Relative-Permeability Hysteresis: Trapping Models and Application To Geological CO2 Sequestration

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/266671387
Relative-Permeability Hysteresis: Trapping Models and Application to

Geological CO2 Sequestration
Article · October 2005

DOI: 10.2118/96448-MS
CITATIONS READS
58 520
4 authors, including:
Elizabeth Spiteri
Oil Search Alaska LLC
4 PUBLICATIONS 838 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Multicomponent, multiphase flow in porous media with temperature variation View project
All content following this page was uploaded by Elizabeth Spiteri on 12 June 2015.
The user has requested enhancement of the downloaded file.

SPE 96448
Relative Permeability Hysteresis: Trapping Models and Application to
Geological CO2 Sequestration
Elizabeth J. Spiteri, SPE, Stanford U.; Ruben Juanes, SPE, Stanford U.; Martin J. Blunt, SPE, Imperial College;
Franklin M. Orr, Jr., SPE, Stanford U.
Copyright 2005, Society of Petroleum Engineers, Inc. Introduction

This paper was prepared for presentation at the 2005 SPE Annual Technical Con- Hysteresis refers to irreversibility or path dependence. In
ference and Exhibition held in Dallas, Texas U.S.A., 9 – 12 October 2005.
multiphase flow, it manifests itself through the dependence
This paper was selected for presentation by an SPE Program Committee follow-
ing review of information contained in an abstract submitted by the authors(s).
of relative permeabilities and capillary pressures on the
Contents of the paper, as presented, have not been reviewed by the Society of saturation path and saturation history. From the point
Petroleum Engineers and are subject to correction by the author(s). The material,
as presented, does not necessarily reflect any position of the Society of Petroleum
of view of pore-scale processes, hysteresis has at least two
Engineers, its officers, or members. Papers presented at SPE meetings are subject sources: (1) contact angle hysteresis; and (2) trapping of
to publication review by Editorial Committees of the Society of Petroleum Engi-
neers. Permission to copy is restricted to an abstract of not more than 300 words.
the nonwetting phase.
Illustrations may not be copied. The abstract should contain conspicuous acknowl- The first step in characterizing relative permeability hys-
edgement of where and by whom the paper was presented. Write Librarian, SPE,
P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435. teresis is the ability to capture the amount of oil that is
trapped during any displacement sequence. Indeed, a trap-
ping model is the crux of any hysteresis model: it deter-
mines the endpoint saturation of the hydrocarbon relative
Abstract permeability curve during waterflooding.
The complex physics of multiphase flow in porous media Extensive experimental and theoretical work has focused
are usually modeled at the field scale using Darcy-type on the mechanisms that control trapping during multi-
formulations. The key descriptors of such models are the phase flow in porous media.1–3 Of particular interest to
relative permeabilities to each of the flowing phases. It is us is the influence of wettability on the residual hydro-
well known that, whenever the fluid saturations undergo carbon saturation. Early experiments in uniformly wet-
a cyclic process, relative permeabilities display hysteresis ted systems suggested that waterflood efficiency decreases
effects. with increasing oil-wet characteristics.4, 5 These experi-
In this paper we investigate hysteresis in the relative per- ments were performed on cores whose wettability was al-
meability of the hydrocarbon phase in a two-phase system. tered artificially, and the results need to be interpreted
We propose a new model of trapping and waterflood rela- carefully for two reasons: (1) reservoirs do not have uni-
tive permeability, which is applicable for the entire range form wettability, and the fraction of oil-wet pores is a func-
of rock wettability conditions. The proposed formulation tion of the topology of the porous medium and initial water
overcomes some of the limitations of existing trapping and saturation,6 and (2) the core-flood experiments were not
relative permeability models. The new model is validated performed for a long enough time, and not enough pore-
by means of pore-network simulation of primary drainage volumes were injected to drain the remaining oil films to
and waterflooding. We study the dependence of trapped achieve ultimate residual oil saturation. Other core-flood
(residual) hydrocarbon saturation and waterflood relative experiments, in which many pore volumes were injected,
permeability on several fluid/rock properties, most notably the observed trapped/residual saturation did not follow
the wettability and the initial water saturation. a monotonic trend as a function of wettability, and was
The relevance of relative permeability hysteresis is then actually lowest for intermediate-wet to oil-wet rocks.7–9
evaluated for modeling geological CO2 sequestration pro- Jadhunandan and Morrow10 performed a comprehensive
cesses. Here we concentrate on CO2 injection in saline experimental study of the effects of wettability on water-
aquifers. In this setting, the CO2 is the nonwetting phase, flood recovery, showing that maximum oil recovery was
and trapping of the CO2 is an essential mechanism after achieved at intermediate-wet conditions.
the injection phase, during the lateral and upward migra- An empirical trapping model typically relates the
tion of the CO2 plume. We demonstrate the importance of trapped (residual) hydrocarbon saturation to the maxi-
accounting for CO2 trapping in the relative permeability mum hydrocarbon saturation, that is, the hydrocarbon
model for predicting the distribution and mobility of CO2 saturation at flow reversal. In the context of waterflooding,
in the formation. We conclude that a proper treatment of a trapping model defines the ultimate residual oil satura-
the nonwetting phase trapping leads to a higher estimate tion as a function of the initial water saturation. The most
of the amount of CO2 that it is safe to inject. widely used trapping model is due to Land.11 It is a single-
2 E. J. SPITERI, R. JUANES, M. J. BLUNT AND F. M. ORR, Jr. SPE 96448
parameter model, and constitutes the basis for a number questration in heterogeneous saline aquifers. In the final
of relative permeability hysteresis models. Other trap- section of this paper we gather the main conclusions and
ping models are those of Jerauld12 and Carlson.13 These anticipate ongoing and future work.
models are suitable for their specific applications but, as
we show in this paper, they have limited applicability to Summary of model equations
intermediate-wet and oil-wet media.
In this section, for clarity, we present the trapping and
Land11 pioneered the definition of a “flowing satura-
hysteresis model. A detailed description of the tools used
tion”, and proposed to estimate the imbibition relative per-
and a physical interpretation of the results is provided in
meability at a given actual saturation as the drainage rel-
the subsequent sections.
ative permeability evaluated at a modeled flowing satura-
tion. Land’s imbibition model11 gives accurate predictions A trapping model attempts to capture the residual
for water-wet media,14 but fails to capture essential trends (trapped) oil saturation Sot after waterflooding, given an
when the porous medium is weakly or strongly wetting to initial oil saturation Soi achieved during primary drainage
oil. The two-phase hysteresis models that are typically of oil into a water-filled medium. Therefore, the trapping
used in reservoir simulators are those by Carlson13 and model may be viewed as a curve on a diagram of initial oil
Killough.15 A three-phase hysteresis model that accounts saturation versus residual oil saturation (initial–residual or
for essential physics during cyclic flooding was proposed IR curve). The development of a new trapping model is
by Larsen and Skauge.16 These models have been eval- motivated by the fact that, in contrast with most com-
uated in terms of their ability to reproduce experimental mon trapping models,11, 12 initial–residual curves obtained
data17, 18 and their impact in reservoir simulation of water- from pore-network simulations are not monotonic when
alternating-gas injection.18, 19 Other models are those by the medium is not strongly water-wet. This non-monotonic
Lenhard and Parker,20 Jerauld12 and Blunt.21 More re- trend has been confirmed experimentally.10
cently, hysteresis models have been proposed specifically The simplest functional form that allows us to reproduce
for porous media of mixed wettability.22–24 a non-monotonic behavior is the quadratic expression:
All of the hysteresis models above require a bounding 2
drainage curve and either: (1) a waterflood curve as input; Sot = αSoi − βSoi . (1)
or (2) a calculated waterflood curve using Land’s model.
The task of experimentally determining the bounding wa- Parameters α and β correspond to the initial slope and the
terflood curves from core samples is arduous, and the de- curvature of this curve, respectively. They must satisfy the
velopment of an empirical model that is applicable to non following requirements:
water-wet media is desirable. In this paper, we introduce
a relative permeability hysteresis model that does not re- 0 ≤ α ≤ 1, β ≥ 0. (2)
quire a bounding waterflood curve, and whose parameters
may be correlated to rock properties such as wettability The model contains two parameters, α and β, which must
and pore structure. be calibrated for each rock–fluid system. Therefore, they
Since it is difficult to probe the full range of relative will depend on the pore-space morphology and the wetta-
permeability hysteresis for different wettabilities experi- bility characteristics of the medium. Typical dependence
mentally, we use a numerical tool –pore-scale modeling– of these two parameters on the wettability of the medium
to predict the trends in residual saturation and relative is discussed and explained later in the paper.
permeability. As we discuss later, pore-scale modeling is Inspired by Land’s model,11 we propose to calculate the
i
currently able to predict recoveries and relative permeabil- waterflood relative permeability kro at the actual oil sat-
d
ities for media of different wettability reliably.25–30 We will uration So as the drainage relative permeability kro at a
use these predictions as a starting point to explore the be- different flowing saturation:
havior beyond the range probed experimentally.
i d
In summary, this paper contains two main results: kro (So ) = kro (Sof ). (3)
1. A new model of trapping and waterflood relative
The flowing saturation is evaluated by means of the fol-
permeability, which is able to capture the behavior
lowing expression:
predicted by pore-network simulations for the entire
range of wettability conditions. ·
1
Sof = (α − 1)
2. A demonstration of the importance of nonwetting- 2β
phase trapping and relative permeability hysteresis for p
¸
2
+ (α − 1) + 4β[So − Sot + γ(So − Sot )(So − Soi )] ,
the assessment of geological CO2 sequestration pro-
cesses.
(4)
In the next section we present a summary of the equa-
tions of the proposed model. We follow with an overview of where Sot is the trapped oil saturation given by Eq. (1),
pore-scale modeling of trapping and relative permeability and γ is a parameter related to the presence of film flow.
hysteresis. We then present a new model of trapping and It is dependent on the wettability characteristics of the
waterflood relative permeability. As an application of the medium, and is typically negative for strongly water-wet
importance of trapping and hysteresis we perform three- media and positive for weakly and strongly oil-wet media.
dimensional simulations of realistic scenarios of CO2 se- This behavior will be discussed in detail later.
SPE 96448 RELATIVE PERMEABILITY HYSTERESIS: TRAPPING MODELS AND CO2 SEQUESTRATION 3
Pore-scale modeling of trapping and hysteresis a cube of 3×3×3 mm3 containing 12,349 pores and 26,146
In pore-network modeling the pore space is described by throats. The absolute permeability of the rock is 2287 mD
a network of pores connected by throats with an idealized and the net porosity is 0.183, where 0.0583 is clay-bound
geometry. A set of physically-based rules describe the con- or micro-porosity.
figuration of the fluids within each pore and throat, as well We investigated sequences of two consecutive displace-
as the mechanisms for the displacement of one fluid by an- ments: primary drainage (oil invasion) and waterflooding.
other. This approach was pioneered by Fatt31 and has re- During primary oil drainage the network, which is initially
ceived increasing attention over the past decade. Blunt32 filled with water, is assumed to be strongly water-wet with
and Blunt et al.33 provide a detailed description of the a receding contact angle θr = 0◦ . As the oil invades the
fundamentals and applications of pore-network modeling, largest pores first in piston-like displacement, the water re-
together with an extensive literature review. One of the cedes and is squeezed to the crevices and pore throats until
successful application areas of pore-network models is the a very high capillary pressure or a target oil saturation is
prediction of multiphase flow properties, such as capillary reached. At this point, the surface of the rock in con-
pressure and relative permeability. This success hinges on tact with oil will undergo wettability alteration, while the
the following: corners and elements that still contain only water remain
strongly water-wet. Wettability alteration is accounted for
1. The ability to reproduce the essential geometric fea- by changing the contact angle. In principle, one could
tures of the pore space of real rocks. A realis- change the advancing contact angle θa and the receding
tic three-dimensional pore-space characterization may contact angle θr independently. In this work, however, we
be obtained in a variety of ways: assembly of two- used a correlation proposed by Morrow43 to link both the
dimensional sections to form a three-dimensional im- advancing and receding contact angles with an intrinsic
age;34 direct X-ray microtomography of the three- contact angle θi . This relationship is shown in Fig. 1.
dimensional pore space;35 stochastic 3D modeling
with statistics inferred from two-dimensional thin sec-
tions;36 process-based reconstruction in which grain
deposition, compaction and cementation are mod-
eled.37, 38
2. The ability to capture wettability effects. Most pore-
network models used today introduce wettability ef-
fects based on the pore-level scenario of wetting pro-
posed by Kovscek et al.6 Their model mimics the
saturation change typical of a hydrocarbon reservoir.
The medium is initially filled with water, and the rock
surfaces are water-wet. During oil migration, the oil
invades the pore space, altering the wettability of the
solid surface in contact with the oil. In this fashion,
the network displays mixed wettability: a fraction of
a pore or throat may be oil-wet, while the corners and
crevices not in contact with the oil remain water-wet.
The combination of realistic pore geometry/topology and Fig. 1— Relationship between receding and advancing contact
correct characterization of displacement and trapping angles on a rough surface, as a function of intrinsic contact
mechanisms has allowed pore-network models to pre- angle measured at rest on a smooth surface.43 [Figure from
dict hysteretic capillary pressure and relative permeability Valvatne and Blunt.28 ]
curves under a wide range of wettability characteristics.
Experimental measurements of hysteretic relative per- During waterflooding there are several physical mech-
meability for mixed-wet and oil-wet media are scarce.10, 39 anisms by which the water can displace the oil in place.2
Given the success of pore-network models to reproduce ex- These mechanisms include piston-type displacement, coop-
perimental data,26–28 in this work we have adopted the use erative pore-body filling and snap-off. The predominance
of pore-network modeling as a way to investigate the full of any given displacement mechanism is strongly depen-
spectrum of wettability conditions. Pore-network simula- dent on the wettability (specified by the advancing contact
tions results are taken as “data” to develop and validate angle). These displacement processes and their implemen-
empirical trapping and hysteresis models. tation are described in detail in the literature.28, 29, 40, 41
After individual displacement events the transport proper-
Description of the pore-network simulations. We ties are calculated. The equations for absolute permeabil-
used the two-phase flow pore-network simulator developed ity, relative permeability and other transport parameters
by Valvatne and Blunt.28 The model has similarities with can also be found in the literature.28, 40, 41, 44 Following
other network models.29, 40, 41 A full description of the this procedure, pore-network simulations have been shown
model is given in the references above and will be omit- to reproduce experimental capillary pressure and relative
ted here. permeability curves both in primary drainage and water-
We used a three-dimensional pore-network of a Berea flooding40 and for a variety of wettability conditions.26–28
sandstone developed by Bakke and Øren.42 The model is We investigated the full range of wetting conditions after
wettability alteration, by choosing average intrinsic con-

