You are on page 1of 140

8 Earth Imaging in Depth

• Introduction • Lateral Velocity Variations • Layer Replacement • Wave-Equation Datuming • Poststack


Layer Replacement • Prestack Layer Replacement • Field Data Example • 2-D Poststack Depth Migration •
Image Rays and Lateral Velocity Variations • Time versus Depth Migration • Iterative Depth Migration • Iteration
with Zero-Offset Data • Iteration with CMP-Stacked Data • Iteration with Prestack Data • Iteration in Practice
• 2-D Prestack Depth Migration • Shot-Geophone Migration • Shot-Profile Migration • Sensitivity of Image
Accuracy to Velocity Errors • Field Data Examples • 3-D Poststack Depth Migration • 3-D Poststack Time
versus Depth Migration • Two-Pass versus One-Pass 3-D Poststack Depth Migration • Implicit versus Explicit 3-D
Poststack Depth Migration • 3-D Poststack Datuming • 3-D Prestack Depth Migration • Kirchhoff Summation
• Calculation of Traveltimes • The Eikonal Equation • Fermat’s Principle • Summation Strategies • Migration
Aperture • Operator Antialiasing • 3-D Common-Offset Depth Migration • Exercises • Appendix H: Diffraction
and Ray Theory for Wave Propagation • The Kirchhoff Integral • The Eikonal Equation • Finite-Difference
Solution to the Eikonal Equation • References

8.0 INTRODUCTION

Strong lateral velocity variations associated with com- Time migration yields an inaccurate image of the base-
plex overburden structures require earth imaging in salt (Figure 8.0-1b).
depth. Examples of complex overburden include di- Figure 8.0-2a is a field data example of an imbricate
apiric structures formed by salt tectonics, imbricate structure associated with overthrust tectonics. Time mi-
structures formed by overthrust tectonics and irregu- gration is adequate for imaging a target within the im-
lar water-bottom topography. All three are character- bricate structure itself, but would not yield a correct
ized as structure-dependent lateral velocity variations. image of a target below it (Figure 8.0-2b).
There also exist structure-independent lateral velocity Refer to the field data example for a case of irreg-
variations, often associated with facies changes; for in- ular water-bottom topography shown in Figure 8.0-3a.
stance, changes in lithology from shale to sandstone to Note the false structures along the unconformity T be-
carbonate induce lateral changes in acoustic impedance. cause of irregular water-bottom topography. These are
Figure 8.0-1a shows a field data example of a di- especially noticed below midpoint locations A, B, and
apiric structure associated with salt tectonics. Note C. Sagging of the reflection that is associated with the
how the base-salt (event B) and subsalt reflections at unconformity at these locations is attributed to the low-
about 2 s are pulled up in the middle of the section. velocity overburden, here, the water layer. Once again,
1214 Seismic Data Analysis

FIG. 8.0-1. (a) A CMP stacked section over a salt-dome structure, (b) time migration. Note the similarity of the overmigration
at the base-salt event in this section with that seen in the time-migrated section of the synthetic data in Figure 8.2-7.

time migration does not provide the solution for such a strong lateral velocity variations, an earth image in
complex overburden as shown in Figure 8.0-3b. Specif- time derived from time migration is not accurate;
ically, distortions along the unconformity still exist. and thus, it is imperative to obtain an earth image
Earth imaging in depth is achieved by depth mi- in depth by depth migration.
gration. The following is a cause-and-effect relation be-
(b) Strong lateral velocity variations cause significant
tween various factors with regard to depth migration:
ray bending at layer boundaries.
(a) Complex overburden structures often give rise to (c) This then gives rise to nonhyperbolic behavior of
strong lateral velocity variations. In the presence of reflection times on CMP gathers that correspond
Earth Imaging in Depth 1215

FIG. 8.0-2. (a) A CMP stacked section from an area with overthrust tectonics, (b) time migration.
1216 Seismic Data Analysis

FIG. 8.0-3. (a) Conventional CMP stack from an area with irregular water-bottom topography, (b) time migration. (Data
courtesy Hispanoil.)

to layer boundaries below a complex overburden when depth migration is needed, again in principle,
structure. it must be done not only before stack but also in
(d) As a result, amplitudes and traveltimes associ- three dimensions.
ated with the reflection events with nonhyperbolic
moveout are distorted during conventional CMP Compare the results of depth imaging in 2-D and
stacking which is based on the hypberbolic move- 3-D using post- and prestack data shown in Figures
out assumption. 8.0-4 through 8.0-7. Note the improved imaging of the
(e) This causes CMP stack to depart from an ideal fault planes and the reflectors within the faulted zone
zero-offset wavefield. Therefore, when depth migra- attained by combining the advantages of prestack mi-
tion is needed, in principle, it must be done before gration and 3-D migration. Figure 8.0-4 shows 2-D post-
stack and not after. stack depth migration of the 3-D DMO stack in Figure
(f) Finally, complex overburden structures often ex- 7.0-16a derived from common-cell gathers along an in-
hibit three-dimensional (3-D) behavior. Therefore, line from a 3-D marine survey. Velocities vary in the
Earth Imaging in Depth 1217

FIG. 8.0-4. (a) 2-D poststack depth migration of the 3-D DMO stack shown in Figure 7.0-13a derived from common-cell
gathers along an inline from a 3-D marine survey; (b) the same section as in (a) displayed with reverse polarity.
1218 Seismic Data Analysis

FIG. 8.0-5. (a) 2-D prestack depth migration of the data as in Figure 7.0-13a; (b) the same section as in (a) displayed with
reverse polarity.
Earth Imaging in Depth 1219

FIG. 8.0-6. (a) An inline section from the image volume derived from 3-D poststack depth migration of the data as in Figure
7.0-16a; (b) the same inline section as in (a) displayed with reverse polarity.
1220 Seismic Data Analysis

FIG. 8.0-7. (a) An inline section from the image volume derived from 3-D prestack depth migration of the data as in Figure
7.0-16a; (b) the same inline section as in (a) displayed with reverse polarity.
Earth Imaging in Depth 1221

lateral direction across the steep faults in the middle contain discontinuities, an interval velocity-depth model
portion of the image. The imaging within the faulted can include discontinuities associated with layer bound-
zone can be improved by prestack depth migration as aries.
shown in Figure 8.0-5. Nevertheless, the 3-D behavior A velocity-depth model is the seismic representa-
of the fault planes requires 3-D imaging (Figure 8.0-6). tion of an earth model in depth. An earth model and
The complete solution to the imaging problem in the the earth image created from it are an inseparable pair
presence of lateral velocity variations and the 3-D be- of products of seismic inversion. To obtain an earth im-
havior of reflector geometries, of course, is 3-D prestack age in depth, one has to first estimate an accurate earth
depth migration (Figure 8.0-7). Since lateral velocity model in depth. We shall defer estimation of earth mod-
variations can change the polarity of fault-plane reflec- els in depth to the next chapter where we shall demon-
tions, the image sections in Figures 8.0-4 through 8.0-7 strate the use of various inversion methods and depth
are displayed both with normal and reverse polarities. migration itself to estimate the earth model parameters.
Although the migration methods discussed in Aside from earth modeling and imaging in depth, depth
Chapter 4 are based on a layered media assumption, migration also is used to verify and update velocity-
simple modifications of the basic algorithms make them depth models. In this chapter, we shall be concerned
accurate for situations with mild lateral velocity varia- only with depth migration as a tool for earth imaging
tions. For example, rms velocities can be varied laterally in depth.
in Kirchhoff migration. In the finite-difference method, Historically, depth migration was used in an it-
as long as lateral velocity variations are mild, the thin- erative fashion. Starting with an initial velocity-depth
lens term can be dropped (Section D.3), and the velocity model, depth migration is performed and results are
function used in the diffraction term can be varied lat- interpreted for updating layer velocities and reflector
erally. In the frequency-wavenumber methods such as geometries. Using the updated velocity-depth model,
depth migration is repeated until such time as what
Stolt migration, lateral velocity variations are accom-
is inferred from depth migration as the velocity-depth
modated by varying the stretch factor between 0 and 1.
model matches with the velocity-depth model input to
Even when velocity varies, the output from these three
depth migration. To achieve rapid convergence to a fi-
methods is still a time section, thus, the term time mi-
nal velocity-depth model, only one set of parameters —
gration.
usually, reflector geometries, are altered from one iter-
The situation is different when strong lateral veloc-
ation to the next. We shall review the iterative depth
ity variations are encountered. In that case, simple algo-
migration procedure for zero-offset, CMP-stacked and
rithmic modifications no longer provide adequate accu-
prestack data, and draw conclusions as to when such a
racy, and depth migration (Judson et al., 1980; Schultz procedure yields an acceptable image in depth (Section
and Sherwood, 1980), rather than time migration, must 8.2).
be done. While both migration types use a diffraction In this chapter, we shall review two methods of
term for collapsing energy along a diffraction hyperbola two-dimensional (2-D) prestack depth migration (Sec-
to its apex, only depth migration algorithms implement tion 8.3). In 2-D migration, we assume that the seismic
the additional thin-lens term that explicitly accounts for line is along the dip direction and that the recorded
lateral velocity variations (Section D.4). Unlike time mi- wavefield is two dimensional. Shot-geophone migration,
gration, the output from the migration algorithms that which is based on downward extrapolation of common-
include the thin-lens term is a depth section, thus the shot and common-receiver gathers using the double-
term depth migration. square root equation (Section D.1), focuses primary re-
Sometimes, although not so often, the complex flection energy to zero-offset. The migrated section is
overburden can be defined by a single layer and the obtained by retaining the zero-offset traces and aban-
boundary between the overburden and the substratum doning the nonzero-offset traces. Shot-profile migration
can be determined in the form of an irregular interface is based on migrating each common-shot gather, indi-
with a significant velocity contrast. In that case, the vidually. In this method, the migrated section is ob-
layer replacement method described in Section 8.1, in tained by sorting the migrated common-shot gathers
lieu of depth migration, can be used to remove the dele- into common-receiver gathers and summing the traces
terious effect of the overburden on the geometry of the in each receiver gather.
underlying reflections. We shall review 3-D post- and prestack depth mi-
Time migration requires an rms velocity field, grations in Sections 8.4 and 8.5, respectively. While
whereas depth migration requires an interval velocity- 3-D poststack depth migration can be performed using
depth model. A velocity-depth model usually is defined a wide variety of algorithms based on finite-difference
by two sets of parameters — layer velocities and re- (Sections G.1 and G.2), frequency-wavenumber (Sec-
flector geometries. While an rms velocity field does not tions G.3 and G.4), and Kirchhoff integral solutions
1222 Seismic Data Analysis

(Section H.1) to the 3-D scalar wave equation, 3-D this case, is coincident with the CMP location of the
prestack depth migration often is done using the Kirch- point diffractor.
hoff integral (Section H.1) or the eikonal equation (Sec- Consider what happens when the point diffractor is
tion H.2). This is because Kirchhoff migration is conve- lowered into the second layer as shown in Figure 8.0-9.
niently adaptable to source-receiver geometry irregular- Raypaths from the scatterer to the surface are bent at
ities in 3-D seismic surveys. Topographic irregularities the interface between the first and second layers accord-
also are conveniently accommodated in Kirchhoff mi- ing to Snell’s law of refraction. The zero-offset section
gration. is now approximately hyperbolic. From Section 3.1, we
The output of prestack depth migration consists recall that in a horizontally layered earth model, travel-
of image gathers, which may be likened to moveout- times are governed by the hyperbolic moveout equation.
corrected CMP gathers with the vertical axis in depth. However, moveout is hyperbolic only within the small-
Image gathers, however, consist of traces in their mi- spread approximation. The velocity associated with this
grated positions. A stack of image gathers represents approximate hyperbola is the vertical rms velocity down
the earth image in depth obtained from prestack depth to the diffractor. Suppose the velocity-depth model of
migration. If the velocity-depth model used in prestack Figure 8.0-9 is replaced with that of Figure 8.0-10,
depth migration is correct, then, events on an image where the velocity of the first layer now corresponds to
the rms velocity of the diffractor of Figure 8.0-9 (1790
gather would exhibit a flat character with no move-
m/s). The zero-offset section associated with this new
out. An erroneously too low or too high velocity would
model is an exact hyperbola (Figure 8.0-10).
cause a residual moveout on the image gather. The ini-
The traveltime trajectory in the zero-offset section
tial velocity-depth model can be updated by analyzing
derived from the original model (Figure 8.0-9) differs
this residual moveout and correcting for it (Section 9.5).
negligibly from the hyperbolic traveltime trajectory in
the section associated with the replacement model (Fig-
ure 8.0-10) only at the far flanks. Therefore, it is rea-
Lateral Velocity Variations sonable to assume that the zero-offset traveltime tra-
jectory for a point diffractor in a horizontally layered
Lateral velocity variations often are associated with earth model is a hyperbola. The apex of this approxi-
steep dips. Hence, a depth migration algorithm must mate hyperbola in Figure 8.0-9 coincides with the sur-
not only handle lateral velocity variations but also must face projection of the diffractor, indicated by the ar-
image steeply dipping events accurately. The steep-dip row. Therefore, only time migration is needed to image
implicit and explicit frequency-space migration algo- a diffractor that is buried in a horizontally layered earth
rithms described in Section 4.4 are particularly suit- model. This time migration can be performed either by
able to accommodate lateral velocity variations. In the Kirchhoff summation, which uses the rms velocity, or
by frequency-space or frequency-wavenumber methods,
case of the implicit schemes, the action of the thin-lens
which honor raypath bendings at the interfaces associ-
term, which accounts for lateral velocity variations, is
ated with horizontal layers.
achieved by a complex multiplication of the wavefield in
Now consider the point diffractor situated in the
the frequency-space domain with a velocity-dependent
third layer as shown in Figure 8.0-11. Now we no longer
exponential term (Sections D.3 and D.4). In the case of
have even an approximate hyperbolic diffraction re-
the explicit schemes, lateral velocity variations are ac- sponse. The traveltime trajectory is skewed so that the
counted for by designing a velocity-dependent, laterally apex A does not coincide with the lateral position B
varying extrapolation filter and convolving it with the of the diffractor. As expected, time migration partially
wavefield in the frequency-space domain (Section D.5). focuses the energy toward its apex A, which is to the
The problem of lateral velocity variations will be left of the actual lateral position of diffractor B.
studied using a point diffractor buried in a medium To properly focus the energy and place it laterally
with five different types of velocity-depth models. The to its true subsurface position B, depth migration must
first velocity-depth model is shown in Figure 8.0-8. be performed as shown in Figure 8.0-11. The depth-
The corresponding zero-offset section consists of an ex- migrated image is aligned with the true subsurface po-
act diffraction hyperbola. Therefore, only the diffrac- sition B. This lateral positioning is accomplished by the
tion term is needed in imaging the scatterer (Section thin-lens term. The amount of lateral shift is the lateral
D.3). The surface projection of the point diffractor (the distance AB between the apex A of the skewed diffrac-
source) coincides with the location of CMP 240, indi- tion traveltime trajectory and the actual location B of
cated by the vertical arrow, and is aligned with the apex the diffractor. This shift depends on the amount of ray
of the diffraction hyperbola. After time migration, the bending that occurs at the interfaces above the point
diffraction hyperbola is collapsed to its apex which, in diffractor.
Earth Imaging in Depth 1223

FIG. 8.0-9. The response of a point diffractor buried in a


FIG. 8.0-8. The response of a point diffractor buried in a layered medium (top frame) is approximately a hyperbola
constant-velocity medium (top frame) is a hyperbola (mid- (middle frame). Time migration (bottom frame) still is ad-
dle frame). Time migration (bottom frame) is adequate in equate for imaging the diffractor. The arrows indicate the
imaging this diffractor. The arrows indicate the surface pro- surface projection of the true lateral position of the diffrac-
jection of the true lateral position of the diffractor. tor.
1224 Seismic Data Analysis

FIG. 8.0-10. The response of a point diffractor buried in FIG. 8.0-11. The response of a point diffractor buried in a
a layered medium as in Figure 8.0-9 is nearly equivalent to medium with strong lateral velocity variation (top frame) is
the response of another point diffractor buried in a constant- a skewed hyperbola with its apex shifted to the left of the
velocity medium (top frame), where the constant velocity is true position. Time migration no longer is a valid process;
the rms value of the layered velocity function at the depth instead, depth migration is needed.
of the original diffractor as in Figure 8.0-9. Traveltimes as-
sociated with the zero-offset sections in Figures 8.0-9 and
8.0-10 depart from one another only at the very far flanks
of the diffraction curve. Time migration (bottom frame) is
adequate for imaging the scatterer.
Earth Imaging in Depth 1225

FIG. 8.0-12. The response of a point diffractor buried in FIG. 8.0-13. The response of a point diffractor buried in
a medium with mild to moderate lateral velocity variation a medium with severe lateral velocity variation (top frame)
(top frame) is a slightly skewed hyperbola with its apex is a distorted traveltime curve that implies false structural
shifted to the left of the true position. Time migration still features. Time migration no longer is acceptable; instead
depth migration is imperative.
can be an acceptable way to image the diffractor.
1226 Seismic Data Analysis

From Figure 8.0-11, note that the apex of the the image ray starting at CMP location 140 reaches
skewed traveltime trajectory A coincides with the sur- horizon 4 approximately beneath CMP location 180; a
face position of the ray that emerges vertically. This lateral shift of 40 midpoints. Proper imaging of these
special ray is called the image ray, and it was first rec- two horizons is achievable only by depth migration.
ognized by Hubral (1977). The image ray associated From Figure 8.0-11, note that time migration col-
with the point diffractor in Figure 8.0-11 is roughly at lapses the energy to apex A of the diffraction curve
midpoint 200. The diffractor itself is located beneath that coincides with the image ray location at the sur-
midpoint 240. Therefore, the lateral shift is equivalent face. In principle, the time migration output then can
to 40 midpoints. be converted to depth along the image rays, rather than
There is no lateral shift for the horizontally layered along the vertical rays (Hubral, 1977). Mapping along
earth model (Figure 8.0-9), since there is no lateral ve- the image rays performs some of the action that is asso-
locity variation. The image ray in this case emerges at ciated with the thin-lens term. Remember that at each
the surface location of CMP 240 coincident with that downward-continuation step, the action of the thin-lens
of the diffractor. For a mild to moderate lateral veloc- term is equivalent to a time shift that depends on spa-
ity variation, as in Figure 8.0-12, the lateral shift is less tial velocity variation. Since the thin-lens and diffrac-
than 10 midpoints. For some objectives, this small lat-
tion terms are applied in an alternate manner as the
eral shift and the complete focusing may not be critical;
wavefield is downward continued in depth, the effects of
therefore, time migration may be acceptable in lieu of
these two terms are strongly coupled when the lateral
depth migration. In those cases, the coefficient of the
velocity variation is as severe as shown in Figure 8.0-13.
thin-lens term is negligibly small, which is why we of-
When lateral velocity variation is moderate to
ten can get away with time migration in areas of mild
strong, these two terms can, in principle, be sepa-
to moderate lateral velocity variations.
The imaging problem is complicated when the over- rated and applied consecutively without significant er-
burden is as complex as that in Figure 8.0-13. Here, ror (Larner et al., 1981). Full separation implies that
because of the bowties, a distorted diffraction travel- a correction for the effects of the thin-lens term can
time trajectory indicates a false structure. The result- be done either before or after time migration. If the
ing complexity can give rise to more than one image correction is done after time migration, image-ray map-
ray. In this case, three image rays emerge at around ping should be used. If the correction is done before
midpoints 160, 250, and 370. Time migration fails here time migration, mapping using vertical time shifts usu-
and imaging of this scatterer can only be achieved by ally is applied. In practice, a correction before time mi-
depth migration. gration often performs better, since it tends to provide
We studied some examples of lateral velocity vari- a better focused migration result. Nevertheless, all of
ations in Figures 8.0-8 through 8.0-13. The image ray the above conjectures are related only to now outdated
behavior and the quality of focusing determine whether migration algorithms since contemporary implementa-
time or depth migration should be performed. If the tions of depth migration algorithms are based on split-
starting and end points of the image ray have the same ting, and not separation, of the diffraction and thin-
CMP location (Figure 8.0-9), only time migration is lens terms (Sections D.3 and D.4). It is not a common
needed. A small amount of lateral deviation of the im- practice to correct for the effect of the thin-lens term
age ray (Figure 8.0-12) usually implies a well-focused by time-to-depth conversion of time-migrated data us-
time migration result and hence, a good representation ing image rays. However, it is a common practice to do
of the geometric form of the subsurface. Large image-ray time-to-depth conversion of time horizons interpreted
deviations imply grossly incorrect focusing, thus requir- from time-migrated data by using image rays (Section
ing depth migration rather than time migration (Figure 9.4).
8.0-11). Finally, if more than one image ray is associated
with a subsurface point (Figure 8.0-13), depth migra-
tion is imperative.
These observations on the point diffractor models 8.1 LAYER REPLACEMENT
are extended to a velocity-depth model that involves
the reflecting interfaces in Figure 8.0-13. The image rays In this section, we demonstrate that the problem of a
associated with this model are shown in Figure 8.0-14. complex overburden that involves only one layer bound-
There is no deviation from the vertical along the image ary, such as an irregular water bottom, at which sig-
rays down to horizon 2. Therefore, depth migration is nificant ray bending takes place can sometimes be ad-
not needed to image this horizon. On the other hand, dressed by prestack layer replacement followed by NMO
the image rays significantly deviate from the vertical correction, CMP stacking, and poststack time migra-
as they travel down to horizons 3 and 4. For example, tion. In both cases, the approaches are based on the
Earth Imaging in Depth 1227

FIG. 8.0-14. Image rays through the velocity-depth model in Figure 8.0-13. Note the deviation of the image rays from the
vertical as they travel through the complex structure.
1228 Seismic Data Analysis

