|
|
Figure 1
Text
The Diana field is situated in
the western Gulf of Mexico 260 km (160 mi) south of Galveston in
approximately 1430 m (4700 ft) of water (Figure
1). ExxonMobil is the operator with 66% interest, whereas British
Petroleum (BP) holds a 33% interest. Diana is the second largest of
several discoveries recently made in the Diana Subbasin and has in
excess of 100 MMBOE of recoverable hydrocarbons from the upper Pliocene
A-50  reservoir . The turbiditic sandstones and mudstones that comprise
the A-50 reservoir at the Diana field were deposited as a lowstand fan
in an intraslope basin setting. The field is located on the east flank
of a north-south trending
salt-cored ridge. Hydrocarbons are trapped by a combination of structure
and stratigraphic onlap, with a large gas cap (>305 m [1000 ft] column
height) and a relatively thin oil column (approximately 75 m [240 ft]
column height). The Diana Subbasin is relatively large, consisting of
two narrow feeder corridors to the north, which open into a large
low-relief basin approximately 32 km (20 mi) wide by 32 km (20 mi) long.
It is about three to four times the size of the next largest updip
intraslope basin.
The challenge at the Diana field
was to predict the production performance of a channelized deep-water
reservoir with a relatively thin oil rim and a large gas cap (Figure
1). Associated development costs are high, requiring an optimization
program to ensure a successful project. These predictions were
challenged further by variable-quality seismic data, a reservoir
thickness expressed by a single-cycle seismic event, only limited
appraisal wells, and the likelihood for subseismic reservoir variability
that could control the economic viability of the project. To assist with
reserve assessments and optimization of depletion strategies, deep-water
outcrop analog data were integrated with seismic and well data to
produce a detailed object-based model for more accurate reservoir
characterization.
In the Diana study, two outcrop
analogs were found to be most applicable to the penetrated subsurface
reservoirs based on similarities in grain size, facies associations, and
interpreted sand-body architecture. These were the Lower Permian
Skoorsteenberg Formation in the Tanqua Karoo Basin, South Africa, and
the Upper Carboniferous Ross Formation in the Clare Basin, western
Ireland. These deep-water turbidite successions have been studied widely
in recent years by Collinson et al. (1991), Bouma and Wickens (1994),
Chapin et al. (1994), Sullivan et al.
(1998), Bouma (2000), Elliot (2000), Martinsen et al. (2000), Morris et
al. (2000), and Sullivan et al.
(2000a, b). The main purpose of this current outcrop study was to
provide the data necessary to help assess future prospects and newly
discovered fields with analogous reservoir characteristics.
To better understand and apply
the observations and learnings from this outcrop analog study,
normal-incidence forward seismic models were constructed for both the
Skoorsteenberg and Ross Formations. These models illustrate both the
seismic facies of the individual outcrops using Diana subsurface rock
properties (density and velocity) and the resolution limits of typical
seismic data.
By combining dimensional and
architectural data from outcrops with seismic, well-log, and core data
from the subsurface, it is possible to construct more accurate reservoir
models for deep-water turbidite sandstones for Diana and other fields.
Such studies are important because they greatly reduce the uncertainties
associated with reservoir assessment parameters for economically
important deep-water turbidite sandstones.
Figures 2-7
|
 |
FIGURE 2. The
deep-water deposits of the Skoorsteenberg Formation in Tanqua
Karoo Basin, South Africa, can be subdivided into proximal,
medial, and distal fan settings, each with their own key
characteristics. Note the outline of the Diana Subbasin and
Diana field projected on to outline of fan 3 of the
Skoorsteenberg Formation. The outline of fan 3 is modified from
Bouma and Wickens, 1994. |
|
 |
FIGURE
3. Proximal fan deposits of the Skoorsteenberg Formation are
comprised of erosionally based, narrow (low aspect ratio)
channels and interchannel sheets. Ten major channels are
identified, which range from 300 to 500 m (984 to 1630-ft) wide
and are typically 5-10-m (16-32-ft) thick. Channels are filled
from axes to margins by amalgamated, thick-bedded massive
sandstones. The interchannel areas are characterized by
nonamalgamated thin- to thick-bedded current-ripple-laminated
sandstones and interbedded laminated silty mudstones (location:
Ongeluks River) (modified from Sullivan et al., 2000a). |
|
 |
FIGURE
4. The proximal fan channels in the Ross Formation are
dominated by highly amalgamated massive to cross-bedded
sandstones in an axial position that are replaced by
progressively thinner-bedded, less-amalgamated massive
sandstones toward the margins. Note the distinct inclined
surfaces present in the east portion of the channel. These
inclined surfaces are interpreted to represent channel-margin
migration and suggest that this channel is somewhat sinuous.
