You are on page 1of 236

10 Structural Inversion

• Introduction • Subsalt Imaging in the North Sea • Estimation of the Overburden Model • Estimation of
the Substratum Model • Model Verification • Subsalt Imaging in the Gulf of Mexico • Layered Earth Model
Estimation • Structure-Independent Model Estimation • Imaging Beneath Irregular Water Bottom in the
Northwest Shelf of Australia • Earth Modeling and Imaging in Depth • Imaging Beneath Volcanics in the
West of Shetlands of the Atlantic Margin • Earth Modeling and Imaging in Depth • Imaging Beneath
Shallow Gas Anomalies in the Gulf of Thailand • Earth Modeling and Imaging in Depth • 3-D Structural
Inversion Applied to Seismic Data from the Southern North Sea • Estimation of the Overburden Model
• Model Representation by Tessellation • 3-D Coherency Inversion • 3-D Poststack Depth Migration • Estimation
of the Substratum Model • 3-D Structural Inversion Applied to Seismic Data from the Central North
Sea • 3-D Coherency Inversion Combined with 3-D Poststack Depth Migration • 3-D Stacking Velocity Inversion
Combined with 3-D Image-Ray Depth Conversion • 3-D Structural Inversion Applied to Seismic Data from
Offshore Indonesia • Model Building • Model Updating • Imaging in Depth • Volume-Based Interpretation •
3-D Structural Inversion Applied to Seismic Data from the Northeast China • 3-D DMO Processing •
3-D Prestack Time Migration • From rms to Interval Velocities • Structural Inversion • Structural and Stratigraphic
Interpretation • Exercises • Appendix K: Seismic Modeling • Zero-Offset Traveltime Modeling • Zero-Offset
Wavefield Modeling • Nonzero-Offset Wavefield Modeling • Elastic Wavefield Modeling • References

10.0 INTRODUCTION

A recorded seismic wavefield represented by a shot eltimes. In Chapter 8, we discussed prestack Kirchhoff
gather has two components — traveltimes and ampli- migration, which again, essentially is based on comput-
tudes. Direct inversion of a seismic wavefield to esti- ing diffraction traveltimes. Traveltime inversion thus
mate elastic parameters of the earth demands numer- yields a structural model of the earth represented by a
ically intensive computations. Instead, most practical set of layer velocities and reflector geometries, which
methods of inversion are applied to seismic traveltimes can then be used to derive a structural image of the
and amplitudes, separately. In Chapter 9, we discussed earth by depth migration. The term structural inversion
methods of layer velocity estimation — Dix conver- may be appropriately used to describe the process of
sion, stacking velocity inversion, and coherency inver- structural modeling and imaging by way of inversion of
sion, which essentially are based on inversion of trav- traveltimes.
1558 Seismic Data Analysis

In Chapter 11, we shall discuss poststack ampli- growth faults, often enveloped by major listric faults
tude inversion to estimate an acoustic impedance model and sometimes accompanied by counter-regional faults.
of the earth and prestack amplitude inversion to derive As the sand-shale deposition of Miocene and subsequent
the amplitude variation with offset (AVO) attributes. ages increased the overburden pressure, some diapirs
Amplitude inversion thus yields a stratigraphic model took dike-like forms and some formed overhangs. Ad-
of the earth represented by a combination of acoustic ditionally, some of the salt diapirs were squeezed along
impedance and AVO attribute changes within the lay- the fault planes and some even were moved laterally as
ers themselves. The term stratigraphic inversion may be far up as the water bottom, eventually detaching them-
appropriately used to describe the processes of estimat- selves from the source salt layer situated within the
ing the acoustic impedance and AVO attributes by way deeper strata and forming tabular salt bodies. Some
of inversion of amplitudes. of these tabular bodies joined together to form salt
In this chapter, we shall discuss case studies in canopies. The subsequent deposition on top then de-
structural inversion of 2-D and 3-D seismic data. These formed the shape of the salt canopies, giving rise to a
case studies relate to structural complexities caused by rugose geometry along the top and the base. The target
zones are, for some cases, immediately below the base-
(a) extensional tectonism as in the cases of salt diapirs salt, and for some other cases, somewhat deeper, they
of the North Sea and the Gulf of Mexico, are below an overpressured zone.
(b) compressional tectonism as in the cases of over- The third 2-D case study deals with irregular
thrust belts of the Middle East and Rocky Moun- water-bottom topography associated with a reef body.
tains, and The reef causes severe distortions of reflection travel-
(c) wrench tectonism as in the cases of the pull-apart times associated with underlying target horizons. The
basins of offshore Indonesia and Venezuela. fourth 2-D case study deals with imaging beneath a vol-
canic layer. Finally, the fifth 2-D case study is from the
Results from inversion of a 2-D seismic data set Gulf of Thailand. Shallow gas anomalies cause difficulty
must be evaluated within the bounds of 2-D imaging, in imaging multileveled reservoirs along faults that are
and 3-D effects must always be kept in mind (Section abundant in the area.
8.0). A structural inversion project can be a futile ex- The first 3-D case study is from the Southern Gas
ercise if the 2-D data set has not been recorded along a Basin of the North Sea. The objective is to delineate
dominant dip direction with minimal 3-D effects. the geometry of the base-Zechstein layer which forms
The first 2-D case study is from the Southern Gas a three-dimensionally complex diapiric structure. The
Basin of the North Sea. The North Sea Basin has second 3-D case study is from the Central North Sea.
been subjected to extensional tectonics, primarily in We demonstrate the use of the combination of the in-
the northeast-southwest direction with some rotational version methods listed in Table 9-1 to estimate layer
component. As a result, salt diapiric structures were velocities and delineate reflector geometries within the
formed and the overlying strata were subjected to fault- overburden above the Zechstein formation. The third 3-
ing. Continuing extensional tectonics and salt move- D case study is from offshore Indonesia. The objective
ments caused further faulting of the overlying strata, is to image the complex fault blocks caused by pull-
thus forming collapsed structures especially above the apart tectonism. In this case study, we also demonstrate
apexes of the salt diapirs. Common structural targets structural and stratigraphic interpretation of the im-
in the North Sea are Permian sands of the Rotliegen- age volume derived from 3-D prestack depth migration.
des and Carboniferous substrata below the Zechstein di- Finally, the fourth 3-D case study is from Northeast
apiric formation. We want to obtain an accurate depth China. With this case study, we demonstrate a complete
structure map of the top Rotliegendes formation in the time-and-depth sequence for processing, inversion, and
survey area. This requires an estimate of the velocity- interpretation that involves both 3-D prestack time and
depth model above the Zechstein diapiric formation and depth migrations.
removal of its deleterious effect on the underlying Per-
mian sands of Rotliegendes and deeper targets.
The second 2-D case study is from the Gulf of Mex-
ico. The basin in the Gulf of Mexico has been subjected 10.1 SUBSALT IMAGING IN
to extensional tectonics, primarily in the north-south di- THE NORTH SEA
rection. As a result, the Jurassic salt loaded by the over-
burden sand-shale sequence began to be deformed first The first 2-D case study for structural inversion is from
in the edges of the basin, forming diapiric structures. the Southern Gas Basin of the North Sea. Figures 10.1-
The extensional tectonism also gave rise to a series of 1 and 10.1-2 show selected CMP gathers from the two
Structural Inversion 1559

FIG. 10.1-1. The Southern Gas Basin line: (a) selected CMP gathers; (b) CMP stack from the left-half of a 2-D line from
the North Sea. Events A through D are segments associated with the base Zechstein. (Data courtesy Mobil North Sea Ltd.)
1560 Seismic Data Analysis

FIG. 10.1-2. The Southern Gas Basin line: (a) selected CMP gathers; (b) CMP stack from the right-half of a 2-D line from
the North Sea. Events E through K are segments associated with the base Zechstein.
Structural Inversion 1561

FIG. 10.1-3. The Southern Gas Basin line: time horizons intepreted from the unmigrated CMP-stacked section. See text for
details.

segments of the line accompanied by the corresponding sists of distinct layer boundaries with significant veloc-
CMP stacked sections. The left segment (Figure 10.1- ity contrast. Also, there exist vertical velocity gradients
1) contains a salt diapir with relatively steep flanks and in layers above the salt formation. A typical velocity-
the right segment (Figure 10.1-2) contains a salt di- depth model for the Southern Gas Basin may be con-
apir which has a broader base. The Zechstein forma- sidered in two parts:
tion comprises a halite unit with a velocity of 4400
m/s and anhydrite-dolomite rafts of various sizes and (a) An overburden above the salt layer with some fault-
shapes with a velocity of 5900 m/s. The strong velocity ing where time migration is applicable, and ray-
contrast across the top-salt boundary and the presence paths associated with nonzero-offset traveltimes
of the anhydrite-dolomite rafts within the diapiric for- yield a moveout behavior that makes it possible
mation itself give rise to raypath distortions and thus to estimate layer velocities with sufficient accuracy
consitute a complex overburden structure. The effect of using Dix conversion, stacking velocity inversion or
the complex overburden on the underlying target zone coherency inversion, and
— base Zechstein and the underlying Carboniferous se- (b) a substratum that includes the salt layer and the
quence, is evident on the CMP gathers and the CMP layers beneath, where depth migration is impera-
stacks. Specifically, note events with complex moveout tive.
on CMP gathers 481-881 below 1.5 s (Figure 10.1-1)
and the traveltime distortions along the base-Zechstein The boundary between the overburden and the sub-
reflection on the CMP stack (events A, B, C, and D in stratum is defined by the top of the salt layer where
Figures 10.1-1, and events E, F, G, H, and K in Figure the most severe ray bending takes place. To estimate a
10.1-2). velocity-depth model for a typical Southern Gas Basin
The subsurface geology in the Southern Gas Basin subsalt target, the following procedure composed from
can be represented by a layered earth model that con- the list of inversion methods in Table 9-1 can be used:
1562 Seismic Data Analysis

(a) Coherency inversion to estimate layer velocities and deeper horizon. Spurious peaks on semblance
and 2-D poststack depth migration to delineate re- curves and rapid lateral variations in velocity
flector geometries within the overburden, and should be avoided. Accordingly, smoothing is ap-
(b) constant half-space velocity analysis of image gath- plied to the velocity profile as much as geologically
ers from prestack depth migration to estimate the plausable, but not excessively so as to retain lateral
substratum velocity and stacking of image gath- velocity variations that are realistic.
ers to delineate the reflector geometry of the base- (b) Create a gridded velocity-depth model that con-
Zechstein (top-Rotliegendes) target horizon. sists of two parts — the known part on top, with
the n − 1 layers already established, and the un-
This procedure is applied layer-by-layer starting from known part underneath, defined as a half-space
the surface to resolve layer velocity and reflector geom- with its velocity equal to the velocity of the nth
etry of one layer before moving onto the next (Section layer derived in step (a).
9.4). Such an approach enables us to establish the ac- (c) Perform poststack depth migration using the grid-
curacy of the model one layer at a time and minimizes ded velocity-depth model from step (b) down to a
accumulation of errors as we proceed down to the target depth just below the layer of interest.
zone. (d) Interpret the depth horizon associated with the
base of the layer under consideration from the
depth-migrated section.
Estimation of the Overburden Model (e) To verify the accuracy of velocity estimation from
coherency inversion and, if needed, to update the
Start the analysis by interpreting the time horizons layer velocity (Section 9.5) derived in step (a),
from the unmigrated CMP stack that correspond to perform prestack depth migration to create im-
layer boundaries with significant velocity contrast. age gathers at some interval along the line. The
These horizons are denoted in Figure 10.1-3 — wa- velocity-depth model used for prestack depth mi-
ter bottom (H1), base Miocene Unconformity (H2), gration is the same as that used for poststack depth
base Upper Tertiary (H3), base Lower Tertiary (H4), migration in step (d).
base Cretaceous Chalk (H5), base Upper Triassic (H6a),
base Lower Triassic (H6b), and base Zechstein (top Repeat steps (a) through (e) for all the layers within the
Rotliegendes) (H7). The time horizons are assumed to overburden down to top-Zechstein boundary (Horizon
be equivalent to zero-offset reflection times that are used 6b), and thus establish a velocity-depth model for the
in coherency inversion (Section 9.1). overburden (Figure 10.1-4).
Horizon H6b is the top-salt boundary, which sep-
arates the overburden and substratum parts of the
model. Although interpreted, horizon H7 — base-salt
Estimation of the Substratum Model
boundary, is not included in the analysis sequence for
modeling the overburden. Instead, it is dealt with as
part of modeling the substratum. We now want to estimate a velocity field for the sub-
The depth horizon associated with the water bot- stratum region — Zechstein and the underlying strata.
tom is obtained simply by normal-incidence depth con- Again, consider the velocity-depth model in two parts
version of the time horizon (H1 in Figure 10.1-3). Then, — the overburden, the known part, and the substratum,
the following sequence was applied to horizons H2-H6b, the unknown part, which is defined as a half-space.
one layer at a time, starting at the top. For the sake
of the discussion here, assume that the velocity-depth (a) Shown in Figure 10.1-4 are four different velocity-
model for the first n − 1 layers already have been deter- depth models with the same overburden but differ-
mined, and that the nth layer is under consideration. ent constant velocities assigned to the substratum.
These constant velocities span the range that corre-
(a) Perform coherency inversion along the horizon un- spond to the velocity variations within Zechstein.
der consideration, pick semblance maxima, and de- The anhydrite-dolomite rafts (5900 m/s) floating
rive an interval velocity profile as a function of the within the halite formation (4400 m/s) have little
midpoint location along the line. Available velocity thickness (less than 100 m), but can influence the
gradient information is incorporated into the inter- velocity of the Zechstein unit as a whole depending
val velocity estimation. As demonstrated in Section on the spatial distribution of the rafts.
9.1, the accuracy in velocity estimation degrades (b) Perform prestack depth migration using the four
with shorter effective cable length, faster velocity earth models in Figure 10.1-4, and generate a set
Structural Inversion 1563

of image gathers along the line (Figures 10.1-5 and (h) Convert the depth-migrated sections from depth to
10.1-6). Examine the image gathers from the four time using the final velocity-depth model from step
velocity panels at the same location and note that (f) to apply poststack spiking deconvolution, band-
they all represent the same image associated with pass filtering and AGC scaling. Finally, convert the
the overburden down to a depth that corresponds sections back to depth. It usually helps improve the
the the top-salt boundary. Nevertheless, the event vertical resolution of depth-migrated data to apply
associated with the base Zechstein exhibits dif- poststack deconvolution. This also is needed for a
ferent moveout characteristics depending on the fair comparison of images from prestack depth mi-
half-space velocity. For instance, at midpoint lo- gration and poststack depth migration, since the
cation 321, flatness for the base-Zechstein event latter normally would be performed using stack
at a depth of 3500 m is achieved with the 4700- with poststack processing applied.
m/s substratum velocity (event A in Figure 10.1-
5). Similarly, at midpoint location 1681, flatness Selected image gathers and the depth images using
for the base-Zechstein event at a depth of 3950 m the final velocity-depth model (Figure 10.1-9) from the
is achieved, again, with the 4700-m/s substratum two segments of the line are shown in Figures 10.1-10
velocity (event B in Figure 10.1-6). In principle, one and 10.1-11. Compare the depth image obtained from
may be able to determine velocity nodes from these prestack depth migration with the images obtained from
image gathers based on the flatness of the event poststack depth and time migrations (Figures 10.1-12
that corresponds to the base-Zechstein boundary. and 10.1-13). We should not expect significant differ-
It is not just flatness, but event strength and con- ences between the three images within the overburden
tinuity across the offset axis of the image gather, region where time migration often yields acceptable re-
that we take into consideration when picking ve- sults. Note, however, the differences in the substratum
locity nodes. If the event is weak, it could mean region. Specifically, the base-Zechstein horizon (event A
that the velocity assigned to the layer above is er- in Figure 10.1-10b and events E and F in Figure 10.1-
roneously too low or too high, causing migration 11b) cannot be delineated from poststack depth migra-
errors. To gain confidence in the velocity determi- tion, nor can it be delineated accurately from poststack
nation, the image gathers are used in combination time migration (events G and H in Figure 10.1-12b,
with the image-gather stacks as in the next step. and event K in Figure 10.1-13). The anhydrite-dolomite
(c) Stack the image gathers as from step (b) to gen- rafts (such as event D in Figure 10.1-11b) and the 3-D
erate the depth images (Figures 10.1-7 and 10.1-8) behavior of the diapiric structure limit the accuracy of
associated with the earth models in Figure 10.1-4. the final image obtained from prestack depth migration.
The stack power associated with the base-Zechstein Specifically, note the poor image of the base Zechstein
event can be used as an additional criterion in com- represented by events B and C in Figure 10.1-10b.
bination with the flatness criterion for image gath- Image rays associated with the base-Zechstein
ers (Figures 10.1-5 and 10.1-6) to derive the veloc- boundary (Figure 10.1-14), however, clearly indicate the
ity profile for the Zechstein formation. need for prestack depth migration for accurate imag-
(d) Assign the velocity field for the substratum derived ing of the substratum region. Note the significant lat-
from step (c) to the half-space below the overbur- eral shifts on the image rays, especially beneath the
den. Then, combine this velocity field for the half- salt diapirs — demonstrative evidence of the presence
space with the overburden model to construct a of strong lateral velocity variations associated with a
new velocity-depth model. complex overburden.
(e) Next, perform prestack depth migration using the
new velocity-depth model from step (d).
(f) Interpret the depth image from step (e) to delineate
the base-Zechstein horizon. Subsequently incorpo- Model Verification
rate this horizon into the velocity depth-model, and
assign a velocity of 5400 m/s to the subsalt region The final stage in earth modeling and imaging is the
based on well data to construct a final velocity- verification of the accuracy of the model itself (Section
depth model (Figure 10.1-9). 9.5). For an earth model in depth to be acceptable, it
(g) Perform prestack depth migration once more us- has to pass the following two tests:
ing the final velocity-depth model from step (f) to
obtain the image gathers and their stacks which (a) Image gathers from prestack depth migration using
represent the final image in depth. the earth model in question must exhibit flat events
(text continues on p. 1574)
1564 Seismic Data Analysis

FIG. 10.1-4. The Southern Gas Basin line: four different velocity-depth models with the same overburden down to the
top-Zechstein boundary, but with different constant velocities assigned to the half-space below — 4100, 4400, 4700, and 5000
m/s.
Structural Inversion 1565

FIG. 10.1-5. The Southern Gas Basin line: selected image gathers from the left-half of the line as in Figure 10.1-1 from
prestack depth migration using the four different earth models in Figure 10.1-4. See text for details.
1566 Seismic Data Analysis

FIG. 10.1-6. The Southern Gas Basin line: selected image gathers from the right-half of the line as in Figure 10.1-2 from
prestack depth migration using the four different earth models in Figure 10.1-4. See text for details.
Structural Inversion 1567

FIG. 10.1-7. The Southern Gas Basin line: depth images from prestack depth migration for the left-half of the line as in
Figure 10.1-1 using the four different earth models in Figure 10.1-4. See text for details.
1568 Seismic Data Analysis

FIG. 10.1-8. The Southern Gas Basin line: depth images from prestack depth migration for the right-half of the line as in
Figure 10.1-1 using the four different earth models in Figure 10.1-4. See text for details.
Structural Inversion 1569

FIG. 10.1-9. The Southern Gas Basin line: final velocity-depth model.

