shonbeck

SPECIAL
B r z a iSECTION:
l
B r a z i l
&KDOOHQJHVLQSUHVDOWGHSWKLPDJLQJRIWKH
GHHSZDWHU6DQWRV%DVLQ%UD]LO
YAN HUANG, DECHUN LIN, BING BAI, STAN ROBY, and CESAR RICARDEZ, CGGVeritas
S
everal discoveries, such as Tupi,
Bem-Te-Vi, Parati, and Guara,
have been announced in Santos Basin
off the coast of Brazil, mostly in presalt
layers. These layers were well imaged
by a salt-flood volume in 2003, but
distortions in the base of salt (BOS)
and presalt layers were still obvious.
Therefore, a constant velocity model
is not adequate to capture the velocity
variation inside the salt bodies, which
include mobile salt and evaporites. A
depth migration with a complete salt
model is necessary to correctly position
the reservoir structures. In this paper,
we discuss the challenges in building
such a velocity model and share the
lessons we learned while working on
a data set from the deepwater Santos
Basin.
In order to obtain high-quality
subsurface images, building an accurate velocity model and using the
optimal migration algorithm for the
geology are paramount. For presalt
imaging, the presalt velocities, over- Figure 1. (a) Stack section overlaid with the velocity model of a salt-flood migration without
burden velocities i.e., sediment veloci- Albian layer tomography. (b) CIG of a salt-flood migration without Albian layer tomography.
ties, salt velocities, and salt geometry Overcorrected events indicate faster velocities are needed. (c) Stack section overlaid with the
are critical. The unique aspects of local velocity model of a salt-flood migration with Albian layer tomography. (d) CIG with Albian
geology in Santos Basin made building layer tomography shows flatter events.
the velocity model challenging. Simply using the regular Gulf of Mexico (GOM) depth velocity Zhang and Sun, 2009). However, RTM is computationally
model building flow was inadequate (Siddiqui et al., 2003). intensive and the computing load increases with frequency.
We modified the flow to accommodate the three major differ- Despite this extra cost, high-frequency RTM is essential to
ences: (1) the presence of a thin Albian layer above the salt; (2) image presalt structures and stratigraphy in the deepwater
the sensitivity of the top of salt (TOS) picking to presalt imag- Santos Basin. Any attempt to output low-frequency images to
ing; and (3) the existence of evaporite layers within the salt. save computer time defeats the purpose of using RTM in the
Albian layer tomography, iterative TOS interpretation, and first place. Our study in the Tupi area showed that an abrupt
intrasalt tomography were introduced to the velocity model amplitude change appeared when applying low-frequency
building flow to improve the accuracy of the velocity model RTM. The amplitude break dissipated as higher frequencies
and the presalt image.
were migrated.
Kirchhoff migration can easily handle high-frequency
Further improvement of the presalt images may be posinformation and therefore allows high-resolution seismic im- sible by introducing anisotropy in the model building and miages. However, its single raypath limitation makes it difficult gration. A lack of publicly available well information in Santos
to image complex geology. In general, Santos Basin salt bodies Basin makes it difficult to determine the anisotropy level, but
have a fairly simple BOS, especially compared to the often ru- seismic data indicate that anisotropy does exist in this area.
gose TOS structure. Kirchhoff migration is incapable of imag- Ignoring the anisotropic effects may incorrectly position salt
ing presalt events under the rugose TOS. On the other hand, flanks and distort the BOS and presalt structures. Furtherthe more advanced imaging algorithm, reverse time migra- more, the dip angles of some deep basins can reach more than
tion (RTM), is, by nature, better suited to these complex ar- 50˚. With such high-dip bedding, tilted transverse isotropy
eas (Baysal et al., 1983; McMechan, 1983; Whitmore, 1983; (TTI) may significantly improve positioning of these events.
820
The Leading Edge
July 2010
Downloaded 24 May 2012 to 216.52.185.73. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
B r a z i l
Figure 2. (a) TOS is deeper. The BOS and presalt events have some
undulation that mirrors the TOS shape. (b) TOS moved slightly
shallower. The BOS and the presalt events are flatter.
Figure 3. (a) TOS is shallower. The BOS and presalt events have a
sag. (b) TOS is slightly deeper. The sag is reduced and the BOS and
presalt events become straighter and better focused.
In the following sections, the prestack depth migration
(PSDM) velocity model building flow for deepwater Santos
Basin data is presented with the three key steps in the model
building flow highlighted: Albian layer tomography, iterative
TOS interpretation, and intrasalt tomography. Then, Kirchhoff migration and RTM comparisons illustrate the benefits
of the latter. Finally, a TTI PSDM test demonstrates the benefit of TTI.
