You are on page 1of 182

5 Dip-Moveout Correction and

Prestack Migration
• Introduction • Salt-Flank Reflections • Fault-Plane Reflections • DMO and Stacking Velocities • Turning-Wave
Reflections • Principles of Dip-Moveout Correction • Prestack Partial Migration • Frequency-Wavenumber
DMO Correction • Log-Stretch DMO Correction • Integral DMO Correction • Velocity Errors • Variable Velocity
• Turning-Wave Migration • Dip-Moveout Correction in Practice • Salt Flanks • Fault Planes • DMO and
Multiples • DMO and Coherent Linear Noise • Other Considerations • Aspects of DMO Correction — A Summary
• Prestack Time Migration • DMO Correction and Common-Offset Migration • Salt Flanks • Fault Planes •
Common-Reflection-Point versus Common-Reflection-Surface Stacking • Migration Velocity Analysis • Prestack
Stolt Migration • Common-Offset Migration of DMO-Corrected Data • Prestack Kirchhoff Migration • Velocity
Analysis Using Common-Reflection-Point Gathers • Focusing Analysis • Fowler’s Velocity-Independent Prestack
Migration • Exercises • Appendix E: Topics in Dip-Moveout Correction and Prestack Time Migration
• Reflection Point Dispersal • Equations for DMO Correction • Log-Stretch DMO Correction • The DMO Ellipse •
Nonzero-Offset Traveltime Equation • Prestack Frequency-Wavenumber Migration • Velocity Analysis by Wavefield
Extrapolation • References

5.0 INTRODUCTION

Dip-moveout (DMO) correction is applied to normal- event is intersected by a dipping event, we can only
moveout corrected prestack data to preserve conflicting choose a stacking velocity in favor of one of these events,
dips with different stacking velocities during stacking. not both. Therefore, conventional CMP stacking does
As a result, DMO correction yields an improved stacked not equally preserve events with conflicting dips with
section that is a closer representation of a zero-offset different stacking velocities. This is not the case for
section compared to conventional CMP stack based on a zero-offset section, for it contains all events, regard-
normal-moveout correction, only. This then enables us less of dip. Thus, in the presence of conflicting dips, a
to apply the theory of zero-offset migration discussed in stacked section is not identical to a zero-offset section.
Chapter 4 to stacked data with greater confidence. Since a CMP-stacked section strictly is not equiv-
We remind ourselves from Chapter 3 that stacking alent to a zero-offset section, we expect that migration
velocities are dip-dependent (equation 3-8). When a flat after stack would not produce a crisp image in the pres-
656 Seismic Data Analysis

ence of conflicting dips with different stacking velocities. and is accurate as long as the vertical velocity gradient
To circumvent the problem of conflicting dips, it can be is moderate. Jacubowicz (1990) developed a conceptu-
suggested that migration should be done before rather ally appealing technique that involves first dip decom-
than after stack. position of the input data and application of a DMO
A practical alternative to migration before stack operator to each component, individually. Hale (1983),
is to correct for the dip effect on moveout veloci- Hale and Artley (1992), and Artley and Hale (1994) ex-
ties implied by Levin’s equation (3-8) prior to stack- tended the DMO theory to accommodate vertical veloc-
ing. It can be suggested that prestack data first can ity variations. French et al. (1984) developed a partial
be moveout-corrected using flat-event velocities. This migration technique that tries to account for variations
normal-moveout correction (NMO) then is followed by a in source-receiver azimuths, a form particularly suitable
dip-moveout correction (DMO) to account for the dip ef- for 3-D applications. Biondi and Ronen (1987), Cabrera
fect on moveout. Stacking of NMO- and DMO-corrected and Levy (1989), Granser (1994) and Zhou et al. (1996)
CMP gathers yields a section that is a closer approxima- designed DMO operators applied to shot profiles. All of
tion to a zero-offset section than a conventional CMP- these techniques are confined to vertical velocity vari-
stacked section based on NMO correction only. ations for which time migration is appropriate. How-
Conflicting dips with different stacking velocities ever, DMO correction cannot solve stack imperfections
often are encountered in two geological situations — caused by lateral velocity variations. Principles of and
reflections from steeply dipping fault planes conflicting techniques for DMO correction are described in Section
with reflections associated with gently dipping strata, 5.1, and practical aspects of DMO correction are cov-
and diffractions and reflections off the flank of salt ered in Section 5.2.
domes conflicting with, again, reflections associated While the practical solution to the problem of con-
with gently dipping strata. flicting dips with different stacking velocities is DMO
The conflicting dips problem has been investigated correction combined with poststack time migration, the
extensively. Doherty (1975) first introduced the wave rigorous solution is prestack time migration. A theory
extrapolation equations for nonzero-offset data. Sher- for imaging nonzero-offset data based on the double-
wood et al. (1978) devised a method for mapping square-root equation is provided in Section D.1. Re-
nonzero-offset data to zero offset in the presence of con- call from Section 4.1 that poststack migration, in prin-
flicting dips with different stacking velocities. Then, Yil- ciple, is based on summation of amplitudes along a
maz and Claerbout (1980) suggested a prestack partial zero-offset hyperbolic traveltime curve and placing the
migration (PSPM) technique for solving the problem summed amplitude to the apex of the hyperbola. Sim-
of conflicting dips. Specifically, they developed a wave ilarly, prestack time migration, in principle, involves
theory to account for the difference between migration summation of amplitudes along the nonzero-offset trav-
before stack and the result of conventional processing eltime curve in midpoint-offset coordinates and plac-
that includes moveout correction, CMP stacking and ing the summed amplitude at the apex of the sur-
migration after stack (Section D.1). They recognized face. The nonzero-offset traveltime equation can be de-
the fact that DMO correction actually is a partial mi- rived by stationary-phase approximation to the double-
gration process applied to moveout-corrected common- square-root equation (Section D.2). As with the zero-
offset data. Nevertheless, the PSPM theory had one im- offset case, the velocity field dictates the curvature of
portant drawback. Although valid for the layered earth the nonzero-offset summation paths. Each common-
velocity model, it was based on a small-offset approx- offset section can be imaged separately; the results
imation. Deregowski and Rocca (1981) recast the the- then can be superimposed to produce the migrated
ory for PSPM in a form similar to Kirchhoff migra- section. In practice, however, processing sequence for
tion. Ottolini (1982) developed the PSPM equations in prestack time migration often incorporates DMO cor-
Snell-midpoint coordinates, the domain of constant-ray- rection and repicking velocities after migration. Fowler
parameter sections (Section F.2). The technique is the- (1984) developed a velocity-independent prestack imag-
oretically accurate for layered medium as well as for ing technique that incorporates a step to correct for dip-
all offsets and dips. This method was followed by an- dependency of stacking velocities applied to constant-
other unique approach that involves offset continuation velocity stacked data. Gardner et al. (1986) combined
(Bolondi et al., 1982, 1984; Salvador and Savelli, 1982; DMO correction with a velocity-independent prestack
Bolondi and Rocca, 1985); that is, mapping a far-offset imaging technique based on constant-time slices of
section to a near-offset section, thereby collapsing all prestack data in midpoint-offset coordinates. Bancroft
offsets to zero offset. Hale (1983, 1984) formulated a and Geiger (1994) developed a prestack imaging tech-
DMO method in the f −k domain. This method is exact nique also based on constant-time slices of prestack
for constant velocity, it can handle all dips and offsets, data, but in shot-receiver coordinates. Practical aspects
Dip-Moveout Correction and Prestack Migration 657

of prestack time migration are discussed in Section 5.3, fault and the growth faults in comparison with the con-
while techniques for migration velocity analysis are de- ventional CMP stack (Figure 5.0-2a). Migration of the
scribed in Section 5.4. DMO stack (Figure 5.0-2d) yields a crisp image of the
fault planes associated with the growth faults in con-
trast with the image from migration of the conventional
Salt-Flank Reflections stack (Figure 5.0-2b).

Consider the stacked section in Figure 5.0-1a. While


picking velocities in favor of the flat reflections, the DMO and Stacking Velocities
steep flank of the diffraction hyperbola off the tip of
the salt dome and the steeply dipping reflection off the We now closely examine an aspect of dip-moveout cor-
flank of the salt dome are not stacked with sufficient
rection in relation to stacking velocities. Consider the
strength. This inadequate definition of the diffraction
CMP stack in Figure 5.0-3a and the DMO stack in
and the steeply dipping reflection then causes migration
Figure 5.0-3b, and note that the steeply dipping re-
after stack to respond somewhat poorly (Figure 5.0-1b).
The stack with dip-moveout correction preserves the flections are better preserved by the latter. We know
diffraction and the steeply dipping reflection as well as from equation (3-8) that the steeper the dip the higher
the flat reflections (Figure 5.0-1c). As a result, migra- the stacking velocities. Figure 5.0-4 shows constant-
tion yields an image of the salt diapir that stands out velocity-stack (CVS) panels in the neighborhood of mid-
clearly in contrast with the surrounding strata (Figure point A as denoted in Figure 5.0-3a using CMP gath-
5.0-1d). ers without and with DMO correction. Note that gen-
It is important to evaluate the two migrated sec- tly dipping reflections and associated multiples stack at
tions (Figures 5.0-1b and d) from the viewpoint of an low velocities, whereas steeply dipping reflections stack
interpreter. Surely, by using the termination points of at high velocities. After DMO correction, the velocities
the nearly flat events in Figure 5.0-1b, one can trace for the steeply dipping events have been corrected for
the salt boundary without the need for proper stack- the dip effect; hence both gently dipping and steeply
ing and imaging the reflection off the flank. However, dipping reflections stack with equal strength at low ve-
by preserving this reflection on the stacked section and locities.
migrating it properly (Figure 5.0-1d), we attain a clue Consider velocity analysis at midpoint A in Fig-
to migration velocity errors. Specifically, if the steeply ure 5.0-3a where the steeply dipping reflections are in
dipping event that defines the salt boundary and the
conflict with the reflections associated with the gently
surrounding flat events cross over each other, we know
dipping strata. The dip effect on stacking velocities can
that overmigration has occurred, and as a result, the
be clearly observed on the velocity spectrum shown in
salt diapir has not been delineated accurately.
Figure 5.0-5a. The group of semblance peaks on the
spectrum in Figure 5.0-5a denoted by A represent the
steeply dipping events which conflict with the nearly-
Fault-Plane Reflections flat events as seen in the stacked section in Figure 5.0-
3b in the neighborhood of midpoint A. We would have
The section in Figure 5.0-2a contains a major listric a problem when trying to pick a velocity function from
fault that dips down from left to right. Also, there are this velocity spectrum. We normally would pick along
many auxiliary faults that intersect this main fault and the predominant velocity trend as denoted in Figure
several antithetic faults which do not intersect it. These 5.0-5a. This leads to rejection of the picks associated
fault patterns commonly occur in areas in which exten-
with the steeply dipping reflections and a significant re-
sion has taken place. We see reflections associated with
duction in their amplitudes on the stacked section (Fig-
the sedimentary strata, as well as faint indications of
ure 5.0-3a). Following DMO correction, the duality in
fault-plane reflections. Along major fault zones, such as
those in Figure 5.0-2a, there is often a clear case of con- velocity picks are eliminated, and the velocities are cor-
flicting dips. Migration of the stacked section (Figure rected for dip as shown in Figure 5.0-5b. Also, note the
5.0-2b) positions the events dipping down from right improved velocity trend after DMO correction.
to left with an acceptable accuracy. Nevertheless, the Figure 5.0-6 shows a close-up view of the CMP
listric fault can be inferred and traced only by follow- gather at midpoint A as in Figure 5.0-3a without DMO
ing the termination points of the dipping events. correction, the velocity spectrum, and the gather after
The DMO stack shown in Figure 5.0-2c has pre- NMO correction using the velocity function denoted in
served the conflicting dips associated with the listric the velocity spectrum. The overcorrected events on the
658 Seismic Data Analysis

FIG. 5.0-1. Conflicting dips associated with salt flanks: (a) CMP stack without dip-moveout correction; (b) time migration
of the stack in (a); (c) the stack with dip-moveout correction; (d) time migration of the stack in (c).
Dip-Moveout Correction and Prestack Migration 659
660 Seismic Data Analysis

FIG. 5.0-3. (a) A CMP stack, and (b) a DMO stack. See Figures 5.0-4 through 5.0-11 for the results of velocity analyses
performed at midpoint locations A and B using CMP gathers before and after DMO correction.
Dip-Moveout Correction and Prestack Migration 661
662 Seismic Data Analysis

FIG. 5.0-5. Velocity spectrum at midpoint location A as in Figure 5.0-3 without (a) and with (b) DMO correction. The
group of semblance peaks on the spectrum in (a) denoted by A represents the steeply dipping events which conflict with the
nearly flat events as seen in the stacked section in Figure 5.0-3b in the neighborhood of midpoint A.

moveout-corrected gather represent the steeply dipping The case in the neighborhood of midpoint A in
reflections, and the undercorrected events represent the Figure 5.0-3a is that of reflections with conflicting
multiple reflections. dips. Often, on stacked data, diffractions and reflections
The close-up view of the CMP gather in Figure also form conflicting dips. Figure 5.0-8 shows constant-
5.0-6 after DMO correction is shown in Figure 5.0- velocity-stack (CVS) panels in the neighborhood of mid-
7. Note the more rigorous delineation of the velocity point B as denoted in Figure 5.0-3a using CMP gathers
trend in the spectrum as compared to that in Figure without and with DMO correction. Note that gently
5.0-6. The absence of the overcorrected events in the dipping reflections stack at low velocities, whereas the
moveout-corrected gather is a convincing evidence that diffraction off the tip of the salt diapir stack at high
the velocities of the steeply dipping reflections have velocities. After DMO correction, the velocities for the
been corrected for the dip effect as inferred by equation steeply dipping events have been corrected for the dip
(3-8). Where are the overcorrected events that we see effect; hence both gently dipping reflections and the
in Figure 5.0-6c? Are they still present in the moveout- diffraction stack with equal strength at low velocities.
corrected gather after DMO correction shown in Figure Consider velocity analysis at midpoint B as in Fig-
5.0-7c? They should not be, because both flat events ure 5.0-3a where the diffraction off the tip of the salt
and dipping events could not be flattened simultane- diapir are in conflict with the reflections associated with
ously in the same gather. Moreover, the semblance peak the gently dipping strata. The dip effect on stacking
in the velocity spectrum denoted by A in Figure 5.0-5a velocities can be clearly observed on the velocity spec-
is absent in Figure 5.0-5b after dip-moveout correction. trum shown in Figure 5.0-9a. The semblance peak on
Did it merge with the semblance peaks aligned with the the spectrum denoted by B represents the diffraction off
velocity trend denoted in Figure 5.0-5b? No, it did not; the tip of the salt dome which conflicts with the nearly
instead, the energy represented by the semblance peak flat events as seen in the stacked section in Figure 5.0-
A in Figure 5.0-5a actually has moved to another CMP 3b in the neighborhood of midpoint B. Again, we would
location in the updip direction. So, we begin to have normally pick along the predominant velocity trend as
a clue to another aspect of dip-moveout correction — denoted in Figure 5.0-9a. This leads to rejection of the
that it may not be a moveout correction after all. We pick associated with the diffraction, which would mean
shall examine this clue closely later in Section 5.1. failure of preserving it on the stacked section (Figure
Dip-Moveout Correction and Prestack Migration 663

FIG. 5.0-6. A close-up view of (a) the CMP gather at midpoint A as in Figure 5.0-3 without DMO correction, (b) the
velocity spectrum, and (c) the CMP gather as in (a) after NMO correction using the velocity function denoted in the velocity
spectrum in (b).

FIG. 5.0-7. A close-up view of (a) the CMP gather at midpoint A as in Figure 5.0-3 with DMO correction, (b) the velocity
spectrum, and (c) the CMP gather as in (a) after NMO correction using the velocity function denoted in the velocity spectrum
in (b).
664 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 665

FIG. 5.0-9. Velocity spectrum at midpoint location B as in Figure 5.0-3 without (a) and with (b) DMO correction. The
semblance peak on the spectrum in (a) denoted by B represents the diffraction off the tip of the salt dome which conflicts
with the nearly flat events as seen in the stacked section in Figure 5.0-3b in the neighborhood of midpoint B.

5.0-3a). Following DMO correction, the duality in veloc- cause the steep events associated with the diffraction
ity picks is eliminated, and the velocities are corrected energy and the dipping reflections have been preserved
for dip as shown in Figure 5.0-9b. in the DMO stack (Figure 5.0-3b), time migration of
Figure 5.0-10 shows a close-up view of the CMP this section produces an image of the salt dipair with
gather at midpoint B as in Figure 5.0-3a without DMO its boundaries clearly delineated (Figure 5.0-12b). In
correction, the velocity spectrum, and the gather after contrast, the boundaries of the salt diapir in the CMP
NMO correction using the velocity function denoted in stack can be delineated only by the terminations of the
the velocity spectrum. The overcorrected event at 3.75 s gently dipping strata (Figure 5.0-12a).
on the moveout-corrected gather represents the diffrac-
tion, and the undercorrected events represent the mul-
tiple reflections.
Turning-Wave Reflections
The close-up view of the CMP gather as in Figure
5.0-10 after DMO correction is shown in Figure 5.0-
11. Note the more rigorous delineation of the velocity When salt tectonism is at an advanced stage, it can
trend in the spectrum as compared to that in Figure 5.0- cause overhang structures. If velocities within the
10. The overcorrected event associated with the diffrac- surrounding sedimentary strata increase rapidly with
tion is absent in the moveout-corrected gather — once depth, it is possible to record reflections from the un-
again, convincing evidence that the diffraction velocity derside of the salt overhang. These reflections are as-
has been corrected for the dip effect. The diffraction sociated with turning waves that first travel downward
energy, however, that gave rise to the peak denoted by after they bounce off the underside of the salt overhang
B in Figure 5.0-9a is absent in the moveout-corrected before returning to the surface. Turning-wave reflec-
gather after DMO correction shown in Figure 5.0-11c. tions give rise to a dipping event on the stacked section
Dip-moveout correction, unlike conventional normal- that conflicts with the reflections associated with the
moveout correction, has caused energy to move from surrounding gently dipping strata. Nevertheless, dip-
one CMP location to another. moveout correction alone does not strictly preserve the
Figure 5.0-12 shows time migrations of the CMP- dipping event associated with the turning waves, since
stacked and DMO-stacked sections in Figure 5.0-3. Be- these waves propagate at angles greater than 90 degrees.
666 Seismic Data Analysis

FIG. 5.0-10. A close-up view of (a) the CMP gather at midpoint B as in Figure 5.0-3 without DMO correction, (b) the
velocity spectrum, and (c) the CMP gather as in (a) after NMO correction using the velocity function depicted in the velocity
spectrum in (b).

FIG. 5.0-11. A close-up view of (a) the CMP gather at midpoint B as in Figure 5.0-3 with DMO correction, (b) the velocity
spectrum, and (c) the CMP gather as in (a) after NMO correction using the velocity function depicted in the velocity spectrum
in (b).
Dip-Moveout Correction and Prestack Migration 667

FIG. 5.0-12. Time migrations of (a) the CMP stack, and (b) the DMO stack as in Figure 5.0-3.
668 Seismic Data Analysis

FIG. 5.0-13. (a) Conventional poststack time migration; (b) turning-wave migration. (Courtesy Schlumberger Geco-Prakla.)

They can be preserved only by making use of its ab- zero-offset, the data can then be migrated either before
normal moveout behavior during stacking and imaged or after stack using the zero-offset theory for migration
properly by making use of the evanescent energy associ- as described in Chapter 4.
ated with dips greater than 90 degrees during migration Figure 5.1-1a depicts the nonzero-offset record-
(Section 5.1). ing geometry associated with a dipping reflector. The
Figure 5.0-13 shows a field data example of turning- nonzero-offset traveltime t = SRG/v is measured along
wave migration. The salt boundary from the image the raypath from source S to reflection point R to re-
based on phase-shift migration of the conventional stack ceiver G, where v is the velocity of the medium above
can be inferred only from the termination of the reflec- the dipping reflector. This arrival time is depicted on
tors associated with the surrounding sedimentary se- the time section in Figure 5.1-1b by point A on the
quence. Whereas, the image based on the turning-wave trace that coincides with midpoint yn . We want to map
migration of the stack that preserves the turning-wave the amplitude at time t denoted by the sample A on
energy shows the overhang structure on either side of the trace at midpoint yn of the common-offset section
the salt intrusion, distinctively. with offset 2h to time τ0 denoted by the sample C on
the trace at midpoint y0 of the zero-offset section. We
achieve this mapping in two steps:
5.1 PRINCIPLES OF DIP-MOVEOUT
CORRECTION (a) Normal-moveout correction that maps the ampli-
tude at time t denoted by the sample A on the
The objective we want to achieve with the combina- trace at midpoint yn of the common-offset section
tion of normal-moveout and dip-moveout correction is with offset 2h to time tn denoted by the sample
mapping nonzero-offset data to the plane of zero-offset B on the same trace at midpoint yn of the same
section. Once each common-offset section is mapped to common-offset section.
Dip-Moveout Correction and Prestack Migration 669

Zero-offset migration then maps the amplitude at time


τ0 denoted by the sample C on the trace at midpoint
y0 of the zero-offset section to the amplitude at time τ
denoted by the sample D on the trace at midpoint ym
of the migrated section. Note that the combination of
NMO correction, DMO correction, and zero-offset mi-
gration achieves the same objective as direct mapping
of the amplitude at time t denoted by the sample A on
the trace at midpoint yn of the common-offset section
with offset 2h to the amplitude at time τ denoted by the
sample D on the trace at midpoint ym of the migrated
section. This direct mapping procedure is the basis of
algorithms for migration before stack (Section 5.3).
The important point to note is that the normal-
moveout correction in step (a) is performed using the
velocity of the medium above the dipping reflector.
The NMO equation (3-8) defines the traveltime t
from source location S to the reflection point R to the
receiver location G. This equation, written in prestack
data coordinates, is
4h2 cos2 φ
t2 = t20 + , (5 − 1)
v2
where 2h is the offset, v is the medium velocity above
the reflector, φ is the reflector dip, and t0 is the two-way
zero-offset time at midpoint location yn .
Dip-moveout correction of step (b) is preceded by
zero-dip normal-moveout correction of step (a) using
FIG. 5.1-1. (a) The geometry of a nonzero-offset recording
of reflections from a dipping layer boundary; (b) a sketch the dip-independent velocity v:
of the time section depicting the various traveltimes. NMO 4h2
correction involves coordinate transformation from yn − t to t2 = t2n + , (5 − 2)
yn − tn by mapping amplitude A at time t to B at time
v2
tn on the same trace. DMO correction involves coordinate where tn is the event time at midpoint yn after the NMO
transformation from yn − tn to y0 − τ0 by mapping ampli- correction. Event time tn after the NMO correction and
tude B at time tn on the trace at midpoint location yn of event time t0 are related as follows (Section E.2)
the moveout-corrected common-offset section to amplitude
C at time τ0 on the trace at midpoint location y0 of the 4h2 sin2 φ
zero-offset section. Zero-offset migration involves coordinate t2n = t20 − . (5 − 3)
v2
transformation from y0 − τ0 to ym − τ by mapping ampli-
tude C at time τ0 on the trace at midpoint location y0 of At first glance, equations (5-2) and (5-3) suggest a
the zero-offset section to amplitude D at time τ on the trace two-step approach to moveout correction:
at midpoint location ym of the migrated section. Migration
before stack involves direct mapping of amplitude A at time (a) Apply a dip-independent moveout correction using
t on the trace at midpoint location yn of the common-offset
equation (5-2) to map the amplitude at time t de-
section to amplitude D at time τ on the trace at midpoint
location ym of the migrated section. See text for the rela- noted by the sample A on the trace at midpoint yn
tionships between the coordinate variables. of the common-offset section with offset 2h to time
tn denoted by the sample B on the same trace at
(b) Dip-moveout correction that maps the amplitude midpoint yn of the same common-offset section.
at time tn denoted by the sample B on the trace (b) Apply a dip-dependent moveout correction using
at midpoint yn of the moveout-corrected common- equation (5-3) to map the amplitude at time tn
offset section with offset 2h to time τ0 denoted by denoted by the sample B on the trace at midpoint
the sample C on the trace at midpoint y0 of the yn of the moveout-corrected common-offset section
zero-offset section. with offset 2h to time t0 denoted by the sample
670 Seismic Data Analysis

B on the same trace at midpoint yn of the same the DMO correction is given by
common-offset section.
∆yDM O = |yn − y0 |, (5 − 8a)
This two-step moveout correction is equivalent to the which can be expressed by way of equations (5-4a) and
one-step moveout correction using equation (5-1) to (5-5) as
map event time t directly to event time t0 . h2 2 sin φ
Our goal, however, is to map event time t not to ∆yDM O = . (5 − 8b)
tn A v
t0 — the two-way zero-offset time associated with mid-
point yn between source S and receiver G, but to τ0 The difference between the input time tn and the
— the two-way zero-offset time at midpoint location y0 output time τ0 is defined by
associated with the normal-incidence reflection point R
∆tDM O = tn − τ0 , (5 − 9a)
(Figure 5.1-1). The relationships between (yn , tn ) co-
ordinates of the normal-moveout-corrected data and which can be expressed by way of equations (5-4b) and
(y0 , τ0 ) coordinates of the dip-moveout-corrected data (5-5) as
are given by (Section E.2):
1
∆tDM O = tn 1 − . (5 − 9b)
h2 2 sin φ A
y0 = yn − , (5 − 4a)
tn A v Finally, as sketched in Figure 5.1-1, the reflec-
and tion point dispersal ∆ = N R is defined by the dis-
tn tance along the dipping reflector between the normal-
τ0 = , (5 − 4b) incidence points N and R associated with midpoints yn
A
where and y0 , respectively. By way of equations (E-18) and
(5-8a) it follows that (Section E.1)
2
h2 2 sin φ
A= 1+ . (5 − 5) h2 sin 2φ
t2n v ∆= . (5 − 10)
tn A v
For completeness, the relationship between event times
Note from equation (5-10) that reflection point dispersal
tn and t0 is given by (Section E.2)
is nill for zero offset, and increases with the square of
t0 = tn A. (5 − 6) the offset. Also, the larger the dip and shallower the
reflector, the larger the dispersal.
Note from equation (5-5) that A ≥ 1; therefore, τ0 ≤ tn
A direct consequence of equation (5-10) is that a
(equation 5-4b) and t0 ≥ tn (equation 5-6).
reflection event on a CMP gather is associated with
Refer to Figure 5.1-1 and note that the normal-
more than one reflection point on the reflector. Follow-
moveout correction that precedes the dip-moveout cor-
ing DMO correction, reflection-point dispersal is elimi-
rection maps the amplitude at sample A with coordi-
nated and, hence, the reflection event is associated with
nates (yn , t) to sample B with coordinates (yn , tn ). So,
a single reflection point at normal-incidence (point R in
the midpoint coordinate is invariant under NMO correc-
Figure 5.1-1). While prestack data before DMO cor-
tion. The difference between the input time t and the
output time tn is defined by rection can be associated with common midpoints, and
thus sorted into common-midpoint (CMP) gathers; af-
∆tN M O = t − tn , (5 − 7a) ter DMO correction, the data can be associated with
which can be expressed by way of equation (5-2) as common reflection points, and thus can be considered
follows in the form of common-reflection-point (CRP) gathers.

