SPE 102239
Modeling Low-Salinity Waterflooding
G.R. Jerauld, SPE, C.Y. Lin, K.J. Webb, SPE, and J.C. Seccombe, SPE, BP
Copyright 2006, Society of Petroleum Engineers
This paper was prepared for presentation at the 2006 SPE Annual Technical Conference and
Exhibition held in San Antonio, Texas, U.S.A., 2427 September 2006.
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, as
presented, have not been reviewed by the Society of 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 Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes 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 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.
Abstract
Low salinity waterflooding is an emerging EOR technique in
which the salinity of the injected water is controlled to
improve oil recovery over conventional higher salinity
waterflooding. Corefloods and single well chemical tracer
tests have shown that low salinity waterflooding can improve
basic waterflood performance by 5 to 38%. This paper
describes a model of low salinity flooding that can be used to
evaluate projects, shows the implications of that model,
demonstrates its use to represent corefloods and single well
tests as well as field scale simulations, and gives insight into
the reservoir engineering of low salinity floods.
The model represents low salinity flooding using salinity
dependent oil/water relative permeability functions resulting
from wettability change. This is similar to other EOR
modelling and conventional fractional flow theory can be
adapted to describe the process in one dimension for
secondary and tertiary low salinity waterflooding. This simple
analysis shows that while some degree of connate water
banking occurs it need not hinder the process. Because mixing
of injected water with in situ water delays the attainment of
low salinity, potentially preventing attainment of low salinity
all together if very small slugs of low salinity water are used,
care must be taken in representing mixing appropriately in
interpreting data and in constructing models. The use of
numerical dispersion to represent physical dispersion in 1D,
radial and pattern simulations of this process is demonstrated,
i.e. coarse simulations are shown to give the same result as
fine grid simulations with appropriately large physical
dispersion. In many applications, the fine grid simulation
necessary to represent appropriate levels of dispersion is not
practical and pseudoization is necessary. We demonstrate that
this can be done by changing the salinity dependence and
shapes of relative permeability curves.
Introduction
Waterflooding is widely used to improve recovery from oil
reservoirs but, except to avoid formation damage, is largely
designed without regard to the composition of the brine
injected. Yildiz and Morrow1 showed that showed that
changes in injection brine composition can improve recovery,
thereby, introducing the idea that the composition of the brine
could be varied to optimize waterflood recovery. Tang and
Morrow2-5 built on this idea by demonstrating the benefit
lowering brine salinity has on oil recovery. BP has carried out
an extensive research programme on low salinity injection
which has thus far included more than 20 reservoir condition
core floods on a range of sandstone reservoirs from its global
portfolio both in secondary and tertiary mode, more than 10
single well chemical tracer tests (SWCTT), and a log inject
log test. This program has resulted in a series of publications5-8
and the registration of the LoSal EOR process trademark.
These tests have shown improvements of waterflood process
efficiency by 5% to 38% by using low salinity water, or
corresponding reductions in residual oil saturation of 3 to 17%
pore-volume. The purpose of this work is to present a simple
extension to waterflood simulators that can be used to translate
corefloods or SWCTT into field scale estimates of low salinity
waterflood oil recovery and demonstate this with examples
from a sandstone reservoir.
Mechanism
A recent summary of the understanding of the mechanism
involved in the liberation of additional oil by low salinity
flooding has been given by Lager et al.8. Results of the work
of Tang and Morrow2 suggest that it is a wettability change
towards water wet with increasing waterflood recovery
corresponding to greater rates of spontaneous imbibition of
brine. This is also indicated by the direction of change of the
relative permeability in that there is a lower water relative
permeability and a higher oil relative permeability at a given
water saturation. While residual oil saturation is lower, water
relative permeability at residual oil is roughly the same.
Several studies have been undertaken to investigate potential
mechanisms2-4, 7, and the results from these studies have been
used to infer potential mechanisms for increased oil recovery
with low salinity waterflooding. Results from these studies
showed that:
Initial water saturation required- if no connate water
saturation is present no benefit is seen.
Crude oil is required- no impact is seen with refined
(depolarized) oils. This is also consistent with wettability
change because crude oils contain naturally occurring
surface active agents, acids and bases, which change the
wettability away from water-wet.
2 SPE 102239