FIG. 8.1-1. (a) Velocity contrast between the overburden and the substratum causes raypath bending at the interface between
the two. (b) Replacing the overburden velocity with the velocity of the substratum eliminates raypath bending.

philosophy of revising velocity estimates and obtaining the overburden velocity with the substratum velocity
an improved unmigrated stacked section. can be a viable alternative to using depth migration
Consider the velocity-depth model in Figure 8.1- to remove the deleterious effects of a complex overbur-
1a. Note that complex geometry of the boundary (hori- den, such as an irregular water-bottom topography, on
zon 2) between the overburden and the substratum and the substrata. This technique is known as layer replace-
the significant velocity contrast across this boundary ment.
cause the severe ray bending. This in turn causes dis- A technique for layer replacement (Yilmaz and Lu-
tortions and disruptions of the underlying target reflec- cas, 1986) based on wave-equation datuming (Berry-
tions. Without the velocity contrast (Figure 8.1-1b), the hill, 1979, 1984) is presented here. Berryhill’s datuming
rays would not bend and there would be no need for technique involves extrapolating a known wavefield at
depth migration. Figure 8.1-1 suggests that replacing a specified datum of arbitrary shape to another datum,
Earth Imaging in Depth 1229

FIG. 8.1-2. Layer replacement by Kirchhoff summation. (a) Input zero-offset section (at datum level z = 0) based on a
velocity-depth model consisting of three point scatterers buried at depths of 800, 1300, and 1900 m. Interval velocities are
indicated on the right side of the section. (b) Step 1: Downward continuation of the wavefield from z = 0 to datum level
z = 800 m using a velocity of 2000 m/s. (c) Step 2: Upward continuation of (b) back to z = 0 using a velocity of 2500 m/s.
(d) The zero-offset section derived independently using interval velocities that are indicated on the left side of the section.
This section should be compared to (c).

also of arbitrary shape. Wave extrapolation is per- m/s) and compute the wavefield at the first interface,
formed using the Kirchhoff integral solution to the z = 800 m. The result is shown in Figure 8.1-2b. As ex-
scalar wave equation. It incorporates both the near-field pected, the hyperbola associated with the shallow point
and far-field terms (Section H.1). The velocity used in scatterer largely collapses to its apex since this scatterer
extrapolation is that of the medium confined between is located at the first interface. Because the receivers
the input datum and the output datum. now are closer to the other two deeper scatterers, events
associated with them also are compressed. Figure 8.1-
2b shows the zero-offset section that would have been
recorded if the receivers were placed along the first in-
Wave-Equation Datuming
terface. The energy from the shallow point scatterer on
this section now arrives at t = 0 because the datum
Figure 8.1-2 shows a simple case of datuming. A zero- for this section is the interface at which the scatterer is
offset section is computed over three point scatterers located.
buried beneath midpoint location A in a medium with Now, extrapolate the wavefield at the first inter-
the layered velocity structure as denoted between Fig- face (z = 800 m) shown in Figure 8.1-2b back up to
ures 8.1-2a and b. The point diffractors are situated at the surface (z = 0) using the velocity of the second
the layer interfaces at 800, 1300, and 1900 m depths. layer (2500 m/s). Figure 8.1-2c shows the result, which
The traveltime trajectory associated with the shallow is compared to the zero-offset section in Figure 8.1-2d.
point diffractor is a hyperbola. The traveltimes associ- The latter was derived independently from the velocity-
ated with the deeper diffractors are nearly hyperbolic depth model denoted between Figures 8.1-2c and d.
(Section 3.1). With this two-step wave extrapolation, the first layer
Extrapolate the zero-offset wavefield at z = 0 (Fig- with the 2000-m/s velocity was replaced with the sec-
ure 8.1-2a) using the velocity of the first layer (2000 ond layer with the 2500-m/s velocity.
1230 Seismic Data Analysis

FIG. 8.1-3. Poststack layer replacement involves two steps. Step 1: The zero-offset section (a) is extrapolated down to the
water bottom (horizon 2 in Figure 8.1-1a) using the water velocity. Section (b) is obtained when the receivers are placed
along the water bottom. Step 2: This intermediate wavefield is extrapolated back up to the surface using the velocity of the
stratum below the water bottom (2000 m/s). The resulting zero-offset section (c) can be compared against the zero-offset
section derived independently (d) using the same velocity-depth model as in Figure 8.1-1a, except the overburden velocity is
the same as that of the substratum (2000 m/s) as shown in Figure 8.1-1b.

In principle, wave-equation datuming can be per- model (Section 4.0). Horizon flattening involves down-
formed by any extrapolation method based on phase- ward continuing from one marker horizon to another.
shift, finite-difference, or Kirchhoff summation tech- When this is done successively for all marker horizons
niques. Nevertheless, the Kirchhoff summation is more in a section, the technique is useful in reconstructing the
convenient in handling datum surfaces with arbitrary past structural history in a given survey area (Taner et
shapes. al., 1982). Seismic modeling involves a series of upward
It is important to distinguish between datuming continuations through a specified set of velocity inter-
and migration. Datuming produces an unmigrated time faces starting at the bottom of the model and ending at
section at a specified datum z(x), which can be arbi- the top. An example is provided in Section K.1.
trary in shape. Migration involves computing the wave-
field at all depths from the wavefield at the surface. In
this respect, datuming is an ingredient of migration,
when migration is done as a downward-continuation Poststack Layer Replacement
process. In addition to downward continuation, migra-
tion, of course, requires invoking the imaging principle
(t = 0) (Section 4.1). Now consider another practical application of wave-
Wave-equation datuming has several practical ap- equation datuming to 2-D surface seismic data — re-
plications — horizon flattening, forward modeling of moving the degrading effect of an irregular water-
seismic wavefields, and layer replacement. These are bottom topography on the continuity and geometry of
performed in either the prestack or poststack mode. reflections below. This problem is particularly severe in
The main difference between the two implementations areas with a strong velocity contrast between the water
is that the velocity must be halved when doing post- layer and the substratum. Despite the usual 3-D nature
stack datuming to conform to the exploding reflectors of the problem, the 2-D interpretation of the target re-
Earth Imaging in Depth 1231

observing any arrival-time departures of the event


from t = 0, the validity of an estimated velocity-
depth model can be verified.
(b) The second step in poststack layer replacement
involves upward continuation of the intermediate
wavefield (Figure 8.1-3b) back to the surface z = 0
using the velocity of the substratum (2000 m/s).
Figure 8.1-3c is the zero-offset section at z = 0
after layer replacement.

A zero-offset section was created from the same


velocity-depth model (Figure 8.1-1a) as for Figure 8.1-
3a, except that the first layer velocity was set to 2000
m/s (Figure 8.1-1b). Compare this zero-offset section as
shown in Figure 8.1-3d with the output of layer replace-
ment as shown in Figure 8.1-3c, and note that the two
sections are largely equivalent. Both layer replacement
and depth migration are processes aimed at removing
the effects of the complex overburden. However, note
that layer replacement only requires accurate represen-
tation of the overburden (horizon 2 in Figure 8.1-1a),
FIG. 8.1-4. (a) A modeled common-shot gather from the while depth migration requires accurate representation
complicated part of the velocity-depth model in Figure 8.1- of the entire velocity-depth model (Figure 8.1-1a). Also
1a before (a) and after (b) layer replacement. Limitations in note that the output from depth migration is a migrated
modeling with ray tracing cause abrupt terminations along depth section, while the output from layer replacement
the moveout curves and amplitude glitches in (a). In turn, is an unmigrated time section (Figure 8.1-3c). After
these caused the spurious diffractions in (b) during the up- eliminating the complex overburden effect, this section
ward continuation steps. only requires time migration.

flections often can be improved by replacing the velocity


of the water layer with the velocity of the substratum. Prestack Layer Replacement
Consider the zero-offset section shown in Figure
8.1-3a based on the velocity-depth model shown in Fig- As with depth migration, layer replacement after stack
ure 8.1-1a. Poststack layer replacement involves two ex- can remove the effect of a complex overburden, pro-
trapolation steps: vided the input section accurately represents a zero-
offset section. However, the complex overburden causes
(a) The first step in poststack layer replacement in- raypath distortions that generate anomalous, nonhy-
volves downward continuing the wavefield at the perbolic moveout patterns in prestack data. Poststack
surface (Figure 8.1-3a) to the water bottom (hori- depth migration does not produce a completely accu-
zon 2 in Figure 8.1-1a) using the water velocity in rate image of the subsurface, even when the velocity-
extrapolation. The intermediate result is shown in depth model is known accurately. Similarly, poststack
Figure 8.1-3b. Note that in this horizon-flattened layer replacement does not remove the effect of complex
section, the water-bottom reflection is at t = 0, overburden entirely, even if its geometry is known ac-
which means that all receivers are situated on the curately, because the input stacked section differs from
irregular water bottom. If we specified the overbur- the zero-offset section. Nevertheless, based on simple
den velocity or the water-bottom topography incor- ray tracing we can determine whether poststack layer
rectly, then the water-bottom reflection would not replacement can delineate the underlying structure. If
be at t = 0. In this respect, the intermediate sec- the effort yields unsatisfactory results, layer replace-
tion becomes a useful diagnostic tool before moving ment should be performed before stack.
to the next step. In fact, wave-equation datuming Complex moveout is evident in the modeled
actually can be applied layer by layer for struc- common-shot and CMP gathers in Figures 8.1-4a and
tural model restoration. At each layer boundary, by 8.1-5a, respectively. These gathers were modeled from
examining the flatness of the event at t = 0 and the velocity-depth model in Figure 8.1-1a by using ray
1232 Seismic Data Analysis

(c) Downward continue all shots to the same output


datum using the overburden velocity.
(d) Upward continue all shots back to the surface using
the velocity of the substratum.
(e) Sort the data back to common-shot gathers.
(f) Upward continue all receivers back to the surface
using the substratum velocity.

This series of operations eliminates the traveltime


distortions associated with the water bottom, as seen
in the common-shot and CMP gathers (Figures 8.1-4b
and 8.1-5c, respectively). Despite the undesirable effects
caused by the limitations of modeling with ray trac-
ing, the complexity of the reflections associated with
the two events (horizons 3 and 4 in Figure 8.1-1a) was
reduced by layer replacement. Once the complex over-
burden effect is removed, these events have the hyper-
bolic moveout as seen in Figure 8.1-4b. Events A, B,
and C are associated with horizons 2, 3, and 4, respec-
tively, as shown in Figure 8.1-1a. Note that horizon 3
is flat. Therefore, the apex of the hyberbola is at the
near-offset trace (event B). On the other hand, horizon
4 dips down to the right (Figure 8.1-1a). Therefore, the
apex of the hyberbola shifts up-dip (as indicated by the
arrow in Figure 8.1-4b).
Compare the velocity spectra in Figures 8.1-5b and
d, and note the improvement in the velocity estimates
after layer replacement for reflections beneath the com-
plex overburden. The velocity analysis before layer re-
placement yields a good pick for the water-bottom re-
flection A. However, picks B and C, which are associ-
ated with the deeper layers (horizons 3 and 4 in Figure
8.1-1a), are not distinct. Similarly, events B and C also
are indistinct on the maximum correlation curve plot-
ted on the right side of the velocity spectrum (Figure
8.1-5b). After layer replacement (Figure 8.1-5d), note
FIG. 8.1-5. (a) A modeled CMP gather and (b) velocity that the picks (denoted by ×) associated with the three
spectrum from the complicated part of the velocity-depth events are distinct in the velocity spectrum and the cor-
model in Figure 8.1-1a before layer replacement, (c) the relation curve. The dipping water-bottom reflection A
same gather as in (a) and (d) velocity spectrum after layer now has considerably higher moveout velocity, 2300 m/s
replacement.
(Figure 8.1-5d), as compared to the original velocity
1600 m/s (Figure 8.1-5b). The velocity pick for the flat
event B from Figure 8.1-5d is 2000 m/s, which is the
tracing that neither includes diffractions nor properly velocity of the medium above this reflector after layer
modeled amplitudes. The offset range is 50 to 2387.5 m replacement.
with 12.5-m receiver spacing. A total of 437 shot gathers Just as it is true for any wave extrapolation-based
was generated, each with 192 traces. The coverage is process, the wave-theoretical layer replacement tech-
uniform along the line and is 96-fold. nique described here suffers from spatial aliasing. With
Starting with the common-shot gathers, prestack modern data acquisition, we can record with many
layer replacement involves the following steps: channels at small shot and group intervals. Therefore,
spatial aliasing should not be an issue with more recent
(a) Downward continue all receivers to the output da- data. With this provision, prestack layer replacement
tum using the overburden velocity. followed by poststack time migration is a practical al-
(b) Sort the data to common receiver gathers. ternative to depth migration before stack in areas where
Earth Imaging in Depth 1233

FIG. 8.1-6. (a) Velocity-depth model derived from the constant-velocity (1475 m/s) Stolt migration of the CMP stack in
Figure 8.0-3a. (b) The stacked section in Figure 8.0-3a after extrapolating from the surface to the water bottom [Horizon 1
in (a)] using the water velocity. The water-bottom reflection is at t = 0. This is the first step in poststack layer replacement.
(c) Upward continuation of the wavefield in (b) from the water bottom back to the surface using the substratum velocity just
below the water bottom. This is the second step in poststack layer replacement. (d) Time migration of the section in (c). The
replaced water layer has a velocity of 2500 m/s.
1234 Seismic Data Analysis

FIG. 8.1-7. Steps involved in prestack layer replacement. These shot gathers are associated with the stacked data in Figure
8.0-3a. The near-surface model is shown in Figure 8.1-6a. (0) Common-shot gathers at the surface, (1) downward continuation
of receivers to the water bottom with v = 1475 m/s, (2) downward continuation of shots to the water bottom with v = 1475
m/s, (3) upward continuation of shots to the surface with v = 2500 m/s, (4) upward continuation of receivers to the surface
with v = 2500 m/s.
Earth Imaging in Depth 1235

FIG. 8.1-8. CMP gathers associated with common-shot gathers from steps 0 and 4 in Figure 8.1-7 (a) before and (b) after
layer replacement.
1236 Seismic Data Analysis

FIG. 8.1-9. Velocity analysis in the vicinity of midpoint C in Figure 8.0-3a (a) before and (b) after layer replacement.

FIG. 8.1-10. (a) CMP stack derived from the gathers (Figure 8.1-8b) after layer replacement. Compare this with the
conventional stack and poststack layer replacement results in Figures 8.0-3a and 8.1-6c, respectively. (b) Time migration of
the stacked section in (a). Compare this with time migration of the poststack layer replacement result in Figure 8.1-6d.
Earth Imaging in Depth 1237

simple geology is overlain by a single-layer overburden substratum. The section after upward continuation is
with a strong lateral velocity variation, such as an ir- the result of layer replacement and is shown in Figure
regular water bottom. 8.1-6c. After eliminating the effect of the complex over-
The mathematical details of wavefield extrapola- burden, only time migration is needed to image this
tion based on the Kirchhoff integral are given in Sec- section (Figure 8.1-6d).
tion H.1. For poststack layer replacement, we assume The results of prestack layer replacement now are
that the stack is a zero-offset section. A zero-offset sec- examined. Starting with the common-shot gathers and
tion is equivalent to the exploding reflectors model, pro- using the velocity-depth model in Figure 8.1-6a, we per-
vided the medium velocity is halved (Section 4.0). We form downward and upward continuations of common-
assume that sources already are situated along the re- shot and common-receiver gathers. The intermediate
flectors in the subsurface. Thus, we only need to move steps for selected common-shot gathers are shown se-
the receivers from one datum to another during down- quentially in Figure 8.1-7. Note the arrival time of the
ward and upward continuation steps. For prestack layer water-bottom reflection at t = 0 on the gathers in step 2
replacement, each common-shot and common-receiver after downward continuing both the shots and receivers
gather is extrapolated, independently. In particular, the
to the water bottom.
wavefield at a point on the output datum is computed
A selected set of CMP gathers before and after
using all the traces in the input gather. The output
layer replacement, sorted from the shot gathers in steps
gather should be computed beyond the lateral extent
0 and 4, respectively, is shown in Figure 8.1-8. Because
of the input gather to prevent possible loss of steeply
the data contain strong diffracted multiples, it is diffi-
dipping events. For prestack layer replacement, the ve-
locity used in the extrapolation is that of the medium cult to evaluate the layer replacement results. As in any
between the input datum and the output datum. other one-way extrapolation technique, wave-equation
datuming treats multiples as primaries. From the veloc-
ity analysis in Figure 8.1-9, we are encouraged by the
picking we can do on the velocity spectrum after layer
Field Data Example replacement. However, the stack has the final say and
is shown in Figure 8.1-10a. Again, once time-migrated
Now consider the field data example in Figure 8.0-3a. (Figure 8.1-10b), the result of layer replacement should
Layer replacement will be done on these data before provide an accurate image of the substratum, free from
and after stack to remove the effect of the irregular wa- the effects of complex overburden (compare Figure 8.1-
ter bottom. First, we must define the geometry of the 10b with Figures 8.0-3d and 8.1-6d).
overburden, in this case, the water-bottom topography. The interpretation of the unconformity T from the
Since the overburden velocity is constant (1475 m/s), we results of prestack layer replacement (Figure 8.1-10b)
migrate the CMP-stacked section (Figure 8.0-3a) using closely agrees with the proposed velocity-depth model
the constant-velocity Stolt method, digitize the water in Figure 8.0-3c. The unconformity is continuous be-
bottom, and convert it to depth as shown in Figure 8.1-
neath midpoint C, where there is a structural high. Be-
6a.
low and to the left of midpoint A, the unconformity
First, consider poststack layer replacement. We
extends down to the right with some tensional faults
downward continue the CMP stack (Figure 8.0-3a) by
into the ancient continental shelf below midpoint B,
using the water velocity from the surface to the wa-
then finally extends to the continental slope of that age
ter bottom defined in Figure 8.1-6a to get the horizon-
flattened section in Figure 8.1-6b. As usual, we as- to the right of the structural high below midpoint C.
sume that the CMP stack is a zero-offset wavefield. The The present-day situation is just the opposite, with the
water-bottom reflection is approximately flat and is sit- continental shelf on the right. Note that shallow reflec-
uated at t = 0, which indicates that the velocity-depth tor R now is conformable with the rest of the shallow
model for the water layer in Figure 8.1-6a is fairly ac- section above the unconformity.
curate. Although incorrect in shape, substratum reflec- For the irregular water-bottom case, it was shown
tions are quite continuous on this section, indicating that prestack layer replacement followed by NMO cor-
that we achieve a focusing effect by downward extrap- rection, stack, and time migration after stack with
olation. The bottom of the section reflects the mirror depth conversion basically is equivalent to depth migra-
image of the water-bottom topography. tion before stack. The situation may be more compli-
The next step is to take this wavefield back to the cated when dealing with a complex overburden problem
surface, this time using the velocity of the substratum that involves more than one interface. In principle, da-
(2500 m/s). Luckily, the velocity derived from the seis- tuming can be applied layer by layer and the effect of
mic data is fairly constant across the section for the the overburden can be removed. Nevertheless, there is
1238 Seismic Data Analysis

no practical alternative to depth migration under those Figure 8.2-4 are the normal-incidence rays from each of
circumstances. the three layer boundaries — top-salt, base-salt and the
flat reflector below, and the computed zero-offset trav-
eltimes. Superposition of the modeled traveltime tra-
8.2 2-D POSTSTACK DEPTH MIGRATION jectories yields the zero-offset traveltime section associ-
ated with the velocity-depth model as shown in Figure
8.2-4. The zero-offset section in Figure 8.2-1 represents
Figure 8.2-1 shows a velocity-depth model for a salt
a modeled zero-offset wavefield, whereas the zero-offset
pillow. The aspect ratio of the horizontal and vertical
section in Figure 8.2-4 represents a modeled zero-offset
axes is 1; hence, the diagram exhibits the true shape
traveltime profile.
of the diapiric structure. The model can be treated in
three parts — the constant-velocity overburden above
the salt, the salt diapir itself, and the substratum that
includes the flat reflector below. So far as the flat reflec- Image Rays and Lateral Velocity Variations
tor is concerned, the salt diapir constitutes a complex
overburden structure with strong lateral velocity varia- Normal-incidence rays are associated with zero-offset
tions. Note the significant velocity contrast across the traveltimes and therefore can be used to examine the de-
top-salt boundary and the undulating reflector geome- gree of complexity in velocity-depth models as demon-
try of the base-salt boundary — both give rise to ray strated in Figure 8.2-4. For a quantitative assessment
bending that can only be handled by imaging in depth. of lateral velocity variations, however, image rays need
A total of 154 shot records were modeled along the to be examined as shown in Figure 8.2-5. By definition,
lateral extent of the velocity-depth model in Figure 8.2- image rays emerge at the right angle to the surface.
1 using the two-way acoustic wave equation. Shot and As shown in Figure 8.0-11, the lateral shift between the
receiver group intervals both are 50 m, and the trace point of departure of the image ray at the reflector posi-
spacings of the CMP-stacked and zero-offset sections are tion and the point of emergence of the image ray at the
25 and 50 m, respectively. In Figure 8.2-1, the velocity- surface provides a measure of lateral velocity variation.
depth model, the CMP-stacked and zero-offset sections Consider the image rays departing from the top-
have been displayed with the correct lateral position salt layer boundary in Figure 8.2-5. These rays show no
with respect to one another. Note the focusing and de- lateral shift, and therefore, imaging the top-salt bound-
focusing of the reflection amplitudes associated with the ary does not require depth migration; instead, it can
base-salt boundary on the zero-offset and CMP-stacked be achieved by time migration. The image rays from
sections. the base-salt boundary, however, show significant lat-
Selected shot records are shown in Figure 8.2-2. eral shifts, especially beneath the flanks of the diapir.
Each shot record consists of 97 channels associated with The stronger the lateral velocity variations, the more
a split-spread geometry. The shot interval is 50 m and the lateral shifts in image rays. This behavior of the
the receiver group interval is 50 m. The maximum off- image rays indicate that the lateral velocity variations
set is 2350 m. Note the complexity of reflection times on caused by the salt diapir require depth migration to
shot records located over the salt diapir and the vari- image the base-salt boundary, accurately.
ations in reflection amplitudes caused by the focusing The image rays associated with the flat reflector
and defocusing of rays. below the salt diapir also show significant lateral shifts
The complexity of reflection traveltimes surpris- (Figure 8.2-5). Again, this reflector can only be imaged
ingly is not as apparent on the CMP gathers (Figure 8.2- accurately by depth migration, rather than time migra-
3) as it is on the shot records (Figure 8.2-2), although tion. Note that image rays do not sample the reflector
variations in reflection amplitudes can be observed with boundaries uniformly — there are regions that contain
ease. CMP stacking, however, clearly shows the effect densely and sparsely populated image rays.
of nonhyperbolic moveout (Figure 8.2-1). Compare with In principal, an earth image in depth can be ob-
the zero-offset section and note the traveltime and am- tained by first migrating a stacked section in time, then
plitude distortions on the CMP-stacked section. Depar- converting the time-migrated section to depth along
ture of the stacked section from the true zero-offset sec- image rays using the appropriate velocity-depth model
tion imposes a limitation on the accuracy of the image (Hubral, 1977; Larner et al., 1981). This ray-theoretical
we get from poststack depth migration. two-step depth migration to obtain an earth image in
We now closely examine the degree of ray bending depth is rarely used in practice. However, it is common
at layer boundaries in the salt diapir model. Shown in practice to perform time-to-depth conversion of time
(text continues on p. 1244)
Earth Imaging in Depth 1239

