Vous êtes sur la page 1sur 22

Hydrodynamic aquifer or

reservoir compartmentalization?
Ann Muggeridge and Hisham Mahmode

ABSTRACT
Changes in oil-water contact (OWC) depth across a field are
often a signature of reservoir compartmentalization as a result
of faulting or stratigraphic barriers, such as shales, but may
equally well be caused by an underlying hydrodynamic aquifer. In particular, the pressure-depth data obtained from an
aquifer whose flow is changing over time can look very similar
to that obtained from a compartmentalized reservoir. Misunderstanding which of these mechanisms causes the observed changes in OWC across the field may result in poor
estimates of oil in place and reduced recovery. To address this
problem, an analytic expression is presented to estimate the
time taken for a steady state tilted OWC to be established once
an aquifer starts flowing. A comparison with simulations of
hydrodynamic aquifers in homogeneous, compartmentalized,
and heterogeneous reservoir models shows that this expression can be used in combination with the one derived by M. K.
Hubbert, for the steady state tilt of the OWC, to clarify whether
a reservoir contains barriers or baffles to flow or may simply
have not yet reached equilibrium.

INTRODUCTION
Changes in oil-water or gas-water contact depth across reservoirs have been observed in many fields around the world.
These are typically observed by analysis of data from repeat
formation testers (RFTs, Goetz et al., 1977) or modular formation dynamics testers (MDTs, Badry et al., 1993) and may
be associated with spatially varying pressures in the hydrocarbon column or the aquifer and sometimes spatially varying
temperature. These changes in contact depth may indicate

Copyright 2012. The American Association of Petroleum Geologists. All rights reserved.
Manuscript received October 28, 2010; provisional acceptance April 5, 2011; revised manuscript received
May 10, 2011; final acceptance June 14, 2011.
DOI:10.1306/06141110169

AAPG Bulletin, v. 96, no. 2 (February 2012), pp. 315336

315

AUTHORS
Ann Muggeridge  Imperial College
London SW7 2AZ, United Kingdom;
a.muggeridge@ic.ac.uk
Ann Muggeridge is a reader in reservoir physics
in the Department of Earth Science and Engineering, Imperial College, London. She holds a
B.Sc. degree in physics from Imperial College
and a Ph.D. in atmospheric physics from the
University of Oxford. Her research interests include reservoir characterization, determining the
impact of reservoir heterogeneity on fluid flow,
and enhanced oil recovery processes.
Hisham Mahmode  Imperial College
London SW7 2AZ, United Kingdom; present
address: BP Exploration and Production,
Wareham, Dorset BH20 5JR, United Kingdom;
hisham.mahmode08@imperial.ac.uk
Hisham Mahmode is currently a petroleum
engineer working on the Wytch Farm oil field.
He worked as a reservoir engineer at RPS Energy
before joining BP. He holds a B.Eng. degree in
civil engineering from City University, London,
and an M.Sc. degree in petroleum engineering
from Imperial College, London. His interests include artificial lift, well performance, and understanding dynamic reservoir behavior.

ACKNOWLEDGEMENTS
We thank Schlumberger-GeoQuest for providing
the reservoir simulation software used in the
course of this work. We also thank Jason Go for
his helpful comments on the manuscript.
The AAPG Editor thanks the following reviewers
for their work on this paper: John B. Curtis and
William A. Hill.

either barriers to flow within the reservoir (e.g.,


Weber, 1987; Brehm, 2003; Guscott et al., 2003;
Muggeridge et al., 2005; Talukdar and Brusdal,
2005; Sweet and Sumpter, 2007; Underschultz
et al., 2008; Bakker et al., 2009) or a hydrodynamic aquifer (e.g., Pelissier et al., 1980; Zawisza,
1986, 2004; Berg et al., 1994; Thomasen and
Jacobsen, 1994; Dennis et al., 2000; Underschultz,
2005; Tozer and Borthwick, 2010), although in
the Ghawar oil field in Saudi Arabia, they have also
been ascribed to thermal convection (e.g., Stenger,
1999; Stenger et al., 2001).
Distinctly different oil-water contact (OWC)
depths, otherwise known as perched OWCs, are
associated with reservoir compartmentalization and
often occur in reservoirs in formations that have
undergone significant faulting (Weber, 1987; Brehm,
2003; Guscott et al., 2003; Talukdar and Brusdal,
2005; Bakker et al., 2009). In the Niger Delta, oil
appears to have migrated into these traps via the
fault system, probably during one of several faulting episodes (Weber, 1987). The OWCs may also
be associated with changes in observed oil pressures
in different wells or abnormal reservoir pressures.
Continuously tilting OWCs occur in reservoirs
underlain by hydrodynamic aquifers (e.g., Pelissier
et al., 1980; Zawisza, 1986, 2004; Berg et al., 1994;
Thomasen and Jacobsen, 1994; Dennis et al., 2000;
Underschultz, 2005; Tozer and Borthwick, 2010).
These very slow subsurface flows are commonly
natural, resulting from meteoric waters recharging
aquifers via outcrops (Hubbert, 1967) or expulsion
of water from porous sediments during basin subsidence (Hubbert, 1967; Grosjean et al., 2009),
sometimes combined with aquifer discharge at surface outcrops (Tozer and Borthwick, 2010). In these
cases, the reservoir is commonly normally pressured and the observed oil pressure is constant between wells. Occasionally, tilted contacts are man
made, caused by pressure depletion in neighboring
fields (e.g., Van Kirk, 1976; Coutts, 1999; Hortle
et al., 2010).
Stenger (1999) and Stenger et al. (2001) proposed that tilted contacts may be caused by lateral
temperature gradients within reservoirs. They demonstrated that temperature gradients within the
Haradh Arab D reservoir of the Ghawar field cor316

Hydrodynamic Aquifer or Reservoir Compartmentalization?

related with varying oil densities across the field and


hypothesized that these, in turn, resulted in changing
OWC depths. Natural convection was invoked as a
mechanism for preventing gravitational overturning
(as described by England et al., 1995), equalizing
these horizontal density gradients. However, they
did not discuss the possibility that the lateral temperature gradients may actually be an indication of
a hydrodynamic aquifer (Anderson, 2005).
Being able to distinguish between these possibilities during appraisal is important because
they will result in a different topology of the contact depth across the prospect (and thus different
values of estimated hydrocarbons in place) as
well as different models of the reservoir connectivity. Undiagnosed reservoir compartmentalization can have a significant adverse impact on oil
recovery (Dromgoole and Speers, 1997; Smalley
and Muggeridge, 2010). In contrast, it is generally
possible to mitigate the impact of such compartmentalization on recovery provided that it is identified during appraisal (e.g., Talukdar and Brusdal,
2005; Bakker et al., 2009). Aquifer hydrodynamics
and potential barriers to flow are also important
considerations when designing subsurface carbon
dioxide storage schemes (Bachu et al., 1994; Hortle
et al., 2010; Larkin, 2010).
Unfortunately, distinguishing between changes
in contact depth resulting from compartmentalization and those resulting from a hydrodynamic
aquifer can be difficult, particularly if the lateral
pressure gradients causing aquifer flow have changed
in the recent past (e.g., Underschultz, 2005; Hortle
et al., 2010). A typical signature of compartmentalization is assumed to be different oil pressures in
different parts of the reservoir, but this can also be
evidence that the system has not yet reached steady
state (Dennis et al., 2000; Dennis et al., 2005;
Underschultz, 2005). Similarly, the existence of a
horizontal pressure gradient in the aquifer but no
such gradient in the oil leg is commonly assumed to
be indicative of a hydrodynamic aquifer and good
lateral communication; however, it is equally possible that pressures may have equilibrated through
a low-permeability baffle on geologic time scales
but would not equilibrate through such a baffle
on production time scales (Dennis et al., 2000;

