50 Oilfield Review

From the earliest days of exploration,prospectors associated salt with oil andgas—but not always for the right reasons. Inthe 1920s, so many successful wells weredrilled around salt domes that logging meth-ods were tuned to identify the high-salinitywater in formations overlying pay zones.1By 1923, gravity and seismic methodsbecame successful in spotting salt domes,and the industry was on its way to under-standing the structural role played by salt.Today, interpreters can view and tour saltstructures with the help of powerful graph-ics workstations (next page, top).

Salt is one of the most effective agents innature for trapping oil and gas: as a ductilematerial, it can move and deform surround-ing sediments, creating traps; salt is alsoimpermeable to hydrocarbons and acts as a

For help in preparation of this article, thanks to MarkBogaards, Cliff Kelly and Mark Puckett, Wireline & Test-ing, Houston, Texas, USA; Bob Godfrey, Colin Hulme,Tore Karlsson, Jane Lam, Dominique Pajot, John Ulloand Öz Yilmaz, Geco-Prakla, Gatwick, England; George Jamieson, Geco-Prakla, Houston, Texas; and Ron Roberts, Amoco, Denver, Colorado, USA.

Paul FarmerGatwick, England

Douglas MillerRidgefield, Connecticut, USA

Andy PieprzakJeff Rutledge Richard WoodsHouston, Texas, USA

Exploring the Subsalt

60°

30°

0°

30°

nDistribution of offshore salt sheets. [Adapted from Ward RW, MacKay S, Greenlee SMand Dengo CA: “Imaging Sediments Under Salt: Where are We?” The Leading Edge 13,no. 8 (August 1994): 834.]

Advances in seismic imaging have changed the way explorationists

view salt bodies. Once seen as impenetrable barriers to geophysical

probing with some flanking pay zones, many salt structures are now

proving to be thin blankets shielding rich reserves. Geophysicists are

developing new methods to see through salt, illuminating the reservoirs

below. This new vision of subsalt is impacting E&P decisions from well

planning and drilling to field delineation and development.

1. Allaud LA and Martin MH: Schlumberger The Historyof a Technique. New York, New York, USA: JohnWiley & Sons (1977): 68-69.

2. Western PG and Ball GJ: “3D Prestack Depth Migra-tion in the Gulf of Suez: A Case Study,” GeophysicalProspecting 40 (1992): 379-402.

Charisma seismic interpretation system, KUDOS 3Dvelocity modeling system, and SALTBOND cement sys-tem are marks of Schlumberger. CM-5 is a mark ofThinking Machines Corporation. GeoDepth is a mark ofParadigm Geophysical. InDepth is a mark of WesternGeophysical.

51Spring 1996

nFlying through a seismic interpretation. The top of a salt feature (yellow surface) has been interpreted on a seismic workstation. Alsoshown is a panel of seismic data (background), a reflector above the salt (brown surface), seismic velocities at vertical well locations(multicolored vertical logs) and deviated well trajectories (blue lines).

Passive—No Space Problem

Active—Diapir Creates Space

Reactive—Extension Creates Space

Thinning,arching

Radial orsubparallelfaults

Fan ofnormalfaults

nStyles of salt intru-sion. When the over-lying sediments offerlittle resistance (top),salt can rise, oftendragging flankinglayers up with it. Ifthe overburden doesresist, salt pressuredfrom below (middle)can still pushthrough, doming theoverburden and cre-ating radial faults inthe process. In thecase of regionalextension (bottom)faulting in the rigidoverburden can openthe way for salt torise. [Adapted fromJackson MPA, Vendev-ille BC and Schultz-ElaDD: “Salt-Related Struc-tures in the Gulf ofMexico: A Field Guidefor Geophysicists,” TheLeading Edge 13, no. 8(August 1994): 837.]

seal. Most of the hydrocarbons in NorthAmerica are trapped in salt-related struc-tures, as are significant amounts in other oilprovinces around the world (previous page).Many reservoirs in the North Sea are belowsalt, as are large fields in the Gulf of Suez.2

A product of seawater evaporation, saltaccumulation can reach thousands of feet inthickness. Salt retains a low density of2.1g/cm3 even after burial. However, the sur-rounding sediments compact and at somedepth become denser than the salt—anunstable situation. If the overlying sedimentsoffer little resistance, as is sometimes thecase in the Gulf of Mexico, the salt rises, cre-ating characteristic domes, pillows andwedges that truncate upturned sedimentarylayers (right). If the overburden does resist,salt can still push through, creating faults inthe process. If tectonic conditions are right,extensional faulting in the rigid overburdencan open the way for salt ascent. Much ofthe Zechstein salt pervasive in the North Seahas been mobilized this way.

In contrast to salt’s low density is its highseismic wave velocity—4400 m/sec (14,432ft/sec)—often more than twice that of sur-rounding sediments. The strong velocitycontrast at the sediment-salt interface actslike an irregularly shaped lens, refracting and

reflecting seismic energy. Early data process-ing techniques treated this contrast like amirror, resulting in images that portrayed saltfeatures as bottomless diapirs extending tothe deepest level of seismic data (left). In the1980s, seismic processing began to correctlyimage the steeply dipping and sometimesoverhanging faces of salt where hydrocar-bons could accumulate.