FIG. 8.2-1. (Top) An earth model in depth with a salt diapir; (middle) CMP-stacked section derived from the modeled
prestack data in Figure 8.2-3; (bottom) the modeled zero-offset section. The stacked and zero-offset sections are appropriately
aligned in the lateral direction with respect to the earth model above. Trace spacings in the stacked and zero-offset sections
are 25 m and 50 m, respectively. No amplitude scaling has been applied to the sections. The aspect ratio of the horizontal
and vertical axes in the velocity-depth model is 1.
1240 Seismic Data Analysis

FIG. 8.2-2. A selection of modeled common-shot gathers associated with the earth model in Figure 8.2-1. Numbers on top
represent the nearest CMP location. No amplitude scaling has been applied to the data.
Earth Imaging in Depth 1241
1242 Seismic Data Analysis

FIG. 8.2-4. (Left column) Velocity-depth model as in Figure 8.2-1 with normal-incidence rays from each of the layer bound-
aries; (right column) corresponding zero-offset traveltimes. The bottom-right frame shows the superposition of the zero-offset
traveltimes associated with the three layer boundaries. The vertical axis in the traveltime sections is two-way zero-offset time.
Earth Imaging in Depth 1243

FIG. 8.2-5. Image rays associated with the salt diapir model in Figure 8.2-1.
1244 Seismic Data Analysis

horizons using image rays. Specifically, 3-D volume of migration was the true model. Note, for instance, the
stacked data first is migrated in time and selected time subtle distortions on the base-salt event and the not-
horizons are interpreted. These time horizons are then so-flat reflector in the depth-migrated section. This is
converted to depth horizons along image rays, again, a direct consequence of the fact that the CMP-stacked
using an appropriate velocity-depth model. Creating section is only a close representation of the zero-offset
depth structure maps using this procedure is called map wavefield in the presence of strong lateral velocity vari-
migration. ations associated with complex overburden structures.
In conclusion, by examining the behavior of image Since poststack migration algorithms are based on the
rays through the salt diapir model, we can judge as to zero-offset wavefield theory (Section 4.1), application of
which layer boundary requires imaging in depth (Fig- zero-offset migration to a CMP-stacked section would
ure 8.2-5). The image rays down to the top-salt bound- produce less-than-ideal results. To circumvent this defi-
ary are not distorted laterally; therefore, time migra- ciency in CMP stacking, and to correctly image the sub-
tion is adequte for imaging the overburden above the stratum that includes the base-salt boundary and the
salt diapir. Singificant ray bending, however, takes place flat reflector, strictly, one needs to do prestack depth
at the top-salt boundary; this results in lateral distor- migration. Prestack depth migration is reviewed in the
tions of the image rays down to the base-salt boundary next section; however, for comparison, results of the
and the deeper reflector. Depth migration, therefore, is zero-offset, poststack and prestack depth migrations are
needed for accurate imaging of the base-salt boundary shown in Figure 8.2-8. Note that prestack depth migra-
and the subsalt region. tion produces an image that is free of the distortions
observed on the image produced by poststack depth
migration. Imaging accuracy is similar to that of the
Time versus Depth Migration zero-offset section.
A way to minimize departure of a stacked section
We continue our discussion with the salt diapir model from a zero-offset section — that is, to minimize trav-
data. Figure 8.2-6 shows time and depth migrations of eltime and amplitude distortions caused by nonhyper-
the zero-offset section associated with the salt diapir bolic moveout during CMP stacking, is to use partial
model shown in Figure 8.2-1. Ignore the dispersive noise stacking. Figure 8.2-9 shows a portion of a full-fold
in the vicinity of the top-salt event on the migrated sec- CMP-stacked section that contains a salt diapir struc-
tions; this is caused by the differencing approximations ture. (The CMP fold is 30.) Note the spurious horst-like
made to the differential operators in the implicit scheme structure at the base-salt boundary B. This data set is
used here. While the top-salt boundary is imaged ac- from the Red Sea where such structures are common.
curately by both time and depth migration, note the Hence, at first, this horst structure appears to be ge-
overmigration exhibited by the base-salt and the deeper ologically plausable. Note, however, the horst block at
event in the time-migrated section. Depth migration, on the base-salt is replaced with a continuous event on the
the other hand, images these two reflectors, accurately, single-fold near-offset section. What may seem to be a
if the velocity-depth model input to depth migration is horst block actually is no more than a manifestation
the true model as in the case of Figure 8.2-6. Although of amplitude and traveltime distortions caused by the
we used the true velocity-depth model, time migration full-fold stacking that spans the entire offset range of
produced the incorrect image of the base-salt boundary the recorded data. The conclusion we can draw from
and the subsalt region. Time migration algorithms do this observation is that it is not necessarily the full-fold
not include the term that accounts for strong lateral ve- stack, but the near-offset section that better resembles
locity variations as manifested by the image rays shown a zero-offset section. Hence, it is the near-offset section,
in Figure 8.2-5. On the other hand, depth migration al- and not the full-fold section, that is an appropriate in-
gorithms include this term and thus are able to correct put to poststack depth migration. An immediate ob-
for the lateral shift in image rays. jection to this conclusion is that the near-offset section
Figure 8.2-7 shows time and depth migrations of contains a significant amount of multiple energy that
the CMP-stacked section associated with the salt di- was naturally attenuated during CMP stacking by way
apir model shown in Figure 8.2-1. Again, the top-salt of velocity discrimination between primaries and mul-
boundary is imaged accurately by both time and depth tiples. Also, note the loss of continuity of the base-salt
migration. As expected, however, time migration fails to event on the near-offset section; in contrast, CMP stack-
produce a correct image of the base-salt boundary and ing improves the signal-to-noise ratio. A way to benefit
the deeper reflector. Note that even depth migration from the power of CMP stacking to attenuate multiples
fails to image these two reflectors with sufficient accu- and improve the signal-to-noise ratio while circumvent-
racy, although the velocity-depth model input to depth ing the nonhyperbolic moveout effect is to do partial
Earth Imaging in Depth 1245

FIG. 8.2-6. Time and depth migrations of the zero-offset section associated with the salt diapir model. Trace spacing is 50
m, which in this case, gives rise to some spatial aliasing noise along the steep flanks of the salt diapir as seen on the migrated
sections.
1246 Seismic Data Analysis

FIG. 8.2-7. Time and depth migrations of the CMP-stacked section associated with the salt diapir model. Trace spacing is
25 m.
Earth Imaging in Depth 1247

FIG. 8.2-8. Depth migrations of the zero-offset, CMP-stacked, and prestack data associated with the salt diapir model using
the correct velocity-depth model.

stacking. By a simple series of tests, one can judge as velocity-depth model causes a poor image produced by
to what portion of the cable — near offsets, mid-range not just poststack depth migration, but also by zero-
offsets or far offsets, provides this optimum stack as in- offset and prestack depth migration.
put to poststack depth migration. It is important to
note that, while poststack depth migration may require
a stack based on a subset of offsets, prestack depth mi- Iterative Depth Migration
gration requires all offsets.
We must remind ourselves that an accurate im- Historically, depth migration has been used in an it-
age from depth migration is attainable only when the erative manner to obtain an earth image in depth
velocity-depth model is correctly defined, independent from CMP-stacked data. When performed iteratively,
of the input data type — zero-offset, stack, or prestack. depth migration is done using an initial velocity-
Figure 8.2-10 shows results of the zero-offset, post- and depth model and the result is interpreted for the layer
prestack depth migrations of the salt diapir model data boundaries included in the model. The velocity-depth
using an incorrect velocity-depth model. An incorrect model then is modified accordingly and depth migration
1248 Seismic Data Analysis

FIG. 8.2-9. A portion of a full-fold CMP-stacked section (top), the near-offset section (middle), and partial-fold CMP-stacked
section (bottom) in which the nearest one-third of the cable was used. B is the base-salt event.
Earth Imaging in Depth 1249

FIG. 8.2-10. Depth migrations of the zero-offset, CMP-stacked, and prestack data associated with the salt diapir model
using an incorrect velocity-depth model.

is performed once more. The process is continued until eled zero-offset traveltimes match with the observed
convergence is achieved. reflection traveltimes on the stacked data associated
Convergence means that what is input to depth with the layer boundaries included in the velocity-depth
migration as the velocity-depth model matches with model. Convergence and consistency are the two neces-
the velocity-depth model inferred from the output from sary, but not sufficient, conditions for an earth model
depth migration. With iterative depth migration, we to be certified as a valid, geologically plausable solu-
know that we have achieved convergence when differ- tion from seismic inversion. For a velocity-depth model
ences between the velocity-depth models from two con- to be valid, a further requirement is that it also needs
secutive iterations are minimal. We shall demonstrate to be consistent with prestack data; thus, the strategic
that, by way of convergence, the final velocity-depth requirement for doing prestack depth migration.
model from iterative depth migration, albeit not guar- In this section, we shall examine iterative depth
anteed to be accurate, can be made at least consistent migration for three data types — zero-offset, CMP
with the input data. Consistency means that the mod- stack, and prestack. We shall consider six different
1250 Seismic Data Analysis

FIG. 8.2-11. Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model A (same as the true
velocity-depth model as in Figure 8.2-1).

initial velocity-depth models, each having errors in The result indicates that the geometry of the base-
layer velocities and/or reflector geometries. Conver- salt boundary and the flat reflector below has changed
gence rates and the end results for each starting model (top of the right column). Interpret all the layer bound-
will be evaluated and some practical conclusions about aries — top-salt, base-salt, and the deeper reflector,
iterative depth migration will be drawn from these ex- from the result of depth migration and create an up-
periments. dated velocity-depth model (center of the left column).
Use this new model and perform depth migration once
Iteration with Zero-Offset data more (center of the right column). Interpret the new im-
age from depth migration, again, for all the three layer
Depth migration of the zero-offset section in Figure 8.2- boundaries and obtain an updated velocity-depth model
1 using Model A shown in Figure 8.2-11 produces an (bottom of the left column). Finally, perform depth mi-
accurate image of the salt diapir and yields the same gration for the third time (bottom of the right column)
output model as the input model in a single iteration. and interpret for the three layer boundaries, once more.
This happened because the input data set is the true After this third iteration, we find that the velocity-
zero-offset section and the input velocity-depth model depth model does not change from the previous iter-
(Model A in Figure 8.2-11) is the same as the true ation; hence, convergence is achieved. Nevertheless, the
velocity-depth model (Figure 8.2-1). final solution has converged to a velocity-depth model
Now consider Model B in Figure 8.2-12 in which the
that is different from the true model. We conclude from
salt velocity and the base-salt boundary are specified in-
this experiment that convergence and consistency, al-
correctly. There are practical reasons behind these de-
beit required for a model to be certified as a valid so-
liberate errors introduced into the model. In practice,
often the top-salt boundary is determined with reason- lution from depth migration, do not guarantee that the
able accuracy by way of time migration. However, ac- solution is the true model. The result of iterative depth
curate delineation of the base-salt boundary is almost migration obviously is dictated by the parameters of the
impossible with time migration. Additionally, the salt initial velocity-depth model.
velocity may vary because of dolomite or shale intru- Consider now an initial model in which the ge-
sions, and these variations are often difficult to deter- ometry of only one reflector — that associated with
mine. the base-salt boundary, is in error while layer ve-
Start with Model B as the initial velocity-depth locities are specified correctly (Figure 8.2-13). This
model and perform depth migration (Figure 8.2-12). may be possible in practice if there is an abundance
(text continues on p. 1258)
Earth Imaging in Depth 1251

FIG. 8.2-12. Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model B that contains errors in
salt velocity and base-salt reflector geometry. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is
superimposed on the final velocity-depth model (bottom of left column).
1252 Seismic Data Analysis

FIG. 8.2-13. Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model C that contains errors in
base-salt reflector geometry. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is superimposed on
the final velocity-depth model (bottom of left column).
Earth Imaging in Depth 1253

FIG. 8.2-14. Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model D that contains errors in
salt velocity and base-salt reflector geometry. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is
superimposed on the final velocity-depth model (bottom of left column).
1254 Seismic Data Analysis

FIG. 8.2-15. Part 1: Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model E that contains
errors in overburden and salt velocities, and top-salt and base-salt reflector geometries. For comparison, the true velocity-depth
model (Model A in Figure 8.2-11) is superimposed on the final velocity-depth model (bottom of left column).
Earth Imaging in Depth 1255

FIG. 8.2-15. Part 2: Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model E that contains
errors in overburden and salt velocities, and top-salt and base-salt reflector geometries. For comparison, the true velocity-depth
model (Model A in Figure 8.2-11) is superimposed on the final velocity-depth model (bottom of left column).
1256 Seismic Data Analysis

FIG. 8.2-16. Iterative depth migration using the zero-offset section as in Figure 8.2-1 and Model F that contains errors
in top-salt and base-salt reflector geometries. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is
superimposed on the final velocity-depth model (bottom of left column).
Earth Imaging in Depth 1257

FIG. 8.2-17. Top left: initial velocity-depth model for iterative depth migration; top right: final image from iterative depth
migration; center left: final model from iterative depth migration; center right: true model; bottom left: zero-offset traveltime
section created by forward modeling using normal-incidence rays through the final model (center left) from iterative depth
migration; bottom right: zero-offset wavefield section input to prestack depth migration.
1258 Seismic Data Analysis

of well data in the area of investigation that enables a Figure 8.2-18 shows the initial velocity-depth mod-
reliable specification of layer velocities. els (Figures 8.2-11 through 8.2-16) that contain various
It often is preferable to specify a simple, in this case types of errors in model parameters. Figure 8.2-19 shows
a flat reflector, geometry for the base-salt boundary. the final solutions derived from iterative depth migra-
We shall let iterative depth migration produce a final tion using the initial models in Figure 8.2-18. Note that
geometry for the base-salt boundary. Now, start with those initial models with correctly specified layer veloc-
Model C as the initial velocity-depth model and perform ities but incorrectly defined reflector geometries (Mod-
iterative depth migration (Figure 8.2-13). Convergence els A, C, and F) converge to the true model (Model
is achieved after three iterations. The final solution has A). Those initial models with incorrect layer velocities
converged to a velocity-depth model that is fairly close (Models B, D, and E), however, do not converge to the
to the true model. We conclude that true geometry of true model.
reflectors can be recovered by iterative depth migration Zero-offset traveltime sections computed from the
provided layer velocities are correctly specified in the final velocity-depth models are shown in Figure 8.2-20.
initial velocity-depth model. Compare these traveltime sections with the input zero-
The case of Model D shown in Figure 8.2-14 is sim- offset wavefield section (Figure 8.2-11), and note that
ilar to that of Model B (Figure 8.2-12). While the layer
they all are consistent with the latter. Consistency, how-
velocity for the salt diapir in Model B has been speci-
ever, does not guarantee that the final solution from it-
fied erroneously too low, it has been specified in Model
erative depth migration yields the true model. In prac-
D erroneously too high. As a result, the final velocity-
tice, we will never know which one of the final solutions
depth model, once again, departs from the true model,
shown in Figure 8.2-19 corresponds to the true model.
significantly.
Sometimes time migration may yield an inaccurate
image of the top-salt boundary because of the incorrect
overburden velocity field. Such is the case in Model E Iteration with CMP-Stacked Data
(Figure 8.2-15) where not only the overburden velocity,
and therefore, the top-salt boundary are incorrect, but Depth migration of the CMP-stacked section in Figure
also the salt velocity and the base-salt boundary are in 8.2-1 using Model A required two iterations to achieve
error. This model has too many errors in layer velocities convergence (Figure 8.2-21). The resulting image and
and reflector geometries. It has taken four iterations to the model inferred from it have some inaccuracies. It
achieve convergence. The final solution radically departs appears that starting with the true model (Model A),
from the true model. iterative poststack depth migration does not exactly
If the initial model contains significant errors but reproduce the true model. For the zero-offset section,
only in reflector geometries, and layer velocities are
starting with the true model, convergence was achieved
specified correctly and thus are not altered from one
after one iteration and the resulting model was the same
iteration to the next, then convergence to the true
as the input model (Figure 8.2-11). Again, the reason
velocity-depth model can be achievable (Figure 8.2-16).
for the less-than-ideal performance of poststack depth
To recap the results of the analysis of the six mod-
migration is that the CMP stack is only an approxima-
els (Figures 8.2-11 through 8.2-16), refer to Figure 8.2-
17. Starting with an initial velocity-depth model (top tion to the zero-offset wavefield. The more the stacked
left), which contains some errors in model parameters, section departs from the zero-offset wavefield, the more
perform iterative depth migration. After n iterations, the final model will depart from the true model.
obtain an earth image in depth (top right) that corre- Now consider Model D in Figure 8.2-22 in which
sponds to an earth model in depth (center left) that is the salt velocity and the base-salt boundary are in-
different from the true model (center right). Neverthe- correctly specified. After the third iteration with the
less, forward modeling of zero-offset traveltimes (bot- stacked section, we find that convergence is achieved.
tom left) using the estimated model (center left) from Nevertheless, the final solution has converged to a
iterative depth migration yields results that are consis- velocity-depth model that is different from the true
tent with the traveltimes on the zero-offset wavefield model.
section (bottom right) used as input to iterative depth If the initial model contains errors in reflector ge-
migration. We have met the convergence and consis- ometries only, and layer velocities are specified cor-
tency criteria for the final solution from iterative depth rectly, then convergence using the stacked section to the
migration, but we have only obtained a solution, and true velocity-depth model is nearly achievable (Figure
not the solution — the true model (Figure 8.2-11). 8.2-23).
(text continues on p. 1265)
Earth Imaging in Depth 1259

FIG. 8.2-18. Initial velocity-depth models used in iterative depth migrations illustrated in Figures 8.2-11 through 8.2-16.
1260 Seismic Data Analysis

FIG. 8.2-19. Final velocity-depth models from iterative depth migrations illustrated in Figures 8.2-11 through 8.2-16.
Earth Imaging in Depth 1261

FIG. 8.2-20. Zero-offset traveltime sections computed from the final velocity-depth models shown in Figure 8.2-19.
1262 Seismic Data Analysis

FIG. 8.2-21. Iterative depth migration using the CMP-stacked section as in Figure 8.2-1 and Model A (same as the true
velocity-depth model as in Figure 8.2-1).
Earth Imaging in Depth 1263

FIG. 8.2-22. Iterative depth migration using the CMP-stacked section as in Figure 8.2-1 and Model D that contains errors
in salt velocity and base-salt reflector geometry. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is
superimposed on the final velocity-depth model (bottom of left column).
1264 Seismic Data Analysis

FIG. 8.2-23. Iterative depth migration using the CMP-stacked section as in Figure 8.2-1 and Model F that contains errors
in top-salt and base-salt reflector geometries. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is
superimposed on the final velocity-depth model (bottom of left column).
Earth Imaging in Depth 1265

FIG. 8.2-24. Iterative depth migration using the prestack data as in Figure 8.2-2 and Model A (same as the true velocity-
depth model as in Figure 8.2-1).

