Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Geophysical Prospecting, 2005, 53, 183–203

The 4D seismic signature of oil–water contact movement


due to natural production in a stacked turbidite reservoir
C. MacBeth,1∗ K.D. Stephen1 and A. McInally2 †
1 Reservoir Geophysics Group, Institute of Petroleum Engineering, Heriot-Watt University, Riccarton, Edinburgh EH14 4AS,
and 2 Shell UK Exploration and Production, Aberdeen AB12 3FY, UK

Received December 2003, revision accepted July 2004

ABSTRACT
Combined time-lapse reservoir simulation and seismic modelling has been performed
on both 1D and 3D models of a channelized turbidite reservoir. The models have
been built using core, log, laboratory and seismic data from the Nelson Field (central
North Sea) as a template. Oil and water movement in the main channels, channel
margins and interchannel regions is investigated, with a particular focus being the
effect of poor net-to-gross. The analysis confirms that saturation effects dominate the
response whilst stress-sensitivity effects play a minor role. The trough–peak signature
in the seismic difference volumes formed by the sweep of the water can be continued
and mapped slightly further than the channel margins. This characteristic 4D signa-
ture remains roughly intact, despite the complicated depositional architecture, and
accurately delineates the area of moved fluid, but it needs additional calibration to
combat the detrimental influence of the low net-to-gross. Signal strength is largely
dependent on reservoir quality, but is also compromised by the net-to-gross, fluid
distribution and, more critically, by the exact timing of the seismic survey. For ex-
ample, a region of bypassed oil zone remains undetected as it forms early during the
production. This work demonstrates that to understand fully the 4D signature at a
quantitative level requires adequate knowledge of the fluid properties, but also, more
critically, the geology.

in the 4D sense by a strong signature from clean, relatively un-


INTRODUCTION
consolidated sands with moderate-to-high porosity and an ex-
Stacked turbidite reservoirs are amongst the most commonly cellent contrast in acoustic properties between the displacing
cited 4D successes. Examples are the Palaeocene sands of and displaced reservoir fluids. However, turbidite reservoirs
the West of Shetlands area of the North Sea (Parr, Marsh are also known in the geological sense for the many challenges
and Griffin 2000), the Nelson Field, central North Sea (McI- and surprises offered to reservoir management (Dromgoole
nally et al. 2003) and the Pliocene sands of the South Tim- et al. 2000), due to their complicated depositional architec-
balier Field, Gulf of Mexico (Burkhart, Hoover and Flem- ture, shale occurrence, stratigraphic variations in net-to-gross,
ings 2000). 4D seismic potential for fields lying in deep water and the geometry and internal character of the deposits. In-
has also been demonstrated for turbidite developments in the deed, heterogeneity exists on a variety of scales that will
Oligocene/Miocene of the Campos Basin, Brazil (Santos et al. strongly influence both vertical and lateral connectivity, and
2003) and the Palaeocene of the Girassol Field in West Africa hence fluid-flow behaviour during production or injection.
(Beydoun et al. 2002). Most of these fields are characterized This heterogeneity could compromise our quantitative un-
derstanding of the 4D seismic signature. To investigate this
∗ E-mail: colin.macbeth@pet.hw.ac.uk point further, the monitoring of water sweep during natural
†Now at: Kerr-McGee North Sea (UK) Ltd, Aberdeen AB12 3LG, UK. (primary) production is taken as an example. In this situation,


C 2005 European Association of Geoscientists & Engineers 183
184 C. MacBeth, K.D. Stephen and A. McInally

Table 1 Snapshot of 4D characteristics for the Nelson Field, central North Sea. This field is used in the current study as a data source and a
template to guide the modelling

Field Geology
Discovered 1986 Channelized turbidite
First production 1994 Palaeocene, Upper Forties sands
30 production wells Broad, low relief anticline with four-way dip closure
Natural drive, aquifer supported Three main sand fairways, running NW–SE
Flank water injection from four wells
Rocks Fluid
23–28% porosity 38◦ API (light)
50–1000 mD permeability 550 scf/stb initial GOR
Stress sensitivity – see Appendix C Initial pressure 3343 psi (23 MPa)
Rock compressibility 0.04 GPa−1 Bubble pressure 1700 psi (12 MPa)
Caprock: Sele shale Brine salinity 100 000 ppm
Temperature 107◦ C
Reservoir Seismic
Top Reservoir 7200 ft TVDSS 30 Hz peak frequency
270 ft oil column Surveys 1990, 1997, 2000
Swc = 0.20 (on average) Offset range 0–8000 ft
Sor = 0.15–0.25 280 ft water depth
Production 4D changes
18 000 bpd in well of interest Oil-sand → water-sand, P = 0: IP = 11%, IS = 43%
1994/3343 psi (23 MPa) → Water-sand (at 3343 psi) → water-sand (at 2343 psi):
1997/2343 psi (16 MPa) over 3.5 years IP = 1%, IS = −4%
Reservoir pressure stable at 1997 level Oil-sand (at 3343 psi) → oil-sand (at 2343 psi): IP = 1.5%, IS = −5%

the expectation is for a well-defined oil–water contact (OWC) To define the underlying physics for oil–water contact move-
movement due to sand-on-sand contact between beds. Gener- ment, this study firstly undertakes log-based 1D modelling,
ally, this is the case in the main channel sands; however, vertical and follows this with the construction of a 3D geological and
connectivity progressively diminishes towards the margins of reservoir model containing similar depositional architecture to
the depositional system as the shale volume and lateral conti- Nelson. From this, the resultant near- and far-offset 4D seismic
nuity of the channel beds increases. In these poor net-to-gross responses are generated through petro-elastic transformation
regions at the margins of the channel sands, oil is still present of the simulation output followed by seismic modelling, and
but contact movements (or saturation changes) and bypassed the fidelity of the key indicators for water movement is stud-
patches of oil are more difficult to detect due to the increased ied. In the latter step, consideration is given to the role of
complexity of the 4D signature. stress sensitivity of the sands and shales. For our purposes, it
The objective of this work is to develop a better under- is assumed that the reservoir is produced naturally through
standing of the fluid-flow in stacked turbidite reservoirs un- a strong basal aquifer drive. Although applicable only to an
dergoing primary production, with a focus on the impact of oil–water system, this analysis is equally appropriate for the
net-to-gross variations and their effect on the resultant 4D gas–water case such as that of Alsos et al. (2003). Further
signature. For the purposes of this current study we use data issues relating to other reservoir connectivity scenarios and
from the Palaeocene sands of the Nelson Field in the central stacking patterns in turbidite sands are discussed in a separate
North Sea as a general framework to guide the modelling and study (Al-Maskeri and MacBeth 2003).
analysis. Geological and reservoir details of this field are given
in Table 1. Figure 1(a) shows an example of the seismically es-
PREDICTED TIME-LAPSE EFFECTS FROM
timated oil–water contact movement across the Nelson Field
1D MODELLING
compared with the simulator prediction (after Reid, MacBeth
and McInally 2003). Figure 1(b) illustrates, in vertical section, The submarine turbidite sands of the Forties sandstone mem-
the characteristic trough–peak signature in the 4D seismic as- ber form three distinct axial fairways: the eastern chan-
sociated with this movement. nel, central channel and western channel complexes. These