But in the last five years, a new image ofsalt has emerged. In some areas, not only isthe top of salt clearly visible, but the bottomalso. Geologists hypothesize that in theseareas of allocthonous salt—found awayfrom its original depositional position—con-ditions allow the salt, having reached verti-cal equilibrium, to begin flowing horizon-tally (above). In the Gulf of Mexico, thisoccurs mainly in deep water beyond thecontinental shelf, where sediment cover isnot as thick as it is near shore (bottom left).Wells drilled through thin salt sheets haveencountered oil-bearing sediments below.

However, knowledge of the existence ofhydrocarbons below salt is insufficient rea-son to start drilling. Drilling salt is risky (see“Drilling and Completions Through Salt,”page 54). The salt itself is weak and under-goes continuous deformation. Belowintruded salt, sediment layers are often dis-rupted and overpressured. And most impor-tant, unless seismic data have been pro-cessed to image through the salt, theposition of the target is unknown.

52 Oilfield Review

nEarly imaging results around salt. Seismic data processing resulted in images of bot-tomless salt diapirs (left). Enhancements in processing began to correctly image thesteeply dipping and sometimes overhanging faces of salt (right). [Reprinted with permis-sion from Ratcliff DW, Gray SH and Whitmore ND: “Seismic Imaging of Salt Structures in theGulf of Mexico,” The Leading Edge 11, no. 4 (April 1992): 15 and 22.]

Houston

Lake Charles

New Orleans

Enchilada

Mahogany

Teak Gemini

MickeyMouse

DiscoveryPlugged and abandonedSalt sheets

nSalt sheets mapped in the Gulf of Mexico. Recent exploration wells correspond to wells mentioned in table (next page).

1 2 3 4

Evolution of a Salt Wall

1 2 3 4

Evolution of a Salt Diapir

nEvolution of salt intrusions. Salt walls and diapirs are initiated at instabilities on extensive salt layers. As the salt rises and thenflows horizontally, the walls and diapirs change shape. Eventually some salt features become completely detached from the parentsalt layer.

Two-

way

tim

e

Distance Distance

0

1.0

2.0

3.0

Two-

way

tim

e, s

ec

A few operators have announced signifi-cant oil discoveries beneath salt in the Gulfof Mexico, rekindling a spirit of explorationin the Gulf. Phillips Petroleum Company, inpartnership with Anadarko Petroleum Cor-poration and Amoco Production Company,announced the first commercial Gulf ofMexico subsalt discovery with theMahogany prospect in 1993, and attributedthe success to the imaging technique calledprestack depth migration.3 Drilled in 375 ft[114 m] of water to a depth of 16,500 ft[5030 m], the well produces from sedimentlayers beneath a salt sheet 3000 to 8000 ft[915 to 2439 m] thick.

Since the Mahogany find, many morewells have been drilled in the area, withother operators experiencing similar suc-cess (left). Before prestack depth migration,the success ratio in the subsalt play wasaround 5%. The new technique is increas-ing that to 25%. Depth migration is alsobringing first-time details to light in some ofthe many North Sea reservoirs that producefrom below salt, and operators plan explo-ration campaigns in the Red Sea using thesame method.4

What is this imaging technique and howdoes it help illuminate subsalt reservoirs? Theanswers are found in a review of the familyof imaging methods, including prestackdepth migration, that are bringing subsaltand other complex structures to light.5

53Spring 1996

3. Westcott ME, Leach MC, Wyatt KD, Valasek PA andBranham KL: “Mahogany: Seismic Technology Lead-ing to the First Economic Subsalt Field,” ExpandedAbstracts, 65th SEG International Meeting and Exposi-tion, Houston, Texas, USA (October 8-13, 1995):1161-1164.

4. Salpukas A: “Anadarko Planning to Drill in Red SeaUsing Computers,” New York Times, September 29,1995.

5. For more on subsalt imaging topics: The Leading Edge13, no. 8 (August 1994).

Subsalt Scorecard in January 1996

Mickey MouseMississippiCanyon 211

Prospect

MahoganyShip Shoal 349

Amoco 1South MarshIsland 169

MesquiteVermillion 349

Ship Shoal 250

TeakSouth TimbalierAddition 260

Ship Shoal 360

Ship Shoal 368

South Timbalier 289

Exxon andConoco

Operators/Partners

Phillips, Anadarkoand Amoco

Amoco

Phillipsand Anadarko

Japex and Vastar

Anadarko(originally withPhillips)

Unocaland Conoco

Amerada Hessand Shell

Consolidated Natural Gas andLouisiana Land & Exploration

Noncommercialdiscovery

Result

Commercialdiscovery

Dry hole

Dry hole

Dry hole

Potentiallycommercialdiscovery

Pluggedand abandoned

Dry hole

Dry hole

1991

Date

1993

1993

1994

1994

1994

1994

1994

1994

EnchiladaGarden Bank 128

South AnaVermillion 308

Garden Banks 119

AlexandriteShip Shoal 337

MonaziteVermillion 375

AgateShip Shoal 361

South Timbalier 231

North LobsterSouth Timbalier 308

GeminiMississippiCanyon 292

No NameSouth MarshIsland 97

Bald PateGarden Banks 260

Shell Offshore Inc.,Amerada Hessand Pennzoil

Amocoand Vastar

Oryx

Phillips, Anadarkoand Amoco

Anadarko

Phillipsand Anadarko

Louisiana Land & Exploration,Anadarko and Agip

Marathon

Texacoand Chevron

Pennzoil, OXYand Total

Commercialdiscovery

Dry hole

Dry hole

Dry hole

Drilling in 1996

Drilling in 1996

Drilling started in 1995

Potentiallycommercialdiscovery

Drilling in 1996

Developmentunder way

1994

1994

1995

1995

Spuddedin 1995

1995Oryx andAmerada

Commercialdiscovery

1996

1996

1995

Hydrocarbons present

nSubsalt drilling scorecard in the Gulf of Mexico. Since the successful well drilled byPhillips and partners in 1993, subsalt exploration in the Gulf of Mexico has blossomed.[From Taylor G: “Subsalt Returns to the Top,” AAPG EXPLORER 17, no. 2 (February 1996): 8.]