GOM
Santos Basin
Water-bottom determination
Water-bottom determination
Sediment tomography
Sediment tomography
Velocity model building flow in Santos Basin
A top-down process was used to build the velocity model
for the Santos Basin. Table 1 compares the PSDM velocity
model building flow for deepwater Santos Basin data with
the regular GOM flow (Siddiqui et al., 2003). The first step
is to determine the water-bottom surface. The water depth in
this study area is approximately 2 km. A range of water velocities was tested through water-flood migrations, and 1495
m/s was chosen based on the flatness of the water-bottom
BOS interpretation
BOS interpretation
Subsalt velocity update
Subsalt velocity update
Albian layer tomography
TOS interpretation
Iterative TOS interpretation
Salt flood
Salt flood
Salt layer tomography
Table 1. Comparison of GOM and Santos Basin velocity model
building flows.
events in the common image gathers (CIGs). The water bottom was then interpreted on a water flood migration volume,
resulting in less than 3-m mis-ties when compared with well
measurements.
July 2010 The Leading Edge
Downloaded 24 May 2012 to 216.52.185.73. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
821
B r a z i l
Figure 5. (a) Image from the 2003 velocity model shows a broken
BOS and unfocused presalt events. (b) Image from the new velocity
model. The BOS and presalt events are more continuous and better
focused. The BOS structure is less undulating. Also, the events inside
the salt layer are more accurately imaged.
Figure 4. (a) Velocity model after intrasalt tomography. (b) Gather
with constant salt velocity. (c) Gather with intrasalt tomography.
Events are flatter.
Tomographic inversion (Zhou et al., 2003) was then used
to derive the velocity of the sediment region. A sediment
velocity model which was derived in 2003 by vertical updates was used as the initial velocity model for PSDM. Event
curvatures picked from a grid of CIGs and event dip angles
measured on stack sections were fed into the tomographic
inversion to update the velocity model. A relatively smooth
sedimentary velocity field was obtained after several iterations
of the tomographic updates.
In Santos Basin, an Albian layer is directly above the Aptian salt (Modica and Brush, 2004; Rosenfeld and Hood,
2006). Inside the study area, the Albian layer is thin and its
velocity gradient is higher than the normal sediment compaction trend. The global sediment tomographic velocity update
cannot resolve it for several reasons. First, during the sediment velocity update, a horizon 150 m above the estimated
TOS horizon was used as a mask to constrain the update
within the sediment layer and eliminate TOS stretch in the
CIGs. The thin Albian layer could be masked out in some
places during the sediment tomography. Second, although
the top of the Albian layer is a bright seismic reflector, other
822
The Leading Edge
events in the layer have relatively weak amplitudes compared
to the sediment beddings. Moreover, only a few events are in
the layer due to its small thickness. If optimal event picking
parameters are chosen for the sediment layers, most Albian
layer events would not be picked during curvature picking for
the tomographic update. Third, the Albian layer has a higher
velocity gradient relative to the sediment compaction trend,
and it requires larger velocity perturbations than permitted
by global sediment velocity update thresholds (normally
5–10%). These difficulties had to be addressed before an accurate interpretation of the complicated TOS structure was
feasible.
To resolve these three problems, a layer-constrained tomographic velocity update between the top of Albian (TOA)
and TOS was performed. First, the TOA and TOS were
picked on the sediment velocity flood migration. Only the
region between the TOA and TOS was updated. Second, the
event-picking parameters were optimized specifically to handle the weak amplitude in the Albian layer. Last, the maximum velocity update limit was opened up to 20%. Figure
1a shows a stack section overlaid with the velocity model of
a salt-flood migration without the Albian layer tomography.
The Albian layer thickness at this line is less than 400 m, and
only 2–3 weak events are inside. The CIG (Figure 1b) shows
July 2010
Downloaded 24 May 2012 to 216.52.185.73. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
B r a z i l
Figure 6. (a) Kirchhoff shows relatively consistent amplitudes in the
BOS and presalt events, with slight amplitude shadows. (b) 30-Hz
RTM. The BOS and presalt events on the left are much weaker than
those on the right with an amplitude break in the middle near the
Tupi well location. (c) 40-Hz RTM. The amplitudes of the BOS
and presalt events are comparable to the Kirchhoff result with a slight
improvement. (d) 60-Hz RTM shows better amplitudes at the BOS
and presalt events compared to the Kirchhoff result.
overcorrected curvatures, which indicate a velocity speedup is
needed. The results with the Albian layer update are shown in
Figures 1c and 1d. The CIG in Figure 1d shows flatter events
in the Albian layer and the salt layer. This indicates a more
accurate velocity model was obtained after the Albian layer
tomography.