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 185

Figure 1 (a) Map showing picked oil–water contact (OWC) movement between the 1990 (pre-production) and 2000 (monitor) surveys on
Nelson, compared with the moved oil–water contact predicted from the reservoir simulation (after Reid et al. 2003). 4D seismic has proved to
be a valuable tool in monitoring the drainage patterns in Nelson, with predictions from the reservoir simulator being in good overall agreement
with the seismic estimates. The largest movements correlate with the production. (b) Section showing the trough (red) and peak (blue) signature
as defined by the 1990–2000 4D volume. The trough represents the moved oil–water contact (MOWC) and the peak represents the original
oil–water contact (OOWC).

sand-rich fairways run in a northwest–southeast direction


Main channel sands
across the structure, and are well imaged on the 4D seis-
mic. For the purposes of this current study a comparison will On Nelson, the 4D seismic is known to be dominated by the
be made between the main channel sands and the more het- saturation changes, due principally to the contrast between
erolithic channel-margin facies, which are less well resolved the bulk fluid properties of the oil and the formation wa-
by 4D seismic. ter, and in particular the upward movement of the oil–water


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
186 C. MacBeth, K.D. Stephen and A. McInally

Figure 2 Schematics illustrating the possible


effect of reduced net-to-gross on the differ-
ence (base – monitor) signature due to nat-
ural production and an oil–water contact
movement of 75 ft. (a) Main channel sand.
(b) Sands at the margins of the channel. Un-
desirable events from within the sand pack-
age are produced in this low net-to-gross
case. Internal multiples from layers inside
the swept zone interfere with other multi-
ples whose raypaths cross through the swept
zone. This process has the potential to gen-
erate a whole sequence of trough–peak dou-
blets that interfere with the main signal of
interest.

contact (Boyd-Gorst, Fail and Pointing 2001). In the main duces a second change in the saturation profile (and hence
sand-channel fairways, the direct signature of production is a elastic properties), due to the partially swept sand (as seen
characteristic trough–peak doublet on the difference (base – in Fig. 3a). To describe this 4D signature in more detail, an
monitor) data (Fig. 2a), formed by the interference of reflected elastic model is calculated for idealized pre- (1990) and pro-
events from the original and moved oil–water contacts, which duced (1997) saturation distributions, and reservoir pressures
can thus be used to indicate and quantify oil–water contact are included for completeness. Pressure-sensitive oil and brine
movement (Redondo-López et al. 2002). Before production, acoustic properties under the specific reservoir conditions be-
there is complete water saturation in the water-sand and ir- fore and during production (Table 1) are determined using
reducible, connate water saturation Swc in the oil-sand. Pro- the appropriate equations of Batzle and Wang (1992) with
duction sweeps oil by an upward movement of the water leg laboratory correlations established by Enterprise Oil replac-
in response to the pressure depletion, but with the residual ing relationships for the bulk moduli κ brine and κ oil . It is as-
(and any bypassed) oil remaining. Water sweep thus intro- sumed that production occurs sufficiently slowly for the fluid


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 187

Figure 3 (a) Idealized changes in elastic


properties for water sweep due to natural
production. (b) Baseline (1990) and monitor
(1997) seismic, together with 4D signatures
corresponding to the upper model. Compu-
tations use a zero-phase source wavelet of
30 Hz. The characteristic trough–peak dou-
blet is apparent in all cases. Note that when
computing the elastic properties, the effect
of the pressure gradient within the reservoir
can be assumed to be negligible and it is
thus held at a fixed value for the reservoir
depth range. The oil–water capillary tran-
sition zone is approximately 3–15 ft, and
can thus be simulated at seismic frequencies
as an abrupt saturation discontinuity com-
pared with the seismic wavelength of 300
ft. For this clean-sand example, irreducible
water saturation (Swc ) and residual oil satu-
ration (Sor ) are set at 20% and 15%, respec-
tively.

constituents to be homogeneously distributed at the sub- proximately Top Forties) events mostly cancel out in the 4D
seismic (≤ λ/5) scale, so that the Reuss average (Wood’s equa- signature, which is made up of a peak corresponding to the
tion) can be used to mix the fluid constituents. The exact va- original oil–water contact (OOWC) event combined with a
lidity of this assumption for natural production in turbidites trough associated with the moved oil–water contact (MOWC)
is examined elsewhere. The contribution of the bulk modulus event. As the incident seismic wavelength is approximately
and fluid density to the overall rock properties is then obtained 300 ft, for sweeps of up to 70 ft the trough and peak in
through application of the conventional Gassmann’s equation this composite interfere destructively in an identical way to
and the mass balance equation, respectively. In this particular conventional tuning and thus the amplitude will scale in pro-
section only one common rock-frame and porosity pressure portion to the sweep length. Consequently, visual observation
response is used, this being the result of averaging a series could prove sufficient to monitor the sweep. Indeed, studies
of in-house measurements on different specimens (MacBeth using the seismic data have revealed that 15–20 ft movements
2004). Sele shale properties are also taken from an average of in the oil–water contact may be resolved in the presence of
Enterprise Oil laboratory results. 1–3 ms misregistration noise in 30 Hz seismic. As both the
Synthetic seismograms are calculated for pre-production original oil–water contact and the moved oil–water contact
and produced models using the isotropic reflectivity method events brighten with offset due to the Poisson’s ratio contrast
of Kennett (1983). This method includes all internal multi- (in contrast to the amplitude-variation-with-offset (AVO) be-
ples and P-S, S-S and S-P conversions. Difference signatures haviour of the Top Forties), the signature is more prominent on
for the near-offset (0–4000 ft) and far-offset (4000–8000 ft) the far-offset stack. This suggests enhanced fluid detection for
stacks are obtained between production start-up in 1994 and a the far-offset data. The seismic response, calculated using logs
post-production monitor survey in 1997 (3.5 years after pro- taken from a well in the main channel fairway, has a similar
duction) (see Fig. 3b). As expected, the Top Reservoir (ap- character.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
188 C. MacBeth, K.D. Stephen and A. McInally