FIG. 10.1-10. The Southern Gas Basin line: (a) selected image gathers from the left-half of the line as in Figure 10.1-1 from
prestack depth migration using the final velocity-depth model in Figure 10.1-9; (b) stack of the image gathers — depth image
from prestack depth migration for the left-half of the line as in Figure 10.1-1. See text for the labeled events.
1570 Seismic Data Analysis

FIG. 10.1-11. The Southern Gas Basin line: (a) selected image gathers from the right-half of the line as in Figure 10.1-1
from prestack depth migration using the final velocity-depth model in Figure 10.1-9; (b) stack of the image gathers — depth
image from prestack depth migration for the right-half of the line as in Figure 10.1-1. See text for the labeled events.
Structural Inversion 1571

FIG. 10.1-12. The Southern Gas Basin line: (a) depth image from poststack depth migration using the same velocity-depth
model in Figure 10.1-9 as for the depth image from prestack depth migration shown in Figure 10.1-10b; (b) time image from
poststack time migration. These images are from the left-half of the line as in Figure 10.1-1. See text for the labeled events.
1572 Seismic Data Analysis

FIG. 10.1-13. The Southern Gas Basin line: (a) depth image from poststack depth migration using the same velocity-depth
model in Figure 10.1-9 as for the depth image from prestack depth migration shown in Figure 10.1-11b; (b) time image from
poststack time migration. These images are from the right-half of the line as in Figure 10.1-1. See text for the labeled events.
Structural Inversion 1573
1574 Seismic Data Analysis

(Figures 10.1-10a and 10.1-11a). Events associated 10.2-2). Nevertheless, accurate imaging of the base-salt
with multiples and converted waves (Section 11.6) boundary and the subsalt region is only possible by way
are not expected to be flat. Nevertheless, even with of prestack depth migration.
good models, usually, there also are some primary Aside from the water layer, a Gulf of Mexico
events that do not meet the flatness criterion. Vi- velocity-depth model is typically represented in two
olation of this criterion may occur because of er- parts:
roneous layer velocities or 3-D effects that are not
accounted for by 2-D modeling and imaging. Other (a) A background velocity field with vertical velocity
sources of departures from flatness include strong variations characterized by gentle variations in the
lateral velocity variations that are much less than gradient, the absence of distinct layer boundaries,
a cable length and effect of anisotropy on layer ve- and mild-to-moderate lateral velocity variations.
locities. (b) Tabular and diapiric salt bodies with different
(b) For an earth model in depth to be an acceptable shapes, but with a constant velocity of 4450 m/s,
representation of the subsurface geological model, embedded into the background velocity field.
it must be consistent with the seismic data used to
estimate the model in question. To check for con- To estimate a velocity-depth model for a Gulf of Mex-
sistency, perform ray-theoretical modeling of zero- ico structural target below the tabular salt bodies, the
offset traveltimes associated with the layer bound- following procedure composed from the list of inversion
aries included in the model (Figure 10.1-14). Then, methods in Table 9-1 is used:
superimpose the modeled traveltimes on the CMP-
stacked section and observe any discrepancy be- (a) Dix conversion of stacking velocities to estimate the
tween the modeled and the actual traveltimes (Fig- background velocity field,
ure 10.1-15). Here, we are assuming that the re- (b) Model updating and verification of the velocity
flection traveltimes observed on a CMP-stacked field within the suprasalt region (Section 9.5),
section can be closely approximated by two-way (c) Poststack or prestack depth migration to delineate
zero-offset traveltimes. Actual traveltimes inter- the top-salt boundary,
preted from the unmigrated CMP-stacked section (d) Assignment of the salt velocity into the half-space
are shown in Figure 10.1-3. below the top-salt boundary,
(e) Prestack depth migration to delineate the base-salt
Aside from these two criteria, the estimated earth boundary,
model needs to be validated by examining it for consis- (f) Assignment of the the background velocity into the
tency with the structural model applicable to the area half-space below the base-salt boundary (the sub-
of interest. For instance, the faulting and folding im- salt region),
plied by the the model must have the same patterns as (g) Prestack depth migration to obtain and verify the
in the true subsurface situation. Calibration to well tops final earth image in depth.
is also part of the model validation procedure (Section
9.4). In this case study, we shall first test the procedure used
for the Southern Gas Basin line to estimate a layered
earth model in depth for comparison with the procedure
10.2 SUBSALT IMAGING IN outlined above.
THE GULF OF MEXICO

The second 2-D case study for structural inversion is Layered Earth Model Estimation
from the Gulf of Mexico. Figure 10.2-1 shows the DMO-
stacked section and Figure 10.2-2 shows the poststack Start the North-Sea style analysis (Section 10.1) by in-
time-migrated section of a 2-D data set from the Gulf terpreting a set of time horizons from the time-migrated
of Mexico. The high-amplitude event (such as event A CMP stack (Figure 10.2-3a). These time horizons are
in Figure 10.2-1) with complex traveltime is the top- then used to obtain the unmigrated time horizons (Fig-
salt reflection. Conflicting dips associated with the fault ure 10.2-3b) by way of a ray-theoretical forward mod-
blocks within the overburden and the rugose top-salt eling scheme, sometimes called demigration. Horizon A
boundary are preserved by way of DMO correction (Fig- in Figure 10.2-3a represents the top-salt boundary. The
ure 10.2-1), and accurate imaging of the suprasalt region modeled time horizons are assumed to be equivalent to
can be achieved by poststack time migration (Figure zero-offset reflection times, which are used in coherency
Structural Inversion 1575

FIG. 10.2-1. The Gulf of Mexico line: unmigrated DMO-stacked section. (Data courtesy Schlumberger Geco-Prakla.)

FIG. 10.2-2. The Gulf of Mexico line: poststack time-migrated DMO-stacked section.
1576 Seismic Data Analysis

FIG. 10.2-3. The Gulf of Mexico line: (a) time horizons interpreted from the time-migrated section as in Figure 10.2-2; (b)
modeled zero-offset traveltimes associated with the time horizons in (a) superimposed on the stacked section as in Figure
10.2-1.
Structural Inversion 1577

inversion. Alternatively, time horizons could have been with moveout; they often are associated with shale in-
interpreted from the unmigrated stacked section, di- trusions. These events could also be associated with
rectly. multiples or converted waves (Section 11.6).
To estimate a layered earth model, the following As part of a verification procedure, the estimated
procedure was implemented: earth model (Figure 10.2-4) needs to be tested for con-
sistency with the input seismic data. Figure 10.2-8a
(a) Starting from the top, for each layer within the shows the modeled zero-offset traveltimes superimposed
suprasalt region, perform coherency inversion to es- on the unmigrated stacked section as in Figure 10.2-1.
timate the layer velocity, and These traveltimes are associated with the layer bound-
(b) 2-D poststack depth migration to delineate the re- aries included in the velocity-depth model of Figure
flector geometry associated with the layer bound- 10.2-4. They should be compared with the traveltimes
ary. derived from the forward modeling of the time horizons
(c) Repeat steps (a) and (b) for all the layers picked from the time-migrated section (Figure 10.2-8b).
within the suprasalt region, and thus establish the Note that the two sets of traveltimes are fairly consis-
velocity-depth model for that region. tent within the suprasalt region. However, some discrep-
(d) Assign the salt velocity into the half-space below ancy exists for the top-salt event (red horizon) and the
the top-salt boundary, and base-salt event (yellow horizon).
(e) perform 2-D prestack depth migration to delineate Compare the depth images derived from prestack
the base-salt boundary. depth migration (Figure 10.2-6) and poststack depth
(f) Then, assign a vertically varying velocity function migration (Figure 10.2-9) using the same velocity-depth
into the subsalt region (sometimes referred to as model (Figure 10.2-4). While the images within the
salt flooding), and suprasalt region are comparable, prestack depth migra-
(g) perform 2-D prestack depth migration to obtain tion certainly yields a superior image of the base-salt
and verify the final earth image in depth. boundary and the subsalt region.
(h) Convert the image-gather stack from step (g) from
depth to time using the velocity-depth model from
step (d), apply poststack spiking deconvolution,
Structure-Independent Model Estimation
band-pass filtering, and AGC scaling, then convert
back to depth.
We shall now implement the procedure outlined at the
beginning of this section.
Figure 10.2-4 shows the velocity-depth model based
on the procedure outlined above. Note that there are
three detached salt bodies. Also note the lateral veloc- (a) Start with the stacking velocity field (Figure 10.2-
ity variations, detected by coherency inversion, within 10). Note that already a pattern is inferred about
each layer in the suprasalt region. The velocity field in the velocity-depth model — an upper region where
the subsalt region is referenced to the water bottom, the velocity field shows a consistent increase with
since the salt bodies do not have much influence on the depth and moderate lateral variations, and a lower
vertical variations in the background velocity field. To region where anomalous zones exist. Actually, the
appreciate the raypath distortions caused by the salt upper region coincides with the suprasalt region
bodies, examine the image rays down to the base-salt and the lower region coincides with the salt sills
boundary in Figure 10.2-5. The image rays clearly indi- and the subsalt region.
cate the need for prestack depth migration for accurate (b) Apply smoothing to the stacking velocity field in
imaging of the subsalt region. Figure 10.2-10 to derive what may be considered
Figure 10.2-6 shows the depth image from prestack an rms velocity field (Figure 10.2-11) for Dix con-
depth migration using the velocity-depth model in Fig- version.
ure 10.2-4. Note the rugose top-salt boundary, and rel- (c) Perform Dix conversion to derive an initial velocity-
atively smoother base-salt boundary. The accompany- depth model in the form of an interval velocity
ing image gathers are shown in Figure 10.2-7. Note the field (Figure 10.2-12). The upper half of this model
high-amplitude events (A and B) associated with the appears geologically plausable — a velocity field
top- and base-salt boundaries. Flatness of events both typical of a sedimentary sequence without any dis-
in the suprasalt and subsalt regions provides the evi- tinct layer boundaries with significant velocity con-
dence that the model in Figure 10.2-4 is geologically trast. The lower half exhibits anomalous behavior
acceptable. Note the presence of intrasalt events (C) that geologically is not meaningful. It is this re-
(text continues on p. 1583)
1578 Seismic Data Analysis

FIG. 10.2-4. The Gulf of Mexico line: layered earth model in depth estimated by a combination of coherency inversion
to estimate layer velocities and poststack depth migration to delineate reflector geometries within the suprasalt region, and
prestack depth migration to delienate the base-salt boundary.

FIG. 10.2-5. The Gulf of Mexico line: image rays down to the base-salt boundary.
Structural Inversion 1579
1580 Seismic Data Analysis

FIG. 10.2-7. Part 1: The Gulf of Mexico line: selected im- FIG. 10.2-7. Part 2: The Gulf of Mexico line: selected im-
age gathers from prestack depth migration using the model age gathers from prestack depth migration using the model
shown in Figure 10.2-4. The depth image derived from stack- shown in Figure 10.2-4. The depth image derived from stack-
ing of the image gathers is shown in Figure 10.2-6. ing of the image gathers is shown in Figure 10.2-6. A and
B are events associated with top- and base-salt boundaries,
and C is an event within the salt body.
Structural Inversion 1581
1582 Seismic Data Analysis

FIG. 10.2-8. The Gulf of Mexico line: (a) modeled zero-offset traveltimes associated with the layer boundaries included in
the earth model of Figure 10.2-4. (b) the zero-offset traveltimes of (a) superimposed on the zero-offset traveltimes as in Figure
10.2-3b. Both sets of traveltimes are superimposed on the stacked section as in Figure 10.2-1.
Structural Inversion 1583

FIG. 10.2-9. The Gulf of Mexico line: depth image from poststack depth migration using the model shown in Figure 10.2-4.

gion where Dix conversion would fail because of the interval velocity field (Figure 10.2-12) used to
the presence of salt sills with rugose boundary that create the image gathers themselves.
causes severe raypath distortions. (g) Now perform conventional velocity analysis using
(d) Using the initial velocity-depth model from step the selected image gathers in time (Figure 10.2-
(c) (Figure 10.2-12), perform prestack depth mi- 14) and create vertical rms velocity functions at all
gration and generate image gathers (Figure 10.2- analysis locations along the line.
13). Flatness of events indicates that the initial (h) Create a new rms velocity field from the verti-
velocity-depth model (Figure 10.2-12) based on Dix cal functions as shown in Figure 10.2-15. Compare
conversion of rms velocities is remarkably accurate with the initial rms velocity field in Figure 10.2-
within the suprasalt region (above event A which 11 and note the details introduced to the lateral
corresponds to the top-salt boundary). variations in the upper region.
(e) Nevertheless, small adjustments may be required (i) Using the updated rms velocity field from step
to the velocity-depth model within the suprasalt (h), perform Dix conversion and create an updated
region before moving down to the salt and subsalt velocity-depth model (Figure 10.2-16). Note that
regions. To update the model in the suprasalt re- the upper region now closely resembles the depo-
gion, we shall perform residual moveout analysis of sitional sequence in the suprasalt region as seen in
image gathers. Specifically, a variation of the proce- the time-migrated stacked section (Figure 10.2-2).
dure described in Section 9.5 will be followed here. Also note that the top-salt boundary is now more
First, convert the image gathers at selected loca- evident in this model. The lower region still has to
tions along the line from depth (Figure 10.2-13) be considered as geologically implausable.
to time domain (Figure 10.2-14) using the initial (j) Using the updated velocity-depth model (Figure
velocity-depth model (Figure 10.2-12). The image 10.2-16), perform prestack depth migration. Se-
gathers now are equivalent to moveout-corrected lected image gathers are shown in Figure 10.2-17
CMP gathers, except that they are in their mi- and the depth image derived from stacking of the
grated positions. image gathers is shown in Figure 10.2-18. From the
(f) The next step is the application of inverse moveout flatness of events on image gathers, we may con-
correction to the image gathers from step (e) as vince ourselves that the model in Figure 10.2-16
shown in Figure 10.2-14. The velocity field used for can now be considered as final for the suprasalt
inverse moveout correction is the rms velocity field region. Any further iteration of steps (e) through
shown in Figure 10.2-11, which is consistent with (j) would only yield insignificant refinement to the
(text continues on p. 1596)
1584 Seismic Data Analysis

FIG. 10.2-10. The Gulf of Mexico line: stacking velocity field.

FIG. 10.2-11. The Gulf of Mexico line: rms velocity field derived from the stacking velocity field in Figure 10.2-10.
Structural Inversion 1585

FIG. 10.2-12. The Gulf of Mexico line: an initial velocity-


depth model based on Dix conversion of the rms velocity
field in Figure 10.2-11.

FIG. 10.2-13. Part 1: The Gulf of Mexico line: selected im-


age gathers from prestack depth migration using the initial
velocity-depth model shown in Figure 10.2-12.
1586 Seismic Data Analysis

FIG. 10.2-13. Part 3: The Gulf of Mexico line: selected im-


age gathers from prestack depth migration using the initial
velocity-depth model shown in Figure 10.2-12.
FIG. 10.2-13. Part 2: The Gulf of Mexico line: selected im-
age gathers from prestack depth migration using the initial
velocity-depth model shown in Figure 10.2-12.
Structural Inversion 1587

FIG. 10.2-14. The Gulf of Mexico line: (top) selected image gathers from prestack depth migration after converting from
depth (Figure 10.2-13) to time domain using the initial velocity-depth model shown in Figure 10.2-12; (bottom) after applying
inverse moveout correction using the rms velocity field shown in Figure 10.2-11.
1588 Seismic Data Analysis

FIG. 10.2-15. The Gulf of Mexico line: updated rms velocity field derived from the velocity analysis of the image gathers in
Figure 10.2-14.

FIG. 10.2-16. The Gulf of Mexico line: updated velocity-depth model based on the Dix conversion of the updated rms
velocity field shown in Figure 10.2-15.
Structural Inversion 1589
1590 Seismic Data Analysis

FIG. 10.2-18. The Gulf of Mexico line: depth image from prestack depth migration using the updated velocity-depth model
shown in Figure 10.2-16. Image gathers are shown in Figure 10.2-17.

FIG. 10.2-19. The Gulf of Mexico line: updated velocity-depth model as in Figure 10.2-16 with the top-salt boundary
interpreted from the depth image in Figure 10.2-18.
Structural Inversion 1591

FIG. 10.2-20. The Gulf of Mexico line: new velocity-depth model as in Figure 10.2-19 with the salt velocity introduced to
the region below the top-salt boundary.

FIG. 10.2-21. The Gulf of Mexico line: depth image from prestack depth migration using the new velocity-depth model
shown in Figure 10.2-20.
1592 Seismic Data Analysis

FIG. 10.2-22. The Gulf of Mexico line: new velocity-depth model as in Figure 10.2-20 with the base-salt boundary interpreted
from the depth image in Figure 10.2-21.

FIG. 10.2-23. The Gulf of Mexico line: final velocity-depth model.


Structural Inversion 1593
1594 Seismic Data Analysis

FIG. 10.2-25. Part 1: The Gulf of Mexico line: selected FIG. 10.2-25. Part 2: The Gulf of Mexico line: selected
image gathers from prestack depth migration using the final image gathers from prestack depth migration using the final
velocity-depth model shown in Figure 10.2-23. The depth velocity-depth model shown in Figure 10.2-23. The depth
image is shown in Figure 10.2-24. image is shown in Figure 10.2-24.
Structural Inversion 1595

FIG. 10.2-25. Part 3: The Gulf of Mexico line: selected image gathers from prestack depth migration using the final velocity-
depth model shown in Figure 10.2-23. The depth image is shown in Figure 10.2-24.
1596 Seismic Data Analysis

FIG. 10.2-26. A 3-D perspective view of detached salt sills from the Gulf of Mexico. The silver surface represents the top-salt
boundary and the gold surface represents the base-salt boundary. (Courtesy Schlumberger Geco-Prakla.)

model. The model updating described in steps (e) (n) Interpret the base-salt boundary from the depth
through (j) is only valid if the velocity errors in the image (Figure 10.2-21) and insert it as a layer
initial model are fairly small (Section 9.5). boundary into the velocity-depth model (Figure
(k) Interpret the top-salt boundary from the depth im- 10.2-22).
age (Figure 10.2-18) and insert it as a layer bound- (o) Finally, introduce the background velocity field
ary into the velocity-depth model (Figure 10.2-19). into the subsalt region in the form of a vertically
Often, it is adequate to use the depth image from varying velocity function (Figure 10.2-23). As for
poststack depth migration, rather than prestack the layered earth model in Figure 10.2-4, the veloc-
depth migration as in this case, to interpret the ity field in the subsalt region is referenced to the
top-salt boundary. water bottom, since the salt bodies have very lit-
(l) Assign the salt velocity into the region below the tle influence on the vertical variations in the back-
top-salt boundary (Figure 10.2-20). Assume that ground velocity field.
the velocity within the salt bodies is constant (4450 (p) Using the final velocity-depth model (Figure 10.2-
m/s). This assumption in some cases may not be 23), perform prestack depth migration.
valid — shale intrusions into halite crystalline rock (q) Convert the image-gather stack from step (p) from
can alter the velocity within the salt sills, substan- depth to time using the final velocity-depth model,
tially. apply poststack spiking deconvolution, band-pass
(m) Using the new velocity-depth model (Figure 10.2- filtering, and AGC scaling, then convert back to
20), perform prestack depth migration to get a new depth. Figures 10.2-24 and 10.2-25 show the depth
depth image (Figure 10.2-21). image and selected image gathers, respectively.
Structural Inversion 1597

