. However, similar technology has been at the forefront of increasedintegration acrossteamsand disciplines. Data volumes can be displayed in cinema-style surroundings, allowing many people to view and comment on the data simultaneously(fig- 7.9). Each person brings his own expertise,concemsand solutions,allowing for more informed decisions to be madeandensuringrapid resolution of outstandingissues-It is also an ideal setting for reviewing prospectsv/ith senior managementsince it allows for rapid interaction betweenthe audienceand the data.

-

Relerence Stark.T. J., Dom, G. A. & Cole, M, J. (zm0). ARCO and immersiveenvironmerts:lhe 6rst two generfii.nl,.ls. TheLeadingEdge,19, 52Gj2.

I

8

seismic Time-lapse

We saw in chapter5 how seismicdata can be used in favourablecasesto infer the natureof fluid fill (gas,oil or brine) in a reservoir.An applicationof this is to follow the way that fluids move through a reservoir during production,by carrying out a baselineseismic survey before production begins and then repeat surveysover the productionlifetime. Where3-D surveysarerepeatedin this way,they areoftenreferred to as 4-D seismic,with the idea that time is the extra dimensionover standard3-D. Differencesin seismicamplitudesor travel-timesbetweenthe surveyscan revealthe movementof fluid contacts(e.g. where producedoil has beenreplacedby brine) or the extentofpressurechangesthat affectreservoirproperties.As sketchedin figs. 8.1 and 8.2, it is not alwaysstraightforwardto separateout the causesof the changes.At a producingwell, the pressurcdrops,and if it dropsfar enoughthen gas will come out of solution.This will result in a decreasein both P velocity and in density,resulting in a drop in acousticimpedance.On the other hand,the drop in pore pressurecauses an increasein P velocity and density.At a water injector (fig. g.2), we replaceoil by water,which in itself would increaseP velocity anddensity;however,the injectionalso increasesthe pore pressure,which causesthe reverseeffect on P velocity and density. By comparinga pre-productiontracevolume with a post-productionvolume, we can at leastseewherethe changesarehappening,evenif we do not completelyunderstand them initially. Figure 8.3 showsthree snapshotsectionsof the reservoirsandin a UK Palaeocene field. we seean initial brighteningin the upper sandowing to gascoming out of solution,followed by dimming which may be causedby gasbeing re-dissolved or by waterencroachment. This type of informationcan be usedas a tool for reservoir management,for exampleby monitoringthe expansionof a gascap,with the intention of managingproduction so as to preventit from reachingproducing wells and thus decreasingthe oil productionrate. Reservoirmodels are built using well and seismicdata. The well data have high vertical resolution,but in many casesthere are a limited number of wells available, especiallyin a high drilling-costenvironmentsuchasdeep-wateroffshore.Interpolation betweenthe wells dependson seismicdata,which havelimited verticalresolutioneven after attemptshavebeenmadeto improveit using the methodsdescribedin chapter6. The result is that theremay be considerableuncertaintyover the possiblebarriersto 180

181 -

Time-lapse seismic

Producingwell in oil saturatedreservoir

Fig. 8.1

Pressure effect

Gas ex-solvedfrom solutionas pressure dropsbelowbubble point.Seismicvelocity (P) and density decrease

Decrease in pore pressureincreases seismic(P)velocity anddensity

Rock property changes around a producing well.

Water injectorwell in oil saturatedreservoir ".N,r,

Fig, 8.2

Saturationeffect

Saturationeffect

Pressureeffect

Water replacesoil around injector, increasingdensityand seismic(P) velocity

Increasein pore pressurereduces seismic(P) velocity and density

Rock property changes around a water injector.

flow, both vertically (e.g. shalebandstoo thin to resolveon seismic)and horizontally (e.g.faults that may be visible on seismicbut whosesealingcapacitymay not be easy to estimate).Although someunderstandingof thesebarrierscan be obtainedfrom the observedfluid flow in producingwells,in manycasesthis will not give sufficientlateral

182 -

Time-lapse seismic

Fig. 8.3 Time-lapse sections. Note the initial brightening in the upper level, followed by slight dimming.

resolutionto understandthe 3-D subsurfaceflow pattems.Also, it would be useful to have someway to anticipategas or oil breakthroughinto an oil-producing well before it happens.Time-lapseseismiccanprovidea way to get almostreal-timedataon these fluid movements.Although the idea hasbeenknown for many yearsand can be applied to isolated2-D seismiclines,it is only in the last few yearsthat the availabilityof highquality 3-D seismichasmadethe techniqueroutine.For example,BP acquiredits first 3-D survey primarily for 4-D effects and for commercial as opposedto experimental purposesin 1999, and'thenbuilt up activity quickly to acquirefive surveysfor 4-D effects in200l intheNorthSeaalone(Ritchie eta1.,2002).shellhadacquired 25 4-D surveys worldwide by the end of 2001, andacquired a further 25 in2002 (de waal & calvert, 2002). Like other users,Shell regardstime-lapseas proven technology for monitoring fluid movementsin thick clastic reservoirs offshore. The challenge is to extend the method to pressuremonitoring, to carbonateand to thin-beddedclastic reservoirs,and to the onshore(wheresignificantseismicdataquality issuesarise). There are severalstrandsto be combined in any time-lapse study. ' The rock physicsthat predictswhat amplitudeandtravel-time effectswill be produced by changesin fluid fill andreservoirpressure. ' Seismicacquisitionand processingissues:we needto distinguishgenuinechanges in the reservoir from differencesin the seismicdatadue to differencesin acquisition,

18ti| rrrl

Rockphysics

so to what extent are surveysrepeatable? what can be done to match one survey to anotherif acquisitionparametersare not precisely identical? ' Field managementissues: when do the time-lapsedataneedto be availableto have an impact on real-worlddecisionssuch as whereto placean infilr well? An excellentsurveyof the topic hasbeen given by Lack(1997). -

8.1

Rockphysics Modelling of the effecton rock properties of changesin fluid fir, pressureand temper_ atureis an importantfirst step.The issues havebeenwell summarisedby wang (Lggl). In general,time-rapseis most rikely to work whentherearelargechangesin fluid com_ pressibilitywith production,andin rocks with low elasticmoduli (poorly consolidated or with open fractures).For example, time-lapsehas beenvery eft-ectivein following steaminjection for enhancingrecovery ofhigh-viscosity oil from high_porosity sands (Jenkinset al'' 1997)'In othercases, it may not be certainwhethera change in seismic responseis greatenoughto be rneasured until aftera time-rapsesurveyhasbeencarried out' Effectsof changesin fluid saturation aremodelledusing the Gassmannmethodol_ ogy explainedin chapter5' An empirical observationis thatfluid effectsareoften larger than the Gassmanntheory would predict. This may be becauseclay contentis often not sufficientlytakeninto accountwhen assessingthe propertiesof the rock frame, or becausehydrocarbonsaturationis patchy rather than uniiorm. For example,injected water will move preferentialry along high-permeability layers within a reservoir and produceapatchymixture of hydrocarbo.r uno water attheseismicscale. Understandingthe effectsof pressure changeis more difficult. If the reservoirpressureis allowedto drop,thenthenet overburden pressureon thereservoirrock increases. For unconsoridated sandsor somehigh-porositylimestones, this may leadto a largere_ ductionin porevolumeandthusa significant increas"t;;;"d*"". i- *.u_cemented sands,the pressureeffect will be smalr. To estimatethe likely effect in any particu_ lar case,the main sourceof information is raboratoryn'"urur",nent of rock velocity undercontrolledpressureconditions. Most laboratorymeasurements aremadeat ultrasonicfrequencies(aroundI MHz) so that resultscan be obtainedfrom small samples. unfortunately,the Gassmanntheoryis rimited to low frequencies;real rocks showcon_ siderablechangeof velocity with frequency (usualryreferredto asdispersion) between the seismicandurtrasoniccases.At high frequencies,thereis not enoughtime for pres_ sureto equilibratebetweenporesas the seismicwavepasses,and rocks appear stiffer thanat low frequencies.A methodof predicting the sizeof this effectby modifying the dry rock modurihasbeendescribedby Mavko et at. (r99g).An additionalcomplication of laboratorymeasurements on core samplesis that they may not replicate in situ be_ haviour,mainly becauseclay minerals can be modified by interactionwith pore fluids or by drying.

188

Time-lapse seismic

I

in the near-surfacevelocitiesin the land case.Timing differencesof as little as 2 ms energy can give significantresidualenergyin the differenceplot. Root-mean-square required, even is usually variant scalar balancingusing a smoothedtime and space if the baselineand monitor survey were nominally acquiredin an exactly identical way. Spectralmatchingbetweenthe two surveyswill also, in general,be requiredto ensurethat they both havethe samefrequencycontent.Finally, differencesin seismic phasemust be removed.A phasedifferenceof as little as 15' will be significantfor the differenceplot; suchsmall phaseshifts aretypically too small to be apparentto the unaidedeye,and rely on softwarefor their detectionand correction. -

8.5

Examples Two interesting examples were presented by Hatchell et al. (2002), who discussed the process of comparing time-lapse results with predictions from a reservoir simulator. Figure 8.4 shows the time-lapse difference for both real and synthetic data over the Schiehallion oil field in the West Shetland area.The main sourcesof time-lapse response are pressureand saturationeffects near lhe wells. Near the water injectors.reservoir pressures have increased from 17 MPa to 41 MPa, which causes brightening on the seismic. Near producing wells, the pressure drop causesan amplitude decreaseas long as the pressure is still high enough for gas to remain in solution; once the pressureis low enough for free gas to be present, the amplitude brightens. All these effects were visible on both the real and the synthetic data, and detailed examination of the real data allowed the mapping of pressure compa.rtmentsseparatedby sealing faults, some of which are too small to be resolved by conventional imaging. Another example is from the Maui sas field in New Zealand. The reservoir interval contains laminated sands and shales,

Fig, 8,4 Schiehallion field: difference in amplitudesbetween pre- and post-production for both real (left) and synthetic (right) seismic. Reproducedwith permission from Hatchell et al. (2002)

189

Examples

I

Maui?ypoLag

ffi'rlF%l

Fig' 8.5 Maui fleld: type log acrossthe reservoir interval, and comparison of models to real time-lapse seismic. Reproducedwith permission from Hatchell et al. e002\.

and the questionwas whetherwateris simply displacinggasvertically throughoutthe reservoir or whether the water is preferentiallymoving through the uppermostand most permeablesands,over-ridingunproducedgasin the lesspermeablesandsbelow. Figure 8.5 showsa comparisonbetweenthe real time-lapsedataand syntheticmodels for the two cases.Clearly,the real data arein closeagreementto the prediction with waterover-ride. Stammeijeret al. (2002)havepublishedthe exampleshownin fig. g.6. This shows time-lapseresultsoveramaturefield in theNorthemNorth Seathatis beingproducedby water-flood.The baselinesurveywasacquiredin 1995,somesix yearsafterproduction began,and the repeatsurvey was acquiredin 2000. In this case,connectedvolume analysis(asdiscussedin chapter7) was usedto identify the zoneswhereoil had been swept from the reservoirin the period betweenthe two surveys.Someof the zones whereno changehasoccurredcanbe identifiedasareascompletelysweptbefore 1995. The othersrepresentunsweptareas,which can now be targetedby an infill drilling programme. Repeatseismicsurveysover a field in the UK West Shetlandareaallowed a flood front to be seen(fig. 8.7). Shortly after the repeatseismicsurveyshowedthat the flood front wasapproachingthe producer,significantwaterproductionbeganto be observed.

