SPE 113937

Time-Dependent Injectivity During CO2 Storage in Aquifers

McMillan Burton, SPE, Navanit Kumar, SPE, and Steven L. Bryant, SPE, The University of Texas at Austin
Copyright 2008, Society of Petroleum Engineers

This paper was prepared for presentation at the 2008 SPE/DOE Improved Oil Recovery Symposium held in Tulsa, Oklahoma, U.S.A., 1923April2008.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.


Abstract
Economic considerations are likely to drive operators of geologic storage projects to inject as much CO
2
as possible into the
smallest number of wells. In practice, then, a CO
2
injection well is likely to run at the largest bottomhole pressure that is safe.
However, a constant-pressure well exhibits a varying rate of CO
2
injection because of classical multiphase flow effects, and
also because long-term injection of CO
2
dries out the near-wellbore region. Drying removes water but precipitates dissolved
salts, so the permeability of the dry rock need not equal the initial aquifer permeability. In addition to mobility of CO
2
in the
dried rock, the injectivity depends strongly on the mobility of the two-phase flow region. The difference in injectivity for
seven different measured CO
2
-brine relative permeability curves is substantial. Characterizing relative permeability will
therefore be an important consideration for the practical implementation of CO
2
storage projects. The analytical expressions
developed here show that injectivity variation can be understood in terms of phase mobilities and the speeds of saturations
fronts. The expressions can refine the estimate of the number of wells needed for a target overall injection rate. Well count in
turn strongly affects the economics of sequestration projects.

Introduction
Anthropogenic CO
2
emissions now exceed 7 Gt C per year. Geologic sequestration of CO
2
can mitigate the effect of these
emissions, but very large injection rates will be needed. For example, injecting 1 Gt C per year at typical deep saline aquifer
conditions would correspond to a reservoir conditions flow rate similar to the current global production of oil (over 80 million
barrels per day). Thus, high injectivity will be essential to reduce the cost of geologic sequestration at this scale.
The first order effect on injectivity is the product of formation permeability and formation thickness, kh for vertical wells,
or kL for horizontal wells, L being the length of the horizontal section. All else being equal, subsurface formations with large
kh will be preferred targets for storage. The main message of this paper is that all else is not equal, and kh (or kL) should not be
the only criterion for site selection. Mutual solubility (of CO
2
in the aqueous phase and H
2
O in the CO
2
-rich phase) and the
relative permeabilities to brine and CO
2
can alter injectivity by a factor of four in the formations examined here. Moreover, the
injectivity will vary with volume injected.
To understand injectivity of CO
2
into a brine aquifer, we identify three regions of flow that will develop in the aquifer and
then evaluate the mechanisms that control flow in those regions. The regions are (1) dry CO2 in the near-wellbore region; (2)
the Buckley-Leverett or two-phase region, and (3) the brine region far from the well. As shown in Fig. 1, the three regions are
separated by two fronts that move at characteristic speeds. Because of mutual solubility, the situation differs from classical
two-phase immiscible flow, which has a single front separating two regions.
We develop a simple 1D model of injection at constant pressure using Darcys Law and a modified form of Buckley-
Leverett fractional flow theory, which accounts for partial solubility. The front speeds and region mobilities can be evaluated
directly from the inputs of phase behavior, relative permeability curves, and viscosities. We use this model to evaluate
sensitivities to features of the relative permeability curves (endpoints, curvature, etc.); phase behavior (temperature and
pressure dependence of solubility); and salt precipitation due to drying. We apply the model to a sandstone formation in
Alberta, and then evaluate the effect of the relative permeability on injectivity using laboratory measurements on several
formations in the Alberta Basin. The data were reported by Bennion and Bachu (2005) for CO
2
and brine (both saturated) on
sandstones and carbonates from different target aquifers. These measured relative permeability curves differed significantly
between samples.
In the modified fractional flow theory developed by Noh et al. (2007), all phase behavior is approximated by retardation
factors that account for the mutual solubility. These factors have a small effect on the speed of the Buckley-Leverett front and
a large effect on the speed of the drying front. We examine the sensitivity of these factors to salinity of the brine, pressure, and
2 SPE 113937
temperature. Our simplified model captures the key features of the drying front documented by Ozah et al. (2005) with a
compositional reservoir simulator (Ngheim et al., 2006) that accounts for geochemical reactions and full phase behavior.
Different authors have noted the potential for porosity and permeability reduction due to mineral precipitation during dry
gas injection (Motealleh and Bryant, 2007; Zuluaga, 2005). As the drying front moves deeper into the formation, large
saturations of brine will be evaporated precipitating all remaining salt which will in turn decrease the permeability to the dry
gas phase. We examine the impact of this effect and a method to measure it in the laboratory.

Modeling Approach
Overview and Assumptions. We use a simplified reservoir model to develop our theory for CO
2
injection. The reservoir is
radial and one-dimensional, Fig. 1, extending to a drainage radius of r
e
= 10,000ft. It has homogeneous properties. We assume
constant pressure boundaries at the wellbore and the drainage radius. The injection rate, q, will vary with time as the Buckley-
Leverett and drying fronts advance. We assume that the pressure gradients can be computed from the flow rate at any instant
as though the flow were at steady state. We further assume that fractional flow theory can be applied to this situation of
variable flow rate.
Compressibility is ignored, and viscosities of each phase are assumed constant and independent of composition. Wherever
the CO
2
-rich and aqueous phases are in contact, we assume they are in chemical equilibrium. Temperature is assumed
constant, and no geochemical reactions (e.g. between the CO
2
-saturated brine and minerals comprising the formation) occur.
Precipitation of the dissolved solids is considered after the basic theory is developed.