0.5
tact angles between 20◦ and 160◦ . Due to pore-scale in- o
θ = 20
homogeneities in the rock minerals and surface roughness, 0.45
θ = 60o
assigning a uniform contact angle throughout is unrealis- 0.4 θ = 90o
tic. Therefore, we assigned contact angles throughout the
Trapped oil saturation (Sot)

o
0.35 θ = 100
network randomly within ±20◦ of the average value. Non−monoticity of
θ = 110o trapping relation
In this work we assume that the contact angle distri- 0.3
θ = 160 o
bution in originally oil-filled pores is independent of Soi . 0.25

In reality, a higher Soi represents a larger initial capillary
0.2
pressure that may cause protective water films to collapse
in oil-filled pores, causing a more significant wettability 0.15
alteration than for low Soi .6 However, this simple char- 0.1
acterization of wettability has been shown to be sufficient
0.05
to predict trends in recovery seen experimentally.28 Also
we consider systems where there is a relatively uniform 0
0 0.2 0.4 0.6 0.8 1
distribution of wettability – we do not consider mixed-wet Initial oil saturation (Soi)
media where initially oil-filled pores may be water-wet or
oil-wet with two distinct distributions of contact angle in Fig. 2— Trapped oil saturation Sot as a function of the initial
the same rock. oil saturation (Soi ) for different intrinsic contact angles (θ).
For each contact angle distribution we performed a series
of displacement pairs (oil invasion and waterflooding), with
a different target oil saturation Soi ranging from a very initial water saturation increases, we have more water fill-
small value (almost no oil invasion) to the maximum value ing from the pores and throats in connected patches from
possible (connate water conditions). Next, we present re- throats that are initially water-filled. When these patches
sults from an extensive set of simulations for different con- join up, there can be trapping, as oil becomes stranded
tact angles and different target oil saturations. between these clusters. Low initial water saturation (i.e.
high initial oil saturation) implies that there are few wa-
Pore-network predictions of trapping. The use of a ter clusters and little chance for bypassing. Trapping in-
pore-network simulator allows us to quantify the effects creases as the number of clusters increases and then de-
of wettability and initial oil saturation on the trapping of clines again as there is less oil to trap in the first place.
oil during waterflooding. The main results of the pore- For intrinsic contact angles above 90◦ (advancing contact
network simulations are compiled in Fig. 2, where we plot angles above 120◦ ) we observe the same behavior, but with
the initial–residual (IR) curves for different intrinsic con- even less trapping. The reason is the presence of oil lay-
tact angles. For a given curve (i.e. for a specific value ers within the network. Water fills the largest pores and
of the intrinsic contact angle after wettability alteration), throats in an invasion percolation-like process. Oil remains
each point denotes the trapped oil saturation Sot that cor- connected in layers sandwiched between water in the cen-
responds to a particular initial oil saturation Soi . The ini- ter of an element and water in the corners. These layers
tial oil saturation is dictated by the point at which primary lead to very little trapping, although the oil relative per-
drainage ceases (and waterflood starts), and the trapped meability is very low. These layers are stable until the two
oil saturation is the value of unrecoverable (or residual) oil water/oil interfaces meet. High initial water saturation
after waterflooding. means that water bulges out in the corners causing these
The most noteworthy characteristic of the IR curves interfaces to meet, trapping more oil than for low initial
shown in Fig. 2 is that they do not display a monotonically water saturation. Again, for sufficiently large Swi , (low
increasing behavior for mixed-wet and oil-wet media. This Soi ), there is less trapping simply because there is less oil
means that, for oil-wet media, higher oil saturation after to trap.
the initial oil invasion may lead to lower residual oil satura- We should also mention that the extremely low trapped
tion after waterflooding. This counterintuitive behavior is oil saturations for very high initial oil saturations are an
analyzed and explained below, in terms of pore-scale fluid artifact of the criterion used for ascertaining the stability of
configurations and displacement mechanisms. oil layers.28 In the future we plan to incorporate a stability
For water-wet media (small contact angles) trapping criterion based on free energy balance that predicts that
during waterflooding is controlled by snap-off. As the ini- oil layers become unstable before the two water/oil inter-
tial oil saturation is increased, oil is pushed into smaller faces touch each other.45 For practical purposes, however,
and smaller pores. During waterflooding, water fills the such high initial oil saturations are never achieved during
smallest pores first and snaps off more and more oil. The migration of oil into realistic reservoirs—the initial water
amount of trapped oil increases monotonically with in- saturation is typically much higher than 5%.
creasing initial oil saturation simply because there is more In conclusion, the trapping mechanisms that we have
oil to trap. As the contact angle increases there is a indicated allow for a physical explanation of the nonmono-
crossover from trapping by snap-off to trapping by bypass- tonic behavior of the initial–residual curves. A complete
ing, as the water tends to advance in a connected front with picture of the trapping relation is given in Fig. 3 as a trap-
piston-like advance on throats and cooperative pore-filling. ping surface, that is, a surface that describes the residual
We begin to see a non-monotonic behavior because, as the oil saturation as a function of the initial oil saturation and
the intrinsic contact angle after wettability alteration.