(location: Rinevilla Point) (modified from Sullivan et al.,
2000b). |
|
 |
FIGURE
5. The transition from proximal to medial fan settings is
dominated by compensationally stacked, very broad (high aspect
ratio) channels as much as several thousand feet wide and 8-13 m
(26-42 ft) thick for both the Skoorsteenberg and Ross
Formations. These channelized sand bodies can be subdivided
further into distinct channel-axis and channel-margin
associations. Highly amalgamated massive sandstones characterize
channel-axis deposits. Away from the axis, beds become
distinctly less amalgamated and extremely continuous to produce
laterally extensive, layered wings at the channel margins
(location A: Rehy Cliffs; location B: Loskop) (modified from
Sullivan et al., 2000b). |
|
 |
FIGURE
6. Overall this is a very layered reservoir with most of the
amalgamation of these broad channels occurring where erosional
axes cut into underlying sand bodies. Bed-length data based on
detailed bed-by-bed correlations show these sandstones have
lengths much greater than the length of the outcrop. In
contrast, 80% of all mudstones are less than 0.3 m (1 ft) thick
and have bed lengths less than 200 m (655 ft). Although this is
a relatively high net-to-gross reservoir type with excellent
lateral continuity, the vertical continuity would be moderate to
low. Bed continuity, connectivity, and vertical and lateral
facies variability data such as this were collected from a
variety of channel types to condition the object-based geologic
model for the Diana field (location: Kilbaha Bay) (modified from
Sullivan et al., 1998). |
|
 |
FIGURE 7. Medial
fan deposits also are similar for both the Skoorsteenberg and
Ross outcrops and are comprised of extremely broad channels or
what can be more correctly termed as sheets. Individual sheets
are 3-7 m (10-23 ft) thick with narrow, amalgamated axes and
more sheetlike, layered margins. Although the sandstone bed sets
are highly continuous, they consist of individual lenticular and
compensatory sandstones (location A: West Kilcholer Cliffs;
location B: Loophead; location C: Grootfontein) (modified from
Sullivan et al., 2000b). |
Text
Outcrop analogs span a critical gap in both scale and resolution between
seismic and wellbore data. The integration of appropriate outcrop
analogs, core, well-log, and seismic data can provide the detailed
geometric properties required for interpreting the reservoir
architecture at a subseismic or flow-unit scale. The Lower Permian
Skoorsteenberg Formation in Tanqua Karoo Basin, South Africa, and the
Upper Carboniferous Ross Formation in the Clare Basin, western Ireland,
are both composed of stacked turbiditic sandstones and mudstones
deposited in a channelized basin-floor fan setting. These laterally
continuous outcrops provide an excellent opportunity to characterize
detailed bed-scale reservoir architectures and internal heterogeneities
that affect the producibility of deep-water sand bodies in both
depositional strike and dip perspectives. The selection of appropriate
outcrop analogs, however, is extremely important. The criteria for
choosing the outcrops of the Skoorsteenberg and Ross Formations as
analogs for A-50 reservoir at Diana was based on comparison of key
reservoir characteristics such as grain size, lithofacies, net-to-gross
(ratio of sandstone vs. mudstone), and sand-body architecture. Because
of the strong similarities between the selected outcrop analogs and the
A-50 reservoir, dimensional and architectural data from these outcrops
can be used to help constrain an object-based three-dimensional (3-D)
geologic model to predict well performance, connected volumes, and
recovery efficiencies for the Diana development.
Based on detailed characterization, the deep-water sandstones present in
the outcrop localities can be divided into proximal, transition from
proximal to medial, and medial fan settings (Figure
2). Distal fan deposits are also present but were not a focus of
this study because of their limited reservoir potential and relatively
poor exposure. The presented proximal-to-distal subdivision is for an
idealized slope-to-basin transition. It is recognized, however, that it
is the change in the slope gradient that ultimately controls the degree
of channelization (Imran et al., 1998). Therefore, appropriate outcrop
analogs for subsurface data sets need to be selected based on
similarities in interpreted sand-body architecture and not on
interpreted similarities in location in a slope-to-basin profile.
The
most proximal exposures of both the Skoorsteenberg and Ross Formations
are dominated by compensationally stacked, erosionally confined channels
and interchannel sheets. These narrow proximal fan channels are
typically less than 400 m (1300 ft) wide and 5-12
m (16-39
ft) thick, with aspect ratios (width vs. thickness) ranging from 30:1 to
80:1 (Figures 3, 4).
Net-to-gross ratios for individual measured sections range from 70 to
95%, with an average of approximately 90%. Two distinct styles of
channel fills are recognized. The proximal fan channels of the
Skoorsteenberg Formation are typically filled from axes to margins by
amalgamated, thick-bedded (>30
cm), fine- to medium-grained, massive sandstones (Figure
3). Massive sandstones commonly grade upward into
thick-bedded, fine-grained, planar-stratified sandstones and rare
thin-bedded (<30
cm), very fine- to fine-grained, current-ripple-stratified
sandstones. The interchannel strata are comprised of nonamalgamated
thin- to thick-bedded current-ripple-laminated
sandstones and interbedded laminated silty mudstones. In contrast, the
proximal fan channels in the Ross Formation exhibit a lateral
degradation in reservoir quality. They are dominated by highly
amalgamated massive to cross-bedded sandstones in an axial position that
grade laterally into progressively thinner-bedded, less-amalgamated
massive sandstones toward the margins (Figure 4).
This
difference in the lateral degree of amalgamation, from axis to marginal,
for the proximal fan channels of the Skoorsteenberg and Ross Formations
may suggest differences in the scale and size of the turbidity currents
that deposited the sandstones. The vertically and laterally amalgamated
massive sandstones, which comprise the channel fills in the
Skoorsteenberg Formation, are interpreted to have been deposited by
high-concentration turbidity currents that completely filled the
channels and, therefore, display no variations from axis to margin (Figure
3). Overbanking of these turbidity currents is interpreted to have
produced the distinct interchannel association dominated by
low-concentration turbidites (planar- and current-ripple-stratified
sandstones). By contrast, the distinct axis-to-margin variations in the
Ross Formation suggest that the turbidity currents that deposited these
sandstones were underfit relative to the channels (Figure
4). This conclusion is also supported by the general lack of
low-concentration turbidite-dominated interchannel deposits in the Ross
Formation.
