Seismic migration:

basic concepts and popular methods

Downloaded 09/11/14 to 201.102.102.23. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Part 1

By IRSHAD R. MUFTI
Mobil Researchand DevelopmentCorporation
Dallas, Texas

Shortly after the introduction of the seismic reflection The propagation of seismic energy into the subsurface
method, it was realized that the various events identi- is a very complicated phenomenon which depends on a
fied in a seismic section do not, in general, represent the large number of factors. Any attempt to migrate the
true locations of the subsurface structures. The reflected seismic data observed at the surface that takes into ac-
energy reaching the surface propagates in the form of count all these factors becomes hopelessly involved. In
ever-expanding wavefronts. In a nonuniform medium, order to make things tractable, migration is accomplished
the variations in velocity tend to influence and modify by introducing the following simplifying assumptions:
the direction of propagation and are accompanied by
mutual interference of wavefronts. In the zones of sudden l The earth behaves like a fluid medium in which
changesalong the geologicinterfaces such as faults, a por- variations in density can be ignored. In that case, we can
tion of this energy undergoes diffraction. Consequently, use the simplest form of the wave equation for investi-
the record of events constituting a seismic section repre- gating the migration problem.
sents a distorted image of the subsurface reflectors, an
image which has undergone a complicated process of 9 The real earth can be treated as a two-dimensional
focusing, defocusing, interference and diffraction. A medium. It is admittedly true that in some situations, in-
numerical procedure aimed at correcting for these prop- cluding migration, this assumption can lead to serious
agation effects is known as migration. errors. On the other hand, the acquisition of field data
The first comprehensive and practically meaningful is usually done along linear profiles. Moreover, the
algorithm for doing migration was put forward by J. G. assumption of a two-dimensional earth is almost invari-
Hagedoorn in 1954. He developed a method of reposi- ably inherent at all stages of seismic data processing.
tioning the various events identified in a time section
l Every subsurface reflector is made up of a set of
such that the transformed section would correctly repre-
elementary diffractors. This means that the superposition
sent the geometry of the subsurface reflectors. Hage-
of the wave fields generated by the individual diffractors
doorn’s method utilizes the concepts of ray theory and
is approximately equivalent to the energy reflected by
yields reasonably good results in the absence of strongly
the corresponding subsurface horizon.
dipping events or lateral changes in the subsurface velo-
cities. Moreover, it is quite efficient and can be easily ex- l A common depth point (CDP) seismic section is the
tended to three dimensions. produc! of upward propagating waves only. In the pro-
In 1971, Jon Claerbout introduced a new method of cessof stacking the observed seismic data, the multiples
migration which makes use of an approximate version tend to cancel. A key assumption is that a seismic sec-
of the wave equation. Such an approach is directly based tion is completely free from multiples. Under these con-
on the distribution of energy of the seismic record, and ditions, the seismic sourceswhich are actually located at
it opens up the possibility of taking into account all or near the surface of the ground can be replaced by the
types of phenomena associated with the propagation of subsurface reflectors which explode simultaneously at
waves such as diffraction and interference. The impor- an initial time t = 0. The energy emanating from the
tance of Claerbout’s work was readily recognized; it reflectors travels only in the upward direction without
followed a period of intense effort in research and de- generating any multiples on its way to the surface. This
velopment in the area of wave equation migration. simplification reduces the traveltime by one-half. On a
It is not the intention of this summary to review and seismic section, it amounts to reducing the value of the
evaluate every article written on this subject. Rather, it time sampling interval by one-half.
shall be restricted to the most common migration algo-
rithms in current use. Some basic ideas and assumptions The concept of exploding reflectors is a rather crude
underlying seismic migration are a logical place to way of explaining things and it seemsto have no counter-
begin. part in other disciplines; however, it works reasonably

24 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION AUGUST 198.5

 well for migrating the seismic sections.
Before going into the details of migration, it would be
useful to examine how some simple structures appear on
a seismic section.