1
Oil relative permeability (kro(w))

0.8
S
o,max
0.6 0.6
S
oi i
0.4 0.4 k
ro(w)
d
k
ro(w)
0.2
0.2
0 S
ot
S
ot,max
150 0
0 0.2 0.4 0.6 0.8 1
Co 0.8 Oil saturation (S )
nta100 0.6 o
ct S )
an
gle 50 0.2
0.4
r a t ion ( oi
satu
(θ) 0 l oil Fig. 4— Waterflood relative permeability curves that cross
Initia due to the nonmonotonic trapping relationship.
value. The trapping mechanisms at this point should not

be generalized for the entire waterflood process. Because
different competing trapping mechanisms may dominate
0.6 at different capillary pressure levels, this affects the shape

0.5 of the relative permeability curve.
0.4 In water-wet systems we have already mentioned that
trapping is primarily due to snap-off. After a certain
0.3
point during waterflooding, the flowing oil phase becomes
0.2 trapped due to the invading water phase. However, this
0.1 mechanism does not operate right away. At the beginning
of the waterflood process, the nonwetting phase relative
0
permeability is slightly higher than the drainage relative
0 permeability. This is seen in the two-phase experiments
Co 50 0.2
nta )
ct a 100 0.4 n (S oi performed by Oak46 in water-wet Berea sandstone for both
ngl 0.6 t u ratio
e (θ 0.8 s a water/gas and water/oil systems (see Fig. 5).
) 150 nitial oil
I During primary drainage, the oil preferentially fills large
pores, leaving water residing in narrower throats and in the
corners of the pore space. At the end of primary drainage,
Fig. 3— Two views of the trapping surface. Trapped oil sat- many oil-filled pores have only a single connecting throat
uration depending on the initial oil saturation (Soi ) and rock that is also oil-filled. These dead-end pores may contain a
wettability in terms of the intrinsic contact angle (θ). large saturation but do not contribute to the connectivity
of the oil. During waterflooding, pore-filling is favored in
The trapped oil saturation dictates the endpoint of pores that have many surrounding water-filled throats –
the relative permeability waterflood curves. An impor- essentially these dead-end pores.2 Thus initially there is a
tant practical consequence of the nonmonotonic relation of cascade of pore-filling, where the oil saturation decreases
trapped vs. initial oil saturation for mixed- to oil-wet media with little decrease in oil relative permeability. This pro-
is that waterflood scanning curves will cross, as sketched cess competes with snap-off which traps oil and in con-
in Fig. 4. trast leads to a large decrease in relative permeability. For
Both of these features—nonmonotonic trapping relation water-wet media, the former process generally is more sig-
and crossover of waterflood relative permeability curves— nificant at high oil saturation, giving the typical hysteresis
are not present in existing empirical models. This moti- patterns seen in Fig. 5. The pore-network model is able
vates the development of new empirical trapping and hys- to reproduce this behavior with a quantitative agreement
teresis models that reproduce the observed behavior. with experiment,28 see Fig. 6.
In contrast, the trapping mechanisms that control the
Pore-network predictions of waterflood relative shape of the oil waterflood relative permeabilities in oil-wet
permeability. It is important to understand the trapping media are very different from those of a water-wet rock. At
mechanisms that ultimately define the shape of the rela- the beginning of the waterflood, water percolates through
tive permeability curves during waterflooding. The trap- the largest pores, leading to a significant reduction in the
ping model determines the endpoint residual saturations oil relative permeability. The oil, which remains connected
when the capillary pressure is lowered to an extremely low through oil films and layers, drains down to very low sat-
urations but at a low rate due to the small conductance

Gas−Water experiment (200 md) of these layers. The shape of these relative permeability
1 curves, as predicted by pore-network simulations, is shown
Drainage in Fig. 7.
Gas relative permeability (k )
rg
imbibition
0.8
Contact angle (180)
1
0.6 drainage
Oil relative permeability (kro)

imbibition
0.8
0.4
0.6
0.2
0.4
0
0 0.2 0.4 0.6 0.8 1
Gas saturation (S ) 0.2
g
Oil−Water experiment (1000 md)

1 0
0 0.2 0.4 0.6 0.8
Drainage
Oil relative permeability (k )
Oil saturation (S )
ro
Imbibition o
0.8
Fig. 7— Oil relative permeabilities generated from pore-
0.6 network simulations in strongly oil-wet media.
0.4
Development and validation of a new model of
relative permeability hysteresis
0.2
In the previous section we used pore-network modeling to
highlight the following features of wettability effects on the
0 waterflood relative permeability:
0 0.2 0.4 0.6 0.8 1
Oil saturation (S )
o 1. The initial–residual (IR) curves are not monotonic for
media that are not strongly water-wet (Fig. 2).
Fig. 5— Relative permeability curves of the nonwetting phase
for water/gas (top) and water/oil (bottom) systems from
Oak’s46 experimental data in water-wet Berea sandstone. 2. For intermediate-wet and oil-wet media, the scanning
curves of oil relative permeability may cross (Fig. 4).
3. In strongly water-wet media, the trapped oil satura-

Contact angle (0) tion is high but the waterflood relative permeability
0.8 may be higher than the drainage relative permeability
drainage at high oil saturations (Fig. 6).
0.7 imbibition
0.6 4. In contrast, for strongly oil-wet media, the trapped

oil saturation is low but the waterflood relative per-
0.5 meability decreases sharply at high oil-saturations
(Fig. 7).
0.4
0.3 Clearly, this markedly different behavior in water-wet and

oil-wet media needs to be incorporated in the empirical
0.2 model. We start by describing a new trapping submodel
and we follow with the proposed waterflood relative per-
0.1
meability model.
0
0 0.2 0.4 0.6 0.8 The trapping model. We begin by reviewing some of
Oil saturation (S )
o the existing trapping models. These models were originally
designed to account for gas trapping, but for consistency
Fig. 6— Oil relative permeabilities generated from pore- we will treat them for oil trapping. We then formulate a
network simulations in strongly water-wet media. new model and assess its performance for the full spectrum
of wettability conditions.
Land trapping model. The Land model11 is the most

widely used empirical trapping model. Most relative per- 1
Soi
)
meability models that incorporate hysteresis12, 15, 16, 20–22
ro
are based on this trapping model. It was developed to 0.8
Oil relative permeability (k

predict trapped gas saturation as a function of the initial
gas saturation based on published experimental data from
water-wet sandstone cores.47–49 0.6 kdro
In this model, the trapped nonwetting phase saturation
is computed as: kiro
0.4
Soi ∆ So
Sot (Soi ) = , (5) Sof So
1 + CSoi 0.2
where Soi is the initial oil saturation, or the saturation at Sot,max
the flow reversal, and C is the Land trapping coefficient. 0
The Land coefficient is computed from the bounding oil 0 0.2 0.4 0.6 0.8 1
invasion and waterflood curves as follows: Oil saturation (S o)
1 1
C= − , (6) Fig. 9— Geometric extrapolation of the oil relative permeabil-
Sot,max So,max
ity and trapped saturation during waterflooding, as proposed
where So,max is the maximum oil saturation, and Sot,max is by Carlson.13
the maximum trapped oil saturation, associated with the
bounding waterflood curve. All these quantities are illus-
trated in Fig. 8. The value of the Land trapping parame- This model is adequate if the intermediate scanning
ter is dependent on the type of rock and fluids.18 curves are almost parallel and there is little curvature in
the waterflood curve. The model is problematic when the
system is oil wet. The large curvature of the bounding wa-
1 terflood relative permeability curve at low saturations does
not allow prediction of intermediate relative permeability
)
0.8
curves since any shifting will make the endpoint trapped
ro
gas saturation negative, a nonphysical value.

Oil relative permeability (k
S
o,max
0.6 d
kro Jerauld trapping model. Jerauld’s trapping model12
is an extension of the Land trapping model that accounts
ki
for the “plateau” observed in the initial–residual (IR)
0.4 ro
curves for mixed-wet rocks.50 The trapped nonwetting-
S phase saturation is given by:
oi
0.2
Soi
Sot = 1+b/C
. (8)
S (S )
ot oi
Sot,max 1 + C Soi
0
0 0.2 0.4 0.6 0.8 1
Oil saturation (S ) In the original publication,12 the expression of the trapped
o
saturation Sot was given in terms of the residual satura-
tion achieved when Soi = 1, Sor = 1/(C + 1). Jerauld
Fig. 8— Parameters required in the evaluation and application introduced a second tuning parameter b in addition to the
of the Land trapping model. Land coefficient. If this parameter is set to zero, Jerauld’s
model reduces to the Land trapping model. When this pa-
rameter is equal to one, the trapping curve has a zero slope
Carlson trapping model. A simplified hysteresis at Soi = 1. Although Jerauld argued that the IR curves
model proposed by Carlson13 implicitly defines a trapping should not have a negative slope, his model allows for such
model. The Carlson model requires the bounding drainage behavior if b ≥ 1.
and waterflood curves. The trapped oil saturation is de- Although the fit of Jerauld’s model to the pore-network
termined by shifting the bounding waterflood curve to in- data was good for water-wet and intermediate-wet condi-
tersect the intermediate initial oil saturation at the flow tions (for which the model was designed), it was not as
reversal. The idea behind Carlson’s interpretation is to satisfactory for strongly oil-wet media. One of the reasons
use the model of the waterflood relative permeability scan- is that Jerauld’s model assumes that the IR curve has a
ning curves as being parallel to each other. This geometric unit slope near the origin. This behavior does not conform
extrapolation procedure is illustrated in Fig. 9. to pore-network predictions (see the curves corresponding
The trapped nonwetting-phase saturation is computed to θ = 110◦ and 160◦ in Fig. 2).
as
Sot = Sot,max − ∆So , (7) A new trapping model. We notice that the shapes
where ∆So is the shift in the waterflood scanning curve of the trapping curves (Fig. 2) may be fit to a parabola.
with respect to the imbibition bounding curve (see Fig. 9). We establish the following simple quadratic relationship
Contact angle 20 Contact angle 40 Contact angle 60 Contact angle 80
0.5 0.5 0.5 0.5