Phase Mobility and Relative Permeability Curves. The mobility M of a phase j is given by
=
rj
j
j
k
M

(1)
where k
r
is relative permeability and is viscosity. The fractional flow of phase j is given by
=
+
j
j
j i
M
f
M M
(2)
where i denotes the other phase.
Relative permeability is assumed to depend only upon phase saturation, so the fractional flow of a phase also depends only
on saturation. Bennion and Bachu (2005) measured relative permeability curves on seven cores from deep formations in
western Canada. The experiments were performed with CO
2
-saturated brine and brine-saturated CO
2
to insure that drying
fronts did not affect the measurements. We fit their reported data to the following Corey-type equations:
1
1

=

g
rw
wr
m
S
k
S
(3)
and,

1

=

g
o
rg rg
wr
n
S
k k
S
(4)
where w and g indicate aqueous and CO
2
-rich phases, respectively.

Relationship Between Pressure Drop and Flow Rate. Injecting anhydrous CO
2
into a deep saline aquifer produces three
distinct regions of flow (Fig. 1). Farthest upstream (closest to injection well) is single-phase dry CO
2
flowing in dry rock. The
drying front separates this region from the two-phase region, where the phase saturations vary continuously with position. The
Buckley-Leverett front separates this region from the single-phase brine.

CO
2
Region. The total pressure drop will be the sum of the pressure drops across these three regions. The pressure drop
across the single-phase CO
2
region is found using Darcys law:


, 1
1
ln
2 2
=

= = =




dry dry
w w
p r
g g dry
dry
r Sg w
p r
q q r
p dp dr
kh r hkk r


(5)
where
, 1 = r Sg
k is the permeability reduction due to salt precipitation.
SPE 113937 3
The flow rate q is a function of time, but for conciseness we do not write that dependence explicitly. The final expression
on the right hand side has been modified to include the factor k
r,Sg=1
. This factor satisfies the following bounds:

, , 1
,
1
=
< <
r g r Sg
g dry
k k
S
(6)
The lower bound corresponds to the relative permeability of the CO
2
-rich phase just downstream of the drying front (see
Fig. 1). This bound arises because the precipitated salt cannot occupy more pore space than the water originally containing the
salt. The upper bound corresponds to zero reduction in absolute permeability due to precipitation. Experiments (Zuluaga,
2005) suggest that typical values of
, 1 = r Sg
k are 0.7 to 0.85.

Permeability Reduction Model. We illustrate a simple estimate of k
r,Sg=1
by assuming that the salt will be distributed
evenly in the pore space. (In practice it is more likely to precipitate around the pore throat and at grain contacts (Motealleh and
Bryant 2007)). The salinity (S) expressed in parts per million (ppm) in solution can be converted to volume fraction (V
f,salt
) if
we know the density of the salt (
salt
) and the density of the solution (
sol
).

,
6
10

=

sol
f salt
salt
S
V

(7)
The reduced porosity is found by subtracting the volume of salt in the remaining volume of water from the original
volume:

, ,
(1 ) =
o o g dry f salt
S V (8)
The permeability reduction as described by Kozeny-Carman is:

3 2
, 1
1
1
=

= =

o
r Sg
o o
k
k
k


(9)
The permeability ratio in Eq. 9 is just the factor
, 1 = r Sg
k .

Two-phase Region. The pressure drop across the two-phase Buckley-Leverett region can be approximated by:

1 1
,
1
ln
2 2



= = + +





BL BL
dry dry
p r
rg rg BL
rw rw
BL
g w g w
dry
p r
S
g avg
k k r q q
k k
p dp dr
kh r kh r


(10)
The final expression on the right hand side simplifies the analysis considerably. The approximation consists of replacing
the radial position-weighted mobility with the mobility at some average saturation within the two-phase region. This is
reasonable if the range of saturations in the two-phase region is narrow. This is the case for the measured relative permeability
curves used in this work. For this work, we set
, , ,
( ) / 2 = +
g avg g dry g BL
S S S , where S
g,dry
is the CO
2
phase saturation in the
two-phase region just downstream of the drying front, and S
g,BL
is the CO
2
phase saturation in the two-phase region just
upstream of the Buckley-Leverett front.

Brine Region. The pressure drop across the single-phase brine region is:

1
ln
2 2

= = =



e e
BL BL
p r
w w e
BRINE
BL
p r
q q r
p dp dr
kh r kh r


(11)
The total pressure drop across the reservoir yields the following:

1
, 1
,
ln ln ln
2

=



= + + = + + +






g dry
rg e BL
rw
dry BL brine w
g w
r Sg sf dry BL
S
g avg
r
k r r q
k
p p p p
kh k r r r

(12)
Recall that the total pressure drop is prescribed, and our goal is to find the instantaneous flow rate q. This is easily
accomplished if the positions of the fronts r
dry
and

r
BL
are known. The frontal positions are determined in the next section from
the dimensionless cumulative injected volume (t
D
) and the fractional flow curves. Thus Eq. 12 is an implicit equation for q.