Examine the image gathers in Figure 10.2-25 associated procedure. Note that the reef has a low-velocity thin
with the final velocity-depth model (Figure 10.2-23) and cap and a high-velocity interior. The weight of the reef
observe that there still are some events that do not meet mass may have caused the sagging of the layer bound-
the flatness criterion. Violation of this criterion, in this aries (H2 and H3) associated with the shallow, uncon-
case, largely is due to 3-D effects that are not accounted solidated sediments.
for by 2-D earth modeling and imaging. In the Gulf of Figure 10.3-5 shows the depth image from prestack
Mexico, the salt sills can have complex shapes as shown depth migration using the velocity-depth model in Fig-
in Figure 10.2-26. Only by 3-D structural inversion of ure 10.3-4. Note that the pull-up effect below the reef
3-D data, we can hope to delineate such salt bodies and seen in the time-migrated section (Figure 10.3-3) has
image the subsalt regions. been removed. The shelf slope dipping up from left to
right is evident from the substratum reflector geome-
tries.
10.3 IMAGING BENEATH IRREGULAR Selected image gathers along the line shown in Fig-
WATER BOTTOM IN THE ure 10.3-6 contain flat events associated with primaries
NORTHWEST SHELF OF AUSTRALIA and events with large moveout associated with multi-
ples. No attempt was made in this case study to at-
The third 2-D case study for structural inversion is from tenuate multiples. Conventional CMP stacking attenu-
the Northwest Shelf of Australia. This case study deals ates multiples based on velocity discrimination between
with the deleterious effect of an irregular water-bottom primaries and multiples. The image gathers in Figure
topography, which is associated with a reef, on the ge- 10.3-6 can be considered similar to CMP gathers that
ometry of the underlying reflectors. Shown in Figure have been moveout corrected using primary velocities.
10.3-1 is the CMP-stacked section. The elevation differ- As a result, primaries are flattened and multiples are
ence between the top and base of the reef is nearly 500 undercorrected. Therefore, stacking of image gathers to
m, and the lateral extent of the reef is about 7.5 km. obtain a depth image has a bonus effect of attenuat-
Note the typical pull-up effect on the reflections below. ing multiples. However, multiples can make analysis of
The water-bottom reef causes anomalous behavior image gathers for residual moveout a difficult task.
in the stacking velocity field as seen in Figure 10.3-2. Flatness of the primary events implies the accu-
Time migration using a smoothed form of the stacking racy of the velocity-depth model (Figure 10.3-4) and the
velocity field produces a distorted image of the sub- accuracy of the depth image (Figure 10.3-5). Compare
surface (Figure 10.3-3). Note, for instance, the typical the depth images derived from prestack depth migration
overmigration character below the reef edges. (Figure 10.3-5) and poststack depth migration (Figure
10.3-7) using the same velocity-depth model (Figure
10.3-4). The overmigration effect at the reef edges is
quite evident on the poststack depth-migrated section.
Earth Modeling and Imaging in Depth
On the other hand, we remind ourselves the fundamen-
tal notion about earth modeling and imaging in depth
To correct for the effect of the irregular water-bottom from Section 9.0 — the depth image in Figure 10.3-
topography associated with the reef on the geometry of 5 can only be considered a plausable representation of
target reflectors within the substratum, we shall per- the subsurface geology, and it is not the only representa-
form earth modeling and imaging in depth. The proce- tion. Three-dimensional effects increase the ambiguity
dure for earth modeling includes coherency inversion in the earth model estimated from 2-D seismic data.
combined with poststack depth migration. Start the
analysis by interpreting a set of six time horizons (in-
cluding the water bottom) within the 0-2.5 s time win-
dow from the unmigrated CMP stack (Figure 10.3-1). 10.4 IMAGING BENEATH VOLCANICS
The time horizons are superimposed on the stacking ve- IN THE WEST OF SHETLANDS
locity field in Figure 10.3-2. These time horizons were OF THE ATLANTIC MARGIN
assumed to be equivalent to zero-offset reflection times
and, as such, were used in coherency inversion. The fourth 2-D case study for structural inversion is
The combination of coherency inversion to estimate from the West of Shetlands of the Atlantic Margin.
the layer velocities and poststack depth migration to de- This case study deals with earth modeling and imag-
lineate the reflector geometries was applied to the six ing beneath volcanics. Shown in Figure 10.4-1 is the
horizons, one layer at a time, starting at the top. Fig- CMP-stacked section in two parts. The stratigraphic
ure 10.3-4 shows the velocity-depth model based on this column comprises shallow depositional sequences that
(text continues on p. 1607)
1598 Seismic Data Analysis
Structural Inversion 1599
1600 Seismic Data Analysis
Structural Inversion 1601
1602 Seismic Data Analysis
Structural Inversion 1603

FIG. 10.3-6. The Offshore Australia line: selected image gathers associated with the depth image from prestack depth
migration shown in Figure 10.3-5.
1604 Seismic Data Analysis
Structural Inversion 1605

FIG. 10.4-1. The West of Shetlands line: CMP stack in two parts. Event A corresponds to a major unconformity and event
B corresponds to the top of the volcanic layer. (Data courtesy BP Exploration and Production.)
1606 Seismic Data Analysis

FIG. 10.4-2. The West of Shetlands line: poststack time-migrated CMP stack.
Structural Inversion 1607

are bounded at the base by a major unconformity (event (e) Sort the image gathers into velocity panels as
A). Below the unconformity, is a volcanic layer (event shown in Figures 10.4-4 through 10.4-8.
B) with varying thickness. The exploration objective is (f) At a specific midpoint location, examine the move-
the potential structural high below 2 s at midpoint lo- out of events for flatness. Identify the velocity for
cation 3440 and the faulted zone below 3.5 s between which there exists a group of flat events within a
midpoints 440-1240. Water-bottom and peg-leg multi- depth range. The velocity-depth picks are indicated
ples associated with the unconformity were largely at- by the left-arrows in Figures 10.4-4 through 10.4-8.
tenuated using the Radon transform technique (Section Note that for a velocity-depth pick to be valid, not
6.4) before prestack depth migration. Poststack time only the event at that depth must be flat but also
migration yields an image of the subsurface that falls all events above it must be flat. If we were to pick
short of being acceptable for interpretation within the an rms velocity function at an analysis location as
zone of interest below the unconformity (Figure 10.4- was demonstrated in Figures 5.4-16 through 5.4-19,
2). To obtain an improved image of this deeper zone, we for the present case, we would only need to create
shall perform earth modeling and imaging in depth. The one velocity panel. As an example, the rms veloc-
procedure for earth modeling includes coherency inver- ity picks are denoted by the right-arrows in Figure
sion combined with normal-incidence depth conversion 10.4-5.
applied within the shallow depositional sequences (the (g) Combine the picks for the velocity-depth pairs that
overburden), and prestack depth migration and image- meet the flatness criterion at each analysis location
gather analysis applied within the zone below the un- and create a layer velocity profile (labeled as VL1
conformity (the substratum). in Figure 10.4-3a) and a layer boundary.
(h) Insert the “velocity layer” (labeled as VL1 in Fig-
ure 10.4-3b) into the velocity-depth model and de-
Earth Modeling and Imaging in Depth fine a new half-space region below (Figure 10.4-3b).
(i) Then, repeat steps (d) through (h) for the remain-
Start the analysis by interpreting a set of six time hori- ing velocity layers — VL2, VL3, and VL4.
zons, including the water bottom and the unconformity (j) Perform prestack depth migration (Figure 10.4-9)
itself, within the overburden from the time-migrated using the final velocity-depth model shown in Fig-
CMP stack (Figure 10.4-2). Then, use these time hori- ure 10.4-3b.
zons to obtain the unmigrated time horizons by way
of a zero-offset traveltime modeling scheme based on Compare the results of from prestack depth migration
image-ray tracing. The modeled time horizons are as- (Figure 10.4-9) and poststack time migration (Figure
sumed to be equivalent to zero-offset reflection times, 10.4-2), and note that an event (labeled as C) within the
which are then used in coherency inversion. Except for substratum has been uncovered. This event is at a depth
the water bottom, the following sequence is applied to of 3 km at the left edge of the section and dips down
the five horizons, one layer at a time, starting at the to the right. Also note that the complex structure —
top. which may have a 3-D character — below 4 km between
midpoints 440-1240 is imaged within the accuracy of 2-
(a) Perform coherency inversion along the line to esti- D migration.
mate the layer velocities, and Selected image gathers along the line (Figure 10.4-
(b) Perform image-ray depth conversion of the time 10) largely contain flat events associated with primaries,
horizons to delineate the reflector geometries as- since multiples were attenuated by way of the Radon
sociated with the layer boundaries. transform prior to prestack depth migration. Flatness
(c) Repeat steps (a) and (b) for all five layers within of the primary events demonstrates the accuracy of the
the overburden down to the unconformity and cre- velocity-depth model (Figure 10.4-3) and the accuracy
ate a velocity-depth model for the overburden (the of the depth image (Figure 10.4-9). Note that a spatially
upper portion of the model that includes the top varying mute has been applied to image gathers (Figure
five layers in Figure 10.4-3b). 10.4-10) to attain an optimum depth image.
(d) Assign a range of constant velocities to the half- Figure 10.4-11 shows depth-to-time conversion of
space that corresponds to the substratum (the re- the depth image shown in Figure 10.4-9. The conversion
gion below the unconformity A in Figure 10.4-3b) was done using the same velocity field that was used for
while using the same overburden model (the region prestack depth migration (Figure 10.4-3). This section
down to the unconformity A in Figure 10.4-3b) and should be compared with the time image derived from
perform prestack depth migration to generate im- poststack time migration shown in Figure 10.4-2. Again,
age gathers at some interval along the line. note the event at 3 s (labeled as C) in Figure 10.4-11;
(text continues on p. 1620)
1608 Seismic Data Analysis

FIG. 10.4-3. The West of Shetlands line: (a) layer velocity profiles as functions of midpoint location for the velocity layers
— VL1, VL2, VL3, and VL4, within the substratum region below the unconformity labeled as horizon A in the velocity-
depth model shown in (b). Above the unconformity, the model was estimated using coherency inversion and image-ray depth
conversion; and below the unconformity, the model was estimated using prestack depth migration and moveout analysis of
image gathers (as in Figures 10.4-4 through 10.4-8).
Structural Inversion 1609

FIG. 10.4-4. The West of Shetlands line: prestack depth migration velocity panels at midpoint 241 used in estimating the
substratum model (the region below unconformity A in Figure 10.4-3). Shown here are the velocity panels associated with
four velocity layers within the substratum — VL1, VL2, VL3, and VL4. The left-arrows ← denote the interval velocity-depth
picks and the right-arrows → denote the rms velocity picks. See text for details.
1610 Seismic Data Analysis

FIG. 10.4-5. The West of Shetlands line: prestack depth migration velocity panels at midpoint 321 used in estimating the
substratum model (the region below unconformity A in Figure 10.4-3). Shown here are the velocity panels associated with
four velocity layers within the substratum — VL1, VL2, VL3, and VL4. The left-arrows ← denote the interval velocity-depth
picks. See text for details.
Structural Inversion 1611

FIG. 10.4-6. The West of Shetlands line: prestack depth migration velocity panels at midpoint 401 used in estimating the
substratum model (the region below unconformity A in Figure 10.4-3). Shown here are the velocity panels associated with
four velocity layers within the substratum — VL1, VL2, VL3, and VL4. The left-arrows ← denote the interval velocity-depth
picks. See text for details.
1612 Seismic Data Analysis

FIG. 10.4-7. The West of Shetlands line: prestack depth migration velocity panels at midpoint 721 used in estimating the
substratum model (the region below unconformity A in Figure 10.4-3). Shown here are the velocity panels associated with
four velocity layers within the substratum — VL1, VL2, VL3, and VL4. The left-arrows ← denote the interval velocity-depth
picks. See text for details.
Structural Inversion 1613

FIG. 10.4-8. The West of Shetlands line: prestack depth migration velocity panels at midpoint 801 used in estimating the
substratum model (the region below unconformity A in Figure 10.4-3). Shown here are the velocity panels associated with
four velocity layers within the substratum — VL1, VL2, VL3, and VL4. The left-arrows ← denote the interval velocity-depth
picks. See text for details.
1614 Seismic Data Analysis

FIG. 10.4-9. The West of Shetlands line: depth image from prestack depth migration using the velocity-depth model shown
in Figure 10.4-3. Poststack time-variant filtering and AGC scaling were applied. For event C, compare with Figure 10.4-2.
Structural Inversion 1615
1616 Seismic Data Analysis

FIG. 10.4-11. The West of Shetlands line: depth image from prestack depth migration shown in Figure 10.4-9 converted to
time domain using the velocity-depth model shown in Figure 10.4-3. Poststack time-variant filtering and AGC scaling were
applied. For event C, compare with Figure 10.4-2.
Structural Inversion 1617

FIG. 10.4-12. The West of Shetlands line: depth image from prestack depth migration using the velocity-depth model shown
in Figure 10.4-3 and prestack data without Radon-transform multiple attenuation. Poststack time-variant filtering and AGC
scaling were applied. Compare with the depth image in Figure 10.4-9 derived from data with Radon-transform multiple
attenuation.
1618 Seismic Data Analysis
Structural Inversion 1619

FIG. 10.4-14. The West of Shetlands line: depth image from prestack depth migration shown in Figure 10.4-12 converted to
time domain using the velocity-depth model shown in Figure 10.4-3. Poststack time-variant filtering and AGC scaling were
applied. Compare with the results in Figure 10.4-11.
1620 Seismic Data Analysis

this event is almost impossible to follow in Figure Note, for instance, the fault below CMP 2170 within 1-2
10.4-2. s time window, and the v-shaped structure below CMP
This case study also included a test of the effect 1370. The subsurface appears to be so simple that one
of multiples on prestack depth migration. The image may be tempted to convert this section to depth using
gathers shown in Figure 10.4-10 and the correspond- vertical raypaths and Dix-converted stacking velocities.
ing depth image shown in Figure 10.4-9 are based on However, the anomalous behavior of the stacking ve-
prestack data with Radon-transform multiple attenu- locities will cause Dix conversion to yield meaningless
ation. Using the same velocity-depth model (Figure interval velocities.
10.4-3), but prestack data without multiple attenuation,
prestack depth migration yields the depth image shown
in Figure 10.4-12. Note the overwhelming presence of
Earth Modeling and Imaging in Depth
multiples in this depth image; in contrast, the depth
image in Figure 10.4-9 is largely free of multiples.
Now examine the image gathers with and without To resolve the effect of shallow gas anomalies, we shall
multiple attenuation (Figures 10.4-10 and 10.4-13). Pri- perform earth modeling and imaging in depth. The pro-
mary events are flat or nearly flat, whereas events as- cedure for earth modeling includes coherency inversion
sociated with multiples have significant moveout. The to estimate layer velocities combined with image-ray
multiples dominate the image gathers so much that the depth conversion to delineate reflector geometries. Start
moveout analyses shown in Figure 10.4-4 through 10.4- the analysis by interpreting a set of eight time horizons
8 could not have been conducted using data without within the 0-2.5 s time window from the time-migrated
multiple attenuation. CMP stack (Figure 10.5-3). These time horizons are
Figure 10.4-14 shows depth-to-time conversion of used to obtain the unmigrated time horizons by way of
the depth image shown in Figure 10.4-12. Again, as in ray-theoretical zero-offset modeling. The modeled time
Figure 10.4-11, the conversion was done using the same horizons are assumed to be equivalent to zero-offset re-
velocity field that was used for prestack depth migration flection times, which are then used in coherency inver-
(Figure 10.4-3). Note that the event that is distinctively sion.
identified in Figure 10.4-11 (labeled as C) is somewhat Figure 10.5-4 shows the velocity-depth model based
contaminated by the strong multiples in Figure 10.4-14. on the procedure outlined above. Note the lateral veloc-
ity variations within the layers in the neighborhood of
the faulted zones. Figure 10.5-5 shows the depth image
10.5 IMAGING BENEATH SHALLOW from prestack depth migration using the velocity-depth
GAS ANOMALIES IN model in Figure 10.5-4. Note the crisp image of the fault
THE GULF OF THAILAND below midpoint 2170 and the structural closures along
the fault line between 1.5-2.5 km. Another feature of
The fifth 2-D case study for structural inversion is from interest is the v-shaped structure centered around mid-
the Gulf of Thailand. This case study deals with the point 1370. The boundaries of the closure defined by
deleterious effect of shallow gas anomalies on a multi- the two faults that merge at a depth of approximately
leveled reservoir zone along growth faults. The velocity- 2.4 km below midpoint 1370 are not coincident with
depth model is characterized by near-horizontal layers the changes in the reflector geometries because of fault-
with structure-independent velocity variations. Shown ing. This most likely is caused by the 3-D behavior of
in Figure 10.5-1 is the CMP-stacked section. The the fault, which is not imaged properly by 2-D depth
shallow gas anomalies are represented by the high- migration.
amplitude piecewise-continuous reflections within the Selected image gathers along the line shown in Fig-
first 500 ms of the section. ure 10.5-6 convincingly demonstrate the accuracy of the
The shallow gas anomalies give rise to fluctuations velocity-depth model (Figure 10.5-4). The flatness of
in the stacking velocity field as seen in Figure 10.5-2. events on the image gathers also provides evidence of
For a horizon below the shallow anomaly zone, CMP the accuracy of the depth image (Figure 10.5-5). Com-
raypaths at some locations travel through an overbur- pare the depth images derived from prestack depth mi-
den with laterally varying velocities associated with the gration (Figure 10.5-5) and poststack depth migration
shallow gas anomalies. As a result, traveltimes associ- (Figure 10.5-7) using the same velocity-depth model
ated with deeper reflections are distorted. (Figure 10.5-4). Although both are better than time mi-
Time migration (Figure 10.5-3) using a smoothed gration (Figure 10.5-3), prestack depth migration yields
form of the stacking velocity field (Figure 10.5-2) pro- a better image of the faults and the deep, subtle struc-
duces a seemingly accurate image of the subsurface. tural closures.
Structural Inversion 1621

FIG. 10.5-1. The Gulf of Thailand line: unmigrated CMP stack. (Data courtesy Total Thailand.)

FIG. 10.5-2. The Gulf of Thailand line: stacking velocity field.


1622 Seismic Data Analysis

FIG. 10.5-3. The Gulf of Thailand line: poststack time-migrated CMP stack.

FIG. 10.5-4. The Gulf of Thailand line: velocity-depth model based on a layer-by-layer application of coherency inversion
and image-ray depth conversion.
Structural Inversion 1623

FIG. 10.5-5. The Gulf of Thailand line: depth image from prestack depth migration using the velocity-depth model shown
in Figure 10.5-4. Poststack deconvolution and AGC scaling were applied.
1624 Seismic Data Analysis
Structural Inversion 1625

FIG. 10.5-7. The Gulf of Thailand line: depth image from poststack depth migration using the velocity-depth model shown
in Figure 10.5-4. Depth image from prestack depth migration is shown in Figure 10.5-5.
1626 Seismic Data Analysis

This case study demonstrates that seismic inver- red horizon) — a classic case of the deleterious effect
sion for earth modeling and imaging in depth is not of a complex overburden with strong lateral velocity
required just for targets below obviously complex struc- variations. The complex overburden includes the col-
tures (Sections 10.1 and 10.2), but can also be required lapsed structure above the apex and to the left of the
for resolving low-relief structures in the presence of sub- salt diapir, and the complex geometry of the top-salt
tle, shallow velocity anomalies. boundary (the brown horizon) which is where the most
severe ray bending takes place.
Figure 10.6-3 shows the time surfaces derived from
10.6 3-D STRUCTURAL INVERSION the interpretation of the 3-D volume of unmigrated
APPLIED TO SEISMIC DATA stack data. These time horizons are considered equiva-
FROM THE SOUTHERN NORTH SEA lent to zero-offset time horizons, and as such, are used
in coherency inversion to estimate layer velocities. Hori-
zons TH1-TH6, in ascending order, correspond to base
The first 3-D structural inversion case study is from
Upper and Lower Tertiary, base Cretaceous chalk, base
the Southern Gas Basin of the North Sea. It involves
Upper and Lower Triassic, and base Zechstein. The wa-
a complex overburden associated with salt tectonism.
ter depth in the survey area is shallow and the Upper
The survey area is approximately 90 km2 . The surface
Tertiary velocities near the water bottom are very close
area is roughly a square with dimensions of 10 km in
to the water velocity. Therefore, the water layer was
the crossline direction and 9 km in the inline direc-
considered as part of the Upper Tertiary layer. In fact,
tion. Survey statistics include more than 9 million traces
we observe that there is no appreciable velocity con-
recorded and nearly 300 000 stacked traces. The nomi-
trast at the base Upper Tertiary, and hence, the whole
nal fold of coverage is 32, the inline trace spacing is 12.5
Tertiary section can be considered as one single layer in
m, and the crossline trace spacing is 25 m before trace
earth modeling.
interpolation and 12.5 m after trace interpolation.
To estimate a 3-D velocity-depth model, the fol-
lowing procedure composed from the list of inversion
methods in Table 9-1 is used: Estimation of the Overburden Model

(a) Coherency inversion to estimate layer velocities Based on the volume of stacked data (Figure 10.6-2)
and 3-D poststack depth migration to delineate re- and the well data from the area, we make the following
flector geometries within the overburden, and characterization of the earth model in depth:
(b) constant half-space velocity analysis of image gath-
ers from prestack depth migration to estimate the (a) We may consider the earth model in depth in two
substratum velocity with stacking of image gath- parts — the overburden above the salt diapir and
ers to delineate the reflector geometry of the base- the substratum that includes the salt mass and the
Zechstein (top-Rotliegendes) target horizon. underlying strata.
(b) Time migration may be acceptable within the over-
The procedure outlined above is applied layer-by- burden, but depth migration is imperative within
layer starting from the surface to resolve layer veloc- the substratum.
ity and reflector geometry of one layer before moving (c) The top-salt boundary (the brown horizon in Fig-
onto the next. This minimizes the influence of any er- ure 10.6-2), which also is the boundary between the
rors made in one layer on the parameter estimates for overburden and the substratum, is where the most
the next layer. severe ray bending takes place.
Figure 10.6-1 shows selected crosslines from the un- (d) The overburden above the Zechstein formation
migrated 3-D volume of DMO-stacked data. Superim- comprises layers with significant velocity contrast
posed on the displays in Figure 10.6-2 are the time hori- — Tertiary, Cretaceous chalk, and Triassic units.
zons that correspond to layer boundaries with signifi- (e) As a result of extensional tectonics, the overburden
cant velocity contrast. These are, starting from the top, is intensely faulted. Note the collapsed structures
base Upper and Lower Tertiary, base Cretaceous chalk, above the apex and to the left of the salt diapir.
base Upper and Lower Triassic, and base Zechstein (the (f) Vertical velocity gradients within the overburden
red horizon). The target zones are the Rotliegendes layers are significant and vary spatially.
sands just beneath the base Zechstein, and the underly- (g) The base-Zechstein reflection times (the red hori-
ing Carboniferous Westfalien sequence. Note the trav- zon in Figure 10.6-2) are severely distorted by the
eltime distortions along the base-Zechstein event (the complex overburden.
Structural Inversion 1627