Trapped oil saturation (Sor)

0.4 0.4 0.4 0.4
0.3 0.3 0.3 0.3
0.2 0.2 0.2 0.2
0.1 0.1 0.1 0.1
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S )
oi oi oi oi
Contact angle 100 Contact angle 120 Contact angle 140 Contact angle 160
0.5 0.5 0.5 0.5


0.4 0.4 0.4 0.4
0.3 0.3 0.3 0.3
0.2 0.2 0.2 0.2
0.1 0.1 0.1 0.1
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S ) Initial oil saturation (S )
oi oi oi oi
Fig. 10— Performance of the new trapping model: initial–residual curves calibrated against pore-network simulation data.
between the trapped oil saturation Sot and the initial oil
saturation Soi :
2
Sot = αSoi − βSoi . (9)
2
The parameters α and β correspond to the initial slope α
and the curvature of this curve, respectively. These pa- β
rameters were tuned to minimize the least squared error
1.5
between the model prediction and the pore-network sim-
ulation data. The optimization is constrained by the fol-
lowing restrictions:
1
0 ≤ α ≤ 1, β ≥ 0. (10)
For an initial slope greater than 1, the trapping model 0.5

would predict more trapped oil than what was originally
present, which is not physically possible. The “optimal”
parameters α and β are shown in Fig. 11 as functions of 0
50 100 150
the intrinsic contact angle. Contact angle (θ)
The performance of the optimization is illustrated in
Fig. 10. Notice that for water-wet media, the model Fig. 11— Dependence of parameters α and β of the proposed
tends to slightly overestimate the trapped oil saturation quadratic trapping model on the intrinsic contact angle of ini-
when the initial oil saturation is high. This is because tially oil-filled pores for a network model of a Berea sandstone.
of the constraint in the optimization, and the inability
of the model to achieve the desired curvature at the de-
sired location. The important consideration is that the
trapping model reproduces the observed trapping behavior we propose is an extension of Land’s model to account
for all wettability conditions, even if it may slightly over- for the different pore occupancies at different wettability
estimate the trapping of the bounding waterflood curve conditions.
(Soi = So,max = 1 − Swc ). We should also mention that
the pore-network simulator is likely to underestimate the Land waterflood model. As a prelude to the devel-
trapping for the bounding curves due to an overly opti- opment of the new waterflood relative permeability model
mistic criterion for the stability of oil layers. If a new, proposed in this work, we revisit the derivation of Land’s
free-energy based stability criterion is implemented,45 we relative permeability model. The basis of Land’s formula-
expect a better agreement between the trapping model and i
tion is to express the waterflood relative permeability kro
the pore-network predictions. at a given oil saturation (So ) as being equal to the drainage
d
When the parameters calculated from the optimization permeability kro evaluated at a flowing oil saturation Sof
are employed, the resulting trapping surface is shown in (see Fig. 13):
Fig. 12. This surface should be compared with the one i d
kro (So ) = kro (Sof ). (11)
obtained from pore-network simulations (Fig. 3).
Oil relative permeability (kro(w))

0.8
0.6 d S
kro(w) oi
0.6
0.4
0.4 ki
ro(w)
0.2
S
of
∆S S
0.2 o o
0
150 S (S )
ot oi
Co 100 0.8 0
nta 0 0.2 0.4 0.6 0.8 1
0.6
ct
an 0.4 n( S oi) Oil saturation (S )
o
gle
50 0.2 ratio
(θ) satu
0 l oil
Initia
Fig. 13— Trapped and flowing saturations used to determine
the waterflood relative permeability from the drainage relative
permeability.
At any bulk saturation So , the flowing oil saturation Sof

and the trapped saturation ∆So are related by
0.6
0.5 ∆So = So − Sof . (12)
0.4
Land makes the assumption that the trapped saturation
0.3 ∆So is the cumulative trapped saturation at a given point
0.2 in the waterflood process and that this quantity increases
0.1
as more of the flowing saturation becomes trapped. He
assumes that the maximum amount of cumulative trap-
0
0 ping, equal to the trapped saturation determined by his
50 0.2 ) trapping model (Equation (5)), occurs when the flowing
Con 0.4 n S oi
(
tac t an100 0.6 t u r atio saturation becomes zero (So = Sot (Soi )). It is important
a
gle (
θ 150 0.8 ial oil s
)
Init to note that in Land’s formulation it is necessary to obtain
the maximum trapped oil saturation Sot,max from a core-
flood experiment in order to extract the appropriate Land
Fig. 12— Trapping surface determined with the new trapping
trapping coefficient C.
model.
The intermediate trapped saturation ∆So is equal to the
cumulative trapped saturation Sot minus the amount of oil
that is still flowing and will eventually be trapped:
The waterflood relative permeability model.
Most existing relative permeability hysteresis mod- ∆So = Sot − Sot (Sof ), (13)
els12, 15, 16, 20, 21 either require a bounding waterflood curve
or model this curve according to Land’s11 waterflood rel- where
ative permeability model. The development of his model Soi
is described below. The new relative permeability model Sot ≡ Sot (Soi ) = (14)
1 + CSoi
o o o o
θ = 20 , C = 1.703 θ = 40 , C = 1.812 θ = 60 , C = 1.807 θ = 80 , C = 26.88
1 d 1 1 1
k (data)
rg
i

k (model)
ro
rg
0.8 o i
k (data) 0.8 0.8 0.8
rg
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (S ) Oil saturation (S ) Oil saturation (S ) Oil saturation (S )
o o o o
o o o o
θ = 100 , C = 293.7 θ = 120 , C = 408.5 θ = 140 , C = 415.3 θ = 160 , C = 426
1 1 1 1

ro
0.8 0.8 0.8 0.8
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Oil saturation (So) Oil saturation (So) Oil saturation (So) Oil saturation (So)
Fig. 14— Comparison of Land’s trapping model for the bounding relative permeability curves with pore-network simulation data.
and The Land trapping model predicts the experimental data

Sof fairly well for water-wet media, but is unable to capture the
Sot (Sof ) = (15)
1 + CSof convex shape of the waterflood curve characteristic of oil-
wet conditions. Indeed, for intrinsic contact angles greater
Substituting Equations (13)–(15) in Equation (12) one ob-
than 80◦ , Land’s model predicts reversible relative per-
tains
meability curves—waterflood relative permeability curves
2 1 coincide with the primary drainage curves.
Sof − (So − Sot )Sof − (So − Sot ) = 0. (16)
C
A new waterflood model. Land’s waterflood model
Solving this quadratic equation for Sof and taking the pos- hinges on the assumption that the trapped saturation in-
itive root: creases monotonically during waterflooding. This assump-
" r # tion does not allow reproduction of the convex shape of the
1 4 observed relative permeability curves in intermediate-wet
Sof = (So − Sot ) + (So − Sot )2 + (So − Sot )
2 C and oil-wet media. We modify Equation (13) as follows:
(17) ∆So = Sot (Soi ) − Sot (Sof ) − γ(So − Sot (Soi ))(So − Soi ).
Although Land’s assumptions are generally valid for (18)
water-wet media, they do not hold for oil-wet media. In The last term in this equation is designed to capture
water-wet media, we noticed that the experimental and the convexity of the waterflood curves in oil-wet media.
pore-network waterflood curves are sometimes above the It satisfies the following essential requirements: (1) the
drainage curves for high oil saturations (Figures 5 and 6). flowing saturation Sof equals zero when the bulk satura-
Land’s model will generally underestimate the relative per- tion reaches the ultimate trapped saturation Sot (Soi ); and
meabilities in this region under the assumption that the (2) the flowing saturation is equal to the bulk saturation
hydrocarbon phase will immediately be subjected to snap- (Sof = So ) at the beginning of the waterflood (So = Soi )
off. In oil-wet media, the ultimate residual saturations are and no oil has yet been trapped. The parameter γ is an
often very low due to oil layer drainage. Initially, there additional parameter of the formulation, which should de-
is a sharp decrease in the waterflood relative permeability pend on rock type and wettability characteristics.
(Fig. 7). However, at low oil saturations, oil layer drainage Substituting Equation (18) into Equation (12), we ob-
is the dominant mechanism, which leads to low oil resid- tain:
ual saturations achieved at very low oil relative permeabil-
ities. In Fig. 14 we compare Land’s waterflood relative Sof = So − Sot (Soi )+Sot (Sof )+γ(So − Sot (Soi ))(So − Soi ),
permeability model to the pore-network simulation results. (19)
where the trapped saturation is given by the new trapping In Fig. 17 we compare the new waterflood model with
model: pore-network simulated data for a set of intermediate scan-
2
Sot (Soi ) = αSoi − βSoi . (20) ning curves. Model predictions were obtained using the
We substitute Equation (20) in Equation (19) and solve same parameters determined from the trapping curves and
for Sof , to obtain a new model for the flowing oil satura- the bounding waterflood curves.
tion:
· Wettability correlations. In this paper we have con-
1 sistently used the intrinsic contact angle θi as a measure
Sof = (α − 1)
2β of wettability. This parameter is almost impossible to de-
termine with any certainty in the laboratory since most
¸
p
+ (α − 1)2 + 4β[So − Sot + γ(So − Sot )(So − Soi )] . rocks are characterized by a large range of contact angles.
(21) Moreover, we have used a particular model43 that links
the intrinsic contact angle with the receding and advanc-
This expression of the flowing saturation is then used in ing contact angles.
Equation (11) to evaluate the waterflood relative perme- Ideally, one would correlate the trapping parameters α
ability. and β and the waterflood parameter γ with a measure
The tuning parameter γ can be obtained from fitting of the overall wettability characteristics of the rock that
the model to experimental bounding waterflood curves. In can be determined in the lab. In fact, previous investiga-
this investigation, we used bounding waterflood curves ob- tions26, 28 have shown that pore-network models are able to
tained from pore-network simulation. The dependence of perform quantitative predictions of laboratory wettability
the parameter γ on the intrinsic contact angle is shown in measurements.
Fig. 15. The trends in this relationship are the ones ex- Common measures of wettability are the Amott wetta-
pected. For water-wet media, the parameter γ is negative, bility indices,9 Iw and Io . A strongly water-wet medium is
indicating that the Land trapping model overestimates the associated with Iw = 1 and Io = 0, whilst values of Iw = 0
trapped saturation and subsequently underestimates the and Io = 1 correspond to a strongly oil-wet medium. The
relative permeability. For oil-wet media this parameter Amott–Harvey index Iwo is probably the most popular
takes positive values, which allows for the model water- measure of wettability and is defined as:
flood curve to be below the one predicted by the Land
model. Iwo = Iw − Io , (22)
2.5 which ranges between −1 and 1.