4 SPE 113937
Modified Buckley-Leverett Theory
The speed of the fronts and the phase saturations on either side of the fronts can be determined by using the fractional flow
curve, constructing tangents to that curve which account for conservation of mass of CO
2
in the fluid phases (Noh et al., 2007).
Buckley-Leverett theory for radial flow states:

2
=
D
g g
D g
g
r
df
t dS
S
S
(13)
We apply it to the breakthrough saturation, S
g,dry
or S
g,BL
. The dimensionless cumulative injected volume, t
D
, is defined as:

2
( )
=

t
D
o
e
q t
t dt
r h
(14)
The dimensionless radius, r
D
, evaluated at one of the breakthrough saturation, S
g,dry
or S
g,BL
, defined as:
=
D
e
r
r
r

(15)
where r is the frontal position of the dry or the Buckley-Leverett region.
We also define an addition term called dimensionless velocity. This term is the velocity of the fronts in dimensionless time
and radius as shown in Fig 2.

2
=
D
D
D
r
v
t
(16)
From Fig. 3, the dimensionless velocity of the drying front as defined by Noh et al. (2007) is:

,
,
, ,

= =

BL dry
g dry g
D dry
g
BL dry
g dry g dry
f D
df
v
dS
S D
S
(17)
Also, the dimensionless velocity of the Buckley-Leverett front is:

,
,
, ,

= =

brine BL
g BL g
D BL
g
brine BL
g BL g BL
f D
df
v
dS
S D
S
(18)
Similar to S
g,dry
and S
g,BL
, f
g,dry
is the fractional flow just upstream of the drying front, and f
g,BL
is the fractional flow just
downstream of the Buckley-Leverett front. D
BLdry
and D
brineBL
are defined by the phase concentrations in the different
regions (Noh et al., 2007).

2
2 2
,
, ,

BL
CO a
brine BL
BL BL
CO a CO g
C
D
C C
(19)

2 2
2 2
, ,
, ,

dry BL
CO g CO a
BL dry
BL BL
CO g CO a
C C
D
C C
(20)
The concentrations, C (with units of mol/volume), have a superscript which defines the region and a subscript which
defines the phase (i.e. CO
2
,g is CO
2
in the gas phase and CO
2
,a is CO
2
in the aqueous phase). Because D
BLdry
and D
brineBL

are functions of concentration and concentration is a function of phase behavior, D
BLdry
and D
brineBL
are functions of
temperature, pressure, and salinity. We used PVTSim software (2004) for the concentration calculations which models the
phase behavior with a Peng-Robison equation of state. As pressure in the reservoir will change with time, some judgment was
required for the determination of the concentrations. The pressure in the dry region was assumed to be the bottomhole
injection pressure, and the pressure in the Buckley-Leverett region was assumed to be 200psi above the drainage radius
pressure (see footnote in Table 1). As will be shown later, changes to pressure, temperature, and salinity have only a relatively
small influence on injectivity.
SPE 113937 5
Now that we can evaluate the fractional flow curve in Fig. 3, we can understand how the fronts will be realized in the
reservoir (see Fig. 4). We also remark that the speed of the fronts will remain constant with time. Since the speeds remain
constant, their ratio, v
BL
/v
dry
, is also constant.
The radial positions needed for Eq. 12 can be solved for:

,
=
dry e D D dry
r r t v (21)

,
=
BL e D D BL
r r t v (22)
The ratio of the drying front position and the Buckley-Leverett position, r
BL
/r
dry
, is also independent of the time.

,
,
=
D BL
BL
dry D dry
v
r
r v
(23)
The ratio of the frontal positions will be an important factore in our discussion of effective mobility.

Effective Mobility of the Formation. As we inject CO
2
into our formation, the effective mobility, M
eff
, of the formation will
change because of the combined effect of each of the three regions. If the effective mobility of the formation is increasing with
time, the injectivity must increase with time. We show in the Appendix that the trend for CO
2
injectivity will typically be
monotonically increasing, the only exception being when severe permeability reduction is experienced in the dry region.

Application of Reservoir Simulation Software. Using CMG (Ngheim et al., 2006), we created a radial model with 500
radial grid blocks with two wellsone at the center and one in the grid block at the drainage radius. GEM-GHG models the
phase behavior using a Peng-Robison equation-of-state. Kumar et al. (2005) fit solubility data to the Peng-Robison equation.
For the relative permeability, this software uses a lookup table. If an endpoint relative permeability, k
o
rg
, for aqueous
saturations less than the irreducible saturation, S
wr
, are not given to the simulator, it will assume the endpoint relative
permeability for all aqueous saturations below irreducible saturation.

Results and Discussion
We use an example of the Viking sandstone of Alberta, Canada, to demonstrate the how to implement the modified Buckley-
Leverett theory. First, we discuss why mobility in each of region is the most important mechanism to understand in the
prediction of injectivity. Next, we show the sensitivity of the flow rate to relative permeability. Last, we look at the changes to
the D-terms of Eqs. 19 and 20which would be caused by temperature, pressure, and salinitywhich in turn affect the flow
rate.