Muggeridge et al., 2005). Areal variations in aquifer


permeability may result in significant variations in
contact tilt across a field (Dennis et al., 2005), which
might be interpreted as being caused by barriers to
flow when in fact, good communication exists.
Although many articles describe field evidence
for hydrodynamic aquifers, few articles investigating the influence that reservoir and aquifer properties and changing aquifer flux have on contact
tilt have been published since the classic work of
Hubbert (1953). Hubbert (1953) provided a relationship that related the fluid and rock properties
and the hydrodynamic head to the final steady state
inclination of the oil-water (or, indeed, gas-water)
contact; however, he did not provide any expression for determining the time scales over which
that steady state is established. Dennis et al. (2000,
2005) showed that it could take 20 k.y. for steady
state to be established after the onset of aquifer
flux in a model of the Pierce field (central North
Sea, United Kingdom). Underschultz (2005) estimated, using Darcys law, that it should take between 46 k.y. and 110 k.y. to reach steady state in
the Zeepard and Griffin (North West Shelf, Australia) reservoirs, respectively; however, geologic
evidence suggested that the aquifer pressure field
had last changed at 5.6 m.y. ago. To achieve this time
scale, they had to reduce the aquifer permeability
from 500 md (490 1015 m2) to 10 md (9.9
1015 m2) or invoke capillary barriers to flow.
In this article, we used an analytic expression,
originally derived by Gardner et al. (1962) for the
gravitational overturning of miscible fluids, and later
applied to gravitational overturning in oil reservoirs by England et al. (1995), to estimate the time
scales for tilted OWCs to reach steady state after an
aquifer starts or stops flowing. This expression is
compared with the time scales that were predicted
using numerical simulation to ensure that the reservoir simulation was properly modeling the very
low flow rates in the aquifer and the response of
the contact. Numerical simulation was then used
to investigate the impact of a range of reservoir
heterogeneities, such as a low-permeability fault, a
water-wet fault forming a capillary barrier to oil
flow, a tar mat, and a shale, on the final steady state
geometry of the OWC and the time scales to reach

that steady state. The results suggest that the time


scales for tilted contacts to be established after an
aquifer starts flowing or to relax back to the horizontal after the aquifer stops flowing may be several
hundred thousand years. Examination of the associated pressure-depth profiles highlights the fact that
use of such data in isolation may suggest that a reservoir is compartmentalized when it is still relaxing
back to steady state or that good communication
on production time scales exists when in fact this is
only the case on geologic time scales. Estimation of
time scales to reach steady state using the analytic
expression given here in combination with regional
investigations into changes in aquifer flux may reduce this uncertainty and highlight when further
detailed modeling or data acquisition is required.

HUBBERTS STEADY STATE ANALYSIS


First, we review Hubberts (1953) steady state
analysis for the tilt of the OWC resulting from a
hydrodynamic aquifer flowing at a constant rate in
the absence of capillary pressure. As discussed in
the introduction, this flow may result from meteoric waters entering the aquifer via an outcrop or
discharge of an aquifer at the surface.
Figure 1 shows a vertical cross section through
an oil reservoir underlain by a hydrodynamic aquifer. A pressure gradient across the aquifer in the
reservoir, DPaq, resulted in an aquifer flux from left
to right with a mean interstitial velocity in the
aquifer of v (in meters per second). By Darcys law,
the mean interstitial velocity of water in the aquifer,
v, is given by
v

kres DPaq
mw L

where kres is the effective aquifer permeability for


the whole reservoir (in square meters), L is the
length of the reservoir in the direction of aquifer
flow (in meters), and mw is the water viscosity (in
pascal-seconds).
Two vertical appraisal wells in the reservoir are
separated by a distance, L12 m. By Darcys law, the
observed pressure difference between the wells in
Muggeridge and Mahmode

317

Figure 1. A schematic diagram


showing the vertical cross section
through an oil reservoir underlain by a hydrodynamic aquifer
flowing from left to right.

the aquifer will be related to the interstitial water


velocity by
DP12 P1  P2

nmw L12
k12

where k12 is the permeability of the rock in the


aquifer between the wells (in square meters) that
may be different locally from that over the whole
field because of heterogeneity.
As the water is flowing very slowly in the aquifer,
we can assume that the vertical pressure gradient
at any location is hydrostatic (Coats et al., 1971).
In this case, we can write
P1 P0 + ro gh1 + rw gH  h1
P2 P0 + ro gh2 + rw gH  h2

where the OWC has been measured at a depth of


h1 (in meters) below the reservoir top in well 1 and
at a depth of h2 (in meters) below the reservoir top
in well 2. We have assumed a datum pressure, Po
(in pascals), at the top of the reservoir, which is H
(in meters) above the depth of observation. The
density of oil is ro (in kilograms per cubic meter)
and the density of water is rw (in kilograms per cubic meter). The acceleration caused by gravity is g
(in meters per square second).
Thus, substituting equation 3 into equation 2
and combining the result with equation 1, we obtain an expression relating the angle of dip of the
contact, q, to the water-oil density difference, Dr,
the effective permeability seen between wells in
318

Hydrodynamic Aquifer or Reservoir Compartmentalization?

the aquifer, k12, the water viscosity, mw, the interstitial velocity of the water in the aquifer, v, and
the acceleration caused by gravity, g:
tan q

h2  h1
keff DPaq

k12 Drg L
L12

This is the same as the original expression derived by Hubbert (1953), except that he expressed
the dip in terms of the gradient of the potentiometric surface in the horizontal direction.
Examining equation 4, we see that for a constant pressure drop across the reservoir, contact
tilt will be greater when the density difference between the hydrocarbon and water is lower. The
contact tilt will be higher in regions of the aquifer
where permeability is lower and lower in regions
of higher permeability. The OWC tilts have been
observed ranging from 3 to 200 m/km equivalent
to angles of dip ranging from 0.1 to 10 (Dennis
et al., 2000). The resulting interstitial velocities in
the aquifer are on the order of centimeters per year.

TRANSIENT ANALYSIS
Although hydrodynamic aquifers are common in
hydrocarbon reservoirs (e.g., Pelissier et al., 1980;
Zawisza, 1986, 2004; Berg et al., 1994; Thomasen and
Jacobsen, 1994; Dennis et al., 2000; Underschultz,
2005; Tozer and Borthwick, 2010), it is highly unlikely that they have flowed at the same rate since

the reservoir filled (Dennis et al., 2000, 2005;


Underschultz, 2005). Aquifer flux will be affected
by any geologic events that change the regional
pressure distribution; for example, erosion can result in flow in a previously static aquifer, whereas
subsidence or uplift may cause horizontal pressure
gradients within a basin. In addition, human activity can result in changing regional pressure gradients when, for example, a nearby oil field is depleted (Hortle et al., 2010) or an aquifer is used to
supply water to nearby cities. It is, therefore, of interest to establish how long it takes for a tilted contact to reach the steady state described by Hubbert
(1953) or, indeed, for such a contact to relax back
to the horizontal when aquifer flux ceases.
Aquifer flux through a reservoir is driven by
the regional pressure gradient in the aquifer. This
results in a horizontal pressure gradient being established in the aquifer underneath the oil reservoir. This occurs very rapidly on a geologic time
frame (a few years, Muggeridge et al., 2005). These
horizontal pressure changes are also rapidly transmitted vertically into the oil leg, so that initially, a
horizontal pressure gradient exists in the oil leg,
although the contact is still horizontal. As a result
of the horizontal pressure gradient in the oil leg, the
oil starts to flow. In some cases, this can lead to
the shifting of the oil pool to a position away from
the original structural trap, resulting in a so-called
hydrodynamic trapping (e.g., Berg et al., 1994).
This may also be an important consideration during
geologic sequestration of carbon dioxide in saline
aquifers (Bachu et al., 1994; Larkin, 2010). More
commonly, the oil simply moves within the original
trap, resulting in the OWC tilting until, at steady state,
the horizontal pressure gradient in the oil is zero.
Gardner et al. (1962) derived an analytic expression to describe the time over which an initially vertical contact between two miscible fluids
of different densities returned to the horizontal.
Following England et al. (1995), for fluids whose
viscosity ratio is less than 10, the time t (in seconds)
taken for the interface to become approximately
horizontal is given by
 2
L
25f
mH
t
H
4kh gDr

where H is the reservoir thickness (in meters), f is


the porosity, kh is the horizontal permeability (in
 is the arithmetic mean vissquare meters) and m
cosity of the two fluids (in pascal-seconds). Note
that the time scale is controlled by the horizontal,
instead of the vertical, permeability because at a later
time, overturning is dominated by horizontal flow.
Equation 5 was used by England et al. (1995) to
assess the connectivity between Forties and South
East Forties (central North Sea, United Kingdom).
They did this by comparing the time needed for
the two oils of different densities to overturn in the
absence of barriers to flow as determined by the
Gardner et al. (1962) expression with the actual time
since the reservoir filled. Their calculations showed
that the observed differences in oil density should
have equalized within approximately 100 k.y.,
whereas basin modeling suggested that the reservoir
had filled more than 1 m.y. ago. Numerical modeling showed that the present-day observed oil density
difference could only be preserved over these time
scales if a reduction in transmissibility by a factor of
0.01 exists between the two regions of the reservoir.
England et al. (1995) also examined the time
that it would take for an initially tilted OWC to relax
back to the horizontal, assuming a static aquifer and
a baffle to flow that was 0.01 the permeability of
the reservoirs. They showed that the time scale for
overturning of the contact was much quicker than
that for the gravitational overturning for the different oils because of the higher density contrast
between oil and water. In the case of Forties and
South East Forties, this meant that the two reservoirs would have the same OWC depth, although a
low transmissibility baffle exists between that had
prevented mixing of the oils.
In the rest of this article, we shall examine
whether the same expression (equation 5) can be
used to estimate the time taken for an initially horizontal OWC to reach the steady state tilt predicted
by Hubbert (1953).