Comparison of oil recoveries from low salinity and high VIP. The low salinity model consists of (in addition to the
salinity waterfloods on fired and acidized core material (a conventional waterflood capabilities):
process which inactivates the clays), are similar. 1) Salt is modelled as an additional single lumped component
in the aquous phase which can be injected and tracked. The
The effect has only been seen for clastics. viscosity and density of the aquous phase is dependent on
In some cases, pH changes have been observed in the salinity.
produced brine indicating a chemical interaction of the 2) Relative permeability and capillary pressure made a
injected brine with the connate brine. But in other cases, function of salinity.
particularly, when the pH is low to begin with, there is This dependence disappears at high and low salinities.
little change in pH and, in particular, it does not reach the High and low salinity relative permeability curves are
levels normally associated with either wettability change or inputs. Shapes are interpolated inbetween.
low interfacial tension mechanisms of caustic flooding. Residual oil saturation as a function of salinity is input as
a table to allow for more definition of the dependence.
While Morrow and co-workers did report produced fines 3) Portions of the connate water are made inaccessible to
along with the high pH in the produced fluids, the floods demonstrate the impact of the banking of connate water on the
done at the BP labs have not shown produced fines and process.
have roughly the same water relative permeability at high 4) Hysteresis between imbibition and secondary drainage
salinity and low salinity residual oil saturation. In water relative permeability is included to accurately model oil
situations when fines are mobilized, repeated low salinity bank development.
corefloods on the same core plug, show progressively 5) A model of dispersion within the water phase to help study
lower benefit from low salinity flooding. the impact of disperion.
The ideas behind these features along with the uncertainty
The recovery benefit upon injecting lower salinity brine
in them and the implications of the model are discussed in the
appears to increase with the abundance of some clays and
subsequent sections of this paper.
other minerals and consequently variations in lithology are
important to include in reservoir models. The Berea sandstone
Salinity Dependence
used by Morrow and co-workers for many of their expermints
Coreflood results show that the level of incremental oil
had predominantly Kaolinite clay and quartz. Increases in oil
recovery depends on the salinity of the brine but appears not to
recovery with increasing Kaolinite content was found from a
be simply proportional. Tests have shown that above a certain
series of SWCTT described below. A number of studies have
threshold, recovery does not depend on salinity and likewise,
shown that Kaolinite is wet by crude oil9-12. The components
below a certain level of salinity there is little dependence. Just
of crude oil are thought to be ionically adsorbed, particularly
where these thresholds occur appears to depend to some
to clays because they have a large surface area. Morrow and
degree on the system used and at the current level of
coworkers have found that varying the ionic content of both
understanding must be established through core or other
the injected and connate brine affects oil recovery and thus it
measurements. A predictive theory about the form of the
is clear that details of the brine chemistry are important. In
dependence has not been developed so we use a simple
some experiments chasing with brines that are richer in
empirical dependence in simulation with thresholds at high
divalents has lead to an apparent stop in oil production. Both
and low salinity as show in Fig. 1. The shape of the relative
Sharma and Filoco13 and Lager et al8 report that there is no
permeability is assumed to be linearly dependent on salinity
decrease in residual oil saturation when pure sodium chloride
between the thresholds and constant beyond. The residual oil
brine is used. This observation, coupled with the observation
can be specified over the entire range, but in most applications
than many surfaces are negatively charged at reservoir pH
will be specified at two extremes (0.2 and 0.81 in this
levels, has lead Lager et al8 to conclude that the mechanism
example).
involves ion-bridging with the divalent cation binding the
naturally occurring acidic surfactants to negatively charged
surfaces or possibly exchange of adsorbed cationic surfactants
with other ions. Ion exchange is typically rapid so these
mechanisms are also consistent with the rapid changes seen in
floods.
The detailed relationships between wettability and
surfactant adsorption and the multicomponent adsorption
isotherms are not yet available. In this work, we use the simple
approach of relating relative permeability directly to salinity. Fig. 1 Schematic of salinity dependence of relative permeability
used in the model.

