Sie sind auf Seite 1von 7

GEOPHYSICS, VOL. 64, NO.5 (SEPTEMBER-OCTOBER 1999); P 1546-1552,10 FIGS.

Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Manual seismic reflection tomography

Gary E. Murphy* and Samuel H. Gray$

ABSTRACT through both the current velocity model and the veloc-
ity model plus suggested model perturbations. The dif-
Prestack depth migration needs a good velocity model ferences between those sets of traveltimes are related
to produce a good image; in fact, finding the velocity to differences in depth, allowing the user to preview the
model is one of the goals of prestack depth migration. Mi- approximate effects of a velocity change on the CRP
gration velocity analysis uses information produced by gathers without remigrating the data. As with automatic
the migration to update the current velocity model for tomography, the computed depth differences are essen-
use in the next migration iteration. Several techniques tially backprojected along raypaths through the model,
are currently used to estimate migration velocities, rang- yielding a velocity update that flattens the gathers. Un-
ing from trial and error to automatic methods like reflec- like automatic tomography, in which an algebraic inverse
tion tomography. Here, we present a method that com- problem is solved by the computer for all geologic layers
bines aspects of some of the more accurate methods into simultaneously, our method estimates shallow velocities
an interactive procedure for viewing the effects of resid- before proceeding deeper and requires substantial user
ual normal moveout corrections on migrated common intervention, both in flattening individual CRP gathers
reflection point (CRP) gathers. The residual corrections and in deciding the appropriateness of the suggested ve-
are performed by computing traveltimes along raypaths locity updates in individual geologic units.

INTRODUCTION vide an indication that the velocity/depth function is correct,


because using the correct velocity/depth function will give mi-
Geophysicists use prestack depth migration in structural grated images that are kinematically identical from one off-
settings to obtain an accurate subsurface image and to esti- set to another. In principle, velocity analysis would appear to
mate geologic velocities. While migration algorithms and com- be easier after migration than before migration because the
putational advances have allowed us to achieve the first of migration has supplied most of the normal moveout (NMO)
these objectives, the second objective remains elusive. Sort- correction, leaving only residual NMO corrections to be per-
ing prestack migrated images into gathers indexed by surface formed. In practice, however, the residual NMO present in mi-
location [called migrated common depth point (CDP) gathers, grated CRP gathers has been notoriously difficult to remove
common image gathers, or, in this paper, common reflection when structural complexities are present. The major reason
point (CRP) gathers] allows us to perform velocity analysis af- for this difficulty is the effect of the interval velocity on the
ter migration, just as velocity analysis can be performed before migrated images. In stacking velocity analysis for unmigrated
migration on unmigrated common midpoint (CMP) gathers. or prestack time-migrated gathers, we choose at each point in
Changing the velocity/depth model before remigrating will re- the earth a velocity value that will best flatten the gathers; this
sult in changes both in the migrated image and in the moveout imaging velocity is independent of the imaging velocity chosen
on the CRP gathers. The correct changes yield a geologically at other—even nearby—locations. The resulting velocity/time
plausible model that causes the residual moveout to disappear function is not expected to relate to geology—only to imaging.
upon remigration, leaving flattened CRP gathers that stack In prestack depth migration, on the other hand, changing an
into a clearly focused final image. Flattened CRP gathers pro- interval velocity value at a point will affect the remigrated CRP

Manuscript received by the Editor March 3, 1998; revised manuscript received January 15, 1999.
*Formerly BP/Amoco, Tulsa, Oklahoma 74102; currently Consultant, 8912 South College Place, Tulsa, Oklahoma 74137. E-mail: gary_murphy
@bigfoot.com.
$Formerly Amoco Canada Petroleum Co., Calgary AB T2P 2H8, Canada; currently Veritas DGC Inc., 715 Fifth Avenue S.W. Suite 2200, Calgary,
Alberta T2P 5A2, Canada. E-mail: Sam_Gray@veritasdgc.com.
© 1999 Society of Exploration Geophysicists. All rights reserved.
1546