Iteration with Prestack Data Iteration in Practice

To compare with iterative depth migration applied to In this section, we tested iterative depth migration of
zero-offset and CMP-stacked data, we shall examine zero-offset, CMP-stack and prestack data using six dif-
the performance of iterative depth migration applied ferent initial velocity-depth models. Aside from the true
to prestack data. Techniques for prestack depth migra- model, each model was corrupted by errors in layer ve-
tion itself, however, will be reviewed in the next section. locities and/or reflector geometries. We summarize the
Depth migration of the prestack data using the true results of the model experiments as follows:
model (Model A) shown in Figure 8.2-11 required only
one iteration to achieve convergence (Figure 8.2-24). (a) When will the image from iterative depth migra-
Next, consider Model D in Figure 8.2-25 in which tion converge to the true velocity-depth model?
the salt velocity and the base-salt boundary are in- The answer to this question is many-fold, and it
depends on the type of input data and errors in
correctly specified. After the third iteration with the
the initial velocity-depth model (Figure 8.2-27). If
prestack data, we find that convergence is achieved.
the input data set is zero-offset section and if the
Nevertheless, the final solution has converged to a
input velocity-depth model is the true model (an
velocity-depth model that is different from the true impossible case in practice), then the answer is yes.
model. For prestack data, the answer also is yes, whereas
If the initial model contains errors in reflector ge- for stacked data, the answer is nearly. If the ini-
ometries only, and layer velocities are specified cor- tial velocity-depth model is in error of reflector ge-
rectly, then convergence using the prestack data to true ometries only, the answer is most likely, very likely,
velocity-depth model is achievable (Figure 8.2-26). and likely for zero-offset, prestack and stacked data,
1266 Seismic Data Analysis

FIG. 8.2-25. Iterative depth migration using the prestack data as in Figure 8.2-2 and Model D that contains errors in
salt velocity and base-salt reflector geometry. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is
superimposed on the final velocity-depth model (bottom of left column).
Earth Imaging in Depth 1267

FIG. 8.2-26. Iterative depth migration using the prestack data as in Figure 8.2-2 and Model F that contains errors in top-salt
and base-salt reflector geometries. For comparison, the true velocity-depth model (Model A in Figure 8.2-11) is superimposed
on the final velocity-depth model (bottom of left column).
1268 Seismic Data Analysis

FIG. 8.2-27. Performance of iterative depth migration for different types of input data and errors in the initial velocity-depth
model.

respectively. Finally, if the initial velocity-depth (c) In iterative depth migration, only one set of param-
model is in error of layer velocities, no matter what eters — either layer velocities or reflector geome-
the input data set is, the answer is invariably no. tries, should be modified from one iteration to the
Needless to say, iterative depth migration never next. Changing the parameter type at some inter-
guarantees that the solution converges to a true mediate iteration will only divert the solution to a
velocity-depth model. different end result and will cause the convergence
to that end result to take longer. It is advisable to
(b) It is wrong to discontinue an iterative application
keep layer velocities unaltered from one iteration
of depth migration without achieving convergence
to the next, and only modify reflector geometries
— the intermediate output does not represent a by interpreting the output image from depth mi-
valid earth image in depth and neither does it in- gration.
fer a valid earth model in depth. Zero-offset trav- (d) Number of iterations depends on how much the ini-
eltimes associated with this earth model are not tial model departs from the true model. If the ini-
consistent with the traveltimes in the input sec- tial model contains errors in layer velocities, only,
tion. fewer iterations are required than if the model con-
Earth Imaging in Depth 1269

FIG. 8.2-28. (a) Portion of a CMP-stacked section, and (b) the image from time migration.
1270 Seismic Data Analysis

FIG. 8.2-29. Part 1: Iterative poststack depth migration applied to field data shown in Figure 8.2-28. See text for details.
Earth Imaging in Depth 1271

FIG. 8.2-29. Part 2: Iterative poststack depth migration applied to field data shown in Figure 8.2-28. See text for details.
1272 Seismic Data Analysis

FIG. 8.2-30. (a) Final velocity-depth model derived from the iterative poststack depth migration procedure (Figure 8.2-29),
(b) poststack depth migration of the data in Figure 8.2-28a, (c) modeled zero-offset traveltimes superimposed on the stacked
section as in Figure 8.2-28a.
Earth Imaging in Depth 1273

tains errors in reflector geometries. If the model depth image derived from it are shown in Figures 8.2-
contains errors both in layer velocities and reflector 30a,b. Convergence criterion can be verified by normal-
geometries, a large number of iterations is required incidence modeling of the zero-offset traveltimes that
to achieve convergence. correspond to the reflectors associated with the layer
(e) Iterative depth migration converges to an an- boundaries included in the velocity-depth model. Su-
swer, which almost never corresponds to the true perimpose the modeled traveltimes on the unmigrated
velocity-depth model. Therefore, the final velocity- stacked section and note that the modeled and the ac-
depth model estimated from iterative depth migra- tual traveltimes are in good agreement.
tion needs to be calibrated to well data. Specifi- While the consistency of the modeled and actual
cally, layer velocities are adjusted so as to match at zero-offset traveltimes verifies that the final velocity-
well locations the depth values of the layer bound- depth model (Figure 8.2-30a) from poststack iterative
aries included in the final velocity-depth model depth migration meets the convergence criterion, the
with the well tops corresponding to those layer model is not guaranteed to be accurate. In fact, as de-
boundaries. mostrated by the synthetic data examples in this sec-
tion, there exists not just one but a multiple number
Figure 8.2-28 shows a portion of a CMP-stacked of velocity-depth models that are consistent with the
section and its time migration. The deepest layer stacked data. An acceptable model is that which also is
with an abundance of diffractions represents salt with consistent with prestack data; thus, one strategic rea-
anhydrite-dolomite rafts. The objective is to obtain an son for prestack imaging is resolving the uncertainty
accurate geometry for the base-salt boundary repre- in the acceptable velocity-depth models and reducing
sented by the strong, deepest reflection. the many possible models to a few that are geologically
To illustrate poststack iterative depth migration, plausable.
start with a simple, horizontally layered earth model
with constant layer velocities as the initial velocity-
depth model (Figure 8.2-29a) and migrate the stacked
8.3 2-D PRESTACK DEPTH MIGRATION
section to obtain the depth image shown in Figure 8.2-
28b. Superimpose the flat depth horizons associated
with the initial velocity-depth model on this image sec- In this section, we shall review two methods of prestack
tion, and note that the reflector geometries implied by depth migration applied to 2-D data. These methods
the depth image show discrepancy with the flat depth both have historical and conceptual significance. An
horizons. Discard the latter and interpret the image sec- early method of prestack depth migration is based on
tion to derive a set of structurally consistent depth hori- downward continuation of sources and receivers (Sec-
zons (Figure 8.2-28c). This completes the first iteration tion D.1). The method commonly is known as shot-
of poststack depth migration and model updating. geophone migration. A common-shot gather represents
Next, keep the layer velocities the same as for the a wavefield and thus can be extrapolated in depth at dis-
initial velocity-depth model (Figure 8.2-28a), but use crete intervals. As a result, receivers are lowered from
the updated depth horizons (Figure 8.2-28c) to build one depth level to the next. By invoking reciprocity,
a new velocity-depth model as shown in Figure 8.2- we may also consider a common-receiver gather repre-
28d. Now, perform poststack depth migration once more senting a wavefield. Thus, by extrapolating a common-
and note that the reflector geometries implied by the receiver gather, sources are lowered from one depth
depth image (Figure 8.2-28e) are still in discrepancy level to the next. By alternating between extrapola-
with the intermediate velocity-depth model (Figure 8.2- tion of common-shot and common-receiver gathers at
28d). Discard the depth horizons and reinterpret them each depth level, all sources and receivers are lowered
from the image section (Figure 8.2-28f). This completes from the surface to each of the reflectors in the sub-
the second iteration of poststack depth migration and surface. While sources and receivers are lowered verti-
model updating. cally downward from one depth level to the next, the
Repeat the steps for model updating, poststack recorded waves are back-propagated along the raypaths
depth migration and re-interpreting the depth horizons from source to a reflector back to receiver locations at
for the third (Figures 8.2-28g,h,i), and fourth time (Fig- the surface. When sources and receivers are lowered to
ures 8.2-28j,k,l) until convergence is achieved; that is, the reflector, they coincide and the traveltime dimin-
the velocity-depth model input to depth migration is ishes to naught. This satisfies the imaging condition.
consistent with the reflector geometries inferred by the When the maximum depth of extrapolation is
depth image. The final velocity-depth model and the reached, traces at zero-offset from each of the resulting
1274 Seismic Data Analysis

common-shot gathers are extracted and placed side-by- A bonus effect of prestack migration — be it time
side to produce the image from prestack depth migra- or depth, is its ability to attenuate multiples. Note, for
tion. Nonzero-offset traces are abandoned since all pri- instance, in Figure 8.3-3a the pegleg (m) on the left
mary energy has collapsed to zero-offset provided the flank of the diapir. This multiple has been retained dur-
velocity-depth model is correct. ing CMP stacking and moved with the primary velocity
Note that in practical implementation of shot- during poststack migration (Figure 8.3-3b). However, it
geophone migration, data need to be sorted from one is not on the section derived from prestack depth migra-
gather type to another (common-shot and common- tion. To explain this bonus effect of prestack migration,
receiver) at each depth level. This sorting operation refer to the selected shot records after prestack depth
increases the cost of the method, formidably. Another migration (Figure 8.3-4). Since the migration velocity
practical aspect of this method is that missing near- field is associated with the primary events, the primary
offset traces in common-shot gathers are filled in with energy has focused on the zero-offset traces, while the
zero traces before downward continuation is started. multiple energy has been dispersed to nonzero-offset
A more popular method of prestack depth migra- traces. The focusing of the primary energy also is seen
tion is based on migration of shot records, individually. on the common-receiver gathers after migration (Figure
This is plausable since a shot record is a wavefield gener- 8.3-5). The image from prestack depth migration (Fig-
ated by a single source. The method commonly is known ure 8.3-3) is formed by compiling the zero-offset traces
as shot-profile migration. The migrated shot records from the shot records after migration (Figure 8.3-4). All
are then sorted into common-receiver gathers. Finally, the remaining nonzero-offset traces in the shot records
traces in each receiver gather are summed to construct are simply abandoned.
the image below the receiver location. By placing the The quality of focusing at zero offset by shot-
traces that result from this summation side by side, we geophone migration understandably depends on the ac-
obtain the image from shot-profile migration. curacy of the velocity field used in migration. Erro-
A practical advantage of shot-profile migration is neously too low or too high velocities would cause a par-
its ability to handle irregularities in recording geome- tial collapse of the primary energy to zero offset. Note
try. Since each shot record is migrated independently, the curvature along some of the events in the vicinity
missing shots, duplicate shots or just irregularities in of the zero-offset traces on the shot records (Figure 8.3-
shot spacing are irrelevant. 4) and receiver gathers (Figure 8.3-5). Some exhibit a
curvature upward (erroneously low velocity above the
event) and some exhibit a curvature downward (erro-
Shot-Geophone Migration neously high velocity above the event). Theoretically,
by measuring this curvature as if it is associated with
Figure 8.3-1 shows selected common-shot gathers as- residual moveout, velocities may be updated at each
sociated with a land line recorded over a salt diapiric shot or receiver location. Such residual moveout anal-
structure. Note the presence of coherent reflections on ysis has been developed for shot-profile migration (Al-
shot records to the left and to the right of the salt Yahya, 1989) and is discussed later in this section.
diapir, and poor reflection quality of the shot records We shall extend the discussion on shot-geophone
above the salt diapir. Following the sorting to common- migration by comparing prestack depth migration with
receiver gathers, we observe the missing shots along the prestack time migration results. Shown in Figure 8.3-6
line (Figure 8.3-2). are the images obtained from post- and prestack time
Figure 8.3-3a shows the CMP-stacked section with migration. First, note that both yield an incorrect image
the salt diapir in the middle. The base-salt reflection of the base-salt boundary. Since the problem of imag-
with the velocity pull-up typical of salt diapirs is the ing the base-salt boundary is associated with a complex
most coherent event in this portion of the section. The overburden structure, it only can be handled by depth
poor imaging of the base-salt boundary by poststack migration. Strictly speaking, it should be handled by
depth migration is a direct consequence of the fact that prestack depth migration (Figure 8.3-3).
the CMP-stacked section is only an approximation to Second, prestack time migration actually can pro-
a zero-offset section (Figure 8.3-3b). Compare with the duce, as in this example, a poorer image of the base-salt
image in depth derived from shot-geophone migration boundary compared to poststack time migration (Fig-
shown in Figure 8.3-3c. The geometry of the base-salt ure 8.3-6). Doing the migration before stack does not
boundary indicates the presence of fault blocks. Also necessarily solve the problem of imaging beneath the
note some coherent events, albeit weak, in the subsalt salt diapir; doing it as depth migration is the key to
region. solving this problem. The poor performance of prestack
(text continues on p. 1280)
Earth Imaging in Depth 1275
1276 Seismic Data Analysis

FIG. 8.3-3. (a) The CMP stack associated with the data shown in Figure 8.3-1; (b) poststack depth migration; (c) prestack
depth migration. Trace spacing of the prestack depth-migrated section (the same as the shot or receiver group interval) is
twice that of the poststack depth-migrated section (the same as the CMP interval). Event labeled as m is peg-leg multiple.
Earth Imaging in Depth 1277
1278 Seismic Data Analysis

FIG. 8.3-6. (a) The CMP stack associated with the data shown in Figure 8.3-1; (b) poststack time migration; (c) shot-
geophone prestack time migration. Compare with the results from depth migration shown in Figure 8.3-3. Trace spacing of
the prestack time-migrated section (the same as the shot or receiver group interval) is twice that of the poststack time-migrated
section (the same as the CMP interval).
Earth Imaging in Depth 1279
1280 Seismic Data Analysis

time migration can be verified by examining the focus- flank, and one other on the right flank of the diapir. The
ing of energy at zero offset on common-shot (Figure 8.3- velocity-depth model (Figure 8.3-12b), although verti-
7) and common-receiver gathers (Figure 8.3-8). Specif- cally exaggerated, is the same as in Figure 8.2-1. The
ically, note that the focusing of energy to zero offset same shot records after shot-profile migration using the
by prestack time migration is not as good as it is by true velocity-depth model (Figure 8.3-12b) are shown in
prestack depth migration (Figures 8.3-4 and 8.3-5). Figure 8.3-13a. Note that each shot record after migra-
We now complete our discussion on shot-geophone tion represents partial image of the subsurface within a
migration by examining the effect of missing data on limited lateral extent.
imaging quality. Figure 8.3-9 shows a model of a salt Now, imagine all 154 shot records after migra-
diapir. A total of 193 shot records was created by using a tion placed at their corresponding shot locations along
nonzero-offset ray-theoretical modeling procedure. The the line. Then, consider one specific receiver loca-
receiver cable is split-spread with an offset range of 50- tion. There will be traces from a number of migrated
1200 m, and 48 receivers at 50-m interval. Putting aside shot records that will coincide with this receiver lo-
the inadequacies of ray theory for modeling traveltimes, cation that are common to all. These traces consitute
we should be able to make use of the modeled shot a common-receiver gather after shot-profile migration.
records for investigating the missing-data problem. Selected common-receiver gathers sorted from the mi-
Selected shot records before and after shot- grated common-shot gathers are shown in Figure 8.3-
geophone migration using the true velocity-depth model 13b. The number of traces in a common-receiver gather
(Figure 8.3-9a) are shown in Figure 8.3-10. Note that obviously is determined by the recording geometry.
the events have focused onto and at the vicinity of the If the velocity-depth model used in shot-profile mi-
zero-offset traces. By extracting the zero-offset traces gration coresponds to the true model, then each shot
and placing them side by side, we obtain the depth im-
record should yield a correct image of the subsurface,
age in Figure 8.3-9b. (The amplitude weakening along
albeit limited in lateral extent (Figure 8.3-13a). Then,
the top-salt event is caused by the limitations in the
traces from various shot records after migration at the
ray-theoretical modeling.)
same receiver location should represent the identical im-
A total of 15% of the traces from the modeled shot
age below that receiver location. Stated differently, a
records was discarded arbitrarily and replaced with zero
common-receiver gather should contain flat events if the
traces. As a result, some shot records contained few zero
velocity-depth model used in shot-profile migration is
traces while some contained all zero traces. Shown in
correct. The final step of shot-profile migration involves
Figure 8.3-11 are the same shot records as in Figure 8.3-
10 with missing traces. Note that, after prestack depth the summation of the traces in each receiver gather to
migration, focusing of the energy to zero-offset and its create the image in depth (Figure 8.3-13c). Note that
vicinity (Figure 8.3-11) has been achieved in a manner the partial images with limited lateral extent from each
comparable to the case of the complete data set (Figure individual shot record (Figure 8.3-13a) coincide with
8.3-10). Likewise, the prestack depth-migrated section the complete image within the lateral extent of the line
derived from the missing data (Figure 8.3-9c) is very in its entirety.
similar to the section derived from the complete data set Examine the common-receiver gathers in Figure
(Figure 8.3-9b). We may conclude that shot-geophone 8.3-13b, closely. The receiver locations are labeled as
depth migration can accommodate missing data result- 179, 209, and 249. Note that the events associated with
ing from recording geometry irregularities. the top-salt and base-salt boundaries, and the flat re-
flector below are positioned differently in relation to the
receiver location of each gather. In case of a flat layer
Shot-Profile Migration boundary, the event is positioned symmetrically with
respect to the receiver location (179). In the case of a
We shall review a method of shot-profile migration dipping layer boundary, such as the top-salt, the event
(Reshef and Kosloff, 1986) using the salt diapir model is positioned to the right of the receiver location (209)
shown in Figure 8.2-1. A wave theoretical modeling or to the left of the receiver location (249), but always
scheme based on the two-way acoustic wave equation in the updip direction.
was used to generate 154 shot records. Selected shot
records are shown in Figure 8.2-2. Shot spacing is 50
m and receiver spacing is 50 m. Each shot record con- Sensitivity of Image Accuracy
tains 97 traces corresponding to a split-spread recording to Velocity Errors
geometry with a maximum offset of 2350 m.
Figure 8.3-12a shows three shot records — one lo- Accuracy of the velocity-depth model used in prestack
cated away from the main diapiric body, one on the left depth migration can be checked by examining event
Earth Imaging in Depth 1281

FIG. 8.3-9. (a) A velocity-depth model for a salt diapir; (b) image obtained from prestack depth migration using shot records
with no missing traces as in Figure 8.3-10; (c) image obtained from prestack depth migration using shot records with missing
traces as in Figure 8.3-11.
1282 Seismic Data Analysis

FIG. 8.3-10. (Top) Selected synthetic shot records associated with the velocity-depth model in Figure 8.3-10; (bottom) same
records after prestack depth migration.
Earth Imaging in Depth 1283

FIG. 8.3-11. (Top) Selected synthetic shot records as in Figure 8.3-10 with missing receivers; (bottom) same records after
prestack depth migration.

curvature on common-receiver gathers. Figure 8.3-14a Use three different trial constant velocities for the
shows three common-receiver gathers sorted from the overburden assigned to all the layers in the subsur-
shot records that were migrated using the constant over- face and migrate the shot records. Consider a common-
burden velocity above the salt diapir (Figure 8.3-14c). receiver gather at one specific location (179 in Figure
Note that the top-salt event exhibits flat character at 8.3-14b). The top-salt event exhibits a curvature up-
all three receiver locations since the velocity used for
ward with erroneously low velocity (2500 m/s), flat
migration is the same as the layer velocity above the
top-salt boundary (3000 m/s). Whereas the base-salt character with the correct velocity (3000 m/s), and
event and the event corresponding to the flat reflector downward curvature with erroneously high velocity
below do not exhibit flat character since the migration (3500 m/s). Moreover, the event depth changes from
velocity, in this case, is erroneously lower than the true one trial velocity to another — the low velocity causes
layer velocities. the event to appear at shallow depth.
(text continues on p. 1294)
1284 Seismic Data Analysis

FIG. 8.3-12. (a) Three synthetic shot records over the salt diapir model (b). Numbers on top of the shot profiles and the
velocity-depth model correspond to CMP locations.
Earth Imaging in Depth 1285

FIG. 8.3-13. (a) Shot records as in Figure 8.3-12a after shot-profile migration using the true velocity-depth model (Figure
8.3-12b); (b) selected common-receiver gathers; (c) final image from shot-profile migration. Numbers correspond to CMP
locations.
1286 Seismic Data Analysis

FIG. 8.3-14. (a) Three common-receiver gathers sorted from shot records that were migrated using the constant overburden
velocity above the salt diapir in (c) assigned to all the layers in the subsurface; (b) single common-receiver gather sorted from
shot records that were migrated using an erroneously low velocity (2500 m/s), correct velocity (3000 m/s), and an erroneously
high velocity (3500 m/s) for the overburden assigned to all the layers in the subsurface.
Earth Imaging in Depth 1287

FIG. 8.3-15. Velocity-depth models represented by constant-velocity half-spaces: (a) 2500 m/s, (b) 3000 m/s, and (c) 3500
m/s. Results of prestack depth migrations are shown in Figure 8.3-16.
1288 Seismic Data Analysis

FIG. 8.3-16. Shot-profile depth migrations using the three velocity-depth models shown in Figure 8.3-15. For each trial
velocity-depth model, shown on the left column is the selected receiver gather (location 249) and on the right column is the
depth image obtained by stacking the common-receiver gathers sorted from migrated common-shot gathers. Top, middle, and
bottom rows correspond to the results of prestack depth migrations using the velocity-depth models in Figure 8.3-15a, b, and
c, respectively.
Earth Imaging in Depth 1289

FIG. 8.3-17. Velocity-depth models with the same overburden layer and constant-velocity half-space below with three
different velocities: (a) 4500 m/s, (b) 5000 m/s, and (c) 5500 m/s. Results of prestack depth migrations are shown in Figure
8.3-18.
1290 Seismic Data Analysis

FIG. 8.3-18. Shot-profile depth migrations using the three velocity-depth models shown in Figure 8.3-17. For each trial
velocity-depth model, shown on the left column is the selected receiver gather (location 249) and on the right column is the
depth image obtained by stacking the common-receiver gathers sorted from migrated common-shot gathers. Top, middle, and
bottom rows correspond to the results of prestack depth migrations using the velocity-depth models in Figure 8.3-17a, b, and
c, respectively.
Earth Imaging in Depth 1291

FIG. 8.3-19. Velocity-depth models with the same overburden and diapir layers, and constant-velocity half-space below with
three different velocities: (a) 3500 m/s, (b) 4000 m/s, and (c) 4500 m/s. Results of prestack depth migrations are shown in
Figure 8.3-20.
1292 Seismic Data Analysis