Model
All the cases tested to date have shown that a substantial
The Low salinity model is based on established modelling
change in salinity is necessary to see an effect, probably less
approaches for chemical EOR. In essence, it uses new
than 25% of the connate salinity and consistently at 10% of
information and understanding to address long recognised
connate salinity with further increases seen in lowering below
approximations embedded in conventional waterflood
that level. The feature that there is no dependence at high
modelling. In particaular, it was added to BPs version of
salinity is consistent with industry experience where injected
SPE 102239
salinity was not normally regarded as an important variable in
designing water floods. In virtually all cases tested, substantial
low salinity effects have been seen for salinities in the range of
1000-2000 ppm so in most cases the thresholds are above and
below this range, for the high and low salinity thresholds.
Tang and Morrow3 showed examples which showed an
increase in recovery when the salinity of the connate and
injected brine was reduced from 15,000 ppm to 1,500 ppm and
then a further increase when the salinity was further reduced
to 150 ppm. A reduction to 10% of connate salinity was
effective and reducing it by a further factor of 0.1 was more
than twice as effective. Webb et al.7 showed a coreflood where
there was no production benefit between the formation water
at 80,000 ppm TDS and seawater at 30,000 ppm but a huge
benefit at 1,000 ppm TDS level. Fig. 2 shows reservoir
condition corefloods that demonstrate that a reduction to 20%
of seawater salinity, i.e. down to 5600 ppm gave a small
improvement whereas reducing the salinity to 5% or 1,400
ppm gave a substantial improvement and that much of the
benefit is delayed water break through. Lastly, a SWCTT was
run in Prudhoe Bay, well N-01A which started with a salinity
of 23,000 ppm, found no benefit at 7,000 pm and a 4% PV
response at 1,700 ppm5.
Fig. 2 Dependence of coreflood oil recovery on salinity (after
7
Webb et al ).
There is less data available for the value of a low salinity
threshold. Morrow and co-workers reported that for Dagang
crude in Berea core a reduction to 10% of the connate level or
2417 ppm gave a substantial increase in oil recovery whereas
decreasing to 1% or 242 ppm gave little further increase in
recovery. Corefloods reported here have shown improved
recovery for a tertiary flood with a salinity of 1250 ppm (4%
of connate) but no additional recovery for a subsequent
reduction down to 250 ppm (0.8% of connate).
Mixing of connate and Injected water
Injected water is expected to either displace connate water,
mix with it or potentially by-pass it. The bulk of the evidence
in the literature suggests that all of the water in the pore-space
is displaced by the injected water but the two types of water
mix to some degree, so that the displacement is not piston-
like. There is some evidence to say that under some
circumstances, the injected water does not directly displace all
the connate water but instead relies on diffusion for salt to
move from some dead-end pore-space to the bulk flowing
pore-space, i.e. that the process may not be treated as a simple
one-dimensional system with only displacement with
3
dispersion but requires three-dimensional diffusion. Exchange
with connate water is important in all kinds of chemical
flooding and based on the available evidence the standard
assumption is that connate water is displaced by injected water
with some degree of mixing. These models have been used to
match coreflood results and match field-scale projects, and
this is the approach used in the current model.
Salter and Mohanty14 studied two-phase steadystate flow
in Berea core to better understand how to model flow in
sandstones with application to chemical flooding. They ran
tracer floods in steadystate flow and matched the results with
the Coats-Smith model, which models pore-space as
consisting of a flowing fraction with dispersion in that portion
and a dead-end fraction which interacts with the flowing
fraction through diffusion. They found that while at
intermediate saturations there appears to be a portion of the
pore-space which is not flowing but a dendritic or dead-
phase, constituting perhaps up to 25% of the water-phase, both
near connate water saturation and residual oil saturation the
water-phase is entirely flowing. They also found that
dispersivities tend to be larger in multiphase flow than in
single-phase flow, by up to an order of magnitude. Thus, while
all the water is displaced, there is a significant amount of
mixing. Salter and Mohtanty also found that the measured
dispersion coefficients increase roughly linearly with velocity
rather than being constant, indicating that dispersion not
diffusion governs the flow within the flowing wetting phase.
Wang15 reported similar findings using a similar technique but
on Loudon core and Berea core made mixed-wet by aging in
Loudon crude. The results indicate that in mixed-wet rock the
water is also well connected. He also reported that the
dispersion levels within the water phases are as high for
mixed-wet as for water-wet systems. Based on these
observations and others, it appears that we should expect to
see large amounts of mixing but that connate water will be
displaced from the rock. This is important because it
influences how we interpret coreflood results. Because
connate water as well as oil is displaced when low salinity
water is injected, then the first water to be produced from a
core will be connate water and the connate water will form a
bank which starts out at connate salinity and eventually
decreases. The presence of this means that the standard
approaches to interpreting waterfloods cannot be used if
salinity dependent relative permeability is being used in
subsequent simulations. This is illustrated in Fig. 3.
Fig. 3 shows the impact of both grid resolution or
dispersion and connate water banking on oil recovery in one-
dimensional simulations of secondary high and low salinity
waterflooding. The blue bounding curves for the base (high
salinity) waterflood and the low salinity (LSWF) are
simulations run without the use of low salinity logic. Thus, the
curves that are used are equivalent to starting with the low