FIG. 10.6-1. Selected crosslines from the 3-D volume of unmigrated DMO-stacked data. The crossline direction is northwest-
southeast, with the northwest being on the right-hand side of the displays. The crosslines are from southwest to northeast
over the survey area with increasing line numbers. (Data courtesy Schlumberger Geco-Prakla and Amoco UK Exploration
and Production Company.)
1628 Seismic Data Analysis

FIG. 10.6-2. Selected crosslines as in Figure 10.6-2 from the 3-D volume of unmigrated DMO-stacked data with the super-
imposed color-coded horizons that correspond to layer boundaries with significant velocity contrast.
Structural Inversion 1629

FIG. 10.6-3. Time horizons interpreted from the volumes of unmigrated DMO-stacked data as in Figure 10.6-2. Horizons
TH1-TH6, in ascending order, correspond to base Upper and Lower Tertiary, base Cretaceous chalk, base Upper and Lower
Triassic, and base Zechstein.
1630 Seismic Data Analysis

Based on these characteristics, an appropriate pro- inversion involves the same steps as for 2-D coherency
cedure for estimating an earth model in depth involves inversion (Section 9.1), except for 3-D ray tracing. For
layer-by-layer application of 3-D coherency inversion to a range of trial constant velocities for the half-space de-
estimate layer velocities and 3-D poststack depth migra- fined by the unknown part of the velocity-depth model,
tion to delineate reflector geometries within the over- follow the steps below:
burden. This is then followed by 3-D prestack depth
migration for estimating the velocity field within the (a) Perform 3-D normal-incidence traveltime inver-
substratum. sion of the time horizon associated with the layer
boundary (Figure 10.6-3) under consideration us-
ing the trial constant velocity assigned to the layer
Model Representation by Tessellation above.
(b) Compute 3-D CMP traveltimes using the known
overburden 3-D velocity-depth model.
Assume that we already have estimated the velocity-
(c) Measure the discrepancy between the modeled and
depth model for the Tertiary section, and that we want
actual CMP traveltimes by way of semblance.
estimate the layer velocity for the Cretaceous chalk
(d) Assign the trial constant velocity that yields the
layer and delineate the reflector geometry associated
maximum semblance as the velocity of the layer
with the base of that layer. Hence, we may consider the
above.
subsurface velocity-depth model made up of two parts
— the known part that includes the Tertiary section,
and the unknown part that includes the chalk layer and For each layer, velocity estimation using 3-D co-
the layers beneath. herency inversion is done at grid locations 1 km apart
By combining the layer velocities and reflector ge- in the inline and crossline directions. The panels in Fig-
ometries associated with the known part of the model, ure 10.6-5 show the CMP gathers at six analysis loca-
create a 3-D velocity-depth model represented in the tions and the semblance curves derived from coherency
form of a tessellated volume (Section 9.0). In a tessel- inversion. The modeled traveltime trajectories associ-
lated model, the volume associated with each layer is ated with the selected velocities that correspond to sem-
divided into a set of tetrahedra, the size and shape of blance maxima are plotted on the CMP gathers. For a
which depend on the geometry of layer bondaries (Fig- given CMP location, note from Figure 10.6-6 that the
ure 10.6-4). As in the present case study, the estimated center data window exhibits a flat event coincident with
velocity and vertical velocity gradient from sonic logs the velocity selection made in Figure 10.6-5.
are assigned to each corner of the tetrahedra within Vertical velocity gradients, which were available
the known part of the model. In the unknown part of from sonic logs, were accounted for in the ray tracing
the model, tetrahedra are populated with constant trial to model the CMP traveltimes that are required in co-
velocities used in the next step (coherency inversion). herency inversion (Section 9.1). Additionally, we want
Tessellated velocity fields can be desirable for efficient to use available sonic logs to guide coherency inversion
ray tracing used in prestack traveltime inversion for ve- and choose an optimum range of constant velocities in
locity estimation. Shown in Figure 10.6-4 is the base- computing semblances. As a reminder, coherency inver-
Lower Tertiary layer boundary above which the model sion also honors ray bending at layer boundaries (Sec-
is known, and below which is the half-space that in- tion 9.1). Nevertheless, application of coherency inver-
cludes the layers yet to be determined. Note that the sion should be confined to parts of a subsurface velocity-
volumes within the known and unknown parts of the depth model where no severe raypath distortions may
model have been subdivided into a set of tetrahedra. occur (Section 9.4).
We circumvent picking of prestack reflection travel-
times in coherency inversion by measuring the discrep-
ancy between modeled and actual traveltimes based on
3-D Coherency Inversion semblance. Although picking of time horizons is done
from the 3-D DMO-stacked volume of data, note that
Perform 3-D coherency inversion to estimate velocity the common-cell gathers used in the analysis are not
nodes for the layer under consideration. Accompanying processed for dip-moveout (DMO) correction. This is
the velocity-depth model (Figure 10.6-4), use the time because we model 3-D prestack traveltimes from source
horizon (Figure 10.6-3) associated with the layer bound- locations at the surface to the actual reflection points
ary picked from the 3-D unmigrated CMP-stacked data along each layer boundary and back to receiver loca-
volume (Figure 10.6-2) and CMP gathers at analysis tions at the surface associated with traces in a CMP
locations (Figure 10.6-5). Three-dimensional coherency gather.
(text continues on p. 1637)
Structural Inversion 1631

FIG. 10.6-4. A tessellated earth model in depth — above the surface that represents the base-Tertiary is the known part of
the model; below the surface is the unknown part yet to be estimated.
1632 Seismic Data Analysis

FIG. 10.6-5. Coherency inversion at six analysis locations. Top and bottom frames in each panel show the semblance curves
and the CMP gathers, respectively. See text for details.
Structural Inversion 1633

FIG. 10.6-6. Data windows along the modeled traveltime trajectories used in coherency inversion. The semblance curves are
shown in Figure 10.6-5. See text for details.
1634 Seismic Data Analysis

FIG. 10.6-7. Velocity field created from the velocity nodes denoted by the green squares which were derived from coherency
inversion applied to a layer within the overburden.
Structural Inversion 1635
1636 Seismic Data Analysis

FIG. 10.6-9. (a) Depth horizon strands for base Cretaceous chalk obtained from the interpretation of the 3-D poststack
depth-migrated volume of data (Figure 10.6-8); (b) the surface fit; (c) the triangulation of the surface; (d) another perspective
view of the surface.
Structural Inversion 1637

Results of coherency inversion are used to pick ve- algorithms (Section 7.3) based on the splitting of the
locities at analysis locations. In a 3-D structural inver- 3-D operator into inline and crossline components, does
sion project, quality control of the velocity nodes at not cause azimuthal positioning errors.
analysis locations is imperative for deriving a spatially The image volume derived from 3-D poststack
consistent layer velocity field. We select the layer ve- depth migration is used in an interpretation session to
locity at a given location based on a combination of incorporate the layer under consideration into the earth
the following factors to make sure that the estimated model.
velocities are geologically plausable:
(a) We interpret the base of the layer under consid-
(1) The maximum of the semblance curve derived from eration — the Cretaceous chalk, to delineate the
coherency inversion as a function of trial constant reflector geometry in depth (Figure 10.6-8). Inter-
velocity (Figure 10.6-5), pretation of the depth horizon is done on crosslines
(2) the flatness of the event in the data windows along from the 3-D volume of poststack depth-migrated
the modeled traveltime trajectories (Figure 10.6-6), data at every tenth line.
(3) the fit of the modeled traveltime trajectory over- (b) By using the interpretation results from the
layed on top of the actual event on the common-cell
crosslines, horizon strands are created (Figure 10.6-
gather (Figure 10.6-5),
9a). These strands then are spatially interpolated
(4) the lack of anomolous behavior in the raypath asso-
to create the surface that represents the reflector
ciated with the modeled traveltime trajectory, and
geometry associated with the base of the layer (Fig-
(5) the magnitude of the velocity node with respect to
the neighboring velocity nodes. ure 10.6-9b). A way to represent the surface in the
computer is by a set of triangles, the size and shape
of which vary depending on the complexity of the
We create the velocity field for the half-space by
reflector geometry (Figure 10.6-9c). The reflector
spatial interpolation between the nodes (Figure 10.6-7).
geometry is taken into account during ray tracing
This is followed by the application of some smoothing
used in coherency inversion to honor ray bending
to the velocity field.
at layer boundaries, and in 3-D poststack depth mi-
We then create a gridded velocity-depth model
that includes the known and unknown part of the gration to account for lateral velocity variations.
model. Gridded velocity fields are desirable for doing 3- (c) Finally, the base Cretaceous chalk surface is added
D poststack depth migration based on finite-difference to the model (Figure 10.6-10), and the procedure
schemes. Gridding recognizes layer boundaries and ver- that includes 3-D coherency inversion and 3-D
tical velocity gradients. poststack depth migration is repeated for the next
layer within the overburden.

3-D Poststack Depth Migration Figure 10.6-11 shows a cross-section of the 3-D ve-
locity field in the crossline direction at each iteration of
the procedure described above. Starting from the top,
By using the gridded velocity field, we perform 3-D
note how the velocity field is updated as the velocity
poststack depth migration down to a depth just below
estimate for the next layer is included in the model.
the current layer under consideration (Figure 10.6-8).
The horizons indicated in Figure 10.6-11 correspond to
The algorithm used in the present case for 3-D post-
stack depth migration is based on a frequency-space base Tertiary (TH2), Cretaceous chalk (TH3), and Up-
explicit scheme and the McClellan transform for design- per Triassic (TH4).
ing a 3-D extrapolation operator (Section 7.3). It can Figure 10.6-12 shows selected crosslines from 3-D
handle arbitrary vertical and lateral velocity variations poststack depth migrated volumes of data after each
and is accurate for dips up to 80 degrees. A 2-D stable iteration. The reflector geometries associated with the
explicit operator is converted into a 3-D operator by layer boundaries included in the overburden model are
applying the McClellan transform coefficients. For each delineated from such cross-sections — base Lower Ter-
frequency component and a given velocity ratio, the ex- tiary (TH2) from Figure 10.6-12a, base Cretaceous
plicit operator is stored in a table and fetched as needed chalk (TH3) from Figure 10.6-12b, base Upper Triassic
at each step of downward continuation. The McClellan (TH4) from Figure 10.6-12c, and base Lower Triassic
transform coefficients are optimized to attain a near- (TH5) from Figure 10.6-12d.
perfect circular symmetry for the impulse response of Figure 10.6-13 shows the results of earth model-
the 3-D operator. This means that the algorithm, un- ing using the procedure described above. The left col-
like the conventional one-pass 3-D poststack migration umn shows the map view of the layer velocities and the
1638 Seismic Data Analysis

FIG. 10.6-10. Tessellated earth model in depth that includes the base-Tertiary (top surface) and base-Cretaceous layer
boundary (bottom surface).

right column shows the perspective view of the base- Estimation of the Substratum Model
layer boundaries. Starting from the top, we see layer
velocities and reflector geometries for Lower Tertiary, After the completion of model estimation for the over-
Cretaceous chalk, Upper and Lower Triassic. Note the burden and before moving down to the substratum re-
collapsed zone (A) on the base-Tertiary surface and the gion, the overburden model must be verified. Model ver-
auxiliary structural feature (B) that is oblique to the ification usually is done by examining selected image
axis of the collapsed zone. The latter becomes more gathers from prestack depth migration along selected
prominent on the surface associated with the base- crosslines. If the overburden model is correct, events
Cretaceous chalk (C). The ridge of the salt diapiric from the overburden on image gathers should exhibit a
structure (D) and the collapsed zone (E) to the left of flat character with negligible moveout. If the residual
the salt diapir are evident on the surface associated with moveout is significant, then it must be corrected for by
the base-Upper Triassic. Finally, note that the base- residual moveout analysis (Section 9.5), and the over-
Lower Triassic (equivalent to the top Zechstein) is rep- burden model must be updated prior to estimating the
resented by a multisegmented surface. This is because substratum model.
the Lower Triassic section is missing in the collapsed To estimate the substratum model, we perform a
zone (E) to the left of the salt diapir. series of 3-D prestack depth migrations using velocity-
(text continues on p. 1647)
Structural Inversion 1639

FIG. 10.6-11. Cross-section of the 3-D velocity field along a specific crossline traverse at each stage of the layer-by-layer
analysis. See text for details.
1640 Seismic Data Analysis

FIG. 10.6-12. Part 1: Crossline 401 from 3-D poststack depth-migrated volumes of data after each iteration of the procedure
to estimate the overburden model. The reflector geometries associated with the layer boundaries included in the overburden
model are delineated from such cross-sections: (a) base Lower Tertiary (TH2), (b) base Cretaceous chalk (TH3), (c) base
Upper Triassic (TH4), and (d) base Lower Triassic (TH5).
Structural Inversion 1641

FIG. 10.6-12. Part 2: Crossline 481 from 3-D poststack depth-migrated volumes of data after each iteration of the procedure
to estimate the overburden model. The reflector geometries associated with the layer boundaries included in the overburden
model are delineated from such cross-sections: (a) base Lower Tertiary (TH2), (b) base Cretaceous chalk (TH3), (c) base
Upper Triassic (TH4), and (d) base Lower Triassic (TH5).
1642 Seismic Data Analysis

FIG. 10.6-12. Part 3: 561 Crossline from 3-D poststack depth-migrated volumes of data after each iteration of the procedure
to estimate the overburden model. The reflector geometries associated with the layer boundaries included in the overburden
model are delineated from such cross-sections: (a) base Lower Tertiary (TH2), (b) base Cretaceous chalk (TH3), (c) base
Upper Triassic (TH4), and (d) base Lower Triassic (TH5).
Structural Inversion 1643

FIG. 10.6-13. Layer velocities and reflector geometries derived from layer-by-layer application of 3-D coherency inversion
and 3-D poststack depth migration. See text for details.
1644 Seismic Data Analysis

FIG. 10.6-14. Part 1: Selected image gathers (every 30th) along selected crosslines from 3-D prestack depth migration.
Structural Inversion 1645

FIG. 10.6-14. Part 2: Selected image gathers (every 30th) along selected crosslines from 3-D prestack depth migration.
1646 Seismic Data Analysis

FIG. 10.6-14. Part 3: Selected image gathers (every 30th) along selected crosslines from 3-D prestack depth migration.
Structural Inversion 1647

FIG. 10.6-15. Image gathers as in Figure 10.6-14 sorted into constant half-space velocity panels.

depth models that have the same overburden but differ- sections of the depth image along selected crosslines
ent constant velocities assigned to the half-space that (Figure 10.6-19).
includes Zechstein and the underlying strata. We exam- The algorithm for prestack depth migration used in
ine the image gathers (Figure 10.6-14) to derive velocity the present case study is based on the Kirchhoff summa-
nodes for the substratum. As shown in the close-up dis- tion (Section 8.5). Output can be a subsurface image in
play in Figure 10.6-15, image gathers associated with depth along an arbitrary traverse, inlines and crosslines,
different half-space velocities differ only at and below or within a specified volume. Also, the output can be im-
the base Zechstein (shown by the arrow) since they all age gathers at specified locations for updating layer ve-
have been derived using the same overburden velocity- locities. The algorithm involves computing traveltimes
depth model. The event associated with the base Zech- and summation of amplitudes as implied by the Kirch-
stein exhibits a different moveout character from one hoff integral. Traveltimes for all shot and receiver loca-
half-space velocity to another. Based on the flatness tions at the surface and reflection points in the subsur-
criterion (Figure 10.6-15) and the behavior of the ge- face are computed by efficient ray tracing. Ray bending
ometry of base-Zechstein that can be traced from the at layer boundaries with velocity contrast is honored in
image-gather stacks (Figure 10.6-16), we determine ve- computing traveltimes. Migrated data are obtained by
locity nodes at selected locations for the Zechstein for- summation along maximum-amplitude traveltime tra-
mation. jectories. Prior to summation, data are treated for op-
We expect some spatial variation in Zechstein erator antialiasing (Section 8.5).
velocities because of the effect of the high-velocity Figure 10.6-20 shows selected image gathers along
anhydrite-dolomite rafts within the halite unit. By spa- the crosslines shown in Figure 10.6-19. Except for some
tial interpolation between the nodes, we create a veloc- events on a few of the image gathers, events exhibit
ity field for the half-space defined by the substratum re- a flat character, which indicates that the earth model
gion of the model (Figure 10.6-17). We then combine the (Figure 10.6-18) and the earth image (Figure 10.6-19)
overburden and the substratum regions of the velocity- are fairly accurate. The base-Zechstein event (denoted
depth model. Figure 10.6-18 shows cross-sections of by the arrow in Figure 10.6-19) exhibits an improved
the complete 3-D velocity-depth model along selected continuity below the complex zone associated with the
crosslines. Finally, we perform 3-D prestack depth mi- salt diapir in the center as compared to the image from
gration using this velocity-depth model to obtain cross- 3-D poststack depth migration (Figure 10.6-21).
1648 Seismic Data Analysis

FIG. 10.6-16. Part 1: Stacks of image gathers as in Figure 10.6-14 along selected crosslines from 3-D prestack depth migration.
Structural Inversion 1649

FIG. 10.6-16. Part 2: Stacks of image gathers as in Figure 10.6-14 along selected crosslines from 3-D prestack depth migration.
1650 Seismic Data Analysis

FIG. 10.6-16. Part 3: Stacks of image gathers as in Figure 10.6-14 along selected crosslines from 3-D prestack depth migration.
Structural Inversion 1651

5.7 million stacked traces, and 11.4 million traces input


to 3-D poststack time migration after crossline trace
interpolation. The nominal fold of coverage is 40, the
inline trace spacing is 12.5 m, and the crossline trace
spacing is 25 m before trace interpolation and 12.5 m
after trace interpolation.
We want to conduct a layer-by-layer earth model
estimation, and use the following combinations of inver-
sion methods:

(a) Coherency inversion to estimate layer velocities


and 3-D poststack depth migration to delineate re-
flector geometries.
(b) Stacking velocity inversion to estimate layer veloc-
ities and image-ray depth conversion of time hori-
zons interpreted from the time-migrated volume of
data to delineate reflector geometries.