Dominating the transition from proximal to medial fan settings for both
the Skoorsteenberg and Ross Formations are compensationally stacked,
very broad (high aspect ratio) weakly confined channels. These channels
are as much as 1000 m (3270 ft) wide and 8-13 m (26-42 ft) thick (Figures
5, 6). Their bases tend to be
nonerosional, suggesting that they are primarily aggradational in
origin. In general, these channels do not infill erosional scours;
instead, they are compensationally stacked because of preexisting highs
related to underlying channels. Individual channelized sand bodies can
be further subdivided into distinct channel-axis and channel-margin
facies associations. Highly amalgamated, massive sandstones characterize
channel-axis deposits. Away from the axis, beds become distinctly less
amalgamated and extremely continuous to produce laterally extensive,
layered wings at the channel margins (Figure 5).
Net-to-gross ratios for these weakly confined to unconfined channels
range from 70 to 90%, with an average of 80%. Detailed bed-by-bed
correlations show that these sandstones have bed lengths much greater
than the dimensions of the outcrop, whereas mudstones have much shorter
bed lengths (Figure 6). Therefore, the key
to understanding reservoir continuity and internal heterogeneities that
affect reservoir performance in this instance is knowing the thickness
and lengths of mudstone barriers and not the distribution of sandstones.
Eighty percent of these mudstones are less than 0.3 m (1 ft) thick and
have bed lengths less than 200 m (655 ft). Although this is a relatively
high net-to-gross reservoir type with excellent lateral continuity, the
vertical continuity would be moderate to low. This is caused by the
preserved interbedded mudstones, although only 5-10% of these mudstones
have lengths approaching or greater than 700 m (2295 ft). It is these
continuous mudstones that would likely produce significant vertical
barriers to fluid flow.
Medial fan deposits are also similar for both the Skoorsteenberg and
Ross outcrops and are comprised of extremely broad, unconfined channels
or sheets (Figure 7). The bases of these
sand-prone sheets tend to be nonerosional, comparable to the broad
channels of the proximal to medial fan transition. They also appear to
be compensationally stacked or laterally offset because of depositional
highs related to underlying sand bodies. Individual sheets are 3-7 m
(10-23 ft) thick, with narrow, amalgamated axes and more sheetlike,
layered margins. The sheet axes are locally erosionally based and are
typically 100-160 m (328-525 ft) wide. They are comprised of highly
amalgamated massive sandstones. Away from the axes, progressively
thinner-bedded, less-amalgamated sandstones replace amalgamated massive
sandstones (Figure 7). The sheet margins
have estimated widths in excess of 1600 m (5245 ft) on either side of
the axes. Massive sandstones represent the dominant facies type and are
interpreted to have been deposited rapidly from suspension from
high-concentration turbidity currents in an unconfined setting. Although
the sandstone bed sets are highly continuous, they consist of individual
lenticular and compensatory sandstone packages. Overall, this produces a
very layered architecture, with most of the amalgamation of sheet
complexes occurring where axes locally cut into underlying sand bodies.
Net-to-gross values for these deposits typically range from 65 to 80%,
with an average of approximately 70%.
Figures 8-10
|
 |
FIGURE 8. Forward seismic model
for medial fan sheets from the Ross Formation built using GXII
seismic modeling software. (A) Outcrop photo of the layered
sheets, which dominate the medial fan deposits of the Ross
Formation at Loophead (Figure
7). (B) Depth model constructed for
this outcrop by applying subsurface rock properties (density and
velocity) from deep-water reservoirs in the Gulf of Mexico to
the digitized interpretation of this outcrop. Also shown in
overlay is a detailed measured section for this outcrop. (C) The
high net sandstone section at the top of the outcrop can only be
fully resolved on the seismic model generated using a 60-Hz
wavelet. At this higher frequency, this interval is also
acoustically transparent because of the high percentage of
sandstone and resulting lack of internal reflectivity. (D) On
the forward seismic model generated using a 30-Hz frequency
wavelet, the upper sandstone-prone interval has a high-amplitude
continuous seismic response, whereas the lower package has a
low-amplitude continuous response (modified from Sullivan et
al., 2000b). |
|
 |
FIGURE 9. Forward seismic model
for the proximal- to medial-fan transition channels of the Ross
Formation built using GXII seismic modeling software. (A)
Outcrop photo of the broad, compensationally stacked channels
that dominate the transition from proximal to medial fan setting
at Rehy Cliffs (Figure
5). These channels are highly
amalgamated in an axial position but become less amalgamated and
lower net-to-gross toward the margins. (B) Depth model
constructed for this outcrop by applying subsurface rock
properties (density and velocity) from deep-water reservoirs in
the Gulf of Mexico to the digitized interpretation of this
outcrop. The depth model was extended approximately 1 km beyond
the outcrop control by continuing the observed patterns of
deposition. (C and D) The measured net-to-gross for various
points along the model are plotted across the top and can be
used to locate seismic models relative to the depth model.