190 -

Time-lapse seismic

Fig, 8,6 Northem Norlh Sea field. Perspectivedisplay of field structure, with wells and reservoir bodies indicated. Blue bodies: oil swept between 1995 and 2000. Reproducedwith permission from Stammeijer et al. (2002).

FloodFrontMonitoring Wl Startup

1999 Seismic

2000Seismic

2000 - 1993:FloodFrontVisible

Fig, 8,7 In the upper graph, green is oil production rate, blue is water production rate and red is the gas-oil ratio ofthe produced fluid. The difference map ofamplitudes between the 1999 survey and the 1993 pre-production baseiine (not shown) revealedbrightening due to gas coming out of solution. The difference map between 2000 and 1993 shows brightening (green) in the peripheral areasowing to gas evolution, but dimming (blue) around the water injector. The blue zone therefore shows the position of the flood front in 2000, and is consistentwith the subsequentwater breakthrough at the producing well.

191 -

References

1999Data Pre-Production

2000Data Post-Production

Amplitude increase suggests pressure drop(flow)alongentire welllength

Fig' 8'8

Amplitude increase(green) suggeststhat pressuredrop is occumng everywherein the vicinity of the producer well; there had been concem that it was producrng only from some segmentsof the faulted and channelisedreservoir. A 4-D seismic survey was cheaperthan establishingthe flow from each section of the well by direct measurement,which would have interfered with production.

In anotherfield in the samearea,repeatsurveysestablished that a well waseffectively drainingthe whole areaaroundit (fig. S.g). I-

References de waal, J. A. & calvert, R. (2002). 4D seismic all the way - implementing time lapse reseryoir monitoring globally. Abstract H-0r, EAGE 64th Annuar Meeting, Frorence. Hatchell, P., Kelly, S., Muerz, M., Jones, C., Engbers, p., van der Veeken, J. & Staples,R. (2002). Comparing time-lapse seismic and reservoir model predictions in producing oil and gas fields. Abstract A-24, EAGE 64th Annual Meeting, Florence. Jack'r' (1997)'Time-lapse seismic in Reservoir Managemenr. Distinguished Instructor Short course, Society of Explorarion Geophysicists. Jenkins, s. D., waite, M. w. & Bee, M. F. (1gg7). Time rapsemonitoring of the Duri steamflood: a pilot and casesrudy. The Leading Edge,16, 1267J4. Landro, M. (1999). Repeatability issuesof 3_D VSp data. Geophysics,66, 1673_9. Mavko, G', Mukerji' T. & Dvorkin, J. (199s). The Handbook of Rock physics.cambridge university Press.

184 -

Time-lapse seismic During conventional oil production, temperature changes in the reservoir are usually fairly small. Injection of cold water (sometimes undertaken to maintain pressure and improve sweep efficiency) may cause detectable changes. Where steam is injected into a reservoir to improve recovery of viscous oils the temperature changes can be large, particularly around the injector wells. By combining the rock physics data with the output from a reservoir simulator, it is possible to generate synthetic seismic models that show the changes to be expected in the seismic response over time. This is important for planning a time-lapse survey. Knowing the size of the changeswill indicate whether time-lapse is feasible at all. A dialogue is also neededwith the reservoir engineers.To be useful, a time-lapse survey needs to be available at a moment when it can influence reservoir management, e.g. in deciding the location of an infill well. However, the size of time-lapse effects depends on how much production has taken place. It may not be feasible to detect a time-lapse signal during a very early stage of production, because there has not been enough fluid movement or pressurechangeto be visible. Choosing when to shoot a time-lapse survey requires collaboration between geophysicists and reservoir engineers. When the survey has been acquired and interpreted, it is usual for the results to be different from the synthetic seismic prediction derived from the reservoir model. The model then needs to be modified so as to give a synthetic seismic response that matches the time-lapse survey. Given the uncertainties in the model and in the seismic data, a range of modifications may match the observed results. Again, dialogue is needed to arrive at a result that can be used to guide reservoir interventions such as modifying oil offtake or water injectionpattems.

-

8.2

measurements Seismic If we havea baselineand a repeatsurvey,what can we actuallymeasurethat will tell us about the differencesbetweenthem?There are severalpossibilities,as follow. In any particular caseit will be necessaryto calculatethe expectedchangefor a given reservoir,fluid fill andpressurechangeto seehow big the seismicchangesmight be. (a) TWT to a reflector.Abovethereservoir,thereshouldideallybe no TWT differences betweenthe baselineand repeatsurveys.In the reservoiritself, seismicvelocities will changewith pressureor changingfluid fill. The TWT thicknessof thereservoir will thuschangewith production.Seismiceventsbelow the reservoirwill therefore changein TWT betweenthe two surveys.This effect can be quite large,ranging from a few up to about 10 ms for a thick reservoirinterval. In principle such a time shift is easily detectable.Where seismicdatahavereasonablygood signalto noiseratio, it shouldbe possibleto pick a reflectorto an accuracyof I or 2 ms. Wherewe are dealingwith a shift betweenone suiteof reflectors(abovethe reservoir) and another(below the reservoir),thencorrelationmethodsshouldgive even

185 -

Seismic measurements better accuracy. Such time shifts are often the timeJapse effect that is most robust against differences in acquisition and processing between baseline and repeat surveys. (b) Amplitudes of reflectors. As explained in chapter 5, the amplitude of a reflector (e.g. at top reservoir) may be diagnostic of its fluid fill. changes in amplitude between baseline and repeat survey may therefore be diagnostic of changes in fluid fill. As we shall see,amplitude changesbetween the surveys may also be causedby several factors related to data acquisition and processing, but it can still be possible to extract useful time-lapse information. (c) Seismicvelocities.Subsurfacevelocities areroutinely estimatedas a stepin stacking and migrating seismic data, and in principle they will change with change in fluid fill. However, the accuracy with which stacking velocities can be estimated is limited; it will often be hard to get a higher accuracy than lvo.In a realistic case, the effect of a change in fluid fill on stack or migration velocity to a base-reservoir reflector will usually be less than this. The problem is that the velocity change is confined to a fairly thin layer, but stacking velocities are an average velocity from surface to the reflector concerned, and are dominated by the thick unchanged overburden. Also, as we saw in chapter 3, stacking velocity may change with acquisition azimuth owing to genuine anisotropy or to dip and curvature of the reflector; it will also change with azimuth and maximum offset due to near-surface effects' It is therefore sensitive to acquisition differences, particularly when very high accuracy is wanted. All this makes velocities hard to use as a time_lapse diagnostic. If we are intending to make use of differences in TWT and amplitude between surveys, there are two possible approaches. where fluid or pressure effects are large, and fluid movement follows a simple pattern, it may be sufficient to make maps of rwr and amplitude for a reservoir-related reflector on both surveys, and compare them. If, for example, hydrocarbons cause obvious bright spots at top reservoir, then the extent of the bright spots on amplitude maps of the two surveys can be compared visually. This does not require that the acquisition or processing should be exactly the same in the two surveys. It should certainly follow good practice for obtaining reasonably accurate amplitudes, but what we are looking for is a lateral change in amplitude on each map independentry (marking the fluid contact), followed by a comparison of the positions of these amplitude changes on the maps of the two surveys. If, on the other hand, we are looking for small effects, then we may have to subtract one trace volume from the other in order to reveal subtle differences, such as small amplitude changes within a reservoir interval caused by interfingering of oil and water in the producing zone. This is much more difficult to do, because there are many possibre reasons for differences between surveys that are related to acquisition and processing differences. These differences will appear as unwanted noise in the difference volume, and their elimination requires careful processing.