The simplest example of a seismic image is due to a

point diffractor. But since a point diffractor can be used
Downloaded 09/11/14 to 201.102.102.23. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

as a building block to construct a reflector of any arbi-

trary geometry, this example is also the most important.
Figure 1 shows a point diffractor P and a set of geo-
phones G,, . . . . G5 located along the surface. The energy
emanating from P propagates upwards in the form of
an expanding wavefront. Three different positions of
the wavefront corresponding to times t,, t,, and t2 are
indicated as partially drawn concentric circles. In accor-
dance with the well known Fermat’s principle, the trans-
fer of energy from P to the various geophones will take Figure 1. Various stagesof evolution of an expanding wave-
place along a set of linear paths or rays as indicated in front originating at the point P.
Figure 1. Let the location of the middle geophone G,
correspond to the origin of coordinates (X = 0, z = 0)
SEISMIC TRACES
with positive t-axis pointed downward; z denotes the
3
depth of the point diffractor and x represents the hori- 00 7

zontal position of any of the geophones from the origin. t

The time (usually referred to as arrival time) needed for
the energy to travel from the point P to the various geo-
phones is given by
I = distance from the point diffractor to the geophone
velocity of the medium

where the numerator of the fraction is equal to (9 + z’)‘:.

If, using this relationship, t is plotted as a function of
x for different geophones, the resulting curve will have
the form of a hyperbola. Let us use this relation to com-
pute the arrival times of the wavefront to the various 201 1
geophones by setting z = 2,000 m and velocity = 4,000 3.000 - 2.000 -1000 0 1.
HORIZONTAL DISTANCE (M)
m/set. The results are shown in Figure 2; it represents
the seismic image of the point diffractor. Figure 2. Seismicimage of a point diffractor located at a depth
It is obvious that as we increase the depth of the point of 2,000 m in a medium of velocity 4,000 m/set.
diffractor, it will take the wavefront longer to reach the
surface, losing more and more of its curvature during its
upward path. A flatter front implies that the arrival PROFILE AXIS G, G2 G, G, GS
0
times to the various geophones will differ more slowly -..
_”
_ - -
and the corresponding seismic event will appear as a
broader hyperbola., By the same argument, the shal-
lower the depth of the diffractor, the narrower the event
becomes.

F. lgure 3 shows a horizontal reflector approximated by

a set of equidistant point diffractors PI, . . . , P,. The cor-
responding seismic section consists of five identical but
laterally displaced hyperbolas. It is obvious that a proper
seismic representation 6f the reflector requires a much
larger number of point diffractors. In that case, the
various hyperbolas lose their identity due to destructive
interference of the corresponding wavefronts and the re-
sulting seismic section is reduced to a continuous line or
the envelope running tangentially along the apexesof the
hyperbolas. But note that as we approach either end of 2.0 I
0 1 2 ; 4 5 6 7 8
the reflector, the effect of interference gradually re- HORIZONTAL DISTANCE (KM)
duces. Consequently, the seismic image of the reflector
is identical to the shape of the geologic model except for Figure 3. Approximate representationof horizontal linear re-
a gradual tapering of energy along the hyperbolic arcs. flector and the correspondingseismicimage.

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION AUGUST 1985 25

 Figure 4 shows a reflector PIP2 dipping at an angle the seismic section belongs to the hyperbola associated
a = 30 degrees. We shall employ the above mentioned with the point P, of the geologic model.
procedure to construct its seismic section. Since it is a l For any given combination of velocity and depth,
linear reflector, it would suffice to compute only two the corresponding hyperbola representsa curve of max-
hyperbolas associated with the end points P, and P,. imum convexity. This implies that the two hyperbolas
Note that the hyperbola associated with the point P, is shown in Figure 4 cannot be interchanged. Either of
broader than that for P1 due to differences in depth. them is as convex as it could be for the respective depth.
The line P,‘P,’ which meets the two hyperbolas tangen- In the case of real data obtained in an area where
Downloaded 09/11/14 to 201.102.102.23. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

tially at P,’ and P2I is the required envelope and repre- reliable velocity control exists, one can identify the dif-
sents the image of the reflector. fraction patterns by measuring the curve of the hyper-
Most of the energy leaving the reflector P,P, will propa- bolic type features. Conversely, one can look for and
gate to the surface along the direction normal to the line use such features for estimating subsurface velocities.
P,P,. Thus, the energy arriving at the geophone G,‘will l Formation of the seismic section involves a rolling
be mostly due to the point source P, and will reach the down of the geologic depth points along their hyper-
surface at the time which corresponds to the two-way bolas. Thus in Figure 4, the point PI rolls down to the
travel along the raypath between PI and G ,‘. Since on a point P,‘. The distance along which this rolling down
time section, the traces are plotted at right angles to the takes place increases with depth.
direction of the profile, the event associated with P, will l The seismic section can be regarded as a more
appear on the time trace of the geophone G,‘such that or less distorted and displaced image of the geologic
PIG,’ = P,‘G,‘. A simple way to determine P,‘is to draw a model. and usually appears larger than the true geo-
circular arc through P, with G,‘as its center; P,‘cor- logic model. This apparent increase in size amounts to a