Channel margins Although the data set used to derive (2) is predominantly
(but not exclusively) sandstone, it also adequately represents
As expected, the major controls on the seismic property dis-
shale for the purposes of this current study. In the modelling,
tribution are the step-changes at the moved oil–water contact
reservoir heterogeneity is treated by assuming that each in-
and original oil–water contact due to the contrast between
dividual point in the wireline log represents a thin homoge-
the light oil and formation water (Fig. 3b). In the margins, oil
neous bed with its own particular Swc . Once water satura-
is bypassed by the shale interbeds which exhibit low vertical
tions have been updated in the swept leg, the 4D signature is
connectivity, leading to a non-uniform saturation and poorer
computed using full-wave modelling as before. The reflectivity
recovery than would be predicted by laboratory-measured
approach permits preservation of the large-scale resolution,
residual saturations (see Appendix A). This is due to capil-
and thus upscaling of the properties is not necessary. More
lary pressure and relative permeability effects combined with
accurate saturation (and pressure) updating in the presence
between-bed heterogeneity. An additional effect is that the
of 3D heterogeneity will be addressed later in the reservoir
shales slow down the rise of water, thus these low net-to-
simulation.
gross areas are not likely to show significant water movement
As in the previous case, there is no definable Top Forties
until much later in field life. This uneven, meso-scale non-
event on the calculated 4D seismic. However, due to the shale
uniformity of saturation means that it is difficult to define a
interbedding, distinct reflection events at the fluid contacts are
sharp oil–water contact. This situation contrasts with the pre-
not individually identified on the seismic. The seismic is now
vious section. Figure 2(b) illustrates what events might occur
responding more to small changes in the reflectivity of beds
to provide additional complexity. Separate calculations reveal
in the vicinity of the saturation change, rather than a (seis-
that the internal multiple contribution for Nelson sands and
mically) sharp change across a single discrete interface. It is
shales is low, and the main character of the signal appears to
observed that increased shale fraction lowers the average seis-
be preserved. The trough and peak are sufficiently close to
mic velocity and hence reduces the overall P-wave impedance
interfere destructively, and their separation is not related di-
contrast at the Top Forties, and also between the water- and
rectly to the sweep. To simulate this effect more completely,
oil-sands. (Lower impedance shale is not unusual for fields in
logs from a well in a known low net-to-gross region are used
this part of the North Sea. These generally exhibit a type I
to calculate the seismic response due to production. The elastic
or II AVO response and dim out with offsets. This response
properties for the model are given in Fig. 4(a) using the same
contrasts with the classical Gulf of Mexico bright spots that
recipe as in the previous section. Again, a 75-ft water move-
show the opposite behaviour.) Figure 4(c) indicates the typical
ment is induced in the saturation profile. To establish a precise
amount of amplitude reduction that may be expected. Despite
saturation updating, the swept oil leg needs correction for the
these effects, it is still possible to observe the trough–peak sig-
maximum attainable Sw at each depth as the sweep progresses.
nature, albeit at a much reduced amplitude. This means that
At the pore scale, residual oil saturation is approximately con-
a larger sweep is necessary to induce a similar magnitude of
stant with permeability and rock type (Hamon 2003). Thus,
seismic signature to the main channel sands. The geological
for the purposes of this modelling, residual oil saturation (Sor )
heterogeneity has degraded the characteristic trough–peak in-
has been set at 15%. However, the connate water satura-
dicator. Fluctuations in Poisson’s ratio also reduce the promi-
tion Swc varies strongly with permeability. A low-permeability
nence of the far-offset response, although the 4D signature
rock with narrow pore throats is more fully water-saturated
still appears fractionally stronger than at near-offsets (see re-
when compared with a higher-permeability rock. To deter-
mark above). Thus, in principle, sweep determination could
mine this value, a laboratory-based empirical relationship
be continued into the channel margins despite the drop-off in
is used to connect directly to permeability (Ringrose et al.
net-to-gross. The 4D signature will faithfully follow the fluid
1993):
movement into the more complicated geological setting. How-
Swc = 0.600 − 0.165 log10 (k) , (1) ever, to use this satisfactorily the values would require further
careful calibration to provide a truly quantitative measure.
where k is in millidarcies (mD). This equation has been applied
Otherwise, good sweep in poor-quality sand could be mis-
generally in the past to both sands and shales. Finally, to assign
taken for a small sweep in good net-to-gross. Repeatability
permeability, the following porosity transform from core data
noise in the field data would ultimately determine whether
is used:
this drop-off in 4D amplitude outside the main channel axis is
k (in mD) = 0.00632 × 100.1667(%) . (2) significant.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 189

Figure 4 Saturation profiles with cor-


responding Poisson’s ratio and P-wave
impedance calculated from logs recorded in
a typical low net-to-gross marginal chan-
nel setting. Saturation updating for a 75-ft
sweep (as in Fig. 3) is achieved by calculating
Sor from an empirical Sor versus permeabil-
ity relationship (Pickup, pers. comm.). (a)
Pre-production: red lines; post-production:
black lines. (b) Synthesized near- and far-
offset stacks for baseline (1990) and moni-
tor (1997) surveys, together with their differ-
ence, indicate preservation of the basic 4D
signal of water sweep but lack the previous
correlation with sweep magnitude. (c) Near-
offset amplitude of the difference signature
as a function of net-to-gross for randomly
defined sand/shale realizations guided by log
statistics in the reservoir.

to retain realism whilst capturing the differing geologi-


S I M U L AT O R T O S E I S M I C 4 D M O D E L L I N G
cal scales. Two types of facies are considered in this par-
To understand further the relationship between the seis- ticular modelling: channel and interbed, both with dis-
mic response and fluid-flow for this channelized turbidite tinct property variations at the sub-seismic scale. There are
system, 3D geological and reservoir modelling are per- also two types of net-to-gross to consider, cell-averaged
formed. This requires that known net-to-gross distribu- and the larger-scale expression of the depositional ar-
tions be simulated using statistics derived from the Nelson chitecture. A workflow is developed with the following
wireline logs, cores and outcrop analogue data. This helps components.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
190 C. MacBeth, K.D. Stephen and A. McInally

