SPE-170765-MS

Trapped Gas Saturation Measurements: New Perspectives

N. Bona, L. Garofoli, F. Radaelli, C. Zanaboni, A. Bendotti, and D. Mezzapesa, eni e&p

Copyright 2014, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Amsterdam, The Netherlands, 27–29 October 2014.

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
When a gas reservoir is overlying an aquifer, water will imbibe into the gas-saturated zone with the onset
of gas production, trapping a certain amount of gas. This mechanism is expected to occur in a super giant
field in Mozambique that hosts over 80 TCF of natural gas in place. The observed laboratory variations
in trapped gas saturation represent one of the main uncertainties in the recoverable reserves of the field.
Two laboratory methods for estimating this quantity are compared: one is the classic test in which core
samples are immersed in a liquid and undergo a spontaneous imbibition measurement. The other is a new
method that involves a series of forced imbibitions and combines centrifuge and NMR measurements.
While the standard method produces countercurrent imbibition data, the new method generates co-current
results. These are more representative of the reservoir process and tend to be more optimistic. The data
analysis process involves the estimation of two critical Bond numbers: one is the classic value for the
mobilization of trapped blobs of residual gas, the other sets the boundary between countercurrent and
co-current imbibition regimes. Representative trapped gas saturation results are obtained only for Bond
numbers lying between the two critical values. The new method is validated by the log data acquired in
the two reservoirs under examination below the gas water contact, where immobile gas saturations are
observed. Being caused by the rise of a paleo-contact, the observed saturations are considered to be
representative of the trapped saturations that will form in the reservoir when production starts and taken
as a reference for the lab measurements.

Introduction
The goal of this study is to determine the amount of trapped gas that will be left behind in the aquifer at
the onset of production in one of the largest discoveries of the last decades, a super giant gas field in the
offshore northern Mozambique. The field hosts over 80 TCF of natural gas in place. It is formed by
turbidite channel and lobe deposits of several deepwater basin floor fan accumulations and consists of
coarse and massive sandstones that are extraordinarily thick, clean and homogeneous compared to other
deep water plays such as those in the Gulf of Mexico or West Africa. Porosity ranges between 20% and
30%, and permeability is of the order of hundreds of millidarcies. During the field depletion, the water
beneath the Gas-Water Contact will encroach the gas zone leaving behind its front a quantity of gas
trapped in the porous system. In the context under examination, trapped gas saturation is a critical
parameter for the evaluation of the field production and the final gas recovery. Numerical simulations
2 SPE-170765-MS

show that one percentage point corresponds to approx. 0.15% of the producible gas, which is a huge
amount considering the volume of the field. There is also a considerable impact on the produced water
volume and the duration of the plateau. The former decreases by approximately 5% for a reduction of one
percentage point in trapped gas saturation.
The majority of publications reporting experimental data focus on identifying correlations between
trapped and initial gas saturation, while less emphasis is placed on evaluating the experimental techniques
being used. We show that the choice of the laboratory method has a strong impact. There are systematic
deviations in the results depending on whether laboratory experiments are performed under co-current or
countercurrent conditions. In an imbibition process the proportions of countercurrent and co-current flows
basically depend on the applied Bond number. If imbibition is spontaneous, the Bond number is low and
part of the flow will be countercurrent no matter which boundary condition is used. In order to have a
totally co-current flow, like in the reservoir, the Bond number must be increased by a factor of at least 10.
That calls for a centrifuge experiment.
The first part of the paper reviews those effects that may cause spontaneous imbibition experiments to
be not entirely representative of the reservoir process. Then we introduce a new trapped gas saturation
method that relies on the combined use of the centrifuge and 1D NMR saturation profiling. The method
produces approximately 200 couples of initial saturation versus trapped saturation data per analyzed core
sample for a wide range of Bond numbers. We show that the trapped saturations obtained at too low Bond
numbers overestimate the true values whereas those obtained at too high Bond numbers underestimate
them, and present a methodology to identify the valid, intermediate Bond number range. The measure-
ments conducted on twelve core samples taken from the field under study are discussed. Each sample was
tested by using both the classic spontaneous imbibition method and the new one. We compare the
respective trapped gas saturation results and show that those from the new method are consistent with log
measurements, whereas the spontaneous imbibition experiment gave slightly pessimistic values.
Standard CCI test
The classic laboratory methodology to estimate trapped gas saturations is based on spontaneous imbibition
measurements (Pickell et al, 1966). Some labs call it CCI – Countercurrent imbibition test – and we will
maintain this name throughout the paper. A liquid saturated core is exposed to air and part of the liquid
is allowed to evaporate in order to achieve a predetermined initial air saturation (Sgi) value. Then the core
is immersed in the liquid and its saturation monitored versus time by weighing. The air saturation
remaining in the core at the end of imbibition is corrected for diffusion and the resulting value, which is
conventionally named SgrM, is assumed to represent the trapped gas saturation that is left in the reservoir
behind the aquifer front. Normally four evaporation/imbibition cycles are made in which Sgi is set at
100%, 75%, 50% and 25%. The output of the test is represented by four couples of Sgi-SgrM values.
Typically toluene is used as the liquid phase. Organic solvents are preferred over water because there are
no issues with salinity and density variations during the evaporation stages.
This CCI experiment has some drawbacks. First of all, gas displacement is predominantly counter-
current (see for instance the experiments conducted by Baldwin and Spinler (1999). Because the reservoir
process is strictly co-current, the resulting trapped gas saturations may not be entirely representative of the
reservoir values. Second, there is a risk of Sgi being unevenly distributed in the core at the end of
evaporation, with more liquid in the center of the core than near the core faces. Since Sgr depends on Sgi,
this may cause an additional error. Finally, during imbibition the core is invaded by less than one pore
volume of water, whereas in the reservoir gas zone an identical volume of rock may be flooded by
hundreds of pore volumes of aquifer water. That leads to a further error. Each of these errors tends to
produce an overestimated value of trapped saturation.
It is extremely difficult to mimic the reservoir process in the lab with cheap, conventional experiments.
For example, the invasion of many pore volumes of water would call for a water injection test. However,
SPE-170765-MS 3