186 -

Time-lapse seismic

8.3

repeatability Seismic If we arethinking of comparingtravel-times,amplitudesor seismicvelocitiesbetween a baselineand a repeatsurvey,there are severalpossibleother causesof differences betweenthe surveysthat will interferewith our comparison: (a) Noise.Ambient noisecan easily vary from one surveyto another.For example,if the repeatsurveyis beingcarriedout over a producingfield, therewill probablybe many sourcesof industrialnoisethat were absentwhen the pre-productionsurvey was shot.This neednot be a significantproblemso long asthe signalto noiseratio at reservoirlevel remainshigh. (b) Accessibility.A commonproblemis that someareasthat were shotin the baseline surveyarenot accessibleto the repeatsurveybecauseofproduction facilities. For example,the presenceof a productionplatform offshorewill createa hole in the surveyacquisitionbecauseit will not be possiblefor a surveyvesselto approachit very closely for safetyreasons.This meansthat over part of the areaof interest,it will not bepossibleto acquirea surveythatcloselyreplicatestheoriginalacquisition pattem. (c) Near-surfaceeffects.There may be quite large changesin the near-surfaceover time, for exampleas a result of changesin the depth to the water-tablewith the seasonof theyear.Although in principleit is possibleto allow for theseeffects,they introducean additionaluncertaintywhen comparingtwo surveys.The shifts may be comparableto or greaterthan the expectedtime-lapsetime shifts,but of course the latterwould be confinedto the intervalbelowthereservoirwhereasnear-surface effectswill causea shift of the entire trace.In the marinecase,similar effectscan be causedby a changein the soundvelocity in the sea-watercolumn, owing to temperaturechangescausedby large-scalecurrents.Tidal effectsalso needto be allowedfor. (d) Sourcesignature.This maydiffermarkedlyfrom onesurveyto another.Ideally,seismic data shouldhavebeenconvertedto zero-phaseduring processing(chapter2). If this hasbeensuccessful,it will removemuch of this problem. (e) Acquisition parameters.There may be differencesin the geometryof the source and receiverpattem betweenthe surveys.Jack (1997) quotessomeexamplesof effectsthat were simulatedby taking a well-sampledsurveyand removing some of the data,followed by processingof the decimateddataset.For example,when an 80-fold marine datasetwas decimatedto 4O-fold,then the differencesection betweenthe final 80- and 40-fold sectionsalong a typical dip line had an rms amplitude4OVoof the original section.In sucha test, becausethe decimateddata 'baseline' survey.positioning were producedby merely dropping tracesfrom a repeatabilityis not an issue.In rcaldata, two problemsarise.One is that, evenif we knew the positionsof sourcesand receiverspreciselyfor the baselinesurvey,

187

T

processing Seismic it might not be possible to replicate them. This is typically the case for a marine survey, where streamer locations are affected by surface currents; recent technology allows steeringof the individual streamers,which will reduce,but not eliminate, the problem. Secondly,there will always be inaccuraciesin position-fixing of shotsand receivers.Considerableimprovement in positional accuracyof marine surveyshas been achievedfrom about 1990onwards,owing to satellitepositioning, the tracking of streamertailbuoys, and the use of acoustictranspondernetworks. If the baseline survey is old enough to have been shot before these improvements, then there may be quite large uncertainties on the exact locations of sources and receivers. One simulation showed that errors of 80 m in the location of the tail of a 4000 m marine streamer could give rise to difference sections with an rms amplitude 20-30o/o of the baselinedata. Study of VSP data by Landro (1999) showed rhat shifring a shot location by 10 m resulted in an rms change in the record (after static alignment) of 20Vo,and a20 m shift resulted in a 30Vochange. The records were dominated by the downgoing signal, so thesechangesare causedby transmissionresponsevariations between neighbouring ray-paths. One way of reducing acquisition differences is to leave receivers permanently in place on the seabed, as has been tried in the BP/Shell Foinaven Field. Even so, difference sections between surveys showed rms amplitude differencesup to 35Voof the origin al data(Jack, 1997). (f) Processingparameters.There are many stepsin the processingsequencethat can introduce differencessimilar to the time-lapse signal we are looking for. These include staticscorrection, mute design,pre-stackdeconvolution,stack and migration velocity derivation, and amplitude balancing. It is often necessaryto reprocessthe baseline survey together with the repeat survey,as discussedin the next section.

-

8.4

processing Seismic Specialised processing can alleviate theseproblems of repeatability. Where the baseline survey is a 'legacy' dataset,not originally acquired with time-lapse in mind, it will be beneficial to reprocess it in parallel with the monitor survey. This ensuresthat the same algorithms are used at each stage,with so far as possible the same choice of parameters. At each step, the surveys can be compared and action taken to match them if necessary. This might include time- and space-variant amplitude and spectral trace matching betweenthe two datasets.A detailed discussionis given by Ross et al. (1996). The idea is to match the two surveys over the intervals outside the reservoir, where no time-lapse changes are expected. The quality of the match can be seen from the difference volume between the baselineand monitor surveys,which should ideally be zero except where the production-related effects are present. The matching process can include several elements. Time corrections are needed where there are systematic shifts, for example as a result of salinity variations in the water column in the marine case, or changes

192 -

Time-lapse seismic

Ritchie,B., MacGregor,A., Strudley,A. & Goto, R. (2002).The impactof new 4D seismictechnology on the Magnus Field. Abstract A-20, EAGE 64th Annual Meeting, Florence. black box. The Ross,C. P., Cunningham,G. B. & Weber,D. P. (1996). Inside the crossequalisation Leading Edge, 15, 123340. bstractA-26, Stammeijer,J., Cooke,G. & Kloosterman,H. J. (2002).4-D drivenassetoptimisation.,4 EAGE 64th Annual Meetins, Florence. Wang, Z. (1997). Feasibility of timeJapse seismic reservoir monitoring: the physical basis. Tle Leading Edge, 16, 1327-9.

Appendix1

Workstation issues

This appendix covers briefly the main issuesinvorved in managing hardware,software and data to createan environmelt for the interpretationof 3-D seismicdata.No attempt will be madeto discuss specilic detailsof individual vendors,offerings,as they changevery quickly; rather,the objectiveis to give a generalovefliew ofthe requirementsfor the creationofa successfulintemretationenvironment.

-

A1.1 Hardware The volume of seismic data in a 3-D survey is often large. As we saw in chapter3, a modest survey may conlain 500000 traceseach ol 1000 samples.A large suwey might contain tens of millions of traces.It is quite usual to have severarversionsof the data vorume (e.g. near and rar trace stacksfor AVo analysis,perhaps severaldifferent inve$ion rcsults, and calcuratedattibure volumes sucrras coherence). Storage requirements for the trace data may amount to several to a few tens of gigabyte (lcbyte: 10' byte). If a company has interests in a number of licences,each with its own 3_D sufley, the total volume of seismicdata may runountto a few terabyte( I Tbyte : l0r2 byte). Storing such a volume on disk is possible,but the cost will be less if some of the data are held offline on magnetic tape, preferably in a high-density fomat (e.g. helical scan) that will allow large volumes (e.9. 100 Gbyte) to be stored on a single rape.Rapidly changing conrmercialpriorities will dictare that the archive of data on tape is not static, but wilr need to be retrieved and reworked, maybe with only a day or two's notice. Robust data administrationproceduresfie neededto make this possible. A further complication arisesin a large integratedproject, where severaluse6 may needaccessto the same seismic dataset.Insteadof each interpreterhaving his or her own workstation which runs intelpretationsoftwareon data held in local disk storage,issuesof cost and version conto] may then dictate that all data a.e held centrally and accessedby software that may also be run on the central seryer or may be run on the interpreter'sworkstation. Computer power and dala accesstime on the central server and bandwidth of the network between it and the interpreter'sseat w r then dictate systemperfomance, and needcareful considerationto get good pedormancein a large installation. As with all computer systems,backup of work in progress is critical. The interpreter,swork needs to be backed up onto tape, preferably at the end of every working day. Many installations have found that proper backup is more certain to be canied out if it is managed cenhally rather than by the individual interpreter This is another driver towards centrally held and managed data stores. !\rherc there are many interpreters involved, the amount of data to be backed up each day will becomelarge enoughto needdetailedplanning; high-densitytape ddves are a suitablemedium, but an emcient systemis neededto make sure that all the required backup is completed in the time available.

193

194 -

Appendix 1 Workstation issues The interpreterspendsa laige part of his or her working day looking closely at data on a screen display.High-quality display screensarethereloreimportant.As with any work involving sustaineduse ofa screen,mouseandkeyboard,the?hysicaldesignand siting ofthe units needscarefulconsideration to avoid possibledamageto health. I

E

/

A1.2 Software Decisions on which software should be made availableon the system depend on considerationof functionality and cost, and need to be assessedin detail for each individual installation. The main problem from the point ofview of the systemadministratoris usually the needto provide datatransfer betweenapplicationsfrom different vendors.This can of coursebe reducedby restricting soliware purchasesto a single vendor, but even then there is likely to be a need to interface with proprietary applicationsand databases.Where applications from severalvendors are in use, system upgrades (e.9.major updatesto operatingsystems)can be very arduous,requiring updatesto both softwarc and databases.Pafticularly when there is doubt about the reliability of hardwareor software in the new environment,it may be necessaryto run both old and new environmentsin parallel, with a phased transferfrom one to the other; this needscareful liaison with interpretersand other usersof seismic data to avoid seryiceinterruptionsat momentsthat are critically important to the business. E

41.3 Datamanagement Managementof seismic trace data is for the most part not difficult, except for the huge volumes involved. The data are organisedin a very rcgular fashion, as values of amplitude (or other attdbute) on a regular 3 D grid. When data are retrieved,it is almost always along a plane or line or sub-cube of the 3 D grid (e.g. as sectionsor traces).There is no rcquitement to identify individual samples that satisfy complex selectioncriteria. This meansthat the data do not have to be held in a complex databasestructure.Once the data have beenloaded, usually from SEGY tapes,the files require little lurther maintenance.Dataloading is usually lairly easyfor 3-D seismictraces.The main problem is the specificationof the spatiallocation ofthe traces,which is often doneby manuaUytyping in real-world co-ordinatesofpafticular points,e.g.of threeconerpoints ofthe rectangularsurveygrid. This process is potentially eror prone; small mistakesare difliculr to spot once the data havebeenloadedbut could havedisastrousconsequences for accuratewell positioning.Furthermore,thereis scopefor confusion betweendifferent co-ordinatesystems(projection,spheroidand datum). Sometimesposition data are suppliedon one projection systemand needto be tmnsformedto anothet to match againstcorporate standardsor to mergewith anotherdataset.It is useful to have the servicesofa topographicsuNeyor to ensurethat the co-ordinatedata as loaded are corect and in the desiredprojection system. Loading seismic tracesfor a set of 2-D lines is much more difficult. The main problem is to work out the relationshipbetweentracenumberon tape and position data on amap; this can be quite simple where the tracesand positions are numbered sequentiallyalong a line, but where a line has been shot in severalparts with discontinuitiesin numbedng of tracesand surfacelocations,it can be very time consumingto work out the relationshipsunlessthere is excellentpaper documen[ation. Managementofseismic interpretationdata(i.e. pickedho zonsand faults)is lesssimple.Becauseit is easyto createnew horizons,aminterpreterat the end of a mappingprojectmay havedefinedhundreds of them. Most of them will be trial efforts to investigatedifferent featuresof the data,perhapsover very limited areas;usually a few dozenhorizonswill contain all the significantinformation-Cleaning