NUMERICAL SIMULATION METHODOLOGY


Numerical simulation, using the commercial reservoir simulator Eclipse 100, was used to investigate
Muggeridge and Mahmode

319

Figure 2. The two-dimensional simulation model used to investigate the impact of various parameters on the time for a tilted contact
(oil-water contact [OWC]) to be established and the factors affecting that tilt itself. (A) The model with an initially horizontal contact
(hydrostatic equilibrium). (B) The model after being subjected to a potentiometric gradient (hydrodynamic flow in the aquifer). See Table 1
for abbreviations.

the factors affecting tilted contacts and the time


scales for the steady tilt to be established. A simple model representing a two-dimensional (2-D)
vertical cross section through a reservoir was built

and gridded using a Cartesian blockcentered grid


(Figure 2). The grid dimensions were chosen to
be between 50 20 and 50 100, depending on
the magnitude of the expected tilt of the OWC:

Table 1. Mathematical Symbols, Property Descriptions, and Units Used*


Symbol
L12
P1
P2
P0
V
kaq
k12
H1
H2
L
W
H
ho
f
kh
kv /kh
Co
Cw
Cr
mo
mw
ro
rw
Sor
Swc
n

Property

Value and Units

SI Equivalent

Distance between wells 1 and 2


Pressure measured in well 1 in aquifer
Pressure measured in well 2 in aquifer
Datum pressure
Interstitial velocity of water in aquifer
Effective permeability in aquifer
Effective permeability measured between wells 1 and 2
Depth of oil water contact in well 1
Depth of oil-water contact in well 2
Length of reservoir
Width of reservoir
Thickness of reservoir
Thickness of oil leg
Porosity
Reservoir permeability
Vertical-to-horizontal permeability ratio
Oil compressibility
Water compressibility
Bulk rock compressibility
Oil viscosity
Water viscosity
Oil density
Water density
Irreducible oil saturation
Connate water saturation
Corey exponent

ft
psi
psi
psi
ft day1
md
md
ft
ft
10,000 ft
1 ft
200 ft
20 ft
0.2
100 md
0.1
1 105 psi1
1 106 psi1
2.8 106 psi1
1.0 cp
1.0 cp
39.6 lb ft3
61.2 lb ft3
0.2
0.2
2

m
Pa
Pa
Pa
m s1
m2
m2
m
m
3048 m
0.3048 m
60.96 m
6.096 m

*Values used in the standard model are given where appropriate. All fluid data are at reservoir conditions.

320

Hydrodynamic Aquifer or Reservoir Compartmentalization?

1013 m2
1.45 109 Pa1
1.45 1010 Pa1
4.06 1010 Pa1
103 Pa s
103 Pa s
630 kg m3
980 kg m3

Table 2. Range of Properties Investigated


Property

Lower Range

kh (md)
H (ft)
Co (psi1)
Cw (psi1)
Drw (lb ft3)
Dro (lb ft3)
L (ft)
ho (ft)
F
mw (cp)
mo (cp)

10
200
107
107
10
10
1000
20
0.05
0.5
0.5

Upper Range
1000
500
104
104
40
40
100,000
60
0.60
2.5
2.5

the smaller the tilt, the larger the vertical resolution that was needed to model it accurately.
Initially, we investigated the impact of changing rock and fluid properties in a homogeneous
reservoir on the steady state contact tilt and the
time scales for that tilt to be established and compared results with those predicted by equations 4
and 5 to ensure that the numerical model was properly modeling the very slow flow rates and contact
responses seen in reservoirs with hydrodynamic

aquifers. We then used numerical simulation to investigate the impact of various generic heterogeneities on the development of aquifer tilt over time
and its final geometry. The base case properties are
given in Table 1, and the range of properties investigated is given in Table 2. Table 3 lists the different
heterogeneities investigated and the ranges of their
properties used in simulations.
Each model was initialized with a flat OWC
(Figure 2A). A zero capillary pressure was assumed
so that the OWC was coincident with the free
water level. Water injection and production wells
were used at the opposite end of each of the models
(in the aquifer) to represent the aquifer flux and
thus produce a tilted OWC (Figure 2B). For base
case simulations, the wells were set at a constant
injection and production rate corresponding to an
influx of 3.5 105 ft3 yr1 (106 m3 yr1). This resulted in a pressure drop across the model of 3.75 psi
(25,900 Pa). These flow rates and pressure drop
were chosen to be typical of those encountered in
real reservoirs. Typical potentiometric gradients in
the North Sea range from 0.5 to 9 psi per thousand
feet (11204 kPa km1), (Moss et al., 2003), corresponding to a groundwater interstitial velocity of

Table 3. Description of Heterogeneities and Ranges Investigated


Property
Fault: fault transmissibility
(ft md cp1)
Fault: threshold pressure (psi)

Fault: thickness (ft)

Change in sand quality (md)

Shale with gap position (ft)

Shale with gap width (ft)

Description

Lower Range

The transmissibility of the fault was varied, and the flow rate was
kept constant. The fault had a thickness of 100 ft and a threshold
pressure of zero.
The threshold pressure of the fault was varied, and the flow rate
was kept constant. The fault had a thickness of 100 ft and a
transmissibility of 0.01 (ft md cp1).
The thickness of the fault was varied and the flow rate was kept
constant. The fault transmissibility was set to 0.01 (ft md cp1)
with a threshold pressure of zero.
The permeability of the right side of the reservoir was varied and
the left side was kept constant at 100 md. A constant pressure
drop was maintained across the reservoir.
A zero permeability horizontal shale with a gap of 200 ft was placed
at different heights from the top of the reservoir (the oil-water
contact was at 20 ft). The flow rate was kept constant.
A zero permeability horizontal barrier was placed at 20 ft from the
top of the reservoir (the oil-water contact was at 20 ft), and the
gap width was varied. The flow rate was kept constant.

0.00001

Upper Range
0.01

150

100

0.1

10

25

200

6000

Muggeridge and Mahmode

321

0.002 to 0.4 ft yr1 (0.00060.12 m yr1), as calculated by Darcys law. Using equation 4 and a
typical medium gravity oil, these potentiometric
gradients correspond to a 3 to 60 ft OWC tilt per
thousand feet (i.e., 120 m/km). The simulations
were assumed to have reached steady state when
the maximum change in pressure in any grid cell
was less than 0.01 psi (<69 Pa) between time steps.
This criterion was determined based on the minimum detectable pressure change in common pressure measurement tools. The location of the injection and production well completions below the
OWC was chosen such that when the contact
reached its steady state position, no oil was produced. Similarly, the thickness of the oil leg was
chosen so that the contact did not reach the top
of the reservoir (consistent with the derivation of
equation 5 in Gardner et al., 1962). A very high
productivity index was used for both the injection
and production well to minimize the pressure drop
between the well and the aquifer and thus reduce
nonlinear flow effects near the well that would result in a curvature of the contact near the well.
To investigate the range of values for which
Hubberts analysis is valid, the base model was varied
one property at a time while maintaining the constant pressure drop of 3.75 psi (25,900 Pa) across
the model. Initially, the reservoir was modeled as
being homogeneous in which common reservoir
parameters were varied by an order of magnitude.
After that, a series of common idealized heterogeneities were introduced into the base model to
examine their effect on contact tilt. These heterogeneities included vertical baffles to flow (faults) and
horizontal discontinuous barriers to flow (shales
or tar mats), as well as a sudden change in sand
quality from one side of the reservoir to the other
(see Table 3 for more information).