Figure 1—Air bubbles emerging from the bottom face of one of the core samples investigated in this study during water imbibition. The core was
sealed laterally, with the bottom end in contact with water and the other end open to air. Water imbibition was expected to occur as a co-current
capillary rise, but this was not the case (from Bona et al (2014)).

if a large volume of water is injected into a core at ambient conditions, part of the gas will pass in solution
and this may dramatically affect the results. Guaranteeing a co-current gas displacement is another
problem. With the CCI test most of the air in the core is displaced counter-currently. Unfortunately,
co-current flow conditions cannot be obtained with a spontaneous imbibition experiment even if different
boundary conditions are used. For example, sealing the curved surface of the core and suspending it over
the liquid so that the bottom end stays immersed in the liquid while the top end is exposed to air does not
prevent the formation of air bubbles at the bottom end, especially if the permeability is high. This is what
we observed in our experiments and represents an indisputable indication of the presence of countercur-
rent mechanisms (Fig.1).
Surface effects in spontaneous imbibition
Many of the processes that occur in a totally or partially air saturated core when it is immersed in a liquid
do not occur in the reservoir. This is the main reason why the CCI experiment does not always provide
representative trapped gas saturation results. If we submerge a core under a liquid and do this operation
quickly enough, the liquid will “incapsulate” the air inside the core. Because the rock has a strong wetting
preference to the liquid, the liquid will penetrate into the core from pores of any size, forming a superficial
liquid-saturated layer throughout. The velocity with which the liquid moves is a function of the pore sizes:
although the dynamics of the process are quite complex and still not well understood, there is general
consensus on the fact that the liquid advances faster in the smaller pores (Mason and Morrow, 2013). The
liquid velocity variations from one point to another cause a certain amount of air to be trapped/bypassed
by the liquid. During the early instants of imbibition, the entry of a certain volume of liquid into the core
does not correspond to the expulsion of an identical volume of air. As a result, the air in the core is
compressed. When the air pressure becomes sufficiently large, a number of air channels that connect it to
the outside of the core will form. During this stage the air movement is totally counter-current (the air
moves toward the outside, whereas the liquid moves in the opposite direction). The air flow concentrates
in regions where pores are large and the capillary forces opposing the air advance are weaker. This is
supported by the observation that air often emerges from the core face as a stream of small bubbles at well
defined locations. As noted by Rangel-German et al (2004), the air channels are not always completely
connected to the bulk of the air phase; they connect and disconnect as the liquid advances and displaces
the air, thus generating a dynamic and ever-changing topology.
4 SPE-170765-MS