Field Example. The reservoir properties used for the calculations of the Viking sandstone are found in Table 1. The relative
permeability for the Viking sandstone as measured by Bennion and Bachu (2005) is shown in Fig. 5. Using this information,
the fractional flow curves can be constructed and evaluated (see Fig. 6).
Using the derived equations and the properties in Table 1, we predict the injection rate for the reservoir and pressure drop
across each region as a function of time. A constant pressure drop of 1,000 psi is applied to the reservoir for the 10,000 days of
injection. Eq. 12 is solved at each time step by varying the flow rate using a spreadsheet solver until convergence is reached.
The process is then repeated for each additional time step. The pressure drop across each region is shown in Fig. 7. The largest
pressure drop is across the Buckley-Leverett region after a short period of time. In Fig. 8, the flow rate is shown for our theory
and from a CMG simulation of the same system. The curves differ by only 2%. Most importantly, the major trend is certainly
captured by the simple theory.

Effect of Drying Front on Injectivity. If mutual solubility between the CO
2
and brine were neglected, the drying front would
not occur and injectivity would be determined by the mobility of the two-phase flow region, i.e. by a classical Buckley-
Leverett solution. For the relative permeability curves of the Viking sandstone, the injection rate would decrease with time, the
opposite of the trend when the drying front is accounted for (see Fig. 9). Clearly, the drying front is important for estimating
the number of wells required. (See discussion of Neglecting the Drying Front in the Appendix for further details.)

Mobility. The most important mechanism affecting injectivity is the mobility in each of the three regions of flow (see Fig.
10). If M
dry
is less than M
brine
, injectivity will be increasing. If M
dry
is greater than M
brine
, injectivity will be decreasing.
What effect does the Buckley-Leverett region have on injectivity? Looking at several relative permeability curves with
different curvatures, endpoint permeabilities, and irreducible saturations, it would be unclear which curve would result in the
best injectivity. Using the evaluations techniques presented in this paper, we can now compare them directly using mobility.
The higher the mobility of the Buckley-Leverett region, the higher the flow rate, all thing being held equal (see Figs. 11 and
12). In the Viking sandstone base case, the average mobility of the Buckley-Leverett region (evaluated at the average
6 SPE 113937
saturation, S
g,avg
) is less than the brine mobility. There are instances where the average mobility of the Buckley-Leverett region
is not below the brine mobility (see Fig. 12).
We show the effective mobility as defined in the Appendix (Eq. A.1) for the Viking sandstone case in Fig. 13. The
effective mobility of the reservoir increases with time as CO
2
is injected. Originally, it begins at the mobility of the brine
region. It does not approach the mobility of CO
2
or the mobility of the Buckley-Leverett region, but is an average of all three
regions.

Sensitivity of Injectivity to Relative Permeability Curves. Two studies are carried out to show the strong influence of the
relative permeability on the mobility of the Buckley-Leverett region and, therefore, injectivity. The first uses the same Viking
sandstone reservoir as described; the only difference is a completely different relative permeability. The second study is
varying the curvature, end-point permeability, and end-point saturation parameters for the same relative permeability curves.
For the first study, we show the effect different relative permeability can have on injectivity. Each of the seven relative
permeability curves were measured by Bennion and Bachu (2005). Each was then applied to the Viking Sandstone with
properties in Table 1 to obtain Fig. 11. Notice the four-fold difference in injection rate between the Cooking Lake carbonate
relative permeability curve and the Wabamun low permeability relative permeability curve. The actual formations have
different absolute permeabilities and thicknesses as well as different relative permeabilities, so the actual variation in
injectivity need not be the same as depicted in Fig. 11. The point of this comparison is that the variation in relative
permeability will have a non-negligible effect on the injectivity. The wide range of injection rates demonstrates the need for
determining the range of relative permeability in a target reservoir of interest. Although the relative permeability curves vary,
the more specific reason for the difference is the mobility of the Buckley-Leverett region. The lower the mobility of this region
the less the injectivity (see Figs. 12 and 13).
For the second study, each parameter of the Viking sandstone relative permeability curve was varied (i.e. m, n, k
o
rg
, S
wr
from Eqs. 3 and 4). We show in Fig. 14 the variation in flow rate arising from changes to parameters of interest. Small changes
in some parameter can lead to an important change (20%) in the flow rate. Therefore, uncertainty in relative permeability leads
to uncertainty in flow rate and well count.
We also point out that the S
g,avg
used in Eqs. 10 and 12 is only an approximation of the mobility of the Buckley-Leverett
region. The studies here have shown strong sensitivity to the Buckley-Leverett region mobility which account for the four-fold
difference in injectivity. The sensitivity to S
g,avg
, which we use to determine the mobility of the Buckley-Leverett region, is
important. A perturbation of 0.02 to S
g,avg
results in a 10% change in flow rate.