∆tN M O = tn (An − 1), (5 − 7b)


where Prestack Partial Migration
2
h2 2
An = 1+ . (5 − 7c) While conventional normal-moveout correction involves
t2n v only a time shift given by equation (5-7b), dip-moveout
Again, refer to Figure 5.1-1 and note that the dip- correction involves mapping both in time and space
moveout correction maps the amplitude at sample B given by equations (5-8b) and (5-9b), respectively. This
with coordinates (yn , tn ) to sample C with coordinates means that dip-moveout correction, strictly speaking, is
(y0 , τ0 ). So, the midpoint coordinate is variant under not a moveout correction in conventional terms; rather,
DMO correction. The lateral excursion associated with it is a process of partial migration before stack applied
Dip-Moveout Correction and Prestack Migration 671

to common-offset data. We therefore may speak of a (d) Compare the values for ∆yDM O and ∆tDM O in
dip-moveout operator with a specific impulse response Tables 5-1 and 5-2, and note that the lower the
as for the migration process itself. Following this par- velocity, the larger the DMO correction. This also
tial migration to map nonzero-offset data to the plane implies that the shallower the event, the more sig-
of zero-offset, each common-offset section is then fully nificant the DMO term, since lower velocities gen-
migrated using a zero-offset migration operator. erally are found in shallow parts of the seismic data.
A dip-moveout operator maps amplitudes on a (e) For a specific reflector dip φ, compare the values for
moveout-corrected trace of a common-offset section ∆yDM O and ∆tDM O in Tables 5-2 and 5-3, and
along its impulse response trajectory. Before we derive note that the larger the offset 2h, the more the
the expression for its impulse response, we shall first DMO correction. Whatever the reflector dip, DMO
make some inferences about the DMO process based correction has no effect on zero-offset data with
on equations (5-8b) and (5-9b). Tables 5-1, 5-2, and
h = 0.
5-3 show horizontal (∆yDM O ) and vertical (∆tDM O )
displacements associated with dip-moveout correction
described by equations (5-8b) and (5-9b), respectively. Table 5-2. Horizontal (∆yDM O ) and vertical (∆tDM O )
Combined with equations (5-8b) and (5-9b), we make
displacements associated with dip-moveout correction
the following observations:
described by equations (5-8b) and (5-9b), respectively,
for a given moveout-corrected time variable tn and ve-
(a) Set φ = 0 in equations (5-8b) and (5-10b), and
note that ∆yDM O = 0 and ∆tDM O = 0. Hence, locity v. The dip angle φ = 30 degrees, half-offset
the DMO operator has no effect on a flat reflector, h = 1500 m. Also given are the corresponding values
irrespective of the offset. The steeper the dip, the for the reflection point smear ∆ as in equation (5-10).
larger the DMO correction.
tn v ∆yDM O ∆tDM O ∆
(b) Note from Table 5-1 that the horizontal dis-
(s) (m/s) (m) (s) (m)
placement ∆yDM O and the vertical displacement
∆tDM O decrease with time tn . This means that
0.5 1800 1, 284 0.243 1, 111
the spatial aperture of the dip-moveout operator,
in contrast with a migration operator, actually de- 1.0 2000 900 0.200 779
creases with event time. 1.5 2200 620 0.135 536
(c) Substitute equation (5-5) into equation (5-8b) and 2.0 2400 446 0.090 386
note that, in the limit tn = 0, ∆yDM O = h. This 2.5 2700 324 0.060 280
means that the largest spatial extent of the DMO 3.0 3000 246 0.036 213
operator equals the offset 2h associated with the 4.0 4000 140 0.016 121
moveout-corrected trace at tn = 0.

Table 5-3. Horizontal (∆yDM O ) and vertical (∆tDM O )


Table 5-1. Horizontal (∆yDM O ) and vertical (∆tDM O )
displacements associated with dip-moveout correction
displacements associated with dip-moveout correction
described by equations (5-8b) and (5-9b), respectively,
described by equations (5-8b) and (5-9b), respectively,
for a given moveout-corrected time variable tn and ve-
for a given moveout-corrected time variable tn and ve-
locity v. The dip angle φ = 30 degrees, half-offset
locity v. The dip angle φ = 30 degrees, half-offset
h = 1500 m. Also given are the corresponding values h = 500 m. Also given are the corresponding values
for the reflection point smear ∆ as in equation (5-10). for the reflection point smear ∆ as in equation (5-10).

tn v ∆yDM O ∆tDM O ∆ tn v ∆yDM O ∆tDM O ∆


(s) (m/s) (m) (s) (m) (s) (m/s) (m) (s) (m)

0.5 2400 1, 170 0.188 1, 013 0.5 1800 241 0.064 209
1.0 2400 793 0.152 687 1.0 2000 121 0.030 105
1.5 2400 575 0.115 497 1.5 2200 74 0.015 64
2.0 2400 446 0.090 386 2.0 2400 51 0.010 44
2.5 2400 363 0.075 314 2.5 2700 37 0.008 32
3.0 2400 305 0.063 264 3.0 3000 27 0.004 23
4.0 2400 230 0.008 199 4.0 4000 15 0.000 13
672 Seismic Data Analysis

(f) Finally, note from Tables 5-1, 5-2 and 5-3 that the The amplitude scaling (2A2 − 1)/A3 in equation (5-14a)
reflection point smear ∆ given by equation (5-10) is by Black et al. (1993), and is represented by A−1 in
decreases in time and for small offsets. the original derivation by Hale (1984). The difference
is due to the fact that Hale (1984) defined the output
time variable for DMO correction as t0 of equation (5-
Frequency-Wavenumber DMO Correction 6), whereas Black et al. (1993) correctly defined the out-
put time variable as τ0 of equation (5-4b). Fortunately,
Refer to Figure 5.1-1 and recall that our objective with
the phase term exp(−iω0 tn A) as in equation (5-14a) is
DMO correction is to transform the normal-moveout-
identical in the case of both derivations. There is one
corrected prestack data Pn (yn , tn ; h) from yn − tn coor-
dinates to y0 − τ0 coordinates so as to obtain the dip- other variation of the amplitude term by Liner (1989)
moveout-corrected zero-offset data P0 (y0 , τ0 ; h). Note, and Bleistein (1990) given by (2A2 −1)/A. Nevertheless,
however, the transformation equations (5-4a) and (5- within the context of a conventional processing sequence
4b) require knowledge of the reflector dip φ to perform which includes geometric spreading correction prior to
the DMO correction. To circumvent this requirement, DMO correction, the amplitude scaling (2A2 − 1)/A3
Hale (1984) developed a method for DMO correction described here preserves relative amplitudes.
in the frequency-wavenumber domain. First, we use the We now outline the steps in dip-moveout correction
relation from Section D.1 in the frequency-wavenumber domain:
vky
sin φ = , (5 − 11)
2ω0
(a) Start with prestack data in midpoint-offset y − h
which states that the reflector dip φ can be expressed coordinates, P (y, h, t) and apply normal moveout
in terms of wavenumber ky and frequency ω0 , which correction using a dip-independent velocity v.
are the Fourier duals of midpoint y0 and event time τ0 , (b) Sort the data from moveout-corrected CMP
respectively. By way of equation (5-11), the transfor- gathers Pn (yn , h, tn ) to common-offset sections
mation equations (5-4a) and (5-4b) are recast explicitly
Pn (yn , tn ; h).
independent of reflector dip as
(c) Perform Fourier transform of each common-offset
h2 ky section in midpoint yn direction, Pn (ky , tn ; h).
y0 = y n − , (5 − 12a)
tn Aω0 (d) For each output frequency ω0 , apply the phase-shift
and exp(−iω0 tn A), scale by (2A2 − 1)/A3 , and sum the
tn resulting output over input time tn as described by
τ0 =
, (5 − 12b)
A equation (5-14a).
where A of equation (5-5) now is of the form (e) Finally, perform 2-D inverse Fourier transform to
obtain the dip-moveout corrected common-offset
h2 ky2 section P0 (y0 , τ0 ; h) (equation 5-14b).
A= 1+ . (5 − 13)
t2n ω02
The frequency-wavenumber domain dip-moveout A flowchart of the dip-moveout correction described
correction process that transforms the normal-moveout- above is presented in Figure 5.1-2.
corrected prestack data with a specific offset 2h from We shall now test the frequency-wavenumber DMO
yn − tn domain to y0 − τ0 domain is achieved by the correction using modeled data for point scatterers and
integral
dipping events. Figure 5.1-3 depicts six point scatterers
2A2 − 1 buried in a constant-velocity medium. A synthetic data
P0 (ky , ω0 ; h) =
A3 set that comprises 32 common-offset sections, each with
× Pn (ky , tn ; h) exp(−iω0 tn A)dtn . 63 midpoints, was created. The offsets range is from 0
(5 − 14a) to 1550 m with an increment of 50 m.
Figure 5.1-4 shows two constant-velocity stacks
Derivation of the integral transform of equation (5-14a) (CVS) of the CMP gathers from the synthetic data
is given in Section E.2. set associated with the velocity-depth model depicted
Once dip-moveout correction is applied, the data in Figure 5.1-3. The offset range used in stacking is
are inverse Fourier transformed
50 − 1550 m. At the apex of the traveltime trajectory
P0 (y0 , τ0 ; h) = P0 (ky , ω0 ; h) for each point scatterer, the event dip is zero. There-
(5 − 14b) fore, stack response is best with moveout velocity equal
× exp(−iky y0 + iω0 τ0 )dky dω0 . to the medium velocity (3000 m/s). Along the flanks
Dip-Moveout Correction and Prestack Migration 673

FIG. 5.1-3. Depth model of six point scatterers buried in


a constant-velocity medium. The asterisks indicate the po-
sitions of the point scatterers.

(2) The stacked section derived from these gathers


(Figure 5.1-5c) is shown in Figure 5.1-4b. Because
medium velocity was used for NMO correction, the
stack response is best for zero dip. Note the poor
FIG. 5.1-2. A flowchart for frequency-wavenumber dip-
moveout correction algorithm. The scalar A is given by equa- stack response along the steeply dipping flanks.
tion (5-13) and B = (2A2 − 1)/A3 as in equation (5-14a). The desired section is the zero-offset section in Fig-
ure 5.1-4a.
(3) We sort the NMO-corrected gathers (Figure 5.1-5c)
of the traveltime trajectories, optimum stack response into common-offset sections for DMO processing.
varies as the event dip changes. The steeper the dip, the These are shown in Figure 5.1-6a.
higher the moveout (or stacking) velocity. (4) Each common-offset section is individually cor-
Selected common-offset sections associated with rected for dip moveout. The impulse responses of
the subsurface model in Figure 5.1-3 are shown in Fig- the dip-moveout operator for the corresponding off-
ure 5.1-5a. The well-known nonhyperbolic table-top tra- sets are shown in Figure 5.1-6b, and the resulting
jectories are apparent at large offsets. Selected CMP common-offset sections are shown in Figure 5.1-6c.
gathers from the model of Figure 5.1-3 are shown in Note the following effects of DMO:
Figure 5.1-5b. Only selected gathers that span the right
side of the center midpoint are displayed, since the
(a) DMO is a partial migration process. The flanks
common-offset sections are symmetric with respect to
the center midpoint (CMP 32). Note that the travel- of the nonhyperbolic trajectories have been moved
times at the center midpoint are perfectly hyperbolic, updip just enough to make them look like zero-
while the traveltimes at CMP gathers away from the offset trajectories, which are hyperbolic. As a re-
center are increasingly nonhyperbolic. sult, each common-offset section after NMO and
The following DMO processing is applied to the DMO corrections is approximately equivalent to
data as in Figure 5.1-5a: the zero-offset section (Figure 5.1-4a).
(b) This partial migration is subtly different from con-
(1) Figure 5.1-5c shows the NMO-corrected gathers, ventional migration in one respect. Unlike conven-
with stretch muting applied. The medium velocity tional migration, note from the impulse responses
(3000 m/s) was used for NMO correction (equa- in Figure 5.1-6b that the dip-moveout correction
tion 5-2), an essential requirement for subsequent becomes greater at increasingly shallow depths.
DMO correction. As a result, the events at and in (c) While it does nothing to the zero-offset section, dip-
the vicinity of the center midpoint (CMP 32) are moveout correction also is greater at increasingly
flat after NMO correction, while the events at mid- large offsets (Figure 5.1-6c).
points away from the center midpoint are increas- (d) Finally, as with conventional migration, the steeper
ingly overcorrected. the event, the greater partial migration takes place,
674 Seismic Data Analysis

FIG. 5.1-4. Stack response of six point scatterers buried in a constant-velocity earth model (3000 m/s) as depicted in Figure
5.1-3: (a) zero-offset section, (b) stack with NMO velocity of 3000 m/s, (c) stack with NMO velocity of 3600 m/s.

FIG. 5.1-5. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-3: (a) common-offset sections with offset range from 50 to 1550 m and an increment of 300 m; (b) CMP gathers
sorted from the common-offset sections as in (a) at midpoint locations from 32 to 63 as denoted in Figure 5.1-3 with an
increment of 3; (c) the CMP gathers as in (b) after NMO correction and muting.
Dip-Moveout Correction and Prestack Migration 675

FIG. 5.1-6. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-3: (a) common-offset sections with offset range from 50 to 1550 m and an increment of 300 m sorted from the
NMO-corrected gathers as in Figure 5.1-4c; (b) impulse responses of the DMO operators applied to the common-offset gathers;
(c) common-offset sections as in (a) after DMO correction; (d) CMP gathers sorted from the common-offset sections as in (c)
at midpoint locations from 32 to 63 as denoted in Figure 5.1-3 with an increment of 3.

FIG. 5.1-7. (a) Zero-offset section associated with the depth model in Figure 5.1-3, (b) stack derived from the CMP gathers
as in Figure 5.1-5c, (c) DMO stack derived from the CMP gathers as in Figure 5.1-6d.
676 Seismic Data Analysis

FIG. 5.1-8. DMO processing of dipping events: (a) zero-offset section with the medium velocity of 3500 m/s; (b) stack using
optimum velocity picks from velocity spectra along the line, such as that shown in Figure 5.1-12a; (c) stack using the medium
velocity of 3500 m/s; (d) DMO stack using velocity picks from velocity spectra along the line, such as that shown in Figure
5.1-12b. Location A refers to an example of events with conflicting dips.

with flat events remaining unaltered (Figure 5.1- (Figure 5.1-7a) than the stacked section without
6c). the DMO correction (Figure 5.1-7b). Note the en-
hanced stack response along the steeply dipping
(5) Following the DMO correction, the data are sorted flanks in Figure 5.1-7c. (The sections all have the
back to CMP gathers (Figure 5.1-6d). Compare the same display gain.)
gathers in Figure 5.1-6d to the CMP gathers with-
out DMO correction (Figure 5.1-5b). The DMO We now examine results of DMO processing of
correction has left the zero-dip events unchanged a modeled data set for dipping events. Figure 5.1-8a
(at and in the vicinity of CMP 32), while it sub- shows a zero-offset section that consists of events with
stantially corrected steeply dipping events on the dips from 0 to 45 degrees with a 5-degree increment.
CMP gathers away from the center midpoint (CMP Medium velocity is constant (3500 m/s). Several veloc-
32). The events on the CMP gathers now are flat- ity analyses were performed along the line; an exam-
tened (Figure 5.1-6d). Also, since DMO correction ple is shown in Figure 5.1-9a. Note the dip-dependent
is a migration-like process, it causes the energy to semblance peaks. Selected CMP gathers are shown in
move from one CMP gather to neighboring gath- Figure 5.1-10a. By using the optimum stacking veloci-
ers in the updip direction. Energy depletion at the ties picked from the densely spaced velocity analyses, we
CMP gathers in Figure 5.1-6d farther from the cen- apply NMO correction to the CMP gathers (Figure 5.1-
ter midpoint occurred because there was no other 10b), then stack them (Figure 5.1-8b). Aside from the
CMP gather to contribute energy beyond CMP 63. conflicting dips at location A, stack response is close to
(6) Stacking the NMO- and DMO-corrected gathers the zero-offset section (Figure 5.1-8a). The DMO pro-
(Figure 5.1-6d) yields a section (Figure 5.1-7c) cessing requires NMO correction using medium velocity
that more closely represents the zero-offset section (Figure 5.1-10c). Stack response using the medium ve-
Dip-Moveout Correction and Prestack Migration 677

as a result, the DMO correction is achieved by a sim-


ple multiplication of the input data with a phase-shift
operator in the Fourier transform domain.
Define the following logarithmic variables that cor-
respond to the time variables τ0 and tn of equation (5-
12b):
T0 = ln τ0 , (5 − 15a)
and
Tn = ln tn , (5 − 15b)
where, for convenience, a constant scalar with its unit
in time is omitted. Hence, the inverse relationships are
given by
τ0 = eT0 , (5 − 16a)
and
tn = eTn . (5 − 16b)
Our goal is to derive equations for DMO correction
in the log-stretch coordinates (y0 , T0 ). The transform
relation between the input log-stretch time variable Tn
and the output log-stretch time variable T0 is given by
FIG. 5.1-9. Velocity analysis (a) without and (b) with
DMO correction at analysis. The stacked sections with- T0 = Tn − ln Ae , (5 − 17a)
out and with DMO correction are shown in Figures 5.1-8b and the expression for the midpoint variable y0 in the
and d.
log-stretch domain is given by

locity (Figure 5.1-8c) clearly degrades at steep dips. By h2 ky


y0 = y n − , (5 − 17b)
applying DMO correction (Figure 5.1-10d) to the NMO- Ae Ω0
corrected gathers (Figure 5.1-10b), we get the improved where
stacked section in Figure 5.1-8d. The DMO stack is clos-
est to the zero-offset section (Figure 5.1-8a). h2 ky2
Ae = 1+ . (5 − 18)
DMO correction also yields dip-corrected velocity Ω20
functions that can be used in subsequent migration. Re- The variable Ω0 is the Fourier transform dual of the
fer to the velocity analysis in Figure 5.1-9b and note variable T0 in the log-stretch domain. Equations (5-
that all events have semblance peaks at 3500 m/s, which 17a,b) and (5-18) correspond to equations (5-12a,b) and
is the medium velocity for this model data set. (5-13) in the log-stretch domain. Mathematical details
of the derivation of equations (5-17a,b) are left to Sec-
tion E.3.
Log-Stretch DMO Correction The log-stretch dip-moveout correction process is
achieved by the following relationship (Section E.3):
The frequency-wavenumber DMO correction (Hale, h2 ky2
1984; Black et al. 1993) described in this section is P0 (ky , Ω0 ; h)=exp −i +iΩ0 ln Ae Pn (ky , Ω0 ; h).
Ae Ω0
computationally intensive. Specifically, for each out- (5 − 19)
put frequency ω0 , one has to apply the phase-shift
exp(−iω0 tn A), scale by (2A2 − 1)/A3 , and sum the re- Note that the relationship of input Pn (ky , Ω0 ; h) to out-
sulting output over input time tn as described by equa- put P0 (ky , Ω0 ; h) given by equation (5-19) computation-
tion (5-14a). A computationally more efficient DMO ally is much simpler than that of equation (5-14a). The
correction can be formulated in the logarithmic time do- log-stretch domain implementation of DMO correction
main (Bolondi et al., 1982; Bale and Jacubowicz, 1987; involves application of a phase-shift given by the expo-
Notfors and Godfrey, 1987; Liner, 1990; Zhou et al., nential in equation (5-19) to the input data; whereas,
1996). The log-stretch time variable enables lineariza- the frequency-wavenumber implementation involves an
tion of the coordinate transform equation (5-12b), and integral transform given by equation (5-14a).
678 Seismic Data Analysis

FIG. 5.1-10. (a) Selected CMP gathers from the dipping-events model shown in Figure 5.1-8; (b) NMO-corrected gathers
using medium velocity (3500 m/s); (c) NMO-corrected gathers using optimum stacking velocities picked from velocity spectra
along the line, such as that shown in Figure 5.1-9a; (d) DMO-corrected gathers.
Dip-Moveout Correction and Prestack Migration 679

To circumvent the logarithmic computation, a vari- can be formulated as an integration process (Deregowski
ation of the phase-shift term in equation (5-19) is given and Rocca, 1981; Deregowski, 1987; Black et al., 1993).
by Notfors and Godfrey (1987). As in most log-stretch The integral DMO correction is particularly the pre-
formulations of DMO correction, this reference assumes ferred method for data with irregular spatial sampling
that under DMO correction the midpoint variable is and 3-D data with large variations in source-receiver
invariant; hence, by way of equation (5-17b), the first azimuths (Section 7.2).
term in the exponential of equation (5-19) drops out. A The traveltime trajectory associated with the dip-
further approximation, ln Ae = Ae − 1, and use of the
moveout correction operator is given by the following
definition for Ae given by equation (5-18) then lead to
equation (Section E.3):
the following expression for DMO correction:

h2 ky2 y02 τ2
+ 20 = 1. (5 − 21)
P0 (ky , Ω0 ; h)=exp iΩ0 1 + 2 − 1 Pn (ky , Ω0 ; h). h 2 tn
Ω0
(5 − 20) Equation (5-21) describes an ellipse with the following
We now outline the steps in dip-moveout correction properties (Figure 5.1-12):
in the log-stretch domain:

(a) Start with prestack data in midpoint-offset y − h (a) Semi-major axis in midpoint y0 direction: a = h.
coordinates, P (y, h, t) and apply normal moveout (b) Semi-minor axis in time τ0 direction: b = tn .
correction using a dip-independent velocity v.
(b) Sort the data from moveout-corrected CMP The ellipse of equation (5-21) in the y0 − τ0 plane de-
gathers Pn (yn , h, tn ) to common-offset sections
scribes the impulse response of a dip-moveout operator
Pn (yn , tn ; h).
applied to nonzero-offset data with offset 2h. In Figure
(c) Apply the logarithmic stretch in the time direc-
tion based on equation (5-15b) so as to map each 5.1-12, the coordinate of midpoint M is yn and y0 coor-
common-offset section Pn (yn , tn ; h) in yn − tn co- dinate is referenced to yn . Note also that the maximum
ordinates to Pn (yn , Tn ; h) in yn − Tn ccordinates. lateral extent of the ellipse — the aperture of the DMO
(d) Perform 2-D Fourier transform of each common- operator, is equal to offset 2h. Figures 5.1-6b and 5.1-
offset section in the log-stretch domain. 11 show the DMO ellipses associated with the impulse
(e) Apply the phase-shift given by the exponential responses of the frequency-wavenumber and log-stretch
in equation (5-20) to each common-offset sec- DMO operators.
tion Pn (ky , Ω0 ; h), and obtain the dip-moveout- Analogous to the semicircle superposition tech-
corrected data P0 (ky , Ω0 ; h) in the log-stretch nique for migration (Section 4.1), DMO correction can
Fourier transform domain. be viewed as mapping of an amplitude A0 at time tn
(f) Perform 2-D inverse Fourier transform to obtain on a normal-moveout-corrected trace at midpoint yn to
the dip-moveout corrected common-offset section an amplitude A1 at time τ0 on a trace at midpoint y0 .
P0 (y0 , T0 ; h) in the log-stretch domain. The vertical excursion ∆tDM O and the horizontal ex-
(g) Undo the logarithmic stretch as in step (c) in the cursion ∆yDM O denoted in Figure 5.1-12 are given by
time direction based on equation (5-16a) so as to
equations (5-9a,b). If trace spacing in the midpoint di-
obtain the dip-moveout-corrected data P (y0 , τ0 ; h).
rection is ∆y, then the lateral excursion is ∆yDM O /∆y
Figure 5.1-11 shows the impulse responses of a log- traces, the maximum excursion being h/∆y traces.
stretch DMO operator based on equation (5-20) for While the kinematics of the DMO correction oper-
1000-m, 2000-m and 3000-m offsets. The impulse re- ator is given by equation (5-21), the question remains
sponses greatly resemble those of the frequency-wave- as to the amplitude and the phase of the operator. Al-
number DMO correction described earlier (Figure 5.1- though the method was first described by Deregowski
6b). Field data examples of DMO correction presented and Rocca (1981), a formal derivation of integral DMO
in this chapter mostly have been created using a log- correction with amplitude preserving characteristics is
stretch algorithm. given by Black et al. (1993). Mathematical treatise of
the problem is quite involved, and we shall only refer to
the results of the analysis.
Integral DMO Correction Rewrite equation (5-21) explicitly in terms of the
normal-moveout-corrected time variable tn as
In Section 4.1, we reviewed the migration process based
on Kirchhoff summation. Dip-moveout correction also tn = ατ0 , (5 − 22)
680 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 681

FIG. 5.1-12. The DMO ellipse. See text for details.


where
1 correction prior to DMO correction, the amplitude scal-
α= , (5 − 23)
y02 ing (2α2 − 1)/h described here preserves relative ampli-
1− tudes. Other DMO amplitude scaling strategies include
h2
those suggested by Sorin and Ronen (1989), and Gard-
and y0 ≤ h. Given the output sample time τ0 on
ner and Forel (1990).
the DMO ellipse, equation (5-22) gives the input sam-
In practical implementations of the integral DMO
ple time tn (Figure 5.1-12). The output sample value
correction, a user-defined aperture commonly is im-
Pout (y0 , τ0 ; h) is computed by summing over the in-
put sample values Pin (yn , tn ; h) over the DMO operator posed on the operator to avoid aliasing along the steep
aperture flanks of the DMO ellipse, especially at late times.
When the DMO ellipse is truncated before it reaches its
∆y 2α2 − 1 √ fullest lateral extent, amplitude distribution along the
Pout = τ0 ρ(tn ) ∗ Pin , (5 − 24)
2π y
h elliptic trajectory is adjusted accordingly, such that the
amplitude is tapered to zero at the truncation point on
where ∆y is the trace spacing in midpoint direction.
the ellipse. Figure 5.1-13 shows the impulse responses of
Equation (5-23) is adapted from Black et al. (1993)
an integral DMO operator based on equation (5-24) for
and is similar to equation (4-5) which describes the
1000-m, 2000-m, and 3000-m offsets. Compare with the
Kirchhoff summation. For the 2-D application of DMO
impulse responses in Figure 5.1-11 and note the trunca-
correction, the ρ(tn ) filter has an amplitude spectrum
√ tion at steep flanks of the DMO ellipses. The problem of
of the form ωn , with ωn being the temporal frequency
associated with the input time variable tn , and a phase spatial aliasing due to undersampling and the adverse
spectrum equal to π/4. effect of irregular sampling on DMO correction are par-
In the integral implementation of DMO correc- ticularly relevant for 3-D data; as such, these issues will
tion by Deregowski and Rocca (1981), and Deregowski be dealt with in Section 7.2.
(1987), the term α is set to unity in equation (5-24).
Moreover, in the Liner (1990) and Bleistein (1990)
implementation of integral DMO correction, the term Velocity Errors
2α2 − 1 in equation (5-24) is replaced with α2 (2α2 − 1).
Nevertheless, within the context of a conventional pro- What happens if the NMO correction that precedes the
cessing sequence which includes geometric spreading DMO correction were applied with the wrong velocity?
682 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 683

FIG. 5.1-14. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-3. See text for details.
684 Seismic Data Analysis

FIG. 5.1-15. (a) Zero-offset section associated with the depth model in Figure 5.1-3; (b) stack derived from the CMP gathers
in Figure 5.1-14a; (c) DMO stack derived from the CMP gathers in Figure 5.1-14d.