Geological modelling sensitive fluid moduli, porosity and the rock-frame moduli
at each particular pressure and saturation are inserted into
A Cartesian model of 40 × 50 × 30 cells is built with an
the conventional Gassmann’s equations. Synthetic seismic vol-
individual cell resolution of 25 × 25 × 2 m. Channels are
umes are calculated using the same method as previously. The
modelled as objects placed in a background of interchan-
seismic are modelled with a 30-Hz zero-phase wavelet to sim-
nel facies with random offset positioning. The channel cross-
ulate the observed peak frequency of the seismic data, and also
section is assumed to be semi-elliptical and is extended in the
for 150 Hz to compare resolution issues (Fig. 6). The 150 Hz
transverse direction with some sinuosity (Fig. 5a). Successive
static pre-production seismic sections for the front face and
channels are created with later channels eroding into early
back face of the model show that the seismic mirrors most
channels up through the sequence. Channel dimensions (as-
of the geological architecture and tracks the high-porosity
pect ratio and thickness) are based on the Ainsa II outcrop in
(net-to-gross) regions. Edges of the channel are well defined,
Spain. Porosity and permeability (kv and kh ) of the individ-
but there are no internal reflections in the channels themselves
ual cells are obtained from cell-scale averaging and applied to
as the eroded sand-on-sand contacts and the gradients in elas-
large-scale property trends (see Appendix B). The interchan-
tic properties are not sufficiently contrasting. The total time
nel facies consist of continuous alternating layers of shale and
thickness of the reservoir is 45 ms (≈λ/2), and it is close to
medium-to-low-quality sand. Parallel layers of laterally con-
tuning for the 30 Hz seismic.
tinuous shales are sandwiched between medium-to-thin, low-
quality sub-seismic sands (less than 10 ft thick), permitting
mainly horizontal flow (low kv /kh ). This contrasts with the
R E S U LT S O F S I M U L AT O R T O
channel facies which consists of alternating layers of thicker
SEISMIC MODELLING
sandstone beds separated by shale eroded by various degrees,
associated with the deposition of each sand bed. The charac- Heterogeneous channel model with trapping
ter of the erosional feature is emulated using the approach of
Natural (primary) depletion with good aquifer support is sim-
Stephen, Clark and Gardiner (2001), and constrained to data
ulated in a reservoir produced by a single vertical well (Fig. 7).
from cores and logs taken from the Nelson Field (Appendix B).
Production rates (18 000 bpd) and time scales are set to be
In this approach, the sand quality is assumed to be highest at
close to those seen on the Nelson Field. As expected, the oil
the base and the centre of the channel and to decline upwards
is swept by an upward movement of the water leg in response
and outwards (Fig. 5a). Hence, kv /kh varies, depending on the
to the pressure drop due to production (Fig. 7). Water sweeps
position in the channel, with the lowest being in the channel
quickly through the relatively high-porosity zones, and the
margins. One realization of this geological modelling is shown
initial and final shapes of the water saturation volume are
in Fig. 5(b).
bounded by the geological architecture. Early in the produc-
tion sequence, oil also becomes bypassed in the less eroded
(lower porosity/net-to-gross) zones of the channel stack. The
Simulator to seismic modelling
flow approaches a steady state in terms of pressure at around
The flow simulation is chosen to have an identical scale to the time of the first 4D survey, and thus between the two
the geological model to avoid upscaling. Relative permeabil- survey dates the main effect is the upward water movement
ity curves are obtained from two-phase upscaling of small- with no further bypassing. The water front moves unequally
scale laboratory-based SCAL measurements. Models are then in three dimensions due to the channel complexity between the
run through the ECLIPSETM black-oil simulator to simulate front and back of the model. There is also an upward welling
two-phase (oil and water) flow with aquifer support. Produc- of the water front due to the strong pressure gradients at the
tion is induced by a single vertical well. After simulation, the well and, after the well hits high water-cut (breakthrough), the
resultant pressure and saturations, together with the geolog- back portion of the water front starts to move more rapidly.
ical and reservoir model properties, are used to calculate the Therefore, between the surveys, a greater saturation change is
seismic response. The acoustic properties of the reservoir flu- expected for the back of the model.
ids are determined by an identical scheme to that used in the Figure 8 shows the resultant time-lapse fluid saturation,
1D modelling. However, the rock-frame is now an average of pressure and seismic signatures for surveys carried out after
sand and shale stress-sensitivity parametrizations of MacBeth 3.5 years of production. Results for the back and front of
(2004) weighted by V shale (Appendix C). For each cell, stress- the model display quite different character due to channel


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 191

Figure 5 (a) Internal structure of a single channel showing the variation in erosion and corresponding contours of amalgamation ratio used
in current modelling. (b) Results from one realization of the stacked channel architecture, which is used as a basis for building the reservoir
model. Horizontal (upper left) and vertical (upper right) permeability distributions are derived for direct input into the reservoir model from
the geological model. Porosity variation is similar to net-to-gross, with NTG = 0.0 giving 14% porosity and NTG = 1.0 giving 23% porosity.
There are six channels in this realization, diverging away from the front face due to their sinuosity.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
192 C. MacBeth, K.D. Stephen and A. McInally

(a) plets at and below the main saturation changes, which appear
to track the areas of high net-to-gross. The patches of ‘no sig-
nal’ are zones where no discernible changes occur. These zones
of bypassed oil cannot be inferred, possibly as they are estab-
lished relatively early during the production. For the 30 Hz
seismic, there is a more continuous signature, although the
patchiness in amplitude still remains. The trough–peak sepa-
ration, and hence the oil–water contact movement, appears to
differ from that predicted by the 150 Hz seismic – a point that
will be returned to in a later section. The ideal trough–peak as
(b)
illustrated in the 1D modelling theory is observed on the crest
2630
and the lower flanks of the channel pattern. There is some
evidence of these signatures extending into the interbeds, even
for the 30 Hz seismic. This is more obvious for the back of
TWT (ms)

the model where the channel structure is more dispersed and


the pressure gradient more horizontal. Although the interbed
signature is prominent in the reservoir modelling, the low net-
to-gross has reduced the seismic visibility. These seismic signa-
2675 tures closely match the regions of maximum saturation change
(c) (upper left of Fig. 8). There are no obvious anomalies that ap-
pear to be associated with pressure. The 4D signature in the
2630
channels is clearly observed to be dominated by saturation
changes, thus supporting earlier work by Enterprise Oil. The
TWT (ms)