(continued on page 56)

54 Oilfield Review

Properties of salt—pseudoplastic flow under sub-

surface temperatures and pressures, and low per-

meability—that make salt bodies effective hydro-

carbon traps also present unique challenges for oil

and gas operators (above). Special considerations,

from selecting drilling fluids and bits to imple-

menting casing programs and cementing proce-

dures, are required to produce long-lasting wells.

Methods developed on the US Gulf Coast and in

the Gulf of Suez, Egypt have improved the effi-

ciency and reliability of drilling and completion

operations in thick salt sections.1

Unlike typical sediment sequences in which hor-

izontal stresses are less than vertical stresses

from overburden, salt is like a fluid, with stresses

in all directions approximately equal to the over-

burden. Therefore, if borehole fluid pressure is

less than in-situ salt strength, stress relaxation

may significantly reduce openhole diameters. In

some cases, relaxation and salt creep can cause

borehole restrictions even before drilling and com-

pletion operations are finished. Undergauge bore-

holes can lead to stuck drillpipe, problems running

casing and ultimately casing failures—ovaling,

bending or collapse.

To maintain near-gauge boreholes, drilling flu-

ids must minimize hole closure and washouts.

Water- and oil-base muds with saturated and

undersaturated salt concentrations, and synthetic

fluids have been used to drill salt, but no single

system works all the time. Water-base muds with

low salt concentrations try to balance salt erosion

and dissolution with creep rate to maintain hole

size. However, because salt creep and dissolution

change across thick salt sections, this can be

problematic and hole size may vary with depth.

High-salt-concentration, water-base muds dissolve

enough salt to offset creep, but can become under-

saturated at high temperatures and enlarge the

hole. Oil and synthetic muds prevent dissolution

and can be used effectively in salt, but are expen-

sive, can leach water, gas and other mineral inclu-

sions out of salt and may not offset creep.2 Eco-

nomic, easy to maintain and adaptable

salt-saturated, water-base muds are often used.

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa aPotential Problems Casing Strings Wellbore Displacement

Salt

Radial stress relaxation

Salt creep ledgesimpinge on drillstring

Borehole wall weakenedby leaching water, gasand other mineralsout of salt

Wellbore enlargementresults from saltdissolution

Accumulated cuttingsjam drillstring

Salt

Salt

Caprock

Shear zone

Saltflow

150°F

200°F

Potentialoverpressure

Unconsolidatedzone

nSpectrum of challenges in subsalt drilling and completion. Drillers have to address factors that cause openhole instability and accompanying problems, includingborehole walls weakened by incompatible muds, restrictions and undergauge hole caused by salt creep, or enlargement due to dissolution (left). In rapidly moving salt,liners cemented inside cemented casing reduce radial pipe deformation and so increase wellbore resistance to nonuniform loads (center). During the life of a well, saltmovement can displace wellbore tubulars, possibly causing casing failure or restricted access (right).

Drilling and CompletionsThrough Salt

Salt is weak and soft, so polycrystalline dia-

mond and other mill-tooth insert cutters, which

make hole by scraping, are used. Stronger inserts

may be needed to penetrate caprock formed on the

top of some salt layers by groundwater leaching of

minerals. Side-cutting, eccentric or bicentered

reamers above bits have been proposed to open up

hole diameters that are larger than the bit and

allow for some salt creep before the borehole

becomes undergauge.3

After drilling into salt, heavier than expected

mud weights may be needed to control salt flow.

Drilling speeds vary among operators, but reason-

ably fast penetration rates—60 to 150 ft/hr [18 to

46 m/hr]—are required, so wells can be cased

quickly. Good hole cleaning and periodic back-

reaming, however, should not be sacrificed just to

make hole faster. Circulating a small volume of

fresh water can remove salt restrictions and free

stuck pipe, but care must be used to prevent

washouts. Enlarged or undergauge holes make

directional control difficult.

Thick salt bodies can affect temperature and

pressure in surrounding formations. Salt thermal

conductivity is high compared to other sediments,

so overlying formations are heated and underlying

formations are cooled. Because salt is a barrier to

basin fluids, if outward flow is insufficient to

achieve normal compaction, high pressure may

develop below salt.4 As disrupted sediments

below salt are penetrated, fluid losses or flow

can occur, depending on mud weight and forma-

tion pressures, unless drillers proceed slowly

and carefully.

Washouts, restrictions, ledges and moving salt

exert nonuniform loads on casing.5 Increasing wall

thickness offers better resistance to these loads

than higher yield strength steels, so heavy-wall

casing can be used if salt creep rates are low and

good cement jobs can be obtained. In more

extreme cases of rapidly moving salt, liners

55Spring 1996

cemented inside cemented casing increase

nonuniform load capacity by reducing casing

deformation. Collapse resistance of properly

cemented concentric strings can equal or exceed

the combined strengths of individual liners and

casings.6 Casing across salt zones is subjected to

tension, compression, burst and hydrostatic loads

combined with nonuniform forces, which must be

included in design calculations.7 Casing can be set

just below salt to save time or in deeper forma-

tions for better support, depending on the salt

interval.8 A diversion stage tool in the casing

string just below the salt may be needed to place

specialized cements across the salt, reduce hydro-

static pressure on weaker subsalt intervals or

ensure efficient slurry placement.