RESULTS
Homogeneous Reservoir with
Good Communication
Here, we examine the impact of changing reservoir rock and fluid properties in a homogeneous
322

Hydrodynamic Aquifer or Reservoir Compartmentalization?

Figure 3. The water saturation distribution predicted by numerical simulation at steady state with a 3.75-psi drop across the
base case reservoir. The overall difference in depth of the contact
(oil-water contact [OWC]) across the model is 25 ft, which agrees
well with the tilt predicted by Hubbert (1953).

reservoir with no baffles or barriers to flow on the


time to establish a tilted contact and the final tilt
of that aquifer. Simulation results were compared
with the predictions of equation 4 for steady state
tilt and equation 5 for the time required to achieve
that steady state. Earlier works have looked at
Hubberts equation and its ability to predict the
final tilt of the OWC by comparing its predictions
with experimental results (Hubbert, 1953; Dennis
et al., 2000), 2-D reservoir simulations (Dennis
et al., 2000), and three-dimensional reservoir simulations (Dennis et al., 2005). The experimental
results of Dennis et al. (2000, 2005) were limited,
however, as only flow rate was varied (inducing
different pressure drops and hence different OWC
tilts), with no consideration of the other parameters
that affect pressure drop. No literature examines
the influence of reservoir properties on the time
scale for a tilted contact to be established. The
closest investigation is that of England et al. (1995),
discussed earlier.
First, we consider our base case model. Using
equation 4 and the base case properties given in
Table 1, we expect there to be a 25-ft (7.6-m) drop
in contact depth from left to right of our model.
This was indeed observed (Figure 3). Using equation 5, we predict a time scale of 17,600 yr for steady
state to be established. This time scale was consistent with that observed in the simulation results.
From Figure 4, we see that the contact had almost
reached its steady state position after 10 k.y., and
little change occurred between this time and 100 k.y.
Figure 5 shows the pressure-versus-depth plots
that would have been observed by RFT or MDT at

Figure 4. Oil-water contact (OWC)


depth across the base case model
reservoir as a function of time, assuming that the aquifer only starts
flowing from left to right at time
t = 0 yr. The stepped appearance of
these graphs is because the contacts were defined to be at the
boundary between the grid blocks
of water saturation 1 and 0. The
contact has almost reached its steady
state position after 10 k.y. This is
consistent with the analytic prediction of time taken to reach steady
state of 17 k.y.

two vertical wells in the model reservoir. Well 1 was


completed 3400 ft (1036 m) from the left edge of
the reservoir, and well 2 was completed 6600 ft
(2012 m) from the left edge of the reservoir. It can
be seen when comparing the data from these two
wells that at intermediate times (20100 yr), a pressure shift is observed in both the oil and water leg
pressures. However, at late times (10 k.y.), when
steady state is almost established, the pressure observed in the oil leg is the same in both wells, but
a shift in pressure occurs in the water leg. This is

consistent with water flowing horizontally in the


aquifer but no flow in the oil leg.
We then varied various rock and fluid properties in our homogeneous base case model. Figure 6
compares the simulated steady state OWC tilt with
that predicted by Hubbert (1953). We see that a
good agreement overall occurred between the predictions of Hubbert (1953) and those obtained by
simulation. The two major discrepancies occurred
when a significant viscosity contrast occurred between the oil and the water (because the Hubbert

Figure 5. Pressure-versus-depth
plots observed at different times
after the aquifer started flowing at
two vertical wells, one 3400 ft
(1036 m) from the left side of the
model (well 1) and the other 6600 ft
(2011 m) from the left side of the
model (well 2). At intermediate
times (1001000 yr), the pressureversus-depth plots of the two
wells show a pressure difference in
both the oil and water legs. After
10 k.y., when steady state is almost
established, the pressures in the
oil legs are the same, but a significant pressure difference occurs
in the water leg. OWC = oil-water
contact.

Muggeridge and Mahmode

323

Figure 6. Comparison between the oil-water contact (OWC) tilt predicted by the steady state analysis of Hubbert (1953) and the OWC
tilt observed in numerical simulations of the base case reservoir model (Table 1) when different rock and fluid properties are varied as in
Table 2. Overall agreement is very good except that the Hubbert formula does not take viscosity contrasts between oil and water into
account and that the reservoir simulation results are influenced by boundary effects for short (1000-ft [304 m]) reservoirs.

analysis assumes equal viscosities) and when the


reservoir model was very short (1000 ft [305 m]).
This latter discrepancy was caused by the fact that
the steady state profile was strongly influenced
by the inlet and outlet boundary conditions used in
the simulation. The simulated pressure field near
the inlet and outlet had a radial character caused
by the well models used in the simulator.
Figure 7 compares the time scales to reach the
steady state inclination of the OWC predicted by
equation 5 (England et al., 1995) with those obtained from simulation. Overall, a reasonable agreement occurred between the analytic method and
the simulation. The main differences were that
(1) the time scales predicted by the England et al.
(1995) solution did not depend on the oil leg thickness, whereas the simulation predicted a small decrease in time to reach equilibrium as the oil leg
thickness increases and (2) the dependency of time
324

Hydrodynamic Aquifer or Reservoir Compartmentalization?

scale on viscosity. The simulation predicted a higher


dependency on oil viscosity and a lower dependence
on water viscosity than equation 5. However, these
are minor effects. Figure 8 compares the ranges of
time scales to reach steady state based on the ranges
for reservoir size and rock and fluid properties given
in Table 2. Permeability and system length have a
major influence on the time taken to reach steady
state. Using the base case dimensions, a 20-md (9.9
1015 m2) reservoir may take 90 k.y. to reach steady
state, whereas a 1000-md (990 1015 m2) reservoir will reach steady state in less than 10 k.y. Similarly, using the base case permeability (100 md, 99
1015 m2), we see that a large reservoir 80,000 ft
(24,000 m) long may take more than 1 m.y. to reach
steady state. Water and oil compressibility have a
negligible effect on time scales.
From these results, we concluded that the simulation model of a dynamic aquifer produced steady

Figure 7. Comparison of time scales for tilted oil-water contact to reach equilibrium as predicted by numerical simulation and England
et al. (1995), equation 5 as a function of (A) porosity, (B) permeability, (C) reservoir length, (D) thickness of oil leg, (E) water viscosity, (F) oil
viscosity, (G) oil-water density difference (Dr), and (H) total thickness of oil + aquifer. Note that in H, the oil leg was always 10% of the total
thickness. In general, reasonable agreement occurred between the simulation and the analytic expression, although (1) the time scales
predicted by equation 5 do not depend on the thickness of the oil leg and the simulation time scales do not depend on system thickness and
(2) equation 5 does not predict as much dependency on oil viscosity.
Muggeridge and Mahmode

325

Figure 8. Summary of
the impact of reservoir size
and rock and fluid properties on time taken for oilwater contact to reach its
steady state tilt using the
ranges of properties given
in Table 2. Reservoir length
and permeability have
the major impact on time
scales. It would appear
that a typical time scale to
reach steady state is between 10 and 100 k.y.

state and transient behavior that were consistent


with Hubberts calculation of the steady state tilt and
the analytic solution originally derived by Gardner
et al. (1962). We, therefore, felt that it was reasonable to use this simulation model to investigate the
impact of heterogeneities on both the steady state
distribution of the OWC and the time scales to reach
that steady state.
Heterogeneous and Compartmentalized
Reservoirs
Reservoirs can be compartmentalized by vertical
barriers to flow such as faults and horizontal barriers to flow such as shales or tar mats (which often
form close to the OWC). Sometimes, these features can be inferred from seismic data or from logs
but, in many cases, the only evidence for their existence is from RFT or MDT data. Typically, engineers assume that a flow barrier is present when a
horizontal pressure gradient exists in the oil leg,
possibly combined with a change in OWC between
two wells. As we have seen in the previous section,
this may also be a signature of an aquifer whose
flux has changed and thus the oil and water are
still moving toward their steady state orientation,
although the time scales for this are relatively short
compared with geologic time, unless the reservoir
326

Hydrodynamic Aquifer or Reservoir Compartmentalization?