far-offset stacked response (not shown) produces very simi-


lar results. As larger offsets have a higher noise level in gen-
eral, this suggests that the far-offset seismic response may not
offer an immediate advantage in assessing the oil–water con-
tact movement.
2675
The modelling above assumes identical rock-frame pres-
Figure 6 Pre-production 2D seismic snapshots of a section taken sure dependence for sands and the shales, which is unlikely
through the front end of the channelized turbidite model in Fig. 5. to be the case. This assumption leads to a cancellation of their
Seismic are stacked responses for a range of near-offsets (0–2000 ft).
stress-sensitivity effects in the reflectivity coefficients, and thus
(a) Net-to-gross variation across the front of the model. (b) 150 Hz
seismic, revealing external and some internal detail. (c) 30 Hz seis- renders the response invariant to pressure. This point is also
mic corresponding to the peak frequency of the actual seismic data. brought out by the full-wave modelling. Another stress-related
Blue denotes a positive event at the top of the channel due to the issue is that in situ stress sensitivity may in fact be much lower
low-velocity Sele shale and the Forties sands. Red denotes a negative than that measured in the laboratory for sandstones (Fjaer and
event at the base. The well position is superimposed for comparison
Holt 1999) and also shales (Holt and Fjaer 2003). The impact
purposes. Both seismic plots have had the colour scale thresholded,
and as a consequence some small, low-amplitude events have been
of this possibility is examined by considering models for (i)
excluded. no shale stress sensitivity, and (ii) no sand and shale stress
sensitivity. Again, little change is observed from the seismic
in Fig. 8. The main reason for these results is the magnitude
distribution, but provide similar conclusions. The 4D re- of the effective pressure, which leads to the reservoir pres-
sponses are mainly enclosed by the channel structure and do sures sampling only the most stress-insensitive portion of the
not extend into the oil-bearing interbed regions. The Top For- stress curves (see Fig. 13). Indeed, visible changes in the 4D
ties events cancel out when the base and repeat seismic surveys signature appear possible only by adjusting Sκ and Sµ (see
are differenced; however, for the 150 Hz seismic, the single Appendix C) to large values close to unity. The result of this
trough–peak signature expected from the 1D modelling is re- is to compensate partially for saturation effects, thus reduc-
placed by a sequence of patchy and irregular trough–peak cou- ing the amplitude of the 4D signature. However, to create a


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 193

Figure 7 Results of forward simulation for natural depletion. The bottom diagram shows water cut and oil production over the full time
scale of the simulation. The saturation profiles above are for four particular times: top left – start-up (hence pre-production); top right – after
5.5 years (the first arrow marks the point at which the first 4D survey is taken); bottom left – after a further 7 years (the second arrow marks
the point at which the second survey is taken); bottom right – after a further 1.5 years. The data have been thresholded to exclude regions with
So > 0.75 in order to reveal the shape of the underlying ‘water front’. The well end point is five cells below the top of the model, and it produces
18 000 bpd.

significant stress-sensitivity effect would take a considerable


Heterogeneous channel model with no trapping
departure from the laboratory-measured results and a signif-
icant enhancement of stress sensitivity. This scenario is con- By ignoring capillary trapping effects for vertical flow us-
sidered unlikely in these turbidite sands, and thus saturation ing the rock relative permeabilities, particularly that of wa-
effects remain a dominant influence. ter through the oil–water mix, the sweep efficiency at the


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
194 C. MacBeth, K.D. Stephen and A. McInally

Figure 8 Time-lapse seismic, together with corresponding saturation and pressure signatures for production monitored over a 3.5-year period
beginning 3 years after start-up. Seismic signatures are differences of 2D sections through the front and back of the model, and are displayed as
true relative amplitude, normalized according to the largest signal. Reference picks for the top and base reservoir are based on the 150 Hz result
in Fig. 6. The time interval for seismic is the same as in Fig. 6, i.e. 2630–2375 ms.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 195

microscopic scale can be changed (Appendix A). This pro- in the water and lower in the oil, and a large pressure
vides a second model for study in which the bypassed zones drop is induced in the water–oil transition. The red band in
are now absent, and the sweep is now more efficient (Fig. 9). Fig. 10 is an expression of this pressure differential. The
The oil–water contact movement is now slower as the amount change in pressure softens the rock-frame, and hence coun-
of residual oil is now smaller, water cut is initially lower teracts the effects of saturation. The effect is not dramatic
and later higher, and production is higher after breakthrough. because of the range of effective pressures, which still ren-
Figure 10 shows the resultant 4D seismic signatures, to- ders the rock-frame fairly insensitive (the effective pressure is
gether with the saturation and pressure differences for the 17 MPa = 49 − 32 MPa as opposed to 42 MPa = 49 − 7 MPa
same survey dates as in the previous example. At face value, as in the case of no trapping). It must thus be concluded that
it is anticipated that the oil–water contact will be sharper, the differently connected ‘trapping’ versus ‘no trapping’ cases
with a cleaner transition in saturation and that the 4D sig- cannot be easily distinguished using only 4D seismic.
nature should therefore be stronger in comparison to the
trapping case (Fig. 8). Generally, however, the 4D signature
Homogeneous channel models
appears slightly weaker in parts although it remains sim-
ilar to the previous case. The cause of the weaker signa- For comparison purposes, another model is run in which the
ture is the time-lapse change in pressure due to the lower shale distribution within the channels is absent, which means
total fluid mobility. A larger pressure is required to push that kv = kh within every channel (Fig. 11). This may be con-
the water front through the oil, the pressure is thus higher sidered as an extreme end member in the general sequence

Figure 9 Thresholded saturation profiles (left-hand column) and oil production (right-hand column) for the previous case with bypassed oil
(bottom) and a model with adjusted relative permeabilities to provide no trapping of the oil and hence a low residual oil (top).


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
196 C. MacBeth, K.D. Stephen and A. McInally

Figure 10 Top: Saturation and fluid pressure differences for the revised reservoir model that produces no trapping. Bottom: 2D seismic time-lapse
signatures corresponding to this model, plotted as in Fig. 8.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 197

Figure 11 Saturation and pressure changes, together with corresponding 4D seismic signatures for the case of single porosity and permeability
in homogeneous sand channels, with no internal shale baffles. Plot format as in Fig. 8.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
198 C. MacBeth, K.D. Stephen and A. McInally

Reservoir simulator estimation 30 Hz seismic picks Figure 12 Estimates of the moved oil–water
contact from two different origins: satu-
ration changes output from the simulator;
and picking of the trough–peak signature di-
rectly from the 4D seismic. The 30 Hz seis-
mic time does not deliver the magnitude of
the sweep movement, but its amplitude does
detect the correct values. The 150 Hz seis-
mic delivers the opposite result. Left-hand
side: back of model; right-hand side: front of
model. Colour bars give the estimated oil–
water contact movement in metres. Dashed
lines indicate the approximate ‘field’ outline.