The DMO process requires an input that is NMO cor- second NMO correction (Figure 5.1-14f). Significant im-
rected using the medium velocity (equation 5-2). Thus, provement is seen when these gathers are stacked (Fig-
we try to pick a vertically varying velocity function ure 5.1-16c). For a fair comparison, refer to the con-
from the flattest part of the section for NMO correcting ventional stack with the 3000 m/s repicked velocity in
the data. The optimum stacking velocities are not used Figure 5.1-16b. Although not shown, similar conclusions
because they depend on dip. However, it is the stack- are reached from tests using velocities that are too low
ing velocities that are picked from conventional velocity for NMO correction before DMO processing.
analyses. There is always the possibility that an accu-
rate dip-independent velocity function will not be deter-
mined for NMO correcting the input data before DMO Variable Velocity
correction. The constant-velocity model in Figure 5.1-3
is used to examine this problem. The preceding discussion was based on a constant-
Assume that the velocity used for NMO correction velocity assumption. To be of practical use, DMO must
is 20 percent higher than the velocity that should be be applicable to data with velocity gradients. Figure
used — the 3000-m/s medium velocity. Start with the 5.1-17 shows the depth model we will use to investi-
CMP gathers in Figure 5.1-5b and apply the NMO cor- gate DMO correction in the case of vertical velocity
rection using the incorrect velocity (3600 m/s). The re- variation. The depth model consists of three point scat-
sults are shown in Figure 5.1-14a. Note the undercor- terers buried beneath the center midpoint (CMP 32)
rection at some gathers due to the high velocity used. in a medium with a horizontally layered velocity-depth
Follow the DMO processing sequence described earlier. model. Mathematical aspects of the variable-velocity
Note that events no longer are aligned after the first DMO theory is quite involved and we shall only refer to
NMO and DMO corrections (Figure 5.1-14d). There- the results of the experiments with the synthetic data
fore, the stack obtained from these gathers is not ex- associated with the depth model in Figure 5.1-17.
pected to be any better than the conventional stack Selected common-offset sections and CMP gathers
derived from the gathers in Figure 5.1-14a. The stacked associated with this subsurface model are shown in Fig-
sections are shown in Figure 5.1-15. ures 5.1-18a and 5.1-18b. The same processing sequence
Perhaps the CMP stack can be improved by repick- is followed as that used for the constant-velocity model
ing the velocities after DMO correction. To test this (Figure 5.1-3). The NMO correction (Figure 5.1-18c)
idea, consider the following procedure. First, apply in- before DMO correction is done using the rms velocity
verse NMO correction (Figure 5.1-14e) to the gathers function indicated in Figure 5.1-17. Selected moveout-
with the velocity function that was used in the first corrected common-offset sections are shown in Figure
NMO correction step (Figure 5.1-14a). Then, assum- 5.1-19a. We shall apply both constant-velocity DMO
ing we pick the correct velocity function, use it on the correction and variable-velocity DMO correction (Hale
Dip-Moveout Correction and Prestack Migration 685

FIG. 5.1-16. (a) Zero-offset section associated with the depth model in Figure 5.1-3; (b) stack derived from the CMP gathers
in Figure 5.1-5c; (c) DMO stack derived from the CMP gathers in Figure 5.1-14f.

and Artley, 1992; Artley and Hale, 1994) to these data. Turning-Wave Migration
The corresponding impulse responses are shown in Fig-
ures 5.1-19b,c. Consider the common case of velocities In an extensional basin, such as the Gulf of Mexico,
increasing with depth in practice. As noted earlier, the salt tectonism in its advanced stage can cause the for-
higher the velocity the less the action of the DMO op- mation of salt diapirs with overhang structures. Con-
erator. Note that at late times the lateral extent of the sider such a salt structure as sketched in Figure 5.1-23.
impulse response of the variable-velocity DMO operator If velocities in the surrounding sedimentary sequence
is less than that of the constant-velocity DMO operator. increase rapidly in depth, downward traveling waves
This is equivalent to modifying the offset value for the change their direction at some depth and travel up-
common-offset section under consideration — making ward. When these turning waves, in their upward travel
it smaller than it is so as to decrease the action of the path, encounter a salt overhang, they reflect from the
DMO operator. underside of the structure and follow a downgoing path
Results of constant-velocity and variable-velocity before turning back upward to the surface. The turning-
DMO corrections are shown in Figures 5.1-20 and 5.1- wave reflection then gives rise to a dipping event on the
21, respectively. Events on the selected CMP gathers are stacked section which conflict with the reflections asso-
better flattened with the variable-velocity DMO correc- ciated with the surrounding gently dipping strata. The
tion. The corresponding stacked sections accompanied concept of imaging turning waves was introduced by
with the zero-offset conventional stacked sections with- Claerbout (1985) and first demonstrated on field data
out DMO correction are shown in Figure 5.1-22. The
improvement with the depth-variable velocity also is
evident on the stacked section (Figure 5.1-22d). Specif-
ically, note that the flanks of the diffraction events are
enhanced with the depth-variable velocity, making it re-
semble much more closely the zero-offset section (Fig-
ure 5.1-22a) as compared to the constant-velocity DMO
stack (Figure 5.1-22c).
In practice, constant-velocity DMO correction af-
ten yields acceptable results so long as the vertical ve-
locity gradient is reasonably small and does not change
rapidly in depth. The constant-velocity DMO correc- FIG. 5.1-17. Depth model of three point scatterers buried
tion also has the bonus effect of attenuating coherent in a vertically varying velocity medium. The asterisks indi-
linear noise as demonstrated in the next section. cate the positions of the point scatterers.
686 Seismic Data Analysis

FIG. 5.1-18. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-17: (a) common-offset sections with offset range from 50 to 1550 m and an increment of 300 m; (b) CMP gathers
sorted from the common-offset sections as in (a) at midpoint locations from 32 to 63 as denoted in Figure 5.1-17 with an
increment of 3; (c) the CMP gathers as in (b) after NMO correction and muting.

by Hale et al. (1992). Many examples of turning-wave described by the hyperbolic traveltime equation:
imaging of salt overhang structures from the Gulf of 4h2
Mexico are given by Ratcliff et al. (1992). t2 = t20 + 2 , (5 − 25a)
vN MO
Dip-moveout correction alone, even if it accounts
for vertical velocity variations, does not preserve the where t is the two-way traveltime associated with a
turning-wave energy on stacked data. This is because source-receiver separation 2h, t0 is the two-way zero-
the normal moveout associated with the turning ray- offset time, and vN M O is the velocity that best flattens
the event after normal-moveout correction.
paths exhibit an abnormal behavior as sketched in Fig-
When, however, the CMP raypaths have an upward
ure 5.1-23. Consider a reflection point E that represents
component associated with turning waves (location A),
all dips on the salt flank. Sketched in Figure 5.1-23 are
the moveout trajectory in the CMP gather exhibits an
the zero-offset raypaths for the reflector dip of less than abnormal behavior which may be described by the trav-
90 degrees (raypath that emerges at location C), ex- eltime equation (Hale et al., 1992)
actly 90 degrees (raypath that emerges at location B),
and greater than 90 degrees (raypath that emerges at 4h2
t2 = t20 − 2 . (5 − 25b)
location A). vN MO
Note that when the CMP raypaths do not have Note that the abnormal moveout equation (5-25b) dif-
an upward component (location C), the moveout tra- fers from the normal moveout equation (5-25a) by a
jectory in the CMP gather exhibits the usual behavior change in the sign of the moveout term 4h2 /vN2
MO.
Dip-Moveout Correction and Prestack Migration 687

FIG. 5.1-19. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-17: (a) common-offset sections with offset range from 50 to 1550 m and an increment of 300 m sorted from the
NMO-corrected gathers as in Figure 5.1-4c; (b) impulse responses of the constant-velocity DMO operators applied to the
common-offset gathers; (c) impulse responses of the variable-velocity DMO operators applied to the common-offset gathers.

FIG. 5.1-20. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-17: (a) common-offset sections as in Figure 5.1-19a after the constant-velocity DMO corection; (b) CMP gathers
sorted from the common-offset sections as in (a) at midpoint locations from 32 to 63 as denoted in Figure 5.1-17 with an
increment of 3.
688 Seismic Data Analysis

FIG. 5.1-21. Intermediate results from DMO processing the nonzero-offset synthetic data derived from the depth model in
Figure 5.1-17: (a) common-offset sections as in Figure 5.1-19a after the variable-velocity DMO corection; (b) CMP gathers
sorted from the common-offset sections as in (a) at midpoint locations from 32 to 63 as denoted in Figure 5.1-17 with an
increment of 3.

FIG. 5.1-22. (a) Zero-offset section associated with the depth model in Figure 5.1-17; (b) stack derived from the CMP
gathers as in Figure 5.1-18c; (c) DMO stack derived from the CMP gathers as in Figure 5.1-20b; (d) DMO stack derived from
the CMP gathers as in Figure 5.1-21b.

There can also exist a circumstance where the reflection sociated with the turning waves during stacking,
off the salt flank does not exhibit any moveout (location and
B in Figure 5.1-23). (b) migrate the stacked data using an algorithm that
To make use of the turning-wave energy in imaging can handle dips beyond 90 degrees.
salt overhang structures, we must
To derive a stacked section that preserves reflec-
(a) preserve the reflections with abnormal moveout as- tions associated with both normal (nonturning) and
Dip-Moveout Correction and Prestack Migration 689

FIG. 5.1-23. Geometry of turning rays in a medium with velocities increasing in depth, and the CMP traveltime trajectories
depicted above at three locations, A, B, and C. At location A, the turning raypaths such as ADE causes an abnormal
moveout. At location C, the raypaths yield a normal moveout, and at location B no moveout is observed on the CMP gather.
(Adapted from Hale et al., 1992).

turning waves, consider the stacking process in two energy is preserved during conventional stacking based
2
parts. First, use equation (5-25a) with positive vN M O to on normal moveout.
handle events with normal moveout and derive a stack Note from Figure 5.1-23 that turning waves propa-
associated with the normal waves. Second, use equation gate at angles greater than 90 degrees. This means that
2
(5-25a) with negative vN M O to handle events with ab- turning waves are evanescent waves and need to be im-
normal moveout and derive a stack associated with the aged using a migration algorithm that can handle dips
turning waves. Finally, add the two sections to obtain beyond 90 degrees. Reverse-time migration (Baysal et
the composite-wave stack. al., 1984) described in Section 4.1 is based on such an
Reflections off the flank of a salt dome, whether algorithm. The field data example shown in Figure 4.3-
they are associated with normal waves or turning waves, 22 suggests that the salt dome at its root at about 3 s
and reflections associated with the surrounding sedi- has a gentle overhang. This character is not identifiable
mentary sequence constitute dipping events with differ- in the image obtained by an algorithm that only han-
ent moveout velocities on a stacked section. As such, dles waves propagating at less than 90 degrees (Figure
DMO correction needs to be applied to data prior to 4.4-24).
stacking both the normal and turning waves. The DMO Kirchhoff migration (Ratcliff et al., 1992) and
operator preferably should account for the strong ver- the phase-shift method (Claerbout, 1985) also can be
tical velocity variations that give rise to turning waves adapted to image turning waves. To develop the con-
(Hale and Artley, 1992; Artley and Hale, 1994). ceptual basis of turning-wave migration, consider the
Figure 5.1-24 (left column) shows a normal-wave raypath segments DA and ED as denoted in Figure
stack, a turning-wave stack, and the composite-wave 5.1-23. The raypath segment DA is associated with the
stack. Note that the normal-wave stack actually con- upcoming wave energy contained in the normal-wave
tains a significant portion of the turning wave energy stack and the raypath segment ED is associated with
represented by the steeply dipping event in the turning- the downgoing wave energy contained in the turning-
wave stack. This is because the turning-wave energy is wave stack (Figure 5.1-24).
predominantly of low frequency. Therefore, despite its Just as the stacking process was treated in two
abnormal moveout behavior, much of the turning-wave parts, the imaging process also can be treated in
690 Seismic Data Analysis

FIG. 5.1-24. Left column: Normal-wave stack (top), turning-wave stack (middle) and composite-wave stack (bottom) created
by adding the normal-wave and turning-wave stacks. Right column: Migrations of the sections on the left column. (Hale et
al., 1992).
Dip-Moveout Correction and Prestack Migration 691

two parts. First, perform phase-shift migration of the combined (Hale et al., 1992). First, substitute equation
normal-wave stack; this involves downward extrapola- (5-28a) into equation (5-28b) to obtain
tion of the upcoming wave energy along the raypath CE
P (ky , z = zE , ω) = P (ky , z = 0, ω)
from the surface z = 0 to a depth zE where the wave
originates (location E in Figure 5.1-23). The equation × exp[−ikz zD − ikz (zD − zE )] .
for the wave extrapolation is (Section D.1): (5 − 29a)
P (ky , z = zE , ω) = P (ky , z = 0, ω) exp(−ikz zE ), Then, combine equations (5-26) and (5-29a) to ob-
tain the exptrapolation equation for the composite-wave
(5 − 26)
stack:
where ky , kz , and ω are the Fourier transform variables
P (ky , z = zE , ω) = P (ky , z = 0, ω)
associated with the coordinate variables for midpoint y,
depth z, and two-way zero-offset traveltime t, respec- × exp(−ikz zE )+exp[−ikz (2zD − zE )] .
tively, and P (ky , z = 0, ω) is the 2-D Fourier transform (5 − 29b)
of the upcoming wavefield at the surface z = 0 repre-
Based on the concepts described above, we now
sented by the stacked data P (y, z = 0, t). The vertical outline the steps involved in a turning-wave migration
wavenumber kz is defined in terms of the horizontal algorithm that makes use of the phase-shift method
wavenumber ky and frequency ω by the dispersion re- (Claerbout, 1985; Hale et al., 1992). First, consider
lation (Section D.1): imaging the normal waves. The process involves down-
2 v 2 ky2 ward continuation of the upcoming waves from the sur-
kz = ω2 − , (5 − 27) face z = 0 to the maximum specified depth zmax at
v 4
discrete depth steps ∆z.
where v is the medium velocity. In equation (5-27), the
region of propagation at angles less than 90 degrees (a) Start with the composite-wave stack — an approx-
is associated with ω > (v/2)|ky |, and the region of imation to the zero-offset section P (y, z = 0, t) and
evanescense at angles greater than 90 degrees is associ- apply 2-D Fourier transform to get the transformed
ated with ω < (v/2)|ky |. wavefield P (ky , z = 0, ω).
Next, perform phase-shift migration of the turning- (b) For each frequency ω > (v/2)|ky | and ω ≤ (v/2)|ky |
wave stack; this involves downward extrapolation of the in the transform domain, extrapolate the wavefield
upcoming wave energy along the raypath AD from the P (ky , z, ω) at depth z with a phase-shift operator
surface z = 0 to a depth zD where the wave turns (lo- exp (−ikz ∆z) to get the wavefield P (ky , z + ∆z, ω)
cation D in Figure 5.1-23). The equation for the wave at depth z+∆z. At each depth, a new extrapolation
extrapolation is operator with the velocity v(z) defined for that z
P (ky , z = zD , ω) = P (ky , z = 0, ω) exp(−ikz zD ), value is computed.
(c) Split the wavefield P (ky , z + ∆z, ω) into its propa-
(5 − 28a) gating Pu (ky , z + ∆z, ω) and evanescent Pd (ky , z +
Since the journey of the turning waves does not end ∆z, ω) components, corresponding to the normal
at the turning point D, we must continue with the wave and turning waves, respectively.
extrapolation until we reach the point where the waves (d) Save the wavefield Pd (ky , z + ∆z, ω) for use later in
originate on the salt flank (location E in Figure 5.1-23). turning-wave imaging.
Hence, perform upward extrapolation of the downgoing (e) As for any other migration algorithm, invoke
wave energy along the raypath DE from the depth level the imaging principle t = 0 upon the wavefield
zD to location E on the salt flank: Pu (ky , z + ∆z, ω) at each extrapolation step to ob-
tain the migrated section from the normal waves
P (ky , z = zE , ω)=P (ky , z = zD , ω)exp[−ikz (zD − zE )], Pu (ky , z, t = 0) in the transform domain. The
(5 − 28b) imaging condition t = 0 is met by summing over
all frequency components of the extrapolated wave-
The wave extrapolations described by equations (5- field at each depth step (equation D-84).
28a,b) are performed in the transform domain only us- (f) Repeat steps (b) through (d) for all depth steps
ing the evanescent energy that corresponds to the region down to a specified z = zm ax to obtain the normal-
ω < (v/2)|ky |. wave image in the transform domain Pu (ky , z, t =
Just as the stacking of normal waves and turning 0).
waves is combined to obtain the composite-wave stack (g) Apply inverse Fourier transform in the midpoint y
(Figure 5.1-24), the migration processes based on the direction to obtain the image Pu (y, z, t = 0) from
extrapolation equations (5-26) and (5-28a,b) can also be the normal waves.
692 Seismic Data Analysis

Now consider imaging the turning waves. The pro-


cess involves upward continuation of the downgoing
waves from the maximum specified depth zmax to the
surface z = 0 at discrete depth steps ∆z. At each
depth, the downgoing waves are updated by adding the
evansecent wave Pd (ky , z, ω) saved in step (d) during the
downward extrapolation of the composite-wave stack.
We shall assume that the downgoing wave at z = zmax
is null.

(a) Add the evanescent wave saved at depth z to


the downgoing wave at the same depth, and for
each frequency ω ≤ (v/2)|ky | in the evansecent
region of the transform domain, extrapolate the
new downgoing wave with a phase-shift operator
exp (−ikz ∆z) to get the wavefield Pd (ky , z −∆z, ω)
at depth z − ∆z.
(b) Invoke the imaging principle t = 0 upon the wave-
field Pd (ky , z −∆z, ω) at each extrapolation step to
obtain the migrated section from the turning waves
Pd (ky , z, t = 0) in the transform domain.
(d) Repeat steps (a) and (b) for all depth steps to ob- FIG. 5.2-1. DMO processing flowchart.
tain the turning-wave image in the transform do-
main Pd (ky , z, t = 0).
(e) Apply inverse Fourier transform in the midpoint y (b) Apply NMO correction using the these flat-event
direction to obtain the image Pd (y, z, t = 0) from velocities.
the turning waves. (c) Sort data to common-offset sections, apply DMO
(e) Add the normal-wave image and turning-wave im- correction and sort back to CMP gathers.
age to obtain the composite-wave image. (d) Apply inverse NMO correction using the flat-event
velocities from step (a).
As described above, it is important to emphasize (e) Perform velocity analysis at frequent intervals as
that the turning-wave energy can be preserved only by needed to derive an optimum stacking velocity
making use of its abnormal moveout behavior during field.
stacking and imaged properly by making use of the (f) Apply NMO correction using the optimum satcking
evanescent energy during migration. velocity field.
Figure 5.1-24 (right column) shows migration of a (g) Stack the data and migrate using an edited and
normal-wave stack using the phase-shift equation (5- appropriately smoothed version of the optimum
28), migration of a turning-wave stack using the phase- satcking velocity field.
shift equation (5-31a), and migration of the composite-
wave stack using the phase-shift equation (5-31b). Note Note that this processing sequence is similar to
the distinctively defined salt boundary obtained from the sequence for residual statics corrections described
imaging the compsite-wave stack. in Figure 3.3-12. Both residual statics and DMO cor-
rections are followed by a revision of velocities so as to
get the most out of these corrections during stacking. In
5.2 DIP-MOVEOUT CORRECTION this section, we shall apply the sequence outlined above
IN PRACTICE to two common cases of conflicting dips with different
stacking velocities — salt flanks and fault planes.
Results of the previous section suggest the general DMO
processing sequence shown in Figure 5.2-1.
Salt Flanks
(a) Perform velocity analysis at sparse intervals and
pick just a few velocity functions with minimal dip Figure 5.2-2 shows selected CMP gathers along a ma-
effects. rine line over a salt structure. CMP gathers 1381 (at 1.5
Dip-Moveout Correction and Prestack Migration 693

s), 1461 (at 2.2 s), and 1701 (at 1.55 s) exhibit cases of from the gather with no DMO correction. The dis-
conflicting dips associated with two events with signifi- tinctive trend is a direct result of the fact that
cantly different moveouts. DMO correction removes reflection-point smearing
The DMO processing sequence includes the follow- by mapping reflection points on a dipping reflector
ing steps. associated with nonzero source-receiver separation
onto normal-incidence reflection point. The partial
(a) Perform velocity analysis sparsely along the line at migration effect of DMO correction has actually
locations with prominently flat events and create moved the dipping event with the 2750-m/s peak
an initial velocity field. to a different midpoint location.
(b) Apply normal-moveout correction using flat-event (f) Create a velocity field using the velocity functions
velocities. Note the events associated with steep picked from the velocity spectra computed from
dips have been overcorrected as demonstrated in the DMO-corrected gathers (Figure 5.2-9b). This
Figure 5.2-3, whereas reflections with no dip or neg- velocity field has more detail than the initial field
ligibly small dip have been flattened. (Figure 5.2-9a) used for NMO correction prior to
(c) Apply partial stacking to CMP gathers to reduce DMO correction (step b). This initial velocity field
the fold from 60 to 30. A partial stack up to a also was used to apply the inverse NMO correction
4:1 reduction in fold usually is acceptable before as in step (d).
DMO correction. While fold reduction provides (g) Apply moveout correction to the DMO-corrected
significant computational savings, it must not be gathers using the velocity field as in Figure 5.2-
done excessively. Following the fold reduction, sort 9b. Selected CMP gathers are shown in Figure 5.2-
the moveout-corrected gathers (Figure 5.2-3) to 10, and the corresponding CMP stack is shown
common-offset sections and perform dip-moveout in Figure 5.2-11. As a result of DMO correction,
correction. Then, sort back to CMP gathers. Com- the steeply dipping salt-flank reflections have been
stacked with as much power as the flat events asso-
pare the selected gathers after DMO correction
ciated with the surrounding strata. Since a DMO
(Figure 5.2-4) with the same gathers without DMO
stack is a closer approximation to a zero-offset sec-
correction (Figure 5.2-3), and note that the dual-
tion in comparison with a CMP stack, time mi-
ity in event moveout on CMP gathers 1381 (at 1.5
gration of the DMO stack yields an image of the
s), 1461 (at 2.2 s), and 1701 (at 1.55 s) has been
salt diapirs with their flanks clearly delineated, es-
removed. This is a direct result of the partial mi-
pecially between 1-1.5 s (Figure 5.2-12). For com-
gration effect of the DMO correction.
parison, conventional CMP stack and its migra-
(d) Apply inverse moveout correction (Figure 5.2-5)
tion are shown in Figures 5.2-13 and 5.2-14, re-
with the same velocity field that was used for the
spectively. Because the reflections off the flanks of
NMO correction prior to DMO correction (Figure
the salt diapirs have not been preserved with ade-
5.2-3).
quate strength on the CMP stack (Figure 5.2-13),
(e) Perform velocity analysis at frequent intervals
time migration yields a poor definition of the salt
along the line and pick velocity functions which
boundaries (Figure 5.2-14).
now are supposed to have been corrected for the
dip effect. Refer to the velocity analysis at midpoint
1381 shown in Figure 5.2-6. Note the improved ve-
locity trend after DMO correction. Close-up dis- Fault Planes
plays of the semblance spectrum and the moveout-
corrected gather before and after DMO correction Figure 5.2-15 shows selected CMP gathers along a ma-
are shown in Figure 5.2-7 and 5.2-8, respectively. rine line over a structure with fault blocks. CMP gather
Note the two semblance peaks at 1.5 s — one at at midpoint location 1688 (at 2.5 s) exhibits a clear case
2050 m/s and the other at 2750 m/s. Because the of conflicting dips associated with two events with sig-
gather was moveout corrected using the denoted nificantly different moveouts.
velocity function that includes the 2050-m/s peak, The DMO processing sequence includes the follow-
the event associated with the 2750-m/s peak has ing steps.
been overcorrected (Figure 5.2-7). DMO correction
has removed the duality in the velocity spectrum (a) Perform velocity analysis sparsely along the line at
at about 1.5 s and yielded a more distinctive trend locations with prominently flat events and create
(Figure 5.2-8) compared to the spectrum derived an initial velocity field.
(text continues on p. 705)
694 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 695
696 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 697
698 Seismic Data Analysis

FIG. 5.2-6. Velocity spectrum computed from a CMP gather (a) without, and (b) with DMO correction, as in Figures 5.2-2
and 5.2-5, respectively.

FIG. 5.2-7. A close-up view of the velocity spectrum shown in Figure 5.2-6a and the associated CMP gather without DMO
correction followed by moveout correction using the velocity function depicted in (a).
Dip-Moveout Correction and Prestack Migration 699

FIG. 5.2-8. A close-up view of the velocity spectrum shown in Figure 5.2-6b and the associated CMP gather with DMO
correction followed by moveout correction using the velocity function depicted in (a).

FIG. 5.2-9. The velocity field used to stack the CMP gathers (a) as in Figure 5.2-3 without DMO correction, and (b) as in
Figure 5.2-5 with DMO correction.
700 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 701

FIG. 5.2-11. DMO stack derived from CMP gathers as in Figure 5.2-10. Compare with Figure 5.2-13.
702 Seismic Data Analysis

FIG. 5.2-12. Migration of the DMO stack shown in Figure 5.2-11. Compare with Figure 5.2-14.
Dip-Moveout Correction and Prestack Migration 703

FIG. 5.2-13. CMP stack derived from CMP gathers as in Figure 5.2-3. Compare with Figure 5.2-11.
704 Seismic Data Analysis

FIG. 5.2-14. Migration of the CMP stack shown in Figure 5.2-13. Compare with Figure 5.2-12.
Dip-Moveout Correction and Prestack Migration 705

(b) Apply normal-moveout correction using flat-event image which shows clearly delineated fault blocks
velocities. Note the event at 2.5 s on CMP gather in the vicinity of CMP 1688 (Figure 5.2-22). For
1688 associated with steep fault-plane reflections comparison, conventional CMP stack and its mi-
has been overcorrected as demonstrated in Figure gration are shown in Figures 5.2-23 and 5.2-24, re-
5.2-16, whereas reflections with no dip or negligibly spectively. Because the fault-plane reflections have
small dip have been flattened. not been preserved with adequate strength on the
(c) Apply partial stacking to CMP gathers to reduce CMP stack (Figure 5.2-23), time migration yields
the fold from 60 to 30, and sort the moveout- a blurred image of the fault blocks (Figure 5.2-24).
corrected gathers (Figure 5.2-16) to common-
offset sections and perform dip-moveout correction.
Then, sort back to CMP gathers and compare the
DMO and Multiples
selected gathers after DMO correction (Figure 5.2-
17) with the same gathers without DMO correction
(Figure 5.2-16). Note that the overcorrected event Multiples can be enhanced or attenuated by DMO cor-
at 2.5 s on CMP gather 1688 has been removed. rection. Consider the situation depicted in the velocity
Again, this is a direct result of the partial migra- spectrum shown in Figure 5.2-25. Suppose you have a
tion effect of the DMO correction. flat primary and a flat multiple. After DMO correction,
(d) Apply inverse moveout correction (Figure 5.2-18) the velocity contrast ∆v between the primary and the
with the same velocity field that was used for the multiple would not change; therefore, DMO correction
NMO correction prior to DMO correction (Figure does not have any impact on the success or failure of a
5.2-16). multiple attenuation technique in such a case. A simi-
(e) Perform velocity analysis at frequent intervals lar situation exists for a dipping primary and a dipping
along the line and pick velocity functions which multiple. If you have a dipping primary and a flat mul-
now are supposed to have been corrected for the tiple — a case where the multiple associated with a flat
dip effect. Refer to the velocity analysis at mid- water-bottom interferes with a deeper dipping primary,
point 1688 shown in Figure 5.2-19. Refer to the then, after DMO correction the flat multiple will not be
velocity spectrum (Figure 5.2-19b) associated with affected. However, the dipping primary will shift to the
the gather without DMO correction (Figure 5.2- left on the velocity spectrum, causing a decrease in the
19a) and note the two semblance peaks at 2.5 s — velocity contrast between the two events. This suggests
one at 2500 m/s and the other at 2750 m/s. The that multiple attenuation based on velocity discrimi-
gather was moveout corrected using the denoted nation between primaries and multiples would be less
velocity function that includes the 2500-m/s peak, effective on DMO-corrected gathers. Finally, the situa-
rather than the 2750-m/s peak. DMO correction tion would favor multiple attenuation when you have a
has partially migrated the steeply dipping event flat primary conflicting with a dipping multiple.
to another midpoint location, and as a direct con- A field data example is shown in Figure 5.2-26.
sequence, has removed the duality in the velocity Note the conflicting primary and multiple events below
spectrum at 2.5 s and yielded a more distinctive midpoint 1716 at 1 s on the section without DMO cor-
trend (Figure 5.2-19d) compared to the spectrum rection (Figure 5.2-26a). In this case, following DMO
derived from the gather with no DMO correction correction, the velocity contrast between the primary
(Figure 5.2-19b). and multiple event has increased, and thus has led to
(f) Create a velocity field using the velocity functions attenuation of the latter during stacking of the DMO-
picked from the velocity spectra computed from the corrected gathers (Figure 5.2-26b).
DMO-corrected gathers. The practical question as to whether a multiple at-
(g) Apply moveout correction to DMO-corrected gath- tenuation technique should be applied before or after
ers using this velocity field. Selected CMP gathers DMO correction is an important one. This question is
are shown in Figure 5.2-20 and the corresponding relevant only for multiple attenuation techniques based
CMP stack is shown in Figure 5.2-21. As a re- on velocity discrimination (Chapter 6). Often, for rea-
sult of DMO correction, the steeply dipping fault- sons of efficiency, the multiple attenuation step precedes
plane reflections have been preserved during stack- DMO correction. This is especially the case in process-
ing. Since a DMO stack is a closer approximation ing 3-D data — following 3-D DMO correction, data
to a zero-offset section in comparison with a CMP often are stacked concurrently without creating DMO-
stack, time migration of the DMO stack yields an corrected gathers.