salinity recovery shown and interpreting the results using
standard core-flooding procedures. The intermediate curves
are generated using the low salinity logic. The labels show the
Pecklet number, the ratio of dispersivity to the length of the
system, associated with each simulation. Dispersion can be
simulated in two ways. Coarse models have an artefact which
is equivalent to dispersion, appropriately called, numerical
dispersion. This occurs because salt is mixing uniformly
4 SPE 102239
within each gridblock. It has been shown that the solution to
the difference equations is approximately equivalent to
rigorously solving the governing differential equations with
dispersion and a dispersivity equal to half the gridblock size16.
Thus, as the equation shows, the Pecklet number is equivalent
to the inverse of twice the number of gridblocks. Some
simulations include both effects, and other tests have shown
that modelling dispersion directly gives equivalent results.
Note firstly, that dispersion has a large impact on the results.
The more dispersion there is in the system, the more delayed
the recovery is. Secondly, for very low levels of dispersion,
the results do not reduce to the same as seen in a simple
waterflood with the low salinity curves because the
interpretation did not include salinity dependence. This
discrepancy occurs because of connate water banking.
Fig. 3 Impact of grid size and connate water banking on oil
recovery.
Connate water Banking
Connate water banking arises because much of the connate
water is displaced by injected water. This was identified as
early as 1947 when Russell, Morgan and Muskat17 conducted
waterflood experiments on oil-saturated cores containing 20
and 35 per cent connate water saturations and found that 80 to
90 per cent of the connate water was produced after only one
pore volume of water was injected. W.O. Brown18 noted that
the connate water in a reservoir is the water that actually
displaces oil from the pores of the rock during a waterflood.
During a waterflood connate water forms a zone which
separates the invading flood water front the continuous oil
phase. He studied waterfloods of 1.8 cp kerosene in
sandpacks of different lengths and showed that as the length of
the pack increased the connate bank formed a sharper front.
This quote has been miss-interpreted by some to mean that the
results of floods only depend on the salinity of the connate and
is probably the reason that many laboratories once used only
the connate brine to conduct waterfloods. This is incorrect
reasoning. as results in the next section further substantiate.
Kelly and Caudle19 also indicated that the connate water
saturation of a reservoir is swept from the pore space ahead of
injection water and that this decreases the oil recovery by
viscous water from that expected if such a bank were not
formed. They found that immediately following breakthrough,
oil and connate water were produced at an increasing water-oil
ratio until the more viscous injection water broke through.
Jones20 measured the connate banking when distilled water
was injected into Berea sandstone with 20,000 ppm connate
water. In another case he increased the viscosity of the
injected water by adding 20% glycerol, which showed slightly
more efficient displacement. He also tried to make Berea not
water-wet by saturating a dry core with decane. This appeared
to induce more mixing of injected brine with connate but
recovery was complete and connate was produced first.
More recently, Nielsen, Olsen and Bech21 doing
experiments on fractured chalk noted that connate water is
mobilized and travels ahead of the injection water. Connate
water is piled up as a bank and reaches a saturation of up to
50%. Graue, et. al.22 saw connate water banking in water-wet
and moderately-water-wet fractured chalk. Ovens, Larsen and
Cowie23 reported log observations that indicated that seawater
injection causes a bank of more saline formation water to be
swept up ahead of the seawater bank in the Dan Field, a chalk
reservoir. Thus, evidence at the core and field scales both
show connate water banking.
Kralik et al.24 in a study of Kuparuk relative permeability
reported that The initial aliquot of produced brine was found
to contain 6030 ppm I-(aq), whereas, analysis of the injected
brine found no detectable iodide. At the point of liquid
breakthrough, I- contaminated connate water was still in the
composite. By the end of the waterflood, however, the connate
water had been flushed from the composite. This sandstone is
strongly oil-wet and tests were run on preserved core,
indicating that connate water banking occurs in naturally
occurring oil-wet systems.
Sorbie et al.25,26 studied connate water banking in a
heterogeneous core system with the particular goal of
understanding how connate water should be handled in
chemical flooding simulators. They studied waterflooding of a
700 md Clashach core with the center removed and replaced
with 12 Darcy ballottini microspheres. Water was doped with
radioactive tracers. The authors reported that clear banking of
connate water was observed in both high and low permeability
layers. They obtained good matches to their experiments by
assuming complete mixing with connate water as we do here.
Extended Buckley Leverett Solution
The shape of the high resolution solution for secondary
waterflooding with low salinity water can be found by
analytical techniques, which make clear what aspects of the
relative permeability drive the solution. The saturation and
production as a function throughput for fixed rate is exactly
analogous to polymer or viscosified water flooding (but
without inaccessible pore-volume or adsorption), see for
example Pope27 and Lake28, with low salinity water changing
fractional flow in a way similar to viscous water.
Fig. 4 shows the major aspects of the solution. Whereas
ordinary Buckley-Leverett (BL) solution consists of a shock
and a spreading wave, the low-salinity solution consists of two
shocks, one corresponding to the transition between low and
high salinity (with a spreading wave as in ordinary Buckley-
Leverett flow), and a second corresponding the transition
between high water saturation and connate at high salinity.
Between the spreading low-salinity solution and the zone
SPE 102239 5