195 -

Reference up rnterpretationprcjectsto throw away all the useless horizonsshouldbe partofthe repofting process at the end ofa proiect; ifthis is not done' it is very difficurt for anotherinterpreterto come to gips wrtlr the interpretation and to uqe it as a basis for further work. To make such re-use easier, there also need to be fairry rigid naming conventionsfor horizons, lioked to rhe stratrgraphiccorumn. over a period of time. a givenir5hmic/olume will be interpreted by a number of people, and different versions may exist of the samestratigraphichorizon (depending, lor example,on the view taken about conelating horizonsinto undrilled fault blocks).In addition,a given geographicalareamay be coveredby several diftbfent inferpretationpfojects' coresponding to different seismic surveysor difterent reprocessed versions of a singre survey. The interpreter may need to review alr of them, and carry forward several of them into rhe next stageof study (e.g. reservoirmodelling), in order tully to reflect the degreeof uncertaintyabout the subsurface(Herron,2001). It may be hard to find out what interprctationsare available,letalonedeciding which is the bestfor any particularpurpose. ldeally, a databaseis required showing availableinterpretationdata by geographic area,stratigraphictevel, etc_;it is nor difnculr ro devisea simple scheme,but mainrainingsuch a aatalase wil hlve a srgniticantmanpowercost. Finally, the interprererneedsto be able to access we' data to tie the wells to the seismic survey. Sonic and density logs, at least, need to be available in digital form, so that well syntheticscan be c'eated ro undemtandseismicresponseand its lateral variation,it is herpfulto haveother wireline rogs as well; for example,the caliper log, which measures boreholediametet is a useful indicator oflog reliability, becausewherethe boreholediameteris locally much largerthan expectedthe logging tools may not be able to record correct formation parameters. The interpreter also needs stratigraphic and lithological information ln general,management of werl data is not easy.stratigraphic information in particularcan be complicated,with repeatedormissing sequencesin any particularwell and lateral cnanges1nnomenclature.A systemdesignedprimarily for useby geologlstsmay be over_complicated fbr the needsof the seismic interprete(.For the creation ofwell slnthetics, it may be useful to set up a separatedatabase(or partition of a larger database) containing sonic and density curves specially edited for syntheticcreation,togetheJwith checkshot information, major formarion tops and perhaps the caliper log and the gzlmma-raylog for correlation with other displays and tbr identification of sand/shaletrends.The issuehere is that, as we saw in chapter 3, the we'synthetic can be sensitive to noise in the sonic and density logs. Where wircline logs are being used pnmarily for fomation evaluation,defectivereadingsin the shaleswill probabty not get paid irucn attention;therefore,there is a need to edit logs specifically for synthetic g.n".ution, lnj on"" edited they might as well be retainedin a simple databasesrrucrure.

r--

Reference Henon, D. A. (2001). problems wirh roo much data. The Leading Edge,20, 11244.

Appendix2

Glossary

Accommodation space The spaceavailablefor potential sedimentaccumulation,controlled by processessuch as changesin sealevel, tectonic movements,compactionofpre-existing sedimentand subsidence.

Acoustic imoedance A property of a rock, definedasthe product ofdensity and seismicvelocity. The impedanceson either side ofan interfacedeterminethe reflection coemcient for seismic waves(section3.1).

Aeolian An aeoliandepositionalprccessis one where the dominant agentis wind. A typical modem example is a hot arid desertsuch as the Sahara.There may be areascoveredwith sand dunes,and other areas of bare exposed bedrock. Large arcas of sand dunes folm sand seas or ergs. Aeolian sandstonescan be important hydrocarbonreservoirs,as for example in the Permianof the North Sea. AGC Automatic Gain Contlol is the processof varying the gain of a seismictlace display with TWT' so as to maintain the averageabsolutelevel constantwithin a time window. If a short time window is used, the process has th€ effect of destroying information about lateml and vertical changes in reflection strength,which is highly undesirableif any attempt is being made to recogniseeffects due to fluid contentor lateral changein lithology. If a long window (e.g. I s) is used,however,the prccessmakes it easierto view seismic displays acrossthe full TWT range,and can even be benefrcialto amplitude studiesby removing effectsof variation in near-surfaceabsorption,for example.

Aliasing Seismic tracesairenot continuousmeasurementsin time; instead,the amplitude is recordeddigitally at a certain sample interyal in time (usually 2 or 4 ms). Nor do we record traces at every possible (t, )) location; rather, we shall have traceson a regular grid, perhaps25 m x 25 m. Al1 our data processingand interpretationassumesthat this is an adequateapproach,in the sensethat a smooth curve through the sampledpoints is a ffue picture ofreality. This would obviously not be true if there were stong high-frequencyvariationspresentin the data,so that in reality there were high-amplitude oscillations happening between the sample points. It can be shown that signals can be recovered from a regularly sampled representation provided that they do not contain ftequencies higher than half of the sampling frequency.Frequencieshigher than this cannot be recovered,and are said to be aliassed;they will mimic the behaviour of a lower frequency signal. Care is therefore needednot to alias signals when they are recordedjif there is a dangerof this, the sampling frequencymust be incrcased or the bandwidth of the incoming signal decreasedbefore it is sampled for tuming into digital torm.

196

197 -

Appendix 2 Glossary Alluvialfan On land, rivers are the main means of ransporting sediment. Alluvial fans arc localised areas of high sediment accumulation, formed at a place where laterally confined llows are able fo expand, tbr example on leaving a gorge to flow into a broad valley floor. The expansionof the flow Ieads to a velocity decreaseand a reduction in ability to transport sedimenl, which is therefore deposited.

Amplilude spectry'm The methodsofFourier analysisallow us to envisagea seismic trace as the sum of a seriesof traces that are pure sinNaves. lf we perform this analysis and then plot the amplitude of the sine waves againstfrequency,&e have the amplitude spectrumof the original signal. Seismic processingoften aims to produce a rcasonablyflat amplitude spectrumin the frequencyrange from, say,8 ro 50 Hz, though what is possibledependson the seismic source,the propagationpath length and the degreeto which the rocks absorbhigh liequencies Angle stack It is often possibleto infer subsurfacelithology or fluid fill from the way that reflection amplitude varies with incidenceangle on a reflector (AVO). One way to investigatethis elTectis to make substacksofthe data,including only datafrom a particularanglerange.Thus, one might comparea stack containing incidenceangles0 15" with a 20-35' stack.These datasetsare producedby stackingthe appropriaterange of offsets.which will vary with TWT: the requiredoffset rangefor a particular set of anglesat a given TWT can be calculatedif the seismic velocity is known as a function of depth, either from well information or seismic velocity analyses.

Anisotropy ln general,anisotropymeansthat aphysical propertyvarieswith the directjon in which it is measured. In our case,the property of interestis usually seismicvelocity. In most cases,this exhibits transverse isotropy; thereis a symmetry axis, and velocity is the samein all directionsin the planeperpendicular to this axis, though different from the value parallel to the axis. Fine layering (e.g. in a sand-shale sequence)of mate als with different velocitieswill producerocks with a vertical symmeffy axis. Vertical cncks, all oriented in the sameplane,would producea horizontal symmetry axis, perpendicular to the cracks.

Attdbutes The lirst measurementsmadeon seismictraceswere offfavel-time to a reflector,with a view to making a structural map of it. As data quality improved, it was realised that the amplitude of a reflection could carry useful information; changesin amplitude along the reflector might relate to lithology or reservoirquality, for example.There are many different ways in which we might measureamplitude. The maximum excursion of a loop is the most obvious, and would, for example, be appropriateif we were looking at an isolatedreI'lectorat the top of a massivesand ove ain by a massiveshale.On anotheroccasion,we might be looking at an interbeddedsand shalesequence,where there would be a whole seriesof reflections,each one may be laterally discontinuous,betweenthe top and the base of the sequence.If we wanted to characterisethe va ability of the sequence,it would be useful to measurethe averageamplitude (without regard to sign, so averageabsolutevalue or lrns average) over the inteNal betweenthe top and base of the package.Or again, we might also get infomation from the shapeof a seismicloop. lbr example where we are trying to follow thin beds near the limit of seismic resolution: the most obvious measuremight be the width of the loop, i.e- its duration in TWT between zero-crossings,but many other shapemeasuresare possible. 'Attribute' is in use as a loose generic term to cover all thesetypes of measurementon a seismic trace. It usually refers to

198 -

Appendix2 Glossary measruementstaken on a limited time-window around a reflector ot interval of interest, and very oftel to measuremerits made on single traces, though some attributes are measured over a small cluster of traces(e.g.the local dip or curuatureofa rcflector).

Autotacker Software to follow a reflector automatically through a seismic trace volume, starting from a limited amount of manually picked data (seedpointsllines).

AVo Amplitude VersusOffset: the variation of reflector amplitudewith source-receiverseparation(see chapter5). Sometimes denotedAVA (AmplitudeVersus(Incidence) Angle). Avulsion OverbanksedimenlLtionftom an activeriver channelcausesthe growth of a ridge abovethe surroundingfloodpl{rn, wifhin which the channelis contained.When a channelbank is breachedin a flood, the river ca\find a new path on a lower part of the floodplain,leadingto the abandonment of theold channel.ThiFprocessis calledavulsion,andoccursin deltadistdbutarychannels aswell as meanderingdvers. Bin The final processingoutput of a 3-D surveywill be a set of traceson a rectangulargrid. It is not feasibleto acquircdataon a perfectly regulargrid. Therefore,a uniform grid of rectangular,ir,r is superposed on the acquisitionmap.Eachacquisitiontraceis assignedto the bin locatedat the point midwaybetweenits sourceandrcceiverpositions.The tracesfalling within a bin will be stacke.d after correctionfor flormal moveout,giving the unifom output grid of fiaces.Somerefinementsof this basicideaarediscussed in chapter2. Brightspot The classichydrocarbonindicatoris the bdght spot, whenpay (usually gas-bearing)sandsare very soft and brine sandsare similar in impedanceto the overburden.The top reservoireventwill then be a strongsoft loop in the pay zone,and a weakeventelsewhere.Suchan amplitudeanomalywill conformto sffucturc,unlessthereis a stronglateralchangein lithology acrosstheareaofhydrocarbon fill. Bulkmodulus If a volumeV of a mediumis actedon by a prcssre P, thevolumewill be decreased by Ay. The dilatationA is definedasA : AylV, andthebulk modulusI is definedas

wheretheminussignis insertedto makeI positive. Checkshot survey In orderto tie well informationto seismicdata,it is importantto know therelationshipbetweendepth and TWT at the well. This can be esrablishedby measuringthe travel-timefrom a sourceon the surfaceto a rcceiverat a known depthin the well. This information is usuallycollectedat a number of receiverdepths.In the caseof a deviatedwell, either the sourcecan be locatedvertically above the receiver(andthus at somedistancefrom the wellhead),or cor.ectionscanbe appliedfor the slant favel path. CDPgalher Synonymfor CMP gath€r(seeCMP gather).