Sensitivity of Injectivity to Phase Behavior. Although the viscosity, density, and composition of the phases change as
functions of pressure, temperature, and salinity, these effects are grouped into the two D-terms of Eqs. 19 and 20. In our
simple model, we have assumed the density and viscosity do not change with pressure and that the composition of the phase
remains fixed in each region. Although we do not explore all these effects to ensure that the D-terms are perfect, we do look at
the sensitivity of the flow rate to theses terms (see Fig. 15). The change of 0.02 to the D-terms is a significant change in terms
of phase compositions but results in a small change in flow rate (<5%). A preliminary sensitivity study when temperature,
pressure, and salinity were perturbed showed less than 5% variation in to flow rate. Because these sensitivities are less than
5% and much less than the relative permeability effects, we consider the phase behavior uncertainty to be less important than
to the uncertainty of relative permeability.

Sensitivity of Injectivity to Permeability Reduction in Dried Rock. The permeability reduction in the dry region is
estimated for the Viking Sandstone using Eqs. 7 through 9. The salinity for the formation is approximately 30,000ppm
(Bennion and Bachu, 2005) and S
g,dry
is 0.40. The density of the solution and of the salt are 1.019 gm/cc and 2.16 gm/cc,
respectively. The permeability reduction, k
r,Sg=1
, is only 3% in the dry region and results in a injection rate reduction of 1%.
The work by Motealleh and Bryant

(2007) and Zuluaga (2005) suggests that the permeability could be reduced much more
than 3%. However, we know that the endpoint permeability at the residual water saturation, S
wr
, is the lowest we would
reasonably expect the permeability reduction in the dry CO
2
region to be. This is because the volume of dissolved solids
cannot exceed the volume of brine. If the permeability of the dry region were reduced to the endpoint saturation permeability
for the Viking sandstone (i.e. k
r,Sg=1
=0.33 in the dry region), the flow rate would be reduced by 42%. We expect a flow rate
reduction between the 1% predicted by Kozeny-Carman and the 42% predicted at irreducible water saturation due to
permeability reduction in the dry CO
2
region. This range of permeability reduction for the Viking sandstone is shown in Fig.
16.
We propose that a measurement of this permeability reduction could be incorporated into the measurement of the relative
permeability curve. After measuring permeability at residual water saturation, many pore volumes of dry CO
2
could be flooded
through the core until all the water has been evaporated. If the salinity of brine used in the flood is similar to the in-situ
salinity, we suspect the permeability reduction measurement in the dry CO
2
region could be made. This effect is obviously
more important at high salinities and higher residual water saturations.

SPE 113937 7
Conclusion
Using Darcys Law and a modified form of Buckley-Leverett theory, we have developed a simple model for determining
injectivity of CO
2
into a homogenous brine reservoir with assumed constant pressure boundaries. This theory identifies the
three regions of flow that affect injectivity under these conditions, and it computes the phase mobilities in each region from the
relative permeability curves and the equation of state for the CO
2
/brine system at reservoir ocnditions. Because CO
2
and H
2
O
are soluble in the aqueous and CO
2
-rich phases, respectively, a drying front occurs in addition to the classical Buckley-
Leverett front. The ratio of the speeds of these two fronts is constant. Consequently, injectivity will vary monotonoically with
time. If the mobility of CO
2
in the dry region exceeds the mobility of brine in the undisturbed aquifer, as will be the case in
most formations, the injectivity will increase steadily. The trends of this simplified model agree well with simulations with full
phase behavior. As CO
2
is injected, the drying region increases with time causing an increase injectivity to increases.
Neglecting the drying front underestimates the injectivity and in most cases predicts that injectivity decreases with time.
The increase in injectivity over time occurs independently of the relative permeability curves. However the curves strongly
affect the absolute value of injectivity. We applied our model with the relative permeability curves measured for seven
formations in the Alberta Basin, assuming that all other reservoir parameters were the same. The differences in relative
permeability curves cause the injection rate to vary by a factor of four. We varied the parameters commonly used to
characterize relative permeability curves (m, n, k
o
rg
, and S
wr
) and found the flow rate varied by up to 20%. Therefore, we
conclude that quantifying the relative permeability curve is very important in determining achievable injection rate and
therefore the well count for CO
2
geologic sequestration projects.
The phase behavior of this model is summarized in two retardation factors (D-terms), which depend on temperature
pressure, and salinity. The variation of flow rate due to changes in the D-terms is less than that of the relative permeability
(Fig. 15).
Last, we recommend that laboratory measurements of relative permeability include measurements for a core dried by CO
2

injection from the state of residual saturation of the native brine. The CO
2
will evaporate the brine and precipitate salts in the
pore space. We have shown that this effect is important (Fig. 16). The permeability after drying out the core along with the
presented theory will help to determine the effect of salt precipitation on injectivity.

Acknowledgements
We are grateful to the sponsors of the Geologic CO
2
Storage Joint Industry Project at UT-Austin: Chevron, CMG, ENI,
ExxonMobil, Shell, and Luminant.