The next step was to interpret the TOS surface. Unlike
the top of salt in the GOM, most TOS events in this area are
buried in a sequence of reflectors, making it difficult to pick
the correct one. Our study demonstrates that the images of
the BOS and presalt events are very sensitive to the position of
the TOS. Figure 2 shows a salt-flood section with two slightly
different TOS interpretations. The BOS and presalt events in
Figure 2a have some undulation that mirrors the TOS shape.
Figure 2b shows a shallower TOS interpretation in two minibasins resulting in straighter BOS and presalt events. Another
example is shown in Figure 3a, where the BOS and presalt
events are sagging. Picking TOS slightly deeper (Figure 3b)
reduces the sag and makes the BOS straighter and better focused. Due to the sensitivity of the presalt image to the TOS
interpretation, the TOS horizon could not be finalized in a
single step. Iterative TOS interpretation was necessary to resolve this ambiguity. The BOS and presalt images were used
Figure 7. (a) Kirchhoff migration barely images the BOS and
presalt events under the rugose TOS. (b) RTM migration gives more
continuous BOS and presalt events under the rugose TOS.
to evaluate the necessity of modifying the TOS interpretation. With every modification to the TOS, another salt-flood
migration was needed to confirm the change.
After finalizing the TOS interpretation, a salt-flood migration with a constant salt velocity of 4530 m/s was performed. In the regular PSDM velocity model building flow,
this volume would be used for BOS interpretation and the
velocity in the salt layer would remain constant in the final
velocity model. Due to a lack of reflectors inside the salt, the
flatness of the BOS is the only constraint for selecting the
salt velocity. However, in the Santos Basin area, the presence
of layered evaporites within the salt body provides an opportunity for updating the salt-layer velocity with tomography,
since the layered evaporites have strong reflections. An intrasalt tomography step was introduced after the constant salt
velocity flood to improve the velocity model in the vicinity
of the layered evaporites. By carefully picking the residual
curvatures of those events, a better velocity field inside the
salt (evaporites) layer was obtained through tomographic inversion. The updated velocity model is shown in Figure 4a
July 2010 The Leading Edge
Downloaded 24 May 2012 to 216.52.185.73. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
823
B r a z i l
the BOS and presalt events are more continuous and better focused. The BOS structure is less undulating. Also, the
events inside the salt layer are more accurately imaged.
Figure 8. (a) Isotropic migration. The BOS is pushed down below the
center of the minibasin where the salt body is thin. (b) TTI migration.
The BOS is flatter and more continuous. Also, the minibasin becomes
narrower and shallower.
with the stack section overlaid. The CIG with the salt layer
tomography update (Figure 4c) is flatter than the one with
the constant salt velocity (Figure 4b).
Compared to the TOS, the BOS was relatively simple to
interpret in the study area. The main criterion for evaluating
the correctness of the velocity field above the BOS was the
assumption that the shape of the BOS should not be geologically complex. Below the BOS, we observed a faster velocity
trend, which was updated through tomographic inversion.
This PSDM velocity model building flow produced a
better velocity field and allowed better presalt images to be
obtained. Figure 5 compares a migration using the 2003
velocity model and the new model produced by the PSDM
flow presented in this paper. The 2003 velocity model was
obtained by several iterations of vertical updates in the sediment section and was flooded with a constant salt velocity of
4600 m/s. The image from the 2003 velocity model (Figure
5a) shows a broken BOS and unfocused presalt events. In
Figure 5b, the migration image with the new velocity model,
824
The Leading Edge
Utilizing reverse time migration (RTM)
RTM migrates data directly using the two-way wave equation. Advantages include preserving true amplitude and
more accurate handling of complex structures without dip
angle limitations (Zhang and Sun, 2009). However, RTM is
computationally intensive, and the computing load increases
with frequency. The study in the Tupi area showed that some
BOS and presalt events contain high-frequency information
which cannot be correctly imaged with low-frequency RTM.
High frequencies are essential for correctly imaging presalt
structures and stratigraphy, so any attempt at low-frequency
RTM to save computer time defeats the purpose of using
RTM. Figure 6a shows the Kirchhoff migration result. Amplitudes on the BOS and presalt events are relatively consistent, but display some amplitude shadows. Figure 6b shows
a 30-Hz RTM result. The BOS and presalt events on the left
are much weaker than the one on the right, with an obvious
amplitude break in the middle near the Tupi well location.
With the 40-Hz RTM (Figure 6c), the amplitude at the BOS
is comparable to the Kirchhoff result with a slight improvement. Increasing the frequency cutoff to 60 Hz (Figure 6d)
further improves the amplitude and resolution in the presalt
image.