Once a connected air channel reaches the core surface, an air bubble starts to form outside the core. For
a certain amount of time the bubble remains attached to the core surface. Since in other points the liquid
continues to penetrate into the core compressing the air more and more, the bubble grows into the liquid
surrounding the core. Experiments conducted by Unsal et al (2009) on an arrangement of cross-connected
tubes show that the air pressure drops during bubble growth. This is in agreement with the Young-Laplace
equation, which states that the difference between the pressure of the air in the bubble and the constant
pressure of the liquid surrounding the bubble is inversely proportional to the radius of curvature of the
bubble. The bubble evolution depends on whether buoyancy forces favor the departure of the bubble or
not. If the bubble emerges and grows from the top face of the core, the bubble evolution will consist of
three principal stages (Di Bari and Robinson, 2013): an early stage during which the shape of the emerging
bubble is essentially spherical; an elongation/expansion stage where the bubble becomes progressively
bigger and more elongated; and a final detachment stage in which a neck forms near the bubble base and
the buoyancy force eventually causes the departure of the bubble. For a bubble emerging from the bottom
face of the core, however, things are different. In this case gravity opposes the bubble growth: the growth
time is longer and when the buoyancy force (proportional to the bubble volume) becomes strong enough
to counteract surface tension stresses which tend to keep the bubble shape spherical, the bubble starts
flattening (Fig.1). The bubble may never depart from the core surface and an equilibrium between the
various forces acting on the bubble may eventually be attained. For the bubbles emerging from the top
face of the core, the time to bubble departure is related to the time it takes for buoyancy to become
significant and start lifting the bubble. Once the bubble is being lifted, the contact angle starts to increase
with time at an increasingly faster rate. Departure occurs when the contact angle has increased to such an
extent that the bubble attains a metastable equilibrium state. Upon further inflow of air into the bubble
there is no longer a shape that can establish a stable state and the bubble detaches from the neck to form
a large detached free bubble and a small bubble attached to the pore rim, which are both stable. Di Bari
and Robinson (2013) measured bubble departure volumes for submerged orifices of radii from 290 to 800
microns. Their experimental observations corroborate the theoretical calculation of the departure volume
that can be derived from the balance of buoyancy and capillary forces (Tate, 1864). The conclusion is that
for sufficiently small orifice radii, the maximum metastable bubble volume which can exist in a given
liquid and gravitation field is given by Vmax ⫽ 2␲r␴/(⌬␳g) Applying this formula to a pore mouth of
radius r⫽10 microns under terrestrial gravity gives a detached bubble with a radius of 0.6 mm. That seems
consistent with visual observations of produced air bubbles from air saturated cores immersed in water.
The detachment volume formula provides us a means to estimate the min and MAX values of the capillary
back pressure, Pcb, that is generated during the growth of the bubble. Pcb is the difference between the air
pressure inside the bubble and the pressure of the liquid surrounding the bubble near the interface. When
the bubble starts to emerge from the core face Pcb is maximum and equals 2␴/r, where r is the radius of
the pore from which the bubble is coming out. Pcb decreases as the bubble grows and reaches its minimum
value when the bubble detaches. At bubble departure, the bubble radius, Rdep, can be estimated from the
relation 4/3·␲Rdep3 ⫽ Vmax and the corresponding capillary pressure is 2␴/Rdep. This gives a capillary
back pressure at departure which equals ␴2/3(⌬␳g/r)1/3. For a pore of radius r⫽10 microns under terrestrial
gravity, in the water-air system, Pcb would consequently drop from 0.14 bar (bubble appearance) to 1.7
mbar (bubble departure). These values are consistent with published experimental and numerical results
(Unsal et al, 2013, Di Bari et al 2013). Since the pressure in the bubble decreases as the bubble grows and
the pressure in the air channel that connects the bubble to the bulk of the air phase decreases toward the
core face, at some point inside the core the air pressure may become so low that it is easier for the liquid
to cut the air channel and disconnect the bubble rather than continuing to push the whole air volume and
feed up the bubble. If the liquid interrupts the air bubble connection, the bubble stops growing and the
channel will temporarily cease to contribute to countercurrent production. With a larger gravity applied
SPE-170765-MS 5

to the system, the pressure gradient in the channel is

larger, and the chances of cutting the bubble off
increase.
The processes described in this section can be
classified as surface effects. They result from the
interplay of capillary, viscous and gravity forces at
the core faces. To a first approximation, the ratio of
counter- to co-currently displaced air volumes dur-
ing imbibition is proportional to the contribution of
surface effects to overall air production. This, in
turn, is inversely proportional to quantity gL, where
L is the core thickness and g is the gravity acceler-
ation. The larger the gL product the more represen-
tative the imbibition experiment. Because there are
obvious limitations in increasing L to a large extent,
the only way to minimize surface effects in a lab
experiment is to use high g values. With a centrifuge
g can be increased by several orders of magnitude
with respect to the terrestrial gravity value.
Description of the CNMR method
The CNMR method is based on a standard primary
drainage process followed by a series of forced
imbibitions. We name it CNMR because it relies
upon the combined use of a Centrifuge and 1D
NMR saturation monitoring. A fully water saturated
Figure 2—Gas saturation profiles measured on one of the analyzed
sample is centrifuged under air. Then the sample is samples
centrifuged under water at increasing speeds. At the
end of each centrifugation, the z-dependent air saturation profile formed inside the sample is measured.
We call Sgi the air saturation achieved after centrifuging in air and Sgr the saturations reached after each
of the subsequent centrifugations in water. The Sgi and Sgr profiles are measured by means of an NMR
spectrometer (resonant frequency: 2 MHz; sequence utilized: Spin-Echo SPI; echo time ⫽1.0 ms; gradient
duration ⫽0.15 ms; spatial resolution ⬇ 1 mm and signal to noise ratio ⬎ 100). Fig.2 illustrates an
example of this measurement.
Basically, the Sgr values achieved at each position along the core length depend on two quantities: the
Sgi value at that position and the applied Bond number (Nb). The Bond number is a measure of the ratio
between gravity and capillary forces and is calculated at each position from the formula Nb ⫽⌬␳gk/␳,
where g ⫽ ␻2R,⫽ 2␲f, f ⫽ number of rotations per second, R ⫽ distance from the axis of rotation in
metres, ⌬␳□ ⫽ water-air density difference in kg/m3 (due to the compressibility of air, ⌬␳□ ⫽ is a
fuction of both position and rotational speed),␴ ⫽ water-air interfacial tension in N/m, and k ⫽ sample
permeability in m2 (it is assumed that permeability is constant throughout the sample). Typically we
perform five rotational steps. With 40 positions, a compilation of 200 Sgi-Sgr-Nb triplets is produced.
Fig.3 is an example of the output of the CNMR analysis. Of course, only part of the data is representative
of the trapped gas saturations that remain in the reservoir as the aquifer encroaches the gas zone. Too low
or too high Bond numbers result in unrepresentative Sgr values. The valid Bond number range is defined
by using both theoretical arguments and experimental results: we will see how in the following section.
6 SPE-170765-MS

Figure 3—Output of the CNMR test. Each data point represent a Sgi-Sgr-Nb triple.