We then want to compare the depth structure maps


for the target horizon base Zechstein (top Rotliegendes)
FIG. 10.6-17. The velocity field for the substratum region.
obtained from the two procedures outlined above.
Figure 10.7-1 shows selected inlines from the unmi-
Note that the structural interpretation of the over-
grated 3-D volume of DMO-stacked data. Superimposed
burden from the images obtained by 3-D post- and
on these displays are the time horizons that correspond
prestack depth migrations should not be different. It is
to layer boundaries with significant velocity contrast.
the substratum region where 3-D prestack depth migra-
These are, starting from the top, base Upper and Lower
tion would yield an improved image. The differences in
Tertiary, base Cretaceous chalk, base Upper and Lower
amplitude characteristics and the bandwidth between
Triassic, and base Zechstein (the red horizon). The tar-
the images from 3-D post- and prestack depth migra-
get zone is the Rotliegendes sands just beneath base
tions are due to the use of different algorithms. The
Zechstein, and the underlying Carboniferous Westfalien
algorithm for 3-D poststack depth migration is based
on an explicit finite-difference scheme, whereas the al- sequence.
gorithm for 3-D prestack depth migration is based on The Zechstein formation comprises two units of
the Kirchhoff summation. Although not done in the anhydrite-dolomite embedded in the halite unit. One
present case, results of prestack depth migration should of the anhydrite units is very close to the top boundary
be treated with a signal processing sequence commonly of Zechstein (the green horizon) and is concordant with
applied to stacked data in the time domain, such as it. The deeper, second anhydrite unit has a very com-
poststack deconvolution and frequency filtering. plex geometry as a result of the intensive salt tectonics
in the area. The abundance of diffractions within the
Zechstein formation is associated with this anhydrite-
dolomite unit. One of the objectives of this case study
10.7 3-D STRUCTURAL INVERSION is to investigate whether the complex geometry of the
APPLIED TO SEISMIC DATA second anhyrite-dolomite unit has any subtle, but dele-
FROM THE CENTRAL NORTH SEA terious effect on the geometry of the target horizon —
base Zechstein (top Rotliegendes).
Figure 10.7-2 shows selected inlines as in Fig-
The second 3-D structural inversion case study is from ure 10.7-1 from the 3-D poststack time-migrated vol-
the Central North Sea; it involves a 3-D survey that ume of data. The velocity field used for 3-D poststack
covers an area of nearly 1800 km2 . The surface area is time migration is a smoothed version of the 3-D stack-
roughly a rectangle with dimensions of approximately ing velocity field. Note the complex geometry of the
70 km in the northwest-southeast direction, which is second anhydrite-dolomite unit within Zechstein. The
coincident with the crossline direction, and 25 km in faults along the base Zechstein act as seals on potential
the inline direction. Survey statistics include 456 million reservoir margins. Accurate positioning of these faults,
traces recorded, 228 million prestack traces processed, and determination of their displacements in depth, are
(text continues on p. 1665)
1652 Seismic Data Analysis

FIG. 10.6-18. Part 1: Cross-sections of the final earth model in depth along selected crosslines. See text for details.
Structural Inversion 1653

FIG. 10.6-18. Part 2: Cross-sections of the final earth model in depth along selected crosslines. See text for details.
1654 Seismic Data Analysis

FIG. 10.6-19. Part 1: Cross-sections of the earth image in depth along selected crosslines from 3-D prestack depth migration.
The arrow indicates base Zechstein.
Structural Inversion 1655

FIG. 10.6-19. Part 2: Cross-sections of the earth image in depth along selected crosslines from 3-D prestack depth migration.
The arrow indicates base Zechstein.
1656 Seismic Data Analysis

FIG. 10.6-20. Part 1: Image gathers along selected crosslines as in Figure 10.6-19.
Structural Inversion 1657

FIG. 10.6-20. Part 2: Image gathers along selected crosslines as in Figure 10.6-19.
1658 Seismic Data Analysis

FIG. 10.6-21. Part 1: Cross-sections of the earth image in depth along selected crosslines from 3-D poststack depth migration
using the same velocity-depth model (Figure 10.6-18) as for 3-D prestack depth migration (Figure 10.6-19).
Structural Inversion 1659

FIG. 10.6-21. Part 2: Cross-sections of the earth image in depth along selected crosslines from 3-D poststack depth migration
using the same velocity-depth model (Figure 10.6-18) as for 3-D prestack depth migration (Figure 10.6-19).
1660 Seismic Data Analysis

FIG. 10.7-1. Part 1: Selected inlines from the 3-D volume of unmigrated DMO-stacked data. The inline direction is northeast-
southwest, with the northeast being on the right-hand side of the displays. The inlines are labeled from northwest to southeast
over the survey area with decreasing line numbers. The color-coded horizons correspond to layer boundaries with significant
velocity contrast. (Data courtesy Schlumberger Geco-Prakla, project courtesy Amoco UK Exploration and Production Com-
pany.)
Structural Inversion 1661

FIG. 10.7-1. Part 2: Selected inlines from the 3-D volume of unmigrated DMO-stacked data. The inline direction is northeast-
southwest, with the northeast being on the right-hand side of the displays. The inlines are labeled from northwest to southeast
over the survey area with decreasing line numbers. The color-coded horizons correspond to layer boundaries with significant
velocity contrast. (Data courtesy Schlumberger Geco-Prakla, project courtesy Amoco UK Exploration and Production Com-
pany.)
1662 Seismic Data Analysis

FIG. 10.7-2. Part 1: Selected inlines as in Figure 10.7-1 from the 3-D poststack time-migrated volume of data. The inline
direction is northeast-southwest, with the northeast being on the right-hand side of the displays. The inlines are labeled
from northwest to southeast over the survey area with decreasing line numbers. The color-coded horizons correspond to layer
boundaries with significant velocity contrast.
Structural Inversion 1663

FIG. 10.7-2. Part 2: Selected inlines as in Figure 10.7-1 from the 3-D poststack time-migrated volume of data. The inline
direction is northeast-southwest, with the northeast being on the right-hand side of the displays. The inlines are labeled
from northwest to southeast over the survey area with decreasing line numbers. The color-coded horizons correspond to layer
boundaries with significant velocity contrast.
1664 Seismic Data Analysis

FIG. 10.7-3. Time horizons interpreted from the volumes of unmigrated DMO-stacked and 3-D poststack time-migrated data
as in Figures 10.7-1 and 10.7-2, respectively. Horizon 1, which is not shown here, corresponds to the water-bottom reflection.
Horizons 2-7, in the ascending order, correspond to base Upper and Lower Tertiary, base Cretaceous chalk, base Upper and
Lower Triassic, and base Zechstein.
Structural Inversion 1665

Table 10-1. Lateral variations in horizon-consistent (c) Velocity contrast at layer boundaries causes signif-
stacking velocities shown in Figure 10.7-4. icant ray bending.
(d) Vertical velocity gradients within the overburden
Layer Velocity layers are significant and vary spatially. For in-
Horizon Boundary Range (m/s) stance, the gradient within the Cretaceous chalk
layer can be as much as 1.5 m/s/m.
2 Base Upper Tertiary 1450 − 1950
3 Base Lower Tertiary 1650 − 2350
4 Base Cretaceous 2150 − 2500 Based on these characteristics, an appropriate proce-
5 Base Upper Triassic 2550 − 2800 dure for estimating an earth model in depth involves a
6 Base Lower Triassic 2850 − 3100 layer-by-layer application of 3-D coherency inversion to
7 Base Zechstein 3200 − 3550 estimate layer velocities and 3-D poststack depth mi-
gration to delineate reflector geometries. Assume that
we already have estimated the velocity-depth model for
the first n − 1 layers, and that we want to estimate the
therefore, central to exploration and development ob- layer velocity for the nth layer and delineate the re-
jectives in the area. flector geometry associated with the base of that layer.
Figure 10.7-3 shows the time surfaces derived from Hence, we may consider the subsurface velocity-depth
the interpretation of the 3-D volumes of unmigrated and model made up of two parts — the known part that
time-migrated stack data. Horizons 2-7, in ascending or-
includes the first n − 1 layers, and the unknown part
der, correspond to base Upper and Lower Tertiary, base
that includes the nth layer and the layers beneath. The
Cretaceous chalk, base Upper and Lower Triassic, and
following sequence is conducted for each of the layers in
base Zechstein. The time horizons interpreted from the
the unknown part of the model.
unmigrated DMO-stacked volume are considered equiv-
alent to zero-offset time horizons. They are used in Dix
conversion of stacking velocities, stacking velocity in- (a) Create a 3-D velocity-depth model represented in
version, and coherency inversion to estimate layer ve- the form of a tessellated volume.
locities. The time horizons interpreted from the time- (b) Perform 3-D coherency inversion to estimate ve-
migrated volume are used in map migration by way of locity nodes for the layer under consideration. Co-
image-ray depth conversion to obtain the corresponding herency inversion requires time horizons picked
depth horizons. from a volume of 3-D unmigrated CMP-stacked
Figure 10.7-4 shows the horizon-consistent stack- data (Figure 10.7-3) and CMP gathers at analysis
ing velocities that were extracted from the 3-D stack- locations. For each layer, estimate layer velocities
ing velocity field along time horizons interpreted from using 3-D coherency inversion at grid locations 1
the unmigrated volume of stacked data shown in Fig- km apart in the inline and crossline directions. Fig-
ure 10.7-3. Table 10-1 shows the range of lateral velocity ure 10.7-5 shows results of coherency inversion at
variations within each layer. one specific analysis location. The bottom frames
are copies of the CMP gather at the analysis loca-
tion, while the top frames are copies of the sem-
3-D Coherency Inversion Combined with blance curves derived from coherency inversion (in
3-D Poststack Depth Migration blue) and stacking velocity inversion (in red). The
3-D coherency inversion semblance curve is a mea-
Examination of the earth image volume in time sure of the discrepancy between the modeled and
(Figure 10.7-2) and the velocity field (Figure 10.7-4) actual CMP traveltimes associated with the reflec-
leads to the following characterization of the earth tion event from the layer boundary under consid-
model in depth: eration. The stacking velocity inversion semblance
curve is a measure of the discrepancy between the
(a) The complex geometry of the second anhydrite- modeled and actual stacking velocities associated
dolomite unit within Zechstein may influence the with the reflection event from the layer boundary
geometry of the target level — base Zechstein, and under consideration. The vertical bars in red are
therefore, it may have to be included in the earth aligned with the selected trial velocities that cor-
model. respond to the center data windows in the panels
(b) Lateral velocity variations within the overburden shown in Figure 10.7-6. The modeled traveltime
layers are, in some parts of the survey area, require trajectories associated with the three selected ve-
imaging in depth. locities are plotted on the CMP gather displays in
1666 Seismic Data Analysis

Figure 10.7-5. Note from Figure 10.7-6a, the center


data window exhibits residual moveout which sug-
gests erroneously too low velocity, corresponding
to the selection made in Figure 10.7-5a. Similarly,
the center data window in Figure 10.7-6c exhibits
residual moveout which suggests erroneously too
high velocity, corresponding to the selection made
in Figure 10.7-5c. The center data window in Fig-
ure 10.7-6b, however, exhibits a flat-event charac-
ter, which suggests that the selected velocity in
Figure 10.7-5b is optimum. This velocity coincides
with the peak value of the 3-D coherency inversion
semblance curve and the minimum of the stacking
velocity inversion semblance curve.
(c) Use the results of 3-D coherency inversion — the
semblance spectrum and the modeled CMP trav-
eltimes (Figure 10.7-5), and the residual moveout
panels (Figure 10.7-6) to pick velocities at analysis
locations.
(d) Assign the velocity nodes to the half space defined
by the unknown part of the model which includes
the layer to be delineated. Then, by spatial interpo-
lation between the nodes, a velocity field is created
for the half-space.
(e) Next, create a gridded velocity-depth model that
includes the known and unknown part of the
model.
(f) By using the gridded velocity field, perform 3-D
poststack depth migration down to a depth just
below the current layer under consideration.
(g) The next step involves interpretation of the base
of the layer under consideration to delineate the
reflector geometry in depth. Interpretation of the
depth horizon is done on cross-sections from the
3-D volume of poststack depth-migrated volume of
data at an interval that accounts for the complexity
of the reflector geometry. By using the interpreta-
tion results from the cross-sections, we perform a
surface fit to obtain the 3-D representation of the
reflector geometry associated with the base of the
layer.

Figure 10.7-7 shows selected inlines as in Figure


10.7-1 from the volume of 3-D poststack depth-migrated
data following the layer-by-layer application of the pro-
cedure described above. Compare with the results of
3-D poststack time migration (Figure 10.7-2) and note
the subtle differences in the geometry of the base-
FIG. 10.7-4. Horizon-consistent stacking velocities that
correspond to unmigrated time horizons shown in Figure
Zechstein boundary.
10.7-3. Horizons 2-7, in the ascending order, correspond to The second anhydrite-dolomite unit of Zechstein
base Upper and Lower Tertiary, base Cretaceous chalk, base was included in the earth modeling as a layer of con-
Upper and Lower Triassic, and base Zechstein. stant thickness (100 m). Note from Figure 10.7-8 the
(text continues on p. 1674)
Structural Inversion 1667

FIG. 10.7-5. Coherency inversion at a specific analysis location. Top and bottom frames show the semblance curves and the
CMP gather, respectively. Three picks are indicated for (a) an erroneously too low velocity, (b) optimum velocity, and (c)
erroneously too high velocity.

FIG. 10.7-6. Data windows along the modeled traveltime trajectories used in coherency inversion that correspond to the
three picks made from the semblance curves in Figure 10.7-5 for (a) an erroneously too low velocity, (b) optimum velocity,
and (c) erroneously too high velocity.
1668 Seismic Data Analysis

FIG. 10.7-7. Part 1: Selected inlines as in Figure 10.7-1 from the 3-D poststack depth-migrated volume of data. The inline
direction is northeast-southwest, with the northeast being on the right-hand side of the displays. The inlines are from northwest
to southeast over the survey area with decreasing line numbers. The color-coded horizons correspond to layer boundaries with
significant velocity contrast.
Structural Inversion 1669

FIG. 10.7-7. Part 2: Selected inlines as in Figure 10.7-1 from the 3-D poststack depth-migrated volume of data. The inline
direction is northeast-southwest, with the northeast being on the right-hand side of the displays. The inlines are from northwest
to southeast over the survey area with decreasing line numbers. The color-coded horizons correspond to layer boundaries with
significant velocity contrast.
1670 Seismic Data Analysis

FIG. 10.7-8. Three views of the second anhydrite-dolomite unit of the Zechstein formation; (a) map view over the 3-D survey
area; (b) from top, looking down; (c) from bottom, looking up.
Structural Inversion 1671

FIG. 10.7-9. Layer velocities and reflector geometries derived from layer-by-layer application of 3-D coherency inversion and
3-D poststack depth migration. Horizons 2-7, in the ascending order, correspond to base Upper and Lower Tertiary, base
Cretaceous chalk, base Upper and Lower Triassic, and base Zechstein.
1672 Seismic Data Analysis

FIG. 10.7-10. Layer velocities and reflector geometries derived from a layer-by-layer application of 3-D stacking velocity
inversion and 3-D image-ray depth conversion (map migration). Vertical velocity gradients were not accounted for in the
analysis. Horizons 2-7, in the ascending order, correspond to base Upper and Lower Tertiary, base Cretaceous chalk, base
Upper and Lower Triassic, and base Zechstein.
Structural Inversion 1673

FIG. 10.7-11. Layer velocities and reflector geometries derived from a layer-by-layer application of 3-D stacking velocity
inversion and 3-D image-ray depth conversion (map migration). Vertical velocity gradients from well data were included in
the analysis. Horizons 2-7, in the ascending order, correspond to base Upper and Lower Tertiary, base Cretaceous chalk, base
Upper and Lower Triassic, and base Zechstein.
1674 Seismic Data Analysis

complex geometry of this unit. Although it has a com- The depth structure map for the target horizon
plex geometry and a significant velocity contrast, this — base Zechstein, has the gross features (Figure 10.7-
unit does not have much effect on the geometry of the 10) similar to those of the map obtained from co-
base-Zechstein boundary. This can be explained by not- herency inversion combined with poststack depth mi-
ing that the thickness of the unit is sufficiently small gration (Figure 10.7-9). Nevertheless, it is the sub-
making the traveltime within the unit insignificant de- tle differences that can influence the interpretation.
spite its very fast velocity (5900 m/s). A way to judge whether imaging can be done by
time migration or needs to be done by depth migra-
Figure 10.7-9 shows the results of earth model-
tion is by examining behavior of image rays. Specif-
ing using the procedure described above. Starting from ically, lateral displacement between the point of de-
the top, we see layer velocities and reflector geome- parture of an image ray at the target horizon and
tries for Upper and Lower Tertiary, base Cretaceous the point of emergence of the image ray at the sur-
chalk, base Upper and Lower Triassic, and base Zech- face is an indication of the degree of lateral veloc-
stein. To obtain an accurate depth structure map for ity variations (Figure 10.7-12a). In the case of 3-
the base-Zechstein boundary that is of interest to the D image rays, aside from lateral displacement, the
explorationist, the layer velocities and reflector geome- azimuthal direction of the displacement is an addi-
tries of the layers above are estimated one layer at a tional attribute that needs to be examined (Figure
time starting from the surface. As a result, the ac- 10.7-12b). Note that the largest displacements occur
cumulation of errors in the earth model parameters along the major fault zone within the survey area.
at the zone of interest (base Zechstein) is minimized. We shall end the discussion on this case study by
comparing the results of the three inversion procedures
that we have applied to the data. Shown in Figure 10.7-
13 are the depth structure map of the target horizon —
3-D Stacking Velocity Inversion Combined with base Zechstein derived from the following procedures:
3-D Image-Ray Depth Conversion
(a) coherency inversion, in which vertical velocity gra-
We now consider a traditional approach for obtain- dients were accounted for, combined with poststack
ing structure maps in depth — stacking velocity in- depth migration,
version to estimate layer velocities and image-ray (b) stacking velocity inversion, in which vertical veloc-
depth conversion of time horizons interpreted from ity gradients were accounted for, combined with
the time-migrated volume of data to delineate re- image-ray depth conversion,
flector geometries. Image-ray depth conversion is the (c) stacking velocity inversion, in which vertical veloc-
usual implementation of map migration (Section 9.4). ity gradients were ignored, combined with image-
ray depth conversion.
Application of stacking velocity inversion to 3-D
data involves the same steps as for 2-D stacking ve-
locity inversion (Section 9.2) except for 3-D ray trac- Shown in Figure 10.7-14 are the difference maps, where
the map of Figure 10.7-13a was taken as the reference
ing. As a reminder, stacking velocity inversion can
map, assuming that it is the most accurate between
be considered accurate for velocity-depth models with
the three maps shown in Figure 10.7-13. These differ-
smoothly varying reflector geometries and lateral ve- ence maps are based on the subtraction of the maps of
locity variations much greater than a cable length. Figures 10.7-13b and 10.7-13c from the reference map.
Figures 10.7-10 and 10.7-11 show the results of Note that there are differences in the results obtained
this alternative procedure. The input to stacking ve- from the different procedures outlined above. While
locity inversion is the set of horizon-consistent stack- procedures (a) and (b) appear to produce results that
ing velocities (Figure 10.7-4). The input to image-ray are comparable, procedure (c) produces results with
depth conversion is the set of time horizons picked from significant discrepancies along the major fault zone.
the time-migrated volume of stacked data (Figure 10.7-
3). As in coherency inversion combined with poststack
depth migration, this procedure was conducted layer- 10.8 3-D STRUCTURAL INVERSION
by-layer starting from the surface. Actually, it is con- APPLIED TO SEISMIC DATA
ducted twice — without (Figure 10.7-10) and with (Fig- FROM OFFSHORE INDONESIA
ure 10.7-11) vertical velocity gradients from well data
included in the analysis. In the latter case, vertical gra- The third 3-D structural inversion case study is from
dients are accounted for in ray tracing to compute the offshore Indonesia; it involves a 3-D survey that covers
CMP traveltimes for coherency inversion (Section 9.1). an area of nearly 150 km2 . The surface area is roughly
Structural Inversion 1675

FIG. 10.7-12. (a) Lateral displacement (in meters) of image rays that start at the base-Zechstein surface; (b) displacement
azimuth in degrees with respect to the true north.
1676 Seismic Data Analysis

FIG. 10.7-13. Depth structure map of the base-Zechstein target horizon using three different procedures: (a) coherency in-
version with vertical velocity gradients accounted for, (b) stacking velocity inversion with vertical velocity gradients accounted
for, and (c) stacking velocity inversion without vertical velocity gradients accounted for.
Structural Inversion 1677

FIG. 10.7-14. (a) Difference map: (a) - (b) of Figure 10.7-13, and (b) difference map: (a) - (c) of Figure 10.7-13.
1678 Seismic Data Analysis

a rectangle with dimensions of approxinmately 15 km in (f) Combine the interval velocity maps from step (e)
the inline direction and 10 km in the crossline direction. with the depth horizon maps from step (d) to con-
Survey statistics include more than 15 million traces struct an initial 3-D velocity-depth model.
recorded and 480 000 stacked traces. The nominal fold
of coverage is 32, the inline trace spacing is 12.5 m, and
the crossline trace spacing is 25 m. Figure 7.4-22 shows Model Updating
three inline sections from the volume of 3-D prestack
time-migrated data from this 3-D survey. The steps for model updating by reflection tomography
are outlined below.