Effective cement fill in the annulus between the

outer casing and borehole minimizes nonuniform

load effects. Long slurry thickening times may

allow salt to encroach on casing before a complete

set occurs, and inadequate displacement across

washouts may cause unequal loading or localized

bending. Adequate fluid-loss control is needed to

prevent excessive loss of slurry mix water that can

dissolve or weaken salt, adversely affect cement

properties or cause annular bridging, loss of

hydrostatic pressure and gas migration (see “Get-

ting to the Root of Gas Migration,” page 36).

Salt-saturated cements prevent salt dissolution,

but are more difficult to mix on surface and extend

slurry set times (over-retardation). Freshwater and

low-salt concentration slurries avoid retardation

problems and are easier to handle, but long-term

exposure to salt may lead to cement failures.

Additives introduced in the late 1980s helped solve

over-retardation and strength development prob-

lems in salt-rich slurries.9 This led to development

of proprietary slurries for cementing across salt

zones like the Dowell SALTBOND cement system,

which provides controllable thickening times,

good early compressive strengths, effective place-

ment rheology, excellent fluid-loss control and

resistance to aggressive brine attack. –MET

1. Barker JW, Feland KW and Tsao YH: “Drilling Long Salt Sec-tions Along the U.S. Gulf Coast,” paper SPE 24605, pre-sented at the 67th SPE Annual Technical Conference andExhibition, Washington, DC, USA, October 4-7, 1992.

Pattillo PD and Rankin TE: “How Amoco Solved CasingDesign Problems in The Gulf of Suez,” Petroleum EngineerInternational 53, no. 11 (November 1981): 86-112.

2. Leyendecker EA and Murray SC: “Properly Prepared OilMuds Aid Massive Salt Drilling,” World Oil 180, no. 4 (April 1975): 93-95.

3. Warren TM, Sinor LA and Dykstra MW: “SimultaneousDrilling and Reaming with Fixed Blade Reamers,” paper SPE30474, presented at the 71st SPE Annual Technical Confer-ence and Exhibition, Dallas, Texas, October 22-25, 1996.

4. O’Brien J and Lerche I: “Understanding SubsaltOverpressure May Reduce Drilling Risks,” Oil & Gas Jour-nal 2, no. 4 (January 24, 1994): 28-34.

5. Cheatham JB and McEver JW: “Behavior of Casing Sub-jected to Salt Loading,” Journal of Petroleum Technology 16(September 1964): 1069-1075.

6. Burkowsky M, Ott H and Schillinger H: “Cemented Pipe-in-Pipe Casing Strings Solve Field Problems,” World Oil 193,no. 5 (October 1981): 143-147.

El-Sayed AAH and Khalaf F: “Resistance of Cemented Con-centric Casing Strings Under Nonuniform Loading,” SPEDrilling Engineering 7, no. 1 (March 1992): 59-64.

7. Recent methods use common Von Mises calculations fornormal loads along with the addition of stresses to accountfor nonuniform collapse. For more on these methods: Hack-ney RM: “A New Approach to Casing Design for Salt Forma-tions,” paper SPE/IADC 13431, presented at the 1985SPE/IADC Drilling Conference, New Orleans, Louisiana,USA, March 6-8, 1985.

8. An article in three parts: LeBlanc L: “Drilling, Completion,Workover Challenges in Subsalt Formations,” Offshore(June 1994): 21-22, 49 (part I); (July 1994): 42-44, 59 (partII); (August 1994): 38-40 (part III).

9. Yearwood J, Drecq P and Rae P: “Cementing Across Mas-sive Salt Formations,” paper 88-39-104, presented at the39th Annual Technical Meeting of the Petroleum Society ofCIM, Calgary, Alberta, Canada, June 12-16, 1988.

ImagingImaging describes the two seismic data pro-cessing steps, stacking and migration, thatbring seismic reflections into focus. Stackingattempts to increase signal-to-noise ratio bysumming records obtained from severalseismic shots reflecting at the same point(above). Energy arrives on each trace at adifferent time, depending on the source-receiver separation, or offset. For a uniform-velocity layer overlying the reflector, seismicrays are straight, and the arrival times definea hyperbola. The set of traces is called acommon midpoint (CMP) gather. Before theCMP gather can be stacked, the traces mustbe shifted to align arrivals. The offset versustime parameter that describes the shiftsdefines the stacking velocity of that layer.Shifting is performed for all reflections visi-ble in the traces. The result of stacking is asingle trace, taken to represent the signalthat would have been recorded in a normal-incidence experiment at the midpoint of thesource-receiver pairs. The basic assumptionin stacking is that velocity does not varyhorizontally over the extent of the gather.

The second component of imaging,migration, redistributes reflected seismicenergy from its recorded position to its trueposition using a velocity model (right ).There are many classes of migration, vary-ing in environment of applicability fromsimple structures and smooth velocity varia-tions to complex structures and rapidlyvarying velocities.6

The main distinctions, for the purpose ofthis article, are the imaging domain—either

time or depth—and the order of migrationin the work flow—poststack or prestack. Toprocess any one survey, combinations ofmigration techniques may be used. Thetrend today, as complex reservoirs comeunder scrutiny, is to use depth rather thantime and prestack instead of poststack.