Selection of valid Bond numbers:

theory and experiments
A trapped gas experiment provides representative
results if the flow is entirely co-current and residual
gas blobs are not mobilized. Both are controlled by
the Bond number. The relationship between the
measured Sgr and the applied Bond number is sche-
matically illustrated in Fig.4. At low Bond numbers,
part of the gas is displaced countercurrently because
surface effects are not negligible. As the Bond num-
ber increases the proportion of gas displaced coun-
tercurrently decreases and co-current mechanisms
become progressively dominant. With a counter-
current process there is a higher chance of isolating Figure 4 —Expected Sgr-Nb relationship

the non-wetting phase, so the gas saturation remain-

ing in the core decreases as the Bond number is
increased. Above a certain critical Bond number value, the flow becomes entirely co-current. In this Bond
number region, the remaining gas saturation should be relatively independent of the applied Bond number.
The region corresponds to the optimal experimental conditions for a trapped gas measurement. If the Bond
number is increased further and the critical value for the mobilization of residual gas is overcome, the gas
saturation remaining in the core will start decreasing again and will do it at quite a fast rate.
Based on the theory of mobilization (see the Appendix for details) the valid Bond number region is
defined by the following conditions:
Condition to avoid residual gas mobilization:
(1)

Condition for co-current flow:

SPE-170765-MS 7

Table 1—Base petrophysical parameters of the reservoirs and core samples under examination

(2)

where Nb ⫽ ⌬␳gk/␴ is the Bond number, Nc ⫽

␮liq␮liq/␴ is the capillary number, ⌬␳ is the fluid
density difference, g is the terrestrial gravity accel-
eration or the centrifugal acceleration, k is the ab-
solute permeability, ␴ is the interfacial tension, uliq
is the Darcy velocity of the liquid, liq is the liquid
viscosity, krliq is the liquid relative permeability at
residual gas saturation, r is the mean pore threshold
radius, L is the thickness of the rock and l is the size
of the largest residual blobs - a quantity of the order Figure 5—Example of Sgr - Nb curve (solid symbols: sample 14, posi-
tion: 2.25 cm; open symbols: sample 58, position: 0.67 cm)
of 1 mm. Tab. 1 gives the values of these parameters
for the two reservoirs examined in this paper
(named A and B). Some parameters were measured; others, in absence of direct measurements, were
simply assumed. The latter include l and krliq. All the values are reported for the reservoir process and the
two laboratory experiments we carried out (CCI and CNMR). As can be seen, in the reservoir process the
Bond number lies within the validity range. With the CCI experiment the Bond number is always lower
than the minimum value and so the resulting trapped gas saturations can overestimate the true trapped gas
saturations as shown in Fig. 4. With the CNMR technique there exists a range of Bond numbers potentially
able to provide representative results.
Fig. 5 shows two of the Sgr- Nb curves measured on the core samples under study as an example. The
intermediate Nb region where, based on Fig. 4, Sgr is assumed to level off is not clearly visible in the
curves. In one curve Sgr decreases gradually as the applied Bond number increases and when the Bond
number exceeds a critical value between 1E-5 and 1E-4 (corresponding to the Nb-MAX value) Sgr
decreases at a much faster rate. In the other curve, Sgr appears to increase slightly for Bond numbers
between 1E-6 and 1E-5, before decreasing at approximately the same rate as in the first curve. The amount
of gas produced at a given position after a centrifugation step is equal to the amount of gas expelled minus
the amount arriving from upstream. The balance can be negative at some locations even though the
rotational speed is increased, which explains why Sgr can increase.
Fig. 6 shows the salient features of the whole Sgr-Nb curve population. The y-coordinate of the plotted
data points corresponds to the slope of the segment that connects, in a given sample and at a given
position, two consecutive Sgr-Log(Nb) points. This is the semilogarithmic derivative of the Sgr-Nb curve.
Our dataset is comprised of 4 segments for each position along the core length, approx. 40 positions per
core, 6 cores per reservoir and 2 reservoirs, so the number of data points in Fig.6 is of the order of 2000.
As can be seen, at low Nb values the slopes of the Sgr-Nb curves tend to be slightly negative, which means
8 SPE-170765-MS

Figure 6 —Sgr-Nb semilogarithmic derivatives

that an increase in the Bond number resulted in a

slight Sgr decrease. This region of Bond numbers
corresponds to the region in Fig. 4 where gas dis-
placement is countercurrent. At very high Nb values,
the slopes are strongly negative and this means that
a small increase in Nb translated into a large de-
crease in Sgr. This high Nb region corresponds to
the region where residual gas blobs are mobilized.
At intermediate Nb values, the semilogarithmic de-
rivatives of the Sgr- Nb curves are partly positive
and partly negative. Their average value approaches
zero, indicating that the curves tend to level off.
This Nb region extends from Bond numbers around Figure 7—Base properties of the analyzed cores