Nomenclature
C = concentration, mol/L, gmol/l
C
1
= positive constant
C
2
= positive constant
C
3
= positive constant
C
4 =
positive constant

2
,
BL
CO aq
C = concentration of CO
2
in aqueous phase in Buckley-Leverett region, mol/L
3
, gmol/l

2
,
dry
CO g
C = concentration of CO
2
in gaseous phase in dry region, mol/L
3
, gmol/l

2
,
BL
CO g
C = concentration of CO
2
in gaseous phase in Buckley-Leverett region, mol/L
3
, gmol/l
D
BLdry
= retardation factor for drying front
D
brineBL
= retardation factor for Buckley-Leverett front
f
j
=

fractional flow of j component
f
g,BL
= fractional flow of gas in twp phase region
f
g,dry
= fractional flow of gas in dry region
h = formation thickness, L, ft
k = absolute permeability, L
2
, md
k
o
= original permeability, L
2
, md
k
r
=

relative permeability
k
o
rg
=

end point relative permeability to gas
k
r,Sg=1
=

relative permeability of gas in dry region
k
rg
= relative permeability of gaseous phase
k
rw
= relative permeability of brine phase
m = saturation exponent to brine
L = Length of horizontal well, L, ft
M
BL
=

mobility Buckley-Leverett region, Lt/m, cp
-1

M
brine
= mobility of brine region, Lt/m, cp
-1

8 SPE 113937
M
dry
=

mobility of dry region, Lt/m, cp
-1

M
eff
=

effective mobility of reservoir, Lt/m, cp
-1

M
i
= mobility in I phase, Lt/m, cp
-1
M
j
= mobility in j phase, Lt/m, cp
-1

n = saturation exponent to brine
r = radius, L, ft
r
BL
= radius of two-phase region, L, ft
r
D
= dimensionless radius
r
dry
= radius of dry region with injector at center, L, ft
r
w =
wellbore radius, L, ft
r
e
= radius of drainage radius, L, ft
p
BL
= pressure at Buckley-Leverett front, m/Lt
2
,psi
p
dry
= pressure at dry front, m/Lt
2
,psi
p
e
=

bottomhole pressure at drainage radius, m/Lt
2
,psi
p
w
= bottomhole pressure, m/Lt
2
,psi
q = total flow rate, L
3
/t , bbl/d
S = salinity of brine, ppm
S
g,dry
=

gas saturation in the dry CO
2
region
S
g,avg
=

average gas saturation
S
g,BL
= gas saturation in Buckley-Leverett region
S
wr
=

residual water saturation
t = time, t, d
t
D
= dimensionless time, cumulative volume injected
v
D
=

dimensionless velocity
v
D,dry
=

dimensionless velocity of drying front
v
D,BL
=

dimensionless velocity of Buckley-Leverett front
V
f,salt
= volume fraction of salt
p
dry
= pressure drop in dry region, m/Lt
2
, psi
p
BL
= pressure drop in two phase region, m/Lt
2
, psi
p
brine
= pressure drop in brine region, m/Lt
2
, psi

g
= gaseous phase viscosity, m/Lt, cp

w
= brine viscosity, m/Lt, cp

sol
= density of solution, m/L
3
, lb/ft
3


salt
= density of salt, m/L
3
, lb/ft
3

= porosity

0
= original porosity

References
Bennion, B. and Bachu, S., 2005. Relative permeability characteristics for supercritical CO
2
displacing water in a variety of potential
sequestration zones in the western Canada sedimentary basin, Paper SPE 95547 at SPE Annual Technical Conference an Exhibition,
Dallas, TX, 9-12 October.
Kumar, A., Noh, M., Pope, G., Sepehrnoori, K., Bryant, S., and Lake, L. 2005. Reservoir simulation of CO
2
storage in deep saline aquifer,
SPEJ 10 (3): 336-348. SPE 89343-PA.
Motealleh, S., and Bryant, S., 2007. Predictive model for permeability reduction by small wetting phase saturations, Water Resource
Research, 43, W12S07.
Nghiem, L., Sammon, P., Grabenstetter, J., and Ohkuma, H. 2006. Modeling CO
2
storage in aquifers with a fully-coupled geochemical EOS
compositional simulator. Paper SPE 89474 presented at 2006 SPE/DOE Symposium on Improved Oil Recovery, Tulsa, Oklahoma, 22-
26 April.
Noh, M. Lake, L., Bryant, and Araque-Martinez, A. 2007. Implications of coupling fractional flow and geochemistry for CO
2
injection in
aquifers, SPEREE 10 (4): 406-411. SPE 89341-PA.
Ozah, R. Lakshminarasimhan, S., Pope, G., Sepehrnoori, K., and Bryant, S. 2005. Numerical simulation of storage of pure CO
2
and CO
2
-
H
2
S gas mixtures in deep saline aquifers. Paper SPE 97255 presented at 2006 SPE Annual Technical Conference and Exhibition,
Dallas, Texas, 9-12 October.
PVTSim R14. 2004. Calsep, Inc. Fluid Handling Software, Houston.
Zuluaga, A., 2005. Geochemical effects in two-phase flow. PhD Dissertation, U. of Texas at Austin, Austin, Texas.