ahead of the front and connate water saturation is the connate
water bank with a constant intermediate water saturation and
water fractional flow. Note that the ordinary BL shock speed
is faster and, in particular, breakthrough of high salinity water
must occur earlier in the ordinary flood than in a low salinity
flood.
permeability, the behavior near breakthrough is dependent on
high salinity relative permeability. The most striking feature
of the solution appears to be the zone of constant fractional
flow.
Note that while this solution shows clearly that different
saturation fronts occur for different salinities (with slower
moving saturation fronts for the low salinity case), even if they
moved in a similar fashion, by overall material balance
considerations, water breakthrough in the low salinity flood
must occur later if the lower salinity floods get to lower
remaining oil saturations. This can be envisioned by doing the
thought experiment shown in Fig. 5. Compare the saturation
profiles in cases when both high salinity and low salinity
floods are at breakthrough. If we assume that the behavior
near the front is the same because of connate water banking so
the profiles overlie. However, the low salinity flood must get
to lower oil saturations so the water saturations in the low
salinity must be schematically like the cyan region added to
the blue region. The recovery at breakthrough is the water
saturation in the core minus the connate water saturation, the
area in blue above the dashed line representing connate.
Therefore, it can be seen to be higher for low salinity case
since it is the area under the curve is larger by the amount in
cyan. For the high salinity case pore-volume at breakthrough
is blue whereas for low salinity it is cyan+blue area. This
clearly demonstrates that the common misconception that later
water breakthrough means that connate water banking is not
occurring is incorrect.