(text continues on p. 716)


706 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 707
708 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 709
710 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 711
712 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 713
714 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 715
716 Seismic Data Analysis

While the diffractions associated with the steep flanks


of the salt diapir are better preserved by DMO correc-
tion, the steep coherent linear noise is attenuated by
this process.

Other Considerations

Dip-moveout correction is most effective at shallow


times where velocities usually are low. Figure 5.2-29
shows a shallow portion of a CMP stack with and
without DMO correction. Note that DMO correction
has preserved the diffractions associated with the fault
blocks and the fault-plane reflections. As a result, mi-
gration then has better imaged the subtle faults in the
subsurface.
Albeit dip-moveout correction becomes relatively
less significant at late times — just the opposite situa-
tion with migration, it can still produce a better stack
than conventional CMP stacking. Figure 5.2-30 shows
a moderately deep portion of a CMP stack with and
without DMO correction. DMO correction has visibly
FIG. 5.2-25. A velocity spectrum which indicates a pri- enhanced the deep reflections and diffractions. Migrated
mary event (P) and a multiple event (M) arriving at about sections (Figure 5.2-31) exhibit comparable imaging of
the same zero-offset time. The multiple event is associated the dipping events and the unconformity which en-
with a shallow primary. Depending on the flat or dipping velopes these events from below. On the other hand, the
character of the primary and multiple reflection, the velocity crispness of the image from DMO stack (Figure 5.2-31b)
contrast ∆v between the two changes after DMO correction also is noticeable.
(see text for details). The response of DMO correction to random noise
is examined in Figure 5.2-32. Typically, random noise
is more prominent at late times. Since DMO correction
becomes increasingly less effective at late times, it may
DMO and Coherent Linear Noise be concluded that it has minimal impact on random
noise. On the other hand, random noise at shallow times
The constant-velocity assumption underlying the most may seemingly be attenuated by DMO correction. This
commonly used DMO algorithms can sometimes atten- action of the DMO process can be attributed to the
uate dipping events. This is the case with a shallow fact that it is a migration process involving spreading
dipping event and a deep flat event (Black et al., 1985). of amplitudes along elliptical trajectories. Nevertheless,
If velocity increases with depth (the usual case), then DMO correction should not be viewed and used as a
these two events can arrive about the same time and process to attenuate noise.
have similar moveouts (Figure 5.2-27). In terms of ve- DMO correction becomes insignificant in a medium
locities, this implies that the moveout (or stacking) ve- with high velocities. Figure 5.2-33 shows a CMP stack
locity v1 / cos θ for the shallow, dipping reflector is ap- with and without DMO correction. Velocities vary from
proximately equal to the moveout velocity v2 associ- 4000 m/s at the surface to 6000 m/s at the bottom
ated with the deep, flat reflector. Following the DMO of the section. Note that, diffractions and nearly flat
correction, nothing happens to the flat event, while the reflections with conflicting dips appear to stack equally,
dipping event shifts to a lower velocity value v1 . The with or without DMO correction. A way to distinguish
dipping event, that may be associated with coherent between the two sections is the relative attenuation of
linear noise, drifts away from the velocity function for the shallow coherent linear noise by DMO correction.
flat events and thus is attenuated during stacking. DMO correction must always be considered within
This characteristic response of DMO can be used the framework of time migration. Specifically, DMO cor-
to our advantage in attenuating coherent linear noise rection is not meant to solve the problem of nonhy-
associated with shallow point scatterers in the water perbolic moveout of reflections below complex overbur-
bottom. A field data example is shown in Figure 5.2-28. den structures which often are accompanied with strong
Dip-Moveout Correction and Prestack Migration 717

FIG. 5.2-26. A portion of a CMP stack, (a) without, and (b) with DMO correction.
718 Seismic Data Analysis

FIG. 5.2-27. Response of DMO correction to a deep flat reflector (F) and a shallow dipping reflector (D). See text for details.

lateral velocity variations. While it is not expected to (g) Apply residual statics corrections to CMP gathers.
solve this problem, fortuitously, DMO correction may (h) Apply the inverse of step (d) to move the shots and
not harm such events. The base-salt event in Figure receivers from the flat reference datum back to the
5.2-28 exemplifies this observation. Note that, the two floating datum.
diffraction-like segments of the base-salt event below (i) Apply inverse moveout correction using velocities
midpoint 1116 between 1.8-2 s does not appear to be from step (c).
influenced by DMO correction.
(j) Perform velocity analysis and apply moveout cor-
For land data, DMO correction can be applied be-
rection.
fore statics corrections. Specifically, land data process-
ing sequence that includes DMO correction is as follows: (k) Apply datum corrections to move the shots and
receivers from the floating datum to the reference
flat datum as in step (d).
(a) Estimate a model for the near-surface layer using
refracted arrivals. The model parameters include (l) Apply mute and stack the data. The stacked sec-
the shape of the refractor (base of the weathering tion is referenced to the flat datum level specified
layer) and the bedrock velocity. in step (d).
(b) Assume a value for the weathering velocity and use
the near-surface layer to apply the shot and receiver The above sequence enables performing velocity analy-
refraction statics (Section 3.4) so as to replace the sis after statics and DMO corrections so as to obtain an
near-surface layer with the bedrock and move the improved velocity field for the subsequent stacking and
shots and receivers from the topographic surface migration.
to a floating datum, which is a smoothed version Can DMO be used for trace interpolation? Con-
of the surface topography. sider the constant-velocity synthetic data set associated
(c) Perform preliminary velocity analysis and apply
with the earth model shown in Figure 5.1-3. Throw
moveout corrections.
away every other trace on each of the common-offset
(d) Apply datum corrections to move the shots and
receivers from the floating datum to a flat datum sections and simulate a coarser trace spacing (Figure
to which the CMP stack is referenced. 5.2-34b). The selected CMP gathers shown in Fig-
(e) Apply DMO correction. ure 5.2-34a also exhibit the discarded traces replaced
(f) Estimate surface-consistent shot and receiver resid- with zero traces. Apply DMO correction to all of the
ual static shifts using methods described in Section common-offset sections (Figure 5.2-34c) and sort back
3.3. to CMP gathers (Figure 5.2-34d). Note that on the
Dip-Moveout Correction and Prestack Migration 719

FIG. 5.2-28. A portion of a CMP stack, (a) without, and (b) with DMO correction. Note the attenuation of coherent linear
noise by DMO correction.
720 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 721

FIG. 5.2-30. A moderately deep portion of CMP stack, (a) without DMO correction, and (b) with DMO correction. Migrated
sections are shown in Figure 5.2-31.

FIG. 5.2-31. Migrated sections, (a) without DMO correction, and (b) with DMO correction. Stacked sections input to
migration are shown in Figure 5.2-30.
722 Seismic Data Analysis

FIG. 5.2-32. A deep portion of a CMP stack, (a) without DMO correction, and (b) with DMO correction.

common-offset sections, the zero traces are filled in at stack from the gathers with missing traces is quite ap-
large offsets by DMO correction, whereas they are not parent (Figure 5.2-35c).
quite filled in at small offsets. So, the amplitude dis-
tribution after DMO correction will not be uniform
from one common-offset section to another. Note also
the aliased energy on the CMP gathers. Stack the Aspects of DMO Correction — A Summary
DMO-corrected gathers and compare the resulting sec-
tion with the desired zero-offset section (Figure 5.2-35). We discussed the principles of DMO correction and
These data have not been subjected to any gain treat- studied its practical aspects using synthetic and field
ment after stack; hence, the relative amplitudes have data. It now is appropriate to compile aspects of DMO
been preserved. The amplitude imbalance in the DMO correction.
Dip-Moveout Correction and Prestack Migration 723

FIG. 5.2-33. A portion of a CMP stack, (a) without DMO correction, and (b) with DMO correction.
724 Seismic Data Analysis

FIG. 5.2-34. (a) Selected moveout-corrected CMP gathers associated with the earth model of point scatteres depicted in
Figure 5.1-3, with amplitudes on every other trace zeroed out; (b) selected common-offset sections; (c) same common-offset
sections after DMO correction; (d) selected gathers as in (a) after DMO correction.

(a) The process of dip-moveout corrects for the dip ef- spectra. Velocity analysis of DMO-corrected data
fect on stacking velocities. alleviates this problem and increases the accuracy
(b) Thus, it preserves conflicting dips with different of picking an unambiguous velocity function from
stacking velocities during CMP stacking. a velocity spectrum.
(c) The DMO stack, therefore, is a closer represen- (f) Velocities estimated from DMO-corrected data are
tation of a zero-offset section as compared to a dip independent; therefore, they are more suitable
conventional CMP stack based on normal-moveout to derive a migration velocity field as compared to
correction, only. velocities estimated from data without DMO cor-
(d) The DMO stack can then be migrated using a zero- rection.
offset migration algorithm with greater accuracy. (g) Dip-moveout correction actually is a process of
(e) Conflicting dips with different stacking velocities partial migration before stack. Specifically, it
give rise to multivalued velocity picks from velocity maps normal-moveout-corrected data to normal-
Dip-Moveout Correction and Prestack Migration 725

FIG. 5.2-35. (a) Zero-offset section associated with the earth model of point scatteres depicted in Figure 5.1-3; (b) the DMO
stack same as in figure 5.1-7c from the gathers as in Figure 5.1-6d without missing alternate traces; (c) the DMO stack from
the gathers with missing alternate traces as in Figure 5.2-34d.

incidence reflection points in the subsurface. As a amplitude variation with offset as well as to obtain an
result, midpoint coordinate is variant under DMO improved migrated stack.
correction. We now review the kinematics of prestack time
(h) As a direct consequence of aspect (g), DMO correc- migration. The theory of extrapolation of nonzero-
tion removes the reflection point dispersal associ- offset wavefields is provided in Section D.1, while the
ated with nonzero-offset recording in the presence stationary-phase traveltime trajectory associated with
of dipping reflectors. nonzero-offset wave propagation is given in Section E.5.
(i) Following DMO correction, prestack data can be Figure E-2 shows a sketch of a nonzero-offset raypath
migrated so as to create CMP gathers in their mi- in a constant-velocity medium from a source location S
grated position (next section). This then enables us to a reflection point R and to a receiver location G. The
to conduct velocity analysis to derive a migration traveltime equation associated with the raypath SRG
velocity field with greater confidence. is given by (Section E.5)
(j) Finally, the CMP gathers from prestack time mi-
gration of DMO-corrected data can be used for am- vt = z 2 + (y + h)2 + z 2 + (y − h)2 , (5 − 32)
plitude variation with offset analysis. where v is the medium velocity, t is the total traveltime
from S to R to G. The medium is represented by the
In this chapter, we discussed the dip-moveout pro- midpoint y and depth z coordinates.
cess within the context of 2-D seismic data. Three- Equation (5-32) can be put into the following al-
dimensional aspects of DMO correction are discussed ternative form (Section E.5):
in Section 7.2.
y2 z2
+ = 1, (5 − 33)
(vt/2)2 (vt/2)2 − h2

5.3 PRESTACK TIME MIGRATION which represents an ellipse in the y − z plane for a con-
stant t with the following parameters:
As stated early in the chapter, the rigorous solution to
the problem of conflicting dips with different stacking (a) Semi-major axis in midpoint y direction: a = vt/2.
velocities is prestack time migration. The robust alter- (b) Semi-minor axis in depth z direction: b =
native in practice is to apply NMO and DMO correc- (vt/2)2 − h2 . √
tions followed by poststack time migration. The migra- (c) Distance from center to either focus: a2 − b2 = h.
tion step can, however, be moved up to precede CMP (d) Distance from one focus to a point on the ellipse
stacking. Specifically, by migrating each of the NMO- to the other focus: vt.
and DMO-corrected common-offset sections, one has
the opportunity to update the velocity field and gen- From its properties, note that this ellipse in the y −
erate CMP gathers which can be used for analysis of z plane can be associated with a recording geometry
726 Seismic Data Analysis

for a source-receiver pair which is situated at the foci using equation (5-33) or diffraction summation over the
with a separation equal to the offset 2h (Figure E-2). traveltime surface described by equation (5-32). The
The ellipse itself describes the geometry of a reflecting traveltime surface, which is shown in Figure 5.3-1a, is
interface such that the reflections recorded by a source- known as Cheops’ pyramid (Claerbout, 1985). The re-
receiver pair situated at the foci have the same arrival sult of summation of amplitudes over the pyramidal
time t. This means that the nonzero-offset traveltime surface is placed at its apex. The question that is of
section from such a recording would contain traces with practical importance is how to define the summation
zero amplitudes, except for one sample at time t on the paths over this surface.
trace at a midpoint location that coincides with the To consider alternative methods of summation, re-
center of the ellipse. Hence, the ellipse of equation (5- fer to the traveltime equation (5-32) that describes the
33) in the y − z plane describes the impulse response of pyramidal surface in Figure 5.3-1a, and, first, make the
a prestack migration operator applied to nonzero-offset change of variables from depth z to event time τ in the
data. migrated position by using the relation z = vτ /2, then
Equation (5-32) describes the nonzero-offset trav-
rewrite this equation in terms of the event position af-
eltime trajectory in the y −t plane for a constant z asso-
ter migration, which is the lateral coordinate ym of the
ciated with a point scatterer. Figure D-5 (Section D.1)
apex of the pyramid
shows the elliptic wavefront in the y−z plane equivalent
to the elliptic reflector geometry described here, and the τ2 (y − ym + h)2 τ2 (y − ym − h)2
table-top traveltime trajectory in the y − t plane. t= + + + .
4 v2 4 v2
When equation (5-33) is specialized to the zero- (5 − 35)
offset case, h = 0, we obtain
Hence, within the context of equation (5-35), the sum-
y2 z2
2
+ = 1, (5 − 34a) mation involves mapping amplitude at a point on the
(vt/2) (vt/2)2 pyramidal surface with coordinates (y, h, t) to the apex
which describes a circle in the y − z plane for a constant with coordinates (ym , h = 0, τ ).
t with a radius vt/2. This circle represents the impulse Whatever the summation strategy, it may not be
response of a poststack migration operator applied to desirable to map the amplitudes on the pyramidal sur-
zero-offset data. face directly onto the apex of the pyramid. Instead, it
When equation (5-32) is specialized to the zero- is desirable first to collapse the pyramidal surface de-
offset case, h = 0, we obtain scribed by equation (5-35) to a traveltime curve that
vt = 2 y2 + z2 , (5 − 34b) passes through the apex of the pyramid at y = ym de-
scribed by
which describes the well-known diffraction hyperbola in
the y − t plane for a constant z. Figure D-6 (Appendix 4h2
t= τ2 + . (5 − 36)
D) shows the circular wavefront in the y−z plane equiv- v2
alent to the circular reflector geometry described here, As a result, amplitude at a point on the pyramidal sur-
and the zero-offset hyperbolic traveltime trajectory in face with coordinates (y, h, t) is mapped onto a point
the y − t plane. with coordinates (ym , h, τ 2 + 4h2 /v 2 ) on the hyper-
In Section 4.1, we discussed semicircle superposi- bolic curve of equation (5-36). Now, you have the oppor-
tion and diffraction summation concepts for zero-offset tunity to perform velocity analysis using equation (5-36)
migration using equations (5-34a) and (5-34b), respec- and refine the velocity field used in the first summation
tively. Specifically, zero-offset migration can be con- step. The second step in the summation involves apply-
ceptualized as spreading of amplitudes on each input
ing NMO correction using equation (5-36), stacking the
stacked trace on the y − t plane along semicircular tra-
amplitudes along the offset axis, and placing the result
jectories on the y − z plane of the migrated section.
at the apex of the hyperbola of equation (5-36) at time
Alternatively, for a given output sample of a trace on
τ and offset h = 0. This apex coincides with the apex of
the z − t plane of the migrated section, amplitudes
along the hyperbolic trajectory on the y − t plane of the pyramidal surface with coordinates (ym , h = 0, τ ).
the input stacked section can be summed and placed on The two most obvious choices of summation paths
that output sample location. The Kirchhoff summation to collapse the pyramidal surface of equation (5-35) to
technique for migration incorporates to the process of the hyperbolic curve of equation (5-36) are described
diffraction summation the amplitude and phase factors below:
described in Section 4.1.
Similarly, prestack time migration can be concep- (a) Summation curves of constant offset: Consider a
tualized either by way of semi-elliptical superposition set of vertical cross-sections of the traveltime pyra-
Dip-Moveout Correction and Prestack Migration 727

FIG. 5.3-1. Left column (adapted from Fowler, 1997): The nonzero-offset traveltime surface associated with a point scatterer
and the various summation trajectories for prestack time migration. Right column: The nonzero-offset traveltime surfaces as
in the left column after DMO correction. See text for details.

mid illustrated in Figure 5.3-1a parallel to the mid- which is formed by combining the apex points Ah of
point axis as illustrated in Figure 5.3-1b. Sum the the constant-offset curves. This hyperbolic travel-
amplitudes along each of the constant-offset table- time curve is orthogonal to the constant-offset sum-
top traveltime curves, independently, and place the mation curves.
result for each at the apex Ah of the summation (b) Summation curves of constant time: Consider a
curve with coordinates (ym , h, τ 2 + 4h2 /v 2 ). The set of horizontal cross-sections of the traveltime
summation collapses the pyramidal surface onto pyramid as illustrated in Figure 5.3-1c (Bancroft
the hyperbolic traveltime curve of equation (5-36), and Geiger, 1994; Bancroft et al., 1997). Sum
728 Seismic Data Analysis

the amplitudes along each of the constant-time So, for the constant-offset and constant-time sum-
curves, independently, and place the result for each mation techniques described above, we may consider
at Ah on the summation curve with coordinates substituting the pyramidal surface with the hyper-
(ym , h, τ 2 + 4h2 /v 2 ). The summation collapses boloidal surface in the following manner:
the pyramidal surface onto the hyperbolic trav-
eltime curve of equation (5-36), which is formed (a) Summation curves of constant offset: Consider a
by combining the points Ah of the constant-time set of vertical cross-sections of the traveltime hy-
curves. This hyperbolic traveltime curve is orthog- perboloid parallel to the midpoint axis. Sum the
onal to the constant-time summation curves. amplitudes along each of the constant-offset hyper-
bolic traveltime curves of equation (5-38) as illus-
The pyramidal surface described by equation (5- trated in Figure 5.3-1e, independently, and place
35) is converted by DMO correction to a hyperboloid of the result for each at the apex Ah of the summa-
tion curve with coordinates (ym , h, τ 2 + 4h2 /v 2 ).
revolution as illustrated in Figure 5.3-1d and described
The summation collapses the hyperboloidal surface
by the following equation (Gardner et al., 1986):
onto the hyperbolic traveltime curve of equation (5-
4(y − ym )2 4h2 36), which is formed by combining the apex points
t= τ2 + + . (5 − 37) Ah of the constant-offset curves. This hyperbolic
v2 v2
traveltime curve is orthogonal to the constant-
As for equation (5-35), within the context of equation offset summation curves.
(5-37), the summation required by prestack time mi- (b) Summation curves of constant time: Consider a set
gration involves mapping amplitude at a point on the of horizontal cross-sections of the traveltime hyper-
hyperboloidal surface with coordinates (y, h, t) to the boloid (Gardner et al., 1986). Sum the amplitudes
apex with coordinates (ym , h = 0, τ ). along each of the constant-time circles of equation
Again, it is desirable first to collapse the hyper- (5-39) as illustrated in Figure 5.3-1f, independently,
boloidal surface described by equation (5-37) to a trav- and place the result for each at Ah on the √ summa-
eltime curve that passes through the apex of the hyper- tion curve with coordinates (ym , h, (v/2) t2 − τ 2 ).
boloid at y = ym described by equation (5-36). As a The summation collapses the hyperboloidal surface
result, amplitude at a point on the hyperbolodial sur- onto the hyperbolic traveltime curve of equation (5-
face with coordinates (y, h, t) is mapped onto a point 36), which is formed by combining the points Ah
with coordinates (ym , h, τ 2 + 4h2 /v 2 ) on the hyper- of the constant-time circles. This hyperbolic travel-
bolic curve of equation (5-36). As for the pyramidal time curve is orthogonal to the constant-time sum-
surface, the second step in the summation involves ap- mation curves.
plying NMO correction, stacking the amplitudes along
the offset axis, and placing the result at the apex of the
hyperbola of equation (5-36) at time τ and offset h = 0. DMO Correction and
This apex coincides with the apex of the hyperboloidal Common-Offset Migration
surface with coordinates (ym , h = 0, τ ).
For constant offset h, rewrite equation (5-37) in the As stated earlier, a desired workflow for prestack time
following form: migration incorporates a step for updating the initial
t2 (y − ym )2 velocity field that was used to migrate the data. Specifi-
− = 1. (5 − 38) cally, we want to create common-reflection-point (CRP)
τ 2 + 4h2 /v 2 h2 + v 2 τ 2 /4
gathers from prestack time migration and use them to
Note that, as a result of the transformation from the perform conventional velocity analysis based on the hy-
pyramidal surface to the hyperboloidal surface, the perbolic moveout assumption. Finally, we can apply
table-top summation curves at constant offset are trans- normal-moveout correction to the CRP gathers using
formed to hyperbolic curves described by equation (5- the updated velocity field and stack them along to ob-
38). tain the prestack time-migrated section.
For constant time t, rewrite equation (5-37) in the In this section, we shall follow a most popular se-
following form: quence for prestack time migration based on a practi-
cal variation of the constant-offset summation technique
v2 2 described above (Marcoux et al., 1987).
(y − ym )2 + h2 = (t − τ 2 ). (5 − 39)
4
Note that the constant-time curves of the pyramidal (a) Perform velocity analysis at sparse intervals and
surface are transformed to circles described by equation pick just a few velocity functions with minimal dip
(5-39). effects.
Dip-Moveout Correction and Prestack Migration 729

(b) Apply NMO correction using these flat-event ve- (right column). These migrated sections represent the
locities. nonzero-offset (prestack) migration impulse responses,
(c) Sort data to common-offset sections and apply and their elliptical trajectories are defined by equation
DMO correction. (5-33).
(d) Migrate each of the common-offset sections us- Once again, consider the section in Figure 5.3-2a
ing your favorite zero-offset migration algorithm (left column) as a common-offset section with assigned
(Chapter 4) and the velocity field based on the flat-
offsets values of 0, 1000, 2000, and 3000 m. If we were
event velocity picks from step (a).
to have migrated each of these common-offset sections
(e) Sort the migrated common-offset data back to
CMP gathers, and apply inverse NMO correction directly using a prestack migration algorithm in lieu
using the flat-event velocities from step (a). of first applying DMO correction then migrating using
(f) Perform velocity analysis at frequent intervals as a zero-offset migration algorithm, we would have ob-
needed to derive an optimum stacking velocity tained the same results as those shown in Figure 5.3-2
field. (right column). Figure 5.3-3 shows the DMO impulse re-
(g) Apply NMO correction using the optimum stacking sponses on the left column of Figure 5.3-2 superimposed
velocity field. on the migration impulse responses on the right col-
(h) Stack the data and perform inverse migration umn of the same figure. For the zero-offset case (Figure
(equivalent to 2-D zero-offset wavefield modeling) 5.3-3a), the DMO impulse response has zero aperture,
using the velocity field from step (d). while for the nonzero-offset case the lateral excursion of
(i) Finally, remigrate the result from step (h) using the DMO impulse response increases with offset and at
the updated velocity field from step (f). shallow times. But, the lateral excursion of the DMO
impulse response always is much less than that of the
This sequence differs from the constant-offset summa- migration impulse response.
tion procedure for prestack time migration (Figure 5.3- This experiment demonstrates that zero-offset mi-
1e) in one respect, only. In case of the latter, step (e)
gration of DMO-corrected common-offset data is equiv-
precedes step (d). Although this is more plausable, in
alent to prestack migration of nonzero-offset data. Of
practice, the two procedures yield comparable results.
Consider a zero-offset section that contains a set course, this equivalence is largely valid only within the
of pulses at 500-ms intervals placed on the center trace bounds of velocity variations judged to be acceptable for
as shown in Figure 5.3-2a (left column). Now, assume time migration. In this section, we shall apply the se-
that this same section represents a common-offset sec- quence outlined above to two common cases of conflict-
tion with 1000-m offset. Apply DMO correction to this ing dips with different stacking velocities — salt flanks
section to obtain the section in Figure (5.3-2b, left col- and fault planes.
umn). Repeat the same exercise by labeling the input
section with pulses (Figure 5.3-2a, left column) as a
common-offset section with 2000-m offset and apply
Salt Flanks
DMO correction to get the section in Figure 5.3-2c (left
column). Repeat this exercise once more with an offset
value of 3000 m to get the section in Figure 5.3-2d (left DMO correction maps moveout-corrected common-
column). The DMO-corrected common-offset sections offset data to zero-offset. As a direct result of this aspect
shown on the left column of Figure 5.3-2 represent the of DMO correction, the process decouples common-
DMO impulse responses for zero-offset, 1000-m, 2000-m offset sections, thus enabling treatment of each of
and 3000-m offset cases. the common-offset sections, independently. A DMO-
Following the sequence described above, now, treat corrected common-offset section can be considered a
each of the DMO-corrected sections on the left column replica of a zero-offset section, and therefore, can be
in Figure 5.3-2 as zero-offset sections and migrate them migrated using a method applicable to zero-offset wave-
using a zero-offset constant-velocity algorithm. Results field.
are shown on the right column of Figure 5.3-2. Migra-
tion of the section in Figure 5.3-2a (left column) yields
a set of concentric circles (right column). This migrated (a) Starting with input prestack data in midpoint, off-
section represents the zero-offset migration impulse re- set and two-way event time in the unmigrated posi-
sponse, and its circular trajectories are defined by equa- tion, apply NMO correction using flat-event veloc-
tion (5-34a). Migrations of the sections in Figure 5.3- ities. These are picked from velocity spectra com-
2b,c,d (left column) yield a set of elliptical trajectories puted sparsely along the line. Figure 5.3-4 shows
(text continues on p. 742)
730 Seismic Data Analysis

FIG. 5.3-2. Left column: Impulse response of a frequency-wavenumber dip-moveout operator with source-receiver offsets (a)
0 m, (b) 1000 m, (c) 2000 m, and (d) 3000 m. Right column: Zero-offset migrations of the sections in the left column. These
sections are equivalent to the results one would obtain from presatck migrations of common-offset sections with source-receiver
offsets (a) 0 m, (b) 1000 m, (c) 2000 m, and (d) 3000 m, hence can be considered as impulse response of a prestack time
migration algorithm.
Dip-Moveout Correction and Prestack Migration 731

FIG. 5.3-3. The impulse response of a dip-moveout operator (left column in Figure 5.3-2) and prestack time migration
operator (right column in Figure 5.3-2) superimposed. The source-receiver offsets for the input common-offset sections are (a)
0 m, (b) 1000 m, (c) 2000 m, and (d) 3000 m.
732 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 733
734 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 735
736 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 737
738 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 739

FIG. 5.3-11. Stack of the gathers as in Figure 5.3-10.