Manual Seismic Reflection Tomography 1547

gathers—not only at the analysis point, but at all other locations of one another, attributing all the residual moveout to veloc-
in a cone beneath the point. Often, a shallow velocity change ity errors directly above subsurface analysis points. However,
inserted to flatten a CRP gather at one depth will have the de- when lateral velocity variations are significant or the geologic
sired effect at that location but will insert unwanted residual structures being imaged are dipping, the breakdown of these
moveout below it. In addition, the interval velocity model is of- assumptions can lead to errors that make the methods even
ten expected to incorporate known velocity values or geologic more problematic than guessing interval velocity updates from
features, and the information provided by these constraints can visual inspection of the CRP gathers. By contrast, Lafond and
affect the velocity estimation process. Levander (1993) and Liu (1997) present methods that allow the
Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Several methods have been proposed to address the difficul- velocity to vary laterally above the body being updated; these
ties of prestack depth migration velocity analysis. These range methods use ray tracing through the current velocity model
from ad hoc manual methods, such as guessing interval veloc- to relate residual moveout, expressed as depth variations as a
ity updates based on visual inspection of the CRP gathers, to function of offset for a reflection event in a CRP gather, with
systematic automatic methods, such as reflection tomography. a velocity update.
Guessing interval velocity updates to input into a remigration The methods of Lafond and Levander (1993) and Liu (1997),
scheme has proven tedious and frustratingly slow to converge. which trace rays through the current velocity/depth model to
Often, several tens of migration iterations are needed before relate depth discrepancies with velocity updates, are much
the geophysicist is satisfied that one of the early iterations is as closer in spirit to tomography than are the other methods.
close to the correct picture (migrated image plus velocity/depth They also have much greater accuracy in the presence of strong
function) as possible. Although the computer expense of cal- lateral velocity variations or significant reflector dip than the
culating all the iterations is much less a factor today than it methods that assume laterally constant velocity and flat reflec-
was a few years ago, the value of geophysicists' time is too high tors. Also, they express the velocity model as distinct units to be
to spend in many fruitless migration iterations. At the other analyzed from shallow to deep. (On the other hand, these meth-
extreme is reflection tomography (Stork, 1992), an automatic ods differ from tomography in that they proceed from shallow
method whose chief objective is to flatten CRP gathers. Using layers to deep layers, while tomography simultaneously cor-
picks along reflecting horizons, tomography automatically ad- rects residual moveout problems on reflectors at all depths. As a
justs the velocity model to minimize the discrepancies between consequence, they are subject to error accumulation at depth.)
the picks for all the horizons upon remigration. Tomography The velocity models resulting from these methods are usually
accomplishes this by setting up an algebraic inverse problem to more geologically plausible in areas where velocity contrasts
convert the residual moveout resulting from the pick discrep- tend to align with rock unit boundaries than those resulting
ancies to velocity modifications (Stork, 1992). Tomography acts from any other velocity estimation method.
globally on the velocity model; it produces a velocity update We present a method that generalizes aspects of many of
intended to remove residual moveout from all events, shal- the procedures described so far, although it most resembles
low and deep, at the same time. Thus, tomography avoids the an interactive, graphical version of the methods of Lafond and
problem of the ad hoc methods where velocity updates often Levander (1993) and Liu (1997). The method treats prestack
introduce residual moveout problems that were not present on depth migration as a model-based NMO correction which
previous iterations. On the other hand, by backprojecting or av- needs revision because of errors in the velocity model. It uses
eraging pick discrepancies over many grid cells in the velocity ray tracing between reflecting horizons and the source and re-
model, tomography can produce unrealistically smoothed ve- ceiver locations to compute the effects of provisional velocity
locity updates, and it is often difficult to constrain geologically. updates on the residual NMO. Without remigrating, it finds ve-
The automatic nature of tomography can also be a disadvan- locities to correct the residual NMO, or flatten the CRP gathers,
tage, in that it can be used to flatten CRP gathers and produce on the shallow events first; then, with the updated velocities set
an image and a velocity model (with very little intervention in the shallow part of the section, it corrects the residual NMO
from an interpreter) that do not relate well to the actual geol- on the deeper events. This is all performed interactively on a
ogy. When used this way—minimizing residual moveout with- graphics workstation. Sometimes it is impossible to flatten all
out incorporating geologic constraints—tomography finds one events on the CRP gathers that are associated with some parts
of many possible local solutions to an optimization problem, of the model. Automatic tomography solves that problem by
with little likelihood of finding the global solution to the geo- compromise: recognizing that the CRP gathers will not all be
logically constrained optimization problem. flat after the operation, it attempts to flatten them as well as
Between the extremes of trial and error and tomography lie possible by using some optimal solution (e.g., least squares) to
a spectrum of methods. Jeannot et al. (1986) describe depth- an ill-posed inverse problem. Our method similarly optimizes
focusing analysis, a systematic method for estimating vertical the solution for the velocity update by allowing the user to de-
velocity updates from partially focused depth images under cide the velocities to be inserted as an update into each part
fairly restrictive assumptions. MacKay and Abma (1992) in- of the model. We refer to our method as manual tomography
vestigate and expand the limitations of depth-focusing analysis. for that reason: it is an interactive tool to backproject residual
Al-Yahya (1989) adapts a similar technique to migrated gath- NMO on CRP gathers into velocity updates, where the inver-
ers, providing a residual velocity analysis method for common sion takes place in the user's eyeibrain system and not in the
shot migration. Deregowski (1990) also presents a method for computer.
residual velocity analysis, to be applied to CRP gathers from As with all the other methods listed above, our method pro-
common offset migration. Common to all these methods, ex- vides updates for the velocity model, to be used in the next
cept for Al-Yahya's, are the assumptions that the velocity does iteration of prestack depth migration. The residual NMO cor-
not vary laterally and that reflectors are horizontal. These as- rection obtained by updating the velocity model using our
sumptions allow CRP gathers to be analyzed independently method serves as a crude approximation to remigration. This