199

Appendix2 ctossary GMPgathel The common Midpoint gather is a collection of traces having different source-recelver oilsets but the same midpoint location between source and receiver. For hodzontal reflectors, the traces of the CMp gather have the same reflection point and can therefore be stacked together alter NMO correction. In the presence of rcflector dip, DMO correction will be needed before stack.

Crevasse splay W1lerc a fiver is confined between levees, from time to time during floods the water level will reach the top ofthe levee and spill over At a place where this happens, a-shallow crevasseis formed on the crest of the levee.From the crevasse,lobes of silty or sanay seaimentspreadonto the floodplain as crevassesplays.similar featuresare fomed in the distributarychan rels of de o".

Crossline _SeeInline.

Dalum on land, there may be apFeciable topogaphy across a survey u,ea. Furthemore, ifexplosive charges are used as the source then they may be buried at varying depths across the survey. Also, there will usually be rateralchangesin the thicknessofthe near-surface low-verocity rayer.To rcmove a'these complications, a reference datum is selected and corrections -" ,nud" to travet times so that they representwhat wourd be recordedif shots and receivers were placed on the datum surlace,it being assumedthat thereis no fi[ther low_velocitylayer below this srrrface.For onshoresurveys,the datum surface may be flat or it may be a more or less smoothed version of the topography. Offshore, mean sea-level provides a convenient datum. Decimation Reduction of the number of tracesin a datasetby sysrematic removal of, for example, every other trace lt is not restricted, as the name suggests it might be, to the removal of every renrn race.

Deconvolulion The seismic trace can be thought of as the result of convolving the earth reflectivity with a wavelet. Ideally, we would like to have a seismic

arenectedsisnarwhich*".,i.pryu;.:x:;ilil::,1ffjffi:il.*;:tr,":.T:",_ offinite length.Funhemore, the sourcesignal will bemodified as it passesthroughthe earth,because of absorption, scattering, and other causes. The result is that our recorded signal will be the sum of the reflectionsof a wavelet frcm the sedesof subsurface reflecton (fig. A2.l-). tuathematically,this processrs representedby the convolution ofthe wavelet with the earth reflectivity. Deconvolution is a signal processingstep that attemptsto undo this convorution,ro leave us with the earth rcflectivity senes,thus improving the resolution ofthe seismic data.

Delta A depositionalsystemformed where a river supplies sedimentto a coast,foming a shorelineprotu_ berance. The input of mud and sand from the river is reworked by marine processes lwave and tide action). The componentsof a delta system(from onshore to offshoreyare, 1tI the aettaplain, largely sub-aerial and consisting of distributary channels separared by interdistributary areas of marsh and swamp; (2) the delta front, where fluvial and marine processes intemct and form features such as sand ba$ at the distributarymouths; (3) the prodelta,a zone of quiet deposirion,typicatty ofmud and fine silt, from suspension.As time goes by, if sedimenrconr;ues to'be suppli;;, ihe delta builds our_ ward from the coast.In vertical succession,the prodelta muds will then Le overlarn by progessively shallower-water deposits, generally with higher silt and sand content.

199 -

Appendix 2 clossary CMPgalher The Common Midpoint gather is a collection of traces having different source-receiver offsets but the same midpoint location between source and receiver, For horizontal rcflectors. the traces of the CMp gather have the same reflection point and can thetefore be stacked together after NMO correction. ln the presence of reflector dip, DMO corection will be needed before stack.

Crevasse splay Where a dver is confined between levees, from time to time during floods the water level will reach the top ofthe leveeand spill over.At a place where this happens,a shallow crevasseis fomed on the crest of the levee. Frcm the crevasse, lobes of silb, or sandy sediment spread onto the floodplain as crevassesplays. Similar feafures are formed in the distributary channels of deltas.

Crossline SeeInline.

Datum On land, there may be appreciable topography across a survey area. Furthermore, if explosive chmges are used as the source then they may be buried at varying depths across the survey. Also, there will usually be lateral changesin the thicknessofthe near-surfacelow-velocity layer.To remove all these complications, a reference datum is selected and corections are made to travel-times so that they represent what would be rccorded if shots and rcceivers were placed on the datum surface, it being assumedthat thereis no further low-velocity layer below this surface.For onshoresurveys,the datum surface may be flat or it may be a more or less smoothed version of the ropography. Offshore, mean sealevel provides a convenientdafum. Decimation Reduction of the number of traces in a dataset by systematic removal of, for example, every other trace.It is not restricted,as rhe name suggestsit might be, to the removal of every tenth tmce.

Deconvolulion The seismic trace can be thought of as the result of convolving the eaith reflectivity with a wavelet. Ideally, we would like to have a seismicsourcewhich gaveout a single sharpspike signal, and record a reflectedsignal which was simply a seriesof spikes.Unfortunately,any real sourcewill emit a signal offinite length.Furthermore,the sourcesignal willbe modilied as itpassesthroughthe eafih, because of absorption,scattering,and other causes.The rcsult is that our tecorded signal will be the sum of the rcflections of a wavelet from the seriesof subsurfacereflectols (fig. A2.1). Mathematically,this process is represented by the corNolution of the wavelet with the eafih reflectivity. Deconvolution is a signal processing step that attempts to undo this convolution, to leave us with the earth reflectivity series,thus improving the resolution of the seismic data. Delta A depositional system formed where a river supplies sediment to a coast, forming a shorcline protu berance. The input of mud and sand from the river is reworked by marine processes (wave and tide action). The components of a delta system (from onshorc to offshore) are: (l) the delta plain, largely sub-aerialand consistingof distributary channelsseparatedby inter-distributaryareasof marsh and swamp; (2) the delta front, where fluvial and marine processesinteract and form features such as sand bars at the distributarymouths; (3) rhe prodelta,a zone of quiet deposition,typically ofmud and fine silt, from suspension.As time goes by, if sediment continues to be supplied, the delta builds outward from the coast.In vertical succession,the prodelta muds will then be overlain by progressively shallower-waterdeposits,genenlly with higher silt and sandcontent.

2OO -

Appendix 2 Glossary

0.1 E

^

-

Recordedreflection Earthreflectivity

-0.3 -0.4 TWT(ms) Fig.42,1 Earthreflectivity, in red, showsthat therearethreereflecto$. The tracein purpleshows what would actuallybe recorded;eachreflectorsendsbacka signalrepresentingthe wavelet,scaled in amplitudeby the rcflection strength,andthe final outputis the sumof thesethreesignaltrains.It is hard to visualisethe earthreflectivity from the recordedtracewhenthe reflectorsareclosely spaced.Deconvolutiontries to removethis confusion.

Depthmigralion An implementationof migrationthat allowsfor my-bendingat interfaceswherethe seismicvelocity changes.Thesimplesttypeof migration(timemigration)cancopewith mild lateralchangesin velocity by changingthe migrationvelocity (andthereforethe cuNatureof the hyperboloidalong which data are summedin the Kirchhoff method).Wheretherearerapid lateralchanges,suchasin the presence of steeplydipping interfacesacrosswhich there are large velocity changes,then the sudacealong which dataought to be summedwill no longer be a hyperboloidand has to b€ calculatedin detail by tracingthe Fopagationof raysthrougha subsurfacemodel.This is substantiallymoredemanding in computationtime. It is alsomore difficult to build the conect subsurfacevelocity model,because changesin velocity at any onelevel will affect rays traversingthat level in waysthat may be hard to visualisewithout detailedcomputation. DHI A Dircct HydmcafbonIndicatoris anyfeatureon seismicdatathat givesevidencefor the presenceof hydrocarbons,and is particularly useful in reducingthe dsk associatedwith dritling an explofation well. 'Ilpical exarnplesare an amplitudeanomalyon the top-reservoirreflectorthat showsconformanceto stuctwe, or aflat spot, ahofizot\tal reflectordueto a gas--oilor oil watercontactthat cuts acrossbedding-plane reflectors. Dip moveout(DMo)correction In thesimplestapprcachto stacking,traceswith thesamesouce-rcceivermidpointareaddedtogether aftercorection for NMO. Theoutputstackedtraceis thenteated asbeinga lower-noiseversionofthe tracethatwould berecordedif thesourceandreceiverwerecoincidentat themidpointlocationusedto selectthe input taces. For reflectionfrom a dipping interfacein the subsurface,the reflectionpoint is not vertically belowthesource-receivermidpoint.It is displacedupdip,by anamountwhich increases as the source-receiverdistanceincreases.Stackingtraceswith the samemidpoint thereforeadds togetherfaces that havebeenrcflectedfrom differcnt points in the subsurface,inevitably smearing theimage.DMO processingcorrects.forthis effect.It carriesout a partialmigrationof thedata,moving energyfor a reflectingpoint from the locationseenon an actualfinite-offsettraceto the locationthat

2o1 -

2 Glossary Appendix it would have for a zero-offset ftace. This involves shifting the eneryy both laterally and vertically. After DMO has been applied, traces can be stacked together at each CMP surface location without smearing the image, because all the traces contain reflected energy ftom the same subsuface point.

Dipsection Seismic section shot parallel to the predominant geological dip direction, and so in the direction of maximum horizontal gradient of reflecting horizons and often perpendicular to the principal faults. Fold As can be seenin fig. 2.7, seismic acquisitionis normally laid out so that tracesrecordedat different sourcFreceiver offsets can be added together ('stacked') to enhance signal to noise ratio. (They will, of course, require correction to be made for the morc oblique ray-paths at the longer offsets.) The number of individual source-receiver pairs that contdbute to the stack is called th€ fold of the data. Around the edge of a survey area, the fold decreasesbecause there will be progressively fewer long 'full offsets available. The fold' arca is therefore surounded by a zone where the fold progressively tapers to l. Because of the increased noise level, the data in the zone of reduced fold aJe of limited use and are often largely excluded from the migration process.