2
These indices can be determined from two capillary pres-
sure curves corresponding to waterflood and subsequent
1.5 oil flood, and can be computed directly from pore-network
1 simulations.26, 28 The variability of the Amott oil and wa-
ter indices and the Amott-Harvey wettability index with
0.5
respect to the intrinsic contact angle is given in Fig. 18
γ
0 and Fig. 19, respectively. Although this was not pur-

−0.5 sued here, one could express the dependence of the trap-
ping and waterflood relative permeability parameters with
−1
respect to the Amott-Harvey wettability index directly,
−1.5 rather than the intrinsic contact angle.
−2
50 100 150 Simulation of geological CO2 storage
Contact angle (θ)
It is well documented that carbon dioxide is a greenhouse
Fig. 15— Dependence of parameter γ of the proposed wa- gas, and one of the main contributors to global warming.51
terflood relative permeability model on the intrinsic contact CO2 sequestration refers to the capture and long-term stor-
angle of initially oil-filled pores for a network model of a Berea age of anthropogenic CO2 in order to limit its emission to
sandstone. the atmosphere.52 Injection into geological formations is
one option to store CO2 .53, 54 Different target formations
The performance of the combined trapping and water- have been identified for this purpose, including depleted
flood models is shown in Fig. 16. Unlike Land’s trapping oil and gas reservoirs,55 unminable coal beds,56 and deep
model, the new model provides a suitable fit to the bound- saline aquifers.57, 58
ing waterflood curves determined by pore-network simu- One of the major concerns in any sequestration project
lations for all contact angles. The dark circle represents is the potential leak of the CO2 into the atmosphere. Pos-
the trapped saturation determined from the new trapping sible causes of leaks are loss of integrity of the cap rock
model. The trapped saturations predicted by the model due to overpressurization of the geological formation, and
do not always match the experimental end-points. This is abandoned wells that may be present. When planning ge-
why the model relative permeabilities do not fully agree ologic sequestration projects in saline aquifers or depleted
with the pore-network results, especially for saturations hydrocarbon reservoirs, it is therefore essential to predict
close to the ultimate residual saturation. the migration and distribution of the CO2 in the subsur-
o o o o
θ = 20 , γ = −1 θ = 40 , γ = −0.75 θ = 60 , γ = −0.1 θ = 80 , γ = 1.4
1 d
1 1 1
k (data)
ro

i
ro
k (model)
ro
0.8 i
0.8 0.8 0.8
o k (data)
ro
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
o o o o
o o o o
θ = 100 , γ = 1.65 θ = 120 , γ = 1.3 θ = 140 , γ = 1.3 θ = 160 , γ = 1.2
1 1 1 1

ro
0.8 0.8 0.8 0.8
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Fig. 16— Comparison of the new waterflood model with pore-network simulation data for the bounding relative permeability curves
(initial saturation Soi = 0.76).
o o o o
θ = 20 , γ = −1 θ = 40 , γ = −0.75 θ = 60 , γ = −0.1 θ = 80 , γ = 1.4
1 d 1 1 1
kro (data)
i
ro
kro (model)
0.8 i
0.8 0.8 0.8
o kro (data)
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
o o o o
o o o o
θ = 100 , γ = 1.65 θ = 120 , γ = 1.3 θ = 140 , γ = 1.3 θ = 160 , γ = 1.2
1 1 1 1

ro
0.8 0.8 0.8 0.8
0.6 0.6 0.6 0.6
0.4 0.4 0.4 0.4
0.2 0.2 0.2 0.2
0 0 0 0
0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1
Fig. 17— Comparison of the new waterflood model with pore-network simulation data for scanning relative permeability curves
(initial saturation Soi = 0.65).
0.8
Wettability index
0.6
0.4
I
0.2 o
I
w
0
50 100 150
Fig. 20— Schematic of the trail of residual CO2 that is left be-
Contact angle (θ) hind due to snap-off as the plume migrates upwards [courtesy
Fig. 18— Amott oil and water wettability indices for all intrin- of Marc Hesse].
sic contact angles.
Basis for trapping in geologic CO2 sequestration.

1 Saline aquifers, predominantly water-wet, are prime candi-
dates for geologic CO2 sequestration. In water-wet media
and a capillary-dominated flow regime, snap-off is the dom-
Wettability index
0.5 inant trapping mechanism. Trapping of the non-wetting

gas phase occurs during waterflooding when the gas sat-
uration is decreasing and the water saturation increases
0 as it invades the pore-space. During the injection of CO2
in the geologic formation, the gas saturation increases in a
drainage-like process. Vertical flow paths are created as the
−0.5 gas phase migrates laterally away from the injection well
and to the top of the aquifer due to buoyancy forces. Once
the injection stops, the CO2 continues to migrate upwards.
−1 At the leading edge of the CO2 plume, gas continues to
50 100 150 displace water in a drainage process (increasing gas satu-
Contact angle (θ) ration), whilst at the trailing edge water displaces gas in
Fig. 19— Amott–Harvey wettability index for all intrinsic con-
an imbibition process (increasing water saturations). The
tact angles. presence of an imbibition saturation path leads to snap-
off and, subsequently, trapping of the gas phase. A trail
of residual, immobile CO2 is left behind the plume as it
face structure so that injection can be maximized while migrates upwards (Fig. 20).
keeping the risk of leakage at minimum. We perform a series of representative simulations to as-
Due to the density difference between the CO2 and the sess the impact of trapping and relative permeability hys-
brine, the low viscosity CO2 tends to migrate to the top of teresis on the migration and distribution of injected CO2
the geologic structure. This upward migration is some- in a sequestration project.
times delayed or suppressed by low permeability layers
that impede the vertical flow of gas. There are several Reservoir description. We carried out simulations of
mechanisms by which the CO2 can be sequestered. These CO2 injection in a synthetic but realistic model of a
include—in order of increasing time scales—dissolution in geologic formation. We selected the PUNQ-S3 model,
the brine, gravity instabilities due to the larger density of which is a geometrically complex and heterogeneous three-
the brine–CO2 liquid mixture, and geochemical binding to dimensional geologic model originally designed as a test
the rock due to mineral precipitation. case for production forecasting under uncertainty. The
In this paper we demonstrate that physical trapping original PUNQ-S3 model is described in detail elsewhere,59
mechanisms during the simultaneous flow of two fluid and the model data is publicly available for download.60
phases through the porous medium, such as snap-off, We modified the original model slightly to study hystere-
have a huge impact on the migration and distribution sis and trapping effects in a CO2 injection scenario. The
of CO2 . Moreover, these trapping mechanisms act at a geometry of the model is characterized by a dome in the
much smaller time scale than other sequestration mecha- center. It is bounded by sealing faults on all sides and
nisms such as aqueous dissolution and mineral precipita- contains five layers of fluvial sand and shale. The top of
tion. Therefore, relative permeability hysteresis becomes the reservoir is at a depth of 7677 ft. The average reser-
an order-one factor in the assessment of CO2 sequestration voir thickness is 92 ft. The reservoir is discretized into
projects. 19 × 28 × 5 grid blocks, of which 1761 blocks are active.
The x and y dimension of each block is 590 ft. The average

1
porosity is 0.2, and the average horizontal permeability is
Drainage kd S =0.69
100 md. The anisotropy ratio is about 3. A map of the rg gi
i
absolute permeability is shown in Fig. 21.
Gas relative permeability

0.8 Imbibition krg
0.6
0.4
0.2
Sgt=0.4
0
0 0.2 0.4 0.6 0.8 1
Gas saturation (S )
g
1
Swc=0.31
)
rw
0.8
Water relative permeaility (k