740 Seismic Data Analysis

FIG. 5.3-12. Inverse migration of the stack shown in Figure 5.3-11 using the velocity field shown in Figure 5.2-9b.
Dip-Moveout Correction and Prestack Migration 741

FIG. 5.3-13. Migration of the stack shown in Figure 5.3-12 using the velocity field shown in Figure 5.3-9b.
742 Seismic Data Analysis

three common-offset sections associated with the 5.3-11) using the initial velocity field (Figure 5.2-
data as in Figure 5.2-3 with offsets 78.5 m, 1078.5 9a). The modeled section is shown in Figure 5.3-
m, and 2078.5 m, following NMO correction. 12 and may be treated as equivalent to the un-
(b) Apply DMO correction to each common-offset sec- migrated stacked section. The next step involves
tion (Figure 5.3-5). migrating the modeled section using the updated
(c) Following DMO correction, each common-offset velocity field (Figure 5.3-9b). The final result of
section is assumed to be a replica of zero-offset sec- prestack time migration sequence described above
tion, and thus, can be migrated using a zero-offset is shown in Figure 5.3-13.
migration algorithm. How could these common-
offset sections be migrated before we even know
migration velocities? The conjecture is that a
smoothed stacking velocity field (Figure 5.2-9a) Fault Planes
can be used to perform the migrations of common-
offset data. As a result, events will be moved spa- Figure 5.3-14 shows a portion of a conventional CMP
tially to locations that are between their unmi- stack without DMO correction. Steeply dipping events
grated and correctly migrated positions, but fairly represent diffractions and fault-plane reflections. Be-
close to the latter. Figure 5.3-6 shows three selected cause no DMO correction has been applied, these events
common-offset sections as in Figure 5.3-5 after mi- have not been preserved with as much strength as the
gration. gently dipping reflections. As a result, migration of this
(d) Sort the migrated common-offset sections into conventional CMP stack yields an inadequate definition
CMP gathers (Figure 5.3-7). Events on these gath- of the intensive fault pattern (Figure 5.3-15).
ers are now assumingly close to correct subsur- We shall use the data set shown in Figure 5.3-14 to
face positions after migration. Note the presence bring together and review all we learned in this chapter
of residual moveout on some events, which suggest about DMO correction and prestack time migration. To
errors in the velocity field used for migration. begin with, examine the dip effect on stacking velocities.
(e) Apply inverse NMO correction and perform ve- Based on velocity analysis at CMP location A, a veloc-
locity analysis. Although event positioning in the ity function appropriate for gently dipping reflections
spatial sense will not be affected by this velocity was picked. By setting this as the center function and
analysis, use of updated velocities will improve the using a range of lower and higher percents of this func-
stacking of the CMP gathers after migration. Fig- tion, a set of multivelocity-function stack panels shown
ure 5.3-8 shows selected CMP gathers as in Figure in Figures 5.3-16 and 5.3-17 was created over a CMP
5.3-7 after inverse moveout correction, and Figure range that includes a zone with gently dipping reflec-
5.3-9b shows the migration velocity field derived tions and steeply dipping events. Note that the gently
from the analysis of such gathers. For comparison,
dipping reflections stack best with 100 percent of the
the DMO velocity field using the DMO-corrected
center function (Figure 5.3-16) and the steeply dipping
gathers as in Figure 5.2-5 is shown in Figure 5.3-9a.
events stack best with 120 percent of the center function
(f) Apply NMO correction (Figure 5.3-10) using the
(Figure 5.3-17). This 20 percent difference in stacking
migration velocity field (Figure 5.3-9b) and stack
velocities between the gently dipping and steeply dip-
the migrated data (Figure 5.3-11). Keep in mind
ping reflections is primarily stems from the dip effect.
that migration actually has been done using an ini-
tial estimate for the velocity field (Figure 5.2-9a). After DMO correction, both gently dipping and steeply
(g) To obtain the migrated section with the updated dipping events stack best with 100 percent of the center
velocity field (Figure 5.3-9b), we can follow two al- function (Figures 5.3-18 and 5.3-19).
ternative approaches. First, compute the residual Removal of the dip effect on stacking velocities by
velocity field (Section D.8) from the initial (Figure DMO correction is further demonstrated by velocity
5.2-9a) and updated (Figure 5.3-9b) velocity fields, analysis before and after DMO correction. Figure 5.3-
and use it to perform residual migration (Section 20 shows the CMP gather and its velocity spectrum at
4.5) for which the input is the stacked section from location A as in Figure 5.3-14. At first impression, the
step (f). The residual migration is acceptable pro- velocity spectrum may suggest the presence of multi-
vided the initial velocity field varies vertically, only, ples. Nevertheless, there actually exist two sets of picks
and the vertical variation in velocity is expressed — the low-velocity picks as shown in Figure 5.3-21 as-
with moderate gradients. The second approach in- sociated with the gently dipping reflections, and the
volves first simulation of an unmigrated stacked high-velocity picks as shown in Figure 5.3-22 associated
section by forward modeling (inverse migration) of with the steeply dipping fault-plane reflections and pos-
the migrated section obtained in step (f) (Figure sibly diffractions. By applying NMO correction to the
(text continues on p. 768)
Dip-Moveout Correction and Prestack Migration 743
744 Seismic Data Analysis

5
Dip-Moveout Correction and Prestack Migration 745
746 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 747
748 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 749

FIG. 5.3-20. CMP gather (left) and the velocity spectrum (right) at location A as in Figure 5.3-16 before DMO correction.
750 Seismic Data Analysis

FIG. 5.3-21. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-20, with moveout correction applied to
the data using the velocity function that is posted on the spectrum. This velocity function is appropriate for the nearly flat
events associated with the sedimentary strata.
Dip-Moveout Correction and Prestack Migration 751

FIG. 5.3-22. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-20, with moveout correction applied to the
data using the velocity function that is posted on the spectrum. This velocity function is appropriate for the steeply dipping
events associated with the fault-plane reflections.
752 Seismic Data Analysis

FIG. 5.3-23. CMP gather (left) and the velocity spectrum (right) at location A as in Figure 5.3-18 after DMO correction.
Dip-Moveout Correction and Prestack Migration 753

FIG. 5.3-24. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-23, with the velocity function of the gently
flat events as in Figure 5.3-21 posted on the spectrum.
754 Seismic Data Analysis

FIG. 5.3-25. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-23, with the velocity function of the
steeply dipping events as in Figure 5.3-22 posted on the spectrum.
Dip-Moveout Correction and Prestack Migration 755

FIG. 5.3-26. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-23, with moveout correction applied to
the data using the DMO-corrected velocity function that is posted on the spectrum.
756 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 757
758 Seismic Data Analysis

FIG. 5.3-29. Portions of three common-offset sections that coincide with the CVS panels shown in Figure 5.3-16 before
DMO correction. Offsets are, from left to right, 500, 1500, and 2500 m.
Dip-Moveout Correction and Prestack Migration 759

FIG. 5.3-30. Portions of three common-offset sections that coincide with the CVS panels shown in Figure 5.3-18 after DMO
correction. Offsets are, from left to right, 500, 1500, and 2500 m.
760 Seismic Data Analysis

FIG. 5.3-31. Portions of three common-offset sections that coincide with the CVS panels shown in Figure 5.3-18 after DMO
correction and common-offset migration. Offsets are, from left to right, 500, 1500, and 2500 m.
Dip-Moveout Correction and Prestack Migration 761

FIG. 5.3-32. CMP gather (left) and the velocity spectrum (right) at location A as in Figure 5.3-18, but after DMO correction
and common-offset migration. The velocity function posted on the spectrum is the same as in Figure 5.3-26 derived from the
data after DMO correction.
762 Seismic Data Analysis

FIG. 5.3-33. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-32 after DMO correction, common-offset
migration and inverse NMO correction using the velocity function posted on the spectrum. This velocity function is the same
as in Figure 5.3-26 derived from the data after DMO correction.
Dip-Moveout Correction and Prestack Migration 763

FIG. 5.3-34. CMP gather (left) and the velocity spectrum (right) as in Figure 5.3-33 after DMO correction, common-offset
migration inverse NMO correction using the velocity function posted on the spectrum in Figure 5.3-33, and finally, NMO
correction using the velocity function posted on the spectrum shown here and derived from the gather shown here after DMO
correction and common-offset migration.
764 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 765
766 Seismic Data Analysis
Dip-Moveout Correction and Prestack Migration 767
768 Seismic Data Analysis

CMP gather using the low-velocity function in Figure the presence of steeply dipping fault-plane reflec-
5.3-21, events that correspond to the gently dipping re- tions. Without DMO correction, CMP stacking
flections are flattened whereas events that correspond to fails to preserve these reflections as shown in Figure
the steeply dipping reflections and diffractions are over- 5.3-14.
corrected. By applying NMO correction to the CMP (b) Apply DMO correction to each common-offset sec-
gather using the high-velocity function in Figure 5.3- tion. Figure 5.3-30 shows portions of the three
22, events that correspond to the steeply dipping reflec- selected common-offset sections as in Figure 5.3-
tions are flattened whereas events that correspond to 29 after DMO correction. Note the steeply dip-
the gently dipping reflections and diffractions are un- ping fault-plane reflections. With DMO correction,
dercorrected. CMP stacking preserves these reflections as shown
The velocity spectrum at the same analysis loca- in Figure 5.3-27.
tion as for Figure 5.3-20 after the application of DMO (c) Following DMO correction, each common-offset
correction shows a single, unambiguous velocity trend section is assumed to be a replica of a zero-offset
as shown in Figure 5.3-23. We have resolved the dip ef- section, and thus, can be migrated using a zero-
fect on stacking velocities and thus eliminated the mul- offset migration algorithm. In this case, a single,
tiple number of velocity picks as in Figure 5.3-20. As a vertically varying velocity function was used to
result of the partial migration effect of DMO correction, migrate the NMO- and DMO-corrected common-
energy is moved from one CMP location to another. offset sections. Figure 5.3-31 shows portions of the
Consequently, reflection-point smearing is removed and three selected common-offset sections as in Figure
events on CMP gathers improvise a horizontally layered 5.3-30 after common-offset migration.
earth model. (d) Sort the migrated common-offset sections into
For comparison, the low-velocity function associ- CMP gathers. Events on these gathers are now as-
ated with the gently dipping reflections in Figure 5.3- sumingly close to correct subsurface positions after
21 is superimposed on the velocity spectrum after DMO migration. Figure 5.3-32 shows a CMP gather after
correction in Figure 5.3-24. As anticipated, the picks im- common-offset migration.
plied by the velocity spectrum after DMO correction are (e) Apply inverse NMO correction using the veloci-
fairly close to the velocity function associated with the ties picked before DMO correction and perform ve-
gently dipping reflections. Similarly, the high-velocity locity analysis. The velocity spectrum computed
function associated with the steeply dipping reflections from migrated data is shown in Figure 5.3-32 with
and diffractions in Figure 5.3-22 is superimposed on the the DMO velocities posted on it for comparison.
velocity spectrum after DMO correction in Figure 5.3- The CMP gather after inverse NMO correction us-
25. In this case, the picks implied by the velocity spec- ing the velocity function shown in Figure 5.3-32 is
trum after DMO correction are significantly lower than shown in Figure 5.3-33.
the velocity picks associated with the steeply dipping (f) Apply NMO correction using the velocity picks af-
reflections. ter migration (Figure 5.3-34). The velocity field
The velocity analysis is repeated after DMO cor- derived from the post-migration velocity picks as
rection as shown in Figure 5.3-26. The picked velocities in Figure 5.3-34 is shown in Figure 5.3-35, and
are then used to stack the data (Figure 5.3-27). Com- the stack based on this velocity field is shown in
pare with the conventional CMP stack in Figure 5.3-14 Figure 5.3-36. This stack indeed is equivalent to
and note that the DMO stack shown in Figure 5.3-27 prestack time-migrated section. Note that, how-
has preserved the events with conflicting dips with dif- ever, common-offset migration actually was done
ferent stacking velocities — the gently dipping reflec- using a single, vertically varying velocity function
tions with the velocity function posted in Figure 5.3-21 as in step (c).
and the steeply dipping fault-plane reflections with the (g) To obtain the migrated section with the velocity
velocity function posted in Figure 5.3-22. The result- field (Figure 5.3-35) derived after common-offset
ing migrated section in Figure 5.3-28 shows a superior migration, first perform inverse migration of the
image of the fault blocks as compared to the migrated resulting stack from step (f) using the same ve-
section without DMO correction (Figure 5.3-15). locity function as in step (c) to obtain a zero-offset
For prestack time migration, we shall follow the section equivalent to an unmigrated stack as shown
same sequence as for the salt-flank data set. in Figure 5.3-37. Then, migrate this zero-offset sec-
tion using the post-migration velocity field shown
(a) Sort the moveout-corrected CMP gathers to in Figure 5.3-35 to obtain the final result from
common-offset sections. Figure 5.3-29 shows por- prestack time migration sequence described here
tions of three selected common-offset sections. Note (Figure 5.3-38).
Dip-Moveout Correction and Prestack Migration 769

Common-Reflection-Point versus can you associate it with a hyperbolic traveltime. In-


Common-Reflection-Surface Stacking stead, when migrating the data, you have to deal with
a complex, distorted traveltime trajectory. And neither
Conventional stacking is based on the notion of a com- can you associate the reflections with hyperbolic trav-
mon midpoint (CMP) and migration is based on the no- eltimes. Instead, when stacking the data, you have to
tion of a common reflection point (CRP). In both cases, deal with a nonhyperbolic moveout trajectory associ-
we assume that reflections are represented by hyperbo- ated with many reflection points scattered around in
las and reflectors are represented by points. Consider a the subsurface (de Bazelaire, 1988).
seismic line from the Alberta Plains of Western Canada. So the simple hyperbolic and point rules of the
The subsurface is just as flat as the surface over which Plains or the Foothills are no longer applicable in the
you have recorded the data. When stacking the data, Thrust Belt. To overcome the first problem — migration
you can almost picture summing of the amplitudes in of data in the presence of strong lateral velocity varia-
a CMP gather over all offsets along a hyperbolic tra- tions associated with complex overburden structures in
jectory associated with a zero-offset time and the re- the Thrust Belt, you may decide to do the imaging in
sulting stacked amplitude being placed at a point re- depth (Section 8.0) instead of imaging in time (Section
flector where the CMP raypaths converge conveniently. 4.0). Earth imaging in depth requires earth modeling
When migrating the stacked data, again, you sum the in depth (Section 9.0) — a challenge much higher than
amplitudes along a hyperbolic trajectory and place the the imaging itself. To overcome the second problem —
result at the apex of the hyperbola. You conveniently stacking of data with nonhyperbolic moveout, you may
associate the apex of the latter hyperbola with a point combine it with the first problem and pursue a rigorous
diffractor situated on the reflector. Whether it is a re- solution by doing the imaging not just in depth but also
flection hyperbola associated with a point reflector or a before stack.
diffraction hyperbola associated with a point diffractor, Note that, insofar as stacking and imaging in time
the process of stacking and migrating the data involves or in depth, we choose to map events to common-
summation of amplitudes along hyperbolas and placing reflection points. By way of DMO correction, we map
the resulting sum to a point in the subsurface. events to common-reflection points in their unmigrated
Now consider a seismic line from the Alberta
positions. Similarly, by way of prestack time migration,
Foothills where the cascaded flanks of the Rocky Moun-
we map events to common-reflection points in their mi-
tains rise steeply. The subsurface is just as steep as
grated positions. De Bazelaire (1988) and Gelchinsky
the surface over which you have recorded the data. A
(1988) pointed out that, rather than associating the
diffractor still is a point whether it is in the Plains
recorded data with common-reflection points, we should
or in the Foothills, and you can still think of diffrac-
associate the recorded data with common-reflection sur-
tions as hyperbolas so long as they are situated below
faces. As such, when stacking the data, rather than con-
a simple overburden. Fortunately, you can also think of
reflections as hyperbolas. No longer, however, can you stricting the summation of amplitudes to within a sin-
associate a reflection hyperbola on your CMP gather gle CMP gather, data may be focused to a common-
with a single point reflector; instead, you have reflec- reflection surface (CRS) using multiple shots and re-
tion points dispersed along the reflector (Section E.1). ceivers (Gelchinsky, 1988; and Höcht, 1998).
This is when you have to introduce DMO correction to The theory and practice of the CRS stacking
account for the reflection-point dispersal (Section 5.1). method (Gelchinsky, 1988; Tygel et al., 1997; Gelchin-
Once the reflection-point dispersal is removed, the re- sky et al., 1999a,b; Höcht, 1998; Berkovitch et al., 1998;
sulting stack can be considered equivalent to a zero- Landa et al., 1999) suggest that it is indeed related to
offset section which you can migrate, again, using the prestack time migration when the medium velocities are
hyperbolic summation rule. Thus you have been able judged to be within the acceptable bounds for the lat-
to overcome the Foothills problem of steeply dipping ter. The CRP gathers from prestack time migration and
events. the CRS gathers both are created by including in the
Finally, consider a seismic line from the Canadian summation aperture multiple numbers of CMP gath-
Thrust Belt west of the Foothills. The subsurface is ers. The difference is that the CRP gathers are associ-
just as complex as it appears on the surface over which ated with events in their migrated positions, whereas the
you have recorded the data. A diffractor situated below CRS gathers are associated with events in their unmi-
the complex overburden structure caused by the over- grated positions. So, migration of the CRS stack should
thrust tectonics stays as a point; no longer, however, resemble the section from prestack time migration.
770 Seismic Data Analysis

FIG. 5.3-39. A common-reflection-surface (CRS) stack. (Data courtesy Saudi Aramco.)


Dip-Moveout Correction and Prestack Migration 771

FIG. 5.3-40. Migration of the common-reflection-surface (CRS) stack shown in Figure 5.3-39.
772 Seismic Data Analysis

FIG. 5.3-41. A common-reflection-point (CRP) stack of the data as in Figure 5.3-40 derived from prestack time migration.
Dip-Moveout Correction and Prestack Migration 773
774 Seismic Data Analysis

FIG. 5.3-43. A common-reflection-point (CRP) gather associated with the CRP stack shown in Figure 5.3-41.
Dip-Moveout Correction and Prestack Migration 775

FIG.5.4-2. Migrations of the CMP gather in Figure 5.4-


1a using velocities that are (a) lower than the medium, (b)
FIG. 5.4-1. (a) A CMP gather containing a single reflec- the same as the medium, and (c) higher than the medium
tor in a constant-velocity medium; (b) velocity spectrum velocity. (Note the wraparound effect in (c) that is inherent
derived by migrating the CMP gather using a number of to the phase-shift method used in this analysis.)
constant velocities and displaying the zero-offset trace from
each migration; (c) velocity spectrum derived by NMO cor- the earth’s surface. In turn, downward extrapolation
recting and stacking the CMP gather using the same range provides a basis for CMP stacking and migration (Clay-
of constant velocities as in (b). Each trace in (b) is a zero- ton, 1978; Yilmaz and Claerbout, 1980). Because the
offset trace, while each trace in (c) is a CMP stacked trace. processes of CMP stacking and migration require ve-
locity information, they also can be used to obtain a
Figure 5.3-39 shows a CRS stack from an area with velocity estimate (Taner and Koehler, 1969; Gardner et
low-relief structures. The field data were recorded with
al., 1974).
240-fold coverage. For CRS processing, the negative off-
For example, consider the problem of conventional
sets were discarded and the near one-half of the positive
velocity estimation for stacking. Figure 5.4-1a shows a
offsets were kept; thus the reduced data used to create
the CRS stack had 60-fold coverage. For prestack time CMP gather over a single horizontal reflector. Select a
migration, the original 240-fold data were first reduced constant velocity, apply NMO correction, and stack the
to 60-fold by partial stacking. The migrated CRS stack traces in the gather. Next, place this stack trace onto
shown in Figure 5.3-40 is comparable to the CRP stack the plane of velocity versus two-way zero-offset vertical
derived from prestack time migration shown in Figure time (v, τ ) as shown in Figure 5.4-1c. By stacking the
5.3-41. In the present example, five CMP gathers were gather with different constant velocity values, the entire
used to create one CRS gather with 300 traces as shown (v, τ ) plane is filled with stacked amplitudes. (In this
in Figure 5.3-42. The 60-fold CRP gather at the same section, the variable τ is used specifically to define the
location as for the CRS gather of Figure 5.3-42 is shown vertical two-way zero-offset time, τ = 2z/v.)
in Figure 5.3-43. We now consider the migration process. For a hor-
Practical refinements to various approaches to CRS izontally stratified earth, as in Figure 5.4-1a, we cannot
stacking and imaging, including extensions to 3-D, are distinguish between a CMP gather and a common-shot
under current investigation. As to what extent the tech- gather (CSG). Moreover, since a CSG is a true wavefield
nique will be applied routinely to seismic data remains created by a single shot and recorded by many receivers,
to be seen. it seems reasonable that the CMP gather in Figure
5.4-1a can be migrated by treating the reflection hy-
perbola as if it were a diffraction hyperbola. Assuming
5.4 MIGRATION VELOCITY ANALYSIS there is no velocity information, we migrate with various
trial constant velocities and evaluate the results. Figure
Velocity estimation, CMP stacking, and migration gen- 5.4-2 shows three different attempts at migrating the
erally are considered independent processes. However, CMP gather of Figure 5.4-1a. In one attempt (Figure
they all have a common theoretical base — the scalar 5.4-2a), too low a velocity was used; hence, the event
wave equation. Solution of this equation allows down- was undermigrated. In another attempt, too high a ve-
ward extrapolation of a seismic wavefield recorded at locity was used and the event was overmigrated (Figure
776 Seismic Data Analysis

FIG. 5.4-3. A flowchart of an algorithm for prestack Stolt


migration.

5.4-2c). When the velocity in the migration equals the FIG. 5.4-4. A flowchart of an algorithm for common-offset
medium velocity, we expect the diffraction hyperbola migration of DMO-corrected data.
to collapse to its apex, which is at the zero-offset trace
(Figure 5.4-2b). apply moveout correction to a CMP gather and stack
What is the implication of this experiment for ve- the traces in the gather using a range of constant ve-
locity estimation? Since the correct velocity produces locities. Results can be used for velocity determination
a well-compressed event at the apex of the hyperbola, either in the form of a set of velocity spectra at selected
this velocity can be estimated by evaluating the quality locations along the line or a set of constant-velocity-
of focusing at zero offset. To evaluate focusing, we pick stack (CVS) panels. To estimate migration velocities, in
out the zero-offset traces from the various attempts at principle, we migrate prestack data using a range of con-
migration with different velocities and place them side stant velocities and output selected gathers in their mi-
by side. This produces a display of velocity versus two- grated positions or constant-velocity migration (CVM)
way zero-offset time as shown in Figure 5.4-1b. panels. The key difference between stacking velocity es-
When Figures 5.4-1b and 5.4-1c are compared, an timation and migration velocity estimation is that the
almost identical character is noted. The resolution of former only requires individual CMP gathers, whereas
velocity information obtained in the two approaches is the latter requires prestack data set in its entirety. The
equally degraded by data limitations such as maximum reason for this is that the processes of moveout correc-
source-to-receiver offset and the absence of short-offset tion and stacking are confined to within a CMP gather,
traces. while the process of migration moves energy spatially
No distinct difference exists between migration and from one CMP location to another.
stacking velocity when the subsurface medium is hori-
zontally layered as in Figure 5.4-1. However, for dipping
reflectors, the two types of velocity differ. Stacking ve-
locity is sensitive to the dip of the reflecting interface Prestack Stolt Migration
(Section C.3), while, in theory, migration velocity is the
medium velocity independent of dip. Therefore, for seis- The first method for migration velocity analysis that we
mic migration, we must use a velocity field that is cor- shall review is based on migrating prestack data using
rected for dips present in the data. As a result, any a range of constant velocities and creating constant-
procedure that obtains velocities suitable for migration velocity migration (CVM) panels (Shurtleff, 1984).
must use data from a number of neighboring CMP gath- Since migrations are performed using constant veloc-
ers. ities, an appropriate choice for the algorithm would
Migration velocity analysis can be formulated in be prestack frequency-wavenumber migration (Section
much the same way as stacking velocity analysis. To E.6). A flowchart for the CVM approach for migration
estimate stacking velocities (Chapter 3), in principle, we velocity analysis is shown in Figure 5.4-3.
Dip-Moveout Correction and Prestack Migration 777

If the medium velocity is constant, then migration (d) Invoke the imaging principle by setting t = 0 and
can be expressed as a direct mapping (Stolt, 1978) from obtain P (ky , kh , ωτ , t = 0).
temporal frequency ω to vertical wavenumber kz (Sec- (e) Sum over the offset wavenumber kh to obtain the
tion E.6). The equation for Stolt mapping is image at zero offset, yet in the transform domain,
P (ky , h = 0, ωτ ; v).
v kz2 − ky2 kh2
P (ky , kh , kz , t = 0) = (f) Perform 2-D inverse Fourier transform to obtain
2 (kz2 + ky2 )(kz2 + ky2 ) the constant-velocity migrated zero-offset section,
P (y, τ ; v).
v
× P ky , kh , 0, (kz2 + ky2 )(kz2 + kh2 ) , (g) Repeat steps (b) through (f) for a range of constant
2kz velocities to obtain the migration velocity volume
(5 − 40) P (y, τ ; v). By viewing this volume, it can be incised
where y, h, and t are the variables for midpoint, offset to obtain the surface of optimum migration veloc-
and event time in the unmigrated position, and ky , kh , ity field with an accompanying image derived from
and ω are the associated Fourier transform variables. prestack time migration.
Note that Stolt migration involves, first, mapping
from ω to kz for a specific ky and kh by using the disper- Practical issues related to prestack Stolt migration
sion relation for prestack wave extrapolation (Section include spatial aliasing along the offset axis and cost
E.6): of Stolt mapping in steps (b) and (c). The spatial sam-
pling along the offset axis often is too coarse for shallow
v
ω= (kz2 + ky2 )(kz2 + kh2 ). (5 − 41) events with low velocity; this gives rise to large moveout
2kz on CMP gathers. A linear moveout may be applied to
The output of mapping is then scaled by the quantity CMP gathers to circumvent spatial aliasing. Equation
(5-40) for Stolt mapping is then modified accordingly
v kz2 − ky2 kh2 (Li et al., 1991).
S= . (5 − 42) The Stolt mapping of amplitudes for prestack data
2 (kz2 + ky2 )(kz2 + kh2 ) involves interpolation of complex numbers in the trans-
form domain. This involves the three input variables ky ,
Stolt migration output normally is displayed in
kh , and ω, and the output variable ωτ , and thus is quite
two-way vertical zero-offset time τ = 2z / v. In prac-
costly when one has to consider as many as 100 or more
tice, mapping in the f − k domain really is from ω to
constant velocities. A way to reduce the computational
ωτ rather than from ω to kz , where ωτ is the Fourier
cost is to perform prestack migration using a set of con-
dual of τ , and is simply kz scaled by v/2. Accordingly,
stant velocities at coarse interval, followed by poststack
equations (5-40), (5-41), and (5-42) are recast in terms
residual constant-velocity migrations of the zero-offset
of ωτ = (v/2)kz when implemented in practice.
sections from prestack migration to fill in between the
Migration velocity analysis based on Stolt’s
coarsely sampled migration velocity panels (Li et al.,
prestack algorithm for constant velocity thus involves
1991).
the following steps:

(a) Starting with prestack data P (y, h, t) in coordi-


nates of midpoint y, offset h and two-way event Common-Offset Migration of
time t in the unmigrated position, perform 3-D DMO-Corrected Data
Fourier transform to obtain the transformed vol-
ume of data P (ky , kh , ω), where ky , kh , and ω are This technique for migration velocity analysis is the ba-
the Fourier transform duals of the variables y, h, sis for the most common sequence used for prestack time
and t, respectively. migration (Section 5.3). A flowchart for the method is
(b) For each trial constant velocity v, use equation shown in Figure 5.4-4.
(5-41) to map the transform variable ω — the
temporal frequency associated with the input data (a) Starting with prestack data P (y, h, t) in coordi-
P (ky , kh , ω), to ωτ — the temporal frequency as- nates of midpoint y, offset h, and two-way event
sociated with the migrated data P (ky , kh , ωτ ; v). time t in the unmigrated position, apply NMO cor-
This mapping of complex numbers is the basis for rection using an initial velocity field with minimal
constant-velocity prestack Stolt migration (Section dip effect.
E.6). (b) Sort the normal-moveout corrected data P (y, h, tn )
(c) Apply the scaling factor of equation (5-42). to common-offset sections P (y, tn ; h).
(text continues on p. 788)
778 Seismic Data Analysis