Appendix
Three Region Mobility. Here we demonstrate that the effective mobility of the reservoir will generally increase as CO
2
is
injected and, hence, injectivity will increase accordingly. The effective mobility of fluid in this reservoir, M
eff
, can be
expressed in terms of harmonic averages of mobility in each region.
SPE 113937 9

ln
ln ln ln








= + +
dry BL
e e
dry
w w BL
eff brine BL dry
r r
r r
r
r r r
M M M M
(A.1)
The other definitions are:

,

= +


rg
rw
BL
g w
S
g avg
k
k
M

(A.2)

1
=
brine
w
M

(A.3)

, 1 =
=
r Sg
dry
g
k
M

(A.4)
Recall, that r
w
is the radius of the well and r
e
is the drainage radius. Adding and subtracting a term to the right-hand side
allows us to manipulate the equation.

ln
ln ln ln ln ln








= + + +
dry dry dry BL
e e
dry
w w w w BL
eff brine BL dry brine brine
r r r r
r r
r
r r r r r
M M M M M M
(A.5)
We can rearrange the terms so that the terms that are constant in time are in the brackets. Recall that the ratio of r
BL
to r
dry
is
constant from Eq. 23 in the Modified Buckley-Leverett theory section of the paper:

ln ln
ln ln
1 1
ln








= + +







BL BL
e e
dry dry dry
w w
eff brine BL brine w dry brine
r r
r r
r r r
r r
M M M M r M M
(A.6)
By grouping all of the constant terms from this Eq. A.6 into positive constants C
1
and C
2
, we can solve for the M
eff
:

( ) 1 2
1
1 1
ln
=

+


eff
dry
dry brine
M
C C r
M M
(A.7)
where

( )
1
ln ln
ln
1 1
ln
ln







+


=



BL BL
e
dry dry
w
w
brine BL brine dry brine
e
w
r r
r
r r
r
r
M M M M M
C
r
r

and

2
1
ln
=



e
w
C
r
r
.
The effective mobility therefore varies with time only because the drying front advances into the reservoir with time. In the
special case of the mobility of the dry region, M
dry
, being equal to the mobility of the brine region, M
brine
, M
eff
would be
constant. For typical deep saline aquifers, the likely case is that M
brine
is less than M
dry
as the viscosity of the CO
2
-rich phase is
expected to be much less than the viscosity of the brine. When M
brine

is less than M
dry
, the denominator of Eq. A.7 decreases as
10 SPE 113937
r
dry
increases. Thus M
eff
can only be increasing, and, hence injectivity can only be increasing. Because M
brine

is generally
expected to be less than M
dry
, we expect the injectivity of CO
2
into a brine reservoir to be generally monotonically increasing.
The only exceptions would be where severe permeability reduction occurs in the dry region reducing the dry region mobility,
M
dry
.

Two Region MobilityNeglecting Drying Front. For comparison we obtain the average mobility for the classical two-
phase immiscible displacement (zero mutual solubility), using the same approach as above. In this situation, the only regions
of flow would be brine (downstream of the Buckley-Leverett front) and two-phase (upstream of the front). The average
mobility, M
avg
is the harmonic average of the phase total mobilities in each region:

ln ln ln



= +
e e BL
w w BL
eff brine BL
r r r
r r r
M M M
(A.8)
The same term is added and subtracted,

ln ln ln ln ln



= + +
e e BL BL BL
w w w w BL
eff brine BL brine brine
r r r r r
r r r r r
M M M M M
(A.9)
The terms are rearranged with the constant terms in the brackets.

ln ln
1 1
ln




= +






e e
w w BL
eff brine w BL brine
r r
r r r
M M r M M
(A.10)
By grouping all of the constant terms from Eq. A.10 into positive constants C
3
and C
4
, we can again solve for the M
eff
:

3 4
1
1 1
ln
=

+


eff
BL
w BL brine
M
r
C C
r M M
(A.11)
where

3
ln
1
ln



= =



e
w
brine
brine e
w
r
r
M
C
M r
r

and

4
1
ln
=



e
w
C
r
r
.
Notice the similarity of Eq. A.7 with Eq. A.11. When M
brine

is greater than M
BL
, M
eff
and injectivity will be monotonically
increasing and for the opposite case, the injectivity will monotonically decrease. In the case of smaller M
BL
and when we
neglect the drying front, the injectivity would decrease monotonically. Neglecting the formation of the drying region will thus
underestimate injectivity and fail to predict its behavior with time.
SPE 113937 11
Table 1Viking Sandstone Properties
Property Value Unit
Permeability 30 md
Height 100 ft
Depth 4100 ft
Temperature 91 F
Initial Pressure 1800
a
psi
Bottomhole Pres. 2800

psi
Salinity 30,000 ppm
Gas Visc.,
gas
0.17 cp
Brine Visc.,
w
0.7 cp
Porosity,
13.5 %
Edge radius, r
e
10,000 ft
D
BL

Brine
1.0469
b

D
dry

BL
-0.0582
b

a
Hydrostatic pressure at depth.
b
2800 psia for dry CO
2
region and 2000psia was used for
brine and Buckley-Leverett region in the calculations of the
D parameters




Fig. 1Three regions of flow develop during CO
2
injection and a pressure drop will result over each region. The fractional flow curve
modified to account for multiphase transport of CO
2
(Noh et al., 2007) determines the position of the fronts and the saturations in the
two-phase Buckley-Leverett region.