FIG. 8.3-20. Shot-profile depth migrations using the three velocity-depth models shown in Figure 8.3-19. For each trial
velocity-depth model, shown on the left column is the selected receiver gather (location 249) and on the right column is the
depth image obtained by stacking the common-receiver gathers sorted from migrated common-shot gathers. Top, middle, and
bottom rows correspond to the results of prestack depth migrations using the velocity-depth models in Figure 8.3-19a, b, and
c, respectively.
Earth Imaging in Depth 1293

FIG. 8.3-21. (a) The zero-offset section associated with the velocity-depth model in Figure 8.3-19b; (b) depth migration of
the zero-offset section; (c) depth migration of prestack data (same as in Figure 8.3-20d). Velocity-depth model used in (b)
and (c) is the true model (Figure 8.3-19b).
1294 Seismic Data Analysis

Event curvature on common-receiver gathers can in Figure 8.3-17, accordingly. Here, we have specified a
be likened to residual moveout on moveout-corrected two-layer model — the overburden above the top-salt
common-midpoint gathers caused by incorrect move- boundary interpreted from Figure 8.3-16d, and the new
out velocities. By measuring the residual moveout, layer half-space below with three different trial velocities ap-
velocities can be updated at each receiver location. propriate for the diapir layer.
Residual moveout analysis can be formulated within the Figure 8.3-18 shows the depth image and a receiver
context of common-receiver gathers derived from shot- gather (location 249) from each of the three prestack
geophone migration (Al-Yahya, 1989), common-receiver depth migrations. Note that the top-salt event is flat in
gathers derived from shot-profile migration (Reshef and all three cases since the overburden layer is the same
Kosloff, 1986; Lee and Zhang, 1992), or common-depth- for all the three models. The base-salt event exhibits a
point gathers (image gathers) derived from common- moderate curvature upward with the trial velocity 4500
offset migration (Deregowski, 1990; Cox and Wapenaar, m/s for the diapir layer under consideration (Figure 8.3-
1992). 18a). We conclude that this velocity is too low for the
We shall examine the sensitivity of image accuracy diapir layer. The base-salt event exhibits a moderate
to velocity errors based on event curvature on common- curvature downward with the trial velocity 5500 m/s
receiver gathers derived from the shot-profile migration for the diapir layer (Figure 8.3-18e). We conclude that
of the salt diapir model data (Figure 8.3-12). This anal- this velocity is too high for the diapir layer. Finally, the
ysis also is a prelude to using the event curvature on base-salt event is flat with the trial velocity 5000 m/s
image gathers as a criterion for layer-by-layer velocity for the diapir layer (Figure 8.3-18c). We conclude that
determination (next chapter). Start with the velocity- this velocity is appropriate for the diapir layer.
depth model defined as a half space with three different Examine the stack power in Figure 8.3-18 for the
trial constant velocities that may be considered appro- top-salt event on the depth images and note that it is
priate for the overburden above the salt diapir (Figure identical in all three cases since the overburden layer
8.3-15). Perform shot-profile migration using each of the is the same for all the three models. The highest stack
velocity-depth models and sort the output to common- power for the base-salt event is attained with the trial
receiver gathers. Then, stack the receiver gathers to ob- velocity 5000 m/s for the diapir layer under considera-
tain the images in depth. tion (Figure 8.3-18d). In contrast, the stack power with
Figure 8.3-16 shows the depth image and a receiver the too-low (4500 m/s) and too-high (5500 m/s) veloci-
gather (location 249) from each of the three prestack ties (Figures 8.3-18b and f) is slightly weaker because of
depth migrations. Note that the top-salt event exhibits the curvature of the base-salt event on common-receiver
a strong curvature upward with the trial velocity 2500 gathers.
m/s for the overburden layer under consideration (Fig- Based on the results of prestack depth migration
ure 8.3-16a). We conclude that this velocity is too low using the three trial velocities (Figure 8.3-18), we as-
for the overburden. The top-salt event exhibits a strong sign the optimum velocity 5000 m/s to the diapir layer.
curvature downward with the trial velocity 3500 m/s Next, we interpret the base-salt event on the corre-
for the overburden layer (Figure 8.3-16e). We conclude sponding depth image (Figure 8.3-18d). Then, we mod-
that this velocity is too high for the overburden. Finally, ify the velocity-depth model in Figure 8.3-17b as shown
the top-salt event is flat with the trial velocity 3000 m/s in Figure 8.3-19, accordingly. Here, we have specified a
for the overburden layer (Figure 8.3-16c). We conclude three-layer model — the overburden above the top-salt
that this velocity is appropriate for the overburden. boundary interpreted from Figure 8.3-16d, the diapir
Examine the stack power in Figure 8.3-16 for the with the base-salt boundary interpreted from Figure
top-salt event on the depth images and note that the 8.3-18d, and the new half-space below with three dif-
highest stack power is attained with the trial velocity ferent trial velocities appropriate for the subsalt region.
3000 m/s for the overburden layer (Figure 8.3-16d). In Figure 8.3-20 shows the depth image and a receiver
contrast, the stack power with the too-low (2500 m/s) gather (location 249) from each of the three prestack
and too-high (3500 m/s) velocities (Figures 8.3-16b and depth migrations. Note that the top-salt and base-salt
f) is weaker because of the curvature of the top-salt events are flat in all three cases since the overburden
event on common-receiver gathers. and the diapir layers are the same for all the three mod-
Based on the results of prestack depth migration els. The deepest event associated with the flat reflector
using the three trial velocities (Figure 8.3-16), we as- within the half-space exhibits a very mild curvature up-
sign the optimum velocity 3000 m/s to the overbur- ward with the trial velocity 3500 m/s for the half-space
den. Next, we interpret the top-salt event on the corre- (Figure 8.3-20a). We conclude that this velocity is low
sponding depth image (Figure 8.3-16d). Then, we mod- for the half-space. The same event exhibits a very mild
ify the velocity-depth model in Figure 8.3-15b as shown curvature downward with the trial velocity 4500 m/s
Earth Imaging in Depth 1295

for the half-space (Figure 8.3-20e). We conclude that Figure 8.3-22 also shows the poststack time-
this velocity is high for the half-space. Finally, the same migrated section. The main thrust plane starts from
event is flat with the trial velocity 4000 m/s for the half- the surface at CMP 340 and dips down to the right. To
space (Figure 8.3-20c). We conclude that this velocity is account for the severe elevation differences, migration
appropriate for the half-space that represents the sub- was performed from the irregular topography (Section
salt region with the flat reflector. 4.6).
Examine the stack power in Figure 8.3-20 for the The recording geometry for this land line has severe
top-salt and base-salt events on the depth images, and irregularities with duplicate shots and missing shots.
note that they are identical in all three cases since the Here, we only want to discuss results of prestack depth
overburden and diapir layers are the same for all three migration and leave out the velocity-depth model esti-
models. It is difficult to distinguish the images of the mation. Selected common-receiver gathers sorted from
deepest event associated with the flat reflector within the migrated shot records are shown in Figure 8.3-23.
the half-space based on stack power (Figures 8.3-20b, Most events exhibit flat character. Note also the symme-
d, and f). Nevertheless, based on the subtle differences try of some events with respect to the receiver location
in stack power in favor of the trial velocity 4000 m/s as seen on gathers in the top row — they are associated
and the implausable undulations in the geometry of the with flat or gently dipping reflectors. The asymmetry
flat reflector that resulted from the trial velocities 3500 of some events with respect to the receiver location is
m/s and 4500 m/s, we may conclude that the optimum seen on gathers in the bottom row in Figure 8.3-23.
velocity for the subsalt region is 4000 m/s. Stacking of the common-receiver gathers yields the
This concludes the estimation of the velocity-depth depth image obtained from shot-profile migration (Fig-
model using the event curvature on common-receiver ure 8.3-24). For comparison, the depth image from post-
gathers. The conceptual appeal of this analysis can stack depth migration using the same velocity-depth
be very attractive for construction of earth models in model is shown also in Figure 8.3-24. Trace spacing
depth. In practice, however, the computational cost can in the prestack depth-migrated section corresponds to
be prohibitive. Instead, residual moveout analysis of im- the average shot spacing, whereas that in the post-
age gathers derived from common-offset migration is stack depth-migrated section corresponds to the CMP
used for updating layer velocities (Chapter 9). spacing. Note from the prestack depth-migrated section
Refer to the common-receiver gathers in Figures (Figure 8.3-24) the improved image below 2 km of the
8.3-16, 8.3-18, and 8.3-20 and note that the deeper the imbricate structure to the right of the main thrust plane
event, the higher the velocity and the shorter the ca- which starts at the surface in the vicinity of CMP 350
ble length, the poorer the resolving power of curvature and dips down to the right.
analysis for velocity determination. This observation is Figure 8.3-25 shows the CMP-stacked section from
comparable to the case of conventional stacking velocity the marine line over two salt diapirs. The target zone is
analysis, and it also applies to residual moveout analysis base-salt and the subsalt region. For comparison, Figure
of image gathers derived from common-offset migration. 8.3-25 also shows the poststack time-migrated section.
Compare the amplitudes on the images from depth Note the velocity pull-up along the base-salt reflection
migration of the zero-offset and prestack data (Figure at 1.9 s below CMP 400 and CMP 800 — typical of salt
8.3-21). Note the differences in amplitude distribution diapirs.
along the flat reflector below the salt diapir. Imaging Selected common-receiver gathers sorted from the
beneath complex structures has certain implications as
migrated shot records are shown in Figure 8.3-26. Note
to the acquisition geometry — specifically, on the choice
that most events exhibit flat character; this suggests
of the cable length.
that the velocity-depth model used in prestack depth
migration is fairly accurate. Flat or gently dipping re-
flectors manifest themselves as events on the common-
Field Data Examples receiver gathers that are symmetric with respect to the
receiver locations. Dipping reflectors, on the other hand,
We now review results of shot-profile migration of a manifest themselves as events that are asymmetic with
2-D land and a 2-D marine line. Figure 8.3-22 shows respect to the receiver locations.
the CMP-stacked section of the land line over an im- Stacking of the common-receiver gathers yields the
bricate structure associated with overthrust tectonics. depth image obtained from shot-profile migration (Fig-
The topographic high is the surface expression of the ure 8.3-27). For comparison, the depth image from post-
imbricate structure. Refraction and residual statics cor- stack depth migration using the same velocity-depth
rections were applied to the data before shot-profile mi- model also is shown in Figure 8.3-27. Trace spacing
gration. in the prestack depth-migrated section corresponds to
(text continues on p. 1304)
1296 Seismic Data Analysis

FIG. 8.3-22. (Top) A CMP-stacked section from a land line shot over a topographic high; (bottom) time migration of the
CMP-stacked section. (Data courtesy Lasmo International.)
Earth Imaging in Depth 1297

FIG. 8.3-23. Common-receiver gathers sorted from common-shot gathers after depth migration. The depth image obtained
from shot-profile migration is shown in Figure 8.3-24.
1298 Seismic Data Analysis

FIG. 8.3-24. (Top) Depth image derived from shot-profile migration of the data associated with the CMP-stacked section
in Figure 8.3-22; (bottom) depth migration of the CMP-stacked section in Figure 8.3-22.
Earth Imaging in Depth 1299

FIG. 8.3-25. (Top) A CMP-stacked section from a marine line over two salt diapirs; (bottom) time migration of the CMP-
stacked section. (Data courtesy Mobil North Sea Limited.)
1300 Seismic Data Analysis

FIG. 8.3-26. Part 1: Common-receiver gathers sorted from common-shot gathers after depth migration. The depth image
obtained from shot-profile migration is shown in Figure 8.3-27.
Earth Imaging in Depth 1301

FIG. 8.3-26. Part 2: Common-receiver gathers sorted from common-shot gathers after depth migration. The depth image
obtained from shot-profile migration is shown in Figure 8.3-27.
1302 Seismic Data Analysis

FIG. 8.3-26. Part 3: Common-receiver gathers sorted from common-shot gathers after depth migration. The depth image
obtained from shot-profile migration is shown in Figure 8.3-27.
Earth Imaging in Depth 1303
1304 Seismic Data Analysis

the shot spacing, whereas that in the poststack depth-


migrated section corresponds to the CMP spacing. Note
from the prestack depth-migrated section the improved
image of the base-salt reflector and the subsalt region.
The base-salt reflector appears to be faulted and shows
a structural closure at approximately 3.5 km below
CMP 630.

8.4 3-D POSTSTACK DEPTH MIGRATION

The fundamentals of 3-D migration are discussed in Sec-


tion 7.3 and its mathematical aspects are provided in
Appendix G. By now, we should be familiar with one-
pass and two-pass, implicit and explicit 3-D time mi-
gration algorithms. In this section, we shall examine
aspects of 3-D poststack migration within the context
of imaging beneath complex structures: (a) 3-D post-
stack time versus depth migration, (b) one-pass versus
two-pass 3-D poststack depth migration, and (c) im-
FIG. 8.4-1. A time slice used as a base map for a synthetic
plicit versus explicit 3-D poststack depth migration. In
3-D survey. The velocity-depth model and the modeled 3-
the next section, we shall extend the analysis to 3-D
D zero-offset data for selected inlines are shown in Figure
depth migration of prestack data. 8.4-2.

3-D Poststack Time versus Depth Migration used in computing the 3-D zero-offset wavefield (Figure
8.4-2). Also, for comparison, pretend that two lines —
Consider a 3-D survey over a hypothetical salt-dome center line I-241 and line I-181 away from the center, are
structure. The base map is shown in Figure 8.4-1. The part of a 2-D survey and perform 2-D time migration.
synthetic 3-D survey data consist of 481 inlines and 481 Note the following observations:
crosslines with 25-m trace spacing in both directions.
Selected cross-sections of the 3-D velocity-depth model
(a) The top of the salt, albeit the vertical scale is in
and the associated zero-offset inline sections are shown
time, has been imaged properly with 3-D time mi-
in Figure 8.4-2. The salt dome has a circular symmetry
gration, while the base of the salt has not. This is
with a flat base. The base map shown in Figure 8.4-1
actually is a time slice at 1200 ms to show the circular because the salt diapir acts as a complex overbur-
symmetry. den.
Keep in mind that the sections in Figure 8.4-2 are (b) The 2-D time migration produced the correct re-
the cross-sections of the 3-D zero-offset wavefield along sult for the top of the salt only along the center
the traverses that coincide with the cross-sections of inline (I-241), because there are no sideswipes on
the 3-D velocity-depth model. In other words, we are this line; therefore, there is no need for 3-D migra-
not starting with the cross-sections of the velocity-depth tion. However, on a line away from the center line,
model in Figure 8.4-2 and generating a set of 2-D zero- say I-181, even the top of the salt has not been im-
offset wavefields from them. Strictly speaking, the only aged properly by 2-D time migration, let alone the
section that does not contain any sideswipe energy is base of the salt. This is because the line contains
the center line I-241. Note the energy reflecting off the sideswipes off the flank of the salt dome.
flank of the salt dome and being recorded on the lines (c) Now you may have second thoughts about 2-D seis-
away from the center line. mic exploration. Fortunately, most structures have
First, we perform one-pass implicit 3-D time mi- dominant strike and dip directions. Properly ori-
gration (Section G.1) on the entire 3-D zero-offset syn- ented 2-D migrated lines often yield an acceptably
thetic data and display the same lines after 3-D time accurate structural picture. However, one lesson to
migration (Figure 8.4-3). The migration velocity field be learned from comparing the 3-D and 2-D time
is based on the true subsurface velocity-depth model migrations of line I-181 shown in Figure 8.4-3 is
(text continues on p. 1313)
Earth Imaging in Depth 1305

FIG. 8.4-2. Velocity-depth model (left column) and cross-sections of the zero-offset 3-D synthetic wavefield (right column)
for a circularly symmetric salt dome. The base map is shown in Figure 8.4-1.
1306 Seismic Data Analysis

FIG. 8.4-3. 3-D time migration (left column) versus a 2-D time migration (right column) of the synthetic data shown in
Figure 8.4-2.
Earth Imaging in Depth 1307

FIG. 8.4-4. 3-D depth migration (left column) versus a 2-D depth migration (right column) of the synthetic data shown in
Figure 8.4-2.
1308 Seismic Data Analysis

FIG. 8.4-5. First-pass time migration along the inline direction of the 3-D zero-offset wavefield of Figure 8.4-2. Results of
the second-pass time migration along the crossline direction are shown in Figure 8.4-6.
Earth Imaging in Depth 1309

FIG. 8.4-6. Second-pass time migration along the crossline direction of the 3-D zero-offset wavefield of Figure 8.4-2. Results
of the first-pass time migration along the inline direction are shown in Figure 8.4-5.
1310 Seismic Data Analysis

FIG. 8.4-7. One-pass versus two-pass 3-D time migration of the salt-dome model data shown in Figure 8.4-2. The one-pass
result is the same as in Figure 8.4-3 and the two-pass result is the same as in Figure 8.4-6.
Earth Imaging in Depth 1311

FIG. 8.4-8. Second-pass depth migration along the crossline direction of the 3-D zero-offset wavefield of Figure 8.4-2. Results
of the first-pass time migration along the inline direction are shown in Figure 8.4-5.
1312 Seismic Data Analysis

FIG. 8.4-9. One-pass versus two-pass 3-D depth migration of the salt-dome model data shown in Figure 8.4-2. The one-pass
result is the same as in Figure 8.4-4 and the two-pass result (first-pass in time followed by second-pass in depth) is the same
as in Figure 8.4-8.
Earth Imaging in Depth 1313

that the migration velocities that yield an accept- display four selected inline sections and four selected
able 2-D migrated section may be quite different crossline sections (Figure 8.4-5). After this first-pass
from the true subsurface velocity model required migration, the imaging of the center inline (I-241) is
by 3-D migration. complete, since there is no sideswipe energy that needs
to be moved out of the plane of this inline. On the other
We now perform one-pass implicit 3-D depth mi- hand, the center crossline (X-241) has not been imaged
gration (Section G.1) of the 3-D zero-offset synthetic at all, because no movement of energy took place in this
data shown in Figure 8.4-2. After 3-D depth migration line as yet after the first-pass migration. On other lines,
of the entire volume of the 3-D zero-offset wavefield us- there is still some more imaging to be done — note, for
ing the true 3-D velocity-depth model, we display the instance, the sideswipe energy on inline I-151.
same lines (Figure 8.4-4). Also, for comparison, pretend Now sort the data into crosslines, perform the
that the two lines — center line I-241 and line I-181 second-pass migration on the already migrated data,
away from the center, are part of a 2-D survey and and display the same inline and crossline sections (Fig-
perform 2-D depth migration, again using the correct ure 8.4-6). These are now the inlines and crosslines after
velocity-depth model. Note the following observations: the two-pass 3-D time migration. Since the overburden
velocity above the salt layer is constant, the two-pass
(a) The top and base of the salt now have been imaged 3-D time migration correctly images the top-salt bound-
properly with 3-D depth migration. The complex ary. Since we have done time migration, whether it is
overburden problem has been solved. (Compare the two-pass or one-pass, we have not imaged the base-salt
left column in Figure 8.4-3 with that in Figure 8.4- boundary correctly.
4.) Figure 8.4-7 gives a summary of the results of one-
(b) The 2-D depth migration produced the correct sub- pass (as in Figure 8.4-3) and two-pass (as in Figure 8.4-
surface model only along the center inline (I-241) 6) implicit 3-D time migration of the salt-dome model
because there are no sideswipes on this line. How- data (Figure 8.4-2). Although not evident from the in-
ever, on a line away from the center line, say I-181, line sections in Figure 8.4-7, the top-salt boundary ac-
neither the top nor the base of the salt have been tually has been imaged more accurately by the two-
imaged properly. This is because the line contains pass migration as compared with the one-pass migra-
sideswipes off the flank of the salt dome. tion. The constant-velocity of the layer above the salt
(c) By performing 2-D depth migration iteratively to enables the two-pass scheme to produce an accurate re-
converge to assumingly correct depth model, we sult (Section G.1). On the other hand, even though the
may be forcing the model to converge to something overburden velocity above the salt layer is constant, the
very different from the truth. This stems from the approximation made in the splitting has caused the one-
treatment of the sideswipes as events that are in pass scheme to produce an incorrect image of the top-
the plane of profile. salt boundary, especially along the directions diagonal
to inline and crossline directions (Section 7.3). Again,
the base-salt boundary has not been imaged correctly
Two-Pass versus One-Pass 3-D Poststack by either one-pass or two-pass migrations, since we have
Depth Migration done time migration rather than depth migration.
We now examine the plausability of a hybrid two-
In Section 7.3, we examined the characteristics of two- pass 3-D migration of the salt-dome synthetic data
pass and one-pass 3-D poststack time migration. We set, wherein the first-pass migration is in time and the
concluded that given the choice between one-pass and second-pass migration is in depth. In areas with a com-
two-pass schemes, 3-D poststack time migration may be plex overburden structure, such as the overthrust belts,
done in two passes provided the vertical velocity gradi- usually velocity varies laterally more in the dip direc-
ent is not excessively large and dips are gentle. Is the tion perpendicular to the thrust fronts than the strike
two-pass strategy also applicable to 3-D poststack depth direction. When that is the case, it might be plausable
migration? We shall experiment with the salt-dome syn- to do time migration in the strike direction with mild
thetic data set (Figure 8.4-2) and conclude that 3-D lateral velocity variations, followed by depth migrations
poststack depth migration has to be done in one pass. of selected lines in the dip direction with strong lateral
First, we consider two-pass implicit finite-difference velocity variations. Such two-step hybrid strategy may
3-D time migration of the salt-dome synthetic data set. be useful in building the 3-D velocity-depth model for
Start with the 3-D zero-offset wavefield (Figure 8.4-2), a subsequent, proper 3-D depth migration of the 3-D
and apply time migration in the inline direction and data.
1314 Seismic Data Analysis