FIG. 5.4-5. A CMP stacked section associated with data from an overthrust belt. (Data courtesy Husky Oil.)
Dip-Moveout Correction and Prestack Migration 779

FIG. 5.4-6. Velocity field associated with the stacked section shown in Figure 5.4-5.
780 Seismic Data Analysis

FIG. 5.4-7. Poststack time migration of the stacked section shown in Figure 5.4-5 using the velocity field shown in Figure
5.4-6.
Dip-Moveout Correction and Prestack Migration 781

FIG. 5.4-8. Part 1: Constant-velocity prestack Kirchhoff time migration panels of the data as in Figure 5.4-5.
782 Seismic Data Analysis

FIG. 5.4-8. Part 2: Constant-velocity prestack Kirchhoff time migration panels of the data as in Figure 5.4-5.
Dip-Moveout Correction and Prestack Migration 783

FIG. 5.4-8. Part 3: Constant-velocity prestack Kirchhoff time migration panels of the data as in Figure 5.4-5.
784 Seismic Data Analysis

FIG. 5.4-9. Part 1: Velocity panels from the overthrust data created by sorting the results of prestack constant-velocity
migrations shown in Figure 5.4-8.
Dip-Moveout Correction and Prestack Migration 785

FIG. 5.4-9. Part 2: Sorted results from Figure 5.4-8. Each panel shows results of constant-velocity migrations in the neigh-
borhood of the CMP location as denoted.
786 Seismic Data Analysis

FIG. 5.4-10. Time migration velocity field created from the velocity function picks made from the analysis panels shown in
Figure 5.4-9.
Dip-Moveout Correction and Prestack Migration 787

FIG. 5.4-11. Prestack time migration using the velocity field shown in Figure 5.4-10.
788 Seismic Data Analysis

(c) To each common-offset section P (y, tn ; h), apply thrust faults. The sole thrust — zone of detachment
DMO correction P (y, τ0 ; h). between the competent rock layers below 2 s and the
(d) Migrate each of the common-offset sections using incompetent rock layers above, can be traced along a
a zero-offset migration algorithm (Chapter 4) and trajectory that starts at approximately 1.75 s on the
the velocity field based on the flat-event velocity left-edge of the section, climbs upward and emerges on
picks from step (a). the surface at approximately midpoint 510.
(e) Sort the migrated common-offset data P (y, τ ; h) We now perform prestack Kirchhoff migration us-
back to CMP gathers P (y, h, τ ) and apply inverse ing a range of constant velocities and obtain a set
NMO correction using the flat-event velocities from of migration panels as shown in Figure 5.4-8. To use
step (a). the constant-velocity migration panels for picking rms
(f) Perform velocity analysis at frequent intervals as velocities, they were split into vertical stripes each
needed to derive an optimum velocity field which comprising 100 midpoints. By grouping the constant-
is used for the subsequent migration of the data. velocity stripes with the same center midpoint, we ob-
tain the velocity panels shown in Figure 5.4-9. Just as a
constant-velocity-stack (CVS) panel (Figure 3.2-10) is
used to pick a stacking velocity function, the constant-
velocity-migration panels in Figure 5.4-9 can be used to
Prestack Kirchhoff Migration
pick an rms velocity function at each center CMP loca-
tion. These functions can then be combined to create a
A practical alternative to prestack Stolt migration that migration velocity field as shown in Figure 5.4-10. Com-
involves all of the prestack data at all times during pare with Figure 5.4-6 and note that the migration ve-
the computation is to migrate prestack data using the locity field indeed exhibits a pattern that conforms with
Kirchhoff summation method. The summation of ampli- the subsurface structure. By using the velocity field in
tudes is done along the table-top trajectories inferred by Figure 5.4-10, the prestack time migrated section shown
equation (5-15), and the necessary amplitude and phase in Figure 5.4-11 was obtained. Compare with the post-
treatment of the summed amplitudes are given by the stack time migrated section (Figure 5.4-7) and observe
integral solution to the scalar wave equation (Section that, in this case, the differences are marginal. The pri-
4.1). mary reason for this is that velocities are generally quite
Prestack Kirchhoff migration provides the flexibil- high, making even DMO correction unnecessary.
ity to output selected CMP gathers in their migrated
positions. Such an option is attractive if one wants to
pick rms velocity functions from the selected output
gathers to create a migration velocity field. This is not Velocity Analysis Using
possible with prestack Stolt migration which generates Common-Reflection-Point Gathers
a volume of prestack migrated data. Another advantage
of prestack Kirchhoff migration is its ability to accom- The next field data example demonstrates the use of
modate irregular topography. constant-velocity prestack time migration to generate
Figure 5.4-5 is a CMP-stacked section associated selected CMP gathers in their migrated positions to
with a seismic traverse over an overthrust structure. determine migration velocities. Figure 5.4-12 shows a
The line direction is approximately orthogonal to the CMP stack without DMO correction; the exploration
thrust planes, and thus yields minimal 3-D effects. This objective requires delineation of fault blocks which
data set is quite deceptive — at first, it gives the im- manifest themselves with phantom diffractions. These
pression that time migration is not a valid strategy for diffractions conflict with the gently dipping reflections
imaging the subsurface. Although the reflector geome- associated with the surrounding strata. Poststack time
tries are quite complex, velocities vary from 4000 m/s migration of the CMP stack without DMO correction
at the surface to 6500 m/s at the bottom of the section falls short of providing a crisp image of the fault blocks
with not much refraction occurring at layer boundaries. (Figure 5.4-13).
The stacking velocity field shown in Figure 5.4-6 Dip-moveout correction preserves the phantom
does not appear to be geologically plausable. Never- diffractions and fault-plane reflections on stacked
theless, use it without much editing for poststack time data as shown in Figure 5.4-14. As a result, mi-
migration. The resulting section shown in Figure 5.4-7 gration of the DMO stack (Figure 5.4-15) en-
already exhibits a fairly accurate image of the struc- ables a better delineation of the fault planes as
turally complex zone above 2 s. We observe imbricate compared with the migration of the conventional
structures, folds, and reverse faults that accompany the CMP stack without DMO correction (Figure 5.4-13).
Dip-Moveout Correction and Prestack Migration 789

Table 5-4. Velocity function derived from the constant- Table 5-6. Velocity function derived from the constant-
velocity migration panel of Figure 5.4-16. velocity migration panel of Figure 5.4-18.

Two-way Zero-Offset RMS Velocity Two-way Zero-Offset RMS Velocity


time, ms Picked, m/s time, ms Picked, m/s

0 1500 0 1500
450 1500 375 1500
700 1600 700 1700
900 1750 950 2000
1050 1900 1100 2200
1200 2200 1350 2350
1500 2500 1500 2450
2000 2800 2500 3000
2350 3000
2500 3000
Table 5-7. Velocity function derived from the constant-
velocity migration panel of Figure 5.4-19.
Table 5-5. Velocity function derived from the constant-
velocity migration panel of Figure 5.4-17. Two-way Zero-Offset RMS Velocity
time, ms Picked, m/s
Two-way Zero-Offset RMS Velocity
time, ms Picked, m/s 0 1500
350 1500
0 1500 650 1750
400 1500 750 1800
750 1700 950 2000
900 1800 1250 2500
1050 2000 1750 2700
1350 2300 2500 3000
1500 2600
2050 2900
2500 3000 means that the image below the midpoint location asso-
ciated with that gather is the same irrespective of offset,
and thus migration velocity is correct. Erroneously too
To derive a migration velocity field, we perform low velocity yields a moveout on a gather from prestack
prestack time migration using a range of constant ve- time migration similar to the moveout of an event on a
locities and generate selected gathers at their migrated CMP gather which has been overcorrected. Similarly,
positions. The interval for output gathers is specified in erroneously too high velocity yields a moveout on a
accordance with the degree of lateral velocity variations gather from prestack time migration much like to the
that can be accommodated by time migration. An inter- moveout of an event on a CMP gather which has been
val equal to a cable length is a reasonable rule to follow. undercorrected. Observe flat, over- and undercorrected
Figures 5.4-16 through 5.4-19 show migration velocity events in the velocity panels shown in Figures 5.4-16
panels at four midpoint locations. The way to use these through 5.4-19, and pick for each CMP location rms ve-
panels to pick rms velocity functions at analysis loca- locity functions. Tables 5-4 through 5-7 list the picked
tions is similar to the use of a CMP gather that has rms velocity functions.
been normal-moveout corrected using a range of con- There is one very important aspect of the constant-
stant velocities (Figure 3.2-11). In both cases, a velocity velocity migration panel as in Figure 5.4-19 and the
function is picked based on the criterion of flatness of constant-velocity moveout panel as in Figure 3.2-11.
events. The flatness of an event on a moveout-corrected When picking a moveout velocity function from the lat-
gather means that the moveout velocity associated with ter, you are tracking the same event at the same reflec-
that event is optimum. Similarly, the flatness of an event tion point from gather to gather in the same panel —
on a CMP gather created by prestack time migration overcorrected at low velocities, flat at optimum velocity
(text continues on p. 798)
790 Seismic Data Analysis

FIG. 5.4-12. A portion of a CMP stack from an area with fault blocks. (Data courtesy BHP Petroleum.)

FIG. 5.4-13. Poststack time migration of the CMP stack shown in Figure 5.4-12. Note the undermigration caused by the
2-D migration of 3-D behavior of the fault-plane diffractions and reflections (Chapter 7).
Dip-Moveout Correction and Prestack Migration 791

FIG. 5.4-14. DMO stack associated with the data set as in Figure 5.4-12.

FIG. 5.4-15. Poststack time migration of the DMO stack shown in Figure 5.4-14. Note the undermigration caused by the
2-D migration of 3-D behavior of the fault-plane diffractions and reflections (Chapter 7).
792 Seismic Data Analysis

FIG. 5.4-16. Migration velocity analysis panel for CMP 1161 of the data set as in Figure 5.4-12. The panel is created by
performing prestack time migration using a range of constant velocities and displaying the results at CMP location 1161.
Dip-Moveout Correction and Prestack Migration 793

FIG. 5.4-17. Migration velocity analysis panel for CMP 1561 of the data set as in Figure 5.4-12. The panel is created by
performing prestack time migration using a range of constant velocities and displaying the results at CMP location 1561.
794 Seismic Data Analysis

FIG. 5.4-18. Migration velocity analysis panel for CMP 1961 of the data set as in Figure 5.4-12. The panel is created by
performing prestack time migration using a range of constant velocities and displaying the results at CMP location 1961.
Dip-Moveout Correction and Prestack Migration 795

FIG. 5.4-19. Migration velocity analysis panel for CMP 2361 of the data set as in Figure 5.4-12. The panel is created by
performing prestack time migration using a range of constant velocities and displaying the results at CMP location 2361.
796 Seismic Data Analysis

FIG. 5.4-20. Time migration velocity field computed from the vertical velocity functions picked from the migration velocity
analysis panels as in Figures 5.4-16 through 5.4-19.
Dip-Moveout Correction and Prestack Migration 797
798 Seismic Data Analysis

FIG. 5.4-22. The stack of image gathers as in Figure 5.4-21 derived from prestack time migration. Compare with Figure
6.0-40.

value and undercorrected at high velocities. Whereas, of DMO correction, are one motivation for doing time
when picking an rms velocity function from the former, migration before stack.
you are not tracking the event at the same reflection Limitations in picking reliable velocity functions
point because of the lateral positioning effect of migra- are the same as those associated with constant-velocity
tion. moveout panels (Figure 3.2-11). Shortening of effective
By interpolating between the picked rms velocity cable length at shallow times because of muting, inter-
functions as listed in Tables 5-4 through 5-7, a velocity ference of linear noise and multiples all impose a limit
field for time migration can be created (Figure 5.4-20). on the picking accuracy (Section 6.0).
By using this velocity field, data are migrated before
stack. The common-reflection-point (CRP) gathers in
Figure 5.4-21 exhibit flatness of events — a way to check Focusing Analysis
the accuracy of the migration velocity field. Stacking
the CRP gathers yields the image from prestack time The idea of a velocity analysis that is based on differ-
migration as shown in Figure 5.4-22. Admittedly, there ential solutions of the scalar wave equation first was
is some undermigration caused by the 3-D geometry of introduced by Doherty and Claerbout (1974). They
the fault planes which can be adequately imaged only by used the 15-degree finite-difference migration algorithm
3-D migration (Section 7.0). Nevertheless, compare with and worked with single CMP gathers. Gonzalez-Serrano
migrations of the conventional stack (Figure 5.4-13) and and Claerbout (1979) later extended the wave equa-
DMO stack (Figure 5.4-15), and note that the faults tion velocity analysis to slant-midpoint coordinates and
have been delineated much more distinctively. Such dif- worked with linearly moveout-corrected CMP gathers.
ferences — conflicting dips with different stacking veloc- The method discussed here (Yilmaz and Chambers,
ities associated with fault blocks and salt diapirs where 1984) operates in the Fourier transform domain using
vertical velocity variations may be beyond the accuracy the exact form of the double-square-root (DSR) opera-
Dip-Moveout Correction and Prestack Migration 799

FIG. 5.4-24. Common-offset data derived from a constant-


velocity earth model consisting of six point scatterers be-
neath midpoint 1, where (a) is zero-offset and (b) is far off-
set.

is given by the envelope of the velocity volume of


data P (y, h = 0, τ = t, v).

We now demonstrate the procedure outlined in


Figure 5.4-23 using a synthetic data set. Figure 5.4-
24 shows two common-offset sections over a number
of point scatterers buried in a constant-velocity earth,
FIG. 5.4-23. A flowchart of an algorithm for focusing anal- where v = 3000 m/s. Using a constant velocity for ex-
ysis. trapolation, ve = 3000 m/s, the t − τ image plane was
produced for each midpoint. Two such planes corre-
tor. Mathematical details of this method are found in sponding to midpoints 1 and 5 denoted in Figure 5.4-24
Section E.7. Figure 5.4-23 summarizes the main compu- are shown in Figure 5.4-25. The v−τ planes (Figure 5.4-
tational steps involved in this migration velocity analy- 26) then were generated from the t − τ image planes by
sis based on wavefield extrapolation. the mapping procedure described in Section E.7. Peak
amplitudes for all events occur at the correct medium
(a) Starting with the prestack data in midpoint y, off- velocity (3000 m/s). We expect the diffraction events
set h, and two-way event time t in the unmigrated in Figure 5.4-23 to migrate to the apexes beneath mid-
position, represented by the wavefield P (y, h, τ = point 1, where the point scatterers are located. Note
0, t) at the surface τ = 0, perform 3-D Fourier that in Figure 5.4-25, almost all the energy is in the
transform. The variable τ is associated with the
direction of wave extrapolation, and is related to
depth z by τ = 2z/v, where v is the medium ve-
locity.
(b) Specify an extrapolation velocity function that
only varies vertically, v(τ ) and apply the extrapo-
lation operator exp(−iωDSRτ /2) to compute the
extrapolated wavefield in the transform domain
P (ky , kh , τ, ω) from the surface wavefield in the
transform domain P (ky , kh , τ = 0, ω).
(c) To obtain the zero-offset image, sum over the offset
wavenumber, and thus obtain P (ky , h = 0, τ, ω).
(d) Apply 2-D inverse Fourier transform to obtain the
zero-offset image P (y, h = 0, τ, t). The image below
a midpoint y is contained in the t − τ plane. FIG. 5.4-25. Image planes corresponding to midpoints 1
(e) Perform mapping of the variables as described in and 5 as indicated in Figure 5.4-24, where (a) is CMP 5 and
Section E.6 from τ to v. The velocity information (b) is CMP 1.
800 Seismic Data Analysis

FIG. 5.4-28. Image planes corresponding to midpoints 1


and 5 as indicated in Figure 5.4-27, where (a) is CMP 5 and
(b) is CMP 1.

FIG. 5.4-26. The v−τ planes corresponding to midpoints 1


and 5 derived from the image planes in Figure 5.4-25, where
(a) is CMP 5 and (b) is CMP 1.

FIG. 5.4-27. Common-offset data based on a horizontally


layered earth model containing three point scatterers located
beneath midpoint 1 on the boundaries between constant- FIG. 5.4-29. The v−τ planes corresponding to midpoints 1
velocity layers, where (a) is zero-offset and (b) is far offset. and 5 derived from the image planes in Figure 5.4-28, where
(a) is CMP 5 and (b) is CMP 1.
image plane corresponding to midpoint 1; just five mid-
points away, at midpoint 5, the migrated energy is very
low. which is just the velocity used in generating the model
How do we interpret the t − τ image planes? If we in Figure 5.4-23. Any displacement of peak energy from
used the true medium velocity in downward extrapola- the t = τ image line means that the velocity value
tion, then, according to the imaging principle, we would used for downward extrapolation differs from that of
see all the events along the diagonal τ = t, the image the event. This displacement is also the basis for map-
line, on the image plane. This happens in Figure 5.4- ping from the t − τ image plane to the v − τ plane by
25, because a 3000-m/s extrapolation velocity was used, equation (E-77).
Dip-Moveout Correction and Prestack Migration 801

FIG. 5.4-30. (a) CMP gather at location 1 as indicated


in Figure 5.4-27; (b) and (c) are velocity spectra derived
from this gather by the methods of Figures 5.4-1b and c,
respectively.

This mapping is investigated further with the mod-


eled data set shown in Figure 5.4-27, in which veloc-
ity increases with depth. In Figure 5.4-28b, note that
the top and middle events fall to the left of the im-
age line suggesting that the velocity used in extrapo-
lation (ve = 3000 m/s) is greater than the velocities
associated with these events. The bottom event falls on
the image line, implying that its velocity is nearly the
same as the extrapolation velocity. These observations
are confirmed in the corresponding v − τ planes in Fig-
ure 5.4-29. While true stacking velocity values for the
three events are 2700, 2850, and 3000 m/s, the velocities FIG. 5.4-32. A CMP stacked section. The center portion
interpreted from Figure 5.4-29b are about 2500, 2800, was used in the migration velocity analysis described in Fig-
and 3000 m/s. Thus, the migration-based velocity esti- ures 5.4-33, 5.4-34, and 5.4-35.
mate for the shallow event is in error by approximately
8 percent.
To determine the reason for the velocity error, of the depth-variable velocity model associated with
we will consider a migration-based velocity analysis the constant-offset sections in Figure 5.4-27. The mi-
of our synthetic data example that does not involve gration velocity analysis on this gather (Figure 5.4-
the approximate mapping step. Figure 5.3-30a shows 30b) was done by extrapolating the surface wavefield
a CMP gather from midpoint 1 in the zero-dip region P (kh , ω, τ = 0) repeatedly with different constant ve-
locities in steps of ∆τ = ∆t (the sampling rate). The
zero-offset trace from each attempt was collected after
this effort, abandoning the rest of the migrated CMP
gather.
Interpretation of the velocity analysis in Figure
5.4-30b reveals correct stacking velocities for the three
events, including the shallowest. Clearly the error ob-
served in Figure 5.4-29 is attributable to the mapping
(equation (E-100). Note that the error does not occur
because of depth variability of velocity, but instead,
because the single extrapolation velocity used differed
FIG. 5.4-31. Focusing analysis for determining migration from the medium velocity. The conventional velocity
velocities. See text for details. analysis for midpoint 1 of this model data set is shown
802 Seismic Data Analysis

FIG. 5.4-33. Image planes associated with the center midpoint from the zone of interest in Figure 5.4-32.

in Figure 5.4-30c for comparison. Note the familiar windowed into 1024-ms time gates with 50 percent over-
NMO stretching that is apparent in the shallow event. lap. The image planes for one particular midpoint are
In other respects, both the results (Figures 5.4-30b and shown in Figure 5.4-33. Different extrapolation veloc-
5.4-30c) are comparable. ities picked from a specified regional velocity function
The departure of an event on the t − τ image plane are used in each time gate. The velocity scan used in
from the t = τ image line is measured by the quantity mapping is then carried out within a corridor around
∆τ as depicted in Figure 5.4-31a. In some practical im- this function. Because different extrapolation velocities
plementations, the t − τ image plane is mapped onto are used in successive segments, a given event appears
the plane of ∆τ versus τ as depicted in Figure 5.4-31c at different values of τ in adjacent time segments.
to determine the rms velocity v(τ ) for time migration The resulting velocity analysis for the central mid-
from the extrapolation velocity ve (τ ). An event with a point is shown in Figure 5.4-34. In conventional prac-
velocity error v(τ ) − ve (τ ) is represented by an energy tice, to improve the quality of velocity picks, velocity
maximum either to the left or to the right of the ∆τ = 0 analyses from a number of neighboring CMP gathers
line. The δτ (τ ) trend can be picked and translated into often are summed. Figure 5.4-34c shows the result of
a velocity trend as depicted in Figure 5.4-31b. This type stacking velocity analysis for data from the six adja-
of analysis has come to be called focusing analysis in the cent CMP gathers indicated in Figure 5.4-34a. For the
industry (Faye and Jeannaut, 1986). It has been used in migration-based method, the v − τ planes correspond-
some cases erroneously to estimate and update velocity- ing to these gathers were summed. The result is shown
depth models used for depth migration. The method in Figure 5.4-34b. The most obvious difference between
can only provide plausable velocity update within the the two results is the lack of shallow information in the
framework of time migration. migration-based v − τ plane. This shortcoming is at-
Figure 5.4-32 is a CMP stack from offshore Texas. tributed to spatial aliasing and lack of long-offset data
A 7000-ft portion (64 midpoints each with 48 offset in the shallow time gate. The problem can be eliminated
traces) of the profile was used for migration velocity partly by increasing the length of the time gate used in
analysis. For computational efficiency, the data were the velocity analysis. With the shortcut time-windowing
Dip-Moveout Correction and Prestack Migration 803

FIG. 5.4-34. (a) A portion of the CMP stacked section in Figure 5.4-32, (b) the velocity spectrum based on the procedure
in Figure 5.4-23, followed by averaging over six midpoints; (c) the conventional velocity analysis as discussed in Section 3.2
followed by averaging over six midpoints.

approach described above, the shallowest time segment Fowler’s Velocity-Independent


did not include the large-offset data necessary for veloc- Prestack Migration
ity resolution. Because the events have dip, the derived
migration velocities are lower (by up to 4.5 percent) We now restate the underlying principle for migration
than the velocities derived from the stacking velocity velocity analysis:
analysis.
The velocity analysis described in this section Starting with the prestack volume of data P (y, h, t)
does not handle lateral variations in velocity. It is in midpoint y, offset h and two-way event time t in
based on a Fourier-transform domain formulation with the unmigrated position, create a velocity cube —
only vertically varying velocity used in extrapolation. volume of data P (y, v, τ ) in midpoint y, migration
This method may be particularly efficient for the dip- velocity v and two-way event time τ in the migrated
corrected velocity estimate needed for time migration. position. Within the context of time migration, the
804 Seismic Data Analysis

tion (4-24b) recast as


2 k2
vmig y
ω0 = ωτ2 + , (5 − 46)
4
where the relationship ωτ = vmig kz /2 is used. The out-
put of this mapping P (ky , ωτ ; vmig ) is then scaled by
the quantity S given by equation (4-24c) recast as
vmig ωτ
S= . (5 − 47)
2 2
vmig ky2
2
ωτ +
4
Again, the relationship ωτ = vmig kz /2 is used to obtain
equation (5-47) from equation (4-24c).
Figure 5.4-35 describes a flowchart for creating the
migration velocity volume P (y, vmig , τ ) by Fowler and
FIG. 5.4-35. A flowchart of an algorithm to create a mi-
gration velocity cube. Stolt mapping.

(a) Start with data P (y, h, t) in coordinates of mid-


output time variable τ is related to depth by way point y, offset 2h and event time t in the unmi-
of vertical stretch: τ = 2z/v. grated position, and create constant-velocity stack
(CVS) volume P (y, vstk , tn ) using a range of veloc-
Although the velocity cube can be created by ities vstk , where tn is the event time after constant-
means of some of the migration velocity analysis tech- velocity normal moveout correction.
niques described in this section, a variation of the (b) Apply 2-D Fourier transform to obtain the CVS
method by Fowler (1984) is particularly efficient and cube P (ky , vstk , ω) in the frequency-wavenumber
elegant. First, we review Fowler’s method to create the domain, where ky and ω are the Fourier transform
velocity cube. Refer to the moveout equation (5-1) and variables associated with the variables y and tn .
recall that stacking velocity vstk is dip dependent: (c) Sort the CVS volume P (ky , vstk , ω) into a set of
4h2 constant-velocity sections P (ky , ω; vstk ).
t2 = t20 + 2 , (5 − 43) (d) Perform the Fowler mapping based on equation (5-
vstk
45) on each of the velocity sections so as to obtain
where the DMO velocity volume P (ky , ω0 ; vDM O ).
vDM O (e) Migrate each of the constant-velocity sections of
vstk = . (5 − 44)
cos φ the DMO velocity volume by performing the Stolt
Use equation (5-11) to establish a relationship between mapping based on equations (5-46) and (5-47)
the dip-dependent stacking velocities vstk and the dip- so as to obtain the migration velocity volume
independent DMO velocities vDM O — velocities esti- P (ky , ωτ ; vmig ).
mated from dip-moveout-corrected data: (f) Apply 2-D inverse Fourier transform to obtain the
vDM O migration velocity volume P (y, τ, vmig ).
vstk = . (5 − 45)
2
vDM O ky A variation of Fowler’s sequence described above
1−
2ω0 involves creating the CVS cube directly from DMO-
This equation is the basis for Fowler mapping of corrected data.
vstk to vDM O . Note that the process is applied to data
in the frequency-wavenumber domain. The Fowler map- (a) Start with data P (y, h, t) in coordinates of mid-
ping is then followed by constant-velocity Stolt map- point y, offset 2h and event time t in the unmi-
ping (Sections 4.1 and D.7) to map the DMO velocities grated position, and apply DMO correction fol-
vDM O to migration velocities vmig . lowed by inverse NMO correction.
Stolt migration of the dip-moveout-corrected (b) Create constant-velocity stack (CVS) volume
data volume in the Fourier transform domain P (y, vDM O , t0 ) using a range of velocities vDM O ,
P (ky , ω0 ; vDM O ) involves, first, mapping from ω0 to ωτ where t0 is the event time after constant-velocity
for a specific ky by using the dispersion relation of equa- normal moveout correction (Figure 5.4-36a).
(text continues on p. 815)
Dip-Moveout Correction and Prestack Migration 805

FIG. 5.4-36. (a) A velocity cube computed from DMO-corrected gathers; (b) the same data as in (a) after Stolt migration
of each of the constant-velocity panels.
806 Seismic Data Analysis

FIG. 5.4-37. (a) A time slice from the migration velocity cube shown in Figure 5.4-36b, (b) the same time slice as in (a)
with velocity picking. Red and dark blue curves correspond to primaries and multiples, respectively.
Dip-Moveout Correction and Prestack Migration 807

FIG. 5.4-38. (a) Velocity strands picked from the time slices of the migration velocity cube shown in Figure 5.4-36b as
demonstrated in Figure 5.4-37b; (b) the color-coded velocity field computed by interpolation of the velocity strands shown in
(a).
808 Seismic Data Analysis

FIG. 5.4-39. (a) The prestack time-migrated section extracted from the migration velocity cube of Figure 5.4-36b coincident
with the surface that corresponds to the velocity field in Figure 5.4-38b, (b) the same section as in (a) with the viewing axis
perpendicular to the plane of the paper.
Dip-Moveout Correction and Prestack Migration 809

FIG. 5.4-40. The prestack time-migrated section as in Figure 5.4-39b collapsed onto a 2-D section.
810 Seismic Data Analysis

FIG. 5.4-41. (a) A migration velocity volume, (b) the same volume viewed end-on showing the vertical variation in rms
velocity at a given CMP location.
Dip-Moveout Correction and Prestack Migration 811