Fresnel zone If we think in tems of rays, then a rcflection comes from a point on a reflecting surface. In terms of wave theory, however, a reflection is made up of energy retuming fiom a finite area ol the reflector. A Fresnel zone is the area ftom which reflected energy arriving at a receiver has a phase difference of no more than half a cycle, and is therefore able to contribute consrucdvely to the reflection. For the case where source alld receiver are coincident and a distance lr above the rcflector, then most of the reflected energy comes from a circular zone of radius r given by 12 : Ir/2, where )' is the wavelengthof the seismic signal. GPSpositioning The Global PositioningSystemdependson time-rangingto a setofsatellites, which are distributedin various orbits so that a user can receive signals from at least four of them at any point on the Earth's surface at any time. Since the satellite positions in their orbits are known to high accuracy, four rangemeasurementsare enoughto calculatethe user'slatitude, longitude and height, plus the timing offset between the user's clock and the satellite system clock. For the highest accuracy, differe[tial GPS (DGPS) is used, where a lixed station monitors its apparent GPS position and the deviations from its klown location are used to refine the apparent position of a user in the field. In this way it is possible to correct for uncertainties in orbital parameters, atmospheric refraction, and deliberate signal de$adation by the system'sowners,giving a positional accuracyof, for example,2 m within 2000 km of the reference station. lnline A 3-D seismic survey consists of traces on a rectangular grid. One of the axes of this gdd is called the inline direction and the other is called the crosslile. Lines in these two directions are numbered, and then co-ordinatesof haces within the survey can be specifiedby meansofthe (inline, crossline) co-ordinate pair. The choice of which direction is called inline and which crossline is sometimes arbitrary. The inline direction may refer to the original shooting direction for the survey, but a bin grid may be used for processing that is not simply aligned to the original acquisition layout.

lnvasioneftects When a borehole is drilled, the drilling fluid will penetrate into the formation being ddlled. When wireline logs are run to measure formatron density and seismic velocity, for example, then they may

2O2

Appendix 2 Glossary record not the true fomation values, but rather those in the zone arcund the borehole where the drilling fluid has rzvaded the formation, replacing the original fluid. Particularly where oil and gas reservoirs have been drilled with a water-based mud, these invasion effects may be signiflcant and will need to be correctedbefore using the well data as a basis for seismicmodelling_

Lacustaine A lacustrine system is one where sediment deposition occurs in a lake. At the present day, only about l7o of the Earth's land surface is covered by lakes, though inland seas such as the Black Sea have been large lakes at times of lowered sealevel. Levee Levees are ddges built on either side of a river chanrlel- They are typically formed from coalesc, ing crevassesplays, and consist of fine-grainedsandsand silts. Similar featuresare seenalong the distributary channels of a delta plain, and along deep,water channels of a submarine fan. Loop A seismic loop is a single wiggle of a seismic trace,from one zerc-crossingto the nex!. Migralion Suppose we make a set of seismic records across an area by keeping the source and receiver together and moving the combined source receiver point around on a reguiar gdd. We could then plot the recorded seismic traces vertically downwards at the proper position on a map of the grid, creating a 3-D volume ofseismic traces.This would not give us a conect pictwe ofsubsurfacereflectorgeometry, becausethe reflection points are not in reality vertically below the sourca-receiver point. If we traced mys Aom a sourceposition, propagatingin all directions into the subsurface,we could find the one that hits a given reflector at right angles.This ray will be reflected back, exactly retracing its path, until it arrives at the receiver We could also determine the time that the ray would take along this path. Now, what we would like to do is to rearrange the traces in our 3 D volume so that the reflected signal at this travel-time, on the as-rccorded trace plotted below the receiver location, is moved laterally and vertically to the real location in spaceofthe reflection point. Migration is this processofmoving the as-recordeddata to the corect location in space.In reality, ofcourse, seismicdata are recordedwith a range of source receiver separations. Ideally, each one should be migrated separately, although it is common to cut down on the computation effofiby stacking databefore migration, which transforms them to the travel-time that they would have for zero sepamtion between source and rcceiver and then sums them. There are fufiher complications in practice becausethere may be several rays from one surface point that hit a given reflecting surface at dght angles, perhaps in widely separated locations.

Migration aperture In order to get a satisfactory subsurface image from the migration process, it is necessaryto have avail able a volume of tracessunounding the point at which we want the image. One way of implemenfing migration is to use the Kirchhoff approachdescribedin section 1.2; for every point in the output image, we want to sum data along a hyperboloid sudace in the unmigrated data set. The aperture is the lateral extent of the traces that we take into this summation. Clearly, a very small aperture (a few traces) will do little to reposition data: on the other hand, a large aperture will include very distant tracesthat will have little influenceon the result becausethe signal at long favel-times will be small. A rule of thumb is that the aperture needs to be twice the lateral distance over which reflections will be moved; a larger aperture is therefore needed for steeper dips, particularly on deep reflecton. When planning a 3-D survey, it is impofiant to acquire the data around the edge of the area of interest that

2O3 -

Appendix 2 Glossarv

<+ 1jne (ms)

Minimum

Zerc

s 'nne (ms)

Minimum

Zeto

Fig, M.2 Two examplesof zero- and minimum-phasewavelets.In each pair, the zero- and minimum phasewaveletshave the sameamplitude spectrum.

will be neededfor thc migration aperture.This might typically add a 2 km fringe otl each edgc ofthe area that is to be well imaged.This makessmali 3-D surveysrclatively expensive;to image an area 9 km r 9 km (81 kmr) would requireacquisitionover 13 km x 13 kn (169 kmr). Migrationvelocity A velocity field used to migrate the seismic data 10 obtain a well focussed and conectly located image-This is usually closerto seisnic velocitiesin the real earththan stackingvelocities,particularly where the velocities have been determired in the courseof pre-stackdepth migration which allows for the complexities ol ray-bending in the overburden.It may still be quite strongly affected by anisotropy,however, and not be well suited to depth-convertingpicked hoaizonswithout further adjustmentlo tie the well data; lateralresolutionofthe velocity field mav also be al1issuefordetai]ed well ties-

Minimum nhase There is an infinite number of seismicwaveletsthat sharethe sameamplitude spectrum.One ofthese is the minimum-phasewavelet,which is constructedso as to start at zero time and then have as much energyas possibleat the earliesttimes. In practice,the wavelet will have a maximum value in the firsf or sccondloop (fi8. A2.2). This type of wavelet is close to what is often generatedby real physical sources.Sornetimesseismic data are processedto minimum-phase final output. In theory, seismic reflectionsshould then be picked at the zero crossingconespondingto the staft ofrhe reflectedsignal from the particular interface.In practice,they are usually picked at a maximum excursion on either the first or secondloop down from this zero-crossingpick, dependingon where the most consistent loop is fbund. Potentially, this can cause confusion about the polarity of rhe data if the picking philosophy is not caretully documented.A further disadvantageof interpretingminimum-phasedata is that interactionsbetweenrcflectionsfiom closely spacedinterfacesare not as easy to visualise as tbr the zero phasecase.

2O4

Appendix 2 clossary Moveout Moveout can refer in severalcontextsto the way that the arrival time ofa signal (e.g.a reflectionfrom a particularhorizon) changessystematicallyacrossa set of seismictraces.In the particularcasewhere the source receiver spacing expandssymmetrically about a common midpoint (fig. 2.7), then the increasein travel-timefrom the zero offset caseis called the normal moveout (NMO). Correctingfor NMo is essentialbelore tracescan be stackedtogether This NMo correctionwill shift the individual samplesof a finite-offset hace upwards by an amount that decreaseswith increasingtravel-time.In consequence,far-offset traceswill be stretched(NMo srretch)and the.efbre shifted towards lower frequencies.At shallow depth and long offset, the effect may distort traces so badly that they are unusableand have to be rcmoved (mute(lout) before stack.

ms Abbreviation for millisecond (l/1000 of a second).

(reflection) Multiple A seismicsignal that has undergonemore than one reflection.For example,the signal from a marine sourcemight be reflectedfrom the seabed,reflectedagain from the sea-surface,then travel down to a deep interface where it would be reflectedback to a surfacereceiver It will obviously arrive later than a signal that has gone straightfiom sourceto deepinlerface to receiver(the primary refiection). Similar multiples can be created by bouncing the signal between any two interfacesoverlying the deep interface,and there could be two or moaebouncesbetweentheseshallowerlevels.not iust one. Deep interfacescan thereforeeasily becomeobscuredby multiples ofinterlaces lying abovethem, if the primary from the deepinterfacearrivesat the sametime asthe multiple ofthe shallowerinterface. For this reason,elimination ofmultiples is often a key objectiveof seismicprocessrng.

Mute Reflection ftaces recorded at long offset and shofi travel time will be strongly contaminatedby varioustypes ofunwanted signal, suchasrefractions,and willbe distortedby the applicationofNMO corection which stretchesthe individual loops. They are usually removedbefore stack by settingto zerc all tracevaluesfor oflsets beyonda specifiedoffset-Twr curve (the mute). sometimesan inner trace mute is employed also; this setsto zero all trace valuesfbr offsetsless than a specifiedoffsetTWT curve, usually to remove traces heavily contaminateclby mulriples, which are often poorly handled at short offsets by demultiple proceduresthat rely on NMO differencesbetweenpdmaries and multiples.

t'tM0 See Moveout. 0ffset Sourceto receiverdistance. Pay (zone) Hydrocarbon bearing reservoir;in geophysicaldiscussion,usually irrespecriveof whetherhydrccar_ bons are producible,economically or at all. PhasesDeclrum When a signal is describedas the sum of a seriesof sine waves by the methodsof Fourier analysis, each component sine wave has an amplitude and phase.The amplitude spectrumdefinesthe peak_ to-peak amplifude of each sine wave as a function of frequency.However, this is not enough to deline the signal. We need also to know what the time alignment of the various sine waves should be. This might for example be seen in the time of the first maximum after time zero. The phase

2OS

Appendix 2 clossary spechumdefinesfhis alignrnenLA phaseofzero meansthat the componentsine wavehas a maximum at zero time, a phase of 90' means a zero-crossingat zero time, and a phase of 180. means a minimum at zero time. The phasespectrumis a plot ofthe phaseof the componentsine wavesagainst frequency.