0.7

0.6
Oil Recovery, fraction OOIP

0.5

0.4
High Salinity
0.3
Low Salinity

0.2

0.1

0
0 0.4 0.8 1.2 1.6
HCPV injection
Fig. 4 Extended Buckley Leverett solution for low salinity
secondary displacement, construction, saturation profile and
production.
Note that data for the fractional flow in the connate water
bank, S2, is not available from conventional high salinity
waterfloods (however, they might be obtained through Fig. 5 Thought experiment illustrating that higher recovery at
steadystate measurements) but is restricted by the condition breakthrough does not indicate that no connate water banking is
occurring.
that fw(S2)<fw(Sf). It is apparent that while the high
While long core experiments with careful measurements of
throughput behavior is dependent on low salinity relative
salinity will be required to validate the solution for low
6 SPE 102239
salinity floods experiments validation of the polymer and
viscous water solutions are available in the literature. Element
et al.29 show a validation of the solution given above for water
displacement with viscositified with polyethalene glycol in
sandpacks aged in North Sea crude. A simple waterflood was
run and then the viscosified waterflood was run. The
waterflood results coupled with the analytical theory predicted
the vicous flood results including the connate water bank.
Similar results were found when the sandpacks were polymer
flooded. Osterloh and Law30 reported the same behaviour for
sandpacks aged in Captain Field reservoir crude and flooded
with polymer solutions of different concentrations.
0.04
0.03
PV oil produced
0.02
0.01
0 somewhat greater in imbibition than secondary drainage but is
0 1 2 3
PV low Salinity Water
Fig. 6 Tertiary flooding Buckley Leverett solution for low salinity
flooding.
Many proposed low salinity floods involve tertiary
displacements either 1) because they are proposed for
improving existing waterfloods, or 2) it is necessary to
determine that aquifer support is not effective enough before
undertaking the water flood, or 3) because the size of the low
salinity water source is less than required for the whole field.
A similar Buckley-Leverett solution exists for tertiary flooding
which is demonstrated in Fig. 6. The principles used to derive
the solution are similar to the secondary flood.
Note that data for the fractional flow in the oil bank, SwHS,
is not available from conventional high salinity waterfloods
(they might be obtained through steadystate measurements,
however). Moreover, because it involves oil bank formation
and therefore increasing oil saturation, this curve is a
secondary drainage fractional flow curve, which may be
different than the imbibition. As in the secondary
displacement case, it is apparent that while the high
throughput behavior is dependent on low salinity relative
permeability, the behavior near breakthrough is dependent on
high salinity relative permeability but in this case secondary
drainage behavior. The most striking feature of the solution
appears to be the zone of constant fractional flow which
constitutes an oil bank and leads to the linear region in the
incremental recovery shown in Fig. 6. Fig. 7 shows a tertiary
coreflood run on reservoir core which displays the features of
the tertiary low salinity flood along with a fit from the model.
Fig. 7 Incremental oil recovery vs. pore-volumes of low salinity
water injected for a tertiary coreflood. After 0.4 of low salinity
injection the oil bank reaches the outlet (constant fractional flow),
at 1 PV it is mostly produced and most of the incremental is
produced after 2PV. Dashed line shows the match with the model.
Many oil-wet and mixed-wet sandstones have high
remaining oil saturations. Because these are the best
candidates for low salinity it is particularly important to be
able to model such situations. Oil-wet and mix-wet systems
often display hysteresis between imbibition and secondary
drainage that leads to fractional flow behaviour that is not
easily described without including hysteresis in relative
permeability. The most extreme situation is the oil-wet
hysteresis in which water relative permeability approaches
zero long before the water saturation reaches the connate level
in a secondary oil-flood, i.e. where there is an apparent
trapped water saturation. Water relative permeability is
4 the same in a subsequent waterflood. This behaviour leads to a
reduced fractional flow at a given water saturation and
therefore earlier oil breakthrough. To match the rapid
breakthrough of oil seen in corefloods, it has sometimes been
necessary to use this hysteresis behaviour. Of course, the
speed of breakthrough is also impacted by the remaining oil
saturation with higher remaining oil saturations leading to
faster breakthrough, all else being equal.
SPE 102239
Impact of Dispersion and Grid Resolution
It is clear from Fig. 3 that the results of simulations are
strongly dependent on the amount of dispersion. This occurs
directly because of the salinity dependence, the mixing of high
salinities and low salinities leading to intermediate salinities
that are less effective than very low salinities in improving
relative permeability behaviour. This has two important
implications. Firsly, in matching corefloods the correct level
of dispersion must be used. Secondly, simulations must be
done with realistic levels of dispersion, or pseudos must be
used which account for the level of dispersion in the system.
This means that we need to estimate the level of dispersion in
both the laboratory and in the field. Sources of information on
dispersion are limited. In core flooding, single phase
dispersivity can be estimated from the tracer floods done to
calibrate material balance but field scale measurements are not
readily available. Mahadevan et al31 addresses the level of
dispersivity to use in reservoir simulation of miscible gas-IOR
processes by using coreflood, SWCTT and other water-phase
tracer data.
Fig. 8 Dispersivity as a function of distance travelled from
31
Mahadevan, Lake and Johns . For the inter-well distances
important to low salinity flooding the dispersivity is roughly 2%-
5% of the distance travelled. Red dots indicate BP data.
Fig. 8 shows a plot of dispersivity measured as a function
of the scale of the measurement along with data from this
study. The plot shows that Inter-well tests and SWCTT appear
to agree and dispersivities appear to grow nearly linearly with
the distance the tracers travel (things apparently flatten at very
large distances but these are large compared to interwell
distances). The dispersivity is roughly 2-5% of the length
traveled, which implies that the number of gridblocks between
injector and producer should be roughly 10-25 based on the
numerical dispersion estimate of single-point upstream
weighting16. This result is approximate and carries a large
amount of uncertainty for many reasons. Firstly, we do not
know the level of mixing in the field very accurately and it is
expected to vary between fields because heterogeneity and
geometry of fields vary. The results in Fig. 8 show roughly an
order of magnitude variation for any distance. Secondly, a
heterogeneous model will have more mixing than a
homogeneous model so greater amounts of resolution will be
required to get the appropriate level of mixing. Miscible flood
simulations are also sensitive to dispersion levels and have
7
shown results that match the field response for 30-40
gridblocks between injector and producer when realistic levels
of heterogeneity are included in the reservoir description32-34.
Thirdly, the data themselves are somewhat controversial
beyond the scale of SWCTT. This is true because one cannot
distinguish between the effects of layers or, more broadly,
streamlines of different lengths, each of which might have
little mixing along their length but are mixed in the well bore,
and true mixing where intermediate compositions exist in the
reservoir. Thus, there is a possibility that levels of mixing are
overstated. This issue has been well recognized in the
literature but no simple ways of addressing it have been
suggested. One advantage of SWCTT is that they do give a
true estimate of mixing because they involve injecting and
then producing fluids through the same well (and because the
flow reverses direction retracing the forward path, the impact
of streamlines of different lengths would be cancelled out). A
disadvantage is that they involve a small distance traveled
compared to interwell distances. Mahadevan, Lake and
Johns31 were specifically trying to understand what level of
mixing to use in simulations and concluded that this level of
mixing is appropriate.
Fig. 8 also shows that corefloods often have a fair amount
of mixing and should not be considered as dispersion free. Fig.
9 that the dispersion levels of the reservoir core sample used to
generate the data in Fig. 7 has a dispersivity of 5% of core
length. The smooth lines show the match to the data. The
profile is reasonably matched indicating that concentration
spreads diffusively.
Fig. 9 Single phase dispersion tests in reservoir core matched
with the convective-dispersion equation. The dispersivity is 5% of
the core length.
Because the dispersion in a core is roughly the same
fraction of the core length as dispersion in the reservoir is as a
fraction of the interwell distance, we can imagine that the
results of corefloods will scale up directly to the field. Thus,
for a secondary waterflood where only low salinity brine is
injected, and where there is little aquifer influx and injection is
into the oil column, there is no need to use a simulation tool
which takes into account the influence of salinity. The curves
interpreted from the laboratory using standard techniques can
be used with conventional reservoir simulations. However, if
the levels of mixing where thought to be different then one can
argue that we must interpret the corefloods and put appropriate
levels of mixing into the field scale simulations. It is more
often the case that high salinity brine mixes with the injected
8 SPE 102239