Forward seismic models for 30- and 45-Hz peak frequencies are
shown. The seismic response for both models is very continuous,
but the net-to-gross variations from axis to margin are directly
related to the amplitude of the seismic response. This indicates
that these distinct lateral changes in sand percentage should be
seismically detectable. The individual channels can only be
resolved on the higher resolution (45 Hz) seismic model, but
there is still substantial interference between the stacked
channels (modified from Sullivan et al., 2000b). |
|
 |
FIGURE 10. Forward seismic model
for proximal fan channels from the Skoorsteenberg Formation,
Tanqua Karoo Basin, South Africa (Figure
3), built using GXII seismic
modeling software. (A) Outcrop photo of the compensationally
stacked, erosionally based, narrow (low aspect ratio) channels
and interchannel sheets that dominate the most proximal
exposures of the Skoorsteenberg Formation at Ongeluks River. (B)
Depth model constructed for this outcrop by applying subsurface
rock properties (density and velocity) from deep-water
reservoirs in the Gulf of Mexico to the digitized interpretation
of this outcrop and assuming that the reservoirs are encased by
mudstones. (C) The seismic response for a 30-Hz wavelet images
these intercutting channels as a single-cycle event. Amplitude,
isochron, and dip variations on the base reservoir can yield
clues to the net sand and channel locations, as the brightest
amplitudes and thickest isochron correspond to the most
channelized and highest net-to-gross portion of the outcrop. The
distribution of interreservoir mudstones, however, is not
resolved by the forward seismic model (modified from Sullivan et
al., 2000b). |
Text
The major uncertainties
associated with exploration and development of deep-water reservoirs are
predrill predictions of net-to-gross and assessment of reservoir
continuity and net-to-gross away from well penetrations. Based on recent
drilling results for deep-water petroleum reservoirs, successfully
estimating reservoir continuity and net-to-gross away from well
penetrations requires correct interpretation of reservoir type.
ExxonMobil's postdrill analyses in several deep-water basins have shown
that the information required to successfully predict these parameters
is often embedded in the seismic response of reservoirs. The challenge
for seismic analysis, therefore, is the proper interpretation of these
seismic responses.
The paucity of well control and
the abundance of high-quality 3-D seismic data at the exploration and
development scales require interpretation of reservoir type, or
environment of deposition, to be performed using detailed seismic facies
analysis. Key criteria of ExxonMobil's deep-water seismic facies scheme
include external geometry (e.g., truncation, onlap, mounding, etc.),
amplitude strength and continuity (e.g., high-amplitude continuous vs.
high-amplitude semicontinuous), and attribute map patterns. Because of
the nonuniqueness inherent in seismic facies analysis, translation of
these seismic facies into depositional environments requires careful
selection of an appropriate analog, be it either subsurface or outcrop
based. In the case of outcrops, architectural analysis can provide the
characteristics of the fundamental units that comprise subsurface
reservoirs.
Different architectural elements
yield different seismic signatures, such as channels vs. sheets and
axial vs. marginal lithofacies associations. Typically, these elements
are at or below seismic resolution. Additional challenges of applying
the outcrop analogs appropriately to a seismic response are related to
the signatures of individual sand bodies, which can vary with the
seismic bandwidth and rock properties. Lastly, most single channels and
sheets stack to form complexes, and the interplay between these
different individual sand bodies can also modify their seismic
signatures.
Normal-incidence forward seismic
models have been constructed using GXII seismic modeling software for
both the Skoorsteenberg and Ross Formations to calibrate the appropriate
outcrop analogs to the Diana subsurface data. These models are shown in
Figures 8, 9,
10 and illustrate both the seismic facies of
individual outcrops using subsurface rock properties (density and
velocity) from the Gulf of Mexico and the resolution limits of typical
seismic data. All forward seismic models were generated using
vertical-incidence ray tracing and a zero-phase Ricker wavelet. A trough
(red) represents a negative impedance boundary, and a peak (black)
represents a positive impedance boundary. Each of the outcrops discussed
in the previous section would be seismically expressed as a single cycle
at the bandwidth and rock properties of the deep-water Gulf of Mexico.
These models provide a link between the architectures observed in
outcrop and seismic data in the same way synthetic seismograms link
well-log and core data to seismic data.
The comparative seismic response
of the medial, transition from proximal to medial, and proximal portions
of the Skoorsteenberg and Ross Formations reveals that the variations in
sand-body architecture and degree of vertical and lateral amalgamation
are manifested as changes in amplitude strength and continuity and
subtle changes in isochron. As would be expected, the layered, extremely
continuous medial fan sheets of the Ross Formation (Figure
8) produce a high-amplitude continuous seismic character at typical
30 Hz seismic frequencies. The high net-to-gross section at the top of
the outcrop can only be fully resolved on the seismic model generated
using a 60-Hz wavelet. At this higher frequency, this interval also is
acoustically transparent because of the high percentage of sandstone and
resulting lack of internal reflectivity.
Individual channels from the
proximal to medial fan transition of the Ross Formation are not
resolvable except at the highest frequency (Figure
9). At the channel-complex scale, however, the lateral change from
high net-to-gross axis to lower net-to-gross margin is reflected clearly
in a lateral degradation of amplitude strength. This indicates that
these distinct lateral changes in sand percentage should be seismically
detectable.
The vertically and laterally
amalgamated, high net-to-gross proximal fan channels of the
Skoorsteenberg Formation also display a high- to moderate-amplitude,
moderately continuous seismic character (Figure
10). In contrast to the medial fan sheets, however, modeling of
these outcrops exhibits greater evidence of variation in isochron
because of the channelized nature of the outcrops.
Each of these forward seismic
models is subtly different. These differences reflect the proximal to
distal variations that are inherent in many deep-water depositional
systems. Integrating the knowledge from detailed analysis of these
deep-water outcrops and forward seismic modeling can provide important
information concerning variations in reservoir architecture and
net-to-gross values that can ultimately control the development
potential of many deep-water reservoirs.