2E-6 to Bond numbers in the 1E-5 - 1E-4 range and

corresponds to the region in Fig. 4 in which gas
displacement is entirely co-current and the residuals are not mobilized. The valid Sgi-Sgr data belong to
this Nb region. Compared to the theoretical values calculated from conditions 1) and 2) in Tab. 1, the
experimental Nb-min and Nb-MAX values derived from Fig. 6 show pretty good consistency. The literature
is rich in studies on the critical Bond number for residual phase mobilization and the majority of them give
values from 1E-5 to 1E-4, consistently with our observations (King et al, 1986; Christainsen, 2008;
Morrow and Songkran,1981). On the other hand, little data have been published on the critical value for
the counter/co-current flow transition: Schechter et al (1994) studied low IFT imbibition using conjugated
solvent systems along tie-lines of a three-component system and scaled the crossover from capillary
forced imbibition to gravity dominated imbibition according to a condition for the Bond number that is
equivalent to our condition 2). Hosein et al (2010) performed numerical investigations of water-oil
imbibition on a 61 cm long core for a water-oil system at various IFT values and confirmed the
counter/co-current transition rule postulated by Schechter et al.
SPE-170765-MS 9

Figure 8 —MICP curves and relationship between principal pore throat radius and sample permeability

Figure 9 —Air saturation evolution with time for one of the investigated
sample
Figure 10 —Durations of the capillary-dominated regimes as a function
of initial air saturation
Materials

Two reservoirs were investigated. They belong

to a large offshore field in Mozambique. Reservoir
A is an Eocene sandstone, reservoir B is an Oligo-
cene sandstone. For both reservoirs the main facies
is described as massive, thick-bedded, and very
coarse to coarse grained. The clays⫹micas content
is very low. We analyzed six core samples from
each reservoir. Each core sample was submitted to
the same tests (CCI and CNMR). Fig. 7 illustrates
the poro-perm characteristics of the samples. Those
from reservoir A have permeabilities from 800 to
1400 millidarcies and porosities around 20%; those Figure 11—Trapped gas saturations at the end of the capillary-
from reservoir B have permeabilities from 200 to dominated regime versus initial air saturations
400 millidarcies and porosities around 30%. Fig. 8a
shows the pore throat distribution curves measured by mercury injection on three of the cores (those
marked with large symbols in Fig. 7). The distributions are essentially monomodal. There is an excellent
10 SPE-170765-MS

Figure 12—Sgi-Sgr- Nb triplets obtained from the CNMR method for the reservoir A cores and Sgi-Sgr couples measured on the same cores with
the standard CCI test

correlation between the pore throat radius value (r) at which the distributions peak and permeability (Fig.
8b).
Discussion of CCI test results
The CCI experiments were carried out using the classic method, that is the cores were immersed in toluene
with all faces open to imbibition after a preliminary controlled evaporation stage. Fig. 9 graphically
illustrates the output of one of the measurements. As is widely documented in the literature, two flow
regimes are visible.
The standard interpretation of this behavior is that air displacement is initially an essentially immiscible
process dominated by capillary forces and then becomes an essentially miscible process dominated by
diffusion mechanisms. Plotting the instantaneous air saturations against the square root of time, the two
regimes appear as two straight lines. The durations of the capillary regime are quite short, of the order of
tens-hundreds of seconds (Fig. 10).
We observed a dependence on initial saturation: the higher the initial amount of liquid in the sample,
the faster the imbibition process. This fact can be explained by considering that with a higher initial liquid
saturation the chance of having snap-off mechanisms ahead the liquid front is greater. These add to the
piston-like air displacement processes that take place behind the front, thus increasing the overall amount
of displaced air. The A cores exhibited slopes of the diffusion-dominated ranging from
. For reservoir B the observed range was . The average
values are and , respectively, which means that after 10, 100 and 1000
seconds of imbibition the respective average losses of air saturation due to diffusion were of 0.3, 0.8 and
2.6 saturation units in the A cores and 0.4, 1.4 and 4.3 saturation units in the B cores. Fig. 11 shows the
measured trapped gas saturations, SgrM, as a function of the initial saturations, Sgi. As can be seen, SgrM
increases as Sgi increases, tending to level out at high Sgi values. This behavior is largely reported in the
literature (see for example Iglauer et al (2011) and the references therein). The plateau values of SgrM are
of the order of 27% for the A cores and range from 30 to 40% for the B cores.
Discussion of CNMR test results
Fig. 12 and Fig.13 illustrate all the Sgi-Sgr couples obtained from the CNMR test. The Sgr values are
corrected for gas diffusion. Because the cores stayed immersed in water for approx. 20 minutes during
each centrifugation, the volume of air passed in solution during this amount of time is not negligible.
Based on the results discussed in the previous section, we estimated it by multiplying the square root of
SPE-170765-MS 11