Reservoir simulator estimation 150 Hz seismic picks

30 Hz seismic amplitude 150 Hz seismic amplitude

of possible turbidite models. In this case, production now


Mapping the oil–water contact movement
gives rise to a strong pressure gradient around the well and
considerable water-front coning. The signature of the water Adequate detection of oil and water movement in turbidite
movement is much more evenly spread across the model, and reservoirs can help to refine and better constrain the reser-
the previous patchiness is now absent. Thus relative channel voir model and determine the optimum locations for addi-
connectivity may be inferred by the continuity of the 4D signa- tional production wells or injectors. Maps of oil–water contact
ture. Unfortunately, despite the strong pressure and saturation movement form an important link in providing supplemen-
differences, the interchannel fluid effects still remain largely tary data on the reservoir model, in addition to production
undetected. logging tool data and time-domain transmissometry data. For


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 199

Figure 13 A selection of the stress-


sensitivity measurements made in the Hoek
cell under isotropic loading conditions.
The initial effective (differential) stress in
the Nelson reservoir is shown for reference
(dashed line).

example, for any given value of stock tank oil initially in place sufficient water to support flow. This condition could be fully
(STOIIP), a smaller oil–water contact movement is anticipated assessed by the well pressures and quantitative 4D seismic. It is
for a reservoir with a more efficient sweep compared with one therefore desirable to use 4D seismic in a quantitative capacity,
in which there is a bypassing of oil by preferential pathways. to map the oil–water contact movement from seismic estima-
Alternatively, if the STOIIP or pore volume is higher than ini- tion and help production data constrain the reservoir model.
tially anticipated in the reservoir model, the oil–water contact Picks of the two-way interval time between the trough and
sweep will also be larger than that predicted. Quantitative 4D peak of the signature, visible across the seismic volume, are one
seismic estimates could be used together with material balance way of providing information on the oil–water contact move-
calculations using production data (if available) to assess these ment. This attribute is then converted into movement in me-
conditions. Other possibilities for a greater than predicted oil– tres and mapped across the reservoir for both the 150 Hz and
water contact movement in the field compared with the model 30 Hz seismic (Fig. 12). As there is no such feature as a water
could be that the aquifer connectivity is low and there is in- front per se, movement predicted from the simulator is based


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
200 C. MacBeth, K.D. Stephen and A. McInally

on evaluation of saturation changes. Thus, for the purposes of directly indicating movement. Exact assessment for individ-
this exercise, movement is calculated by evaluating the thick- ual reservoirs depends on the frequency of the seismic and the
ness of the thresholded saturation-difference volume, where thickness of the sand interval. Both 30 Hz and 150 Hz seis-
only saturation changes above 0.4 are accepted (Fig. 12). It is mic can adequately map the aerial drainage patterns, and thus
observed that the 30 Hz seismic provides a good continuous 4D can be used for optimal positioning of the injectors and
areal resolution of the regions of change, but lacks quantita- producers for drainage. The trough–peak indicator also high-
tive definition due principally to strong tuning effects. Thus, lights areas that are not drained effectively by the producer
whilst the seismic time picks between trough and peak for the well, through lack of continuity in the signature. A more con-
moved oil–water contact remain largely invariant across the tinuous trough–peak signature indicates higher permeabilities
field, the amplitude of the signature demonstrates a good re- within the individual channels. The 4D signal remains strong
lationship to oil–water contact movement. In comparison, the in the lower flanks and could be used to refine the reservoir
150 Hz seismic accurately detects the movement, although it model and determine the optimum locations for additional
also appears to extend the areal coverage, presumably related production wells or injectors. However, it is not possible to
to interferences or multiples. The magnitude of the picked detect production changes in the interbed channel margins,
trough amplitude bears no relation to the sweep; indeed, it even when there is strong connectivity in the reservoir. For
predicts a minimum in the position of a maximum. If the thick- ‘Nelson’ types of reservoir, the rock-frame stress-sensitivity ef-
ness of a reservoir interval is unknown, this analysis suggests fects and also fluid-compressibility effects are small, and hence
that an attribute formed as the product of time thickness and saturation is the dominant cause of the 4D signal. Whilst this
amplitude, such as the pseudo-elastic impedance attribute of does keep some aspects of interpretation simple, it also means
McInally et al. (2003), might form a suitable low-risk indica- that it is difficult to distinguish the overall sweep efficiency
tor of water movement. from seismic alone. There are, however, some indications that
continuity of the water front may provide a qualitative guide.
The monitoring of changes due to production is important
CONCLUSIONS
in the understanding of reservoir drainage and in evaluating
The 4D signature of natural (primary) production is mod- the ultimate reservoir recovery. Overall, this study has shown
elled for a producing turbidite reservoir similar to the Nelson that measurements of the seismically derived oil–water con-
Field, central North Sea. Particular focus is placed on eval- tact movement can be used to constrain the reservoir model as
uating the movement of the ‘oil–water contact’, and in de- part of history-matching, provided the 4D signal is interpreted
tecting the characteristic trough–peak seismic indicator from with the reservoir architecture in mind.
the difference volume. The interaction between the 3D geo-
logical structure, smaller-scale internal heterogeneity, and the
ACKNOWLEDGEMENTS
flow and seismic properties is complex. Understandably, the
4D seismic method is found to work best for the main sand This work was funded by the Edinburgh Time-Lapse Project
fairways and can therefore be used to target potential upside (Phase I) and is published with the approval of its sponsors: BP,
areas. Despite the reservoir complexity, water-front movement Concept Systems, Enterprise Oil, Fairfield Inc., Schlumberger,
can still be detected in variable net-to-gross regions. Indeed, Shell Exploration, Statoil, TotalFinaElf and Landmark. We
Redondo-López et al. (2002) found a correlation between net- also thank the Nelson joint venture partners: Shell, Exxon-
to-gross and the magnitude of sweep over the Nelson Field, Mobil, Total, Intrepid, Svenska & Summit, for permission to
and the presence of mega-cones due to clean zones with high use and show data from the Nelson Field. We thank MSc stu-
net-to-gross. The 1D modelling here suggests, however, that in dent John Oyewole for help in preparing the core data. We are
low net-to-gross regions with complicated fine-scale interbed- grateful to Andy Gardiner and Patrick Corbett for their sup-
ding, the basic signal shape is preserved but the overall am- port and discussions on the geological aspects of this work.
plitude is degraded and additional calibration is required. 3D
modelling appears to suggest that amplitude is the prominent
indication of movement at seismic frequencies, whilst time REFERENCES
separation between the trough and peak may not be satis- Al-Maskeri Y. and MacBeth C. 2003. Model controlled quantitative
factory except at high frequencies. Tuning is the root cause interpretation of 4D signatures in turbidite sands. 6th SGBF con-
of this effect, and compromises the use of picked times for ference, Rio de Janeiro, Time-lapse session.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 201