|
 |
FIGURE 11. (A) Inline 807 is a
depositional strike section through the reservoir (see
Figure 1 for
location). The proximal portion of the
Diana Subbasin, which includes the Diana 2 and Diana 3 well
penetrations, is represented by high-amplitude, continuous
seismic character above the gas-oil contact. The observed
amplitude dimming toward the Diana 3 location is fluid related
(change from gas to oil) and is not associated with variations
in net-to-gross. Subtle variations in isochron are observed for
the A-50 and may suggest a more channelized reservoir, but it is
not conclusive. (B) The forward seismic model of the proximal
fan channels from the Skoorsteenberg Formation is very similar
to the seismic character observed on the seismic line through
the Diana 2 and Diana 3 wells. (C) This similarity between the
forward seismic model of the proximal fan outcrop and the actual
seismic data supports the channelized, proximal fan
interpretation for the A-50 reservoir in this portion of the
basin. |
|
 |
FIGURE 12. (A) Inline 651 is a
depositional strike section through the reservoir (see
Figure 1 for
location). The medial portion of the
reservoir (Diana 1 well penetration) has a distinctly different
seismic character than the updip interpreted proximal portion of
the reservoir (Figure
11A) and is represented by a
high-amplitude, semicontinuous seismic character above the
gas-oil contact. The observed lateral variation in seismic
character of the A-50 sands in the vicinity of Diana 1 suggests
that seismically detectable variations in net-to-gross and
reservoir architecture exist. (B) The forward seismic model of
the channels/sheets, which characterize proximal to medial fan
transition in the Ross Formation, bears a striking resemblance
to the seismic character of the A-50 reservoir where Diana 1 is
drilled. (C) The seismic character of this segment of the
reservoir, therefore, suggests this portion of the A-50
reservoir is less channelized than updip in the vicinity of
Diana 2 and Diana 3. |
|
 |
FIGURE 13. Plain light (left
side) and ultraviolet (right side) core photos from Diana 3
well. The cored interval is comprised of sharp-based,
upward-fining channels (red arrows mark interpreted channel
bases) and individual channel-fill successions that can be
further subdivided into channel-axis, channel off-axis, and
channel-margin associations. Note: core depth is in feet
(modified from Sullivan and Templet, 2002). See
Figure 14 for key to bed
types. |
|
 |
FIGURE 14. Summary plot for EB
946-1 well (Diana 3) illustrating core lithofacies, interpreted
depositional setting, and relationship between log and core
depths for the A-50 reservoir. A statistical foot-by-foot
comparison of gamma-ray logs vs. core-described facies for all
cored intervals was used to interpret the depositional facies in
uncored portions of wells. These facies-blocked wells were then
used to condition the object-based model and control channel
distribution and vertical stacking patterns (modified from
Sullivan and Templet, 2002). |
|
 |
FIGURE 15. (A) Integration of
Diana seismic, well-log, core, and appropriate outcrop analog
data provided the detailed geometric data required for
interpreting the reservoir architecture at a subseismic scale.
(B) Depth model constructed for interpreted reservoir
architecture between Diana appraisal wells by applying
subsurface rock properties (density and velocity) from the A-50
sandstones at the Diana field. (C) The forward seismic model
generated from the interpreted reservoir architecture between
the Diana appraisal wells produces a seismic response extremely
similar to the actual seismic data (see
Figure 11),
suggesting that the appropriate architectural data had been
selected to populate the reservoir model. |
Text
Based on detailed analysis, the
3-D seismic data at the Diana field appears to be of variable quality
and does not allow direct geometric analysis of reservoir elements (Figure
1B). To assist with assessment, deep-water outcrop analog data and
forward seismic modeling were integrated with seismic and well data to
produce a more accurate characterization of the reservoir. Seismic
amplitude extractions for the A-50 reservoir display distinct stripes in
the in-line direction that are interpreted to be related to the
acquisition of the survey. This complicates any quantitative attribute
analysis of the reservoir and restricts the application of seismic
amplitude map patterns to delineate sand-body trends and dimensions.
Qualitative examination of vertical seismic in-lines and cross-lines,
however, provides valuable information concerning the architecture of
the reservoir elements (Figures 11,
12). The A-50 sands are low impedance where
they are hydrocarbon charged and are typically represented by a
single-cycle seismic event (trough-peak pair) with a trough (red =
negative impedance boundary) at the top and a peak (black = positive
impedance boundary) at the base on zero-phase data.
The proximal portion of the Diana
field, which includes the Diana 2 and Diana 3 well penetrations, is
represented by high-amplitude, continuous seismic character above the
gas-oil contact (Figure 11). The observed
amplitude dimming toward the Diana 3 location is fluid related (change
from gas to oil) and is not associated with variations in net-to-gross (Figures
13, 14, 15A).
This suggests that, if variation in net-to-gross and reservoir
architecture exists in this portion of the reservoir, it is below
seismic detection. Furthermore, forward seismic modeling indicates that
both high net-to-gross, amalgamated proximal fan channels (Figure
10) and moderate net-to-gross, layered medial fan sheets can have a
similar high-amplitude continuous seismic response (Figure
8). Subtle variations in isochron are observed for the A-50 interval
and may suggest a more channelized reservoir, but it is not conclusive.
The Diana 2 and Diana 3 wells, however, penetrate a very high
net-to-gross interval dominated by amalgamated high-concentration
turbidites and shale clast conglomerates (Figures
13, 14, 15A).
This association of facies, in conjunction with the seismic character of
the A-50 interval, suggests a relatively channelized reservoir (Figures
11C, 15A).