Model Building (a) Perform 3-D prestack depth migration using the
initial 3-D velocity-depth model (Figure 10.8-6)
To build the model, we shall apply the structure- and generate the image gathers along selected in-
independent inversion strategy discussed in Section 9.4. line traverses. Note from the selected image gath-
The steps for model building are outlined below. ers shown in Figure 10.8-7 that majority of the
events are nearly flat. Nevertheless, residual move-
(a) As for the time-to-depth conversion sequence de- out along some of the events calls for a modification
scribed in Section 9.4, the Dix equation (9-1) is to the initial velocity-depth model. Figure 10.8-8
used to derive a 3-D interval velocity field from shows three inline sections derived from stacking
a 3-D rms velocity field. The latter is preferably the image gathers. The same inline sections with
estimated from prestack time-migrated data. Re- reverse polarity for better identification of some of
call the sequence for 3-D prestack time migration the fault planes are shown in Figure 10.8-9. Com-
described in Section 7.4. Following the application pare these image sections from 3-D prestack depth
of NMO and 3-D DMO correction, each common- migration with those derived from 3-D poststack
depth migration shown in Figures 10.8-3 and 10.8-
offset volume of data is migrated using an initial ve-
4, and note that prestack depth migration has im-
locity field derived from the DMO-corrected data.
proved the image quality within the zone with in-
The migrated data are then used to perform ve-
tensive faulting.
locity analysis, once again, over the survey area at
(b) Assume that residual moveout on image gathers
specified grid locations (Figure 7.4-19). The picked
can be described by a parabolic moveout equation
rms velocity functions are then used to create a
when analyzed in time, and compute the horizon-
3-D rms velocity field. Figure 10.8-1 shows selected
consistent residual moveout semblance spectra for
time slices from the 3-D rms velocity volume.
events on image gathers that correspond to the
(b) Perform Dix conversion to derive the 3-D interval layer boundaries included in the velocity-depth
velocity field (Figure 10.8-2). model. Shown in Figure 10.8-10 are the semblance
(c) By using this structure-independent 3-D interval spectra for three horizons H2, H3, and H4 as in
velocity field, create an image volume from 3-D Figure 10.8-6 along three inline traverses.
poststack depth migration. Figure 10.8-3 shows (c) For each horizon, pick the residual moveout profiles
three inline sections from the image volume. The by tracking the maxima of the semblance spectra
same inline sections with reverse polarity for bet- for each of the inline traverses and create a resid-
ter identification of some of the fault planes are ual moveout map as shown in Figure 10.8-11 asso-
shown in Figure 10.8-4. Selected depth slices from ciated with a reference offset of the image gathers
the image volume shown in Figure 10.8-5 exhibit a — usually the maximum offset.
complex fault pattern. (d) Use the residual moveout maps for the reference
(d) Interpret a set of depth horizons from the image offset from step (c) and compute the residual move-
volume of 3-D poststack depth migration created in out for all offsets in the image gathers based on
step (c). The resulting depth structure maps for six the parabolic moveout assumption. Build the trav-
horizons that correspond to layer boundaries with eltime error vector ∆t of equation (9-7) using the
significant velocity contrast are shown in Figure residual moveout times.
10.8-6. (e) By using the interval velocity and depth-horizon
(e) Intersect the 3-D interval velocity field from step maps associated with the initial velocity-depth
(b) with the depth horizons from step (d) and ex- model (Figure 10.8-6), construct the coefficient ma-
tract a horizon-consistent interval velocity map for trix L in equation (J-88) as described in Section
each layer (Figure 10.8-6). J.6.
(text continues on p. 1690)
Structural Inversion 1679

FIG. 10.8-1. Horizontal slices from the 3-D rms velocity field associated with the image volume shown in Figure 9.4-1.
1680 Seismic Data Analysis

FIG. 10.8-2. Depth slices from the 3-D interval velocity field derived from Dix conversion of the 3-D rms velocity field as in
Figure 10.8-1.
Structural Inversion 1681

FIG. 10.8-3. Three inline sections (from top to bottom: Inlines 105, 155, and 205 with 1250-m distance between them) from
an image volume derived from 3-D poststack depth migration using the 3-D interval velocity field as in Figure 10.8-2.
1682 Seismic Data Analysis

FIG. 10.8-4. The same three inline sections as in Figure 10.8-3 with reverse polarity.
Structural Inversion 1683

FIG. 10.8-5. Four depth slices from the image volume derived from 3-D poststack depth migration as in Figure 10.8-3 (from
top to bottom at 1000, 1200, 1400, and 1600 m). The vertical axis denotes the inlines and the horizontal axis denotes the
crosslines.
1684 Seismic Data Analysis

FIG. 10.8-6. Maps of layer velocities (left column) and reflector geometries (right column) associated with the data as in
Figure 10.8-3 before tomographic update. Refer to Figure 10.8-12 for the same maps after the tomographic update.
Structural Inversion 1685

FIG. 10.8-7. Selected image gathers from 3-D prestack depth migration of the data as in Figure 10.8-3 using the velocity-
depth model based on the layer velocities and reflector geometries shown in Figure 10.8-6.
1686 Seismic Data Analysis

FIG. 10.8-8. Stack of the image gathers as in Figure 10.8-7 along three inlines (from top to bottom: Inlines 105, 155, and
205 with 1250-m distance between them).
Structural Inversion 1687

FIG. 10.8-9. The same three inline sections as in Figure 10.8-8 with reverse polarity.
1688 Seismic Data Analysis

FIG. 10.8-10. Horizon-consistent residual moveout semblance spectra for the three horizons H2, H3, and H4 as in Figure
10.8-6, derived from the image gathers along the three inline traverses as in Figure 10.8-7.
Structural Inversion 1689

FIG. 10.8-11. Residual moveout maps associated with the six layers that correspond to the velocity-depth model represented
by the maps shown in Figure 10.8-6.
1690 Seismic Data Analysis

(f) Estimate the change in parameters vector ∆p by the final image from 3-D prestack depth migration.
way of the GLI solution given by equation (9-7). Shown in Figure 10.8-17 are three inline sections from
The tomography matrix equation (9-7) can be set the image volume and in Figure 10.8-18 are the same
up such that the residual moveout times given by inline sections with reverse polarity for better identifica-
the traveltime error vector ∆t are used to perturb tion of some of the fault planes. Compare these image
interval velocities and reflector geometries for not sections with those derived from 3-D prestack depth
necessarily all of the horizons, but any combination migration before the tomographic update shown in Fig-
of them. In fact, if judged to be appropriate, only ures 10.8-8 and 10.8-9, and note that the tomographic
one set of the parameters — either layer velocities update of the model has improved the image quality
or reflector geometries may be perturbed. within the zone with intensive faulting.
(g) Update the paramater vector p + ∆p. Figure 10.8- Once the image quality from the selected inlines
12 shows the updated interval velocity and depth- is judged to be acceptable, perform 3-D prestack depth
horizon maps. Compare with the interval velocity migration to produce the entire image volume. Shown in
and depth horizon maps associated with the initial Figure 10.8-19 are selected inline sections from the im-
model (Figure 10.8-6) and note that tomographic age volume derived from 3-D prestack depth migration
update was actually applied to the interval veloc- using the final velocity-depth model. While the selected
ities while the reflector geometries represented by inline sections from 3-D prestack depth migration as in
the depth horizon maps were not altered. Figure 10.8-18 were created using the Kirchhoff sum-
(h) Now combine the new set of interval velocity and mation algorithm, the entire image volume represented
depth horizon maps after the tomographic update by the inline sections in Figure 10.8-19 was created us-
(Figure 10.8-12) to construct a new velocity-depth ing the common-offset 3-D phase-shift-plus-correction
model. algorithm (Section 8.5). This choice for creating the
(i) Perform 3-D poststack depth migration to create image volume was primarily for faster turnaround. Se-
the image volume using the updated velocity-depth lected crossline sections from the image volume in depth
model. Shown in Figure 10.8-13 are three inline sec- are shown in Figure 10.8-20. Note from the inline and
tions from the image volume. Again, the same in- crossline sections in Figures 10.8-19 and 10.8-20 the in-
line sections with reverse polarity for better iden- tensive faulting and folding caused by the wrench tec-
tification of some of the fault planes are shown in tonism.
Figure 10.8-14. Compare these image sections with
those derived from 3-D poststack depth migration
before the tomographic update shown in Figures Volume-Based Interpretation
10.8-3 and 10.8-4, and note that the tomographic
update of the model has improved the image qual- Interpretation of 3-D seismic data is based on one of
ity within the zone with intensive faulting. Selected the following two strategies:
depth slices from the image volume are shown in
Figure 10.8-15.
(a) Image in time, interpret in time to create a set
(j) Re-interpret the depth horizons using the image
of time structure maps, and then perform time-to-
volume from step (i) and examine discrepancies
depth conversion to create the corresponding set of
with the previous interpretation. If required, re-
depth structure maps, or
peat steps (a) through (i) until residual moveouts
(b) Image in depth and interpret directly in depth to
for events associated with the layer boundaries in-
create a set of depth structure maps.
cluded in the model have been reduced to negligible
magnitudes.
Whatever the strategy, the resulting depth structure
maps are calibrated to well data.
We demonstrated the first strategy for interpreta-
Imaging in Depth tion in Section 7.5. We now demonstrate the second
strategy for interpretation. Figure 10.8-21 shows se-
Following the tomographic updating described above, lected 3-D views of the image volume from 3-D prestack
perform 3-D prestack depth migration using the final depth migration of the data as in Figure 10.8-19 in the
velocity-depth model (Figure 10.8-12) to generate the inline direction, and Figure 10.8-22 shows selected 3-D
image sections along the selected inline traverses. Shown views of the image volume in the crossline direction as in
in Figure 10.8-16 are selected image gathers along the Figure 10.8-20. Begin the structural interpretation ses-
three inline traverses. Stack of the image gathers yields sion by 3-D visualization of the image volume in depth
(text continues on p. 1703)
Structural Inversion 1691

FIG. 10.8-12. Maps of layer velocities (left column) and reflector geometries (right column) associated with the data as in
Figure 10.8-3 after tomographic update. Compare with the maps in Figure 10.8-6 before the update.
1692 Seismic Data Analysis

FIG. 10.8-13. Three inline sections (from top to bottom: Inlines 105, 155, and 205 with 1250-m distance between them)
from an image volume derived from 3-D poststack depth migration using the updated velocity-depth model based on the layer
velocities and reflector geometries shown in Figure 10.8-12.
Structural Inversion 1693

FIG. 10.8-14. The same three inline sections as in Figure 10.8-13 with reverse polarity.
1694 Seismic Data Analysis

FIG. 10.8-15. Four depth slices from the image volume derived from 3-D poststack depth migration as in Figure 10.8-13
(from top to bottom at 1000, 1200, 1400, and 1600 m). The vertical axis denotes the inlines and the horizontal axis denotes
the crosslines.
Structural Inversion 1695

FIG. 10.8-16. Selected image gathers from 3-D prestack depth migration of the data as in Figure 10.8-3 using the updated
velocity-depth model based on the layer velocities and reflector geometries shown in Figure 10.8-12. Compare with the image
gathers in Figure 10.8-7 before the tomographic update.
1696 Seismic Data Analysis

FIG. 10.8-17. Stack of the image gathers as in Figure 10.8-16 along three inlines (from top to bottom: Inlines 105, 155, and
205 with 1250-m distance between them). Compare with the image sections in Figure 10.8-8 before the tomographic update.
Structural Inversion 1697

FIG. 10.8-18. The same three inline sections as in Figure 10.8-17 with reverse polarity. Compare with the image sections in
Figure 10.8-9 before the tomographic update.
1698 Seismic Data Analysis

FIG. 10.8-19. Part 1: Selected inline sections from the image volume derived from 3-D prestack depth migration using the
updated velocity-depth model based on the layer velocities and reflector geometries shown in Figure 10.8-12.
Structural Inversion 1699

FIG. 10.8-19. Part 2: Selected inline sections from the image volume derived from 3-D prestack depth migration using the
updated velocity-depth model based on the layer velocities and reflector geometries shown in Figure 10.8-12.
1700 Seismic Data Analysis

FIG. 10.8-19. Part 3: Selected inline sections from the image volume derived from 3-D prestack depth migration using the
updated velocity-depth model based on the layer velocities and reflector geometries shown in Figure 10.8-12.
Structural Inversion 1701

FIG. 10.8-20. Part 1: Selected crossline sections from the image volume derived from 3-D prestack depth migration using
the updated velocity-depth model based on the layer velocities and reflector geometries shown in Figure 10.8-12.
1702 Seismic Data Analysis

FIG. 10.8-20. Part 2: Selected crossline sections from the image volume derived from 3-D prestack depth migration using
the updated velocity-depth model based on the layer velocities and reflector geometries shown in Figure 10.8-12.
Structural Inversion 1703

to understand and characterize the subsurface struc- were derived from the structural interpretation. Each
tural model. Combine the visualization of the image layer is then represented by a solid volume bounded by
volume in the inline and crossline directions with the these surfaces as shown in Figure 10.8-35. Explosion of
the scanning of the depth slices (Figure 10.8-23). Note the entire solid model facilitates viewing of the indi-
that the structural model is based on wrench tectonism vidual depositional units represnted by the solid seg-
that has caused the intensive faulting and folding. ments (Figure 10.8-36). The solid segment associated
By using the seed detection technique (Section 7.5), with the zone of interest can be subsequently popu-
we capture a set of surface patches for each of the depth lated by the seismically derived petrophysical proper-
horizons to be interpreted. Where seed detection fails, ties, which are constrained by the well data, to derive a
as in the complex fault zones, the surface paths are com- reservoir model.
plemented with the horizon strands along the selected
inlines and crosslines derived from line-based interpreta-
tion. The combination of seed detection and line-based
10.9 3-D STRUCTURAL INVERSION
interpretation produce a set of control points for each
APPLIED TO SEISMIC DATA
depth horizon as shown in Figures 10.8-24a through
FROM THE NORTHEAST CHINA
10.8-29a. The control points are then used to create
a complete surface for each horizon by grid-fitting (Sec-
tion J.5) as shown in Figures 10.8-24b through 10.8- The fourth 3-D structural inversion case study is from
29b. The sixth horizon, which is the deepest, shown in the onshore northeast China; it involves a 3-D survey
Figure 10.8-29b is deeper in some areas than the max- that covers an area of nearly 150 km2 . The surface
imum depth (4 km) associated with the image volume area is roughly a rectangle with dimensions of approx-
from 3-D prestack depth migration (Figures 10.8-21 and inmately 13 km in the inline direction and 12 km in the
10.8-22). All six depth horizons are shown in the 3-D crossline direction. Survey statistics include more than
view of Figure 10.8-30. 1.8 million traces recorded and nearly 120 000 stacked
Stratigraphic interpretation uses an amplitude ma- traces. The average fold of coverage is 30, the inline
nipulation technique based on removal of opacity (Sec- trace spacing is 25 m, and the crossline trace spacing is
tion 7.5). Application of this technique to the deposi- 50 m.
tional unit bounded by surface DH3 at the top and DH4 The fold of coverage is fairly uniform over the sur-
at its base (Figure 10.8-31a) leads to the discovery of vey area as seen from the map in Figure 10.9-1. There
an ancient channel (Figure 10.8-31b). This buried chan- are 19 swaths of data recorded using four cables that
nel within the depositional unit of Figure 10.8-31a re- are 100-m apart, each with 60 receiver groups at 50-m
sembles the recent channel system at the water bottom interval. The shots were placed in the direction perpen-
which is dramatically manifested by the depth slices dicular to the swaths. A sketch of the receiver cable
shown in Figure 10.8-32. layout is also shown in Figure 10.9-1.
A close-up view reveals the disruption of the buried Shown in Figure 10.9-2 are selected shot records
channel by the faulting associated with the wrench tec- from each of the 19 swaths with a display gain applied.
tonism in the area (Figure 10.8-33a). By using the sub- Each shot record contains 240 traces that correspond
volume detection (Section 7.5), the channel can be iso- to the four 60-channel receiver cables. Examine the shot
lated as shown in Figure 10.8-33b. The color-coded view records and note that the quality of the recorded data is
of the channel shows that elevation changes exist along remarkably good, except for some coherent, dispersive,
the channel caused by the faulting that has occurred low-frequency and high-amplitude noise at near offsets
after the formation of the channel itself (Figure 10.8- that may be associated with the ground-roll energy.
34a). After its subvolume detection, the channel can be With this case study, we shall develop a compre-
extracted from the volume associated with the deposi- hensive, time-and-depth workflow for processing, inver-
tional unit (Figure 10.8-31a) and viewed in combination sion and interpretation of 3-D seismic data. This work-
with the surface that corresponds to the base of the unit flow is applicable to many of the oil and gas provinces
(Figure 10.8-34b). The objective is to gain a complete with low-relief structures and complex structures that
understanding of the stratigraphic prospect represented involve folding and faulting caused by extensional or
by the channel in relation to the subsurface structural compressional tectonism, and moderate-to-strong lat-
model. eral velocity variations. The workflow, however, is not
Finally, the results of the structural interpretation suitable for imaging beneath complex overburden struc-
(Figures 10.8-24b through 10.8-29b) are combined to tures associated with salt tectonism or overthrust tec-
build a solid earth model. Figure 10.8-35b shows the tonism, and strong-to-severe lateral velocity variations
layer boundaries represented by the depth horizons that (Section 8.2).
(text continues on p. 1713)
1704 Seismic Data Analysis

FIG. 10.8-21. Subvolumes of the image volume from 3-D prestack depth migration viewed along the inline direction.
Structural Inversion 1705

FIG. 10.8-22. Subvolumes of the image volume from 3-D prestack depth migration viewed along the crossline direction.
1706 Seismic Data Analysis

FIG. 10.8-23. Selected depth slices of the image volume from 3-D prestack depth migration as in Figures 10.8-22 and 10.8-23.
Structural Inversion 1707

FIG. 10.8-24. Interpretation of depth horizon DH1 from the image volume as in Figures 10.8-22 and 10.8-23: (a) map
view of picks created from a combination of seed detection and line-based interpretation, (b) the surface in depth created by
gridfitting the control points in (a). The displays in (a) and (b) are color-coded independently.
1708 Seismic Data Analysis

FIG. 10.8-25. Interpretation of depth horizon DH2 from the image volume as in Figures 10.8-22 and 10.8-23: (a) map
view of picks created from a combination of seed detection and line-based interpretation, (b) the surface in depth created by
gridfitting the control points in (a). The displays in (a) and (b) are color-coded independently.
Structural Inversion 1709

FIG. 10.8-26. Interpretation of depth horizon DH3 from the image volume as in Figures 10.8-22 and 10.8-23: (a) map
view of picks created from a combination of seed detection and line-based interpretation, (b) the surface in depth created by
gridfitting the control points in (a). The displays in (a) and (b) are color-coded independently.
1710 Seismic Data Analysis

FIG. 10.8-27. Interpretation of depth horizon DH4 from the image volume as in Figures 10.8-22 and 10.8-23: (a) map
view of picks created from a combination of seed detection and line-based interpretation, (b) the surface in depth created by
gridfitting the control points in (a). The displays in (a) and (b) are color-coded independently.
Structural Inversion 1711

FIG. 10.8-28. Interpretation of depth horizon DH5 from the image volume as in Figures 10.8-22 and 10.8-23: (a) map
view of picks created from a combination of seed detection and line-based interpretation, (b) the surface in depth created by
gridfitting the control points in (a). The displays in (a) and (b) are color-coded independently.
1712 Seismic Data Analysis

FIG. 10.8-29. Interpretation of depth horizon DH6 from the image volume as in Figures 10.8-22 and 10.8-23: (a) map
view of picks created from a combination of seed detection and line-based interpretation, (b) the surface in depth created by
gridfitting the control points in (a). The displays in (a) and (b) are color-coded independently.
Structural Inversion 1713

FIG. 10.8-30. All six surfaces in depth as in Figures 10.8-24 through 10.8-29. Each surface is color-coded independently.

We shall execute a unified seismic workflow in five unmigrated stack volume associated with the 3-D
phases as illustrated in Figure 10.9-3 and outlined be- prestack time-migrated data.
low: (3) Stratigraphic inversion with deliverables that in-
clude the AVO attribute volumes derived from
prestack amplitude inversion and an acoustic
(1) 3-D DMO processing with deliverables that include impedance volume derived from poststack ampli-
a set of 3-D DMO gathers, a 3-D DMO stack vol- tude inversion. In this phase, we also derive a 3-D
ume, a 3-D DMO velocity field and an image vol- interval velocity field by Dix conversion of the rms
ume derived from 3-D poststack time migration. velocity field from phase 3 and use it to create an
(2) 3-D prestack time migration with deliverables that image volume from 3-D poststack depth migration.
include a set of common-reflection-point (CRP) (4) Structural inversion that involves the estimation
gathers, an image volume from 3-D prestack time of an initial 3-D velocity-depth model and model
migration, a 3-D rms velocity field and a 3-D zero- updating with deliverables that include a final 3-D
offset wavefield modeled from the image volume. velocity-depth model and an image volume derived
The latter may be treated as the equivalent of an from 3-D prestack depth migration.
(text continues on p. 1720)
1714 Seismic Data Analysis

FIG. 10.8-31. (a) The depositional unit bounded by surface DH3 (as in Figure 10.8-26) at the top and bounded by surface
DH4 (as in Figure 10.8-27) at the bottom; the subvolume has been extracted from the image volume as in Figures 10.8-22
and 10.8-23, (b) the same subvolume as in (a) with the opacity removed.
Structural Inversion 1715

FIG. 10.8-32. Selected shallow depth slices of the image volume from 3-D prestack depth migration as in Figures 10.8-22
and 10.8-23.
1716 Seismic Data Analysis

FIG. 10.8-33. (a) Close-up view of the subvolume as in Figure 10.8-31b, (b) seed detection applied to the lower portion of
the channel.
Structural Inversion 1717

FIG. 10.8-34. (a) The same subvolume as in Figure 10.8-33b with the channel color-coded in depth, (b) the same channel
as in (a) displayed with the base surface DH4 of the depositional unit as in Figure 10.8-31a.
1718 Seismic Data Analysis

FIG. 10.8-35. (a) All six surfaces in depth as in Figures 10.8-24 through 10.8-29; (b) solid model created from the depth
horizons in (a).
Structural Inversion 1719

FIG. 10.8-36. Explosion of the solid model shown in Figure 10.8-35b.