is very large. Conversely, these features may act only


as baffles on geologic time scales, in which case,
there may be little or no evidence of their existence
in pressure data, although they may act as barriers
to flow on reservoir production time scales because
of their low permeability.
In this section, we focus on the impact that
some common idealized barriers and baffles may
have on OWC tilt and on observed synthetic RFT
or MDT data. In addition, we examine how a
change in reservoir quality in the direction of flow
may affect pressure-versus-depth data and thus the
inferred tilt of the OWC. Table 3 gives a brief description of the features investigated, the range of
properties varied, and the boundary conditions used
in the simulations.
Faults are one of the most significant influences on reservoir communication (Yielding et al.,
1999) and can seriously affect petroleum recovery
(Corrigan, 1993). The resolution of data on faults
varies greatly from low-resolution seismic data to
high-resolution core samples taken from the fault
itself. Seismic data can provide information about
the extent of faults and their locations, whereas core
samples can provide information on the percentage
of shale within the samples and the threshold pressure of the fault; however, in many cases, the existence of a fault and its impact on flow can only be

Figure 9. Effect of a clean fault (<15% shale) with a transmissibility of 0.00001 ft md cp1 on oil-water contact (OWC) tilt at steady state.
(A) The OWC tilt across the length of a reservoir. This is a large change in contact depth across the fault compared with the gentle tilt seen
within the reservoir compartments. (B) The pressure as a function of depth in the four wells shown in A. All four wells see the same
pressure with depth in the oil leg, confirming that communication exists across the fault, but the wells on the left see a much higher water
pressure than those on the right as well as seeing a higher contact depth. There is little to distinguish between the pressure versus depth
seen in wells 1 and 2 and similarly between wells 3 and 4.

inferred from well tests, production data, or changes


in OWC depth across a reservoir.
Pressure and total flow are influenced by the
transmissibility of the fault (Manzocchi et al., 1999),
but the flow of oil may be affected differently from
the water flow if a change in wettability occurs between the reservoir rock to fault rock. Typical reservoir rocks exhibit a mixed wettability, whereas
the rocks within the fault zone may well be waterwet, particularly if there has been smearing of clay
(Fisher and Knipe, 1998; Fisher et al., 2001). If
this is the case, then the nonwetting phase (commonly oil) will be trapped until the capillary pressure (threshold pressure) is exceeded. Great uncertainty exists in all fault properties (Lia et al.,
1997; Chambon et al., 2006), and hence, an orderof-magnitude study was performed to investigate
the effect of these properties on the final tilt of
an OWC resulting from aquifer flux through such
a fault and the time scales to reach this steady
state. The fault permeabilities investigated ranged
from than 0.1 to less than 0.0001 md (90.009
1017 m2), whereas the threshold pressure ranged
between 1 and 150 psi (69001 106 Pa). Fault
thickness ranged between 1 and 100 ft (0.330 m).

Figure 9 shows the OWC seen at steady state


for the minimum fault transmissibility examined
(1 105 m2 ft md cp1, 3 1018 m4 s2 kg1), together with the pressures versus depth that would
be seen in four wells, two being completed in different places in each compartment on either side
of the fault. The tilt of the contact was very small
within each reservoir compartment but, as expected,
a huge change in contact depth occurred across
the fault. A very small horizontal pressure gradient
was present across the compartment, as shown by
the small difference between the pressure-versusdepth data seen in wells in the same compartment
because of the small change in contact depth. Comparing the data seen on either side of the fault, it
can be seen that oil pressures fall on the same line
for all wells, but water pressures were much higher
to the left of the fault (upstream direction for aquifer
flux) than to the right. Examination of these data
would suggest that communication exists across
the fault; however, as the time taken for this steady
state to be established was 14 m.y., this fault would
act as a barrier to flow on reservoir time scales. In
this case, it would be possible to infer the existence
of the fault if RFT or MDT data from all four wells
Muggeridge and Mahmode

327

Figure 10. Effect of a clay-rich fault (40%) with a transmissibility of 0.01 ft md cp1 and a threshold pressure of 10 psi (69 kPa) on oilwater contact (OWC) tilt at steady state. The effect of the clay is to allow water to flow through the fault, but oil cannot flow. (A) A gentle
tilt of the OWC exists across the length of a reservoir that does not appear to be influenced significantly by the fault. (B) The pressure as a
function of depth in the four wells shown in A. Although the depth of OWC appears to exhibit a constant tilt across the reservoir, a
different pressure is present in the oil leg with depth seen in wells on either side of the fault, corresponding to the fact that oil cannot flow
across the fault.

shown in the figure existed; however, this would be


more difficult even if data from one well on either
side of the fault existed. Calculations of the large
aquifer interstitial velocity needed to sustain the
observed oil-water tilt based on the observed reservoir sand permeability (1.8 ft yr1, 0.54 m yr1)
would probably warn the engineer or geoscientist
that a significant barrier to flow potentially existed.
Figure 10 shows the results obtained when a
dynamic aquifer flowed through a clay-rich fault.
The fault had a relatively high transmissibility of
0.01 ft md cp1 (3 1015 m4 s2 kg1) but, because of
the clay, had a threshold pressure of 10 psi (69 kPa)
for oil to flow through the fault. A gentle constant
tilt of the OWC occurred across the field in this
case because of the high fault transmissibility to
water flow. In contrast, examination of the pressureversus-depth plots for four wells completed at different locations across the field shows that a large
drop in pressure occurred in the oil leg between
wells on either side of the fault. This was a result of
there being no oil flow through the fault so pressures in the oil could not equalize even when the
contact had reached its steady state tilt. The time
taken to reach steady state in this case was 40 k.y.
There would be communication between compart328

Hydrodynamic Aquifer or Reservoir Compartmentalization?

ments during reservoir production if this resulted


in a differential pressure drop between compartments greater than 10 psi (69 kPa), although the
pressure-versus-depth plots would suggest that
no communication existed.
We also examined the influence of horizontal
baffles and barriers on contact tilt over time. These
could be tar mats or shales. Tar mats form as the
result of biodegradation and are typically found at
the contact between oil and water. They are encountered in many Middle Eastern oil fields and
can be described as sharply limited reservoir levels
enriched with heavy compounds ranging from a
few inches to hundreds of feet (centimeters to tens
of meters) thick (Carpentier et al., 1998) commonly found at the OWC. Their presence in any
reservoir can severely impede the communication
between the oil leg and aquifer (Al-Kaabi et al.,
1988) because of their extremely high viscosity at
reservoir temperatures (Wadman et al., 1979). Unlike tar mats, shales can be found at any depth within
a reservoir because they are formed during the deposition of the reservoir sediments; however, they
are similar in the sense that they result in a thin
barrier to vertical flow. In this work, we examined
the impact of a thin horizontal barrier to flow on

contact tilt as a function of the depth of the barrier in


the reservoir and the size of a gap within that barrier.
Figure 11 shows the steady state location of
the OWC for different depths of the shale or tar
mat and the associated pressure-versus-depth data
that would be observed in two wells 3000 ft (910 m)
apart. The effect of the barrier was to prevent the
upward motion of the aquifer. When the barrier was
within the aquifer before it started to flow, then no
tilt was seen in the contact, although a pressure drop
resulting from the flow of water in the aquifer was
observed in the pressure-versus-depth plots. When
the barrier was in the oil leg, then a tilted contact
was established, unless the water rose up to the
barrier, and again a pressure drop would be observed in the water below the shale resulting from
the flow of the water. In all cases, the gap in the
barrier had little effect; this was because it had
been placed in the center of the model where
little change in pressure occurred when the aquifer
started flowing. The pressures in the oil leg above
the barrier were not influenced by the changes in
pressures in the aquifer during the transient period.
As a result, steady state was established more quickly
the closer the barrier was to the OWC. Steady state
was established in approximately 16 k.y. when the
barrier was near the top of the reservoir but in less
than 1 k.y. when it was lower down near the aquifer.
This model barrier in the oil leg would form a significant impediment to vertical flow on production
time scales, but it would be difficult to identify this
from the pressure-versus-depth plots. The model
barrier in the aquifer would restrict the ability of the
aquifer to provide pressure support, but in this case,
this could be identified from the pressure-versusdepth plots.
Figure 12 shows how the steady state tilt of
the OWC was altered when a larger gap of 2000 ft
(610 m) existed in the shale or tar mat at the original OWC. In this case, the water from the aquifer
spilt backward over the top of the shale into the
original oil leg and, at the same time, oil was displaced below the horizontal barrier, resulting in an
isolated wedge of oil below the barrier at steady
state. Above the barrier, the OWC was horizontal,
but below the barrier, a tilt corresponding to the
background permeability of the reservoir sand