The medial portion of the field
(Diana 1 well penetration) has a distinctly different seismic character
than the updip portion of the reservoir (Diana 2/Diana 3 region) and is
represented by a high-amplitude, semicontinuous seismic character above
the gas/oil contact (Figure 12A). Forward
seismic modeling shows that lateral change from high net-to-gross to
lower net-to-gross should be reflected by a degradation of amplitude
strength (Figure 9). This suggests that the
lateral variation in seismic character of the A-50 sands in the vicinity
of Diana 1 is caused by seismically detectable variations in
net-to-gross and reservoir architecture. Well penetrations confirm this
interpretation, as Diana 2 was drilled in a higher-amplitude portion of
the reservoir and encountered approximately 85% net-to-gross (Figure
15A). Diana 3 also is extremely high net-to-gross (Figures
13, 14, 15A),
but it was drilled in the oil leg and, as a result, has a lower
amplitude. In contrast, Diana 1 is drilled in a lower amplitude within
the gas cap, and the net-to-gross is significantly lower (approximately
65%). Laterally away from the Diana 1 penetration, the amplitudes
brighten, and this is interpreted to reflect more axial, higher
net-to-gross portions of the reservoir (Figure
12A). The seismic character of this segment of the reservoir,
therefore, suggests a less-channelized reservoir than updip (Figure
12C). In fact, the seismic character is very similar to the forward
seismic model for the channelized sheets, which dominate the transition
from proximal to medial fan deposits of the Ross Formation (Figures
9, 12). This supports the proximal fan
interpretation for the updip deposits around Diana 2 and 3 (Figure
11).
Excellent core coverage in the
Diana field also enables close calibration of seismic and well data. The
cored interval is comprised of stacked, sharp-based, upward-fining
channels (Figures 13,
14). Individual channel-fill successions can
be subdivided into channel-axis, channel off-axis, and channel-margin
associations in a similar fashion as the outcrops of the Skoorsteenberg
and Ross Formations (Figures 3,
4, 5,
6, 7).
Channel-axis deposits are characterized by highly amalgamated, massive
sandstones deposited from high-concentration turbidity currents (Figures
13, 14). The channel off-axis
association is composed of stacked, semi- to nonamalgamated, massive to
planar-stratified sandstones and interlaminated mudstones (Figure
14). The channel-margin deposits contain a variety of lithofacies
and are characterized by a heterolithic mixture of interbedded
sandstones and mudstones (Figures 13,
14). Statistical foot-by-foot comparisons of
log curves vs. core-described lithofacies were used to interpret
depositional facies in uncored portions of wells. Blocked wells were
further used to condition an object-based geologic model and to control
the distribution of channel elements and vertical stacking patterns (Figure
14).
Integration of seismic, well-log,
and core data with forward seismic models of deep-water outcrop analogs,
therefore, suggests a more channelized reservoir updip (Figure
11), becoming more distributive and sheetlike downdip (Figure
12). This subsurface data, however, does not have the resolution to
provide the dimensional and architectural data required to populate a
geologic model for flow simulation and well-performance prediction.
Diana
Reservoir Model
Figures 16-20
|
 |
FIGURE 16. Single-channel (facies
body) model. Modeled channels are divided into proximal (p),
medial (m), and distal (d) regions, each with their specific set
of characteristics. Channels are further subdivided into axis
(CA), channel off-axis (COA), and margin associations (CM). The
final model contains more than 100 individual channels generated
from a range of possible widths and thicknesses (modified from
Sullivan et al., 2000a). |
|
 |
FIGURE
17. Based on this modeling effort, notable variability
exists in both the vertical and lateral continuity of the
reservoir facies from updip to downdip. This is illustrated by
the change in cell dimensions from thicker and more equal
dimensional (amalgamated) updip to thinner and more elongate
(nonamalgamated) downdip (modified from Sullivan et al., 2000a). |
|
 |
FIGURE
18. Net-to-gross map generated from the facies model,
assuming a constant net-to-gross of 0.95, 0.85, 0.65, and 0.10
for the channel axis, channel off-axis, channel margin facies,
and overbank facies, respectively. Net-to-gross ranges from
greater than 0.95 in the proximal area (red) to less than 0.40
in the distal area (purple). A low net-to-gross
 crease,
shown just east of the southern stewardship polygons, honors an
interpreted net-to-gross low, visible on seismic amplitude maps. |
|
 |
FIGURE
19. Side scan sonar image for Mississippi Fan illustrating
the detailed sand-body architecture. The bright colors represent
more sand-prone regions of the fan. The map pattern of this
modern fan, which has been rotated to match the orientation of
the Diana net-to-gross model, is extremely similar to the map
pattern produce by the Diana net-to-gross model. This similarity
between the map patterns of the Mississippi Fan and the
object-based model for the A-50 reservoir helps to validate the
modeling effort (modified from Twichell et al., 1995). |
|
 |
FIGURE
20. Oil saturation maps for the Diana oil rim for the first
4 yr of production. Note that the Diana gas cap and aquifer are
not shown. The updip portion of the reservoir has higher initial
oil saturations because of its higher porosities and also starts
making high water cuts earlier than the downdip portion of the
reservoir because of its more amalgamated character and better
reservoir quality. The smaller cell size observed at the
proximal portion of the model reflects the downscaling of the
reservoir relative to the distal portion of the reservoir that
was upscaled (modified from Sullivan et al., 2000a). |
Text
To solve these uncertainties,
dimensional and architectural data (e.g., width vs. thickness
measurements) from the Skoorsteenberg and Ross deep-water outcrops (Figures
3, 4, 5,
6, 7) were
compared to the interpreted thickness data derived from the
Diana-specific seismic, well-log, and core data and were adjusted
accordingly (Figures 13,
14, 15A). From
these measurements, a spectrum of channel dimensions and shapes was
collected. Comparison of the forward seismic models of the
Skoorsteenberg and Ross deep-water outcrops to the actual Diana seismic
data was made to select the appropriate architectural data to populate
the reservoir model (Figures 11,
12, 15). In
addition to the collection of channel dimensions and shapes, bed
continuity, and lateral and vertical facies, variability data also were
gathered from both outcrop analogs and well logs/core to condition the
reservoir model, as these factors ultimately control the reservoir
behavior (Figures 6,
13, 14).
