S E I S M I C A P P L I C A T I O N S

PRESTACK DEPTH MIGRATION WITH

TOMOGRAPHIC ANALYSIS: DEPTH
MODEL, CAUVERY BASIN

Downloaded from http://onepetro.org/JPT/article-pdf/50/05/32/2219509/spe-0598-0032-jpt.pdf/1 by Indian Institute of Technology (ISM) Dhanbad user on 01 April 2024
Nine additional appraisal wells were drilled reservoir section is characterized by a strong
during 1988–91. Two wells, PY-3-3 and PY-3- amplitude package (Fig. 2). However, the
4, tested hydrocarbons, and the rest were event defining the top of the true effective
abandoned as dry holes. This appraisal cam- reservoir appears to be composite in nature
paign, which used an assortment of two- and is affected by the presence of thin sand
dimensional (2D) -seismic lines, led to the stringers almost directly above.
acquisition of a three-dimensional (3D) -seis- The reservoir is sealed by the overlying
mic survey in 1990–91 to delineate the reser- Porto Novo shale that comprises a thick
voir distribution. sequence of calcareous mudstones and silts.
The trap is formed by dip closure to the
GEOLOGIC AL SETTING north and south, by a combination of dip
The PY3 field lies in the Cauvery basin with- and stratigraphic pinchout to the west, and
in the north-northeast trending Tranquebar dip and truncation to the east.
sub-basin. Developed in late Jurassic/early
Cretaceous, the basin straddles the present- SEISMIC INTERPRETATION
day east coast of India. The upper Six main seismic markers were picked dur-
Cretaceous reservoir comprises turbiditic, ing the interpretation.
debris-flow sand bodies that have been sub- 1. Intra-Eocene Event. This event forms a
divided vertically into five units [A (upper- prominent feature toward the south and
Fig. 1—Location map. most) through E]. Sand quality and distrib- east of the area where it appears as the junc-
ution are highly variable, with the best tion between an underlying seismically
reservoir characteristics observed on the opaque sequence and an overlying relative-
The PY3 field is in the Cauvery basin approxi- western side (core area) of the field. In the ly-high-amplitude, layered, on-lapping
mately 10 km off the coast of southeast India core area, porosities of 20% and permeabili- sequence. These sequences are interpreted
(Fig. 1). An accurate depth model has been ties of 30 to 40 md are recorded in the upper to represent a low-stand buildup feature
difficult to define in this field, which has led to sand units. Gross reservoir thickness aver- that was subsequently submerged beneath a
problems with well prognoses and reserve esti- ages approximately 70 m. Seismically, the transgressive (high-stand) sequence.
mations. Depth-conversion difficulties result
from rapidly varying overburden stratigraphy
because the field is beneath a series of stacked
shelf-edge sequences. This phenomenon caus-
es rapid, unpredictable lateral velocity changes
that lead to conversion of time “lows” to depth
“highs.” Discovery Well PY-3-2 was drilled by
Oil and Natural Gas Corp. (ONGC) in 1988
and produced approximately 3,000 B/D of
48°API oil plus 3.45 MMcf/D of associated gas.

This article is a synopsis of paper SPE

39499, “The Application of Prestack
Depth Migration Using Tomographic
Analysis To Aid in the Reduction of
Uncertainty in the Depth Model for the
PY3 Field, Cauvery Basin India,” by
N.K. Oates, Hardy E&P India Inc.;
P. Harinder, Tata Petrodyne; and
C. Willacy, GX Technology, originally
presented at the 1998 SPE India Oil
and Gas Conference and Exhibition,
Fig. 2—Representative seismic line (151) through the PY3 field.
New Delhi, India, 7–9 April.