1720 Seismic Data Analysis

(5) Structural and stratigraphic interpretation, and cal- selected inline and crossline sections, respectively,
ibration to well data with the ultimate objective of from the image volume derived from 3-D poststack
building a reservoir model that is fundamentally a time migration. To circumvent the adverse effect
petrophysical representation of an earth model in of spatial aliasing on migration, especially in the
depth. crossline direction, trace interpolation before mi-
gration was considered. However, this process dete-
In the present case study, we shall execute the entire riorated the stacked data in zones of intensive fault-
time-with-depth workflow outlined above, except for ing associated with extensional tectonism. There-
stratigraphic inversion in phase 3 and reservoir mod- fore, in lieu of trace interpolation, the unmigrated
eling in phase 5. data in step (e) were filtered down to a passband of
6-36 Hz; hence the difference in the frequency con-
tent between the unmigrated data in Figure 10.9-8
and the migrated data in Figure 10.9-10.
3-D DMO Processing
The deliverables from phase 1 — 3-D DMO pro-
Prestack signal processing of the data included geo-
cessing, include a set of 3-D DMO-corrected gathers,
metric spreading correction using a t2 -scaling function
a volume of 3-D DMO-stacked data, a volume of 3-D
(Figure 10.9-4), spiking deconvolution and time-variant
DMO velocity field, and an image volume derived from
spectral whitening (Figure 10.9-5). This parsimonious
3-D poststack time migration.
sequence sufficed to attain a desired flat and broad spec-
trum within the passband (Figure 10.9-6). Note that
the coherent linear noise that dominates the raw shot
record (Figure 10.9-4a) does not stand out in the shot 3-D Prestack Time Migration
gather after spectral balancing (Figure 10.9-5b).
Following the signal processing, the data are ready While the interpreter may have a sneak preview of the
for DMO processing: subsurface image using the volume of 3-D poststack
time-migrated data from phase 1, we move on to phase
(a) Sort the common-shot gathers to common-cell 2 for 3-D prestack time migration.
gathers and perform stacking velocity analysis over
a sparse grid of 2 × 2 km. (a) Sort the 3-D DMO-corrected and moveout-
(b) Apply NMO and 3-D DMO correction to remove corrected data from step (e) of phase 1 to common-
source-receiver azimuth and dip effects from the offset volumes.
stacking velocities. (b) Assume that each of the common-offset volumes
(c) Apply inverse NMO correction and repeat the of data is a replica of a 3-D zero-offset wavefield
stacking velocity analysis over a finer grid of 0.5 × and perform 3-D zero-offset time migration using a
0.5 km. Figure 10.9-7 shows a subset of four veloc- regionally averaged, but vertically varying velocity
ity analysis panels over the survey area. Each panel function derived from the 3-D DMO velocity field
comprises the common-cell gather at the analysis from step (d) of phase 1.
location and the velocity spectrum computed from (c) Sort the migrated common-offset volumes of data
it. Generally, the velocity picking from the DMO- back to common-cell gathers, apply inverse move-
corrected data was consistently reliable over the out correction using the 3-D DMO velocity field
survey area. from step (d) of phase 1, and repeat the velocity
(d) Create a 3-D DMO velocity field from the vertical analysis over a grid of 0.5 × 0.5 km.
functions picked in step (c). (d) Create a 3-D rms velocity field associated with the
(e) Apply NMO correction to the 3-D DMO-corrected migrated data from the vertical functions picked in
data from step (c) using the velocity field from step step (c).
(d) and stack the data. Figures 10.9-8 and 10.9-9 (e) Apply NMO correction to the 3-D common-offset-
show selected inline and crossline sections, respec- migrated data from step (c) using the rms velocity
tively, from the volume of the 3-D DMO stack fol- field from step (d) and stack the data. Figures 10.9-
lowing poststack deconvolution and band-pass fil- 12 and 10.9-13 show selected inline and crossline
ter. sections, respectively, from the image volume de-
(f) Perform 3-D poststack time migration using a lat- rived from the stack of the 3-D common-offset mi-
erally smoothed form of the 3-D DMO velocity field grations. Again, to circumvent the adverse effect of
from step (d). Figures 10.9-10 and 10.9-11 show spatial aliasing, the common-offset volumes of data
(text continues on p. 1742)
Structural Inversion 1721

FIG. 10.9-1. Fold of coverage map and recording geometry of the land 3-D seismic data associated with the case study
presented in Section 10.9.
1722 Seismic Data Analysis

FIG. 10.9-2. Part 1: Selected shot records from the 3-D survey data associated with the case study presented in Section
10.9.
Structural Inversion 1723

FIG. 10.9-2. Part 2: Selected shot records from the 3-D survey data associated with the case study presented in Section
10.9.
1724 Seismic Data Analysis

FIG. 10.9-2. Part 3: Selected shot records from the 3-D survey data associated with the case study presented in Section
10.9.
Structural Inversion 1725

FIG. 10.9-3. A time-with-depth workflow used in the case study presented in Section 10.9.
1726 Seismic Data Analysis

FIG. 10.9-4. (a) A raw shot record from the 3-D seismic data associated with the case study presented in Section 10.9, (b)
after geometric spreading correction.
Structural Inversion 1727

FIG. 10.9-5. (a) The same shot record as in Figure 10.9-4b after deconvolution, and (b) time-variant spectral whitening.
1728 Seismic Data Analysis

FIG. 10.9-6. Amplitude spectrum of the shot record in (a) Figure 10.9-4a, (b) Figure 10.9-4b, (c) Figure 10.9-5a, and (d)
Figure 10.9-5b.
Structural Inversion 1729

FIG. 10.9-7. A selection of four velocity analyses applied to the 3-D data associated with the case study presented in Section
10.9 after the application of 3-D DMO correction. Shown in each panel are the common-cell gather and the velocity spectrum
at the analysis location.
1730 Seismic Data Analysis

FIG. 10.9-8. Selected inline sections from the 3-D DMO stack volume of data associated with the shot records as in Figure
10.9-3.
Structural Inversion 1731

FIG. 10.9-9. Selected crossline sections from the 3-D DMO stack volume of data associated with the shot records as in
Figure 10.9-3.
1732 Seismic Data Analysis

FIG. 10.9-10. Selected inline sections from the volume of 3-D poststack time migration of the 3-D DMO-stacked data as in
Figure 10.9-8.
Structural Inversion 1733

FIG. 10.9-11. Selected crossline sections from the volume of 3-D poststack time migration of the 3-D DMO-stacked data as
in Figure 10.9-9.
1734 Seismic Data Analysis

FIG. 10.9-12. Selected inline sections from the volume of 3-D prestack time migration of the data associated with the shot
records as in Figure 10.9-3. The same inlines from the volume of 3-D poststack time migration are shown in Figure 10.9-10.
Structural Inversion 1735

FIG. 10.9-13. Selected crossline sections from the volume of 3-D prestack time migration of the data associated with the
shot records as in Figure 10.9-3. The same crosslines from the volume of 3-D poststack time migration are shown in Figure
10.9-11.
1736 Seismic Data Analysis

FIG. 10.9-14. Selected inline sections from the volume of the 3-D zero-offset wavefield modeled from the volume of 3-D
prestack time migration of the data as in Figure 10.9-12.
Structural Inversion 1737

FIG. 10.9-15. Selected crossline sections from the volume of the 3-D zero-offset wavefield modeled from the volume of 3-D
prestack time migration of the data as in Figure 10.9-13.
1738 Seismic Data Analysis

FIG. 10.9-16. Selected inline cross-sections from the volume of the 3-D rms velocity field associated with the 3-D prestack
time-migrated data as in Figure 10.9-12.
Structural Inversion 1739

FIG. 10.9-17. Selected inline sections from the 3-D prestack time-migrated volume of data created by 3-D zero-offset
migration of the modeled 3-D zero-offset wavefield volume as in Figure 10.9-14 using the 3-D rms velocity field as in Figure
10.9-16.
1740 Seismic Data Analysis

FIG. 10.9-18. Selected crossline sections from the 3-D prestack time-migrated volume of data created by 3-D zero-offset
migration of the modeled 3-D zero-offset wavefield volume as in Figure 10.9-15 using the 3-D rms velocity field as in Figure
10.9-16.
Structural Inversion 1741

FIG. 10.9-19. Selected inline cross-sections from the volume of the 3-D interval velocity field derived from the 3-D rems
velocity field as in Figure 10.9-16 by Dix conversion.
1742 Seismic Data Analysis

were filtered down to a passband of 6-36 Hz prior titioning of energy associated with an incident com-
to migration. pressional plane wave at a horizontal layer bound-
(f) Model a 3-D zero-offset wavefield from the image ary into its reflected and refracted compressional-
volume derived from the 3-D common-offset mi- and shear-wave components, we want to use the
gration and stack in step (e) using the same verti- CRP gathers from 3-D prestack time migration
cally varying velocity function as in step (b). Fig- that contain events in their migrated positions in
ures 10.9-14 and 10.9-15 show selected inline and prestack amplitude inversion rather than the DMO
crossline sections, respectively, from the modeled gathers that contain events in their unmigrated po-
3-D zero-offset wavefield volume. sitions.
(g) The final step in phase 2 involves remigration of the (b) Apply poststack amplitude inversion (Section 11.3)
to the AVO intercept volume to derive the acous-
stacked data. Specifically, perform 3-D poststack
tic impedance volume. It is assumed that the AVO
time migration for which the input data volume is
intercept attribute is a close representation of the
the 3-D zero-offset modeled wavefield from step (f)
reflection amplitudes at vertical incidence and zero
and the migration velocity field is the 3-D rms ve- offset. Alternatively, poststack amplitude inver-
locity field from step (d). Figure 10.9-16 shows se- sion can be applied to the image volume from 3-
lected inline cross-sections from the 3-D rms veloc- D prestack time migration derived in step (g) of
ity volume, and Figures 10.9-17 and 10.9-18 show phase 2.
selected inline and crossline sections, respectively, (c) Apply Dix conversion to the 3-D rms velocity field
from the image volume derived from 3-D prestack from step (d) of phase 2 to derive a 3-D inter-
time migration based on the above sequence. The val velocity field. Figure 10.9-19 shows selected in-
migration artifacts on the left-hand side of some of line cross-sections from the 3-D interval velocity
the sections are caused by the missing data zone volume. To satisfy the underlying assumptions for
within the survey area (bottom left corner of the Dix conversion (Section J.4), we want to use the
fold map in Figure 10.9-1). rms velocity field derived in conjunction with 3-D
The deliverables from phase 2 — 3-D prestack time prestack time migration of the data in phase 2 that
migration, include a set of CRP gathers, a volume of the is associated with events in their migrated positions
3-D zero-offset wavefield which may be considered as rather than the DMO velocity field in conjunction
equivalent to an unmigrated stack volume, a volume of with 3-D DMO correction of the data in phase 1
the 3-D DMO rms velocity field, and an image volume that is associated with events in their unmigrated
derived from 3-D prestack time migration. positions.
(d) Perform 3-D poststack depth migration of the 3-D
zero-offset wavefield from step (f) of phase 2 using
the 3-D interval velocity field from step (c) of the
From RMS to Interval Velocities present phase of the workflow.

As we note from the list of deliverables from phase 2, we The deliverables from phase 3 include a volume of
are not implementing phase 2 — 3-D prestack time mi- 3-D interval velocity field and an image volume derived
gration, just for imaging the subsurface. We shall make from 3-D poststack depth migration.
use of the 3-D rms velocity field and the 3-D zero-offset
wavefield from phase 2 to move from time to depth do-
main in the analysis. Although it will not be considered Structural Inversion
in the present case study, phase 3 of the generalized
workflow also includes amplitude inversion that uses the
We want to construct a structurally consistent 3-D
CRP gathers from 3-D prestack time migration and the velocity-depth model using the deliverables from phase
image volume itself. 3 and update it to obtain a final 3-D velocity-depth
model. We then want to use this earth model in depth
(a) Apply prestack amplitude inversion (Section 11.2) to create an earth image in depth from 3-D prestack
to the CRP gathers and derive the AVO at- depth migration.
tribute volumes. These include the intercept and
gradient volumes, P -wave and S-wave reflectivity (a) Interpret a set of depth horizons from the image
and impedance volumes, pseudo-Poisson and fluid- volume derived from 3-D poststack depth migra-
factor volumes (Section 11.2). To satisfy the un- tion (Figure 10.9-20a). These depth horizons corre-
derlying assumptions for AVO analysis, and specifi- spond to layer boundaries with significant velocity
cally the Zoeppritz equations that describe the par- contrast.
Structural Inversion 1743

FIG. 10.9-20. (a) Depth horizons interpreted from the volume of 3-D poststack depth migration of the 3-D zero-offset
wavefield as in Figures 10.9-14 and 10.9-15 using the 3-D interval velocity field as in Figure 10.9-19; (b) the same depth
horizons as in (a) with the additional phantom horizons shown in green used to sub-divide each of the layers in (a) into a set
of 10 layers to preserve the lateral and vertical variations in the gradient of the 3-D interval velocity field as in Figure 10.9-19.
1744 Seismic Data Analysis

(b) To preserve the vertical and lateral velocity varia- volume in depth derived from 3-D prestack depth
tions in the 3-D interval velocity field derived from migration.
Dix conversion (Figure 10.9-19), create a set of
phantom depth horizons by subdividing each of The deliverables from phase 4 — structural inver-
the layers bounded by the depth horizons (Fig- sion, include a volume of structurally consistent 3-D in-
ure 10.9-20a) interpreted from the 3-D poststack terval velocity field and an image volume derived from
depth-migrated volume of data into a set of 10 sub- 3-D prestack depth migration.
layers. The phantom horizons can be conformed ei-
ther to the top-boundary or the base-boundary, or,
as in the present case, to both the top- and base-
boundary of the layer under consideration. Figure Structural and Stratigraphic Interpretation
10.9-20b shows the principal depth horizons as in
Figure 10.9-20a interleaved with the phantom hori- By using the techniques we learned in Section 7.5,
zons in green. we now interpret the image volume from 3-D prestack
(c) Intersect the 3-D interval velocity volume (Figure depth migration — structural intepretation based on
10.9-19) from step (c) of phase 3 with the principal picking a set of depth horizons and stratigraphic in-
and phantom depth horizons from step (b) of the terpretation based on amplitude manipulation. Figure
present phase and extract horizon-consistent inter- 10.9-28 shows six depth horizons that coincide with the
val velocity surfaces. Then, combine these layer ve- geologic markers in the area. The top three horizons,
locities with the reflector geometries represented by DH1, DH2, and DH3, were delineated largely by us-
the depth horizons to create an initial structurally ing the seed detection technique that involves connect-
consistent 3-D velocity-depth model. Figure 10.9- ing neighboring voxels with amplitudes that are within
21 shows selected inline cross-sections from the ini- a specified range (Section 7.5). The bottom three hori-
tial 3-D velocity-depth model. zons, DH4, DH5, and DH6, were delineated by line-
(d) Perform 3-D prestack depth migration and gener- based interpretation that involves creating a set of hori-
ate a set of image gathers along every tenth in- zon strands along selected inlines and crosslines.
line, and compute the residual moveout semblance The effect of extensional tectonism is evident on
spectra for model updating (Section 9.5). Figure the top three depth horizons. Note in particular the
10.9-22 shows selected panels of residual-moveout series of faults across the lowlands indicated in green.
analysis. Each panel comprises the image gather at Examine the depth horizons in reverse order, starting
the analysis location and the residual moveout sem- with DH6 and ending with DH1, to reconstruct the
blance spectrum computed from it. While many of structural evolution of the area. In the beginning, there
the events in the image gathers show very small or was an erosional surface cut by a deep channel as seen
no residual moveout (panel I-51), note that some on horizon DH6. There may have been a period of com-
events exhibit significant residual moveout errors pressional tectonism that gave rise to the gently folded
(panel I-41). Perform residual moveout corrections surfaces in the lowlands portion of horizons DH5 and
and update the initial velocity-depth model (Sec- DH4. Then, a reversal from compressional tectonism to
tion 9.5). extensional tectonism began to cause subsidence of the
(e) Repeat the residual-moveout analysis a few times basin, which in turn gave rise to the formation of the se-
until the residual-moveout errors have been re- ries of faults parallel to the ancient shoreline. The high-
duced significantly. Figures 10.9-23 and 10.9-24 lands indicated by the orange on all horizons have re-
show inline and crossline cross-sections, respec- mained as hiatus until the geologic times approximately
tively, from the final 3-D velocity-depth model. The coincident with the age of the overburden above horizon
residual-moveout analysis from the last iteration of DH1.
model updating shown in Figure 10.9-25 indicates Now split the image volume into subvolumes that
significant reduction of the residual-moveout errors represent individual depositional units as shown in Fig-
(compare, for instance, panel I-41 in Figure 10.9-25 ure 10.9-29. The extensional fault patterns are observed
with that in Figure 10.9-22). on the top surfaces of these subvolumes down to hori-
(f) Perform 3-D prestack depth migration using the fi- zon DH4. To investigate the interior of each of these
nal 3-D velocity-depth model from step (e). The in- depositional units, cut them into thin slices as shown
put to 3-D prestack depth migration is the prestack in Figure 10.9-30a. Then, apply transparency (Section
data set with signal processing applied as in Figure 7.5) to each of the thin slices to identify depositional
10.9-5b. Figures 10.9-26 and 10.9-27 show selected features of interest. Note the presence of a stream with
inlines and crosslines, respectively, from the image its traverse orthogonal to the fault patterns. Also shown
(text continues on p. 1778)
Structural Inversion 1745

FIG. 10.9-21. Selected inline cross-sections from the initial 3-D velocity-depth model derived from the depth horizons in
Figure 10.-9-20b and the 3-D interval velocity field as in Figure 10.9-19.
1746 Seismic Data Analysis

FIG. 10.9-22. Residual-moveout analysis along crossline 321 at selected inline locations using image gathers derived from
3-D prestack depth migration of the data as in Figure 10.9-5b using the initial 3-D velocity-depth model as in Figure 10.9-21.
Shown in each panel are the image gather and the residual-moveout semblance spectrum computed at the analysis location.
Structural Inversion 1747

FIG. 10.9-23. Part 1: Selected inline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22. The corresponding cross-sections of the initial
3-D velocity-depth model are shown in Figure 10.9-21.
1748 Seismic Data Analysis

FIG. 10.9-23. Part 2: Selected inline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22. The corresponding cross-sections of the initial
3-D velocity-depth model are shown in Figure 10.9-21.
Structural Inversion 1749

FIG. 10.9-23. Part 3: Selected inline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22. The corresponding cross-sections of the initial
3-D velocity-depth model are shown in Figure 10.9-21.
1750 Seismic Data Analysis

FIG. 10.9-23. Part 4: Selected inline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22. The corresponding cross-sections of the initial
3-D velocity-depth model are shown in Figure 10.9-21.
Structural Inversion 1751

FIG. 10.9-23. Part 5: Selected inline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22. The corresponding cross-sections of the initial
3-D velocity-depth model are shown in Figure 10.9-21.
1752 Seismic Data Analysis

FIG. 10.9-24. Part 1: Selected crossline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22.
Structural Inversion 1753

FIG. 10.9-24. Part 2: Selected crossline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22.
1754 Seismic Data Analysis

FIG. 10.9-24. Part 3: Selected crossline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22.
Structural Inversion 1755