formed; however, this is not immediately apparent


in the pressure-versus-depth plot for well 2. In this
case, the existence of the oil leg underneath the
shale likely would be missed, and it would probably
be inferred that the shale or tar mat was continuous. However, the existence of the same pressures
in the oil leg at a given depth combined with a
horizontal pressure gradient in the aquifer would
be a strong indication that the shale or tar mat was
not continuous. Indeed, the time taken to reach
the steady state in this case was 16 k.y., suggesting
that the barrier had very little influence on the vertical communication.
Finally, we examined the impact that a change in
reservoir quality would have on OWC tilt. The base
reservoir permeability of 100 md (99 1015 m2)
was maintained in the left part of the model, and the
permeability in the right part of the model was reduced. Figure 13 shows the steady state location of
the OWC when the right side had a permeability
of 10 md (9.9 1015 m2). The tilt of the OWC
abruptly changes from that predicted by equation
4 for a permeability of 100 md (99 1015 m2) to
that predicted for a permeability of 10 md (9.9
1015 m2) at the boundary between the two types
of sand. The tilt inferred from the pressure-versusdepth plots for the two wells shown would be midway between the two tilts; however, inspection of
logs might lead the engineer to conclude that a
change of sand quality was present between the
two wells. If this was not correctly identified, then
extrapolating the OWC inferred from the pressureversus-depth plots would result in an underestimation of the oil in place, as shown in Figure 13. The
time taken to reach steady state in this case was
80 k.y.midway between that would be inferred
from equation 5 for a 100-md (99 1015 m2)
reservoir and for a 10-md (9 1015 m2) reservoir.
The effect of reducing the permeability of the
right side to 1 md (9.9 1016 m2) or 0.1 md (9.9
1017 m2) was to increase the time taken to reach
steady state to more than 600 k.y. (1 md [9.9
1016 m2]) and 2 m.y. (0.1 md [9.9 1017 m2]),
as well as to increase the tilt of the OWC in the
right side still further.
Figure 14 compares the times taken to reach
steady state for each of the compartmentalized
Muggeridge and Mahmode

329

330

Hydrodynamic Aquifer or Reservoir Compartmentalization?

Figure 11. Continued.

heterogeneous models with the time taken to reach


steady state in the homogeneous base case model.
In general, heterogeneity tended to increase the time
taken for the OWC to reach its steady state shape, in
many cases, by at least an order of magnitude (fault
thickness, threshold pressure of the same order as
seen in the aquifer, large gap seen in horizontal barrier). In some cases, the time to reach steady state
was greater than 1 m.y. (very low fault transmissibility, <0.0001 ft md cp1 [3 1020 m4 s2 kg1]
or very poor quality reservoir on the left side,
<0.1 md [9.9 1017 m2]). In contrast, in some
cases, the time taken to reach steady state was
much quicker than in the homogeneous reservoir
(threshold capillary pressure greater than pressure drop in aquifer, horizontal barrier at original
OWC) because the reservoir was effectively compartmentalized. In the case of a water-wet fault
with a high threshold pressure, the effective length
of the compartment was halved so the time taken
to reach steady state is quartered (see equation 5).
Although most of the heterogeneities caused a
monotonic change in time to reach steady state as
a function of the change in heterogeneity (e.g.,
decreasing fault transmissibility increased time to
reach steady state), the size of the gap in the hori-

zontal barrier had a decidedly nonlinear impact on


time scale. This was because the time scale was
also related to the thickness of the barrier. Water
could only flow up through the gap (or oil downward) if the horizontal pressure drop along the gap
caused by the hydrodynamic aquifer was greater
than the vertical hydrostatic pressure drop. For
example, when the gap in the shale or tar mat was
1000 ft (305 m) long, then a horizontal pressure
drop of 0.375 psi (2.6 kPa, as the total pressure
drop across the 10,000 ft, 3050 m reservoir was
3.75 psi [26 kPa]) occurs in the gap in the direction
of aquifer flow. As the shale in our model was only
1 ft thick, then the vertical hydrostatic pressure drop
over that thickness caused by water was 0.42 psi
(2.9 kPa). No water flowed through the gap into
the aquifer in the simulation. In contrast, when the
gap was increased to 2000 ft (610 m), the horizontal
pressure drop across the gap was then 0.75 psi
(5.2 kPa), and water was seen to flow through the
gap to form an OWC above the shale on the upstream side of the model, and oil moved down
through the gap on the downstream side (Figure 12).
The model took more than 150 k.y. to reach steady
state in this case compared with a time of 60 k.y.
when the gap was 6000 ft (1800 m) long.

Figure 11. The effect of the position of a single horizontal barrier to flow, with a small gap, on contact tilt and observed pressureversus-depth data at steady state when the barrier is located at different depths in the reservoir. The gap allows the pressures in the oil
leg to stabilize so the oil pressure is the same in both wells at a given depth. A horizontal pressure gradient occurs in the water because of
the hydrodynamic aquifer. The effect of the horizontal barrier is to alter the tilt of the oil-water contact (OWC). It prevents or reduces a tilt
forming when the barrier is in the aquifer.
Muggeridge and Mahmode

331

Figure 12. The effect of a horizontal barrier at the initial oil-water contact (OWC) on final contact tilt when it has a 2000-ft gap in the
middle and the pressure versus depth that would be observed at steady state in two wells on either side of the gap. No observable
contact tilt occurs above the barrier on the side upstream of the gap, but a noticeable tilt occurs on the downstream side below the
barrier. The pressure-versus-depth plots show a significant pressure gradient in the water as a result of the hydrodynamic flux.

DISCUSSION
These results suggest that great care has to be taken
when interpreting pressure-versus-depth data to
identify the cause of different OWCs between
wells. In a homogeneous reservoir in which it is
likely that the aquifer has been flowing for more

than 1 million yr, then it should be possible to


reconcile the pressure drop observed between wells
in the aquifer with the reservoir properties and
dimensions and the observed tilt in the OWC using
Hubberts equation (equation 4), provided that
the wells have the same pressure at a given depth
in the oil leg (Figure 6). If, however, the wells have

Figure 13. The effect of a reduction in reservoir quality on the oil-water contact (OWC) tilt. The aquifer is flowing from left to right, and
the reservoir has 100 md permeability on the left, changing abruptly to 10 md permeability on the right. The tilt of the contact is much
less on the left than on the right, as would be expected from Hubberts (1953) expression (equation 4); however, this change in tilt would
not be seen in pressure-versus-depth plots.
332

Hydrodynamic Aquifer or Reservoir Compartmentalization?

Figure 14. Time taken to reach steady state for the ranges of
properties examined in each of the heterogeneous models. The
time taken to reach steady state in the homogeneous base case
model is shown for comparison. Heterogeneity tends to increase
the time taken, except in the cases of (1) a water-wet fault with a
threshold capillary pressure (Pc) greater than the pressure drop
seen in the aquifer and (2) when a horizontal barrier (shale or
tar mat) is located at the oil-water contact.

a different pressure in the oil at a given depth, then


either a barrier to oil flow is present (e.g., a fault
that acts as a barrier to oil flow, Figure 10) or the
aquifer has only started flowing relatively recently
in the time since the reservoir filled. In this case,
equation 5 could be used to estimate the time
needed to reach steady state, assuming that the
reservoir was homogeneous. If this time is relatively short (perhaps <100 k.y.) compared with the
time since the reservoir filled or the aquifer started
flowing, then it is likely that a barrier to flow is
present. If the time is long (perhaps the reservoir is
large, for example; see Figure 8), then it is entirely
possible that good communication within the reservoir exists and it has not yet reached steady state.
Further studies should be undertaken to confirm this.
A more difficult situation arises when a horizontal pressure gradient is seen between wells in
the aquifer but no such gradient is seen in the oil.
In the case of a very low transmissibility fault (such
as in Figure 9), then application of Darcy's law