32 MAY 1998 •
 S E I S M I C A P P L I C A T I O N S

2. Top-Eocene Marker. This event is 6. Near-Top Basement. This event was a range of values for structural elevation
defined as a strong trough at the top of a mapped to increase regional understanding and to test the drilling-location sensitivity
high-amplitude seismic package. It shows of the basin development. The event was to different depth-conversion treatments.
evidence of slump faulting related to deep- based on seismic character changes
seated structuring. Several good-quality between the predominantly crystalline Interval-Velocity Method. The seismic
sands are in this package, but no prospects Archean basement and the sedimentary sequence down to the reservoir top was
have been identified. infill. Although this event comprises fault- divided into four layers (water; seabed to top-
3. Near-Top-Cretaceous Unconformity. scarp zones, it was picked as a continuous Eocene marker; top-Eocene marker to near-
This event appears as a broad trough that horizon to ease mapping and modeling. top-Cretaceous unconformity; and near-top-
shows, in places (particularly the south and Because of the nature of the mute applied to Cretaceous unconformity to reservoir top).

Downloaded from http://onepetro.org/JPT/article-pdf/50/05/32/2219509/spe-0598-0032-jpt.pdf/1 by Indian Institute of Technology (ISM) Dhanbad user on 01 April 2024
east), marked truncations below and onlaps the 3D data set, it was not possible to pick a Interval velocities within each layer were
above. Toward the north and west of the seabed reflector. Water-depth information was mapped with a combination of acceleration
area, this event becomes parallel to subpar- restricted to well measurements and soundings factors and well-interpolation/-extrapola-
allel with adjacent events and, therefore, is taken during seismic acquisition. The two- tion techniques. A back-calculated average
more difficult to interpret. way-time structure map of the reservoir-top velocity (mean sea level to reservoir top)
4. Effective Reservoir Top. The true level indicates that the field exhibits no closure was created by use of the depth map and
reservoir top is difficult to define seismi- in the time domain. Therefore, the existence of the two-way times. This calculation
cally but, for mapping purposes, the any structural closure can be demonstrated enabled a direct comparison with the stack-
pick was made at a zero crossing (peak only by time-to-depth conversion. ing-velocity approach.
to trough) beneath a pronounced peak.
The reservoir interval itself is characterized DEPTH CONVERSION Stacking-Velocity Method. The stacking
by a single, broad, generally-high-ampli- Historically, depth conversion has been a velocities used for the original processing of
tude trough. major problem in the Cauvery basin and a the 3D data set were loaded as a velocity
5. Reservoir Base. This event is generally significant factor in well failure. The chal- cube. From this cube, an average velocity
defined as a moderate-to-strong peak, but lenge was to devise a depth-conversion from mean sea level to reservoir top was
occasionally becomes indistinct. However, technique that reflects the rapidly varying extracted. This initial field was extremely
it is possible to map this time horizon over nature of the overburden geology adequate- noisy; therefore, the raw field was
most of the study area, even where the ly. For this interpretation project, two smoothed with a 500-m Lowess filter and
reservoir section itself is absent. methods were applied in parallel to provide by manual point editing to remove isolated

34 MAY 1998 •
 S E I S M I C A P P L I C A T I O N S

anomalous values. This smoothed velocity SEISMIC REPROCESSING ing the crossline time migration. This proce-
grid was then calibrated to tie to well val- The original processed data were deter- dure reduced the size of the data set on the
ues. Calibration required gridding of the mined unsuitable for prestack depth migra- computer system and improved the
well values and use of the smoothed veloc- tion because of an abundance of low-ampli- signal/noise ratio of the data. During the par-
ity grid as an external drift. Additional tude random noise and the presence of tial stack, the offset planes were regularized to
manual smoothing was also applied to pro- high-amplitude noise bursts throughout the a 150-m increment, producing superoffset
duce the map. offset range. The data set was reprocessed to planes. The resultant data set had a stacked
The stacking-velocity technique indicated clean it up and to improve the signal/noise pseudofold of 144 and was used for the
higher relief across the central area of the ratio. A combination of channel suppres- remainder of the prestack work flow.
field but predicted steeper and deeper flanks sion and function-of-x deconvolution was Velocity analysis was accomplished with