For the synthetic data set in Figure 8.4-2, start in two dimensions, the image-ray plot (Figure 8.4-12b)
with the results of time migration in the inline direc- through the velocity-depth model (Figure 8.4-12a) veri-
tion (Figure 8.4-5) and perform depth migration in the fies the presence of a complex overburden above horizon
crossline direction. Then, display the same inlines and 8. (See Section 8.2 for a discussion on image rays.)
crosslines as in Figure 8.4-6. After time migration as Note the similarity between the inline velocity-
the first-pass and depth migration as the second-pass depth model (Figure 8.4-12a) and line B after 3-D depth
(Figure 8.4-8), note that we certainly have restored the migration (bottom left, Figure 8.4-10). Despite this sim-
true geometry of the top-salt boundary correctly. This ilarity, there are parts of the survey in which the data
is because the overburden is constant velocity; there- are overmigrated (on line A, bottom left, Figure 8.4-
fore, it did not matter whether we did time or depth 11) and parts in which the data are undermigrated (not
migration. However, we have not been able to restore shown). Differences between the output from depth mi-
the geometry of the base-salt boundary, except for the gration and the velocity-depth model used requires an
center crossline (X-241), because this salt structure is
iterative modification of the velocity-depth model where
truly 3-D in character with no dominant strike or dip
it departs from the output of depth migration (Section
direction.
8.2). For comparison, Figures 8.4-10 and 8.4-11 show
Figure 8.4-9 gives a summary of the results of one-
2-D depth migrations of the selected lines. Note the
pass (as in Figure 8.4-4) and two-pass (as in Figure 8.4-
8) 3-D depth migration of the salt-dome model data obvious difference between 2-D and 3-D depth migra-
(Figure 8.4-2). Again, note that the top-salt boundary tions of line D (Figure 8.4-11). Although the 2-D depth-
actually has been imaged more accurately by the two- migrated section contains an abundance of reflections as
pass migration as compared with the one-pass migra- compared to the 3-D depth migrated section, that re-
tion. The constant-velocity of the layer above the salt flection energy does not belong on line D. From line A
enables the two-pass scheme to produce an accurate re- or B, note that energy should be migrated in the updip
sult. On the other hand, even though the overburden direction from line D to C. Hence, after 3-D migration,
velocity above the salt layer is constant, the approxi- line D is depleted of reflection energy (Figure 8.4-11),
mation made in the splitting has caused the one-pass while line C is enriched (Figure 8.4-10).
scheme to produce an incorrect image of the top-salt The question of how to supply a 3-D velocity field
boundary. The base-salt boundary, however, has not to do 3-D depth migration is beyond the scope of this
been imaged correctly by the two-pass scheme, while chapter. Earth modeling in depth is discussed in Chap-
it has been imaged correctly by the one-pass scheme. ter 9.
The experiments using the salt-dome synthetic
data set (Figure 8.4-2) combined with the model exper-
iments presented in Section 7.3 lead us to the following
Implicit versus Explicit 3-D Poststack
conclusions:
Depth Migration
(a) If the velocity field is judged to be suitable for time
Finally, we compare the performance of the implicit
migration, the two-pass strategy for 3-D time mi-
gration may be acceptable provided the vertical ve- and explicit schemes using the circularly symmetric
locity gradient is not excessively large and dips are salt-dome model data set of Figure 8.4-2. The implicit
not very steep. scheme uses the 45-degree extrapolator in a split mode,
(b) If the velocity field requires depth migration, the and the explicit scheme uses a one-dimensional (1-D)
one-pass strategy for 3-D depth migration is im- explicit filter combined with the 5 × 5 McClellan fil-
perative. ter template (Section G.2). Selected inline sections are
shown in Figure 8.4-13, and depth slices are shown in
Figures 8.4-10 and 8.4-11 show a field data exam- Figure 8.4-14. The top-salt boundary is imaged more
ple of 3-D depth migration using the one-pass scheme. accurately by the explicit scheme because of the near-
The inline (top left) and crossline (top right) stacked circular symmetry of its impulse response (Figure 7.3-
sections in both figures show a structural high, which is 14). Positioning errors by the implicit scheme implied
associated with culminations in an overthrust belt. Fig- by its impulse response (Figure 7.3-3) are better ob-
ure 8.4-12a is an inline cross-section of the 3-D velocity- served on the depth slices. Note in Figure 8.4-14 that
depth model. The deepest horizon on the velocity model the top-salt boundary image by the implicit scheme is
(horizon 8) corresponds to the event below 2 s on the not circularly symmetric; instead, significant undermi-
inline stacked section in Figure 8.4-10. The true geom- gration especially along the two diagonals has resulted.
etry of this horizon has been distorted by the struc- In contrast, the explicit scheme has preserved the cir-
ture above acting as a complex overburden. Although cular character of the salt dome.
Earth Imaging in Depth 1315

FIG. 8.4-10. An inline (top left) and a crossline stacked section (top right) from a land 3-D survey and the corresponding
2-D (center) and 3-D (bottom) depth migration results. A cross-section of the velocity-depth model is shown in Figure 8.4-12.
(Data courtesy Chevron USA, Inc.)
1316 Seismic Data Analysis

FIG. 8.4-11. Another inline (top left) and crossline stacked section (top right) from the same land 3-D survey as in Figure
8.4-10 with their 2-D (center) and 3-D (bottom) depth migrations. A cross-section of the velocity-depth model is shown in
Figure 8.4-12. (Data courtesy Chevron USA, Inc.)
Earth Imaging in Depth 1317

FIG. 8.4-12. (a) An inline cross-section of a 3-D velocity-depth model that was used in 3-D depth migration of the data
shown in Figures 8.4-10 and 8.4-11. (b) Image rays through this velocity-depth model.
1318 Seismic Data Analysis

FIG. 8.4-13. A comparison of implicit and explicit 3-D migrations. Velocity-depth model and cross-sections of the zero-offset
3-D synthetic wavefield are shown in Figure 8.4-2. The base map is shown in Figure 8.4-1. Time slices are shown in Figure
8.4-14.
Earth Imaging in Depth 1319

FIG. 8.4-14. A comparison of implicit and explicit 3-D migrations. Velocity-depth model and cross-sections of the zero-offset
3-D synthetic wavefield are shown in Figure 8.4-2. The base map is shown in Figure 8.4-1. Vertical sections are shown in
Figure 8.4-13.
1320 Seismic Data Analysis

FIG. 8.4-15. Three-dimensional wave-equation datuming (left column) versus 2-D wave-equation datuming (right column)
from the surface down to a 1000-m depth of the synthetic data shown in Figure 8.4-2.
Earth Imaging in Depth 1321

3-D Poststack Datuming for nonhyperbolic moveout caused by lateral veloc-


ity variations. And the compelling reason for doing
As a byproduct of 3-D migration algorithms based on depth migration in three dimensions is to account
wavefield extrapolation, 3-D stacked data can be da- for the 3-D behavior of complex overburden struc-
tumed from the surface to a specified depth or time tures that give rise to lateral velocity variations. In
level in a 3-D sense. This capability can be particularly a strict theoretical sense, in the presence of lateral
useful in reservoir studies. Basically, the surface wave- velocity variations, you need to image the subsur-
field is downward continued to a desired depth without face by migration of seismic data in depth, before
invoking the imaging principle along the way. Figure stack and in three dimensions.
8.4-15 shows the 3-D zero-offset data (Figure 8.4-2) da-
tumed from the surface z = 0 to a 1000-m depth. The Figures 8.5-1 and 8.5-2 show subsurface images
following conclusions can be made: along the same inline traverse obtained from 2-D and
3-D, post- and prestack, and time and depth migra-
(a) Since there are no sideswipes on the center line, tions. Examine the quality of imaging of the fault planes
I-241, datuming in a 2-D or 3-D sense is identical. and the reflectors within the faulted zone in each of
(b) However, there is a significant difference between the sections. The fault-plane reflections and the gently
2-D datuming and 3-D datuming for lines with dipping reflections within the fault blocks give rise to
sideswipe energy — those that are increasingly far- the problem of conflicting dips, and the lateral velocity
ther from the center line, say, Line I-181. variations across the fault blocks themselves give rise
to the problem of nonhyperbolic moveout. These two
The wave-equation datuming described here takes problems also manifest themselves in three dimensions.
the input wavefield from one flat constant horizontal da- Therefore, to get the best possible image, as predicated
tum to another. Prestack and poststack wave-equation by the data set in Figures 8.5-1 and 8.5-2, one should
datuming using arbitrary 2-D datum interfaces is dis- do the migration before stack, in depth, and in three
cussed in Section 8.1. The constant datum level should dimensions.
not be a limitation, particularly in reservoir studies. For What then is an appropriate algorithm for 3-D
example, the 3-D stacked data can be datumed to the prestack depth migration — Kirchhoff-summation,
top of the reservoir level followed by detailed imaging finite-difference, or frequency-wavenumber? The algo-
of the target zone only. rithm of choice must meet the following requirements:

(a) For it to qualify as a depth migration algorithm,


8.5 3-D PRESTACK DEPTH MIGRATION first and foremost, the algorithm of our choice must
be able to image steeply dipping reflectors in the
We remind ourselves of the two prominent circum- presence of lateral velocity variations.
stances that require migration of seismic data before (b) 3-D prestack seismic data invariably suffer from ir-
stack and in three dimensions: regular spatial sampling. Therefore, the algorithm
of choice for prestack migration must cope with ir-
(a) The compelling reason for doing time migration is regularly sampled data.
dipping events. The compelling reason for doing (c) Just as we use prestack time migration to estimate
time migration before stack is to account for con- rms velocities (Sections 5.4 and 7.4), we often wish
flicting dips with different stacking velocities. And to use prestack depth migration to estimate inter-
the compelling reason for doing time migration in val velocities. When used as a velocity estimation
three dimensions is to account for the 3-D behav- tool, we do not have to generate image gathers
ior of fault-plane reflections and reflections within from prestack depth migration at all CMP loca-
fault blocks that give rise to the problem of con- tions along a 2-D line or at all bin locations over
flicting dips. In a strict theoretical sense, in the a 3-D survey area. Instead, it often is sufficient to
presence of conflicting dips with different stacking generate image gathers at sparse intervals along the
velocities, you need to image the subsurface by mi- line, or along selected lines or even on a sparse grid
gration of seismic data in time, before stack and in over the 3-D survey area.
three dimensions.
(b) The compelling reason for doing depth migration is Whatever the type of algorithm, requirement (a)
lateral velocity variations. The compelling reason cannot be waived for depth migration. Requirement (b)
for doing depth migration before stack is to account to handle irregular spatial sampling may be fulfilled
1322 Seismic Data Analysis

FIG. 8.5-1. Application of 2-D migration strategies to a field data set: the vertical axis for the time-migrated sections is in
time (s) and the vertical axis for the depth-migrated sections is in depth (km). Compare with the results of 3-D migrations
shown in Figure 8.5-2.
Earth Imaging in Depth 1323

FIG. 8.5-2. Application of 3-D migration strategies to a field data set: the vertical axis for the time-migrated sections is in
time (s) and the vertical axis for the depth-migrated sections is in depth (km). Compare with the results of 2-D migrations
shown in Figure 8.5-1.
1324 Seismic Data Analysis

by azimuth-moveout (AMO) correction (Biondi et al., (b) scaling and summation of the amplitudes along
1998) or inversion to common offset (ICO) (Chemingui the computed traveltime trajectory based on the
and Biondi, 1999). Once data are spatially regularized Kirchhoff integral solution to the scalar wave equa-
so that the resulting prestack data have uniform fold tion.
of coverage and have been corrected for source-receiver
azimuth, then, in principle, any of the three categories The scaling of amplitudes before summation includes
of migration algorithms — Kirchhoff-summation, finite- application of the obliquity factor cos θ, the spherical
difference, or frequency-wavenumber, can be used to
divergence factor 1/vr, and the amplitude and phase
perform 3-D prestack depth migration. While it has
corrections, |ω| exp(iπ/2), implied by the derivative op-
the advantage of producing an image in depth along
erator ∂P/∂t (Section 4.1). Additionally, as for any mi-
a set of line traverses without having to produce an im-
gration, undersampling of the input data in x and y
age volume in depth, Kirchhoff summation technique
lacks the rigor to handle amplitudes that the frequency- directions needs to be compensated for by a suitable
wavenumber techniques can provide. Finally, the finite- antialiasing filter. Nevertheless, it is the traveltime com-
difference and frequency-wavenumber migration meth- putation that poses numerical accuracy and efficiency
ods are global methods; as such, they are not suitable challenges when implementing the Kirchhoff summation
to meet requirement (c). The better treatment of am- method for 3-D prestack depth migration.
plitudes by the frquency-wavenumber algorithms com-
pared to the Kirchhoff summation technique, however,
has greatly increased their use in practice (Biondi and Calculation of Traveltimes
Palacharla, 1996).
A direct method to compute traveltimes is ray tracing
through the specified velocity-depth model. A bundle
Kirchhoff Summation of rays emerging from a source location at the sur-
face can be sprayed down into the earth and traced
Kirchhoff’s integral solution to the scalar wave equation through the subsurface while accounting for ray bend-
ing caused by changes in velocity gradient and refrac-
∂2P ∂2P ∂2P 1 ∂2P tion at layer boundaries with velocity contrast. Reflec-
2
+ 2
+ 2
= 2 (8 − 1) tion points along each of the raypaths are identified as
∂x ∂y ∂z v (x, y, z) ∂t2
the intersection points of the rays with the layer bound-
gives the pressure wavefield P (x, y, z; t) propagating in aries. The traveltime from the source location at the
a medium with velocity v(x, y, z) at a location (x, y, z) surface and a reflection point at the subsurface is then
and at an instant of time t. As such, the Kirchhoff solu- calculated by integrating the elements of distance along
tion is a mathematical statement of Huygen’s principle
the raypath divided by the velocity associated with that
which states that the pressure disturbance at time t+∆t
element. By applying reciprocity, the traveltime from a
is the superposition of the spherical waves generated by
receiver location at the surface to a reflection point in
point sources at time t (Officer, 1958).
The discrete form of the integral solution to equa- the subsurface can be computed in the same manner.
tion (8-1) as used in practical implementation of Kirch- Finally, for a given source-receiver pair at the surface
hoff migration is given by (Section H.1) and a reflection point in the subsurface, the total trav-
eltime is computed by adding the traveltime from the
source to the reflection point to the traveltime from the
∆x∆y cos θ ∂
Pout = Pin , (8 − 2) reflection point to the receiver.
4π vr ∂t
A The two-point ray tracing described above is con-
where ∆x and ∆y are inline and crossline trace in- ceptually simple, but is computationally intensive. Effi-
tervals, Pout = P (xout , yout , z; τ = 2z/v) is the out- cient ray tracing through complex velocity-depth mod-
put of migration using the input wavefield Pin = els is not a trivial task. Alternatives to two-point ray
P (xin , yin , z = 0; τ = t − r/v) within an areal aper- tracing, however, have been developed and implemented
ture A. The geometry associated with equation (8-2) is with sufficient accuracy. Examples include paraxial ray
given in Figure H-1. tracing (Keho and Beydoun, 1988; Červený et al., 1982)
The Kirchhoff summation method requires and Gaussian beam ray tracing (Červený et al., 1984).
Given the circumstances described above, one can
(a) computing nonzero-offset traveltimes through a identify potential problems with the ray tracing. There
3-D, spatially varying velocity medium, and will not always be a raypath combination associated
Earth Imaging in Depth 1325

with a source-receiver pair and a reflection point. De- (x, y, z) in a medium with velocity v(x, y, z). Specifi-
pending on the recording geometry and the complexity cally, T (x, y, z) = constant represents the wavefront
of the velocity-depth model, there may be zones through of constant phase at an instant of time. The wave is
which some rays may be missed. This complication is propagated from one wavefront to the next by way of
compounded by the computational load involved in ray raypaths, which are perpendicular to the wavefronts.
tracing itself. As such, direct ray tracing is rarely used When the medium velocity is not constant but is
for traveltime computations required for prestack depth an arbitrary function of space variables v(x, y, z), and
migration. the wave amplitude P0 (x, y, z) is not constant but also
Figure 8.5-3 shows the traveltime contours through varies spatially, then the traveltime function T (x, y, z)
a velocity-depth model that includes a salt sill with ve- of equation (8-3) is not a solution to the eikonal equa-
locity higher than the surrounding sediments. The dis-
tion (8-4). For a wave function, such as that given by
continuities along the traveltime contours correspond to
equation (8-3), with spatially varying amplitudes, the
locations where ray bending is implied by ray tracing.
eikonal equation is a good approximation to the wave
Note, however, that the locations where ray bending
equation only at a high-frequency limit (Section H.2).
takes place do not coincide with the salt sill bound-
ary where the largest velocity contrast exists. Problems The high-frequency limit is equivalent to small wave-
with ray tracing through a complex model as outlined lengths. How small should the wavelength be for a solu-
above also cause physically implausable rapid variations tion of the eikonal equation to be a good approximation
in the traveltime contours. to the wave equation? The approximation is valid if the
An alternative to two-paint ray tracing is wavefront fractional change in the velocity gradient is much less
construction (Vinje et al., 1993), which involves tracing than the wave frequency (Officer, 1958). In practice, this
not just one ray but a fan of rays together. As such, the means that the eikonal equation can be used to compute
medium represented by the velocity-depth model used traveltimes if the velocity-depth model does not contain
in ray tracing is covered adequately by controlling the large velocity gradients.
ray density along wavefronts (Lecomte, 1999). In areas Just as a solution to the scalar wave equation can
with low ray density, additional ray bundles may be be formulated using a finite-difference scheme (Section
created by paraxial rays. 4.1), the eikonal equation (8-4) also can be solved using
the finite-difference technique (Section H.3). Consider
the 3-D computation mesh sketched in Figure H-2b. We
The Eikonal Equation want to compute the traveltime T of the eikonal equa-
tion (H-27) at grid point (x + ∆x, y, z + ∆z) using the
Consider a plane wave function P (x, y, z; t) with a spa- known traveltimes at grid points (x, y, z), (x+∆x, y, z),
tially varying amplitude P0 (x, y, z) and spatially vary- and (x + ∆x, y + ∆y, z).
ing traveltime T (x, y, z) (Section H.2) Computing the traveltimes at depth z + ∆z from
those at depth z means extrapolating T in the z-
direction. Rewrite equation (8-4) in the form of an ex-
P (x, y, z; t) = P0 (x, y, z) exp −iω t − T (x, y, z) . trapolation equation as (Reshef and Kosloff, 1986)
(8 − 3)
Assume for the moment that the wave amplitude 2 2
∂T ∂T ∂T
P0 (x, y, z) does not vary spatially, but is a constant. v = 1− v − v . (8 − 5)
Then, the plane-wave solution of the form defined by ∂z ∂x ∂y
equation (8-3) satisfies the scalar wave equation (8-1). As shown in Section H.3, the 3-D eikonal equation (8-5)
Substitution of equation (8-3) into equation (8-1) yields can be solved for the traveltime values T (x, y, z) for the
propagating wavefront through a velocity field v(x, y, z)
∂T 2 ∂T 2 ∂T 2 1 in the subsurface using a finite-difference scheme.
+ + = , (8 − 4)
∂x ∂y ∂z v 2 (x, y, z) Figure 8.5-4 shows the traveltime contours through
which is called the eikonal equation. It is a ray- the same velocity-depth model as in Figure 8.5-3 com-
theoretical approximation to the scalar wave equation puted by using the eikonal equation. These may be con-
(8-1). Derivation of equation (8-4) is provided in Section sidered as wavefronts expanding from a source at the
H.2. surface located at 3000 m from the origin. Note that the
While the solution of the scalar wave equation (8- shape of the traveltime contours honors the boundaries
1) represents the wavefield P (x, y, z; t) at a point in with velocity contrast. Nevertheless, the traveltime con-
space (x, y, z) and at an instant of time t, the solu- tours do not show the abrupt changes as would be the
tion of the eikonal equation (8-4) represents the trav- case at a boundary such as the salt-sediment interface
eltime T (x, y, z) for a ray passing through a point with sharp velocity contrast.
1326 Seismic Data Analysis
Earth Imaging in Depth 1327
1328 Seismic Data Analysis

FIG. 8.5-5. Wavefronts of an expanding wave from a source located at surface point S. The layer above has a lower velocity
than the layer below. The dotted lines indicate the wavefronts associated with the head wave that is refracted at a critical
angle along the interface of the two layers.