Figure 13—Same as Fig.12 but for the reservoir B cores

Figure 14 —Valid Sgi-Sgr- Nb triplets obtained from the CNMR method for the reservoir A cores and Sgi-Sgr couples measured on the same cores
with the standard CCI test

the centrifugation time (in seconds) by 8⫻10-2 for the A cores and by -1⫻10-1 for B cores. The resulting
quantity (expressed in %PV) was added to the air saturation remained in the core at the end of each
centrifugation step to obtain the diffusion-corrected Sgr values. In Fig. 12–13 the Sgi-Sgr data are grouped
by Bond number. As expected, the higher Nb the lower Sgr for a fixed Sgi value (the colors of the symbols
get lighter as Nb increases). Plotted together with the CNMR data, there are also the Sgi-SgrM values
obtained on the same cores from the CCI experiments. CCI and CNMR gave the same results for the same
initial saturations and the same Bond numbers. This consistency of the data is an indication of the good
quality of both measurements.
Fig. 14 and 15 highlight the representative Sgi-Sgr data, extracted by filtering the entire data set
according to appropriate Nbmin and Nb-MAX values, as discussed in the previous section. The resulting
Sgi-Sgr curves show maximum trapped gas saturations between 20% and 25% for reservoir A and from
20% to 30% for reservoir B. All these values were obtained from analyses carried out at atmospheric
pressure.
Every time the centrifuge is stopped, the pressure in the core falls causing the air to expand. As a result,
the reported CNMR air saturations may be higher than those existing while centrifuging. Air expansion
results in a saturation increase if part of the water is expelled from the core. This may be the case for
trapped air clusters, whereas the air that is connected to the outside of the core should come out displacing
12 SPE-170765-MS

Figure 15—Same as Fig.14 but for the reservoir B cores

Figure 16 —Effect of air compressibility on the valid Sgi-Sgr data for the reservoir A cores

little amounts of water and, consequently, with little impacts on air saturation. Depending on the rotational
speed of the centrifuge and the distance from the axis of rotation, we calculate that in the CNMR
experiments during centrifugation the absolute pressure of air ranged from approximately 1.1 bar to
approximately 6.0 bar and when the centrifuge was stopped the air volume increased by a factor ranging
from 1.003 to 5.60. The 1.003 factor describes the expansion of air situated at the core outlet (close to the
axis of rotation) after centrifuging at the lowest rotational speed. The 5.60 factor applies to air at the core
inlet after centrifuging at the highest speed. Correcting for air compressibility the ambient Sgr values
illustrated in Fig. 14 and 15 gives the corresponding “while centrifuging” values that are shown in Fig.
16 and 17. These were obtained by assuming that all the air expansion resulted in the expulsion of water,
so they should represent the minimum possible values of Sgr. As can be seen, in reservoir A the scatter
in the plateau region increases and the average Sgr decreases by approx. 5 saturation units. For reservoir
B, the scatter remains more or less the same, and a similar reduction in the Sgr values is obtained.
Unfortunately, it is really complex to determine how much water is expelled during gas expansion when
the centrifuge is stopped. We can only say that the “true” Sgr values lie somwhere between the two sets
of data. However, we think that because the trapped saturations obtained from the spontaneous imbibition
experiments are not dramatically different from the CNMR saturations and the latter appear to be in line
with the log data (see next section), air compressibility effects did not play a role of primary importance.
SPE-170765-MS 13

Figure 17—Same as Fig.16 but for the reservoir B cores

Figure 18 —Capillary trapping capacity -the product of porosity and trapped saturation- as a function of initial nonwetting phase saturation. Adapted
from Iglauer et al (2011)

In partial support of this assumption, the residual air saturation values obtained from the production
profiles registered during water-air injection tests conducted on sister samples are not very different from
the residual saturations estimated by weighing at ambient pressure after dismounting the samples from the
core holders. In theory, air compressibility has an impact also on the definition of the valid Bond number
range, given that Nb depends on the water-air density difference and the critical value for the mobilization
of trapped blobs is a function of the length of the blobs. Both quantities increase when the centrifuge is
stopped, but to a small extent, and the CNMR results are only negligibly affected.
Fig. 18 compares the CNMR and CCI results with a compilation of data published by Iglauer et al.
(2011). The x-coordinate is the initial non-wetting phase saturation. The y-coordinate is the capillary-
trapping capacity, Ctrap, which is defined as the product of trapped saturation and porosity. Although there
is a considerable scatter in the data due to the heterogeneity of the materials and the experimental
conditions used, the CNMR results appear to be in line with the literature data.
14 SPE-170765-MS

Figure 19 —Log data from below the water-gas contact

Comparison to log data

The log interpretations are presented in Fig. 19. In both reservoirs, below the gas-water contact (defined
as the lowest point at which gas is mobile) there exists a thick interval with non-zero gas saturations. The
gas here is considered to be the result of the rise of a paleo-water contact. Because the petrophysical
characteristics of the rock do not vary above and below the current gas-water-contacts, the immobile
saturations below the GWC should be representative of the trapped saturations that will form above the
GWC at the onset of water encroachment. In reservoir A, the top of the zone containing immobile gas is
at XX84 m. From XX84 m to XX94 m, Sg exhibits a steep trend and passes from 50% to 20%. This is
the effect of the shale barrier at XX62-84 m, which acted as an accumulation point for the gas displaced
by the paleo-contact. Below XX94 m, for a considerable thickness, the Sg profile appears stabilized even
though it is quite noisy. In this zone Sg ranges from 5% to 30% with an average value of 20%. We
SPE-170765-MS 15

consider these saturations as the best indication of trapped gas saturation. The log values are in line with
the CNMR results, whereas the CCI values are slightly overestimated.
Passing to reservoir B, here the current gas-water contact is located at X770 m (Fig. 19-B). The top of
the immobile gas zone is at X796 m and below there are approx. 70 meters of relatively clean rock. Sg
ranges between 5% and 35% with an average value around 20%, like in reservoir A. Again, while CNMR
gave consistent trapped gas saturations, those from CCI appear to be overestimated. Since the clay content
of the rock is very low, the log-derived saturations were obtained by using the Archie model. The m&n
parameters were derived from gas-water primary drainage experiments. In theory, imbibition data should
have been used, but these were not available. Imbibition saturation exponents tend to be lower than the
drainage exponents, so the reported gas saturations might be underestimated to some extent. However, due
to the strong water-wet conditions of the system, the impact is expected to stay within 5 saturation units.
Adding this amount to the reported log-derived gas saturation makes the log data even closer to the CNMR
results.