Pointbar Where a river bends,the maximum flow velocity is close to the outer bank. At the inner bank, the flow is less and sedimentaccumulates10form a point bar. that grows by lateral accretion.Typically the depositsare sands,perhapswith some mud in the upper part.

(retlection) Primary Signal that has travelleddirect from sourceto reflecting interfaceto receiver,in contrastto multiples (q.v.) with theirmore complex pathsinvolving multiple bounoes.The primariescany the information we needto createan image of subsuface structure.

gather Receiver A collection ofthe tracesrecordedat a given rcceiver from all the various shot points that have been recorded at that receiver.This is sometimescalled a common rcteiNer gdther_Creation of such a gather involvesre ordering the tracesrecordedin the field, which will be organisedas common-shot gathers,i.e. the collection oftraces recordedat ali the different receiversfrom each shot. Reflection coefficient When a seismicwave ofamplitude A is incident on an inteface betweentwo different media, it is in general partly rellected and partly transmitted.If the amplitude of the reflectedwave is R, then the reflection coefficient is delined as the ratio R/4. A negativevalue indicatesthat the reflecterlwave is Iu0 out of phaie wilh the Incidenrwa\e.

Refraction record To correct Iand seismictracesfbr static shifts generatedby latenl variationsin the near surfacestructure beneathshotsand receivers,we needto know the thicknessesand velocities of the near-surface layers. Usually, there is a low velocity weatheredlayer near the surface,overlying a higher-velocity layer. To investigatethe thicknessof the low-velocity layer, we can shoot refiaction profiles. These consist of long lines of receivers with a source at each end. With this geometry, the first arrival at each receiver will usually be the head wave. often called a refraction. This travels along the top of the high velocity layer. Geometrically we can think of it as being predicted by Snell's Law, which says that a ray travelling through the interface will be bent (refracted)away from the interface normal. As the angle of incidenceof the ray is increased,there wjll come a point, the critical angle, at which the ray bending would make the ray in the second medium travel along the interface. (The full theory that explains the amplitude of the head wave and its spatial variation is much more complex.) By analysingthe vaiation in head wave arrival times from one receiverto another, it is possible to map the changes in thickness of the low-velocity layer above the high velocity refractor.

Root-mean-square (rms)average The root-mean-square(rms) averageof a set of numbersjs the squareroot of the arithmetic average of their squares. Seismicwaves The most important type ofseismic wave is the P wave, which is an ordinary soundwave.As it moves through the rock, individual particlesmove backwardsand forwards in a direction parallel to that of wave propagation.The other type of wave that can exist in the body of the rock is the shearor S-wave,

20,6 -

Appendix 2 clossary in which particle motion is perpendicularfo the propagationdirection.Both types havevelocitiesthat dependon the elastic moduli and d€nsitv ofthe rock:

v,:[(k]4plJ)/plu2 V"- fut/Plt/2 where f is the bulk modulus, p is the shearmodulus, and p is the density.Other types of seismic wave exist, confined to the vicinity of layer boundaries,but they are not important to the interpreter of seismic data.

Shear modulus Shear deformation of a material involves change in shape without change in volume. The shear modulus ofa material is a measureofits resistanceto shearstress.For liquids, which will flow freely to accommodateany change in shapeof their containing vessselthat does not involve chanqe in volume, the shearmodulus is zero.

Stack In geneml, the adding of a number of t.aces together to improve signal to noise mtio. Most often used to refer to the adding of traceswith different source receiver offsets but a common midpoint (fig. 3.7). The tracesare first coffected for the increasedtravel_timeat the lonser offsets due to the oblique havel path; this is the Normal Moveour iNMO) corecrion. They may also be correctedfor effectsof subsurtbcedip and lateral velocity variation, by some form ofmigration process. Stacking velocity The velocity field that, when used to calculateNMO conection, gives the best alignmentofthe haces acrossthe CMP gather and therefore the highest amplitude in the stackedtrace. It is only loosely connectedto the actual velocitiesof seismicwavesin the earth,owing to effectsofdip and cu*ature of the reflector and the impact of lateral variations in the overburden. Statc cofieclion A static conection is a time shift that is applied uniformly acrossa particular tmce. For example, the effect might be to shift an entire trace downwardsby g ms. A neighbouringtrace might have a different time shift applied. Such shifts are typically neededwhen processingland data! to remove near-surtacetime delaysparticularto eachshotand receiverlocation,asa resultofchansesinelevation and thicknesschangesin a near,surlacelow-velocity layer

Stratigraphic trap The simplestsort ofoil or gas trap is a domal anticline, or four-way dip closure.Variantson this may require an elementof sealingalong a fault, but the possibleexistenceofsuch a Jlrrclr/41trap canbe inferred from the top reseNoir map alone. Sometimes,however,a hap requiresan elementof lateral Iithological changeto work. This might, for example,be lateral transition from sand to shale within the reservoir formatiorr. S'ch a.stratigraphic /r?p cannot be found from the structural map alone; the lateral change in rock propefties needs to be present. It may be infe.red from geological argument alone,but seismicevidencefor the requiredtransition will reducethe (normally high) risk associated with such a prospect.

Slrikesection seismic sectionshot perpendicularto the dip direction of the main reflectors,often sub paralrelto the main faults and thereforehard to interpreton 2-D seismicsectionsbecausesub-horizontalreflections at nearly the same travel-time may relate to different fault blocks on either side of the line or directly below it.

207 --

2 Glossary Appendix Trace

time' conventionally plotted with time Graph showing the amplitude of a seismic signal against different types' eg as-recorded' many of increa.ing uerti.ally divn.r'ards. The signal can be be shown by a conventional would tmce stacked, or migrated. Originally the amplitude of a information' as explained in amplitude *iggly tine, brit increasingly colour is used to convey section3.2.

Transgression

in sealevel' Landward migration ofthe shoreline'owing to relative rise

Turbidites

water flow; such currents are able to ,L turtidity current is a suspensionof sedimentin a turbulent Turbidites are the depositsofthese water' deep move coarse-grainedsedimenttar out to seaand into bedsup to severalmetres individual with turbidity cunJnts. They are widespreaddeep waterdeposits' grained sediment' in thickness,and ranging fiom coarse-to fine taken tbr a seismic signal to travel from the The Two-Way Time to a seismic reflector is the time is the usualvertical scalelbr seismicsection surfaceto the reflectorand back to the sudaceagain This display. Unconfolmity sedimentdeposition:this may be a result An unconfomity is a surfaceacrosswhich there is a gap in ofe'osionolofnondeposition.whefethetime-gapissubstantial'thepropertiesofthesedimentsare prominent seismicreflection Sometimesthe often quite different on either side of it, giving rise to a below it; suchangularunconformities unconfor-ity surtac"cuts acrossbeddingplanesofthe sediment are ollen easily recognisedon seismic displays' Well synthetic make a well synthetic From the wireline log To make it easierto tie seismicto well data,it is useful to of depth by multiplying togetherthe data in the well, acousticimpedanceis calculatedas a function log is calculated'and an expectedseismic recordedvelocity and density logs. From this a reflectivity using checkshotinformation if available' responsecalculatedby convertingit into a function of TWT 3 1 for details' and then convolving it with a seismicwavelet Seesection

Zero-ollset

at the samelocation This implies a simple A zeo-offset traceis one recordedwith sourceand receiver at right angles and is reflected back a reflector seismic geometry in which an outgoing ray strikes along exactly the samepath as it has come

Zero-phase

the same amplitude spectrum One of There are an inlinite number of seismic waveletsthat share about zero time ln practice' the wavelet will theseis the zero phasewavelet' whlch is symmetrical sidelobes (frE A22)' This is not a wavelet have a strong centml loop and a number of smaller would need to begin before the source was that could be generatedby any real source,becauseit generatedby the real source wavelet can be triggered at tiine zero. However, the seismic traces if the wavelet were zero-phase This is m;pulated by processinginto the form they would have usefulbecauseitismucheasiertoundeNtandazero-phaseseismicsection.Evelyreflectingintedace This meansthat a picked horizon producesa signal with its maxlmum value centredon the intelface (positive or negative depending on r"pres.nting an interface follows a maximum loop excursion Where reflections are closely spaced' the impedancechange), which makes autotracking simple

2@ -

Appendix 2 Glossary it is much easierto visualisehow they intcractto modify the seismicamplitudeswhendataare zero-phase. Zoeppritequations Theseequations determinethe amplitudes of the reflectedandrefractedp- andS-wavesgenerated whena planeP-waveis incidenton a planeinterfacebetweentwo mediaof differentdensityand elasticmoduli.