Alsos T., Tondel R., Aanvik F. and Solheim O.A. 2003. Quantifying Ringrose P.S., Sorbie K.S., Corbett P.W.M. and Jensen J.L. 1993. Im-
rise in gas–water contact from time-lapse seismic on the Sleipner miscible flow behaviour in laminated and cross-bedded sandstones.
Ost field. 65th EAGE conference, Stavanger, Norway, Extended Journal of Petroleum Science and Engineering 9, 103–124.
Abstracts, A08. Santos R.A., Sobrinho A.A., Pinto A.C., Dillon L.D., Ritter G.L.S.,
Batzle M. and Wang Z. 1992. Seismic properties of pore fluids. Geo- Silva E.F.F. et al. 2003. 4D integrated technologies for deep-water
physics 57, 1396–1408. turbidite reservoirs: from petrophysics to fluid flow simulation.
Beydoun W., Kerdraon Y., Lefeuvre F., Bancelin J.P., Medina S. and Petroleum Geoscience 9, 73–84.
Bleines B. 2002. Benefits of a 3DHR survey for Girassol Field ap- Stephen K.D., Clark J.D. and Gardiner A.R. 2001. Outcrop-based
praisal and development, Angola. The Leading Edge 21, 1152– stochastic modelling of turbidite amalgamation and its effects on
1155. hydrocarbon recovery. Petroleum Geoscience 7, 163–172.
Boyd-Gorst J., Fail P. and Pointing L. 2001. 4-D time lapse reservoir Stephen K.D., Clark J.D. and Pickup G.E. 2002. Modelling and flow
monitoring of Nelson Field, central North Sea: Successful use of an simulations of a North Sea turbidite reservoir: sensitivities and
integrated rock physics model to predict and track reservoir pro- upscaling. Proceedings of the 13th European Petroleum Confer-
duction. The Leading Edge 20, 1336–1350. ence, Aberdeen, UK, SPE No. 78292.
Burkhart T., Hoover A.R. and Flemings P.B. 2000. Time-lapse (4-D)
seismic monitoring of primary production of turbidite reservoirs at
South Timbalier Block 295, offshore Louisiana, Gulf of Mexico.
APPENDIX A
Geophysics 65, 351–367.
Upscaling relative permeability
Dromgoole P., Bowman M., Leonard A., Weimer P. and Slatt R.M.
2000. Developing and managing turbidite reservoirs – case histories Relative permeability and capillary pressure curves in multi-
and experiences: results of the 1998 EAGE/AAPG research confer- phase flow simulations vary depending on pore-scale geom-
ence. Petroleum Geoscience 6, 97–105.
etry and wettability, on the interaction of lamina and bed-
Fjaer E. and Holt R.M. 1999. Stress and stress release effects on acous-
tic velocities from cores, logs and seismics. 40th SPWLA Annual scale heterogeneity and also on the degree of compensation
Logging Symposium, paper WW. required for numerical dispersion. These properties, however,
Hamon G. 2003. Two-phase rock-typing: another perspective. Pro- will determine saturations and pressures during the produc-
ceedings of the Annual SPE Meeting, SPE No. 84035. tion process, and hence the sensitivity of time-lapse seismic
Holt R.M. and Fjaer E. 2003. Wave velocities in shales – a rock physics
and whether they can be detected. The current work con-
model. 65th EAGE conference, Stavanger, Norway, Extended Ab-
stracts, C13.
siders two scenarios with shales in the channels. In the first
Kennett B.L.N. 1983. Seismic Wave Propagation in Stratified Media. scenario (Fig. 8), we assume that capillary trapping effects
Cambridge Monographs on Mechanics and Applied Mathematics. are strong. In this case, when the water front reaches a shale
Cambridge University Press. layer, the capillary force rapidly draws water into the shale
MacBeth C. 2002. Multi-Component VSP Analysis for Applied Seis- and sweeps mobile oil faster than the remaining oil in the up-
mic Anisotropy. Handbook of Geophysical Exploration. Pergamon
stream sand layer can be fully swept. The shale then becomes
Press, Inc.
MacBeth C. 2004. A classification for the pressure sensitivity proper- a barrier to any further movement of oil and any remaining oil
ties of a sandstone rockframe. Geophysics 69, 497–510. is bypassed. Here, this effect is treated by upscaling using the
McInally A.T., Redondo-López T., Garnham J., Kunka J., Brooks capillary equilibrium method (e.g. Pickup and Stephen 2000),
A.D., Strenstrup-Hansen L., Barclay F. & Davies D. 2003. Opti- which produces a pseudo-relative permeability and capillary
mising 4D fluid imaging. Petroleum Geoscience 9, 91–101.
pressure curve with increased effective residual oil saturation.
Parr R., Marsh J. and Griffin T. 2000. Interpretation and integration
of 4-D results into reservoir management, Schiehallion Field, UKCS.
Stephen, Clark and Pickup (2002) derived the curve we use
70th SEG meeting, Calgary, Canada, Expanded Abstracts, 1464– in this work for vertical flow as an end member for both of
1467. our genetic units. In the second scenario (Fig. 10), the relative
Pickup G.E. and Stephen K.D. 2000. An assessment of steady-state permeability and capillary pressure curves are obtained from
scale-up for small-scale geological models. Petroleum Geoscience the Nelson Field data for vertical flow, therefore ignoring trap-
6, 203–210.
ping. In both cases, the Nelson curves were used for horizontal
Postma G.W. 1955. Wave propagation in a stratified medium. Geo-
physics 20, 780–806. flow.
Redondo-López T., McInally A.T., Brooks A.D. and Kunka J. 2002.
Quantitative integration of 4D seismic for field development. 87th APPENDIX B
AAPG Annual Meeting, Houston, USA, vol. 11, A146.
Reid F.J.L., MacBeth C. and McInally A.T. 2003. 4D signal enhance- Conditioning of cell-scale kv /kh using the Nelson data
ment using singular value decomposition: OWC movement on the
Nelson field. 65th EAGE conference, Stavanger, Norway, Extended Porosity and permeability data in the 3D geological model
Abstracts, A11. are distributed by a combination of simple averaging and


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
202 C. MacBeth, K.D. Stephen and A. McInally