0.6
Fig. 21— Horizontal permeability distribution (md) of the
PUNQ-S3 model.
0.4
9 3
The reservoir pore volume is approximately 10 ft with
an initial reservoir pressure of 1020 psia at the top of the 0.2
structure. Our model has eight injection wells open to the
bottom layer of the reservoir. The injection wells are rate- 0
controlled and operate with a constraint in the maximum 0 0.2 0.4 0.6 0.8 1
bottom hole pressure of 3000 psia. We assume that the Water saturation (S )
w
faults on all sides of the reservoir act as no-flow boundaries
so the effects of regional groundwater flow through the Fig. 22— Relative permeability curves used in the CO2 sim-
reservoir are not included in this study. ulations, taken from Oak’s dataset46 for a water-wet Berea
The relative permeabilities of water and gas are taken sandstone.
from the Oak dataset46 for a water-wet Berea sandstone
and a gas–water system. The relative permeability curves
2. Injection rate, by comparing the results of Case 2
are plotted in Fig. 22. Because the medium is strongly
(CO2 injection during one year) and Case 3 (injection
water-wet, hysteresis effects are significant in the gas rel-
of the same volume over a period of ten years).
ative permeability only. We simulate hysteresis effects by
providing the simulator with the bounding drainage and 3. Injection of water, by comparing Case 4 (injection of
waterflood curves from the experimental data (Fig. 22). 0.05 pore volumes of water following CO2 injection)
These data result in a Land trapping coefficient C ≈ 1. with Case 3, respectively (injection of CO2 alone).
Intermediate scanning curves are computed using Kil-
lough’s15 hysteresis model. The PVT properties of the gas For each case, we show results of the fluid distribution af-
(density, viscosity and formation volume factor) are repre- ter 50 years from the beginning of the injection phase. We
sentative of the properties of CO2 at reservoir conditions. plot three-dimensional views of the water saturation Sw
(saturation of CO2 is simply Sg = 1 − Sw ). Additional
Setup of numerical simulations. The formation is ini- insight into the dynamic behavior of the system is gained
tially filled with brine. CO2 injection is simulated by con- by plotting the evolution of the CO2 saturation at specific
trolling the volume of CO2 that is injected into the for- gridblocks. We chose three “observation” points: one near
mation. A total of 0.15 pore volumes are injected into the the top of the anticline structure (block 13, 18, 1), one at
bottom layer of the aquifer. The injection rate and volume a slightly lower elevation (block 7, 21, 1) and a third one
is the same for all eight injectors. at a lower elevation still (block 11, 11, 1). All of them are
We simulated four different scenarios, summarized in located at the top layer of the formation. The location of
Table 1. They were designed to assess the following fac- the injection wells and the observation gridblocks is shown
tors: in Fig. 23.
1. Hysteresis and trapping, by comparing the results of Effect of hysteresis and trapping. We begin by illus-
Case 2 (in which hysteresis is modeled) and Case 1 (in trating the dramatic effect of relative permeability hystere-
which hysteresis is not modeled). sis on the predictions of the fate of the injected CO2 . We
Table 1— Summary of CO2 sequestration simulations.

Case no. Description Injection scheme Injection time
1 No hysteresis 0.15 PV gas 1 yr
2 Hysteresis 0.15 PV gas 1 yr
3 Hysteresis 0.15 PV gas 10 yr
4 Hysteresis-water 0.15 PV gas, 0.05 PV water 10yr, 1yr
and geochemical binding (more interfacial area between

INJ-07 (11,11,1)
the CO2 and the initial pore water).
INJ-03 Fig. 25 shows the evolution of CO2 saturation with time
INJ-02
(13,18,1) at the three different observation gridblocks in the reser-
INJ-08
voir. For the block at the very top (13, 18, 1) we see the
INJ-06 accumulation of CO2 when hysteresis effects are ignored,
INJ-04 reaching a saturation value close to 0.7 that corresponds
INJ-05
to the connate water saturation. In contrast, very little
INJ-01
accumulation (CO2 saturation of about 0.15) occurs when
hysteresis is accounted for. The evolution of the CO2 sat-
uration in blocks at lower elevations (7, 21, 1 and 11, 11, 1)
displays an interesting behavior. When hysteresis is ig-
(7,21,1)
nored, the gas saturation first increases sharply—the lead-
2340 2350 2360 2370 2380 2390 2400 2410 2420
ing edge of the plume reaches the block during its migra-
tion upwards—and then decreases to a very low value—the
plume travels through the block without leaving any resid-
Fig. 23— Three-dimensional view of the depth (ft) of the top
ual CO2 . On the other hand, simulations that account for
of the PUNQ-S3 formation. Also shown are the injection wells
and the “observation” grid blocks.
hysteresis predict that the CO2 saturation decreases only
to a finite, positive value. This is due to trapping during
the imbibition process that occurs at the trailing edge of
compare the results from Case 1 (no hysteresis) and Case 2 the plume, which results in residual CO2 being left behind.
(with hysteresis). In Case 1, the gas relative permeability
is assumed to be reversible, and only the drainage curve is Effect of injection rate. We now investigate the effects
used. Both cases simulate injection of a total of 0.15 pore of CO2 injection rate on the overall performance of the
volumes of CO2 during one year, and the migration of the sequestration project. We do so by comparing Cases 2
CO2 plume during the next 49 years. and 3, both of which account for hysteresis. We inject the
In Fig. 24 we plot the distribution of water saturation same amount of CO2 in both cases, but over a period of
predicted by both models after 50 years from the beginning 1 year in Case 2 and over a period of 10 years in Case 3.
of injection. In Case 1, because the gas relative permeabil- The saturation distributions for both of these cases are
ity is assumed to be reversible, the model does not predict shown in Fig. 26. Clearly, a slower injection rate leads to
any trapping of CO2 . The CO2 plume migrates upwards more mobile CO2 reaching the top of the aquifer, which
due to buoyancy forces without leaving any residual sat- has an adverse effect for sequestration purposes. These
uration behind. After a sufficiently long time, the model results can be explained as follows.
predicts the formation of a gas cap of mobile CO2 at the Higher injection rates lead to a more radial displacement
top of the formation. This scenario is unfavorable from a pattern and higher gas pressures in the vicinity of wells.
sequestration standpoint: damage in the cap rock could Physically, this higher pressure is responsible for the non-
lead to fractures that might serve as conduits for leaks of wetting gas to invade smaller pores, which have a higher
the mobile CO2 to upper formations and, eventually, the capillary entry pressure. Snap-off occurs readily in smaller
atmosphere. pores during imbibition, resulting in increased macroscopic
The predictions under Case 2 are entirely different. Af- trapping.
ter the injection phase, the model predicts a trail of resid- For lower injection rates, on the other hand, gravity af-
ual, immobile CO2 during the migration of the plume. As fects the displacement pattern of water by CO2 early. The
a result, while there is a net flow of CO2 in the vertical low-viscosity gas forms stable paths through the high per-
direction, trapping prevents the injected CO2 from form- meability regions of the porous medium. Microscopically,
ing a gas cap. In fact, the simulations predicts that, after only the largest pores are invaded, which leads to reduced
50 years or less, almost all the CO2 is trapped in the forma- snap-off during an eventual imbibition process. Moreover,
tion. Accounting for hysteresis effects leads to a spread-out the CO2 reaches the top of the formation before injection
distribution of trapped CO2 , as opposed to a concentrated ceases. Since trapping only occurs during imbibition (wa-
distribution of mobile CO2 . This scenario is in fact much ter displacing CO2 at the trailing edge of the plume), the
more realistic and, importantly, much more favorable for fraction of CO2 that has reached the top of the aquifer
the effectiveness of CO2 sequestration: it minimizes the cannot undergo imbibition and is therefore not subject to
risk of leaks (the gas is immobile) and enhances other se- trapping.
questration mechanisms such as dissolution into the brine The plots of CO2 saturation over time for these two cases
INJ-07 INJ-07
INJ-03 INJ-03
INJ-02 INJ-02
INJ-08 INJ-08
INJ-06 INJ-06
INJ-04 INJ-04
INJ-05
INJ-05
INJ-01
INJ-01
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. 24— Water saturation distributions after 50 years from the beginning of CO2 injection. Left: results from Case 1 (no hysteresis).
Right: results from Case 2 (with hysteresis).
Block (13,18,1) Block (7,21,1) Block (11,11,1)

1 0.2
Hysteresis Hysteresis Hysteresis
No hysteresis 0.6 No hysteresis No hysteresis
0.8
0.15
0.5
0.6 0.4
Sg
Sg
Sg
0.1
0.4 0.3
0.2
0.05
0.2
0.1
0 0 0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
years years years
Fig. 25— Evolution of CO2 saturations at the three “observation” gridblocks for Case 1 (no hysteresis) and Case 2 (with hysteresis).
are shown in Fig. 27. As seen from the plots of grid- to Case 3 and Case 4 are of course identical. Once wa-
blocks 7, 21, 1 and 11, 11, 1, the system undergoes trapping ter injection starts (in Case 4) the system experiences a
of the CO2 in both cases. However, the case with low drastic change in the flow dynamics. Water starts displac-
injection rate shows that CO2 reaches a saturation value ing the CO2 radially away from the wells. This can be
of 0.7 (connate water) at the top of the anticline (grid- seen as a forced imbibition process that leads to almost
block 13, 18, 1), whereas a value of just 0.15 is achieved in immediate trapping of the CO2 phase in large portions of
the high-rate case. the reservoir. It explains why we see a sharp decrease in
CO2 saturations at the observation gridblocks—the CO2
Effect of subsequent water injection. Finally, we in- is being pushed elsewhere—down to some residual satura-
vestigate how the performance of the CO2 sequestration tion. It is worth noting that this saturation value is lower
project is affected by the injection of a slug of water af- than what would be predicted from Land’s trapping model,
ter injection of the CO2 into the aquifer. The motivation Sgt = Sgi /(1+CSgi ), because part of the CO2 is being dis-
is to enhance the imbibition process that naturally occurs placed through existing connected paths and is therefore
at the trail of the nonwetting CO2 plume as it migrates not trapped.
upwards.
In Fig. 28 we compare the fluid distributions of Case 3 Summary. We offer the following brief summary of the
and Case 4. In both cases we inject 0.15 pore volumes findings from the simulations of CO2 sequestration in
of CO2 for the first ten years. In Case 4 this is followed saline aquifers:
by injection of 0.05 pore volumes of water during a 1-year 1. Accounting for trapping and relative permeability hys-
period. These results confirm the expected response: sub- teresis of the nonwetting CO2 phase is essential in or-
sequent water injection induces more trapping and reduces der to correctly characterize the migration and final
significantly the amount of CO2 that accumulates at the distribution of the injected CO2 . Trapping occurs as
top of the aquifer. the trailing edge of the plume ascends (and is replaced
The evolution of the CO2 saturation at the observation by water) after injection stops.
gridblocks (Fig. 29) offers additional insight into the be-
havior of the displacement process. During the CO2 injec- 2. Trapping of the CO2 leads to more favorable scenarios
tion period—the first ten years—the curves corresponding for sequestration purposes: a large fraction of the CO2
INJ-07 INJ-07
INJ-03 INJ-03
INJ-02 INJ-02
INJ-08
INJ-08
INJ-06
INJ-06
INJ-04
INJ-04
INJ-05
INJ-05
INJ-01
INJ-01
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. 26— Water saturation distributions after 50 years from the beginning of CO2 injection. Left: results from Case 2 (injection
over 1 year). Right: results from Case 3 (injection over 10 years).