To explain the smooth behavior of the traveltime tion that has excluded the head wave. Note the sharp
contours at layer boundaries in Figure 8.5-4, refer to changes in the traveltime contours, which at some loca-
the sketch in Figure 8.5-5 of an expanding wavefront tions appear to coincide nicely with the salt-sediment
through a flat interface with a large velocity contrast. layer boundary. Nevertheless, the match is not consisi-
At the critical angle of refraction, waves travel along the tent at all locations and the traveltime contours show
layer boundary with the faster velocity of the underly- rapid fluctuations that are not physically plausable.
ing layer, Eventually, these waves are refracted back Figure 8.5-7a shows traveltime contours that were
into the overlying layer and are recorded in the form computed using a finite-difference solution to the
of first arrivals; thus, they often are called head waves. eikonal equation. The velocity-depth model consists of
Note that the wavefront associated with the head wave layers with velocities that vary from 1500 m/s to 4000
tends to smooth out the sharp change in the expand- m/s. Note that the eikonal solution, although it is as-
ing wavefront as it crosses over the layer boundary with sociated only with the fastest arrival, always yields a
velocity contrast. It is this effect that we see in Fig- raypath from a source at the surface to a point in the
ure 8.5-4 where the traveltime contours make a smooth subsurface through the gridded velocity-depth model.
transition from the sediment to the salt layer. Some of the raypaths are indicated by trajectories that
Ideally, the traveltime contours should pronounce have a solid circle at the end. The smooth behavior of
the sharp layer boundaries. While some solutions to the the traveltime contours is associated with the head wave
eikonal equation include the head waves (Vidale, 1988; that often is the first arrival.
Podvin and Lecomte, 1991) as shown in Figure 8.5- In contrast with the eikonal solution, the wavefront
4, others exclude the head wave (Reshef and Kosloff, construction yields multiple arrivals as shown in Figure
1986; Reshef, 1991). Figure 8.5-6 shows the traveltime 8.5-7b. Again, some of the raypaths are indicated by
contours derived from the solution of the eikonal equa- trajectories that have a solid circle at the end. Note
Earth Imaging in Depth 1329
1330 Seismic Data Analysis

FIG. 8.5-7. Traveltime contours derived (a) from a finite-difference solution to the eikonal equation, (b) and (c) from
wavefront construction (Lecomte, 1999). See text for details.
Earth Imaging in Depth 1331

the multiple raypaths associated with a single point in each of the raypaths may be used in the computation.
the subsurface. Where a head wave or diffraction de- Among the 11 raypaths from the grid point 6 at depth
velops, the wavefront construction leaves a gap in the z to grid points 1 to 11 at depth z + ∆z, choose that
traveltime contour. With wavefront construction, how- which corresponds to the minimum traveltime.
ever, you can be selective in the type of rays that you The process may be continued for all grid points
would like to include in the summation of amplitudes. at depth z, and then from one depth to the next, and
Figure 8.5-7c shows traveltime contours associated with the traveltimes associated with the minimum-traveltime
the reflecting boundary represented by the thick curve. raypaths through each of the horizontal slabs may be
In fact, the solution from wavefront construction in this added to compute the total traveltime from a source or
case includes arrivals associated with reflected waves as receiver point at the surface z = 0 to a reflection point
well as transmitted waves. at some depth in the subsurface.
The eikonal solution does not necessarily yield
the maximum energy along the single-arrival raypath.
In wavefront construction, on the other hand, ampli- Summation Strategies
tudes associated with multiple arrivals can be included
in the summation. Thus, the chance of capturing the
Figure 8.5-10 shows the zero-offset wavefield responses
maximum-energy arrival increases, albeit at an increase
of a point diffractor buried in media with varying de-
in computational cost. By including multiple arrivals
grees of complexity. The traveltime trajectories associ-
in the summation, the likelihood of attaining a com-
ated with the point diffractors buried in a constant-
plete image of a complex structure also is higher. Fig- velocity medium, beneath an overburden with mild
ure 8.5-8a shows raypaths from a single source at the to moderate lateral velocity variations, and beneath
surface that illuminate portions of the steep flank of a an overburden with strong lateral velocity variations,
salt dome. By using the arrivals associated with these all are single-valued. Therefore, ray tracing through
raypaths, only, in Kirchhoff summation, the resulting such models would produce unambiguous traveltimes
partial image shown in Figure 8.5-8b is obtained. By for Kirchhoff summation.
including in the summation the arrivals associated with The zero-offset traveltime trajectory associated
the reflections off the salt flank as well as the arrivals with the point diffractor buried beneath an overburden
associated with the reflections from a deeper interface with severe lateral velocity variations, however, is mul-
that bounce off the salt flank (Figure 8.5-8c), a more tivalued (Figures 8.5-10d). One would have to decide as
complete image can be obtained (Figure 8.5-8d). to what summation path to choose:

(a) The travelpath that corresponds to the first ar-


Fermat’s Principle rivals; that is, minimum-time summation trajec-
tory (Vidale, 1988; Podvin and Lecomte, 1991),
The traveltime along a raypath from one point to an- (b) The travelpath that corresponds to the bowties
other has an extremum value which, for most physical that contain the most significant portion of the en-
problems, is a minimum. This is the formal statement ergy associated with the zero-offset wavefield re-
of Fermat’s principle (Officer, 1958). The raypath also sponse; that is, maximum-energy summation tra-
defines the direction of energy flow. Among a bundle of jectory (Nichols, 1996),
rays from one point to another, Fermat’s principle can (c) The travelpath that corresponds to the shortest
be applied to discard all but one raypath that corre- distance between the source or receiver point at
sponds to a minimum time of travel from one point to the surface and the reflection point at the subsur-
the other. This practical concept can be used to perform face; that is, minimum-distance summation trajec-
traveltime computations for prestack depth migration tory (Moser, 1991), or
(Meshbey et al., 1993; Vesnaver, 1996). (d) The entire multivalued travelpath.
Consider the raypath geometry depicted in Figure
8.5-9. The velocity-depth model may be split into a set From the zero-offset wavefields associated with a
of horizontal slabs with a specified thickness, say 50 to diffractor buried in a medium with varying complex-
300 m. Assume that traveltimes from a source at the ity shown in Figure 8.5-10, it can be inferred that the
surface z = 0 to a depth z through a velocity-depth minimum-time strategy may be suitable for cases of
model already have been computed. Now compute the moderate to strong lateral velocity variations (Figure
traveltime from the grid point 6 at depth z to each 8.5-10b,c), whereas the maximum-energy strategy may
of the grid points at depth z + ∆z within a specified be imperative for a case of a complex overburden with
aperture, say grid points 1 to 11. An average velocity severe lateral velocity variations (Figure 8.5-10d). Ide-
between the grid points at depth z and z + ∆z along ally, it would be desirable not to exclude any portion
1332 Seismic Data Analysis

FIG. 8.5-8. Two different illuminations of the salt flank by a single source and rays associated with (a) primary reflections
which yield the image of the salt flank in (b), and (c) reflections that first bounce off a deeper layer boundary which yield the
image of the salt flank in (d) (Lecomte, 1999).
Earth Imaging in Depth 1333

FIG. 8.5-9. A sketch that illustrates traveltime calculation


using Fermat’s principle. See text for details.

of the traveltime trajectory and use a multivalued sum-


mation path. This, however, can be formidably costly
and often is not needed in practice. Efficient traveltime
calculation and choice of a summation path are impor-
tant considerations for the 3-D prestack depth migra-
tion of large volumes of seismic data (Meshbey et al.,
1993; Sethian and Popovici, 1999).

Migration Aperture

Because of cost considerations in 3-D prestack depth


migration, it is compelling to make a careful choice of
aperture width in the inline and crossline directions.
While an excessively large aperture unnecessarily in-
creases run time, a small aperture can produce a poor
image from 3-D prestack depth migration. Migration
aperture in the Kirchhoff summation was discussed al-
ready in Section 4.2. Recall that a small aperture causes
destruction of steeply dipping events. Excessively small
aperture width also organizes random noise, especially
in the deeper part of the section, as horizontally domi-
nant spurious events. FIG. 8.5-10. Zero-offset wavefield responses of a point
Figure 8.5-11 shows aperture tests for 3-D prestack diffractor buried in (a) a constant-velocity medium, (b) be-
depth migration. A small aperture in both inline and neath an overburden with mild to moderate lateral velocity
crossline directions causes poor imaging of the steeply variations, (c) beneath an overburden with strong lateral
dipping fault planes and the steep reflector that defines velocity variations, and (d) beneath a complex overburden
the base of the sedimentary basin. Note also the orga- with severe lateral velocity variations.
nizing effect of small aperture on random noise.

Operator Antialiasing in Section 4.6 we examined its adverse effects on mi-


gration. Prestack data may be made spatially uniform,
In Section 1.2 we reviewed the phenomenon of spatial for instance, by azimuth-moveout correction (AMO)
aliasing caused by undersampling of recorded data, and (Biondi et al., 1998), thus enabling use of finite-dif-
1334 Seismic Data Analysis
Earth Imaging in Depth 1335

can destroy steeply dipping events (Section 4.2). The


second approach can be taken but at a high cost in the
case of prestack data. Trace interpolation also suffers in
accuracy when applied to data with dipping events that
conflict with one another. The third approach requires
multiple copies of the input trace with different band-
widths; this can be troublesome for very large input
prestack data. Lumley et al. (1994) applied a triangu-
lar filter (Claerbout, 1992) to input traces in the time
domain without creating multiple copies. An improved
form of the triangular filtering method for 3-D Kirchhoff
summation is given by Abma et al. (1999).

3-D Common-Offset Depth Migration

In Section 7.4 we defined a strategy for 3-D prestack


time migration based on the application of 3-D zero-
offset migration theory to 3-D common-offset data.
Specifically, following NMO and 3-D DMO correc-
tion, prestack data will have been corrected for dip
and source-receiver azimuth effects. As a result, the
common-offset volumes of data are decoupled repre-
sentations of a 3-D zero-offset wavefield; thus, each
FIG. 8.5-12. A Kirchhoff summation trajectory that inter- common-offset volume can be migrated independently
cepts multiple samples from input data traces (Claerbout,
using a 3-D zero-offset time migration algorithm. For
1985).
the subsequent processing sequence, refer to Section 7.4.
Decoupling of common-offset data volumes, in a
ference or frequency-wavenumber algorithms for pre- strict theoretical sense, is only possible for events with
stack depth migration. When it becomes necessary to hyperbolic moveout. For events with nonhyperbolic
use Kirchhoff summation algorithm, operator aliasing is moveout that calls for depth migration, each of the
an issue that needs to be dealt with. resulting common-offset volumes following the applica-
Shown in Figure 8.5-12 is a low-velocity hyperbolic tions of NMO and 3-D DMO corrections would not be
summation path. Note that it intercepts more than one a representation of a 3-D zero-offset wavefield. As such,
sample in a given input trace, each sample indicated by and again in theory, a 3-D common-offset volume of
a shaded grid block. Thus, the summation along this data could not be depth-migrated.
path should include more than one sample per input
Despite this theoretical mandate, it is surprisingly
trace. There are three different approaches to handle
pleasant to see that indeed, in practice, the common-
operator aliasing caused by mulitple samples per input
trace included in the summation: offset strategy for 3-D prestack time migration some-
times can also be applicable to 3-D prestack depth
migration. While, following NMO and 3-D DMO cor-
(a) The summation trajectory that represents the kine-
rection, each of the common-offset volumes is time-
matics of the migration operator may be truncated
to exclude the steep flanks that suffer from aliasing. migrated using a 3-D rms velocity field, depth migra-
(b) Trace interpolation may be used to create addi- tion of the common-offset volumes is done using a 3-D
tional traces so as to avoid multiple samples per in- velocity-depth model.
put trace included in the summation (Section 7.2). There are two reasons why you may choose to ap-
(c) The frequency components that are aliased at a ply the common-offset strategy to 3-D prestack depth
given dip along the summation trajectory are fil- migration:
tered out (Gray, 1992).
(a) You may wish to use a 3-D zero-offset depth mi-
The first approach is undesirable since truncation gration algorithm other than the Kirchhoff sum-
is equivalent to limiting the migration aperture which mation, and thus avoid the troublesome task of
1336 Seismic Data Analysis

FIG. 8.5-13. Part 1: Two inline sections from image volumes derived from 3-D prestack depth migration using Kirchhoff
summation (left panels) and the common-offset technique (right panel) described in Section 8.5. The vertical axis is in depth
(km).
Earth Imaging in Depth 1337

FIG. 8.5-13. Part 2: Two inline sections from image volumes derived from 3-D prestack depth migration using Kirchhoff
summation (left panels) and the common-offset technique (right panel) described in Section 8.5. The vertical axis is in depth
(km).
1338 Seismic Data Analysis

FIG. 8.5-13. Part 3: Two inline sections from image volumes derived from 3-D prestack depth migration using Kirchhoff
summation (left panels) and the common-offset technique (right panel) described in Section 8.5. The vertical axis is in depth
(km).
Earth Imaging in Depth 1339

FIG. 8.5-14. Depth image derived from (a) prestack Kirchhoff migration, (b) common-offset depth migration using a workflow
that includes DMO correction, (c) common-offset depth migration using a workflow without DMO correction.
1340 Seismic Data Analysis

computing 3-D nonzero-offset traveltimes needed comparable to the result from the Kirchhoff summation
for the latter. (Figure 8.5-14a).
(b) For line output or for selected image gathers Kirch- As a final note, efficient and accurate implementa-
hoff summation is the appropriate algorithm. For tion of 3-D prestack depth migration is a topic of ac-
volume output from 3-D prestack depth migration tive and intensive research in the industry. A choice of
other algorithms can be more efficiently applied to strategies for 3-D prestack depth migration is summa-
common-offset data. rized below:

Figure 8.5-13 shows selected inline sections derived (a) You may wish to first apply azimuth-moveout
from 3-D prestack depth migration using the Kirch- (AMO) correction (Biondi et al., 1998) to reg-
hoff summation and the common-offset strategy. For ularize the prestack data, then use a frequency-
the latter, a 3-D zero-offset phase-shift-plus-correction wavenumber algorithm (Biondi and Palacharla,
(PSPC) (Kosloff and Kessler, 1987) was used to migrate 1996) to perform the 3-D prestack depth migra-
all 32 volumes of 3-D DMO-corrected common-offset
tion. The method involves the design and applica-
data; thus a full volume of image in depth was created.
tion of a wavefield extrapolation operator on the
Comparison of the sections from the two approaches
whole of the 3-D prestack data without decoupling
indicate that the imaging quality is largely comparable.
the common-offset volumes.
The common-offset strategy for depth migration
(b) Alternatively, you may wish to perform 3-D
will not work strictly for a case of complex overbur-
den structure that gives rise to complex, nonhyperbolic prestack depth migration using the Kirchhoff sum-
moveout. Shown in Figure 8.5-14a is the image ob- mation technique discussed in this section. The
tained from prestack depth migration based on Kirch- Kirchhoff summation implicitly handles geometry
hoff summation with a finite-difference solution to the irregularities associated with the prestack data.
eikonal equation to compute the traveltimes. The syn- (c) A third strategy, which also is described in the lat-
thetic data (O’Brien and Gray, 1996) are associated ter part of this section, is based on migration of the
with a salt sill model (Section K.2). The objective is decoupled 3-D common-offset volumes of data that
to image the subsalt region accurately. Note that the have been corrected for geometry irregularities and
common-offset strategy for prestack depth migration reduced to zero offset by way of 3-D DMO correc-
with or without DMO correction (Figure 8.5-14b,c) in- tion. Again, the preferred choice for the algorithm
cluded in the workflow fails to produce an accurate im- to perform the 3-D common-offset depth migration
age of the subsalt region with a level of accuracy that is is one based on wave extrapolation.

FIG. 8.E-1. (a) A zero-offset section containing all-zero traces except the one shown, which has a single isolated wavelet; (b)
the velocity-depth model for depth migration; (c) the x − z plane associated with the depth-migrated section. See Exercise
8-1.
Earth Imaging in Depth 1341

FIG. 8.E-2. Cross-sections extracted from the image volumes created by (a) 3-D poststack time migration, and (b) 3-D
prestack depth migration. See Exercise 8-4. (Data courtesy Diamond Geophysical.)
1342 Seismic Data Analysis

A general guideline for the choice of strategy for 3-D (c) Finally, if you are dealing with a complex over-
prestack depth migration is given below: burden structure, use of a 3-D prestack depth mi-
gration algorithm without decoupling the common-
(a) If you are dealing with a low-relief structure offset data is imperative.
and moderate lateral velocity variations, use 3-D
prestack depth migration for conducting image-
gather analysis for model verification and update, EXERCISES
but not necessarily for imaging in depth. For the
purpose of model verification and update, you need
Exercise 8-1. Consider the zero-offset x − t sec-
to work with only a sparse grid of image gathers;
and the Kirchhoff summation is the suitable algo- tion depicted in Figure 8.E-1. Using the velocity-depth
rithm to produce such a set of image gathers with- model in the same figure, sketch the depth-migrated
out creating a whole image volume. As for imaging x − z section.
in depth, you may be content with 3-D poststack Exercise 8-2. Identify the prominent sideswipe
depth migration; and a finite-difference (Sections energy on Lines I-181 and I-151 in Figure 8.4-2 (right
G.1 and G.2) or frequency-wavenumber algorithm column).
(Section 7.3) that uses wave extrapolation is the Exercise 8-3. In areas with complex overburden,
suitable algorithm. which one is more important — 3-D poststack depth
(b) If you are dealing with a complex structure and migration or 2-D prestack depth migration?
moderate-to-strong lateral velocity variations, use Exercise 8-4. How would the geometry of the sub-
a 3-D prestack depth migration algorithm based salt target reflectors labeled in yellow in Figure 8.E-2
on decoupling of common-offset volumes of data as behave on a section from an image volume created by
demonstrated in this section. 3-D poststack depth migration?
Appendix H
DIFFRACTION AND RAY THEORY FOR WAVE PROPAGATION

H.1 The Kirchhoff Integral

Kirchhoff’s integral solution to the scalar wave equation


∂2 ∂2 ∂2 1 ∂2
+ + − P (x, y, z; t) = 0 (H − 1)
∂x2 ∂x2 ∂x2 v 2 (x, y, z) ∂t2
is a mathematical statement of Huygen’s principle. In equation (H-1), P (x, y, z; t) is the pressure
wavefield propagating in a medium with velocity v(x, y, z). Huygen’s principle states that the
pressure disturbance at time t + ∆t is the superposition of the spherical waves generated by
point sources at time t.
Consider the geometry in Figure H-1 of a point diffractor at a location S(x, y, z) and an
observation surface A for the diffraction wavefield generated by the source at the diffractor.
Actually, the surface area A is only a portion of a closed surface and is the aperture of the
observation made over that closed surface. For convenience, we shall choose the receiver location
R(0, 0, 0) on the observation surface A to be at the origin of our coordinate system.
Also for convenience, we shall apply Fourier transform to the wavefield in the time direction

P (x, y, z, ; ω) = P (x, y, z; t) exp(−iωt) dt, (H − 2a)

where ω is the angular frequency. The inverse transform is given by

P (x, y, z, ; t) = P (x, y, z; ω) exp(iωt) dω. (H − 2b)

Now apply Fourier transform to equation (H-1) in the time direction to obtain
ω2
∇2 + P (x, y, z; ω) = 0, (H − 3)
v2
where ∇2 is the Laplacian operator
∂2 ∂2 ∂2
∇2 P = + + P.
∂x2 ∂x2 ∂x2
We may intuitively state that what we observe at the surface A is what is generated at
the source S. This statement is mathematically expressed by Gauss’s divergence theorem as
(Officer, 1958; Lass, 1950; Coulson, 1965; Jackson, 1962)
∂P
∇2 P dV = dA, (H − 4)
V A ∂n
where V is the volume of the region enclosed by the surface A and the derivative ∂P/∂n is
taken normal to the surface A in the outward direction. Note that Gauss’s divergence theorem
expressed by equation (H-4) transforms a surface integral into a volume integral. For our specific
application, this means imaging the earth’s interior from seismic waves observed at the earth’s
surface.
We shall solve equation (H-3) for each frequency component ω and sum the resulting so-
lutions over all frequency components to compute the wavefield at the source P (x, y, z; t = 0).
1344 Seismic Data Analysis

FIG. H-1. Geometry of a point diffractor to derive the Kirchhoff integral solution to the scalar wave
equation.

The solution obtained by Kirchhoff in 1882 requires Green’s function which describes the prop-
agation outward from a point source with spherical symmetry as (Coulson, 1965; Lass, 1950)
1 ω
G(r, ω) = exp −i r , (H − 5a)
r v
where
r= x2 + y 2 + z 2 (H − 5b).
is the distance between the observation point and the source location. Green’s function given
by equation (H-5a) is also a valid solution to equation (H-3):
ω2
∇2 + G(x, y, z; ω) = 0. (H − 6)
v2
We now rewrite equation (H-4) by multiplying both sides with Green’s function G of equa-
tion (H-5a) as
∂P
G∇2 P dV = G dA, (H − 7a)
V A ∂n
and rewrite, in turn, equation (H-7a) by interchanging our wave function P with Green’s function
G to obtain
∂G
P ∇2 GdV = P dA. (H − 7b)
V A ∂n
Now subtract equation (H-7b) from (H-7a) to get
∂P ∂G
G∇2 P − P ∇2 G dV = G −P dA. (H − 8)
V A ∂n ∂n
Diffraction and Ray Theory for Wave Propagation 1345

Equation (H-8) is known as Green’s theorem. Albeit the algebraic steps that we took from
Gauss’s theorem given by equation (H-4) to Green’s theorem given by equation (H-8) may
appear informal, the intention here is to provide the basic concept underlying the derivation.
Substitute equations (H-3) and (H-6) into the left-hand side of equation (H-8)
ω2 ω2
G∇2 P − P ∇2 G dV = −G P + P G dV,
V V v2 v2
and hence note that

G∇2 P − P ∇2 G dV = 0. (H − 9)
V

Now we turn our attention to the right-hand side of equation (H-8). Because Green’s func-
tion defined by equation (H-5a) becomes infinite at the source location S, we need to place
it inside an infinitesimally small enclosed surface E. This would then require computing the
right-hand side of equation (H-8) in two parts — once for the surface E and once for the surface
A.
Substitute equation (H-5a) into the right-hand side of equation (H-8) and note, from Figure
H-1, that for the surface E, (∂/∂n) = −(∂/∂r):
∂P ∂G 1 ω ∂P ∂ 1 ω
G −P dE = − exp −i r +P exp −i r r2 dΩ, (H − 10a)
E ∂n ∂n E r v ∂r ∂r r v
where dE = r2 dΩ and Ω is the solid angle around the point source S in Figure H-1. Carry
out the differentiation with respect to r and simplify the right-hand side of equation (H-10a) to
obtain
∂P ∂G ω ∂P ω
G −P dE = − exp −i r r + P + i rP dΩ. (H − 10b)
E ∂n ∂n E v ∂r v
Finally, take the limit r → 0 to obtain the contribution of the surface E:
∂P ∂G
G −P dE = −4πP. (H − 11)
E ∂n ∂n
Now substitute, again, equation (H-5a) into the right-hand side of equation (H-8) and note
from Figure H-1 that for the surface A, (∂/∂n) = −(∂/∂z):

∂P ∂G 1 ω ∂P ∂ 1 ω
G −P dA = − exp −i r +P exp −i r dA. (H − 12)
A ∂n ∂n A r v ∂z ∂z r v

Carry out the differentiation with respect to z while noting from Figure H-1 that ∂r/∂z = cos θ,
and simplify the right-hand side of equation (H-12) to get
∂P ∂G ω 1 ∂P cos θ ω cos θ
G −P dA = − exp −i r + 2 P +i P dA. (H − 13)
A ∂n ∂n A v r ∂z r v r
The total contribution to the right-hand side of equation (H-8) actually is the sum of
equations (H-11) and (H-13). The left-hand side of equation (H-8) vanishes by way of equation
(H-9). Hence, the resulting expression from equation (H-9) is
ω 1 ∂P cos θ ω cos θ
4πP = exp −i r + 2 P +i P dA. (H − 14)
A v r ∂z r v r
Recall that P = P (x, y, z; ω) in equation (H-14) and multiply both sides by exp(iωt) and
integrate over frequency ω. The left-hand side of the equation then becomes P (x, y, z; t) by way
of the inverse Fourier transform as in equation (H-2b). Thus the resulting expression is
1 1 ∂P cos θ ω cos θ r
P (x, y, z; t) = + 2 P +i P exp −iω t − dA dω. (H − 15)
4π ω A r ∂z r v r v
1346 Seismic Data Analysis

Define the variable τ = t−r/v as retarded time. Recall from Appendix A that if P (ω) is the
Fourier transform of P (t), then exp(−iωr/v)P (ω) is the Fourier transform of P (τ = t − r/v).
Also, if P (ω) is the Fourier transform of P (t), then iωP (ω) is the Fourier transform of ∂P/∂t.
Incorporate these relations into equation (H-15) and apply inverse Fourier transform to the
right-hand side to obtain

1 1 ∂P cos θ cos θ ∂P
P (x, y, z; τ ) = + [P ] + dA, (H − 16)
4π A r ∂z r2 vr ∂t

where [P ] means that the integration over the area A is done using the wavefield P at retarded
time τ = t − r/v.
The first term depends on the vertical gradient of the wavefield ∂P/∂z. The second term
is called the near-field term since it decays with 1/r2 . Both terms are neglected in seismic
migration. The remaining third term is called the far-field term and it is the foundation of
Kirchhoff migration. Writing it in discrete form, we have
∆x∆y cos θ ∂
Pout = Pin , (H − 17)
4π vr ∂t
A

where ∆x and ∆y are inline and crossline trace spacings, Pout = P (xout , yout , z; τ = 2z/v) is
the output of migration using the input wavefield Pin = P (xin , yin , z = 0; τ = t − r/v) within
an areal aperture A.