In time migration, the velocity model,sometimes called the velocity field, may varyonly smoothly (next page, bottom). Velocityshould increase with depth, and any varia-tions in the horizontal direction should begradual. The output of the process is a seis-

mic volume with time as the vertical axis.Time migration is most successful whenvelocities are laterally invariant or smoothlyvarying. It is often applicable and, hencechosen in most parts of the world.

In depth migration, the velocity modelmay have strong velocity contrasts verticallyor horizontally. Depth migration is suited forenvironments in which velocities changeabruptly, often the case with complex struc-tures such as steep dips, faults, folds, saltintrusions and truncated layers. The outputvolume has depth as the vertical axis. Depthmigration, though often appropriate, is stillrarely done because of the difficulty in con-structing an accurate velocity model.

Poststack migration is migration appliedafter the seismic traces have been stacked.Stacking enhances the seismic signal, andalso reduces by an order of magnitude thenumber of traces that comprise the stackedseismic volume, so migration poststack isroughly 100 times faster than prestack. Forpoststack migration to be effective, theassumptions made in stacking must bevalid. The amplitude of the stacked tracemust represent that of the normal-incidencetrace and reflected arrivals must be approxi-mately hyperbolic (next page, top). Theseassumptions are valid only when the struc-ture is simple. Otherwise prestack migrationis more suitable.

Prestack migration is run before stacking,and can handle the most complex structures

56 Oilfield Review

6. For a review: Farmer P, Gray S, Hodgkiss G, PieprzakA, Ratcliff D, Whitcombe D and Whitmore D: “Struc-tural Imaging: Toward a Sharper Subsurface View,”Oilfield Review 5, no. 1 (January 1993): 28-41.

nStacking to enhance and focus seismic signals by summing traces reflected midway between several source-receiver pairs. Energyarrives on each seismic record at a different time, depending on the source-receiver separation, or offset. The arrival times define ahyperbola. Before the traces can be stacked, they must be shifted to align arrivals. The offset versus time relationship that describesthe shifts defines the stacking velocity of that layer.

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa0 offset

Offset 1

Offset 2

Offset 3

Offset 4

Common midpoint(CMP)

1 2 3 4

Two-

way

tim

e

1 2 3 4Offset

Hyperboliccurve

Stackingvelocity

+ =

Offset

Corrected CMP gather StackedCMP

+ + + =aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa a aaaaaaaaMigratedtrace

Midpointtrace Receiver

MIG

Source

Salt

Originaldata

nRedistribution of reflected seismic energyby migration. In this simple 2D rendition,migration (MIG) repositions a reflectedtrace from its recorded position to its trueposition using a velocity model. In morecomplex and 3D cases, reflections may beredistributed to positions outside the planecontaining the source and receivers.Energy may also be distributed amongmultiple locations.

57

nThe effects of velocity variations on raytracing and common midpoint (CMP) assumptions. In a flat model with simplestructures and velocities (top left), raypaths are straight and wavefronts are spherical. Arrival times on seismic records canbe fit with a hyperbola (bottom left). In such a case, the CMP and reflection point would be coincident. Inserting a salt wedgeover the flat reflector (top right) gives rise to bent raypaths. The arrivals do not have a hyperbolic shape on seismic records(bottom right). In this case, the CMP would not be coincident with the reflection point. Also visible in the salt case are multi-ples—arrivals from multiply reflected waves—that present additional processing problems. These waveforms and traceswere created with 2D acoustic finite-difference modeling.

Simple velocities + simplestructure = poststack time migration

Complex velocities + simplestructure = poststack depth migration

Simple velocities + complexstructure = prestack time migration

Complex velocities + complexstructure = prestack depth migration

Incr

easi

ng v

eloc

ity

nVelocity models for four migrationclasses: time, depth, poststack andprestack. Poststack models are on the left,prestack on the right. Time-based modelsare on the top, depth-based on the bot-tom. In time migration, the velocitymodel may vary only smoothly or mono-tonically—always increasing with depth.Depth migration is required for more com-plex velocity models. Poststack migrationworks with models of low complexity,while prestack migration can handle themost complex models.

Dep

th, f

tDistance, ft

1000 1000 2000 30000Distance, ft

1000 1000 2000 30000

Distance, ft1000 1000 2000 30000

Distance, ft1000 1000 2000 30000

500

1000

1500

2000

Dep

th, f

t

0

500

1000

1500

2000

0

0

100

200

300

400

500

600

700

Two-

way

tim

e, m

sec

0

100

200

300

400

500

600

700

Two-

way

tim

e, m

sec

Bottom salt

Reservoir topReservoir top

Shot Gather Shot Gather

SnapshotSnapshot

Reservoir

Snapshottime

Snapshottime

Reservoir

Top salt

Saltwedge

Spring 1996

and velocity fields. With the amount of datain modern 3D surveys, the main constraintson this method are the time and skillneeded to construct velocity models and thecomputing power required for reasonableprocessing turnaround time.

Imaging a seismic volume containing asalt body is unlike traditional processing, inwhich thousands of tapes are sent off to aprocessing group that sends back a finishedproduct, ready for interpretation. Subsaltimaging requires several iterations of migra-tion and interpretation. The process is acomplex interplay of many steps (left ).7

Some of the steps, such as the migrations,are run as batch input to mainframe or mas-sively parallel processor (MPP) computers.Others, such as velocity modeling and layerboundary interpretation, require interactiveworkstations.