1 0.2
Hysteresis (1yr) Hysteresis (1yr) Hysteresis (1yr)
Hysteresis (10yr) 0.6 Hysteresis (10yr) Hysteresis (10yr)
0.8
0.15
0.5
0.6 0.4
Sg
Sg
Sg
0.1
0.4 0.3
0.2
0.05
0.2
0.1
0 0 0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
years years years
Fig. 27— Evolution of CO2 saturations at the three “observation” gridblocks for Case 2 (injection over 1 year) and Case 3 (injection
over 10 years).
is trapped and immobile for practical purposes, and the observed behavior: (1) non-monotonicity of the initial–
is more spread out throughout the aquifer, thereby residual curves, which implies that waterflood relative per-
increasing the interfacial area for subsequent dissolu- meabilities cross; and (2) convexity of the waterflood rela-
tion. tive permeability curves for oil-wet media due to film flow
of oil.
3. High injection rates result in more effective seques- We have also investigated the impact of trapping and
tration of the CO2 . A shorter injection period leaves relative permeability hysteresis in the context of CO2 se-
less time for the buoyant CO2 to reach the top of the questration projects in saline aquifers. We have provided
formation, from which it is difficult to immobilize. conclusive evidence that hysteresis is an order-one factor
in the prediction of the migration and final distribution of
4. Injection of a water slugs after CO2 injection increases the CO2 . Trapping of the CO2 occurs during the upwards
the effectiveness of the sequestration project. The in- migration of the CO2 plume, but only after injection has
jected water forces break-up of large connected CO2 stopped and the trailing edge of the plume is naturally be-
plumes, enhancing trapping and immobilization of the ing displaced by water. This imbibition process leads to
CO2 . A similar behavior would be observed if there trapping of the CO2 . A trail of residual CO2 is left be-
were a regional groundwater flow. hind as the plume migrates upwards. We have shown that
hysteresis effects have a very favorable effect on the effec-
Conclusions and outlook tiveness of CO2 sequestration. From the point of view of
the design of a CO2 project, we have shown that trapping
We have presented a new model of trapping and water- of the CO2 can be enhanced by: (1) operating at high in-
flood relative permeability. Development of the model jection rates, and (2) alternatively injecting water and CO2
is motivated by the inability of existing models to cap- in the spirit of classical WAG for enhanced oil recovery.
ture the trends observed for intermediate-wet and oil-wet
media. Due to scarcity of reliable experimental data, we
have used pore-network simulation as a means to predict
the trends in trapping and relative permeability hystere-
sis. The new model is able to capture two key features of
INJ-07 INJ-07
INJ-03 INJ-03
INJ-02 INJ-02
INJ-08 INJ-08
INJ-06 INJ-06
INJ-04 INJ-04
INJ-05 INJ-05
INJ-01
INJ-01
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Fig. 28— Water saturation distributions after 50 years from the beginning of CO2 injection. Left: results from Case 3 (no water
injection). Right: results from Case 4 (water injection after CO2 injection).