PROFILE DIRECTION
G, GI C$ Gz I 14 15 16 17
00 F f t
-0 00'

P
05: b

G IO’

Y
6 15

I
20s -> ,‘_ 14
0 1 2 3 4 5 6 3 8 201 f

HORIZONTAL DISTANCE (KM)

25i 1 I I
Figure 4. A dipping linear reflector P, P, and its seismicimage
P: P: obtained by ray-theory considerations. Figure 5. A simple example of a synthetic seismicsection.

responds to the intersection of this arc and the vertical loss of lateral resolution. Moreover, its displacement is
line passing through G,‘. By a similar procedure, we can approximately along the downdip direction and takes
determine the image Pi for the point P,. place in such a manner that it tends to overlie the object.
Consequently, the seismic section of a salt dome is
S lnce every other structure can be visualized as made broader than the dome and fully enclosesit from above.
Similarly, by visualizing a synclinal structure as made
up of linear segments dipping at different angles, the
importance of the case under consideration cannot be up of linear reflectors dipping at different angles, it can
overemphasized. be easily explained why its seismic section sometimes
gives rise to the familiar “bow-tie” pattern.
Conclusions which can be drawn from this examina-
tion of fundamental migration assumptions are: Some of the basic features of the seismic section and
l Every point belonging to a reflector in the geologic how they differ from the corresponding features in the
geologic model have now been reviewed. The various
depth model acts like a diffractor and gives rise to a
methods of transforming the seismic sections into the
hyperbola on the corresponding seismic section. Conse-
true geometry of the corresponding geologic structures
quently, as long as we ignore multiples, a one-to-one - the procedure of migration - will now be discussed.
correspondence exists between a given set of diffractors
and the corresponding set of hyperbolas. Except in
zones of sudden changes in reflectivity such as faults L et us use our newly acquired skill to compute a syn-
and termination of reflectors, the individual hyperbolas thetic seismic section showing a dipping reflector under-
lose their identity due to destructive interference. lying a diffractor - for convenience we shall use the
l Every diffractor belonging to the geologic model reflector in Figure 4. The results are shown in Figure 5.
represents the apex of its hyperbola in the seismic sec- The seismic responseof the planar reflector appears as a
tion. The seismic image of such a diffractor obtained by chain of events along the dashed line connecting the
simple considerations involving rays is one of the points points P,’ and Pi. The seismic response of the diffrac-
belonging to this hyperbola. Thus in Figure 4, the P,‘of tor can be identified as a hyperbola with its apex P, on

26 GEOPHYSICS: THE LEADING EDGE OF exploration AUGUST 1985

 SEZSMKMIGRATZONI1111)
Downloaded 09/11/14 to 201.102.102.23. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