Different operators and service companiesmay have variants of these methods, but thegeneral processing flow is the same. Thefirst step is to build an initial model of thevelocity in the overburden—the velocities oflayers overlying the salt. In the North Sea,several major velocity contrasts may overliethe salt. Velocity estimates can come fromray-tracing-based velocity analysis on CMPgathers. If the common midpoint geometryis not suitable, such as when velocities varyhorizontally, a CMP gather cannot be used.Instead, a common image point (CIP) gatheris created using a prestack migration tech-nique to assemble all the traces that imagethe depths below a given surface location.8

In the Gulf of Mexico, sediments are typi-cally sand-shale sequences with smallvelocity contrasts between layers. Withoutstrong velocity contrast, CMP-based velocityanalysis is not necessary, so initial velocitiesare taken from stacking velocities. In bothcases, velocities are checked for trends withwell data such as sonic logs or boreholeseismic data.

The second step uses this early velocitymodel to predict reflection arrival times onCMP or CIP gathers at control points. Theshape of the arrival times of the shallowestmajor reflector is analyzed for the velocitythat best flattens the times, and the velocitymodel is updated. This is the most time-intensive step, and requires the interventionof an expert and the versatility of an interac-tive velocity modeling workstation. (For atour of the Geco-Prakla KUDOS 3D veloc-ity model building workstation, see “Foun-dations in Velocity,” page 60.)

58 Oilfield Review

Model validation

Current layer = maximum

Current layer < maximum

3D poststackdepth migration

volume

Prestack depthmigration of

selected offsets

AnalyzeCIP gathers

3D CMP gathers

3D prestack travel-timeinversion from ray-based

velocity analysis

Compute and outputvelocity nodes

Update velocitycomponent of model

3D poststackdepth migration

Delineate geometryof base of layerin depth domain

Update depthcomponent of model

3D prestackdepth migration

Interpretation ofcurrent layer

in time domain

Velocity modelin depth

3D prestackdepth migration

volume

Inputs

3D stackvolume

nSubsalt prestack depth migration flow chart.Some of the steps, such as velocity modeling andlayer boundary interpretation, require interactiveworkstations. The migrations are run on powerfulmainframe or MPP computers.

With the updated velocity model, post-stack or prestack depth migration is applied,and the gathers are recomputed andchecked for arrival flatness. If necessary,these few steps are iterated to obtain anaccurate velocity of the topmost layer. Thenthe process is repeated for as many layers asare identified above the salt.

If the top of salt appears to be structurallysimple based on preliminary time migration,the velocities of the overburden can be usedin a poststack depth migration to image thetop of salt with good precision. An exampleof this is the imaging of the Cavendish 3Dsurvey in the North Sea. The velocity modelindicates a smooth top of the Zechstein salt(bottom left). Encased within the Zechsteinis a thin, complexly folded dolomite, calledPlattendolomit, that causes strong distortionof seismic ray paths before they reach theSilverpit target. An important step in theconstruction of an accurate depth-velocitymodel was characterizing the shape of thePlattendolomit (below right). The complex-ity of the velocity model—high-velocity saltoverlying lower-velocity sediments—sug-gests that depth migration is better suited forimaging than is time migration. Applyingdepth migration makes a dramatic differ-ence in subsalt structure: the dip of subsaltlayers, and so the locations of potentialtraps, changes significantly compared to thetime migration results (left).

59Spring 1996

7. Modified from: Godfrey B, Pieprzak A, Berg K and Yilmaz Ö: “3-D Salt and Sub-Salt Imaging Strategy: A Case History from the Gulf of Mexico,” TechnicalProgram and Abstracts SEG Summer Research Work-shop on 3-D Seismology: Integrated Comprehensionof Large Data Volumes, Rancho Mirage, California,USA, (August 1-6, 1993): 128-134.

8. Common image point gathers are assembled by amethod that has been likened to looking for a needlein a haystack. Every possible source-receiver pair inthe 3D volume of interest is checked to see whether itcontributes to the signal generated by the reflection ata test point in the volume.

nComparing poststack time (top) to poststack depth migration (bottom) on theCavendish survey. The complex velocity model requires depth migration toaccurately image subsalt structures. Without depth migration, the dips on sub-salt layers may be incorrectly imaged.

nSurface of the complexly folded Plattendolomit.

nVelocity model for the North SeaCavendish survey. The Zechstein salttop is relatively smooth, allowing post-stack migration. Within the salt layerthe lower-velocity Plattendolomit canbe seen. The target layer is the Silver-pit formation.

Time Migration

Depth Migration

Distance, mTw

o-w

ay ti

me,

mse

cD

epth

, m

Distance, m

011,250

1700

1800

1900

2000

2100

2200

2300

2400

2500

2600

2700

2800

2500

2750

3000

3250

3500

3750

4000

4250

4500

011,250

Amoco Survey Velocity Section

Base Zechstein

Top Zechstein

Silverpit formation

Plattendolomit

50001500 3250Velocity, m/sec

Plattendolomit Surface

(continued on page 63)

Before the arrival of massively parallel processor

computers, migration was the stumbling block in

prestack depth imaging. Now that MPPs can han-

dle migration in reasonably short order, the con-

struction of an accurate 3D velocity model is the

most time-consuming task. The Geco-Prakla

KUDOS 3D velocity model building system allows

specialists in interpretive processing to construct

and visualize velocity models interactively.