Table 2 Permeability in millidarcies (mD) and porosity values obtain The above relationships are applied to the channel model
from data core by first assuming a linear decrease in θ and A upwards and
outwards from the centre of the channel and a value is assigned
kh (mD) kv (mD) Porosity
to each cell. These linear trends are constrained by the values
obtained from core for the interbeds (θ = 0.3, A = 0) as well
Sand 331.70 250.00 0.24
Shale 0.05 0.05 0.10 as a maximum value obtained for the channels (θ = 0.98,
A = 0.8). The amalgamation ratio is estimated by counting
the number of sand-on-sand contacts for the total number
of sand beds present in the channels. θ is also assumed to
augmentation of the results of Stephen et al. (2001). There are vary stochastically according to a normal distribution with
two genetic units in the model: interbeds and channel sands, the same standard deviation that is applied to the interbeds.
and each is treated separately. kv and kh are then obtained from the above log-linear and
The interbeds constitute the background genetic unit and linear relationships, respectively. To account for variability of
are modelled at the small scale as alternating laterally extensive bed permeability and thickness variation, kv and kh are further
beds of medium-to-good-quality sand and virtually imperme- modified stochastically by application of a normal distribution
able shale. From core data, an average net-to-gross thickness to the grid-cell values with coefficients of variation of 0.01 and
ratio θ of 0.3 was obtained in addition to permeability values 0.02, respectively. This follows results obtained numerically.
for both facies and for each direction (Table 2). θ is assumed Interbed and channel sand beds are observed to have very
to vary laterally and vertically about the mean in each cell ac- similar petrophysical properties.
cording to a normal distribution with a standard deviation of
0.01, and a value is then assigned to each cell stochastically.
APPENDIX C
The thickness-weighted (i.e. θ -dependent) arithmetic and har-
monic averages are used to determine the interbed kh and kv Upscaling the stress sensitivity of the rock-frame
values, respectively, for each cell using the average bed perme-
To guide the assignment of rock-frame stress sensitivity for
abilities (Table 2). The porosity is similarly upscaled using a
this work, ultrasonic velocity measurements are made whilst
thickness-weighted arithmetic average of porosities. To allow
isotropically loading the rock-frame in the laboratory. Here,
for variation in permeability between sand beds, both later-
only sandstones are available to be tested in this way, and
ally and vertically, permeabilities are then modified stochasti-
consequently the contribution from the shales is unknown.
cally, assuming a normal distribution for each direction with a
Figure 13 shows a selection of the rock-physics stress-
coefficient of variation of 0.01.
sensitivity results. These measurements are then transformed
Channel sands are treated in a similar manner. Initially, the
into bulk and shear modulus variations and fitted by the sig-
channel features are distributed as objects within the interbed
moidal pressure relationship of MacBeth (2004):
background, with successive channels eroding those deposited
κ∞
below. The deposition of sand and shale beds within the chan- κ(P) = (C1)
1 + (Sκ /(1 − Sκ )) e−P/Pκ
nels follows the model of Stephen et al. (2001), where suc-
cessive sand beds may erode previously deposited shale. Their and
2D model is extended to three dimensions, and the effects of µ∞
µ(P) = , (C2)
varying bed thickness and permeability on upscaled kv and kh 1 + (Sµ /(1 − Sµ ))e−P/Pµ
are investigated. Similar to previous results, it is found that where κ ∞ and µ∞ are high-pressure asymptotes, Pκ and Pµ
the log of the upscaled kv depended linearly on the amalga- are constants, and Sκ and Sµ define the maximum possible
mation ratio A, which quantifies the degree of shale erosion. A change in the moduli. The average stress-sensitivity param-
stochastic component is also observed, reflecting the bed per- eters are determined by fitting to the six available samples,
meability and thickness variation. A similar linear relationship the results of which are shown in MacBeth (2004). Based on
exists for kh . The kv log-linear trend is constrained by kv at A = these values, the maximum stress sensitivity (at zero effective
0 (the harmonic average of vertical sand and shale vertical per- stress) for the Nelson sands is estimated to be approximately
meabilities) and kv at A = 1 (the vertical sand permeability); 9% per MPa; however, this sensitivity rapidly diminishes with
similarly for kh as a linear trend with A, using the arithmetic increasing effective pressure due to the non-linearity of the
average. velocity–stress curve. Nelson is a normally pressured reservoir,


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203
4D signature of oil–water contact movement 203

with initial pressure of 23 MPa, decreasing to 16 MPa after or


production. Given an overburden pressure of 50 MPa (assum-
 = Vshale shale + (1 − Vshale ) sand . (C4)
ing a lithostatic gradient of 1 psi/ft), the effective (differential)
stress changes from 27 MPa to 34 MPa. Laboratory measure-
Thus, using the factor V shale as a multiplier, for strictly vertical
ments therefore predict that the frame-stiffening component
raypaths, the average elastic moduli are
will fall to 3% per MPa and reduce to only 1% per MPa af-  −1
ter production. This should be compared with the impact of Vshale (1 − Vshale )
(λ + 2µ) = + , (C5)
the fluid component, evaluated to be 43% for a change from (λ + 2µ)shale (λ + 2µ)sand
an oil- to water-sand. Thus, under conditions of a negligible  −1
stress response, the fluid component dominates the velocity
Vshale (1 − Vshale )
µ= + (C6)
µshale µshale
and impedance variations. This prediction is in agreement with
the work of Boyd-Gorst et al. (2001) and is borne out in the and hence
modelling. Although the frame effect is likely to be small, it is 4
κ = (λ + 2µ) − µ̄, (C7)
still taken into account in this work for completeness as it is 3
not completely negligible. where λ and µ are the Lamé parameters. The elastic moduli can
The above conclusions are made from measurements on be the full pressure-dependent moduli of (C1) and (C2), and
core sample size (3.5 cm diameter by 6.5 cm length) elements combine to form the composite stress sensitivity of the sand
of a relatively intact homogeneous rock. To upscale these to plus shale in each cell of the geological model. This particular
the properties of a single grid cell (25 × 25 × 2 m), use is averaging is valid provided the raypaths through the reservoir
made of the same geological model that also underlies the zone are travelling subvertically. However, the fine structure of
permeability calculations in Appendix B. Here the turbidite turbidite deposits with sheet-like sands separated by discrete
system is modelled as a sequence of discontinuous but paral- shales will clearly give VTI anisotropy, and this average will
lel, horizontal layers of shale embedded in sand. As the cell not be correct for far-offset raypaths. A simple adjustment
thickness is sub-seismic (2 m thick versus 100 m wavelength), for this effect can be implemented so that isotropic two-term
vertical averaging of the sand and shale properties can be amplitude-variation-with-offset will closely resemble the VTI
achieved using the conventional long-wavelength equivalent- counterpart. This is achieved by incrementing the averaged
medium theory of Postma (1955), which is appropriate for shear modulus by Thomsen’s δ parameter, calculated for the
two cyclic constituents. For each cell, it is appropriate to use fine-layered system (see MacBeth 2002, p. 37). Thus
V shale (a thickness-weighted metric) to average the elastic prop-
δ
erties. In the modelling, given either a porosity or net-to-gross µ→µ+ (λ + 2µ). (C8)
4
(NTG) value, it is possible to back-calculate for V shale from
As δ is typically 0.01, and µ = 21 GPa and λ + 2µ = 19 GPa,
either
this transformation affects the modulus by only a few percent
Vshale = 1 − NTG (C3) at most in this particular example.


C 2005 European Association of Geoscientists & Engineers, Geophysical Prospecting, 53, 183–203

You might also like