“There3gottabe an easierdownward
continuation
of sourceandreceiver.
*
trace 5. For subsequent discussion,a numberof hyper- with the real size seismicsections,sucha procedureis
bolaspassingthroughthe variouspointshavebeenadded extremely laboriousand is not worth the effort - even
as well as somerandom noiseon trace 12 at time 1.95 at today’s prices of oil and gas. Subsequentworkers
set and on trace 14 at time 1.75 sec. were able to automateHagedoorn’s procedureby mak-
The dashedhyperboliccurvesshownin Figure5 were ing the assumptionthat every point on the seismicsec-
obtainedby usinga velocity 4,000 m/set. Let us start tion can be treated as the apex of a hyperbola. In the
with the hyperbolapassingthrough the point P,. The absenceof a reflector, the sum signalresultsin zero
event which falls on this curve constitutesthe seismic energy except for any residualenergyassociatedwith
responseof the diffractor whose true location corre- noise.
spondsto the point P,. Consequently,the processof The summationprocessusuallyyieldsgood results.
summingthe energyassociatedwith this event and as- However,sinceit is basedon the ray theory approxima-
signingthe sumvalueto the point whichcorresponds to tion, it fails to take into accountsmallchangesin phase
the apexof the hyperbolawill undothe effect of diffrac- and cangive riseto a varietyof computationalnoise.As
tion and the seismicresponsewill collapseto the dif- an example,the sumsignalassigned to the point P, will
fractor. But thisis exactlywhat migrationis expectedto also includethe noiseat the lower portion of trace 12,
accomplish.This procedurecanalsobe usedfor migrat- whereaspoint P, may indicatea weakdiffractingregion
ing the planar reflector.We observethat the curvepass- due to the noisepresenton trace 14. Similarly, point Pa
ing through PI touchesthe seismicevent of the planar will deriveits energyfrom the planarreflectorbelowit.
reflector tangentiallyand passesthrough the response Theseare somesimpleexamplesof the so-calledmigra-
on traces2, 3 and 4. Thus if we sumup the energyalong tion noise.
this curve and assignit to the point PI, we are able to
migrate the left end of the reflector from its apparent The ideaof summingup the diffractionenergyrecorded
position P,‘to the correct position PI. A similar pro- at the surfaceand assigning
it to thecorresponding source
cedureenablesus to migrate P,‘to P,. In this way, we of diffraction is equivalentto derivingthe strengthof
are able to migrate the planar reflector from position the wave field at the location of the sourcefrom the
P,‘P2Ito the correctpositionPIP,. valueof the fieldalonga surfaceinfluencedby the source.
The ingeniousmethoddescribedabovewasintroduced This invokesthe possibilityof doingmigrationby using
by J. G. Hagedoorn.In the absenceof digitalcomputers, someof the well-knownresultsfrom the theoryof optics.
he useda large numberof precomputedcurvesfor dif- Considera wave field definedas a function of space
ferent velocitiesand depths.When it comesto dealing and time over a closedsurface.Then the value of the

GEOPHYSICS: THE LEADING EDGE OF EXPLORATION AUGUST 1985 27

 curvedsurfacewould be too far away to be influenced
by any subsurfacereflectors. Sincethe field measure-
mentsare confinedwithin a finite area, we mustintro-
ducean additionalassumption- that the value of the
field outsidethe surveyedareabecomesnegligiblysmall.
It is not a good assumption,but we cannotavoid it. In
any case,the effective portion of the closedsurfaceis
now confinedto a finite area wherethe field measure-
Downloaded 09/11/14 to 201.102.102.23. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

mentsare available.Keepingin view the conceptof the

exploding reflector, if a subsurfacepoint with coor-
dinates(x, y, z) belongsto a reflector,the value of the
wave field at sucha point at time t = 0, will be anoma-
louslystrongascomparedto the pointswhereno reflec-
torsare present.This is an elegantapproachto migration
and in principle, it should lead to superiorresultsas
comparedto the migration methodsbasedon straight
i
summation. In practice, however, the quality of
Figure 6. A hemispherical surface over which the Kirchhoff migrateddata obtainedby this methodis often inferior
summation is carried out. to that obtainableby other schemesin current use.At
leastsomeof the problemsassociated with thismethod
field at any point insidethe surfaceand at time I = 0 may be the directconsequence of too muchcompromise
can be computedby meansof an expressionassociated betweenthe theory and the practicallimitations. The
with the name of Kirchhoff (seee.g., Baker and Cop- theory impliesthat data to be migratedare basedon a
son, 1950).I shallnot go into further detailswhichcan 3-D seismicsurvey.In actualpractice,the field measure-
be obtainedelsewhere.It may be mentioned,however, mentsare usuallyobtainedalong a profile rather than
that the resultsobtained in this manner remain unaf-
over an area. This necessitatesfurther compromisewith
fectedby the shapeof the closedsurfacewhichmay be the theory. Sometimein the future, when 3-D acquisi-
chosento suit the problemunderinvestigation.For our tion of field data has becomea commonpractice,this
purposes,a goodchoicewould be as shownin Figure6; approximationcouldbe eliminated.Lli
it consistsof a hemisphericalbowl whoseflat surface
represents the zone of field measurements. Let usmake (The concluding part of this article, discussing modern migra-
the radiusof the bowl infinitely large. In that case,the tion techniques, will appear next month.)

28 GEOPHYSICS: THE LEADING EDGE OF EXPLORATION AUGUST 1985