H.2 The Eikonal Equation

The eikonal equation is a ray-theoretical approximation to the scalar wave equation. Its solution
represents wavefronts of constant phase. The wave is propagated from one wavefront to the next
by way of raypaths which are perpendicular to the wavefronts.
Consider a compressional plane wave P (x, y, z; t) in 3-D Cartesian coordinates (x, y, z)
P (x, y, z; t) = P0 exp −iωt + ikx x + iky y + ikz z , (H − 18)
where P0 is the wave amplitude, t is the traveltime, and kx , ky , kz and ω are the Fourier duals
of the variables x, y, z and t, respectively. That this equation is a solution to the 3-D scalar
wave equation
∂2P ∂2P ∂2P 1 ∂2P
2
+ 2
+ 2
= 2 (H − 19)
∂x ∂y ∂z v (x, y, z) ∂t2
can be verified by computing the partial derivatives of the wavefield P (x, y, z; t) given by equation
(H-18) and direct substitution into equation (H-19). The result is the dispersion relation
ω2
kx2 + ky2 + kz2 = (H − 20)
v2
of the scalar wave equation (H-19). Since equation (H-18) satisfies this relation, it is a valid
solution to the scalar wave equation. In equation (H-19), v(x, y, z) is the propagation velocity
of the compressional plane wave P (x, y, z; t).
We shall rewrite the phase term in the plane-wave solution given by equation (H-18) as
kx ky kz
P (x, y, z; t) = P0 exp −iω t − x+ y+ z , (H − 21)
ω ω ω
so that we can define a 3-D traveltime surface T (x, y, z) as
kx ky kz
T (x, y, z) = x + y + z. (H − 22)
ω ω ω
Diffraction and Ray Theory for Wave Propagation 1347

Substitute this definition into equation (H-21) to obtain the expression for the plane wave in
terms of the traveltime surface T (x, y, z)

P (x, y, z; t) = P0 exp −iω t − T (x, y, z) . (H − 23)

Now, we need to verify that this form of the plane wave solution satisfies the scalar wave
equation (H-19). Compute the partial derivative of the wave function P (x, y, z; t) of equation
(H-23) with respect to the space variable x
∂P ∂T
= P0 (iω) exp −iω t − T (x, y, z) (H − 24)
∂x ∂x
and differentiate the result once more to get
∂2P 2 ∂T
2 ∂2T
= −P 0 ω − iω exp −iω t − T (x, y, z) . (H − 25a)
∂x2 ∂x ∂x2
Repeat the differentiation for the space variable y
∂2P ∂T 2 ∂2T
= −P0 ω 2 − iω exp −iω t − T (x, y, z) (H − 25b)
∂y 2 ∂y ∂y 2
and for the space variable z
∂2P ∂T 2 ∂2T
= −P0 ω 2 − iω exp −iω t − T (x, y, z) . (H − 25c)
∂z 2 ∂z ∂z 2
Next compute the partial derivative of the wave function P (x, y, z; t) of equation (H-23) with
respect to the time variable t
∂2P
= −P0 ω 2 exp −iω t − T (x, y, z) . (H − 25d)
∂t2
Substitute equations (H-25a,b,c,d) into equation (H-19) and combine the terms into real
and imaginary parts

∂T 2 ∂T 2 ∂T 2 ∂2T ∂2T ∂2T ω2


ω2 + + − iω 2
+ 2
+ = . (H − 26)
∂x ∂y ∂z ∂x ∂y ∂z 2 v 2 (x, y, z)

Since the term on the right-hand side is real, the imaginary part of the left-hand side has to
vanish, leading to the final expression
∂T 2 ∂T 2 ∂T 2 1
+ + = , (H − 27)
∂x ∂y ∂z v 2 (x, y, z)
which is called the eikonal equation. It gives the traveltime value T (x, y, z) for a ray passing
through a point (x, y, z) in a medium with velocity v(x, y, z).
A solution to the eikonal equation, T (x, y, z) = constant represents the wavefront at an in-
stant of time. Kinematically, a solution to the eikonal equation (H-27) should also be a solution
to the wave equation (H-19). This raises the important question as to under what circumstances
the eikonal equation may be considered a good approximation to the wave equation. To inves-
tigate this matter, we shall consider a plane wave function as in equation (H-23) but with a
spatially varying amplitude P0 (x, y, z) (Officer, 1958):

P (x, y, z; t) = P0 (x, y, z) exp −iω t − T (x, y, z) (H − 28)

and test its validity as a solution to the wave equation (H-19).


Compute the partial derivative of the wave function P (x, y, z; t) of equation (H-28) with
respect to the space variable x
∂P ∂P0 ∂T
= + iωP0 exp −iω t − T (x, y, z) (H − 29)
∂x ∂x ∂x
1348 Seismic Data Analysis

and differentiate the result once more, then simplify to get

∂2P ∂ 2 P0 ∂T 2 ∂P0 ∂T ∂2T


= − ω 2 P0 + iω 2 + P0 2 exp −iω t − T (x, y, z) .
∂x2 ∂x2 ∂x ∂x ∂x ∂x
(H − 30a)
Repeat the differentiation for the space variables y

∂2P ∂ 2 P0 ∂T 2 ∂P0 ∂T ∂2T


= − ω 2 P0 + iω 2 + P0 2 exp −iω t − T (x, y, z)
∂y 2 ∂y 2 ∂y ∂y ∂y ∂y
(H − 30b)
and z
∂2P ∂ 2 P0 ∂T 2 ∂P0 ∂T ∂2T
= − ω 2 P0 + iω 2 + P0 2 exp −iω t − T (x, y, z) .
∂z 2 ∂z 2 ∂z ∂z ∂z ∂z
(H − 30c)
Next compute the partial derivative of the wave function P (x, y, z; t) of equation (H-28) with
respect to the time variable t
∂2P
= −P0 ω 2 exp −iω t − T (x, y, z) . (H − 30d)
∂t2
Substitute equations (H-30a,b,c,d) into equation (H-19), and combine the terms into real
and imaginary parts to obtain

∂T 2 ∂T 2 ∂T 2 ∂ 2 P0 ∂ 2 P0 ∂ 2 P0
ω 2 P0 + + − 2
+ 2
+
∂x ∂y ∂z ∂x ∂y ∂z 2

∂P0 ∂T ∂P0 ∂T ∂P0 ∂T ∂2T ∂2T ∂2T ω2


−iω 2 + + + P0 + 2 + 2 = P0 .
∂x ∂x ∂y ∂y ∂z ∂z ∂x2 ∂y ∂z v2
(H − 31)
Equate the real parts of both sides of equation (H-31) to get

∂T 2 ∂T 2 ∂T 2 1 ∂ 2 P0 ∂ 2 P0 ∂ 2 P0 1
+ + − 2
+ + = , (H − 32a)
∂x ∂y ∂z ω P0 ∂x2 ∂y 2 ∂z 2 v2

and equate the imaginary parts to get

2 ∂P0 ∂T ∂P0 ∂T ∂P0 ∂T ∂2T ∂2T ∂2T


+ + + + + = 0. (H − 32b)
P0 ∂x ∂x ∂y ∂y ∂z ∂z ∂x2 ∂y 2 ∂z 2

Now we analyze the implications of equations (H-32a,b). To reduce equation (H-32a) to the
eikonal equation (H-27), it follows that the second term on the left-hand side of equation (H-32a)
must vanish. This is only possible if we let 1/ω go to zero, or in other words, if we make the
high-frequency assumption. Thus, for a wave function, such as that given by equation (H-23),
with spatially varying amplitudes, the eikonal equation is a good approximation to the wave
equation at the high-frequency limit. Additionally, for T (x, y, z) to be a good approximation to
the solution of the wave equation, the first term on the left-hand side of equation (H-32b) must
be negligible compared to the second term. In fact, the wave function given by equation (H-23)
is a special form of the generalized asymptotic ray series solution given by (Hubral and Krey,
1980)

P (x, y, z; t) = Pn exp −iω t − T (x, y, z) (iω)−n , (H − 33)
n=0
Diffraction and Ray Theory for Wave Propagation 1349

where n = 0, 1, 2, . . .. Note that the terms for n ≥ 1 in equation (H-33) become insignificant
for a high-frequency source. Specifically, in the high-frequency limit, the summation given by
equation (H-33) reduces to the wave function of equation (H-23).
Since λ = 2πv/ω, where λ is the wavelength, the high-frequency limit is equivalent to
small wavelengths. How small the wavelength should be for a solution of the eikonal equation
to be a good approximation to the wave equation is an important consideration in the practical
validity of the eikonal equation itself. The approximation is valid if the fractional change in
the velocity gradient ∆v is much less than the frequency v/λ (Officer, 1958). In practice, this
approximation will not be met across a layer boundary with a sharp velocity contrast or in a
layer with velocity variations that occur within a spatial extent much less than the wavelength.
As a direct consequence of this approximation, the eikonal equation is a good approximation to
the wave equation only for a medium in which velocities do not vary rapidly, and especially, do
not exhibit sharp discontinuities.

H.3 Finite-Difference Solution to the Eikonal Equation

Just as the scalar wave equation itself can be solved using a finite-difference scheme to extrapolate
a wavefield from one depth level to another (Section 4.1), the eikonal equation can be solved
using a finite-difference scheme to compute the traveltimes along wavefronts expanding from
a source (Qin et al., 1992; Reshef and Kosloff, 1986; Vidale, 1988; Podvin and Lecomte, 1991;
Vinje et al., 1993). For simplicity, we shall rewrite the eikonal equation (H-27) for the 2-D case
in (x, z) coordinates
2 2
∂T ∂T 1
+ = . (H − 34)
∂x ∂z v 2 (x, z)
Consider the 2-D computation mesh sketched in Figure H-2a. We want to compute the traveltime
T of the eikonal equation (H-34) at grid point (x + ∆x, z + ∆z) using the known traveltimes at
grid points (x, z) and (x + ∆x, z).
Computing the traveltimes at depth z + ∆z from those at depth z means extrapolating T
in the z-direction. Rewrite equation (H-34) in the form of an extrapolation equation (Reshef
and Kosloff, 1986)
2
∂T 1 ∂T
=± − , (H − 35a)
∂z v2 ∂x
which can be rearranged to take the form
2
∂T ∂T
v = 1− v . (H − 35b)
∂z ∂x
Note that in equation (H-35b) we also have dropped one of the two solutions and only considered
that which yields an increase in traveltime as we march in depth.
Now replace the differential operators with difference operators
2 1/2
v(z + ∆z)T (z + ∆z) − v(z)T (z) v(x + ∆x)T (x + ∆x) − v(x)T (x)
= 1− ,
∆z ∆x
(H − 36a)
and rewrite explicitly in terms of T (z + ∆z)
2 1/2
∆z v(x + ∆x)T (x + ∆x) − v(x)T (x)
T (z + ∆z) = T (z) + 1− , (H − 36b)
v(z + ∆z) ∆x
1350 Seismic Data Analysis

FIG. H-2. Finite-difference mesh used for (a) 2-D and (b) 3-D solution to the eikonal equation. See
text for details.

where we have assumed that v(z)/v(z +∆z) is approximately unity. Also, we have avoided in our
algebra crowding the terms in equation (H-36b) with the dual variables (x, z). Now introduce
the dual variables explicitly to get the final expression for a finite-difference solution to the 2-D
eikonal equation
k k 2 1/2
k+1 k ∆z vi+1 Ti+1 − vik Tik
Ti+1 = Ti+1 + k+1 1 − , (H − 37)
vi+1 ∆x

where Tik is the discrete form of T (x, z) as labeled in Figure H-2a.


Now consider the 3-D computation mesh sketched in Figure H-2b. We want to compute the
traveltime T of the eikonal equation (H-27) at grid point (x + ∆x, y, z + ∆z) using the known
traveltimes at grid points (x, y, z), (x + ∆x, y, z) and (x + ∆x, y + ∆y, z). The 3-D equivalent of
equation (H-35b) is
2 2
∂T ∂T ∂T
v = 1− v − v . (H − 38)
∂z ∂x ∂y
Following the same algebra as for the 2-D case, we obtain the finite-difference solution to the
3-D eikonal equation
2 21/2
k k k k k k k k
k+1 k ∆z vi+1,j Ti+1,j − vi,j Ti,j vi+1,j+1 Ti+1,j+1 − vi+1,j Ti+1,j
Ti+1,j = Ti+1,j+ k+1 1− − ,
vi+1,j ∆x ∆y
(H − 39)
k
where Ti,j is the discrete form of T (x, y, z) as labeled in Figure H-2b.
Diffraction and Ray Theory for Wave Propagation 1351

Note that in equations (H-36), (H-37), and (H-39), unlike the previous authors (Podvin and
Lecomte, 1991; Reshef, 1991), the indices for the velocity v and traveltime T are kept the same.
This was done with the stability of the finite-difference solution of the eikonal equation in mind.
Actually, Kjartannson (1979) has applied the same notion to the finite-difference solution to the
scalar wave equation.

REFERENCES

Abma, R., Sun, J., and Bernitsas, N., 1999; Antialiasing methods in Kirchhoff migration: Geo-
physics, 64, 1783-1792.
Al-Yahya, K., 1989, Velocity analysis by iterative profile migration: Geophysics, 54, 718-729.
Berryhill, J.R., 1979, Wave-equation datuming: Geophysics, 44, 1329-1333.
Berryhill, J.R., 1984, Wave-equation datuming before stack: Presented at the 54th Ann. Internat.
Mtg., Soc. Expl. Geophys.
Biondi, B., Fomel, S., and Chemingui, N., 1998, Azimuth moveout for 3-D prestack imaging: Geo-
physics, 63, 574-588.
Biondi, B. and Palacharla, G., 1996, 3-D prestack migration of common-azimuth data: Geophysics,
61, 1822-1832.
Červený, V., Klimes, L., and Pšenčı́k, I., 1984, Paraxial ray approximation in the computation of
seismic wavefields in inhomogeneous media: Geophys. J. Roy. Astr. Soc., 79, 89-104.
Červený, V., Popov, M. M., and Pšenčı́k, I., 1982, Computation of wavefields in inhomogeneous
media — Gaussian beam approach: Geophys. J. Roy. Astr. Soc., 70, 109-128.
Chemingui, N. and Biondi, B., 1999, Data regularization to inversion to common offset (ICO): 69th
Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1398-1401.
Claerbout, J.F., 1985, Imaging the earth’s interior: Blackwell Scientific Publications.
Claerbout, J.F., 1992, Earth soundings analysis: Blackwell Scientific Publications.
Cox, H. L. H. and Wapenaar, C. P. A., 1992, Macromodel estimation by common-offset migration
and by shot-record migration: J. Seis. Expl., 1, 29-37.
Deregowski, 1990, Common-offset migrations and velocity analysis: First Break, 8, 225-234.
Coulson, C. H., 1965, Waves: A mathematical account of the common type of wave motion: Oliver
and Boyd.
Gray, S., 1992, Frequency-selective design of the Kirchhoff migration operator: Geophys. Prosp.,
40, 565-571.
Hubral, P., 1977, Time migration — some ray-theoretical aspects: Geophys. Prosp., 25, 738-745.
Hubral, P. and Krey, T., 1980, Interval velocities from seismic reflection time measurements: Soc.
Expl. Geophys.
Jackson, J. D., 1962, Classical electrodynamics: John Wiley and Sons.
Judson, D. R., Lin, J., Schultz, P.S., and Sherwood, J.W.C., 1980, Depth migration after stack:
Geophysics, 45, 376-393.
Keho, T. H. and Beydoun, W. B., 1988, Paraxial ray Kirchhoff migration: Geophysics, 53, 1540-
1546.
Kjartansson, E., 1979, Modeling and migration by the monochromatic 45-degree equation: Stanford
Expl. Proj. Rep. No. 15.
Kosloff, D. and Kessler, D., 1987, Accurate depth migration by a generalized phase-shift method:
Geophysics, 52, 1074-1084.
Larner, K. L., Hatton, L., and Gibson, B., 1981, Depth migration of imaged time sections: Geo-
physics, 46, 734-750.
Lass, H., 1950, Vector and tensor analysis: McGraw-Hill Book Co.
Lecomte, I., 1999, Local and controlled prestack depth migration in complex areas: Geophys. Prosp.,
47, 799-818.
Lee, W. B. and Zhang, L., 1992, Residual shot-profile migration: Geophysics, 57, 815-822.
1352 Seismic Data Analysis

Lumley, D. E., Claerbout, J. F., and Bevc, D., 1994, Antialiased Kirchhoff 3-D migration: 64th
Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1282-1285.
Meshbey, V., Kosloff, D., Ragoza, Y., Meshbey, O., Egozy, U., and Cozzens, J., 1993, A method
for computing traveltimes for an arbitrary velocity model: Presented at the 45th Ann. EAGE
Mtg.
Moser, T. J., 1991, Shortest-path calculation of seismic rays: Geophysics, 56, 59-67.
Nichols, D. E., 1996, Maximum-energy traveltimes calculated in the seismic frequency band: Geo-
physics, 61, 253-263.
Chemingui, N. and Biondi, B., 1999, data regularization by inversion to common offset (ICO):
Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 1398-1401.
O’Brien, M. and Gray, S., 1996, Can we image beneath salt?: The Leading Edge, 17-22.
Officer, C. B., 1958, Introduction to the theory of sound transmission: McGraw-Hill Book Co.
Podvin, P. and Lecomte, I., 1991, Finite-difference computation of traveltimes for velocity-depth
models with strong velocity contrast across layer boundaries — a massively parallel approach:
Geophys. J. Int., 105, 271-284.
Qin, F., Luo, Y., Olsen, K. B., Cai, W., and Schuster, G. T., 1992, Finite-difference solution of the
eikonal equation along expanding wavefronts: Geophysics, 57, 478-487.
Reshef, M., 1991, Prestack depth imaging of three-dimensional shot gathers: Geophysics, 56, 1158-
1163.
Reshef, M. and Kosloff, D., 1986, Migration of common-shot gathers: Geophysics, 51, 324-331.
Sethian, J. A. and Popovici, A. M., 1999, 3-D traveltime computation using the fast marching
method: Geophysics, 64, 516-523.
Schultz, P.S. and Sherwood, J.W.C., 1980, Depth migration before stack: Geophysics, 45, 361-375.
Taner, M.T., Cook, E.E., and Neidell, N.S., 1982, Paleo-seismic and color acoustic impedance sec-
tions: applications of downward continuation in structural and stratigraphic context: Presented
at the 52nd Ann. Internat. Mtg., Soc. Expl. Geophys.
Vesnaver, A., 1996, Ray tracing based on Fermat’s principle in irregular grids: Geophys. Prosp.,
44, 741-760.
Vidale, J., 1988, Finite-difference calculation of traveltimes: Bull. Seis. Soc. Am., 78, 2026-2076.
Vinje, V., Iversen, E., and Gjoystdal, H., 1993, Traveltime and amplitude estimation using wave-
front construction: Geophysics, 58, 1157-1166.
Yilmaz, O. and Lucas, D., 1986, Prestack layer replacement: Geophysics, 51, 1355-1369.

You might also like