1548 Murphy and Gray

approximate remigration is performed interactively on a single times along raypaths shot upward from the analysis location
reflecting horizon, working on a few neighboring CRP gathers into the receiver spread (Figure 5). We emphasize that we per-
at a time. Using raypaths through the laterally varying veloc- turb the traveltimes only along the raypath segments within
ity model with its current updates, it varies traveltimes along the active body. But if we have previously updated velocities in
the raypaths by scanning over a range of percentage slow- regions above the active body, those fixed velocities are used
ness (reciprocal velocity) perturbations within a single body. to compute traveltimes through the overburden. By relating
The raypaths themselves are held constant during the slowness the time differences along raypaths to depth differences as
scans within an individual residual NMO calculation. As a re- explained by Al-Yahya (1989), we perform a residual NMO
Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

sult of this linearization, when accumulated velocity updates correction on the migrated gathers using the trial slowness per-
through one or more regions of the velocity model become turbations. This residual NMO correction approximates the ef-
large, the method loses accuracy, requiring an actual remigra- fects of remigration using the trial slowness perturbation plus
tion before further velocity updates can be estimated during the all the updated velocities in the overburden. This is the heart
next iteration. However, when accumulated velocity updates of the method; the limited number of computations—far fewer
are relatively small (usually <10% of the migration velocity than a remigration—allows a large number of possible per-
throughout the entire model), approximating remigration by turbations to be scanned almost instantaneously on a modern
our residual NMO correction is often surprisingly accurate. workstation.
Because of the heavy user intervention it requires, this We then select the slowness perturbation that best flattens
method is much less automatic and much more interpretive the gathers in the vicinity of the analysis location (Figure 6).
and user driven than tomography. This is an advantage over In our example, decreasing the slowness along the raypaths by
tomography when the process is driven by an interpreter and
a possible disadvantage when it is driven by a processor with
very little interpreting expertise.
Although we believe our implementation is unique, the tech-
nique described here resembles interactive methods presented
elsewhere (e.g., Schmid et al., 1995). However, we have not
seen a general description of the method, which we believe
is an important velocity estimation method and will gain in
importance as it is increasingly applied in three dimensions.