Downloaded from http://onepetro.org/JPT/article-pdf/50/05/32/2219509/spe-0598-0032-jpt.pdf/1 by Indian Institute of Technology (ISM) Dhanbad user on 01 April 2024
than the interval-velocity method. Locations used to improve the signal quality and to iterations of travel-time tomography and
for development wells were chosen on the facilitate prestack-domain use of the data. depth focusing. The tomographic inversion
basis of commonality between the two depth Velocity-analysis lines were extracted from uses a common-reflection-point (CRP) -ori-
maps to minimize the risk of drilling a the 3D volume with a proprietary 3D ented ray-tracing technique rather than a
depth-conversion-induced anomaly. crossline-migration program. This program traditional common-midpoint (CMP) or
The joint-venture members felt that, compensates for out-of-plane events and repo- common-receiver method. The CRP tech-
while the well-derived velocity values sitions the 3D prestack data into a user-defined nique traces rays reflecting on a known
could be regarded as hard data, the same set of 2D lines suitable for detailed velocity interface at a known location that originate
was not true of the stacking-velocity data. analysis. The adopted approach involves bifur- from a source/receiver pair with a known off-
Consequently, a trial reprocessing exercise cating the migration operator into crossline set but an unknown CMP. The main advan-
was initiated on only four lines chosen to and inline components. The first workflow tages of CRP ray tracing are the ability to
give a reasonable spatial coverage. This stage (crossline migration) migrates the ener- provide regular sampling along the inter-
trial was designed to generate improved gy in the crossline direction and then is faces (more appropriate for depth inversion)
stacking-velocity data and to produce a applied to all input traces perpendicular to the and the ability to simplify the management
number of high-resolution seismic pro- inline plane. The second stage involves time or of multiple arrivals.
files. To achieve these objectives, the rec- depth migration that relocates the trace energy After tomographic inversion, a 2D Kirchoff
ommendation was that these lines be to its correctly migrated position. The original common-offset depth migration was per-
reprocessed by use of prestack depth- 48-fold data (50-m offset increment) acquired formed on each of the analysis lines.
migration software. over the PY3 field were partially stacked dur- Retuning of the interval-velocity depth

• MAY 1998 35
 S E I S M I C A P P L I C A T I O N S

model was carried out with a series of focus-

ing panels generated during the depth migra-
tion. These fully migrated focusing panels
were built by taking into account adjacent
common-data points within the migration
aperture. The migration-velocity grid is per-
turbed in small increments and decrements
of velocity, and the seismic traces are migrat-
ed and their coherency calculated for each
velocity perturbation. Adjustments to the

Downloaded from http://onepetro.org/JPT/article-pdf/50/05/32/2219509/spe-0598-0032-jpt.pdf/1 by Indian Institute of Technology (ISM) Dhanbad user on 01 April 2024
depth-migration-velocity grid are made by
picking the velocity that corresponds to the
maximum reflector coherency. Several focus-
ing and depth migrations are required before
the velocity model is fully optimized.
Velocity-depth picks were then merged and
interpolated into maps and, subsequently,
into an integrated full-volume, 3D velocity-
depth and structure model (Fig. 3).

RESULTS
The immediate impression from the
reprocessed data is cleaner, sharper defini-
tion compared with the original versions.
Improved resolution was particularly
Fig. 3—Three-dimensional structural model of the main picked depth surfaces.
noticeable at and around the reservoir level
and had a significant effect on the under-
standing of reservoir distribution and prob-
able small-scale faulting. These changes
added many details to the reservoir model.
While the original stacking-velocity field
is noisy (even after attempts to smooth it),
its overall trend is in broad agreement with
the trend derived during the tomographic
analysis. However, if the well-calibrated
velocity profile is compared with that of the
smoothed original stacking velocities, the
apex of the velocity profile on the original
(in this case a velocity low) has shifted lat-
erally some 2.5 km. This shift results from
the variable correction required to tie the
stacking-velocity field to the actual well-
derived values (Fig. 4).
Therefore, the calibration of stacking
velocities with actual well velocities has
resulted in a distortion of the original trend
information from the stacking velocities.
Additional work is required to investigate
the reasons for the wide variance between
stacking and well velocities. In addition,
the choice of geostatistical technique used
to calibrate stacking with well velocities
needs to be reviewed to maintain the maxi-
mum retention of trend information from
the stacking-velocity field.

Please read the full-length paper for

additional detail, illustrations, and ref-
erences. The paper from which the syn-
opsis has been taken has not been peer
Fig. 4—Map showing shifts required to tie stacking velocity to well values.
reviewed.

36 MAY 1998 •