Figure 7 shows another comparison of Kirchhoff migration (Figure 7a) and RTM (Figure 7b) in both inline and
crossline directions. In this section, the TOS rapidly changes causing extremely poor imaging of the BOS and presalt
events in the Kirchhoff result (marked with a circle). In the
RTM result, the blank presalt zone below the rugose TOS
where the Kirchhoff result failed is fairly well imaged. The
BOS and presalt events are more continuous throughout the
section and can be easily interpreted in the RTM image.
Benefits of TTI imaging
Huang (2008) proposed a TTI velocity model building flow
that improved salt-flank images in the GOM. A similar
TTI PSDM test was conducted in Santos Basin and produced some promising results in salt-flank positioning and
the BOS images. In this test, constant values for ¡ and b (¡
= 5.1% and b = 3%) were used in the sedimentary area. The
symmetry axis was assumed to be perpendicular to the sedimentary bedding, so it was spatially variant. V0 was updated
with multiple iterations of 3D TTI tomographic inversion.
After the TTI sedimentary velocity updates, the TOS was
interpreted. Below the TOS, the velocity volume was flooded
with a constant salt velocity. Migrations with the isotropic
and TTI models are compared in Figure 8. For comparison
purposes, a constant salt flood model was used. The BOS
from the isotropic migration (Figure 8a) is pushed down in
the region below the center of the basin where the salt body is
thin. The TTI migration (Figure 8b) makes the BOS flatter
and more continuous. Also the minibasin becomes narrower
and shallower. Our test indicates that a TTI velocity field is
July 2010
Downloaded 24 May 2012 to 216.52.185.73. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
B r a z i l
required for maximum image quality.
Conclusions
In this paper, we presented a PSDM velocity model building flow in the deepwater Santos Basin and highlighted the
three key steps in the model building flow: Albian layer tomography, iterative TOS interpretation, and intrasalt tomography. This flow produced a more accurate velocity model,
and allowed us to obtain better presalt images through the
advanced imaging algorithm RTM. High-frequency RTM is
essential to image presalt structures and stratigraphy. Compared to Kirchhoff migration, a significant presalt image improvement under the rugose TOS was obtained by RTM. An
initial TTI PSDM test showed that anisotropic imaging may
be important in this area.
References
Baysal, E., D. D. Kosloff, and J. W. C. Sherwood, 1983, Reverse time
migration: Geophysics, 48, 1514–1524.
Etgen, J., S. Gray, and Y. Zhang, 2009, An overview of depth imaging in exploration geophysics: Geophysics, 74, no. 6, WCA5–
WCA17.
Huang, T, S. Xu, J. Wang, G. Ionescu, and M. Richardson, 2008,
The benefit of TTI tomography for dual azimuth data in the Gulf
of Mexico: 78th Annual International Meeting, SEG, Expanded
Abstracts 27, 222–226.
McMechan, G. A., 1983, Migration by extrapolation of time-dependent boundary values: Geophysical Prospecting, 31, 413–420.
Modica, C. and E. Brush, 2004, Postrift sequence stratigraphy, paleogeography, and fill history of the deep-water Santos Basin, offshore
southeast Brazil: AAPG Bulletin, 88, 923–945.
Rosenfeld, J. H. and J. Hood, 2006, Play potential in the deepwater
Santos basin, Brazil: Offshore, 66, no 9.
Siddiqui, K., S. Clark, D. Epili, N. Chazalnoel, and L. Anderson,
2003, Velocity model building methodology and PSDM in deepwater Gulf of Mexico: A case history: 73rd Annual International
Meeting, SEG, Expanded Abstracts, 22, 442–445.
Whitmore, D., 1983, Iterative depth migration by backward time
propagation: 53rd Annual International Meeting, SEG, Expanded
Abstracts, 382–385.
Zhou, H., S. Gray, J. Young, D. Pham, and Y. Zhang, 2003, Tomographic residual curvature analysis: The process and its components: 73rd Annual International Meeting, SEG, Expanded Abstracts 22, 666–669.
Zhang, Y., and J. Sun, 2009, Practical issues in reverse time migration:
true amplitude gathers, noise removal and harmonic source encoding: First Break, 26, 29–35.
Acknowledgments: We thank CGGVeritas for permission to publish
this paper; Joseph Cole, Scott Shonbeck, and Sheng Xu for reviewing the paper; and Jerry Young for constructive suggestions and
discussions.
Corresponding author: [email protected]
July 2010 The Leading Edge
Downloaded 24 May 2012 to 216.52.185.73. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
825