DESCRIPTION OF THE METHOD

We use a synthetic example to describe the method; later,


we illustrate its use on a field example. The synthetic example
(Figure 1; Gray and Marfurt, 1995) is derived from some struc-
tural styles in the Canadian Foothills. It depicts a cross-section
25 000 m long and 10 000 m deep, with velocities ranging from
3600 to 6000 m/s.
We begin by performing a prestack depth migration with an
initial velocity model. Our migration is a common-offset Kirch-
hoff migration. Initially, we used a laterally invariant (v(z))
velocity function with a slope different from the background
gradient in the correct model. Horizons along the migrated,
stacked section are picked (Figure 2); these horizons divide
the velocity model into distinct units, or bodies. These bodies
are investigated one at a time, from top to bottom. The method
typically works best with five to ten bodies from top to bottom,
where all the bodies are thick enough to yield useful residual
NMO information. Our example contains a number of thin
bodies, thereby providing an extreme test for the method. As-
suming the velocities have been updated in all the bodies above
the present, or active, body (Figure 3), we update the velocities
in the active body as follows.
First, we select a few neighboring CRP gathers around an
image location at the base of the active body that shows signif-
icant energy on the migrated section (Figure 4). These gathers,
whose horizontal axis is source-to-receiver offset, decreasing FIG. 1. Synthetic data based on geologic structures in the
from 3600 m on the left to 0 m on the right, show a significant Canadian Foothills. (a) The velocity/depth model used to gen-
amount of residual NMO because of the errors in the original erate finite-difference shot records used in the prestack depth
migrations. The cross-section is 25 km long and 10 km deep.
velocity function. The pink line near the top of the section is the topographic sur-
Next, we test the effects of percentage slowness changes face. (b) Prestack depth migration using the correct velocities
within the active body interactively, by perturbing the travel- (from Gray and Marfurt, 1995).

Manual Seismic Reflection Tomography 1549

7% removes most (though not all) of the residual NMO from


the CRP gathers at the analysis location. We can select several
more analysis points at the base of the active body and repeat
the scan over slowness perturbations, obtaining velocity up-
dates along many ray fans (Figure 7). Often, two or more ray
fans from nearby analysis points overlap within the active body,
leading to a possible conflict between slowness updates chosen
at the analysis points. Where tomography resolves these con-
Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

flicts by applying an optimization criterion such as least squares,


the present method forces the user to resolve the conflicts by ac-
cepting one choice or the other for the updated velocity/depth
function within the active body or by making a compromise

FIG. 4.Migrated CRP gathers from five adjacent locations on


the right side of the active body. The ray fan in Figure 5 shows
the center of the five locations. Within each gather, offsets
range from 3600 m for the trace at the left to 0 m for the trace at
the right. These gathers, windowed to show only energy from
the reflector at the base of the active body, show significant
residual NMO after the initial iteration of prestack depth mi-
FIG. 2. Middle part of the initial v(z) migration with picked gration.
horizons superimposed. These horizons are combined to form
bodies, and the velocity will be updated inside each body.

FIG. 5. A ray fan from the analysis location of Figure 4 shot


upward toward the recording spread. The segments of these
FIG. 3. Middle part of the velocity model, with the active body raypaths within the active body are used to compute residual
surrounded by bold red lines. Velocities have been updated NMO interactively, with a number of percentage slowness per-
above the active body but not below it. turbations of the original velocity function.

1550 Murphy and Gray

between the conflicting choices that might incorporate geo- in a body, the body becomes thicker. Thus, reflecting bound-
logic knowledge. aries move as the velocity is updated. Although this stretching
Finally, the slowness perturbation values chosen at the analy- of the body should occur in the direction of the velocity gradi-
sis locations are combined, yielding a laterally varying velocity ent when the reflector are aligned with the velocity field (Stork,
update within the active body that the user is free to accept or 1992), we approximate the stretch with a simple depth stretch,
reject (Figure 8). In our implementation, the velocity update incurring the greatest error at the steepest dips.
can take on any of several functional forms within the active Having updated the velocity within a body, we proceed
body, such as constant, v(0) + kz (possibly different at each downward through the image to the next body without remi-
Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