In the case of the Diana field,
this data was used to help maximize the development of the relatively
thin, yet economically important oil rim. This was accomplished by
building a detailed object-based reservoir model, which integrated both
subsurface and outcrop data. The model was built using ExxonMobil
proprietary code for modeling deep-water reservoirs and the reservoir
modeling system IRAP-RMS object-based modeling tool. This model consists
of discrete objects (facies bodies), each with specific dimensions,
facies juxtapositions, and continuity. This type of modeling is
appropriate in data-limited situations where a facies model is based on
a conceptual interpretation of reservoir architecture. The reason for
choosing this technique to model the Diana reservoir included (1) the
poor quality of the seismic data, (2) limited well penetrations, (3)
interpretation of the reservoir being comprised of channels with
distinct lateral changes in facies (axis to margin), (4) interpretation
of updip to downdip changes in channel architecture and net-to-gross,
and (5) desire to apply a concept-driven geologic model that
incorporated outcrop analog information.
The fundamental object in this
reservoir model is a turbidite-dominated deep-water channel. In the
Diana model, individual channels are narrow updip and become wider and
less amalgamated downdip (Figures 16,
17), as observed in the outcrops of both the
Skoorsteenberg and Ross formations (Figures 3,
4, 5,
6, 7). Modeled
channels are divided into proximal, medial, and distal regions with
their own specific set of characteristics. Channels are subdivided
further into axis, off-axis, and margin associations. Lateral
degradation in reservoir quality from axis to margin is interpreted from
core data, as observed in the outcrops of the Skoorsteenberg and Ross
Formations (Figures 3,
4, 5,
6, 7). Shales
were inserted as objects in a fine-layer (0.1 m) framework. The spatial
distribution of shales in a given facies type is random (e.g., there was
no preferential placement of shale either areally or vertically in the
3-D model volume). The volume of shale added to a facies depended on the
net-to-gross of the facies type (e.g., axis vs. margin). Shale
dimensions were obtained from the Skoorsteenberg and Ross Formations (Figure
6). Shale objects in the model are square or rectangular in shape,
and their dimensions depend on facies type (e.g., axis vs. margin) and
location (e.g., proximal vs. distal).
The final model contains more
than 100 individual channels, each one stochastically generated from a
range of possible widths and thicknesses (Figures
16, 17, 18).
The facies objects were inserted first at the well locations and then
subsequently inserted stochastically into interwell regions according to
geologic constraints (e.g., vertical stacking patterns), until volume
targets were met. Net-to-gross maps, which were generated by calculating
the average value of the sand-shale parameter at a given X, Y
location in the model, provide an indication of how the net sand is
distributed in the model (Figure 18). The
resulting net-to-gross maps strongly resemble modern deep-water systems,
such as the Mississippi Fan (Figure 19), and
further support this integrated study. Each facies and subfacies body
was then populated with petrophysical properties using Gaussian
simulation drawn from subfacies property histograms generated from
available well data. To preserve the facies architecture and
heterogeneity expected in a channel-dominated deep-water setting, the
rock property modeling was performed in individual channel objects.
Based on this modeling effort and
flow simulation, significant variations in reservoir performance exist
from updip to downdip (Figure 20). The
development strategy for the Diana field is to produce oil initially
from horizontal wells high in the oil rim. Once water breaks through in
significant quantities, these wells will be recompleted in the gas cap.
The goal is to maximize oil production while minimizing water production
and movement of oil into the gas cap. Typically, reservoir models are
scaled up for flow simulation. However, in this case, the updip portion
of the reservoir was actually scaled down to preserve its more
channelized and amalgamated nature (Figure 20).
The updip portion of the reservoir has higher initial oil saturations
because of its higher porosities. It also starts producing water earlier
than the downdip portion of the reservoir because of its higher
porosities and more channelized nature. This study, therefore, predicts
significant variations in reservoir producibility that exist across the
Diana field. This information was used to place wells in optimum
locations to maximize the architectural controls on reservoir
performance and has had a significant impact on the final development
strategy for the field.
This study shows the importance
of incorporating outcrop analogs in the analysis of subsurface
reservoirs. Outcrop research is critical because the observed updip to
downdip variability in sand-body geometry, continuity, and net-to-gross
of deep-water reservoirs affects both the exploration and production
potential of these sandstones. Commonly, this variability, as in the
case of the A-50 reservoir at the Diana field, is at or below seismic
resolution, and well penetrations are typically limited. Properly
calibrated deep-water outcrops can provide constrained geometric and
architectural data to fill the gaps between wells or stochastic modeling
uncertainties below the resolution of seismic data. Dimensional and
architectural data from outcrops and forward seismic modeling can
therefore be integrated with seismic and wellbore data to build regional
depositional models to better understand reservoir distribution and
delineate exploration plays. Deep-water outcrop data can also be used to
help populate object-based models that can be used to more accurately
predict well performance, connected volumes, and recovery efficiencies
for newly discovered fields. Furthermore, the integration of seismic,
well-log, core, and outcrop data with object-based models provides the
framework for optimal placement of wells to maximize the architectural
controls on reservoir performance. The bottom-line impact of this type
of integrated analysis has been a significant reduction in the range of
uncertainty attached to reservoir assessment parameters for deep-water
sandstones, both in the Diana Subbasin and in many other areas where
exploration and development of deep-water reservoirs is currently
occurring.
Bouma, A.H., 2000, Fine
grained, mud-rich turbidite systems: Models and comparison with
coarse-grained sand-rich systems, in A.H. Bouma, ed., Fine
grained turbiditic systems: AAPG Memoir 72, p. 9-20.