12 SPE 113937

Fig. 2A time-distance diagram showing the flow regions and
fronts when CO
2
is injected into an aquifer. The slopes of the
lines are the speeds of the corresponding fronts. The distance
axis is plotted as r
D
2
because the flow is assumed radial. The
schematic in Fig. 1 corresponds to a vertical cross-section of
the time-distance diagram.
f
g
S
g,BL
S
g,dry
(D
BLdry
, D
BLdry
)
(D
brineBL
, D
brineBL
)
BL
v
dry
v
1
1
0
S
g
Unit slope
f
g,BL
f
g,dry

Fig. 3This fractional flow curve is evaluated by drawing
tangent lines from the D-terms shown to account for the effects
of dissolution of CO
2
into the water phase and water into the gas
phase (modified from Noh et al. 2007). Notice the small window
of saturations which will be realized in the ideal reservoir
between Sg,BL and Sg,dry. Recall, Sg,avgused to evaluated
Eq. 12is (Sg,dry + Sg,BL)/2 (Modified from Noh et al., 2007).
r
D
S
g,BL
S
g,dry
1
1
0
BL or two-phase Dry CO
2
Brine
S
g
BL D,
S
D
r r
BL g,
=
CO
2
-saturated brine
Water-saturated CO
2
dry D,
S
D
r r
dry g,
=

Fig. 4Using a radial view of the reservoir, we can visualize the
three regions of flow: dry CO
2
, Buckley-Leverett or two-phase,
and brine only. The saturations in the reservoir are Sg=0 in the
brine region, Sg=1 in the dry CO
2
region, and between Sg,dry
and Sg,BL in the two-phyase region. The values are defined by
the fractional flow curve in Fig. 3 (modified from Noh et al.,
2007).

Fig. 5The relative permeability curves for the Viking
Sandstone were measured by Bennion and Bachu (2005).
SPE 113937 13

Fig. 6The fractional flow curve for the Viking Sandstone
constructed from the measurements of Fig. 5. The significant
gas saturations are shown as down arrows and as values. The
gas saturations realized in the two-phase region of the reservoir
will be between 0.28 and 0.40. The average gas saturation,
Sg,avg = 0.34, is used to simplify the calculation of effective
mobility (see Appendix).

Fig. 7During 10,000 days of injection, the total pressure drop
over the reservoir is 1,000 psi. As the dry CO
2
and Buckley-
Leverett regions grow in influence, they contribute more to the
total pressure drop. The Buckley-Leverett pressure drop (BL p)
is the most influential region due to its small mobility (see Fig.
10).

Fig. 8During 10,000 days of injection of CO
2
at a constant
1,000psi pressure differential, the injection rate increases
steadily. The rate determined from our simplified theory
compares well with a full-fledged reservoir simulation for the
same system.

Fig. 9To demonstrate the importance of including the drying
effect, the solution given in Fig. 8 for the Viking sandstone is
compared to the classical fractional flow solution, which
assumes the two phases phase have zero mutual solubility.
Accounting for mutual solubility yields a dry CO
2
region which
leads to an increase in injectivity with time. Without the dry CO
2

region, the injectivity is dominated by the two-phase flow
regionm and injectivity will decrease with time.
14 SPE 113937

Fig. 10The phase mobilities for the three regions occurring in
CO
2
injection into an aquifer are shown. The relative
permeability curves for the Viking sandstone (Bennion and
Bachu, 2005) were used. For typical deep aquifer conditions, the
highest mobility is in the dry CO
2
region. The mobility in the
two-phase Buckley-Leverett region is smaller than in the brine
region. The range of gas saturations in the Buckley-Leverett
region is between Sg,dry and Sg,BL or 0.28 and 0.40. The
mobility evaluated at Sg,avg = 0.34 is 0.62 cp
-1
.

Fig. 11The relative permeability, or more specifically the
effective mobility, has a strong impact on the injection rate.
Seven different relative permeability curves were published by
Bennion and Bachu (2005). Each of those curves was used in
conjunction with the formation properties in Table 1 to obtain
this plot. Different mobility in the Buckley-Leverett region (see
Figs. 10 and 12) are responsible for the differences in injection
rate.

Fig. 12The mobility in the Buckley-Leverett region, evaluated
at the average saturation, Sg,avg, is responsible for the
differences in injection rate (Fig 11). The legend shows the
formations in order of decreasing mobility in the Buckley-
Leverett regions. The order is exactly the same as the
arrangement of decreasing injection rate in Fig. 11.

Fig. 13The effective mobility, as defined in the appendix (Eq.
A.1), increases with time for typical deep aquifer conditions.
This is the underlying mechanism that causes injectivity to
increase with time. The initial effective mobility is near the brine
mobility. The effective mobility does not approach the dry-
region mobility on this time scale.

Fig. 14We varied the parameters for the relative permeability
equations (Eqs. 7 and 8) by the amounts shown. The flow rate
varies by as much as 20% from changes in one parameter.
Small uncertainties in the relative permeability curve can lead to
important uncertainties in the flow rate.
SPE 113937 15

Fig. 15The injectivity is sensitive to temperature, pressure,
and salinity because these parameters affect the the D-terms in
Eqs. 19 and 20. Altering the D-terms by 0.02 changed the flow
rate less than <5%.

Fig. 16The permeability reduction due to salt precipitation in
the drying region will cause a reduction in flow rate. We
consider two limiting cases: Kozeny-Carman, which assumes
salt precipitation will be distributed evenly in the pore space,
and setting the permeability reduction to the measured gas
phase permeability at the irreducible water saturation. We
expect the true permeability reduction to lie somewhere in
between.