lateral location), or a gradient perpendicular to the top or base grating. Ideally, we continue to the bottom of the image. Often,
of the body. Finally, when the velocity is updated in a body, the however, in cases of large velocity updates or poorly imaged
thickness of the body is changed to preserve the traveltime of initial sections, very little reliable information can be obtained
rays through the body. For example, if the velocity is increased from the deeper bodies before the shallow velocities are com-
pletely corrected. In these cases, it is efficient to update only the
shallow velocities in the earliest iterations and then remigrate
before proceeding deeper in the section. When we do this, we
usually find the deeper reflectors to be imaged well enough
after one or two iterations to obtain reliable velocity updates
from deeper bodies. As with all methods that attempt to correct
residual NMO, our method loses effectiveness when applied to
bodies whose depth is much greater than a spread length.
On our synthetic data set, we performed three iterations,
FIG. 6. The results of a ray-based residual NMO correction per- each consisting of prestack depth migration followed by veloc-
formed on the CRP gathers in Figure 4. Reducing the velocity ity analysis using the manual tomography method. Our fourth
along the raypaths of Figure 5 by 7% has nearly flattened the migrated, stacked section is shown in Figure 9 along with our
gathers, indicating that the velocity in the active body above final velocity model. Figure 9 should be compared with Fig-
the analysis point should be reduced in the next migration it- ure 1. Overall, the final velocity model is structurally similar to
eration.

FIG. 7. Choosing several analysis points along the base of the


active body allows residual NMO calculations to be performed FIG. 8. The velocity update suggested by the residual NMO
across the entire active body. The ray fans can intersect each calculations have been inserted into the active body to be ac-
other, possibly giving conflicting information that the user must cepted or rejected by the user. Velocities above and below the
interpret. active body remain unchanged.

Manual Seismic Reflection Tomography 1551

the correct model, although there are many errors in the fine
details of the model. These errors are because of the inabil-
ity of tomography, or any method based on residual moveout
calculations, to resolve thin layers, through which even large
velocity changes provide very little change in NMO. The final
stack has positioned most of the events correctly, although the
image is not as well focused as the image obtained using the
correct velocity model. This image is about as good as one can
Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

hope to obtain using manual tomography, which uses the long-


and medium-wavelength information contained in the CRP
gathers.

FIELD DATA EXAMPLE

In 1995, Husky and Talisman released a data set from the


southern Alberta (Canada) Foothills for the purpose of indus-
try and university investigations. This data set was the subject
of a workshop on structural imaging at the 1995 SEG An-
nual Meeting (Stork et al., 1995). The data quality is good,
with many reflectors clearly visible on the unmigrated records
and on preliminary migrations. As with our synthetic example,
we performed several iterations, each consisting of prestack
depth migration followed by manual tomography. Here, how-
ever, we used geologic knowledge to constrain the procedure.
Specifically, we restricted the basement reflector to lie between
certain depths, and we restricted the amount of variation al-
lowed in the dip of the basement reflector. This strategy is
more difficult to apply than using manual tomography only to
flatten the CRP gathers because it forces us to resolve con-
flicts interpretively between the CRP gathers and the geologic
constraints. Stated differently, we discovered that for this data
set, as for many others, the best imaging velocity does not al-
ways agree with the correct geologic velocity. (In automatic
FIG. 9. Results after three iterations of migration velocity anal-
tomography, geologic constraints can be incorporated into the
ysis. (a) The velocity model. The velocity analysis has captured inversion with a certain amount of effort, resulting in conflicts
most of the large-scale features of the actual velocity model. between the CRP gathers and the geology being resolved with-
(b) The prestack depth-migrated image. out involving the user. This leads to quicker convergence to a

FIG. 10. Prestack depth migration of the Husky/Talisman Canadian Foothills line obtained after several iterations
of migration and manual tomography. Geologic constraints (e.g., constraining the depth and dip of the basement
reflector) were applied.

1552 Murphy and Gray

solution that satisfies the criteria of the mathematical inverse vantage: by working very closely with the seismic data, the
problem at the heart of automatic tomography, although the user becomes much more intimately involved in the interpre-
solution does not necessarily satisfy the criteria of the geologic tive process than is possible with automatic tomography.
problem addressed by the seismic data set.) Velocity estimation remains a major problem in depth imag-
Because of the added interpretational difficulties inherent ing, and model studies have shown us that uniquely determin-
in resolving these conflicts, we needed to perform about ten ing the correct velocity/depth model for prestack depth migra-
iterations. Figure 10 shows the final image obtained with this tion is impossible in many—if not most—cases of structural
procedure. This image compares well with other results ob- complexity. Adding to the uncertainty, and frequently causing
Downloaded 03/07/15 to 137.195.150.201. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