FIG. 10.9-24. Part 4: Selected crossline cross-sections from the final 3-D velocity-depth model following the model updating
based on the residual-moveout analysis of image gathers as in Figure 10.9-22.
1756 Seismic Data Analysis

FIG. 10.9-25. Residual-moveout analysis of the image gathers along crossline 321 at selected inline locations derived from
3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D velocity-depth model as in Figures 10.9-23
and 10.9-24. Shown in each panel are the image gather and the residual-moveout semblance spectrum computed at the analysis
location.
Structural Inversion 1757

FIG. 10.9-26. Part 1: Selected inlines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
1758 Seismic Data Analysis

FIG. 10.9-26. Part 2: Selected inlines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
Structural Inversion 1759

FIG. 10.9-26. Part 3: Selected inlines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
1760 Seismic Data Analysis

FIG. 10.9-26. Part 4: Selected inlines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
Structural Inversion 1761

FIG. 10.9-26. Part 5: Selected inlines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
1762 Seismic Data Analysis

FIG. 10.9-27. Part 1: Selected crosslines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
Structural Inversion 1763

FIG. 10.9-27. Part 2: Selected crosslines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
1764 Seismic Data Analysis

FIG. 10.9-27. Part 3: Selected crosslines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
Structural Inversion 1765

FIG. 10.9-27. Part 4: Selected crosslines from the image volume in depth derived from the stack of the image gathers as
in Figure 10.9-25 that were created by 3-D prestack depth migration of the data as in Figure 10.9-5b using the final 3-D
velocity-depth model as in Figure 10.9-23.
1766 Seismic Data Analysis

FIG. 10.9-28. Part 1: Depth horizon DH1 interpreted from the image volume derived from 3-D prestack depth migration.
Selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
Structural Inversion 1767

FIG. 10.9-28. Part 2: Depth horizon DH2 interpreted from the image volume derived from 3-D prestack depth migration.
Selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
1768 Seismic Data Analysis

FIG. 10.9-28. Part 3: Depth horizon DH3 interpreted from the image volume derived from 3-D prestack depth migration.
Selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
Structural Inversion 1769

FIG. 10.9-28. Part 4: Depth horizon DH4 interpreted from the image volume derived from 3-D prestack depth migration.
Selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
1770 Seismic Data Analysis

FIG. 10.9-28. Part 5: Depth horizon DH5 interpreted from the image volume derived from 3-D prestack depth migration.
Selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
Structural Inversion 1771

FIG. 10.9-28. Part 6: Depth horizon DH6 interpreted from the image volume derived from 3-D prestack depth migration.
Selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
1772 Seismic Data Analysis

FIG. 10.9-29. The subvolume associated with the layer bounded by depth horizons (a) DH1 on top and DH2 at the bottom;
(b) DH2 on top and DH3 at the bottom; and (c) DH3 on top and DH4 at the bottom. The depth horizons are shown in Figure
10.9-28 and selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27, respectively.
Structural Inversion 1773

FIG. 10.9-30. (a) A subvolume associated with a 50-m thick slice bounded by depth horizon DH1 on top. The depth horizon
is shown in Figure 10.9-28 and selected inlines and crosslines from the image volume are shown in Figures 10.9-26 and 10.9-27,
respectively, (b) Map view of the subvolume shown in (a) with opacity removed to enhance the structural and depositional
features.
1774 Seismic Data Analysis

FIG. 10.9-31. (a) An inline section from the image volume derived from 3-D prestack depth migration with the interpretation
of the top and base of the deltaic depositional sequence. Other selected inlines and crosslines from the image volume are shown
in Figures 10.9-26 and 10.9-27, respectively. (b) A close-up view of a crossline section with the horizon strands representing
the sublayers associated with the deltaic sequence.
Structural Inversion 1775

FIG. 10.9-32. Top and base of the deltaic sequence interpreted from the image volume derived from 3-D prestack depth
migration.
1776 Seismic Data Analysis

FIG. 10.9-33. Selected inline cross-sections of the deltaic sequence with its top and base shown in Figure 10.9-32.
Structural Inversion 1777

FIG. 10.9-34. (a) The subvolume associated with the deltaic sequence extracted from the image volume derived from 3-D
prestack depth migration using the top and base surfaces shown in Figure 10.9-32, (b) an inline cross-section of the subvolume
in (a).
1778 Seismic Data Analysis

in Figure 10.9-30b is the map view of the thin slice that plitude variation with offset (Section 11.2) and acoustic
exhibits the stream channel disrupted by the faults. impedance (Section 11.3), by the petrophysical proper-
As indicated in Figure 10.9-3, the seismic pathway ties of the reservoir unit, such as, porosity, permeability
we followed from phases 1 through 5 in this case study and fluid saturation, all of which are allowed to vary
should lead us to constructing a model for the reservoir. laterally within each of the sublayers. In the final chap-
Begin with a detailed delineation of the top and base of ter, we shall review prestack and poststack amplitude
the deltaic sequence that represents the reservoir unit inversion methods to infer petrophysical properties of
shown in Figure 10.9-31a. Structural interpretation of reservoir rocks from seismic data as an aid to reservoir
the image volume derived from 3-D prestack depth mi- modeling.
gration (Figures 10.9-26 and 10.9-27) yields the top and
base surfaces shown in Figure 10.9-32. Then, define the
interior geometry of the reservoir by splitting the deltaic
sequence bounded by the two surfaces shown in Figure EXERCISES
10.9-32 into a set of thin layers as shown in Figure 10.9-
31b. Cross-sections of the reservoir unit along selected Exercise 10-1. Does a zero-offset wavefield mod-
inline traverses after sublayering are shown in Figure eled from exploding reflectors contain multiples?
10.9-33. Exercise 10-2. Consider two image sections de-
Next, extract the subvolume associated with the rived from (a) time migration, and (b) depth migration
reservoir unit from within the image volume derived followed by scaling from depth to time. To perform the
from 3-D prestack depth migration (Figure 10.9-34). depth-to-time scaling, you may use an interval veloc-
You may want to use the subvolume to estimate a set of ity field based on the rms velocity field to do the time
seismic attributes. The sublayers (Figure 10.9-33) then migration or the velocity-depth model to do the depth
are populated, based on any available well data, and migration. Which option would make the two image sec-
results of the analysis of seismic attributes, such as am- tions in time more compatible?
Appendix K
SEISMIC MODELING

K.1 Zero-Offset Traveltime Modeling

Seismic modeling essentially is a simulation of a recorded seismic wavefield, seismic amplitudes,


or seismic traveltimes. The input to seismic modeling is a representation of the earth’s reflectivity
and a velocity-depth model. Seismic migration is a process of estimating earth’s reflectivity from
a recorded seismic wavefield using a velocity-depth model. Therefore, seismic wavefield modeling
may be viewed as the reverse process of seismic migration. As such, both seismic migration and
seismic wavefield modeling algorithms are based on the wave equation.
Given a seismic wavefield P (x, z = 0, t) recorded over time t, at the surface z = 0, and
along the spatial axis x, seismic migration yields the earth’s reflectivity P (x, z, t = 0) based on
a process of wavefield extrapolation in depth z and collecting the image at time t = 0 (Section
4.1). Conversely, the fundamental ingredient of the modeling process is wavefield extrapolation
in time t and collecting the result at depth z = 0. Both processes use wave equation as the basis
for wave extrapolation.
Seismic modeling is different from data modeling (Appendix J). The latter involves, given
an observed data set d, an estimation of a set of parameters p that are used to construct a
model d of the observed data set d, such that the difference between the observed data set d
and the modeled data set d is minimum based on a specific mathematical norm.
Throughout this book, we have seen numerous examples of seismic modeling:

(a) to explain a process such as deconvolution (Figure 2.1-3) or migration (Figure 4.0-8),
(b) to test an algorithm such as predictive deconvolution (Figure 2.4-13) or implicit frequency-
space 3-D poststack time migration (Figure 8.4-2), or
(c) to understand a structural or stratigraphic phenomenon that may be of interest in explo-
ration.

Just as there are several approaches to solving the wave equation for migration, there also
are several types of modeling techniques. There are modeling techniques based on the Kirchhoff
integral (Hilterman, 1970), finite-difference (Kelly et al., 1976), and f − k domain (Sherwood
et al., 1983) solutions to the wave equation. The algorithms based on the scalar (acoustic)
wave equation (Section D.1), which describes P-wave propagation, are suitable for structural
modeling in which amplitudes are not as important as traveltimes. The algorithms based on
the elastic wave equation (Section L.2), which describes both P- and S-wave propagation, are
suitable for detailed stratigraphic modeling in which amplitudes are as important as traveltimes.
Modeling based on one-way wave equations does not include multiples, while modeling based
on two-way wave equations includes multiples in the simulated wavefields. In this appendix,
we shall not discuss the details of specific algorithms for seismic modeling which span a broad
range of applications including 2-D and 3-D, zero-offset and nonzero-offset, acoustic and elastic
simulation. Instead, we shall provide examples of the most common seismic modeling strategies
— zero-offset traveltime modeling, zero-offset and nonzero-offset acoustic wavefield modeling,
and elastic modeling.
Shown in Figure K-1a is a velocity-depth model for a salt diapir with an overhang structure.
The zero-offset traveltime response shown in Figure K-1b is created by normal-incidence ray
1780 Seismic Data Analysis

FIG. K-1. (a) A velocity-depth model of a salt dome with overhang, (b) the zero-offset traveltime response.
Seismic Modeling 1781

tracing. Note that the top-salt boundary has given rise to a complex and multivalued traveltime
trajectory. Note also that the base-salt boundary is flat and continuous in the velocity-depth
model (Figure K-1a), whereas the reflection traveltime follows a discontinuous and multivalued
trajectory (Figure K-1b).
Zero-offset traveltime modeling using normal-incidence rays is a very useful and trivially
simple tool for understanding the complexity of a reflection traveltime in field data. The dis-
ruptive behavior in traveltime trajectory associated with a layer boundary below a complex
overburden as in Figure K-1b is observed also in real data (Figure 10.1-1).

K.2 Zero-Offset Wavefield Modeling

Recall from Section 4.0 that a stacked section often is assumed to be a close representation of
a zero-offset wavefield. A modeled zero-offset wavefield therefore can be used to test poststack
migration algorithms. Zero-offset wavefields can be simulated very efficiently using the exploding
reflectors, also discussed in Section 4.0.
Wave-equation datuming (Berryhill, 1979), which was described in Section 8.1, can be used
to perform the simulation based on exploding reflectors. In particular, the datuming approach
can propagate a wavefield from one irregular interface to another. Consider zero-offset modeling
using the datuming technique of the velocity-depth model shown in Figure K-2a. Horizons 2
and 3 are the top and base of a salt dome. Start with the receivers situated along horizon 3.
The corresponding zero-offset section (Figure K-2b) contains the reflection from the bottom of
the velocity-depth model at z = 4000 m (not shown in Figure K-2a). Take this wavefield and
extrapolate it to a new datum, horizon 2, using the salt velocity (5000 m/s). The resulting
zero-offset section (Figure K-2c) contains the reflection (the deeper one) from the bottom of the
velocity-depth model (z = 4000 m) and the reflection (the shallow one) from the base of the
salt (horizon 3). Finally, extrapolate this wavefield (Figure K-2c) from horizon 2 to the surface
(horizon 1 at z = 0) using the overburden velocity (3000 m/s) to get the 2-D zero-offset section
in Figure K-2d. This section contains reflections from both the top and base of the salt. (The
reflection from the bottom of the model arrives after the latest time shown on this section.)
Note the velocity pull-up along the reflection from the base of the salt dome. Proper imaging of
the top of the salt dome can be achieved by time migration (Section 4.0), while proper imaging
of the base of the salt requires depth migration (Section 8.0).

K.3 Nonzero-Offset Wavefield Modeling

Understanding complexities of recorded wavefields clearly requires nonzero-offset wavefield mod-


eling. A finite-difference technique for modeling acoustic and elastic wavefields is described by
Kelly et al. (1976). Figure K-3 shows an example of acoustic modeling of a complex structure
associated with overthrust tectonics. A seismic line is simulated over a 2-D complex structure
(Figure K-3a). Selected common-shot (Figure K-3b) and CMP gathers (Figure K-3c) from this
simulation show the many complexities in the arrivals. Since this is a two-way acoustic simu-
lation, the modeled gathers contain not only primaries but also multiples. The zero-offset and
stacked sections associated with this nonzero-offset data are shown in Figure K-4. Note the broad
traveltime trajectories associated with the tight imbricate structures in the velocity-depth model
(Figure K-3a).
An example of a nonzero-offset modeling application of wave-equation datuming is provided
in Figure K-5. The shot gathers in Figure K-3b are computed to a flat datum level z = 0. Better
simulation of the actual field conditions requires that the gathers be computed using an irregular
topography. To do this, we can upward continue the shots and receivers to the new irregular
datum represented by the topography shown in Figure K-5a, then compute the shot gathers in
1782 Seismic Data Analysis

FIG. K-2. Wave-equation datuming (Section 8.1) used as a modeling tool. The zero-offset wavefield at
the surface (d) has been upward continued [(b) and (c)] through the velocity-depth model (a).
Seismic Modeling 1783

FIG. K-3. (a) A 2-D velocity-depth model, (b) full acoustic modeling of shot gathers, (c) selected CMP
gathers. The zero-offset and CMP-stacked sections are shown in Figure K-4. (Modeling courtesy Amoco
Production Company.)
1784 Seismic Data Analysis

FIG. K-4. (a) Zero-offset, (b) stacked sections from the nonzero-offset model data shown in Figure
K-3. (Modeling courtesy Amoco Production Company.)
Figure K-5b and sort them to the CMP gathers shown in Figure K-5c. Compare Figures K-3b
and K-3c with Figures K-5b and K-5c, and note the traveltime distortions.
Figure K-6 shows a velocity-depth model associated with a salt sill structure caused by
salt tectonics in the Gulf of Mexico. Note that velocity variations in some parts of the sedimen-
tary section are structure independent and represent overpressured zones. Selected common-shot
gathers shown in Figure K-7 have been created by two-way acoustic wavefield modeling (O’Brien
and Gray, 1996); therefore, they contain both primaries and multiples. Each shot gather repre-
sents a modeled wavefield. Note the complex events in the gathers above and in the vicinity of
the salt sill shown in Figure K-6.
Shown in Figure K-8 are selected CMP gathers sorted from the modeled shot gathers
as in Figure K-7. Observe the events with complex moveout in the gathers above and in the
(text continues on p. 1790)
Seismic Modeling 1785

FIG. K-5. Upward continuation using wave-equation datuming of shots and receivers from a flat datum
at z = 0 to an irregular topography shown above the velocity-depth model in (a). Selected input shot
gathers are shown in Figure K-3b. (b) Shot gathers and (c) CMP gathers along the irregular datum.
(Data courtesy Amoco Production Company.)
1786 Seismic Data Analysis

FIG. K-6. A velocity-depth model of a salt sill structure commonly encountered in the Gulf of Mexico
(O’Brien and Gray, 1996).

FIG. K-7. Selected modeled common-shot gathers using the velocity-depth model shown in Figure K-6
(O’Brien and Gray, 1996).
Seismic Modeling 1787

FIG. K-8. Part 1: Selected CMP gathers associated with the stacked section in Figure K-9b, the
modeled shot records as in Figure K-7, and the velocity-depth model shown in Figure K-6 (O’Brien and
Gray, 1996).
1788 Seismic Data Analysis

FIG. K-8. Part 2: Selected CMP gathers associated with the stacked section in Figure K-9b, the
modeled shot records as in Figure K-7, and the velocity-depth model shown in Figure K-6 (O’Brien and
Gray, 1996).
Seismic Modeling 1789

FIG. K-9. (a) Modeled zero-offset wavefield (O’Brien and Gray, 1996), and (b) stacked section derived
from CMP gathers as in Figure K-8 which are associated with the modeled shot records as in Figure
K-7 and the velocity-depth model shown in Figure K-6.
1790 Seismic Data Analysis

FIG. K-10. Elastic modeling of a water layer on top of an earth model represented by a vertically
varying velocity. The water depths are (from left to right) 5, 10, 15, 20, and 50 m. Identify multiples
(both reflected and refracted) and guided waves. The linear features below 3 s are artifacts of the
modeling program.

vicinity of the salt sill. The zero-offset section obtained by collecting the zero-offset traces from
the modeled shot gathers is shown in Figure K-9a, and the stacked section obtained from the
CMP gathers as in Figure K-8 is shown in Figure K-9b. A zero-offset wavefield simulated by
exploding reflectors does not include multiples, because the exploding reflectors are associated
with simulation based on one-way wave equation (Claerbout, 1985). When the simulation is
based on the two-way acoustic wave-equation as in Figure K-9, the zero-offset and stacked
sections both include primary and multiple reflections. Since it is wavefield modeling, not just
traveltime modeling, the simulated shot gathers (Figure K-7), the associated CMP gathers
(Figure K-8), and sections (Figure K-9) all contain the diffractions caused by the reflector
discontinuities in the velocity-depth model (Figure K-6).

K.4 Elastic Wavefield Modeling

Elastic wavefield modeling primarily is used to understand the effect of lithology and pore
fluids on seismic amplitudes (Sections 11.1 and 11.2). Sherwood et al. (1983) developed an f − k
method for nonzero-offset modeling of elastic waves in a 2-D horizontally layered medium. Figure
K-10 shows five shot gathers derived from an earth model represented by a vertically varying
velocity function that includes a water layer with five different thicknesses. The water depths
are 5, 10, 15, 20, and 50 m. Note the guided wave energy, which is especially prominent in
gathers corresponding to water depths of 5, 10, and 15 m. These gathers contain all primaries,
both P-waves and S-waves, as well as all possible multiples and converted modes. By examining
Seismic Modeling 1791

FIG. K-11. Elastic modeling examples: (a) the response of a clastic section with a low-velocity shallow
layer on the left and the response of the same clastic section with a fast-velocity shallow layer on
the right. What is the low-frequency low-group velocity dispersive energy? (b) Seismic response of an
all-shale section, (c) seismic response of a sand-shale section (see text for details; Sherwood et al., 1983).

such modeled data, we can better understand the nature of coherent noise (guided waves and
multiples) in both land and marine environments.
A more interpretive application of elastic modeling is shown in Figure K-11 (Sherwood et
al., 1983). The synthetic shot gather on the left in Figure K-11a is from a clastic section with a
shallow layer of low P-wave velocity. This layer has been replaced with a fast-velocity limestone
for the record on the right. Note the invasion of the large-offset primary reflection data with
coherent noise that is associated with this limestone layer. In the field, limestone on the surface
often generates a large amount of coherent noise.
Figures K-11b and K-11c show two synthetic shot gathers for which the upper part of the
depth model consists of a 50-ft water layer underlain by a 1020-ft shale section with a P-wave
velocity of 5500 ft/s. The deeper portion of the model for Figure K-11b is an all-shale section
with P-wave velocity increasing from 5600 ft/s on the top to 7700 ft/s on the bottom of the layer.
The primary reflections PP between 380 and 850 ms are associated with this all-shale sequence.
1792 Seismic Data Analysis

Figure K-11c shows the effects of including a 30 percent sand layer between 660 and 850 ms. The
P-wave reflections PP in Figure K-11c show stronger amplitudes at large offsets. An analysis of
amplitudes as a function of offset can provide hints for determining the sand-shale ratio, as well
as fluid content, in some cases. Because the effects are complex, this type of modeling can be
helpful in analyzing amplitude variations with offset. Also note the relatively strong converted
PS and SP waves on the record from the sand-shale model. Modeling of this type also is useful
when analyzing converted waves in multicomponent reflection data (Section 11.6).

REFERENCES

Berryhill, J., 1979, Wave-equation datuming: Geophysics, 44, 1329-1344.


Claerbout, J. F., 1985, Imaging the earth’s interior: Blackwell Scientific Publications.
Hilterman, F. J., 1970, Three-dimensional seismic modeling: Geophysics, 35, 1020-1037.
Kelly, K. R., Ward, R. W., Treitel, S. and Alford, R. M., 1976, Synthetic seismograms: A finite-
difference approach: Geophysics, 41, 2-27.
O’Brien, M. and Gray, S., 1996, Can we image beneath salt?: The Leading Edge, 17-22.
Sherwood, J. W. C., Hilterman, F. J., Neale, R. N. and Chen, K. C., 1983, Synthetic seismograms
with offset for a layered elastic medium; Offshore Technology Conference, Paper 4508.

You might also like