(equation 1) using the permeability inferred from


well logs or cores will predict an unrealistically high
interstitial velocity in the aquifer. Of course, the
high pressure gradient may be caused by degradation of reservoir quality in the aquifer, but it may
also be an indication of a fault between the wells.
Comparison of data between additional wells (e.g.,
adding wells 1 and 4 to the data from wells 2 and 3
in Figure 9) may provide further evidence of a fault.
In Figure 10, there would have been an equally large
pressure gradient in the water between wells 1 and
2 as between wells 2 and 3, if the aquifer permeability was much less than that in the reservoir. In
the case of a horizontal permeability barrier (as
shown in Figure 11), it may also be difficult to reconcile the pressure-versus-depth data showing good
communication between wells with logs showing
the existence of a barrier to vertical flow that can
be correlated between wells. Even a small gap in the
shale or tar mat may allow pressure to be equalized
vertically but the shale or tar mat may still significantly impede vertical flow on reservoir time scales.
Changes in reservoir quality in association with
a tilted OWC are perhaps the most difficult to
interpret. As mentioned previously, even a change
of a factor of 10 in reservoir permeability may significantly alter the tilt of the OWC in that region.
Extrapolating the contact seen in Figure 13 results
in an underestimation of oil in place; however, if
the lower permeability region was upstream of the
higher permeability region, then oil initially in place
would be overestimated. Adding further data from
other wells may improve understanding; alternatively, Monte Carlo simulation using different realizations of reservoir permeability in association
with aquifer flux would enable better management
of reservoir uncertainty.

CONCLUSIONS
This study has investigated the interaction between
a hydrodynamic aquifer and the observed OWC
for homogeneous, heterogeneous, and compartmentalized model reservoirs using analytic methods
and numerical simulation. An analytic expression,
originally derived by Gardner et al. (1962) and
Muggeridge and Mahmode

333

used by England et al. (1995) to predict gravitational overturning, can also be used to estimate the
time taken for a steady state OWC to be established once the aquifer starts flowing. Comparison
with results from simulations have shown that this
can be used in combination with the expression
derived by Hubbert (1953) for the final steady state
inclination of the OWC to clarify whether a reservoir contains barriers or baffles to flow or may
simply have not yet reached equilibrium.
Simulation studies using simple models of typical baffles and barriers to flow have demonstrated
the complexity of the interactions between a hydrodynamic aquifer and those features. In particular, analysis of pressure-versus-depth plots should
be carefully integrated with data from logs and other
sources to reduce uncertainty in the reservoir description. Use of pressure-versus-depth plots in
isolation may suggest that the reservoir is compartmentalized when in fact it is still relaxing back
to its steady state or alternatively that good communication exists between wells when this is only
the case on geologic time scales. On reservoir production time scales, communication may be severely
impaired. Combining this analysis with Hubberts
equation for the steady state tilt and equation 5 for
estimating the time scale for steady state to be established may help identify when further data acquisition or more detailed modeling studies are required. As a rule of thumb, it would appear that
most good-quality reservoirs with hydrodynamic
aquifers will reach steady state within approximately
100 k.y. This article has only considered oil-water
reservoirs, but many of the arguments and analyses
can also be used in gas-water reservoirs with the
proviso that gas is considerably more compressible
than oil and water. As a result, the analysis of time
scales to reach equilibrium may not be applicable
when significant pressure drops occur so that compressibility becomes important.

REFERENCES CITED
Al-Kaabi, A. A., H. Menouar, M. A. Al-Marhoun, and H. S.
Al-Hashim, 1988, Bottomwater drive in tar mat reservoirs: Society of Petroleum Engineers Reservoir Engineering, v. 3, no. 2, p. 395400.

334

Hydrodynamic Aquifer or Reservoir Compartmentalization?

Anderson, M. P., 2005, Heat as a ground water tracer: Ground


Water, v. 43, no. 6, p. 951968.
Bachu, S., W. D. Gunter, and E. H. Perkins, 1994, Aquifer disposal of CO2: Hydrodynamic and mineral trapping: Energy Conversion and Management, v. 35, no. 4, p. 269
279, doi:10.1016/0196-8904(94)90060-4.
Badry, R., E. Head, C. Morris, and I. Traboulay, 1993, New
wireline formation tester techniques and applications: Society of Petrophysicists and Well Log Analysts 34th Annual
Logging Symposium, June 1316, Paper 1993-22, 15 p.
Bakker, P., L. Watts, R. Salakhetdinov, Y.-Y. Liew, and B.
Dale, 2009, Appraisal and development of thin oil rims
using smart field approach, an example from Champion
West, Brunei: Society of Petroleum Engineers Asia Pacific
Oil and Gas Conference and Exhibition, Jakarta, Indonesia, August 46, 2009, SPE Paper 122600, 9 p.
Berg, R. R., W. D. DeMis, and A. R. Mitsdarffer, 1994, Hydrodynamic effects on Mission Canyon (Mississippian)
oil accumulations, Billings-Nose area, North Dakota:
AAPG Bulletin, v. 78, no. 4, p. 501508.
Brehm, J. A., 2003, The North Brae and Beinn fields, Block
16/7a, U.K. North Sea, in J. G. Gluyas and H. M. Hichens,
eds., United Kingdom oil and gas fields: Commemorative millenium volume: Geological Society Memoir 20,
p. 467481.
Carpentier, B., A. Y. Huc, F. Marquis, A. E. R. Badr, A. A. Al
Aldarous, and S. Al-Baker, 1998, Distribution and origin
of a tar mat in the S. Field (Abu Dhabi, United Arab
Emirates): Society of Petroleum Engineers Abu Dhabi
International Petroleum Exhibition and Conference,
Abu Dhabi, United Arab Emirates, SPE Paper 49472,
10 p., doi:10.2118/49472-MS.
Chambon, G., J. Schmittbuhl, A. Corfdir, N. Orellana, M.
Diraison, and Y. Graud, 2006, The thickness of faults:
From laboratory experiments to field scale observations:
Tectonophysics, v. 426, no. 12, p. 7794, doi:10.1016
/j.tecto.2006.02.014.
Coats, K. H., J. R. Dempsey, and J. H. Henderson, 1971, The
use of vertical equilibrium in two dimensional simulation of three-dimensional reservoir performance: Society
of Petroleum Engineers Journal, v. 11, no. 2, p. 6371.
Corrigan, A. F., 1993, Estimation of recoverable reserves:
The geologists job, in J. R. Parker, ed., Petroleum geology of northwest Europe: Proceedings of the 4th Conference, Geological Society, London, p. 14731482.
Coutts, S. D., 1999, Aquifer behavior during Brent depressurization and impact on neighboring fields: Society of
Petroleum Engineers Reservoir Engineering and Evaluation, v. 2, no. 1, p. 5361, doi:10.2118/54749-PA.
Dennis, H., J. Baillie, T. Holt, and D. Wessel-Berg, 2000, Hydrodynamic activity and tilted oil-water contacts in the
North Sea, in K. Ofstad, J. E. Kittilsen, and P. AlexanderMarrack, eds., Improving the exploration process by learning from the past: Norsk Petroleumforening Special Publication 9, p. 171185.
Dennis, H., P. Bergmo, and T. Holt, 2005, Tilted oil-water
contacts: Modeling the effects of aquifer heterogeneity,
in A. G. Dore and B. A. Vining, eds., Petroleum Geology Conference Series: London, Geological Society, v. 6,
p. 145158, doi:10.1144/0060145.