Index

2-D seismic,2 I D image projection, 176 8

calibration of acousticimpedanceto porosjty. I 64 CDP gather, 198

4C data,2'74

check shots,60, 198

4 - D s e i s m i c ,1 2 ,3 4 , 1 8 0

cinema-styledata display, 179 classeso1-AVOresponse,130

accommodationspace, 196

classificationalgorithms, I l2

a c o u s t i ci m p e d a n c e , 3 , 5 81, 2 0 . 1 9 6

clay content,effect on P wave velocity, 146 7

acousticimpedancelog, 59 aeoliar depositionalprocess,196

c l a y s m e a r .1 1 8

ACC. 53. 196

CMP gather 2, 199

AlabasterField, 11

coherenceslice- 87 9. I l0

aliasing, 196 a l l u v i a l f a n ,1 9 7

coherenceslice, offshorc Egypt example. Ll3 I6 c o h e r e n c veo l u m e .8 7 . 1 7 3

amplitude measurement,84

cok ging, 168

amplitude spectrum, 197

common midpoint gather.2

Amplitude VersusAngle, 198

compositedisplay,77

Amplitude VersusOftiet.50, 120

c o m p o s i t el i n e s , 7 8

a n g l es t a c k , 5 0 ,1 3 5 ,1 9 7

connecledvolune analysis, 174. 1tl9

angular unconlbrmity, 207

convertedmodes,28

a n i s o t r o p y . 9 81. 9 7

co ordinate systems,194

annual 3 D seismic acquisition,worldwide total. l2

Comorant Field, 11

altributes,72. 197

conidor stack, 68

A u g e rF i e l d , 1 3 8

crevassesplay, 199

Automatic Gain Control, 53

citical angle, l2l. 205

autotmckers,8 t+, 198

critical porosity model. 150

AVA, I98

c r o s s l i n e1, 9 9

A V O . 5 0 . 1 2 0 ,1 9 8

curvatureof horizon. 86

c l i p p i n g ,o n s e i s m i cd i s p l a y , 7 5 6

AVo classes.110 A V O c r o s s p l o t ,1 3 l . l 3 G 7 AVO grdienl, 125

data volume, 3'D surveys, 193

avulsion, 198

decimation, 199

d a t u m s .1 5 . 1 9 9 deconvolution.39. 199

209

backup of interpretationdata. 193

delta depositionalsystem, 199

Backus averaging.66

delta front, 199

band limiled inversion. 158

delta plain, 199

bin, 198

densities,individual minerals, I44

binning.43 5

d e n s i t y1 o g , 5 8

b r i g h t s p o l . 1 2 6 .1 9 8

densily velocity crossplot. 122

b u l k m o d u l u s ,1 4 0 . 1 9 8

depositionalsystems,106

2'lO -

lndex

depositionalsystems,chamelised, 107

fluid-fill elTects,135, 140 1

depositionalsystems,delta, 107

fbld,201

depositionalsystems,fluvial, 107 depth conversion, 89

fbur-componentacquisition.27 F r e s n ezl o n e , 7 , 2 0 1

depth conversion strategy. 90

f u l l f o l d a r e a .1 8 , 2 0 1

depth conversionunder fault planes,93

Fulmar sand,use ofinversion ro predict porosity,

depth conversionwith lateral variation in overbuden, 94 depth conversionwith velocity gradient,90 deprh migralion, 47, 100, 200 designature.35, 38 detectabiliryof rhin beds, 105 DHI, 125,200 dim spot, 127 dip moveout,49, 200 dip section,201

1644

gas chimney, 129 Gassmannequation, 140. 183 Global Positioning Sysrem,201 GPS,201 gradient (AVO). 130 crcenbery{astagna relation, I 48 gridding and contouring, 8l

dip/stnke shooting,24 dip'azimuth maps, 84 5 Direct Hydrocarbon Indicator, 125, 200

horizon picking, automated,81,4 horizon picking, benelif of workstation,72

dispersionofseismic velocities,59, 139, 183

horizon picking, manual, 77

display,dynamic range,75

horizon picking, on paperprints.7l

display,variable intensity,72 74 display,wiggle trace.72 75

h o r i z o ns l i c e , 7 7 .1 1 0

display scales,74

illumination display oIhorizon, 86

Dix formula, 97

image ray, 100

DMO, 49, 200 downlap geometry interpretation,I 16

immersive visualisation. 177

dry rock elastic moduli, l50 l dry rock Poisson'sratio, 142, 150 2

incidenceangle, approximateformula for. 132 infill shooting,26

impedancedepth trends, 132

inline,20l eanh rellectivity spectrum, 159

instantaneousphase, 109-l I

economic value of 3-D seismic, 12

inslantaneousvelocity, 90 intercepl (AVO), 130

effective medium approximation, 65 elastic impedance,170 elastic moduli, individual minerals, 144

intercept*gradientsection. 138 9 interpolationof shot rccords,40-1 int€rval velocity, 97

tar trace stack, 133-4 fault cutout, 79 fault displacementmapping, 86 fault juxtaposition diagram, I l8 fault pattem, establishing.78 80 fault polygons,79

i n r r \ i o n e f f e c r so n u e l l l o g ' . 5 0 . 6 5 . 1 4 5 .r 0 inversion,bandlimited, 158 inversion.full bandwidth, 163 inversion,slochastic,I6G8 inversion ofreflectivily to impedance.155 inversionofreflectivily to impedance,principles of, 1 5 56

fault sealby clay smeat I 18 faults, manual picking, 78 faults, normal and reverse, 7+80

Kirchhoff migration, 2-D and 3 D,4.7

faults, on dip and strike lines,78-9 feathering(marine cable), 25 5 flat spot, 126

knging, 168

flexibinning, 44

laIId acqDisition,30

fluid factor stack, 137-8 fluid properties, 143 4

land acquisition.geometry,3l

lacustrinedepositionalsystem,202

levee,202

2'11 -

lndex loop, seismic,202 Iow frequencymodel, use of for inversion, 156, 162

Poisson'sratio, 121 Poisson'sratio - acousticimpedancecrossplot, 123, 131

macrolayers,in inversionmodel, 160 Magnus sand, distribution from inveNion, 165 7 Inap-maLing.57 manne acquisition geometry,19 20 marine acquisition geometry,4-component,29 marine seismic acquisition, 18 marine seismic acquisition,mulri-source, multi-streamer 19 21 marine seismic acquisition, non-standard geomerry, 2 I marine seismic sources,18

Poisson'sratio, dry rock, 142,l5c 2 polarity change of reliector at fluid contact, I 26 polarity of display, 14,62 porosity, effect on impedance, for Chalk, 128-9 porosity,effect on seismicvelocity, 146 7 position-fixing inaccuracies,187 positive AVO, 131 pressurechange,eft-ecton seismic impedance,183 pre stack depth migration, 49

marine seismicvessels,19

primary rellections,41-2, 205 processingsequence,typical, 36

Maui Field timelapse example, 188-9

prodelta, 199

migration, 3, 202 migration, example,52

P w a v e ,1 2 1 , 2 0 5

migration, laleral shifts, 98

receivergather,205

migration.pdnciples,46 7 migration, target and acquisition areas,I 8

rcfl ection coeflicient. 58, 205

migration aperture. 202

refraction record. 205

migration velociry, 203

relay ramp, 85

mineral propefties, 145 minimum phase,36-7, 60, 203 misfie after 2-D migration, 6 models for oil and brine sandresponse,126 moveout, 204

rellectionsas chronostratigraphicsurfaces,105

repeatabilityof time-lapsesurveys, 186 residual moveout,elTecton AVO measurement,132 resolution,horizontal, ofseismic data, 106

multi tbld coverage,genemtion,22

resolution,of screendisplay, 72 resolution,vertical, ofseismic data, 102 reversefxults, multi-valu€d horizons at, 79 80 RMS average, 205

multiple cable/sourceacquisirion,21

rms velocity, 96

ms. 3,204

multiples, 39 42, 204

root-mean squareaverage,205

mutes,5l,204

near trace stack, I 33 4 neural net trace classification,112 14 neural nets, I 12 NMO, 2, 2M NMO str€tch, 204 normal incidencereflectivity, 125 normal moveout, 2, 204, 206

SAIL (Seismic Approximate ImpedanceLog), 157 8 sandstonevelocity porosity crossplot, 147 satelliteimaging, display combined with subsurface data, 173 scaling for AVO measurement, 132 SchiehallionField time lapseexample, 188 seedpoints, 83 SEG polarity convention, 14

oceanbottom receivers,27 ofrset, 2M

seismicfacies map. l14 seismicsection,3 seismic time datum. l5

P to S conversion,28

seismic trace,2

paftvane, 23 partial offset stacks, 133

semi lransparentseismic cube, 175 7 sequencestratigraphy,107 shalegouge ratio, ll8

pay zone, 2M peak (seismic loop), 14 phasespecfum, 60-1, 204 point bar, 205

shenrmodulus, 140. 206 shearvelocity, prediction from P velocity, 148 y shearwave, 121, 205

212 -

lndex

shearwave acquisition,26 8

t w o - w a yt i m e , 3 , 2 0 7

shot, 2

TWT, 3. 207

shol interpolation,40 S n e l l ' sL a q 1 2 1

unconformity, 207

sonic log, 58 sparsespike representation,160

undershootingobstructions,26 7 unit conversions,15

sphedcaldivergenceconection, 39

units, systemsof, 14

stack, 2, 46, 206

uphole times, 33

stacking conflict, 48 stacking velocity, 96, 206

variogram, 166

static corrections,33, 206

velocity, labomtory measurement,183

static delays,eftect on srackingvelocity, 9?

velocity, rms, 96

stochasticinversion, 16G8

velocity, stacking,96, 206

stratalslice, 110

velocity iensity relations, 146-8 vertical exaggeration, 74

stratalslice, offshore Louisidna example, 113 15 stratigraphicuap, 206

Vertical Seismic Profile, 66

strike section.206

vibrator (land seismic source),30

strike shooting,preferredover salt dome, 25

visionarium, 178

structuraltrap, 206

visualisation,172

structure-confbming amplitude, 126

Voigt Reuss Hill averageelastic modulus. 1rl4 voxel, 174

struclure-orientedfi ltering, 84 S - w a v e ,l 2 l , 2 0 5 syntheticseismogram,58 56

voxel picking. 174 VSP,66 VSB exampletrace displays,69-70

time depth relation,60 time-lapseseismic 172, 180

VSB upgoing and downgoing waves,66 VSP, watk-above, 68

time-lapseseismic,changesin reflector ampliiude, 1825, 190 1

VSP geometry, 6G7

timelapse seismic,changesin reflector TWT, 184 timelapse seismic,matching surveys, 187

wavelet,zero and minimum phase,37,203 wavel€t extraction,53-4, 62, 161

time lapseseismic, positioning repeatabilityeffecrs,

wavelet scaling in inve$ion, 162

186

weathe.ing layer, 34

timelapse seismic. when to shoot, 184

wedge rnodel, 103-5

time migration, 47, 100

well depth datum, 15

t i m e s l i c e ,7 6 7 , 1 1 0 t^ce,201

well log editing, 64 well synthetic,58,20?

transgression,20?

w e i l t i e s , 5 7 ,1 6 0

transition zone survey,34 trough (seismic loop), l4 tuning, amplitude and thicknessresponse,104 5, 129

well to seismicmatch,53 well to seismicmatch, causesofmismatch, 64 well trajectories,display with seismic volume, 173 Wyllie time-averageequation. 146

tuning, and b€d thicknessestimation, 104 tuning, examplesection, 106

zero oftiet,3,20?

turbidites, 207

zero-phase,14, 3G7, 53. 60, 203, 2U Zoeppritz equations,121, 208

two-passmigmtion, 48

,t|llt|u