low salinity brine decreasing the exposure in the reservoir. In

tertiary floods low salinity water will mix with the secondary
flood water and be delayed in getting to high recovery. Small
slugs of low salinity water will mix with the water in the
reservoir ahead and behind them and, therefore, if mixing is
large the effectiveness of slugs will be deminished.
To contact the largest portion of a reservoir with a limited
supply of low salinity water, it is natural to consider schemes
involving slugs of low salinity water in tertiary recovery. Fig.
10 shows the incremental oil recovery vs. pore-volumes of
low salinity water injected for a range of low salinity slug
sizes for two different grid resolutions in one dimension, 200
gridblocks (solid line) and 10 gridblocks (dashed). The results
demonstrate a strong dependence of incremental oil recovery
on grid resolution or equivalently, physical dispersion. For the
case of 10 grid blocks, a 25% PV slug leads to no incremental
oil recovery and a 50% PV slug leads to less than half that
which could be ultimately recovered. The 200 gridblock case Fig. 11 Incremental oil recovery for low salinity water injection at
leads to significantly higher recoveries with a 50% PV slug 10 PV of total injection since the start of low salinity injection for
yielding much of what might ultimately be recovered and a a range of slug sizes and grid resolutions.
12% PV slug leading to non-trivial incremental oil recovery.
To further quantify the impacts of grid resolution a series Dispersion on Radial Grids
of simulations were run with 30 and 1000 gridblocks. Fig. 11 Interpretation of SWCTT involves dispersion in radial
below shows the incremental oil recovery after 10 PV of total geometry and is an important source of field information about
water injection plotted against the slug size of low salinity the effectiveness of low salinity flooding. Dispersion levels
water. The results show clearly that for progressively smaller can be estimated from the tracers used in the tests or because
grid-block sizes progressively smaller slug sizes are optimum the same estimate of dispersivity that holds in one dimensional
with respect to the amount of incremental oil recovery flow holds in radial flow, matching the SWCTT data
achieved for a given amount of low salinity water injection. determines the grid resolution.
This demonstrates that we must estimate the level of
250 0.53 ft
dispersion in the reservoir to understand the optimal slug size.
1.06 ft
It is also clear from the Fig. 11 that there is a limit to how well
2.12 ft
slugs can do (compare the cases with 200 and 1000 200
Concnetration, IPA

4.23 ft
gridblocks). This occurs because while the recovery due to the
7.86 ft
oil bank shown in Fig. 6 can be obtained from a 150 0.13 ft
infinitesimally small slug if the dispersivity were are also a=0.26
infinitesimally small, the portion of the displacement which is a=0.53
recovered in the spreading wave requires finite throughput. 100 a=1.06
a=2.12
50 a=3.93

0
0 1000 2000 3000 4000 5000
Water Production

Fig. 12 Impact of grid resolution and dispersion level on tracer

Fig. 10 Incremental recovery versus throughput in a 1D simulation
with 200 (solid lines) and 10 (dashed) gridblocks.
production in a simulation of a SWCTT. For a uniform grid the
dispersion level is half the grid blocks size as in one dimension.
Fig. 12 shows simulations of the tracer, isopropyl alcohol,
in simulations of a low salinity SWCTT for different grid
resolutions and for different dispersivity levels and a grid
resolution of 0.132 ft. The results demonstrate not only that
dispersion has a profound effect on the flow, but that grid
block size can be used as a proxy for dispersivity. Matching
the SWCTT associated with low salinity tests gives a range of
dispersivities from 0.33 ft to 1.7 ft, depending on the test,
consistent with Fig. 8. Fig. 13 shows an example of the data
from a SWCTT and the match to a dispersion model. A series
SPE 102239 9

of these SWCTT were run to test the impact of slug size on the
low salinity process.
2500
Concnetration IPA, ppm

2000

1500

1000

500

0
0 1000 2000
Water Production, bbl
Fig. 13 Match to IPA production in a SWCTT. Results match a
dispersion model reasonably well with a dispersivity of 1.7 ft.
Fig. 14 shows a summary of the residual oil saturation
measured in the tests along with simulations of the tests. In
these tests first a high salinity water was injected to establish
high salinity baseline followed by the tracer test, a 0.2 PV slug
of low salinity water with 1 PV high salinity brine followed by
another SWCTT, a 0.3 PV slug of low salinity water followed
by 2 PV of high salinity water and another SWCTT, followed
by 2 PV of low salinity water. The results show that residual
oil saturation reduced by 4 saturation units upon injection of
the 0.2 PV slug and reduced another 7 saturation units with the
injection of a subsequent 0.3 PV of low salinity water. The
final 2 PV slug of low salinity water appeared to mobilize no
additional oil. Fig. 15 shows the relative permeability curves
used to match these tests. They display the characteristic
behavior found in many floods of more favorable fractional
flow behavior and a similar water relative permeability at high
and low salinity residual oil saturation.
0.6
0.5
0.4
Oil Saturation
3000 4000 5000
Fig. 15 High and low salinity relative permeability curves
used in the simultions shown in Fig. 14.
Impact of Dispersion in 2D
This section demonstrates the impact of slug-size of floods in
two dimensions with high resolution to establish that
numerical dispersion can model physical dispersion.
Transverse dispersion has been reported in the literature to be
an order of magnitude smaller than longitudinal dispersion28.
However, numerical dispersion is isotropic (for an isotropic
grid like this one) so the impact of having a significantly
smaller transverse than isotropic dispersion can be
demonstrated.
Simulations were run in a quarter five-spot configuration
with 162 gridblocks in both directions with DX=DY=14.8333
ft. This gives a numerical Peclet number of 0.3%. A physical
dispersivity of 40.75 ft equates to a total Peclet number of 2%.
A transverse dispersion of 0.74 ft was used to get a transverse
dispersivity which is 0.16*longitudinal. A waterflood was run
and then followed by tertiary low salinity slugs of various
sizes Fig. 16 shows a flood with a 0.12 pore-volume of low
salinity water. The color scale is set so that high salinity is
blue, low salinity is magenta and shades of red or higher are