Velocity modeling systems developed by other

service companies, such as InDepth by Western

Geophysical and GeoDepth by Paradigm Geo-

physical, contain similar features.

A velocity model is defined by two sets of

parameters—layer velocities and reflector

geometries. Such models can have either time or

depth as their vertical axis. Models with time as

the vertical axis are relatively easy to derive from

conventional time-domain processing, and are

generally smooth: rays can be traced through the

models with moderate bending at interfaces, so

processing steps such as computing travel times

through the model can be executed rapidly and

nearly automatically.

In contrast, earth models in depth usually have

strong horizontal and vertical velocity variations.

Rays can bend sharply at interfaces and so the

reflector geometry must be known very accu-

rately. Processing must take an interpretive pause

after each layer is built, precluding automation.

Efficient construction of depth-based models is

the aim of the KUDOS velocity modeler.

Traditional velocity modeling programs con-

strain models to be simple—unlike the real earth

—with no abrupt terminations, pinchouts or multi-

ple vertical values. Layers must be continuous

and extend across the entire survey. The KUDOS

system, by contrast, allows models to be built

with any structural complexity. Graphic elements

are rendered on a high-performance workstation,

allowing immediate visualization—a key ability in

velocity model construction and validation.

In the KUDOS system, a modeling volume is

defined that has its vertical dimension in depth.

Surfaces corresponding to the main geological

horizons are inserted into this volume, subdivid-

ing it. Interval velocity fields are derived and

assigned to each subvolume, forming a spatially

variant velocity-depth model.

Layers are added to the model in an iterative

sequence. At each stage the model consists of a

series of layers, each with its own velocity field,

and a halfspace of unknown velocity below the

bottom layer. This halfspace contains the next

horizon to be imaged. The velocity that will cor-

rectly image the next horizon is derived through

ray-based velocity analysis (below). The velocity

of the layer is mapped by interpolating velocities

determined at control points (next page, bottom).

The halfspace is then “flooded” with the velocity

field derived for that next horizon.

The subvolume model is then exported from

the KUDOS workstation as either a tessellation or

60 Oilfield Review

nInteractive ray-based velocity analysis. For a chosen gather (lower left panel) traces can be shifted interactively to test different interval veloci-ties. A plot of semblance—the coherence achieved between traces shifted with a given velocity—shows the best choices for velocities (upperleft). The higher the semblance, the better that velocity flattens the traces. Velocities that are too high leave arrival times drooping at long offsets(upper right). Velocities that are too low produce corrected gathers that swing up at long offsets (middle right). The correct velocities flattenarrival times across the gather (lower right).

Foundations in Velocity

a 3D grid, and sent with the seismic data to the

computer for post- or prestack depth migration.

Tessellation involves dividing the layered

velocity model into tetrahedra (above, left and

right). Interval velocities are stored at each cor-

ner of every tetrahedron, and the topographies of

the depth surfaces are represented by tetrahedral

facets. Tessellated volumes have special proper-

ties; they are especially efficient for modeling

arrival times by raytracing—for generating travel

times for prestack depth migration—and they can

represent realistic geologic models with struc-

tural complexity at all scales (left). The KUDOS

61Spring 1996

nLayered model before (left) and after (right) tessellation. Tessellation divides layer volumes into tetrahedra and assigns a velocity to each corner of every tetrahedron.

nVelocity control pointsfor a chalk reflectorabove the salt. Velocitiesfor the layer immediatelyabove the reflector areinterpolated betweencontrol points (smallcubes) which are colorcoded by interval veloc-ity—blue is faster thangreen. The spatial posi-tion of each control pointis dictated by rays tracedthrough the velocityfields of the overlyinglayers. A 2D slicethrough the seismic vol-ume is displayed withrays contributing toselected control points.

nTessellation of salt volume with structural complexity at many scales.

Velocity Control Points

Layered Model Before Tessellation Layered Model After Tessellation

Tessellated Salt Body

system can also express the velocity model as an

array of evenly spaced 3D grid points. This creates

a volume that may not look as complex as the tes-

sellated volume, but has a velocity representation

more suited for some migration algorithms.

Following migration, the seismic data are

loaded to the interpretation workstation, where

the newly imaged horizon is delineated in depth.

This surface is then incorporated into the KUDOS

model, forming a new base layer. The velocity

field below this layer now needs to be determined,

so the next iteration of velocity analysis begins.

In some areas, such as the Gulf of Mexico, the

background velocity is slowly varying and layer

boundaries are difficult to identify (next page, top).

Instead of proceeding in steps, layer by layer, the

background velocity model is built in just a few

steps, each handling several layers. At selected

locations, CIP gathers are analyzed for the overall

velocity function that best flattens all the arrivals

simultaneously. In the KUDOS system, this

method is called image-based velocity analysis.

The velocity function can be modified interactively

and a corrected gather can be viewed (right).

62 Oilfield Review

nFinding the velocity function that flattens all arrivals simultaneously. Common image point (CIP) gathers(top) obtained from prestack depth migration are converted from depth to time using the current velocitymodel and displayed twice (left and center). The interval (green) and root mean square (RMS) velocity func-tions (red) for this model are shown as a pair of curves on a semblance display (right). Interval velocities canbe modified interactively, automatically adjusting the corresponding RMS velocity function. A new gather isthen computed, and the arrival curvature can be compared to that on the reference gather (left) which remainsunchanged. Other velocities can be tested (bottom). In this example, velocities higher than the referencemodel have been picked (green dots) and applied to the gather (center panel). The new velocities are too high,causing downward curvature to the arrivals. The original velocities remain as black dots on the screen.