Conclusions
1. A lab method for measuring trapped gas saturation from a co-current water imbibition process is
presented. The method combines centrifuge measurements with in situ saturation monitoring by
NMR and produces a comprehensive compilation of Sgi-Sgr-Nb data in a few days with a limited
number of core samples.
2. Representative trapped gas saturations are obtained only for Bond numbers in the 10-6 - 10-5 range.
Bond numbers lower than 10-6 cause part of the flow to be countercurrent. For Bond numbers
higher than 10-5 part of the residual gas will be produced. The two critical Bond numbers are in
line with the predictions of mobilization theory and published data.
3. Standard trapped gas measurements, performed by immersing the core samples in a liquid under
terrestrial gravity, have Bond numbers of the order of 10-7. In these types of experiments a large
fraction of the flow is countercurrent. We have observed that the flow remains countercurrent even
if the core is sealed laterally and suspended over the liquid. The only way to produce co-current
flow conditions is to increase g.
4. With the new method g is increased from 10 to 104. The trapped gas saturations obtained for Bond
numbers within the valid range are slightly lower compared to those obtained from the standard
method.
5. Formation evaluation results calibrated to core and fluids information showed the clear presence
of residual gas below the Gas-Water contact. The log-derived saturations are in good agreement
with the trapped saturations obtained with the new method.

References
Baldwin B.A., Spinler E.A., “In-situ saturation development during spontaneous imbibition”, SCA
1992 proceedings of the SCA Int. Symposium, Golden, CO, Aug. 1992.
Bona N., Garofoli L., Radaelli F., Zanaboni C., Anelli M., Bendotti A. and Mezzapesa D., “An
efficient method for measuring trapped gas saturation under co-current conditions”, SCA 2014-24,
proceedings of the SCA Int. Symposium, Avignon, France, Sept. 2014.
Chatzis I. and Morrow N.R.: “Correlation of capillary number relationships for sandstone”, SPE
Journal, Oct. 1984, pag. 555–562.
Christainsen R.L.: Multiphase Flow thorugh porous Media, KNQ Engineering, Salt Lake City, Utah,
June 2008.
Di Bari S., Lakehal D., Robinson A.J.: “A numerical study of quasi-static gas injected bubble growth:
Some aspects of gravity”, International Journal of Heat and Mass Transfer, (2013) 64, 468 –482.
16 SPE-170765-MS

Di Bari S., Robinson A.J.: “Experimental study of gas injected bubble growth from submerged
orifices”, Experimental Thermal and Fluid Science, (2013) 44, 124 –137.
Dullien, F.A.L: “Porous Media: Fluid transport and Pore Structure”, Academic Press, New York City
(1979).
Hosein Kalaei M., Green D.W., Willhite, G.P., “Numerical modeling of the water imbibition process
in water-wet laboratory cores”, SPE 132645, proceedings of the Western North America Regional
Meeting, Anaheim, CA, May 2010.
Iglauer S., Wulling W, Pentland C.H., Al-Mansoori S.K. and Blunt M.J. “Capillary-trapping capacity
of sandstones and sandpacks”, SPE Journal, Dec. 2011, pag. 778 –83.
Khishvand M., Akbarabadi M. Piri M., “Trapped non-wetting phase clusters: an experimental
investigation of dynamic effects at the pore scale using a micro-CT scanner”, SCA 2013-24, proceedings
of the SCA Int. Symposium, Napa Valley, CA, Sept. 2013.
King M.J., Falzone A.J., Cook W.R., Jennings J.W. and Mills WH,: “Simultaneous Determination of
residual Saturation and Capillary Pressure Curves Utilizing the Ultracentrifuge”, paper SCA 15595
presented at the 1986 SPE ATCE, New Orleans, Oct 5-8.
Mason G, Morrow N.R., “Developments in spontaneous imbibition and possibilities for future work”,
JPSE, 110(2013), 268 –293
Morrow N.R. and Songkran B: “Effect of Viscous and Buoyancy Forces on Nonwetting Phase
Trapping in porous Media”, in Surface Phenomena in Enhanced Oil Recovery, edited by D. O. Shah,
1981.
Pickell J.J., Swanson B.F., Hickman W.B., “Application of air-mercury and oil-air capillary pressure
data in the study of pore structure and fluid distribution”, old SPE Journal, (March, 1966), 51–65.
Raeesi B., Morrow N.R., “Pore network modeling of experimental capillary pressure hysteresis
relationships”, SCA 2013-15, proceedings of the SCA Int. Symposium, Napa Valley, CA, Sept. 2013.
Rangel German E.R., Kovscek A.R.: “Microvisual analysis of matrix-fracture interaction”, SPE
92133, proceedings of the 2004 SPE International Petroleum Conference, Puebla, Mexico, Nov. 2004.
Schechter D.S., Zhou D., Orr F.M., “Low IFT drainage and imbibition”, JPSE, Vol. 11(1994),
283–300.
Sohrabi M., Riazi M., jamiolahmady M., Ireland S. and Brown C., “Mechanisms of oil recovery by
carbonated water injection”, SCA 2009-26, proceedings of the SCA Int. Symposium, Noordwijk, the
Netherlands, Sept. 2009.
Taber J.J.: “Dynamic and static forces required to remove a discontinuous oil phase from porous media
containing both oil and water”, SPE Journal, March 1969, pag. 3–12.
Tate T.: “On the magnitude of a drop of liquid formed under different circumstances”, Philosophical
Magazine (1864) 27, 176 –185.
Unsal E., Mason G., Morrow N.R and Ruth, D.: “Bubble snap-off and capillary back pressure during
counter-current spontaneous imbibition into model pores”, Langmuir, (2009) 25, 3387–3395.
SPE-170765-MS 17