Dromgoole, P., and R. Speers, 1997, Geoscore: A method for


quantifying uncertainty in field reserve estimates: Petroleum Geoscience, v. 3, no. 1, p. 112, doi:10.1144
/petgeo.3.1.1.
England, W. A., A. H. Muggeridge, P. J. Clifford, and Z.
Tang, 1995, Modeling geological mixing rates in petroleum reservoirs to detect flow barriers: Theoretical
considerations and case history from the Forties field
(United Kingdom Continental Shelf ), in J. M. Cubitt
and W. A. England, eds., The geochemistry of reservoirs:
Geological Society (London) Special Publications 86,
p. 185201.
Fisher, Q. J., and R. J. Knipe, 1998, Fault sealing processes in
siliciclastic sediments, in G. Jones, Q. J. Fisher, and R. J.
Knipe, eds., Faulting, fault sealing and fluid flow in hydrocarbon reservoirs: Geological Society (London) Special Publication 147, p. 117134.
Fisher, Q. J., S. D. Harris, E. McAllister, R. J. Knipe, and A. J.
Bolton, 2001, Hydrocarbon flow across faults by capillary leakage revisited: Marine and Petroleum Geology,
v. 18, no. 2, p. 251257, doi:10.1016/S0264-8172(00)
00064-7.
Gardner, G. H. F., J. Downie, and H. A. Kendall, 1962, Gravity segregation of miscible fluids in linear models: Society of Petroleum Engineers Journal, v. 2, no. 2, p. 95
104, doi:10.2118/185-PA.
Goetz, J. F., W. J. Prinz, and J. F. Logar, 1977, Reservoir delineation by wireline techniques: The Log Analyst, v. 18,
no. 5, 43 p.
Grosjean, Y., P. Zaugg, and J.-M. Gaulier, 2009, Burial hydrodynamics and subtle hydrocarbon trap evaluation:
From the Mahakam Delta to the South Caspian seal:
International Petroleum Technology Conference, Doha,
Qatar, December 79, 2009, IPTC Paper 13962, 12 p.
Guscott, S., K. Russell, A. Thickpenny, and R. Poddubick,
2003, The Scott field, blocks 15/21a, 15/22, United
Kingdom North Sea, in J. G. Gluyas and H. M. Hichens,
eds., United Kingdom oil and gas fields: Commemorative millenium volume: Geological Society Memoir 20,
p. 467481.
Hortle, A., C. Trefry, K. Michael, and J. Underschultz, 2010,
Hydrodynamic considerations for carbon storage design
in actively producing petroleum provinces: An example
from the Gippsland Basin: Journal of Geochemical Exploration, v. 106, no. 13, p. 121132, doi:10.1016/j
.gexplo.2010.01.011.
Hubbert, M. K., 1953, Entrapment of petroleum under hydrodynamic conditions: AAPG Bulletin, v. 37, no. 8, p. 1954
2026.
Hubbert, M. K., 1967, Application of hydrodynamics to oil
exploration: 7th World Petroleum Congress Proceedings, Mexico City, v. 1B, p. 5975.
Larkin, R. G., 2010, Hydrodynamic trapping of CO2 geosequestered in saline aquifers: Society of Petroleum Engineers Improved Oil Recovery Symposium, Tulsa, Oklahoma, SPE Paper 128205, 11 p., doi:10.2118/128205-MS.
Lia, O., H. Omre, H. Tjelmel, L. Holden, and T. Egel, 1997,
Uncertainties in reservoir production forecasts: AAPG
Bulletin, v. 81, no. 5, p. 775801.
Manzocchi, T., J. J. Walsh, P. Nell, and G. Yielding, 1999,

Fault transmissibility multipliers for flow simulation


models: Petroleum Geoscience, v. 5, no. 1, p. 5363,
doi:10.1144/petgeo.5.1.53.
Moss, B., D. Barson, K. Rakhit, H. Dennis, and R. Swarbrick,
2003, Formation pore pressures and formation waters, in
D. Evans, C. Graham, A. Armour, and P. Barhurst, eds.,
The millenium atlas: Petroleum geology of the central
and northern North Sea: London, The Geological Society, p. 317329.
Muggeridge, A. H., Y. Abacioglu, W. A. England, and P. C.
Smalley, 2005, The rate of pressure dissipation from abnormally pressured compartments: AAPG Bulletin, v. 89,
no. 1, p. 6180, doi:10.1306/07300403002.
Pelissier, J., A. A. Hedayati, E. Abgrall, and J. Plique, 1980,
Study of hydrodynamic activity in the Mishrif fields offshore Iran: Journal of Petroleum Technology, v. 32, no. 6,
p. 10431052.
Smalley, P. C., and A. H. Muggeridge, 2010, Reservoir compartmentalization: Get it before it gets you, in S. J. Jolley,
Q. J. Fisher, R. B. Ainsworth, P. J. Vrolijk, and S. D.
Delisle, eds., Reservoir compartmentalization: London,
Geological Society (London) Special Publication 347,
p. 2542.
Stenger, B. A., 1999, Regional temperature gradient: A key
to tilted OOWC: Society of Petroleum Engineers Middle East Oil Show, Bahrain, SPE Paper 53197, 15 p.,
doi:10.2118/53197-MS.
Stenger, B. A., T. R. Pham, A. A. Al-Sahhaf, and A. S. AlMuhaish, 2001, Assessing the oil-water contact in Haradh
Arab-D: Society of Petroleum Engineers Annual Technical Conference and Exhibition, New Orleans, Louisiana, September 30October 3, SPE Paper 71339, 16 p.,
doi:10.2118/71339-MS.
Sweet, M. L., and L. T. Sumpter, 2007, Genesis field, Gulf of
Mexico: Recognizing reservoir compartments on geologic and production time scales in deep-water reservoirs: AAPG Bulletin, v. 91, no. 12, p. 17011729, doi:10
.1306/07190707011.
Talukdar, S., and L. H. Brusdal, 2005, Why so many different
types of well on Njord? Society of Petroleum Engineers/
International Association of Drilling Contractors Drilling Conference, Amsterdam, The Netherlands, February 2325, 2005, SPE Paper 92085, 13 p.
Thomasen, J. B., and N. L. Jacobsen, 1994, Dipping fluid
contacts in the Kraka field, Danish North Sea: Society
of Petroleum Engineers Annual Technical Conference
and Exhibition, New Orleans, Louisiana, September 25
28, SPE Paper 28435, 10 p., doi:10.2118/28435-MS.
Tozer, R. S. J., and A. M. Borthwick, 2010, Variation in fluid
contacts in the Azeri field, Azerbaijan: Sealing fault or
hydrodynamic aquifer? in S. J. Jolley, Q. J. Fisher, R. B.
Ainsworth, P. J. Vrolijk, and S. D. Delisle, eds., Reservoir
compartmentalization: Geological Society (London) Special Publication 347, p. 103112.
Underschultz, J., 2005, Pressure distribution in a reservoir
affected by capillarity and hydrodynamic drive: Griffin
field, North West Shelf, Australia: Geofluids, v. 5, no. 3,
p. 221235, doi:10.1111/j.1468-8123.2005.00112.x.
Underschultz, J., R. A. Hill, and S. Easton, 2008, The hydrodynamics of fields in the Macedon, Pyrenees and

Muggeridge and Mahmode

335

Barrow sands, Exmouth subbasin, Northwest Shelf Australia: Identifying seals and compartments: Exploration Geophysics, v. 39, no. 2, p. 8593, doi:10.1071
/EG08010.
Van Kirk, C. W., 1976, Jormar-reservoir performance and
numerical simulation of a classic Denver-Julesberg Basin
oil field: Rocky Mountain Regional Meeting of the Society of Petroleum Engineers, Casper, Wyoming, May
1112, 1976, SPE Paper 5890, 14 p.
Wadman, D. H., D. E. Lamprecht, and I. Mrosovsky, 1979,
Joint geological/engineering analysis of the Sadlerochit
Reservoir, Prudhoe Bay field: Journal of Petroleum Technology, v. 31, no. 7, p. 933940, doi:10.2118/7531-PA.
Weber, K. J., 1987, Hydrocarbon distribution patterns in Nigerian growth fault structures controlled style and stratigraphy: Journal of Petroleum Science and Engineering, v. 1,
no. 2, p. 91104, doi:10.1016/0920-4105(87)90001-5.

336

Hydrodynamic Aquifer or Reservoir Compartmentalization?

Yielding, G., J. A. Overland, and G. Byberg, 1999, Characterisation of fault zones in the Gullfaks field for reservoir
modeling, in A. J. Fleet and S. A. R. Boldy, eds., Petroleum geology of northwest Europe: Proceedings of the
5th Conference, London, Geological Society, p. 1177
1185.
Zawisza, L., 1986, Hydrodynamic condition of hydrocarbon
accumulation exemplified by the Carboniferous formation in the Lublin synclinorium, Poland: Society of
Petroleum Engineers Formation Evaluation, v. 1, no. 3,
p. 286294, doi:10.2118/13345-PA.
Zawisza, L., 2004, Hydrodynamic condition of hydrocarbon
accumulation exemplified by the Pomorsko and Czerwiensk oil fields in the Polish lowlands: Society of Petroleum Engineers Annual Technical Conference and Exhibition, Houston, Texas, SPE Paper 90586, 10 p., doi:10
.2118/90586-MS.

Vous aimerez peut-être aussi