Correct Velocities

High Velocities

If the top of salt is rough, prestack depthmigration must be applied (right). Geolo-gists surmise that such complex topogra-phies indicate instabilities where theupward movement of the salt, once halted,has been reactivated.

Once the top of salt has been imaged, aninterpreter must delineate the top of salt onan interactive seismic interpretation worksta-tion. Then the velocity model is updated byfilling the volume below the top of salt withsalt velocity, assumed to be uniform. Withthis new model, another prestack—or post-stack if overburden velocities are smoothenough—depth migration is performed, andthe bottom of salt comes into focus.

An interpreter then maps the bottom ofsalt. Next, and similar to the first step, veloc-ities of the sedimentary layers below the saltare estimated. These are first approximatedby the velocities of layers at the same depthbut outside the canopy of salt. Then aprestack depth migration is run and sets ofgathers are checked for flat arrivals. Thevelocity model is updated at these controlpoints until all control points show flat

arrivals on CIP gathers. Then the velocitiesare interpolated between control points andthe full-volume velocity model is complete.

The final step is to run a prestack depthmigration using the full-volume velocitymodel. Then individual cuts through themigrated data volume can be displayed forfurther interpretation. With the vertical axisin depth, locations of interpreted featurescan be communicated directly to engineersto guide drilling and well location decisions.

This set of techniques was used to imagethe salt and subsalt layers in a survey forAmoco in the southern North Sea gas basin.Layers were interpreted on the Charismaseismic interpretation system, and theirvelocities were modeled on the KUDOSworkstation. The target layers were theRotleigendes and Westphalian sands below

63Spring 1996

nGulf of Mexico salt sheet with associated velocity functions at controlpoints. The background velocity increases gradually with depth (yellow togreen) making layer boundaries above the salt difficult to pick. The strongcontrast and constant velocity of the salt are depicted by the dark blueband in the velocity functions.

nA structurallycomplex salt fea-ture requiringprestack depthmigration to imageits top and bottom.The top surface iswhite, the bottomis gold. Imagingoverlapping saltbodies, such asthose shown in thisfigure, requiresadditional itera-tions in processing.

50001500 3250Velocity, m/sec

Overlapping Salt Bodies

the Zechstein salt. Comparison of poststackand prestack depth migration shows thegreater clarity of the prestack method infocusing the top and bottom salt reflections(above ). The prestack depth migrationshows a more sharply focused reflection offthe base of salt and more coherently imagedsubsalt strata than does the poststack migra-tion, paving the way for more confidentinterpretation of subsalt layers.

Algorithms for carrying out these classesof migration have been known for sometime.9 But only in the past few years hascomputer power grown sufficiently to allowcommercially acceptable turnaround forprestack depth migration. Massively parallelprocessors have brought the elapsed timerequired to process a “typical” prestackdepth migration down to one month—a ten-fold improvement.10 In this case, typicalmeans an output volume of two to three off-shore US blocks at 9 sq mile [23 km2] each.Specialists estimate that creating an accu-

rate velocity model takes about a week foreach layer in the model. Velocities must beaccurate to within a couple percent to beuseful for guiding subsalt drilling.

Much work remains if subsalt reservoirsare to be understood as fully as other, moreaccessible fields. In general, even the mostcarefully migrated subsalt images fail toexhibit the same signal quality as sectionsimaged in the absence of salt. Up to now,nearly all subsalt features drilled andlabeled commercial successes have beenidentified by structure rather than by ampli-tude or other waveform attributes routinelytracked by interpreters exploring above salt.

Another seismic technique, the boreholeseismic survey, offers subsalt informationunobtainable by other means.11 These sur-veys, with receivers in the borehole, canmeasure subsalt layer velocities with highaccuracy, map reflector locations and mea-sure reflection amplitudes at the subsaltreflectors. Some operators are using bore-hole seismic survey results to update veloc-ity models for reprocessing prestack depthmigrations.

An advance anticipated in the future is themeasurement of sonic velocities whiledrilling, which can be related to seismiclayer velocities. Operators may be willing toupdate seismic velocity models and repro-cess 3D surveys to get a clearer imagebefore drilling deeper.

The future of subsalt exploration anddevelopment promises as many technicalchallenges as in the past. And beyond salt,the same techniques hold the power toimage other complex features such as over-thrust faults, reefs, recumbent folds and sed-iments below high-velocity carbonates.

—LS

64 Oilfield Review

nComparison of poststack (left) and prestack (right) depth migration of the Amoco survey in the North Sea. Poststack migrationproduces a broken image of the top and bottom of the Zechstein salt. Prestack migration better images reflections from thesalt boundaries, and brings subsalt layering into focus.

9. Western PG and Ball GJ, reference 2.10. The Geco-Prakla processing megacenter in Houston,

Texas, USA relies on a Connection Machine CM-5with a 400-Gbyte disk and 512 processing nodes,providing 64 Gigaflops of peak processing power.

11. For a review of borehole seismic applications:Christie P, Dodds K, Ireson D, Johnston L, RutherfordJ, Schaffner J and Smith N: “Borehole Seismic DataSharpen the Reservoir Image,” Oilfield Review 7,no. 4 (Winter 1995): 18-31.

Dep

th, m

Distance, m

1200

00 2250 4500 6750

Distance, m0 2250 4500 6750

2400

3600

Poststack Depth Migration Prestack Depth Migration