1 0.2
Hysteresis (10yr) Hysteresis (10yr) Hysteresis (10yr)
Hysteresis (10yr−WAG) 0.6 Hysteresis (10yr−WAG) Hysteresis (10yr−WAG)
0.8
0.15
0.5
0.6 0.4
Sg
Sg
Sg
0.1
0.4 0.3
0.2
0.05
0.2
0.1
0 0 0
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
years years years
Fig. 29— Evolution of CO2 saturations at the three “observation” gridblocks for Case 3 (no water injection) and Case 4 (water
injection after CO2 injection).
Nomenclature Acknowledgements
EJS, RJ and FMO gratefully acknowledge financial sup-
Roman letters port from the industrial affiliates of the Stanford University
b exponent parameter in Jerauld’s trapping model Petroleum Research Institute for Gas Injection (SUPRI-C)
C Land trapping coefficient and support from the Global Climate and Energy Project
Io Amott oil wettability index (GCEP). MJB would like to thank the sponsors of the
Iw Amott water wettability index Imperial College Consortium on Pore-Scale Modelling and
Iwo Amott–Harvey wettability index the NERC for funding.
krw water relative permeability
d
kro drainage oil relative permeability References
i
kro waterflood oil relative permeability [1] T. M. Geffens, W. W. Owens, D. R. Parrish, and R. A.
Sw water saturation Morse. Experimental investigation of factors affecting lab-
So oil saturation oratory relative permeability measurements. Petrol. Trans.
So,max maximum oil saturation AIME, 192:99–110, 1951.
Sof flowing oil saturation [2] R. Lenormand, C. Zarcone, and A. Sarr. Mechanisms of
Soi initial oil saturation the displacement of one fluid by another in a network of
Sot ultimate trapped oil saturation capillary ducts. J. Fluid Mech., 135:123–132, 1983.
∆So intermediate trapped oil saturation [3] I. Chatzis, N. R. Morrow, and H. T. Lim. Magnitude and
detailed structure of residual oil saturation. Soc. Pet. Eng.
Greek letters J., 23:311–330, April 1983.
α initial slope of IR curve in new trapping model [4] E. C. Donaldson, R. D. Thomas, and P. B. Lorenz. Wet-
β curvature of the IR curve in new trapping model tability determination and its effect on recovery efficiency.
γ film flow parameter in new waterflood model Petrol. Trans. AIME, 246:13–20, 1969.
θ, θi intrinsic contact angle
[5] W. W. Owens and D. L. Archer. The effect of rock wetta-
θa advancing contact angle bility on oil–water relative permeability relationships. J.
θr receding contact angle Pet. Technol., pages 873–878, July 1971.
[6] A. R. Kovscek, H. Wong, and C. J. Radke. A pore-level [23] J.-C. Moulu, O. Vizika, P. Egermann, and F. Kalaydjian.
scenario for the development of mixed wettability in oil A new three-phase relative permeability model for vari-
reservoirs. AIChE J., 39:1072–1085, 1993. ous wettability conditions. In SPE Annual Technical Con-
ference and Exhibition, Houston, TX, October 3–6 1999.
[7] H. T. Kennedy, E. O. Burja, and R. S. Boykin. An investi-
(SPE 56477).
gation of the effects of wettability on the recovery of oil by
water flooding. J. Physical Chem., 59:867–869, September [24] P. Egermann, O. Vizika, L. Dallet, C. Requin, and
1955. F. Sonier. Hysteresis in three-phase flow: experiments,
modeling and reservoir simulations. In SPE European
[8] T. F. Moore and R. L. Slobod. Displacement of oil by wa- Petroleum Conference, Paris, France, October 24–25 2000.
ter — Effect of wettability, rate and viscosity on recovery. (SPE 65127).
Prod. Monthly, 20(10):20–30, August 1956.
[25] A. B. Dixit, S. R. McDougall, K. S. Sorbie, and J. S. Buck-
[9] E. Amott. Observations relating to the wettability of ley. Pore-scale modeling of wettability effects and their in-
porous rock. Petrol. Trans. AIME, 216:156–162, 1959. fluence on oil recovery. SPE Reserv. Eval. Eng., 2:25–36,
[10] P. P. Jadhunandan and N. R. Morrow. Effect of wettability 1999.
on waterflood recovery for crude-oil/brine/rock systems. [26] P. E. Øren and S. Bakke. Reconstruction of Berea sand-
SPE Reserv. Eng., 10:40–46, February 1995. stone and pore-scale modelling of wettability effects. J.
[11] C. S. Land. Calculation of imbibition relative permeabil- Pet. Sci. Eng., 39:177–199, 2003.
ity for two- and three-phase flow from rock properties. [27] M. D. Jackson, P. H. Valvatne, and M. J. Blunt. Prediction
Soc. Pet. Eng. J., 8(2):149–156, June 1968. Petrol. Trans. of wettability variation and its impact on flow using pore-
AIME, 243. to reservoir-scale simulations. J. Pet. Sci. Eng., 39:231–
[12] G. R. Jerauld. General three-phase relative permeability 246, 2003.
model for Prudhoe Bay. SPE Reserv. Eng., 12(4):255–263, [28] P. H. Valvatne and M. J. Blunt. Predictive pore-scale mod-
November 1997. eling of two-phase flow in mixed wet media. Water Resour.
Res., 40:W07406, doi:10.1029/2003WR002627, 2004.
[13] F. M. Carlson. Simulation of relative permeability hys-
teresis to the nonwetting phase. In SPE Annual Technical [29] A. Al-Futaisi and T. W. Patzek. Impact of wettability
Conference and Exhibition, San Antonio, TX, October 5– on two-phase flow characteristics of sedimentary rock: A
7, 1981. (SPE 10157). quasi-static description. Water Resour. Res., 39(2):1042,
doi:10.1029/2002WR001366, 2003.
[14] C. S. Land. Comparison of calculated with experimen-
tal imbibition relative permeability. Soc. Pet. Eng. J., [30] A. Al-Futaisi and T. W. Patzek. Secondary imbibition
11(4):419–425, December 1971. Petrol. Trans. AIME, 251. in NAPL-invaded mixed-wet sediments. J. Contaminant
Hydrol., 74(1–4):61–81, 2004.
[15] J. E. Killough. Reservoir simulation with history-
[31] I. Fatt. The network model of porous media I. Capillary
dependent saturation functions. Soc. Pet. Eng. J.,
pressure characteristics. Petrol. Trans. AIME, 207:144–
16(1):37–48, February 1976. Petrol. Trans. AIME, 261.
159, 1956.
[16] J. A. Larsen and A. Skauge. Methodology for numeri- [32] M. J. Blunt. Flow in porous media—pore network models
cal simulation with cycle-dependent relative permeabili- and multiphase flow. Curr. Opin. Colloid Interface Sci.,
ties. Soc. Pet. Eng. J., 3(2):163–173, June 1998. 6:197–207, 2001.
[17] D. J. Element, J. H. K. Masters, N. C. Sargent, A. J. [33] M. J. Blunt, M. D. Jackson, M. Piri, and P. H. Valvatne.
Jayasekera, and S. G. Goodyear. Assesment of three-phase Detailed physics, predictive capabilities and macroscopic
relative permeability models using laboratory hysteresis consequences for pore-network models of multiphase flow.
data. In SPE International Improved Oil Recovery Con- Adv. Water Resour., 25:1069–1089, 2003.
ference, Kuala Lumpur, Malaysia, October 20–21 2003.
(SPE 84903). [34] R. M. Holt, E. Fjoer, O. Torsoeter, and S. Bakke. Petro-
physical laboratory measurements for basin and reservoir
[18] E. J. Spiteri and R. Juanes. The impact of relative per- simulation. Marine and Petroleum Geology, 13:383–391,
meability hysteresis on the numerical simulation of WAG 1996.
injection. In SPE Annual Technical Conference and Exhi-
[35] P. Spanne, J. F. Thovert, C. J. Jacquin, W. B. Lindquist,
bition, Houston, TX, September 26–29 2004. (SPE 89921).
K. W. Jones, and P. M. Adler. Synchrotron computed mi-
[19] C. A. Kossack. Comparison of reservoir simulation hys- crotomography of porous media: topology and transports.
teresis options. In SPE Annual Technical Conference and Phys. Rev. Lett., 73(2001–2004), 1994.
Exhibition, Dallas, TX, October 1–4 2000. (SPE 63147). [36] M. A. Ioannidis and I. Chatzis. On the geometry and
[20] R. J. Lenhard and J. C. Parker. A model for hys- topology of 3D stochastic porous media. JCIS, 229:323–
teretic constitutive relations governing multiphase flow, 334, 2000.
2. Permeability-saturation relations. Water Resour. Res., [37] S. Bryant and M. J. Blunt. Prediction of relative perme-
23(12):2197–2206, 1987. ability in simple porous media. Phys. Rev. A, 46:2004–
[21] M. J. Blunt. An empirical model for three-phase relative 2011, 1992.
permeability. Soc. Pet. Eng. J., 5(4):435–445, December [38] P. E. Øren and S. Bakke. Process based reconstruction of
2000. sandstones and prediction of transport properties. Transp.
[22] R. J. Lenhard and M. Oostrom. A parametric model for Porous Media, 46(2-3):311–343, 2002.
predicting relative permeability-saturation-capillary pres- [39] M. J. Oak. Three-phase relative permeability of
sure relationships of oil-water systems in porous media intermediate-wet Berea sandstone. In SPE Annual Tech-
with mixed wettability. Transp. Porous Media, 31:109– nical Conference and Exhibition, Dallas, TX, October 6–9,
131, 1998. 1991. (SPE 22599).
[40] P. E. Øren, S. Bakke, and O. J. Arntzen. Extending pre- [59] F. J. T. Floris, M. D. Bush, M. Cuypers, F. Roggero,
dictive capabilities to network models. Soc. Pet. Eng. J., and A. R. Syversveen. Methods for quantifying the un-
3(4):324–336, December 1998. certainty of production forecasts: a comparative study.
[41] T. W. Patzek. Verification of a complete pore network Petrol. Geosci., 7(Sp.Iss.):S87–S96, 2001.
simulator of drainage and imbibition. Soc. Pet. Eng. J., [60] Netherlands Institute of Applied Geosciences. PUNQ Case
6(2):144–156, June 2001. Studies. http://www.nitg.tno.nl/punq/cases/index.shtml.
[42] S. Bakke and P. E. Øren. 3-D pore-scale modelling of
sandstones and flow simulations in the pore networks. Soc.
Pet. Eng. J., 2:136–149, 1997.
[43] N. R. Morrow. Effects of surface roughness on contact
angle with special reference to petroleum recovery. J. Can.
Pet. Technol., 14:42–53, 1975.
[44] M. J. Blunt and P. King. Relative permeabilities from
two- and three-dimensional pore-scale network modeling.
Transp. Porous Media, 6:407–433, 1991.
[45] M. I. J. van Dijke, M. Lago, K. S. Sorbie, and M. Araujo.
Free energy balance for three fluid phases in a capillary of
arbitrarily shaped cross-section: capillary entry pressures
and layers of the intermediate-wetting phase. J. Colloid
Interface Sci., 277(1):184–201, 2004.
[46] M. J. Oak. Three-phase relative permeability of water-
wet Berea. In SPE/DOE Seventh Symposium on En-
hanced Oil Recovery, Tulsa, OK, April 22–25, 1990.
(SPE/DOE 20183).
[47] C. R. Holmgren and R. A. Morse. Effect of free gas sat-
uration on oil recovery by water flooding. Petrol. Trans.
AIME, 192:135–140, 1951.
[48] J. R. Kyte, R. J. Stanclift Jr., S. C. Stephan Jr., and L. A.
Rapoport. Mechanism of water flooding in the presence of
free gas. Petrol. Trans. AIME, 207:215–221, 1956.
[49] S. G. Dardaganian. The displacement of gas by oil in the
presence of connate water. MS Thesis, Petroleum Engi-
neering, Texas A&M University, 1957.
[50] G. R. Jerauld. Prudhoe Bay gas/oil relative permeability.
SPE Reserv. Eng., 12(1):66–73, February 1997.
[51] P. M. Cox, R. A. Betts, C. D. Jones, S. A. Spall, and
I. J. Totterdell. Acceleration of global warming due to
carbon-cycle feedbacks in a coupled climate model. Nature,
408(6809):184–187, 2000.
[52] K. S. Lackner. A guide to CO2 sequestration. Science,
300(5626):1677–1678, 2003.
[53] G. Hitchon, W. D. Gunter, T. Gentzis, and R. T. Bailey.
Sedimentary basins and greenhouse gases: a serendipitous
association. Energy Conv. Manag., 40(8):825–843, 1999.
[54] S. Bachu. Sequestration of CO2 in geological media: cri-
teria and approach for site selection in response to climate
change. Energy Conv. Manag., 41(9):953–970, 2000.
[55] S. Holloway. Storage of fossil fuel-derived carbon dioxide
beneath the surface of the earth. Annu. Rev. Energy En-
viron., 26:145–166, 2001.
[56] G. S. Bromhal, W. N. Sams, S. Jikich, T. Ertekin, and
D. H. Smith. Simulation of CO2 sequestration in coal beds:
The effects of sorption isotherms. Chem. Geol., 217(3–
4):201–211, 2005.
[57] K. Pruess and J. Garcı́a. Multiphase flow dynamics during
CO2 disposal into saline aquifers. Env. Geol., 42(2–3):282–
295, 2002.
[58] S. Bachu. Sequestration of CO2 in geological media in re-
sponse to climate change: capacity of deep saline aquifers
to sequester CO2 in solution. Energy Conv. Manag.,
44(20):3151–3175, 2003.
View publication stats

Relative-Permeability Hysteresis: Trapping Models and Application To Geological CO2 Sequestration

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Relative-Permeability Hysteresis: Trapping Models and Application To Geological CO2 Sequestration

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Relative-Permeability Hysteresis: Trapping Models and Application to

Article · October 2005

The user has requested enhancement of the downloaded file.

Copyright 2005, Society of Petroleum Engineers, Inc. Introduction

wettability alteration, by choosing average intrinsic con-

Trapped oil saturation (Sot)

bution in originally oil-filled pores is independent of Soi . 0.25

the intrinsic contact angle after wettability alteration.

Oil relative permeability (kro(w))

value. The trapping mechanisms at this point should not

0.6 at different capillary pressure levels, this affects the shape

urations but at a low rate due to the small conductance

Oil relative permeability (kro)

Oil−Water experiment (1000 md)

3. In strongly water-wet media, the trapped oil satura-

0.6 4. In contrast, for strongly oil-wet media, the trapped

0.3 Clearly, this markedly different behavior in water-wet and

Land trapping model. The Land model11 is the most

Oil relative permeability (k

gas saturation negative, a nonphysical value.

Contact angle 20 Contact angle 40 Contact angle 60 Contact angle 80

0.5 0.5 0.5 0.5

Trapped oil saturation (Sor)

Trapped oil saturation (Sor)

Trapped oil saturation (Sor)

0.3 0.3 0.3 0.3

0.2 0.2 0.2 0.2

0.1 0.1 0.1 0.1

0.5 0.5 0.5 0.5

Trapped oil saturation (Sor)

Trapped oil saturation (Sor)

Trapped oil saturation (Sor)

0.3 0.3 0.3 0.3

0.2 0.2 0.2 0.2

0.1 0.1 0.1 0.1

For an initial slope greater than 1, the trapping model 0.5

Oil relative permeability (kro(w))

At any bulk saturation So , the flowing oil saturation Sof

Oil relative permeability (kro)

Oil relative permeability (kro)

Oil relative permeability (kro)

Oil relative permeability (k )

0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

Oil relative permeability (kro)

Oil relative permeability (kro)

Oil relative permeability (k )

0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

and The Land trapping model predicts the experimental data

2.5 which ranges between −1 and 1.

0 and Fig. 19, respectively. Although this was not pur-

Oil relative permeability (kro)

Oil relative permeability (kro)

Oil relative permeability (kro)

Oil relative permeability (k )

0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

Oil relative permeability (kro)