Bouma, A.H., and H. de V. Wickens, 1994,
Tanqua Karoo, ancient analog for fine-grained submarine fans, in
P. Weimer, A.H. Bouma, and B.F. Perkins, eds., Gulf Coast Section SEPM
Foundation, Fifteenth Annual Research Conference: Submarine fans and
turbidite reservoirs, p. 23-34.
Chapin, M.A., P. Davies, J.L. Gibson, and H.S.,
Pettingill, 1994, Reservoir architecture of turbidite sheet sandstones
in laterally extensive outcrops, Ross Formation, western Ireland, in
P. Weimer, A.H. Bouma, and B.F. Perkins, eds., Submarine fans and
turbidite reservoirs: Gulf Coast Section SEPM Foundation, Fifteenth
Annual Research Conference, p. 53-68.
Collinson, J.D., O.J. Martinsen, and A.
Kloster, 1991, Early fill of the Western Irish Namurian Basin: A complex
relationship between turbidites and deltas: Basin Research, v. 3, p.
223-242.
Elliot, T., 2000, Depositional architecture
of a sand-rich, channelized turbidite system: the Upper Carboniferous
Ross Sandstone Formation, western Ireland, in P. Weimer, R.M.
Slatt, A.H. Bouma, and D.T. Lawrence, eds., Deep-water reservoirs of the
world: Gulf Coast Section SEPM Foundation, Twentieth Annual Research
Conference, p. 342-373.
Imran, J., G. Parker, and N. Katopodes, 1998,
A numerical model of channel inception on submarines fans: Journal of
Geophysical Research, v. 103, no. C1, p. 1219-1238.
Martinsen, O.J., T. Lien, and R. Walker,
2000, Upper Carboniferous deep water sediments, western Ireland:
Analogues for passive margin turbidite plays, in P. Weimer, R.M.
Slatt, A.H. Bouma, and D.T. Lawrence, eds., Deep-water reservoirs of the
world: Gulf Coast Section SEPM Foundation, Twentieth Annual Research
Conference, p. 533-555.
Morris, W.R., M.H. Scheihing, H. de V.
Wickens, and A.H., Bouma, 2000, Reservoir architecture of deep-water
sandstones: Examples from the Skoorsteenberg Formation, Tanqua Karoo
Sub-basin, South Africa, in P. Weimer, R.M. Slatt, A.H. Bouma,
and D.T. Lawrence, eds., Deep-water reservoirs of the world: Gulf Coast
Section SEPM Foundation, Twentieth Annual Research Conference, p.
629-666.
Sullivan, M.D., and P. Templet, 2002,
Characterization of fine-grained deep-water turbidite reservoirs:
Examples from Diana Sub-Basin, western Gulf of Mexico, in P.
Weimer, M. Sweet, M.D. Sullivan, J. Kendrick, D. Pyles, and A. Donovan,
eds., Gulf Coast Section SEPM Foundation: Deep-water Core Workshop,
northern Gulf of Mexico, CD-ROM, p. 75-92.
Sullivan, M.D., J.L. Foreman, M. Devries, and
A. Khan, 1998, Application of deepwater outcrop analogs to 3-D reservoir
modeling: An example from the Diana field, western Gulf of Mexico, in
M.R. Mello and P.O. Yilmaz, eds., International AAPG Meeting Extended
Abstracts Volume, Rio de Janeiro, Brazil, November 8-11, 1998, p. 24-25.
Sullivan, M.D., J.L. Foreman, D.C. Jennette,
D. Stern, and A. Liesch, 2000a, Application of deepwater outcrop analogs
to 3-D reservoir modeling: An example from the Diana field, western Gulf
of Mexico, in R. Shoup, J. Watkins, J. Karlo, and D. Hall, eds.,
AAPG/Datapages Discovery Series No. 1: Integration of geologic models
for understanding risk in the Gulf of Mexico, p. 1-14.
Sullivan, M.D., G.N. Jensen, F.J. Goulding, D.C. Jennette,
J.L. Foreman, and D. Stern, 2000b, Architectural analysis of deep-water
outcrops: Implications for exploration and production of the Diana
Sub-basin, western Gulf of Mexico, in P. Weimer, R.M. Slatt, A.H.
Bouma, and D.T. Lawrence, eds., Deep-water reservoirs of the world: Gulf
Coast Section SEPM Foundation, Twentieth Annual Research Conference, p.
1010-1032.
Twichell, D.C., W.C. Schwab, and N.H. Kenyon, 1995,
Geometry of sandy deposits at the distal edge of the Mississippi Fan,
Gulf of Mexico, in K.T. Pickering, R.N. Hiscott, N.H. Kenyon, F.
. Lucchi, and R.D.A. Smith, eds., Atlas of deep water environments:
Architectural style in turbidite systems: London, Chapman and Hall, p.
282-286.
The authors would like to thank
Dave Larue, Mike DeVries, Arfan Khan, DeVille Wickens, and Arnold Bouma
for their assistance in collecting outcrop data from the Skoorsteenberg
Formation. Ian Moore, Chris Armstrong, Kevin Keogh, and Trevor Elliot
are also thanked for their assistance in collecting portions of the
outcrop data from the Ross Formation. Permission to publish this paper
was granted by ExxonMobil Upstream Research and by BP Exploration. The
authors would also like to thank Grant Wach, William Schweller, Jim
Borer, Michael Grammer, and Ray Sullivan for reviewing and improving
this paper. In addition, we would like to acknowledge Ed Garza for all
of his assistance in producing the illustrations presented in this paper
Return to top.
|
Print this page
Click
to article in PDF format.