FIG. 5.4-42. (a) A time slice from the migration velocity volume shown in Figure 5.4-41a with the interpretation of the rms
velocity as a function of midpoint location, (b) the migration velocity volume as in Figure 5.4-41a with the picked velocity
strands as in (a).
812 Seismic Data Analysis

FIG. 5.4-43. (a) The color-coded optimum rms velocity surface extracted from the migration velocity volume shown in
Figure 5.4-41a, (b) the same view as in Figure 5.4-41b with the intersection of the rms velocity surface showing the vertical
velocity variation at a given CMP location.
Dip-Moveout Correction and Prestack Migration 813

FIG. 5.4-44. (a) The image surface associated with prestack time migration that was extracted from the migration velocity
volume shown in Figure 5.4-41a by sculpting the amplitudes over the rms velocity surface shown in Figure 5.4-43a, (b) the
image surface as in (a) after collapsing it onto a 2-D section associated with prestack time migration.
814 Seismic Data Analysis

FIG. 5.4-45. A close-up view of the prestack time-migrated section as in Figure 5.4-44b.
Dip-Moveout Correction and Prestack Migration 815

(c) Sort the CVS volume P (y, vDM O , t0 ) into a set of view of the volume that represents the plane of velocity
cosntant-velocity stacked sections P (y, t0 ; vDM O ). versus time.
(d) Apply 2-D Fourier transform to obtain the The migration velocity volume is interpreted by
CVS cube P (ky , vDM O , ω0 ) in the frequency- picking the primary velocity trend from selected time
wavenumber domain, where ky and ω are the slices as shown in Figure 5.4-42a. Note the lateral vari-
Fourier transform variables associated with the ation in velocities which is captured by continuous pick-
variables y and t0 . ing along the midpoint axis. By interpolating the veloc-
(e) Sort the CVS volume P (ky , vDM O , ω0 ) into a set of ity strands resulting from the interpretation of selected
constant-velocity sections P (ky , ω0 ; vDM O ). time slices (Figure 5.4-42a), an rms velocity surface is
(f) Migrate each of the constant-velocity sections of generated. The picked velocity strands are shown in Fig-
the DMO velocity volume by performing the Stolt ure 5.4-42b embedded within the migration velocity vol-
mapping based on equations (5-46) and (5-47) ume, and the rms velocity field is shown in Figure 5.4-
so as to obtain the migration velocity volume 43a as a color-coded surface extracted from within the
P (ky , ωτ ; vmig ). migration velocity volume. A quality control of the rms
(g) Apply 2-D inverse Fourier transform to obtain velocity surface can be made by intersecting it with the
the migration velocity volume P (y, τ, vmig ) (Figure cross-sections of the migration velocity volume at se-
5.4-36b). lected CMP locations as shown in Figure 5.4-43b.
The prestack time-migrated section is a by-product
The migration velocity volume P (y, τ, vmig ) shown of the migration velocity analysis described here. Specif-
in Figure 5.4-36b can be visualized and interpreted to ically, the image surface associated with the prestack
derive a migration velocity field. For spatial consistency, time migration is obtained by sculpting the amplitudes
velocity picking should be done on time slices from the from within the migration velocity volume over the rms
migration velocity volume as shown in Figure 5.4-37. velocity surface as shown in Figure 5.4-44a. The conven-
The resulting velocity strands shown in Figure 5.4-38a tional 2-D display of the prestack time-migrated section
are interpolated to create the migration velocity field is then created by simply collapsing the sculpted image
shown in Figure 5.4-38b. This velocity field then can be surface onto a 2-D plane (Figure 5.4-44b). A close-up
used to extract from the volume the section that cor- view of the prestack time-migrated section shows accu-
responds to prestack time migration as shown in Fig- rate imaging of the steeply dipping event (Figure 5.4-
ure 5.4-39. An enlarged view of this section is shown in 45).
Figure 5.4-40. Note the excellent imaging of the steeply
dipping fault planes which conflict with the gently dip-
ping strata.
EXERCISES
Alternatively, the plane of (vmig , τ ) for each mid-
point y can be inverse transformed to the plane of (h, τ )
associated with the common-reflection-point gather de- Exercise 5-1. Consider the application of DMO
rived from prestack time migration. This is then fol- correction to data referred to a floating datum repre-
lowed by conventional normal-moveout correction and sented by a smooth form of an irregular topographic
stacking. The resulting section, again, represents the surface and data referenced to a flat datum below.
image from prestack time migration. Which DMO correction would have more effect on the
The data example shown in Figure 5.4-36 demon- data?
strates the use of the migration velocity volume in Exercise 5-2. Refer to Fowler’s velocity-
deriving a high-fidelity image of fault planes from independent prestack migration technique for migration
prestack time migration (Figure 5.4-40). The data ex- velocity analysis described in Section 5.4. Suppose you
ample shown in Figure 5.4-41 demonstrates the use of have transformed the prestack data from offset to ve-
the migration velocity volume in imaging steeply dip- locity space using 30 constant velocity values from 2000
ping event. The migration velocity volume was created m/s to 4900 m/s using an increment of 100 m/s. Can
using the procedure described above. Specifically, the you create additional constant-velocity panels such that
DMO-corrected CMP gathers were first NMO-corrected the increment is 50 m/s by poststack migration rather
using a range of constant velocities and a CVS volume than prestack migration or trace interpolation? If so,
was created. Next, each constant-velocity stacked sec- what velocity would you use for poststack migration to
tion was migrated using the Stolt method and the con- create the panel for 3050-m/s velocity.
stant velocity associated with that section. The result- Exercise 5-3. Derive Bancroft’s equivalent off-
ing migration velocity volume is shown in Figure 5.4-41. set equation (E-72b) from the nonzero-offset traveltime
Note the vertical variation in velocities on the end-on equation (E-67).
816 Seismic Data Analysis

Exercise 5-4. Suppose you have two events with


conflicting dips of the same magnitude, but in oppo-
site directions, associated with a reflector within a fault
block and the fault plane itself. Can these two events be
distinguished on a velocity spectrum computed from a
CMP gather before DMO correction at a location just
above the surface point where the two events intersect
one another?
Appendix E
TOPICS IN DIP-MOVEOUT CORRECTION AND PRESTACK TIME MIGRATION

E.1 Reflection Point Dispersal

We start the analysis using the geometry of Figure E-1. Place the source S at the origin (0, 0),
and denote the coordinates of the points of interest as N : (x0 = SK, z0 = KN ), H : (x1 =
SB, z1 = BH), F : (x2 = SA, z2 = AF ), and R : (x3 = SQ, z3 = QR). Our objective is
to compute the coordinates for the normal-incidence reflection point N : (x0 , z0 ) associated
with midpoint M , and the reflection point R : (x3 , z3 ) associated with the nonzero-offset source
receiver separation x = SG.
First, we compute the coordinates for N : (x0 , z0 ). From the geometry of Figure E-1, first,
note that x0 = SK and z0 = KN . Then, from the triangle M KN , we have
MK
= cos α,
MN
which, by way of M K = x0 − x/2, yields
x0 − x/2
MN = . (E − 1a)
cos α
Similarly, from the same triangle M KN , we have
KN
= cos φ,
MN
which, by way of KN = z0 , yields
z0
MN = . (E − 1b)
cos φ
Finally, combine equations (E-1a) and (E-1b) to get a relation between x0 and z0
x0 − x/2 z0
= . (E − 2a)
cos α cos φ
Next, write the equation of the line that represents the dipping reflector in terms of the direc-
tional cosines and the normal d = SC to the line from the origin S
x0 cos α + z0 cos φ = d. (E − 2b)
Solve for z0 from equation (E-2a) and substitute into equation (E-2b)
x cos2 φ
x0 cos α + x0 − = d. (E − 3)
2 cos α
Note that the directional cosines satisfy the relation cos2 α + cos2 φ = 1. By making use of this
relation, simplify equation (E-3) to get the final expression for the coordinate x0
x
x0 = d cos α + cos2 φ. (E − 4a)
2
Finally, substitute x0 back into equation (E-2a) to get the expression for the coordinate z0
x
z0 = d cos φ − cos α cos φ. (E − 4b)
2
818 Seismic Data Analysis

FIG. E-1. Geometry of a dipping reflector to derive the expression for reflection-point dispersal (Section
E.1).

The distance d = SC is given by


d = SU + U C,
which can be rewritten using the geometry of Figure E-1 as follows:
d = SM cos α + M N.
Substitute the definitions SM = x/2 and D = M N , where D is the distance along the normal-
incident raypath from the point N on the reflector to the midpoint location M , to obtain the
relation between d and D
x
d = D + cos α. (E − 5)
2
Make the further substitution of equation (E-5) into equations (E-4a) and (E-4b), and recall
that cos2 α + cos2 φ = 1 to derive the expressions for the coordinates of N : (x0 , z0 ) in terms of
D
x
x0 = D cos α + (E − 6a)
2
and
z0 = D cos φ. (E − 6b)
To derive the expressions for the coordinates of R : (x3 , z3 ), we note that R is the intersection
of the line defined by the reflecting interface, which yields the relation
x3 cos α + z3 cos φ = d, (E − 7a)
and the line SRF , which yields the relation
x3 z3
= , (E − 7b)
x2 z2
where (x2 , z2 ) are the coordinates of F . From the geometry of Figure E-1, first, note that
x2 = SA, z2 = AF , and x3 = SQ and z3 = QR.
Dip-Moveout Correction and Prestack Migration 819

Solve equation (E-7b) for x3 and substitute into equation (E-7a) to get
z2
z3 = d. (E − 8)
x2 cos α + z2 cos φ
The coordinates for F : (x2 , z2 ) are derived in Section C.3 (equations C-18c,d)
x2 = 2(d − x cos α) cos α + x (E − 9a)
and
z2 = 2(d − x cos α) cos φ. (E − 9b)
Substitute equations (E-9a) and (E-9b) into equation (E-8), and replace d via equation
(E-5) to obtain, after some involved algebra, the expression for z3
x2
z3 = D cos φ − cos2 α cos φ. (E − 10a)
4D
Back substitution of equation (E-10a) into equation (E-7a) then yields the expression for x3
x x2
x3 = D cos α + + sin2 α cos α. (E − 10b)
2 4D
Knowing the coordinates of N : (x0 , z0 ) and R : (x3 , z3 ), the distance ∆ = N R can now be
computed
∆= (x3 − x0 )2 + (z3 − z0 )2 . (E − 11a)
Substitute equations (E-10a) and (E-10b) for the coordinates of R : (x3 , z3 ), and equations
(E-6a) and (E-6b) for the coordinates of N : (x0 , z0 ) into equation (E-11a) and carry out the
algebraic steps to get the final expression
x2
∆= sin α cos α. (E − 11b)
4D
Equation (E-11b) represents the reflection point smear, otherwise known as reflection point
dispersal (Deregowski, 1981) — the distance between the normal-incidence reflection point N
associated with the zero-offset raypath from midpoint M , and the reflection point R associated
with the nonzero-offset raypath for a source-receiver pair separated by an offset x. For the case
of zero-offset x = 0, reflection point smear ∆ vanishes.
Define the two-way zero-offset time t0 at midpoint M as the time associated with the
normal-incidence raypath M N by the relation vt0 /2 = D. Then use this relation in equation
(E-11b) to obtain the expression for reflection point dispersal ∆ in terms of t0
x2
∆= sin α cos α. (E − 12)
2vt0
This equation is applicable to the general case of three-dimensional geometry (Figure 3.1-14),
where α is the angle between the normal to the dipping reflector surface and the direction of
the profile line (Levin, 1971).
For the 2-D geometry of the dipping reflector (Figure E-1, note that
sin α = cos φ, (E − 13)
where φ is the dip angle of the reflector. Hence, equation (E-12) is written in terms of the
reflector dip angle φ as follows:
x2
∆= cos φ sin φ. (E − 14)
2vt0
Equation (E-14), by way of equation (5-6), is the same as equation (5-10) of the main text with
offset x = 2h.
820 Seismic Data Analysis

E.2 Equations for DMO Correction

In this section, we shall derive the traveltime equation for dip-moveout (DMO) correction by
using the geometry of Figure 5.1-1. Start with equation (C-22b) of Section C.3 which defines the
traveltime t from source location S to the reflection point R to the receiver location G, written
in prestack data coordinates
4h2 cos2 φ
t2 = t20 + , (E − 15a)
v2
where 2h is the offset, v is the medium velocity above the reflector, φ is the reflector dip, and
t0 is the two-way zero-offset time at midpoint location yn . Dip-moveout correction is preceded
by normal-moveout (NMO) correction using the dip-independent velocity v
4h2
t2 = t2n + , (E − 15b)
v2
where tn is the event time after the NMO correction. To relate event time tn after the NMO
correction and event time t0 , first, write equation (E-15a) as
4h2 4h2 sin2 φ
t2 = t20 + − . (E − 16)
v2 v2
Set the right-hand sides of equations (E-15b) and (E-16) equal and simplify the result to obtain
the dip-dependent moveout equation
4h2 sin2 φ
t2n = t20 − . (E − 17)
v2
Equations (E-15b) and (E-17) suggest that moveout correction can, in principle, be per-
formed in two steps:

(a) Apply a dip-independent moveout correction using equation (E-15b) to map event time t
to event time tn .
(b) Apply a dip-dependent moveout correction using equation (E-17) to map event time tn to
event time t0 .

This two-step moveout correction is equivalent to the one-step moveout correction using equation
(E-15a) to map event time t directly to event time t0 .
Our goal, however, is to map event time t to τ0 — two-way zero-offset time, not at midpoint
location yn associated with the source-receiver pair S −G, but at midpoint location y0 associated
with the normal-incidence reflection point R. This mapping requires coordinate transformation
of moveout-corrected prestack data Pn (yn , tn ; h) to P0 (y0 , τ0 ; h). Therefore, we need to compute
the coordinates (y0 , τ0 ) in terms of the coordinates (yn , tn ).
From the geometry of Figure 5.1-1, note that

y0 = y n − , (E − 18)
cos φ
where ∆ = N R is the distance along the reflector between the normal-incidence reflection point
N associated with the zero-offset raypath from midpoint yn and the reflection point R associated
with the nonzero-offset raypath for a source-receiver pair separated by an offset 2h.
Adapt equation (E-14) of Section E.1 for the prestack data coordinates x = 2h
2h2
∆= cos φ sin φ, (E − 19)
vt0
and substitute into equation (E-18) to obtain
h2 2 sin φ
y0 = y n − . (E − 20)
t0 v
Dip-Moveout Correction and Prestack Migration 821

Solve equation (E-17) for t0 and write the result in the following form
2
h2 2 sin φ
t0 = tn 1+ (E − 21)
t2n v
or
t0 = tn A, (E − 22)
where
2
h2 2 sin φ
A= 1+ . (E − 23)
t2n v
Now, substitute equation (E-22) into equation (E-20) to get the final expression for y0
h2 2 sin φ
y0 = yn − . (E − 24)
tn A v
Also from the geometry of Figure 5.1-1, note that
2∆ tan φ
τ0 = t 0 − . (E − 25)
v
Substitute equation (E-19) for ∆ and carry out the required algebra to get the desired expression
for τ0 in terms of t0
2
h2 2 sin φ
τ0 = t0 − . (E − 26)
t0 v
Now, substitute equation (E-22) into equation (E-26)
2
h2 2 sin φ
τ0 = t n A − . (E − 27)
tn A v
Finally, simplify equation (E-27) by way of equation (E-23) to obtain the desired expression
for τ0 in terms of tn
tn
τ0 = . (E − 28)
A
We now remind ourselves of our objective to transform the normal-moveout-corrected
prestack data Pn (yn , tn ; h) from yn − tn coordinates to y0 − τ0 coordinates so as to obtain the
dip-moveout-corrected data P0 (y0 , τ0 ; h). The transformation is done using equations (E-24) and
(E-28) which relate the input data coordinates yn − tn to output data coordinates y0 − τ0 . Note,
however, equations (E-24) and (E-28) require knowledge of the reflector dip φ. To circumvent
this requirement, we use the relation from Section D.1
vky
sin φ = , (E − 29)
2ω0
which states that the reflector dip φ can be expressed in terms of wavenumber ky and frequency
ω0 , which are the Fourier duals of midpoint y0 and event time τ0 , respectively. The variables y0
and τ0 defined by equations (E-24) and (E-28), respectively, are associated with the zero-offset
section after DMO correction. By way of equation (E-29), these equations are recast explicitly
independent of reflector dip as
h2 ky
y0 = y n − (E − 30)
tn Aω0
and
tn
τ0 = , (E − 31)
A
822 Seismic Data Analysis

where A of equation (E-23), by way of equation (E-29), is of the form

h2 ky2
A= 1+ . (E − 32)
t2n ω02
Since we have switched to the Fourier transform domain in our analysis, our objective now
is to compute the dip-moveout-corrected data P0 (ky , ω0 ; h) in the transform domain. Thus, start
with the 2-D Fourier transform of P0 (y0 , τ0 ; h)

P0 (ky , ω0 ; h) = P0 (y0 , τ0 ; h) exp(iky y0 − iω0 τ0 ) dy0 dτ0 . (E − 33)

A wavefield is invariant under a coordinate transformation; hence, P0 (y0 , τ0 ; h) =


Pn (yn , tn ; h). Use the transform relations given by equations (E-30) and (E-31) and the in-
variance relation to write equation (E-33) in terms of the normal-moveout-corrected data
Pn (yn , tn ; h) (Liner, 1990)
∂y0 ∂τ0 h2 ky ω 0 tn
P0 (ky , ω0 ; h) = Pn (yn , tn ; h) exp iky yn − −i dyn dtn . (E − 34)
∂yn ∂tn tn Aω0 A
From equation (E-30) it follows that
∂y0
= 1. (E − 35)
∂yn
Square both sides of equation (E-31), substitute equation (E-32) for the quantity A and simplify
to get
t4n ω02
τ02 = . (E − 36a)
t2n ω02 + h2 ky2
Next, differentiate the result
∂τ0 4t3n ω02 (t2n ω02 + h2 ky2 ) − 2t5n ω04
= . (E − 36b)
∂tn 2τ0 (t2n ω02 + h2 ky2 )2
Rearrange the terms in equation (E-32) to the form
t2n ω02 + h2 ky2 = t2n ω02 A2 , (E − 36c)
and use in equation (E-36b) together with equation (E-31) for the τ0 in the denominator to
obtain
∂τ0 2A2 − 1
= . (E − 37)
∂tn A3
Finally, substitute equations (E-35) and (E-37) into equation (E-34)
2A2 − 1 h2 ky ω 0 tn
P0 (ky , ω0 ; h) = Pn (y n , tn ; h) exp iky y n − −i dyn dtn . (E − 38)
A3 tn Aω0 A
Now, we turn our attention to the phase term in equation (E-38)
h2 ky ω 0 tn
Φ = ky yn − − .
tn Aω0 A
Rearrange the terms, first, as follows:
ω 0 tn h2 ky2
Φ = ky yn − 1+ 2 2 ,
A tn ω 0
then, by way of the expression for A as in equation (E-32)
Φ = ky yn − ω0 tn A. (E − 39)
Dip-Moveout Correction and Prestack Migration 823

Return to equation (E-38) and substitute equation (E-39) for the terms in the exponential
2A2 − 1
P0 (ky , ω0 ; h) = Pn (yn , tn ; h) exp(iky yn − iω0 tn A) dyn dtn . (E − 40)
A3
Use the Fourier transform integral

Pn (ky , tn ; h) = Pn (yn , tn ; h) exp(iky yn ) dyn , (E − 41)

to rewrite equation (E-40) in the form


2A2 − 1
P0 (ky , ω0 ; h) = Pn (ky , tn ; h) exp(−iω0 tn A) dtn . (E − 42)
A3
Equation (E-42) describes the dip-moveout correction process which transforms the normal-
moveout-corrected prestack data with a specific offset 2h from yn − tn domain to y0 − τ0 domain.
Referring to Figure 5.1-1, we remind ourselves that yn − tn coordinates are associated with
event time tn at midpoint location yn , and y0 − τ0 coordinates are associated with event time
τ0 at midpoint location y0 that corresponds to the normal-incidence reflection. The lateral
excursion applied by dip-moveout correction is ∆yDM O = |yn − y0 | and the vertical excursion
is ∆tDM O = tn − τ0 .
Once dip-moveout correction is applied, the data are inverse Fourier transformed

P0 (y0 , τ0 ; h) = P0 (ky , ω0 ; h) exp(−iky y0 + iω0 τ0 ) dky dω0 . (E − 43)

A flowchart of the dip-moveout correction in the frequency-wavenumber domain described above


is presented in Figure 5.1-2.

E.3 Log-Stretch DMO Correction

A computationally efficient DMO correction can be formulated in the logarithmic time domain
(Bolondi et al., 1982; ; Bale and Jacubowicz, 1987; Notfors and Godfrey, 1987; Liner, 1990; Zhou
et al., 1996). The log-stretch time variable enables linearization of the coordinate transform
equation (E-28), and as a result, the DMO correction is achieved by a simple multiplication of
the input data with a phase-shift operator in the Fourier transform domain.
Define the following logarithmic variables that correspond to the time variables τ0 and tn
of equation (E-28):
T0 = ln τ0 (E − 44a)
and
Tn = ln tn . (E − 44b)
Hence, the inverse relationships are given by
τ0 = eT0 (E − 45a)
and
tn = eTn . (E − 45b)
In Section E.2, we derived the equations for DMO correction to transform the normal-
moveout-corrected data from (yn , tn ) ccordinates to (y0 , τ0 ) coordinates. The objective here is
to derive equations for DMO correction in the log-stretch coordinates (y0 , T0 ).
Square both sides of equation (E-28)
t2n
τ02 = , (E − 46a)
A2
824 Seismic Data Analysis

and subsitute equation (E-23) for A


t2n
τ02 = 2. (E − 46b)
h2 2 sin φ
1+ 2
tn v
Here, we have to make the crucial assumption that A is close to unity. Refer to equation
(E-23) and note that such an assumption implies one of the following: small h (offset), small φ
(dip), large tn (time after moveout) or large v (velocity). As a result, equation (E-46b) can be
approximated as
2
h2 2 sin φ
τ02 = t2n 1 − . (E − 46c)
t2n v
The slope dτ0 /dy0 measured on a zero-offset section in (y0 , τ0 ) coordinates is given by
dτ0 2 sin φ
= . (E − 46d)
dy0 v
Substitute equation (E-46d) into equation (E-46c)
2
h2 dτ0
τ02 = t2n 1 − . (E − 46e)
t2n dy0
Apply the chain rule for differentiation
dτ0 dτ0 dT0
= (E − 47a)
dy0 dT0 dy0
and, by way of equation (E-45a), obtain the relationship
dτ0 dT0
= eT0 . (E − 47b)
dy0 dy0
Next, combine equations (E-29) and (E-46d) to get
dτ0 ky
= . (E − 47c)
dy0 ω0
By analogy, we may write
dT0 ky
= , (E − 47d)
dy0 Ω0
where Ω0 is the Fourier transform variable associated with the log-stretch time variable T0 .
Finally, substitute equation (E-47d) into equation (E-47b) to get
dτ0 ky
= eT0 . (E − 47e)
dy0 Ω0
Now, return to equation (E-46e), and use equations (E-45a,b) and (E-47e) to obtain the
expression
h2 ky2
e2(T0 −Tn ) 1 + = 1. (E − 48)
Ω20
Take the logarithm of both sides and simplify to get the transform relation between the input
log-stretch time variable Tn and the output log-stretch time variable T0
T0 = Tn − ln Ae , (E − 49)
where
h2 ky2
Ae = 1+ . (E − 50)
Ω20
Dip-Moveout Correction and Prestack Migration 825

We now turn our attention to equation (E-24) and derive the corresponding equation in the
log-stretch domain. First, substitute equation (E-23) into equation (E-24) and approximate the
quantity A as before
2
h2 1 h2 2 sin φ 2 sin φ
y0 = yn − 1− . (E − 51a)
tn 2 t2n v v
Substitute equation (E-46d) into equation (E-51a)
2
h2 1 h2 dτ0 dτ0
y0 = yn − 1− . (E − 51b)
tn 2 t2n dy0 dy0
Then, make the substitutions from equations (E-45b) and (E-47e), drop the high-order term,
and simplify to get
ky
y0 = yn − eT0 −Tn h2 . (E − 51c)
Ω0
Note from equation (E-48)
1
eT0 −Tn = . (E − 52)
h2 ky2
1+ 2
Ω0
Use this result in equation (E-51c) accompanied with the definition of equation (E-50) to obtain
the final expression that relates y0 and yn in log-stretch coordinates
h2 ky
y0 = yn − . (E − 53)
Ae Ω0
Equations (E-49), (E-50), and (E-53) correspond to equations (E-31), (E-32), and (E-30) in the
log-stretch domain.
We remind ourselves of the objective to compute the dip-moveout-corrected data
P0 (ky , Ω0 ; h) in the transform domain. Thus, start with the 2-D Fourier transform of P0 (y0 , T0 ; h)

P0 (ky , Ω0 ; h) = P0 (y0 , T0 ; h) exp(iky y0 − iΩ0 T0 ) dy0 dT0 . (E − 54)

A wavefield is invariant under a coordinate transformation; hence, P0 (y0 , T0 ; h) =


Pn (yn , Tn ; h), the latter being the normal-moveout-corrected data in the log-stretch domain.
Use the transform relations given by equations (E-49) and (E-53) and the invariance relation
to write equation (E-54) in terms of the normal-moveout-corrected data Pn (yn , Tn ; h) (Liner,
1990)
h2 ky ∂y0 ∂T0
P0 (ky , Ω0 ; h) = Pn (yn , Tn ; h) exp iky yn − − iΩ0 (Tn − ln Ae ) dyn dTn .
Ae Ω0 ∂yn ∂Tn
(E − 55)
From equation (E-53) it follows that
∂y0
= 1, (E − 56a)
∂yn
and from equation (E-49), we have
∂T0
= 1. (E − 56b)
∂Tn
Substitute equations (E-56a,b) into equation (E-55) and rearrange the terms in the exponential
to obtain
h2 ky
P0 (ky , Ω0 ; h) = exp −iky + iΩ0 ln Ae Pn (yn , Tn ; h) exp(iky yn − iΩ0 Tn ) dyn dTn .
Ae Ω0
(E − 57)
826 Seismic Data Analysis

Assume that the Fourier transform variable Ω0 in the log-stretch domain is independent of Tn .
Then, the double integral on the right-hand side represents the 2-D Fourier transform of the
normal-moveout-corrected data in the log-stretch domain as

Pn (ky , Ω0 ; h) = Pn (yn , Tn ; h) exp(iky yn − iΩ0 Tn ) dyn dTn . (E − 58)

Therefore, by way of equation (E-58), equation (E-57) takes the form


h2 ky2
P0 (ky , Ω0 ; h) = exp −i + iΩ0 ln Ae Pn (ky , Ω0 ; h). (E − 59)
Ae Ω0
Equation (E-59) describes the dip-moveout correction process which transforms the normal-
moveout-corrected prestack data with a specific offset 2h from the log-stretch yn − Tn domain to
y0 − T0 domain. Note that the relationship of input Pn (ky , Ω0 ; h) to output P0 (ky , Ω0 ; h) given
by equation (E-59) computationally is much simpler than that of equation (E-42). Once the
phase-shift in the log-stretch domain given by equation (E-59) is applied to the input and the
result is inverse Fourier transformed, the data are unstretched from (y0 , T0 ) coordinates back to
(y0 , τ0 ) coordinates using the relationship expressed by equation (E-45a).
A variation of the phase-shift term in equation (E-59) is given by Notfors and Godfrey
(1987). As in most log-stretch formulations of DMO correction, this reference assumes that under
DMO correction the midpoint variable is invariant; hence, by way of equation (E-53), the first
term in the exponential of equation (E-59) drops out. A further approximation, ln Ae = Ae − 1,
and use of the definition for Ae given by equation (E-50) then lead to the following expression
for DMO correction:
h2 ky2
P0 (ky , Ω0 ; h) = exp iΩ0 1+ −1 Pn (ky , Ω0 ; h). (E − 60)
Ω20
Various implementations of the log-stretch technique applied to shot records are described
by Biondi and Ronen (1987), Cobrera and Levy (1989), and Zhou et al. (1996). The latter also
includes a method in double-log-stretch domain, which involves stretching both in time and
midpoint coordinates. See Section 5.1 for a practical implementation of the log-stretch DMO
correction described here.