Appendix

Following the theory of mobilization (see for example Taber,

1969, and Chatzis and Morrow, 1984), consider a cluster of
gas surrounded by a liquid (Fig. A-1).
Let r be the average pore threshold radius and l the length
of the cluster along the z-direction (the cluster spans over
several pore distances). The pressure drop across the cluster
can be written as pliq1 – pliq2– ␳ggl, where pliq1 and pliq2 are the
upstream and downstream liquid pressures, ␳g is the gas den-
sity and g is the gravity or centrifugal acceleration. The gas
density is dependent on the experimental setup and varies
along the core length in a centrifuge experiment. Because the
capillary pressure resisting cluster mobilization is of order of
2␴/r (␴ is the liquid-gas interfacial tension), the cluster is
immobile if:
A1)

Assuming that Darcy’s law holds at the cluster length

scale, the Darcy velocity uliq of the liquid surrounding the
cluster can be expressed as: Figure A-1—Trapped gas cluster
A2)

where k is the absolute permeability, krliq is the liquid relative permeability, ␮liq is the liquid viscosity and ␳liq is the liquid
density. Solving Eq. A2) for pliq1 – pliq2 and substituting into A1) gives:
A3)

The left-hand member of A3) is the sum of the capillary number Nc (⫽ ␮liq␮liq/␴) divided by a characteristic liquid relative
permeability value plus the Bond number Nb (⫽⌬␳gk/␴). In terms of Bond and capillary numbers, A3) can be rewritten as:
A4)

Simulations and experimental results (see Raeesi, 2013, Khishvand, 2013, and the references therein) show that at the end
of imbibition, non-wetting phase (NWP) clusters spanning over 1 or 2 pores account for approx. 20% of the total trapped NWP
volume, whereas approx. 60% of this trapped volume is due to clusters of 6 to 10 pores. Based on these data, we consider it
reasonable to assume that the largest trapped clusters may have sizes of the order of 10-3 m. Displacement experiments on 2D
micromodels (Sohrabi et al, 2009) give trapped phase sizes in line with these results. Also, it appears reasonable to assume krliq
⫽ 0.1 in a water-gas system. It is worth noting that quantity 2k/(rl) is of order of 10-5 for the majority of rocks and that the
capillary number is at least one order of magnitude smaller in a capillary-driven imbibition process. That explains why the
majority of publications report critical Bond numbers for residual non-wetting phase mobilization of the order of 10-5.
Critical Bond number for counter/co-current flow transition
The transition between the capillary-dominated (countercurrent) and the gravity-dominated (vertical co-current) flow
regimes may be explored by using the same concepts utilized in the previous section. We now consider a gas cluster having
the same vertical extension as the entire core sample (L). The difference between the upstream and downstream liquid pressures
is ␳liqgL. This situation represents the starting point of a standard trapped gas measurements, in which a core sample containing
a mobile, continuous gas phase is immersed in a liquid. According to A1), the gas will be mobilized upward if
A5)

Rewriting A5) in terms of Bond number gives:

A6)

Quantity 2k/(rL) represents the Bond number crossover between countercurrent and co-current regimes. Unlike residual
phase mobilization, the counter/co-current transition is not widely documented in the literature. Schechter et al (1994) and
Hosein et al (2010) use the dimensionless group and report that when Bo-1 ⬎ 5 imbibition is countercurrent;
18 SPE-170765-MS

when Bo-1 ⬍ 1 imbibition is co-current; for intermediate values a combination of the two imbibition mechanisms is expected.
It is interesting to note that the condition Bo-1 ⬍ 1 coincides with A6) if the average pore threshold radius r is substituted by
quantity 5 . The ⫽ 5 relationship was derived mathematically for a cubic network of capillary tubes
(Dullien, 1979). Although it is strictly valid only for this ideal porous medium, many authors have used it to model also real
rocks. For the three cores submitted to MICP analysis (Fig. 8), there is a good agreement with experiments: the r values at
which the measured pore threshold size distribution curves peak are respectively of 20, 10 and 5 microns; the corresponding
r values calculated from the ⫽ 5 realtionship are of 10, 5 and 4 microns.