tained from this data set (e.g., Zhu and Lines, 1997), and the imaging velocities to differ significantly from geologic veloci-
velocities obtained were geologically plausible. Significantly, ties, are 3-D effects and effects of anisotropy. However, both
however, the seismic image in Figure 10 is not an obvious im- automatic and manual tomographic methods have taken us a
provement over an image (not shown) obtained using only long way past earlier methods of velocity estimation in speed
three iterations of manual tomography without applying the and in accuracy. These methods, combined with the increased
geologic constraints. In our experience, incorporating geologic speed of prestack depth migration brought about by increases
constraints to obtain the most correct image often fails to pro- in computing power, allow us to test more geologic hypotheses
vide the most pleasing image, in the sense of reflector continuity than before, reducing the uncertainties in the final interpre-
or focusing. tation. We will never eliminate these uncertainties completely
using the seismic method, but we can at least envision reducing
DISCUSSION them to an acceptable level for typical amounts of structural
We have presented a method for prestack depth migra- complexity that we may encounter in exploring for hydrocar-
tion velocity analysis, which we call manual tomography. This bons.
method, based on removing residual NMO from migrated ACKNOWLEDGMENTS
CRP gathers, is an interactive variant of methods presented
by Lafond and Levander (1993) and Liu (1997). It resembles We thank Gary Maclean for providing the synthetic data
methods currently in use that have not been described ade- set and Christof Stork for providing the field data set. We
quately in the geophysical literature (e.g., Schmid et al., 1995), also thank our many colleagues within Amoco who have sug-
and we believe that it represents a significant improvement gested improvements to our velocity analysis method, and
in accurate velocity model determination over currently used three reviewers who suggested numerous improvements to our
methods that assume flat reflectors and lateral invariance of exposition.
the model above each analysis point. REFERENCES
Manual tomography, like automatic tomography, attempts
to convert residual NMO on migrated CRP gathers into an up- Al-Yahya, K., 1989, Velocity analysis by iterative profile migration:
Geophysics, 54, 718-729.
dated velocity model for the next migration iteration. While Deregowski, S. M., 1990, Common-offset migrations and velocity anal-
automatic tomography sets up an algebraic inverse problem ysis: First Break, 8, 225-234.
to perform this conversion, providing the updated velocities Gray, S. H., and Marfurt, K. J., 1995, Migration from topography: Im-
proving the near-surface image: Can. J. Expl. Geophys., 31,18-24.
with minimal user intervention, manual tomography forces the Jeannot, J. P., Faye, J. E, and Denelle, E., 1986, Prestack migration
user to perform the inversion by explicitly deciding to accept velocities from depth-focusing analysis: 56th Ann. Internat. Mtg.,
or reject velocity updates in individual bodies proposed by the Soc. Expl. Geophys., Expanded Abstracts, 438-440.
Lafond, C. F., and Levander, A. R., 1993, Migration moveout analysis
program. These decisions might be easy, i.e., merely accepting and depth focusing: Geophysics, 58, 91-100.
velocity updates suggested by the residual NMO calculations. Liu, Z., 1997, An analytical approach to migration velocity analysis:
Geophysics, 62, 1238-1249.
Or they might be more difficult, i.e., deciding between geologic MacKay, S., and Abma, R., 1992, Imaging and velocity estimation with
information and proposed velocity updates. This makes our depth-focusing analysis: Geophysics, 57, 1608-1622.
method much more interpretive than automatic tomography, Schmid, R., Link, B., and Butler, P., 1995, Model manipulation plus
interactive depth image analysis using model based NMO: 22nd Ann.
with advantages and disadvantages. The major disadvantage, Cony., Can. Soc. Expl. Geophys., Expanded Abstracts, 51-52.
as we saw with our field data example, is the prospect of slower Stork, C., 1992, Reflection tomography in the postmigrated domain:
convergence than with automatic tomography because of the Geophysics, 57, 680-692.
Stork, C., Welsh, C., and Skuce, A., 1995, Demonstration of processing
interpretational difficulty of resolving conflicting information, and model building methods on a real complex structure data set:
either between geologic knowledge and seismic velocities or 65th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts,
Workshop 6.
between information contained in individual CRP gathers. The Zhu, J., and Lines, L. R., 1997, Implicit interpolation in reverse-time
major advantage is actually another aspect of the major disad- migration: Geophysics, 62, 906-917.

Das könnte Ihnen auch gefallen