E.4 The DMO Ellipse

The objective is to derive the traveltime trajectory associated with the dip-moveout correction
operator by using the method of stationary phase. Insert from equation (E-42) the expression
for P0 (ky , τ0 ; h) into equation (E-43)
2A2 − 1
P0 (y0 , τ0 ; h) = Pn (ky , tn ; h) exp(−iω0 tn A − iky y0 + iω0 τ0 ) dtn dky dω0 ,
A3
(E − 61)
where the total phase is given by

Φ = −ω0 tn A − ky y0 + ω0 τ0 . (E − 62a)

Substitute equation (E-32) for A

Φ = − ω02 t2n + h2 ky2 − ky y0 + ω0 τ0 . (E − 62b)

The main contribution to the integration in equation (E-61) occurs when the phase in
equation (E-62b) stays nearly constant. We therefore determine the variation of Φ with respect
Dip-Moveout Correction and Prestack Migration 827

to variables ω0 and ky
∂Φ ω0 t2n
=− + τ0 (E − 63a)
∂ω0 ω02 t2n + h2 ky2

and
∂Φ ky h2
=− − y0 . (E − 63b)
∂ky ω02 t2n + h2 ky2

Then set each variation to zero. Rearranging the terms of the resulting expressions, we have
ω 0 tn τ0
= (E − 64a)
ω02 t2n + h2 ky2 tn

and
ky h y0
=− . (E − 64b)
ω02 t2n + h2 ky2 h

Then, square both sides of equations (E-64a,b)


ω02 t2n τ02
= (E − 65a)
ω02 t2n + h2 ky2 t2n
and
ky2 h2 y02
= . (E − 65b)
ω02 t2n + h2 ky2 h2
Finally, sum equations (E-65a,b) to obtain
y02 τ2
2
+ 20 = 1. (E − 66)
h tn
Equation (E-66) describes an ellipse with the following properties:

(a) Semi-major axis in midpoint y0 direction: a = h.


(b) Semi-minor axis in time τ0 direction: b = tn .

The ellipse of equation (E-66) in the y0 −τ0 plane describes the impulse response of a dip-moveout
operator applied to a common-offset section with offset 2h.

E.5 Nonzero-Offset Traveltime Equation

Prestack wave extrapolation is performed using the double-square operator which enables down-
ward continuation of common-shot and common-receiver gathers (Section D.1). Stationary phase
approximation to the double-square root operator yields the nonzero-offset traveltime equation
(D-35) derived in Section D.2
vt = (y + h)2 + z 2 + (y − h)2 + z 2 , (E − 67)
where y, h, t, z are midpoint, offset, two-way traveltime and depth coordinates and v is the
medium velocity. Here, we shall show that this equation represents an ellipse (Figure E-2) in the
y −z plane for a constant t, h, and v, and derive the parameters of this ellipse. The nonzero-offset
two-way time is associated with the raypath from source S to reflection point R to receiver G
as sketched in Figure E-2. The origin of the y − z plane coincides with midpoint M .
828 Seismic Data Analysis

FIG. E-2. The prestack time migration ellipse. See Section E.5 for details.

Square both sides of equation (E-67) to get

v 2 t2 = 2 (y + h)2 + z 2 (y − h)2 + z 2
(E − 68)
+ (y + h)2 + z 2 + (y − h)2 + z 2 .
Combine the second and third terms on the right-hand side and simplify the terms inside the
square root
v 2 t2 = 2 (y 2 − h2 )2 + 2z 2 (y 2 + h2 ) + z 4 + 2(y 2 + h2 + z 2 ). (E − 69a)
Perform further algebraic manipulation to collect the terms in y and z
v 4 t4
− v 2 t2 h 2 .
(v 2 t2 − 4h2 )y 2 + v 2 t2 z 2 = (E − 69b)
4
Finally, normalize by the terms on the right-hand side and rearrange the terms in the denomi-
nators
y2 z2
2
+ = 1. (E − 70)
(vt/2) (vt/2)2 − h2
Equation (E-70) represents an ellipse in the y − z plane for a constant t with the following
parameters (Figure E-2):

(a) Semi-major axis in midpoint y direction: a = vt/2.


(b) Semi-minor axis in depth z direction:√b = (vt/2)2 − h2 .
(c) Distance from center to either focus: a2 − b2 = h.
(d) Distance from one focus to a point on the ellipse to the other focus: vt.

The ellipse of equation (E-70) in the y−z plane describes the impulse response of a nonzero-offset
migration operator applied to prestack data.
When equation (E-70) is specialized to the zero-offset case, h = 0, we get
y2 z2
2
+ = 1, (E − 71a)
(vt/2) (vt/2)2
Dip-Moveout Correction and Prestack Migration 829

which describes a circle in the y − z plane for a constant t with a radius vt/2. This circle
represents the impulse response of a zero-offset migration operator applied to poststack data.
When equation (E-67) is specialized to the zero-offset case, h = 0, we get
vt = 2 y 2 + z 2 , (E − 71b)
which describes the well-known diffraction hyperbola in the y − t plane for a constant z.
Reduction of the double-square-root equation (E-67) to a single-square-root equation (E-
71b) can also be achieved by defining an equivalent offset he such that (Bancroft et al., 1998)
(y + h)2 + z 2 + (y − h)2 + z 2 = 2 h2e + z 2 . (E − 72a)
Solving for he , we obtain (Bancroft et al., 1998; Margrave et al., 1999)
4y 2 h2
h2e = y 2 + h2 − , (E − 72b)
v 2 t2
where t is the two-way nonzero-offset traveltime of equation (E-67).
Poststack time migration can be conceptualized either by way of a semicircle superposition
using equation (E-72a) or a diffraction summation along the hyperbolic traveltime trajectory
using equation (E-72b). Similarly, prestack time migration can be conceptualized either by way of
semi-elliptical superposition using equation (E-70) or diffraction summation over the traveltime
surface described by equation (E-67). The traveltime surface is known as Cheops’ pyramid
(Claerbout, 1985) and is illustrated in Figure E-3a. The result of summation of amplitudes over
the pyramidal surface is placed at its apex. The question that is of practical importance is how
to define the summation paths over this surface.
Four possible choices of summation paths over the pyramidal surface of equation (E-67) to
perform prestack time migration are:

(a) Summation curves of constant offset: Consider a set of vertical cross-sections of the trav-
eltime pyramid illustrated in Figure E-3a parallel to the midpoint axis as shown in Figure
E-3b. Sum the amplitudes along each of the constant-offset table-top traveltime curves,
independently, and place the result for each at the apex of the summation curve. The sum-
mation collapses the pyramidal surface onto a hyperbolic traveltime curve, which is formed
by combining the apex points of the constant-offset curves. This hyperbolic traveltime curve
is orthogonal to the constant-offset summation curves.
(b) Summation curves of constant time: Consider a set of horizontal cross-sections of the trav-
eltime pyramid as shown in Figure E-3c (Bancroft and Geiger, 1994; Bancroft et al., 1997).
Sum the amplitudes along each of the constant-time curves, independently, and place the
result for each at the maximum offset on the summation curve. The summation, again,
collapses the pyramidal surface onto the hyperbolic traveltime curve, which is formed by
combining the maximum-offset points of the constant-time curves. This hyperbolic travel-
time curve is orthogonal to the constant-time summation curves. The event associated with
the resulting hyperbolic moveout trajectory is on the he −t plane of a common-scatter-point
(CSP) gather (Bancroft et al., 1998), where the equivalent offset he is given by equation
(E-72b). The scatter point corresponds to the apex of the traveltime pyramid A0 in Figure
E-3c.
(c) Summation curves of constant shot: Consider a set of vertical cross-sections of the traveltime
pyramid as shown in Figure E-3d (Berryhill, 1996). Sum the amplitudes along each of
these constant-shot curves, independently, and place the result for each at the apex of
the summation curve. The summation collapses the pyramidal surface onto the hyperbolic
traveltime curve, which is formed by combining the apex points of the constant-shot curves.
This hyperbolic traveltime curve is orthogonal to the constant-shot summation curves and
is on the common-receiver plane that passes through the apex of the pyramid itself.
(d) Summation curves of constant angle: Consider a set of slanted cross-sections of the travel-
time pyramid as shown in Figure E-3e (Ottolini, 1982). These slanted traveltime curves are
associated with constant angle of incidence (Section 6.3). Sum the amplitudes along each
830 Seismic Data Analysis

FIG. E-3. The nonzero-offset traveltime surface associated with a point scatterer and the various
summation trajectories for prestack time migration (composed from Fowler, 1997). See Section E.5 for
details.
Dip-Moveout Correction and Prestack Migration 831

of the constant-angle curves, independently, and place the result for each at the apex of
the summation curve. The summation collapses the pyramidal surface onto the hyperbolic
traveltime curve which is formed by combining the apex points of the constant-offset curves.
This hyperbolic traveltime curve is orthogonal to the constant-offset summation curves.

E.6 Prestack Frequency-Wavenumber Migration

We start with prestack data in midpoint y and offset h coordinates, P (y, h, z = 0, t), where t is
the event time in the unmigrated position, and perform 3-D Fourier transform

P (ky , kh , 0, ω) = P (y, h, 0, t) exp(iky y + ikh h − iωt) dky dkh dω, (E − 73)

where ky , kh , and ω are the Fourier duals of the variables of the midpoint-offset coordinates, y,
h, and t.
Extrapolate the prestack data from the surface z = 0 to a depth z by the following extrap-
olation equation which we borrow from Section D.1
P (ky , kh , z, ω) = P (ky , kh , 0, ω) exp(−ikz z). (E − 74)
The vertical wavenumber kz is given by
ω
kz = DSR(Y, H), (E − 75)
v
where, the double-square-root (DSR) operator takes the following form in midpoint-offset co-
ordinates (equation D-22 of Section D.1)
DSR(Y, H) = 1 − (Y + H)2 + 1 − (Y − H)2 . (E − 76)
The variables Y and H are the normalized midpoint and offset wavenumbers, respectively
vky
Y = (E − 77a)

and
vkh
. H= (E − 77b)

Assume a horizontally layered earth model associated with a vertically varying velocity function
v(z). By inverse Fourier transforming equation (E-74), we have

P (y, h, z, t) = P (ky , kh , 0, ω) exp(−ikz z) exp(−iky y − ikh h + iωt) dky dkh dω.


(E − 78)
The imaging principle t = 0 then is applied to get the image volume P (y, h, z, t = 0),

P (y, h, z, t = 0) = P (ky , kh , 0, ω) exp(−iky y − ikh h − ikz z) dky dkh dω. (E − 79)

This is the equation for the prestack phase-shift method. Equation (E-79) involves integration
over frequency ω and 2-D inverse Fourier transformation along midpoint y and offset h axes.
We now consider the special case of constant velocity v. Stolt (1978) devised a prestack
migration technique that involves an efficient mapping in the 3-D Fourier transform domain
from temporal frequency ω to vertical wavenumber kz .
First, combine equations (E-75) and (E-76)
vkz
= 1 − (Y + H)2 + 1 − (Y − H)2 . (E − 80)
ω
832 Seismic Data Analysis

Square both sides


v 2 kz2
= 2 1 − (Y + H)2 1 − (Y − H)2
ω2 , (E − 81)
+ 1 − (Y + H)2 + 1 − (Y − H)2
and simplify
v 2 kz2
− (1 − Y 2 − H 2 ) = 1 − 2Y 2 − 2H 2 + Y 4 − 2Y 2 H 2 + H 4 . (E − 82)
2ω 2
Define
v 2 kz2
K2 = (E − 83)
2ω 2
and square both sides of equation (E-82). After some algebra, it follows that
K 4 − 2K 2 + 2K 2 Y 2 + 2K 2 H 2 + 4Y 2 H 2 = 0, (E − 84)
which can be rewritten as
2K 2 = (K 2 + 2Y 2 )(K 2 + 2H 2 ). (E − 85)
Now substitute equations (E-77a,b) and (E-83) into equation (E-85), and simplify to obtain the
final expression for the dispersion relation for prestack wave extrapolation
v
ω= (kz2 + ky2 )(kz2 + kh2 ). (E − 86)
2kz
By setting the offset wavenumber kh = 0, we obtain the special case of zero-offset dispersion
relation
v
ω= kz2 + ky2 , (E − 87)
2
as in equation (D-85) with x replaced by y.
By keeping the wavenumbers ky and kh unchanged and differentiating equation (E-86), we
get
v kz2 − ky2 kh2
dω = dkz . (E − 88)
2 (kz2 + ky2 )(kz2 + kh2 )

By setting the offset wavenumber kh = 0, we obtain the special case for zero-offset
v kz
dω = dkz , (E − 89)
2 kz2 + ky2

as in equation (D-86) with x replaced by y.


When equations (E-86) and (E-88) are substituted into equation (E-79), we get
v kz2 − ky2 kh2
P (y, h, z, t = 0) =
2 (kz2 + ky2 )(kz2 + ky2 )
v
× P ky , kh , 0, (kz2 + ky2 )(kz2 + kh2 ) exp(−iky y − ikh h − ikz z) dky dkh dkz .
2kz
(E − 90)
Finally, sum over kh to obtain the image at zero offset, h = 0
Dip-Moveout Correction and Prestack Migration 833

v kz2 − ky2 kh2


P (y, h = 0, z, t = 0) =
2 (kz2 + ky2 )(kz2 + ky2 )
v
× P ky , kh , 0, (kz2 + ky2 )(kz2 + kh2 ) exp(−iky y − ikz z) dky dkh dkz .
2kz
(E − 91)
This is the equation for constant-velocity prestack Stolt migration. It involves two operations in
the f − k domain. First, the temporal frequency ω is mapped onto the vertical wavenumber kz
via equation (E-86). Second, the amplitudes are scaled by the quantity
v kz2 − ky2 kh2
S= . (E − 92)
2 (kz2 + ky2 )(kz2 + kh2 )

The zero-offset image is then obtained by summing over the wavenumber kh (equation
E-91), and inverse Fourier transforming in the midpoint y direction.

E.7 Velocity Analysis by Wavefield Extrapolation

A method of migration velocity analysis based on wavefield extrapolation is described in Section


5.4. The main computational steps of this method (Yilmaz and Chambers, 1984) are outlined
below. We work with seismic data in midpoint-(half) offset (y, h) coordinates. We want to
obtain a volume of focused energy at zero offset in (y, v, τ ) coordinates from a prestack data
set in (y, h, t) coordinates. For a midpoint location y, migration velocity function then can be
picked from the corresponding (v, τ ) plane.
First, a 3-D Fourier transformation is applied to the upcoming wavefield P (y, h, τ = 0, t),
which is recorded at the surface

P (ky , kh , τ = 0, t) = P (y, h, τ = 0, t) exp(iky y + ikh h − iωt) dky dkh dω, (E − 93)

where t is the two-way traveltime and


dz
τ =2 (E − 94)
v(z)
is the two-way vertical time equivalent of downward continuation depth z in a medium with
velocity v(z). The variables (ky , kh , ω) are the Fourier duals of (y, h, t).
The surface wavefield given by equation (E-93) then is extrapolated down to depth τ by
ω
P (ky , kh , τ, ω) = P (ky , kh , τ = 0, ω) exp −i τ DSR , (E − 95)
2
where
1/2 1/2
DSR ≡ 1 − (Y + H)2 + 1 − (Y − H)2 − 2, (E − 96)

and Y and H are the normalized midpoint and offset wavenumbers given by equations (E-
77a) and (E-77b). The −2 term puts the expression in retarded time form. (This term was not
included in the previous definition of DSR given by equation E-76.) Equation (E-95) is used
recursively to extrapolate the wavefield from one depth to another in steps of ∆τ .
Next, we transform the extrapolated wave field P (ky , kh , τ, ω) into the space-time domain.
In doing so, we only need to obtain the zero-offset information (h = 0). By summing the
extrapolated wavefield over kh in equation (E-95) , we get the wave field at zero offset, P (k, h =
834 Seismic Data Analysis

0, τ, ω). By doing the 2-D inverse transform over (ky , ω), we obtain

P (y, h = 0, τ, t) = P (ky , h = 0, τ, ω) exp(−iky y + iωt) dky dω. (E − 97)

Here, P (y, h = 0, τ, t) is the zero-offset section at various depth levels from which we want to
extract velocity information.
Suppose that velocity ve were used to extrapolate the surface wavefield down to depth τ .
Equation (E-95) is written with ve and τ as
ω
P (ky , kh , τ, ω) = P (ky , kh , τ = 0, ω) exp −i τ DSR(ve ) , (E − 98a)
2
Now suppose that the true medium velocity v were used to extrapolate the surface wavefield
down to depth τ = t. By rewriting equation (E-95) with v and t, we have
ω
P (ky , kh , t, ω) = P (ky , kh , τ = 0, ω) exp −i t DSR(v) . (E − 98b)
2
Match the two extrapolated wavefields in equations (E-98a) and (E-98b) to get a relationship
between ve , τ , v, and t
τ DSR(ve ) = t DSR(v). (E − 99)
Because of the complexity of DSR [equation (E-96)], equation (E-99) does not provide an
explicit expression for v in terms of the other three variables τ , t, and ve . However, we can
get an approximate expression by expanding the square roots in equation (E-96) in the Taylor
series and retaining the terms with Y 2 and H 2 , only. By using this approximate form and
the definitions given by equations (E-77a) and (E-77b), the following approximate relationship
results
τ ve2 = t v 2 . (E − 100)
This expression suggests that downward continuation with the correct (medium) velocity
to a wrong depth is equivalent to downward continuation to the correct depth with the wrong
velocity (Doherty and Claerbout, 1974).
The derivation of equation (E-100) assumes that ve is constant. When ve is depth-variable,
then the relationship in equation (E-100) still holds because equation (E-95) is valid for a
stratified earth model. However, quantity ve in that equation is replaced by the rms velocity.
Because the approximation made to equation (E-96) is best for small ratios of offset-to-
reflector depth, the accuracy of the mapping procedure based on equation (E-100) degrades
at very shallow depths. Refer to Section 5.4 for the practical considerations of the migration
velocity estimation technique described in this appendix.

REFERENCES

Artley, C. and Hale, D., 1994, Dip-moveout processing for depth-variable velocity: Geophysics, 59,
610-622.
Bale, R. and Jacubowitz, H., 1987, Poststack prestack migration: 57th Ann. Internat. Mtg., Soc.
Expl. Geophys., Expanded Abstracts, 714-717.
Bancroft, J. C. and Geiger, H. D., 1994, Equivalent-offset CRP gathers: 64th Ann. Internat. Mtg.,
Soc. Expl. Geophys., Expanded Abstracts, 672-675.
Bancroft, J. C., Geiger, H. D., and Margrave, G., 1998, The equivalent offset method of prestack
time migration: Geophysics, 63, 2042-2053.
Bancroft, J. C., Margrave, G., and Geiger, H. D., 1997, A kinematic comparison of DMO-PSI and
equivalent offset migration (EOM): 67th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
Abstracts, 1575-1578.
Dip-Moveout Correction and Prestack Migration 835

Baysal, E., Kosloff, D., and Sherwood, J. W. C., 1984, A two-way nonreflecting wave equation:
Geophysics, 49, 132-141.
Berkovitch, A., Keydar, S., Landa, E., and Trachtman, P., 1998, Multifocusing in practice: 68th
Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1748-1751.
Berryhill, J. R., 1996, System and method of seismic shot-record migration: U. S. Patent 5,544,126.
Biondi, B. and Ronen, J., 1987, Dip moveout in shot profiles: Geophysics, 52, 1473-1482.
Black, J., Schleicher, K. L. and Zhang, L., 1993, True-amplitude imaging and dip moveout: Geo-
physics, 58, 47-66.
Black, J.L., McMahon, I.T., Meinardus, H., and Henderson, I., 1985, Applications of prestack
migration and dip moveout: Presented at the 55th Ann. Internat. Mtg., Soc. Expl. Geophys.
Bleistein, N., 1990, Born DMO revisited: 60th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
Abstracts, 1366-1369.
Bolondi, G., Loinger, E. and Rocca, F., 1982, Offset continuation of seismic sections: Geophys.
Prosp., 30, 813-828.
Bolondi, G., Loinger, E. and Rocca, F., 1984, Offset continuation in theory and practice: Geophys.
Prosp., 32, 1045-1073.
Bolondi, G. and Rocca, F., 1985, Normal moveout correction, offset continuation and prestack
partial migration compared as prestack processes: in Fitch, A. A., Ed., Developments in geo-
physical exploration methods, 6, Elsevier Applied science Publ., Ltd.
Cabrera, J. and Levy, S., 1989, Shot dip-moveout with logarithmic transformations: Geophysics,
54, 1038-1041.
Claerbout, J.F., 1985, Imaging the earth’s interior: Blackwell Scientific Publications.
Clayton, R., 1978, Common midpoint migration: Stanford Expl. Proj., Rep. No. 14, Stanford
University.
de Bazelaire, E., 1988, Normal-moveout correction revisited: Inhomogeneous media and curved
interfaces: Geophysics, 53, 143-157.
Deregowski, S. M., 1982, Dip-moveout and reflector-point dispersal: Geophys. Prosp., 30, 318-322.
Deregowski, S. M., 1986, What is DMO?: First Break, 4, 7-24.
Deregowski, S. M., 1987, An integral implementation of dip-moveout: Geophys. Trans. of Geophys.
Inst. of Hungary, 33, 11-22.
Deregowski, S. M. and Rocca, F., 1981, Geometrical optics and wave theory for constant-offset
sections in layered media: Geophys. Prosp., 29, 374-406.
Doherty, S. M., 1975, Structure-independent velocity estimation: Ph. D. thesis, Stanford Unviersity.
Doherty, S.M. and Claerbout, J.F., 1974, Velocity analysis based on the wave equation: Stanford
Expl. Proj., Rep. No. 1, Stanford University.
Faye, J-P. and Jeannaut, J-P., 1986, Prestack migration velocities from focusing depth analysis:
56th Ann. Internat. Mtg. Soc. Expl. Geophys., Expanded Abstracts, 438-440.
Fowler, P., 1984, Velocity-independent imaging of seismic reflectors: 54th Ann. Internat. Mtg., Soc.
Explor. Geophys., Expanded Abstracts, 383-385.
Fowler, P., 1997, A comparative overview of prestack time migration methods: 67th Ann. Int. Soc.
Explor. Geophys. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1571-1574.
French, W.S., Perkins, W.T., and Zoll, R.M., 1984, Partial migration via true CRP stacking: 54th
Ann. Internat. Mtg., Soc. Explor. Geophys., Expanded Abstracts, 799-802.
Gardner, G. F. H. and Forel, D., 1990, Amplitude preservation equations for DMO: Geophysics,
55, 485-487.
Gardner, G.H.F., French, W.S., and Matzuk, T., 1974, Elements of migration and velocity analysis:
Geophysics, 39, 811-825.
Gardner, G. F. H., Wang, S. Y., Pan, N. D., and Zhang, Z., 1986, Dip moveout and prestack
imaging: Expanded Abstracts, 75-84, 18th Ann. Offshore Tech. Conf.
Gelchinsky, B., 1988, The common-reflecting-element (CRE) method: ASEG/SEG Internat. Geo-
phys. Conf., Extended Abstracts, 71-75.
Gelchinsky, B., Berkovitch, A., and Keydar, S.,1999a, Multifocusing homoemorphic imaging: Part
I: Basic concepts and formulas: J. Appl. Geophys., 42, 229-242.
836 Seismic Data Analysis

Gelchinsky, B., Berkovitch, A., and Keydar, S.,1999b, Multifocusing homoemorphic imaging: Part
II: Multifold data set and multifocusing: J. Appl. Geophys., 42, 243-260.
Gonzalez-Serrano, A. and Claerbout, J.F., 1979, Wave-equation velocity analysis: Stanford Expl.
Proj., Rep. No. 16, Stanford University.
Granser, H., 1994, Shot gather DMO in the double-log domain: Geophysics, 49, 1305-1307.
Hale, D., 1983, Dip moveout by Fourier transform: Ph.D. thesis, Stanford University.
Hale, D., 1984, Dip moveout by Fourier transform: Geophysics, 49, 741-757.
Hale, D., 1991, A nonaliased integral method for dip moveout: Geophysics, 56, 795-805.
Hale, D. and Artley, C., 1992, Squeezing dip-moveout for depth-variable velocity: Geophysics, 58,
257-264.
Hale, D., Hill, N. R., and Stefani, J., 1992, Imaging salt with turning seismic waves: Geophysics,
57, 1453-1462.
Höcht, G., 1998, Common-reflection-surface stack: Master’s thesis, Karlsruhe University.
Jacubowicz, H., 1990, A simple efficient method of dip-moveout correction: Geophys. Prosp., 38,
221-245.
Landa, E., Gurevich, B., Keydar, S., and Trachtman, P., 1999, Application of multifocusing method
for subsurface imaging: J. Appl. Geophys., 42, 283-300.
Levin, F.K., 1971, Apparent velocity from dipping interface reflections: Geophysics, 36, 510-516.
Li, Z., Lynn, W., Chambers, R., Larner, K. and Abma, R., 1991, Enhancements to prestack
frequency-wavenumber (f − k) migration: Geophysics, 56 , 27-40.
Liner, C. L., 1990, General theory and comparative anatomy of dip-moveout: Geophysics, 55,
595-607.
Margrave, G., Bancroft, J. C., and Geiger, H. D., 1999, Fourier prestack migration by equivalent
wavenumber: Geophysics, 64, 197-207.
Marcoux, M. O., Godfrey, R. J., and Notfors, C. D., 1987, Migration for optimum velocity evalu-
ation and stacking: Presented at the 49th Ann. Mtg. European Asn. Expl. Geophys.
Notfors, C. D. and Godfrey, R. J., 1987, Dip-moveout in the frequency-wavenumber domain: Geo-
physics, 52, 1718-1721.
Ottolini, R., 1982, Migration of seismic data in angle-midpoint coordinates: Ph.D. thesis, Stanford
University.
Ratcliff, D. W., Gray, S. H., and Whitmore, N. D., 1992, Seismic imaging of salt structures in the
Gulf of Mexico: The Leading Edge, 11, no. 4, 15-31.
Rocca, F., Bolondi, G., and Loinger, E., 1982, Offset continuation of seismic sections: Geophys.
Prosp., 30, 813-828.
Salvador, L. and Savelli, S., 1982, Offset continuation for seismic stacking: Geophys. Prosp., 30,
829-849.
Sherwood, J.W.C., Schultz, P.S., and Judson, D.R., 1978, Equalizing the stacking velocities of
dipping events via Devilish: Presented at the 48th Ann. Internat. Soc. Expl. Geophys. Mtg.
Shurtleff, R., 1984, An F − K procedure for prestack migration and velocity analysis: Presented
at the 46th Ann. Mtg. European Asn. Expl. Geophys.
Sorin, V. and Ronen, J., 1989, Ray-geometrical analysis of dip moveout amplitude distribution:
Geophysics, 54, 1333-1335.
Stolt, R.H., 1978, Migration by Fourier transform: Geophysics, 43, 23-48.
Taner, M.T. and Koehler, F., 1969, Velocity spectra — digital computer derivation and applications
of velocity functions: Geophysics,32, 859-881.
Tygel, M., Müller, T., Hubral, P., and Schleicher, J., 1997, Eigenwave-based multiparameter
traveltime expansions: 67th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts,
1770 - 1773.
Yilmaz, O. and Chambers, R., 1984, Migration velocity analysis by wave field extrapolation: Geo-
physics, 49, 1664-1674.
Yilmaz, O. and Claerbout, J.F., 1980, Prestack partial migration: Geophysics, 45, 1753-1777.
Zhou, B, Mason, I. M., and Greenalgh, S.A., 1996, Accurate and efficient shot-gather dip-moveout
processing in the log-stretch domain: Geophys. Prosp., 43, 963-978.

You might also like