ineffective at changing relative permeability. For this slug size

0.3
it is apparent that by 0.39 PV of total injection (0.12 low
salinity and 0.27 PV of chase high salinity), that there has
been sufficient dispersion that the low salinity slug is all above
0.2 the upper threshold in salinity and ineffective. While a clear
Simulation
SWTT
oil bank (in blue at the bottom, So~0.63) forms and a
0.1 significant portion of the model gets into the low 30s in oil
saturation (red), the oil bank does not maintain its integrity
after the low salinity slug becomes ineffective and incremental
0
oil recovery is slow and small. Fig. 17 shows a similar
8996 9016 9036 simulation for a 25% PV slug. This performs signficantly
Time, days better with the low salinity slug only losing effectiveness after
Fig. 14 Average oil saturation in the region of investigation of a the oil bank has reached the producer and significant amounts
series of four SWCTT assessing the impact of slugsize on low of oil have been produced. Fig. 18 shows results for 0.48 PV
salinity flooding, the green line shows simulation and the data slug of low salinity water most all the available residual oil
points are oil saturation interpreted from SWCTT.
target has been contacted by low salinity water.
10 SPE 102239

dimension appears to be different than in two dimensions with

Fig. 16 Salinity and oil saturation for a 0.12 PV slug of low salinity
water injected in the upper right and produced from the lower left.
the 2D cases showing more recovery at very small slug sizes
and less recovery large slug sizes than 1D. Thus, there appear
to be benefits to going to larger slugs that are seen in multi-
dimensions but are not seen in one-dimension. Fig. 21 shows a
comparison of the oil-recovery versus total water injection
since the start of low salinity injection between the one-
dimensional and high resolution results with longitudinal
dispersion appropriately large and also in an irregular
geometry for a different set of input relative permeability
curves. The results demonstrate both that the timing of EOR
breakthrough, the fraction of the incremental oil produced
over time and the dependence on slugsize all depend on
geometry. While the more irregular models have fast
streamlines which leads to some earlier breakthrough, they
also have slower streamlines and a lower efficiency of
producing the oil that is mobilized.
0.16
cont
0.14

Incremental Oil Recovery, PV

0.09 PV
0.12 0.19 PV

0.1 0.29 PV
0.52 PV
0.08
cont
0.06 0.11 PV
0.04 0.23 PV
Fig. 17 Salinity and oil saturation for a 0.25 PV slug of low 0.34 PV
salinity water injected in the upper right and produced 0.02
from the lower left. 0.52 PV
0
0 0.5 1 1.5 2
PV Injection since start low Salinity
Fig. 19 Incremental oil recovery as a function of pore-volumes of
water injected (low salinity + chase high salinity) for a high
resolution two-dimensional quarter-five spot model for different
low salinity slug sizes (dashed is 25x25).
0.16

0.14
incremental at 2 PV

0.12

Fig. 18 Salinity and oil saturation for a 0.48 PV slug of low salinity
0.1
water injected in the upper right and produced from the lower left.
Fig. 19 shows the impact of using grid size to model 0.08 152x152, at=0.2 al
numerical dispersion rather than physical dispersion. The fine 0.06
grid results are compared with comparable results for a 25x25 25x25
grid, so that the numerical dispersion just equals the physical 0.04 1D
dispersion. The results show good agreement between the
coarse grid and fine-grid simulations over a range of slug sizes 0.02 152x152, at=al
and for the whole range of pore-volumes of injected water.
Fig. 20 shows the incremental recovery versus low salinity 0
slug sizes for this quarter 5-spot model using a variety of 0 0.5 1 1.5 2
approximations to dispersion along with a one-dimensional Low Salinity Slug Size
model with the same level of dispersion (162 cells with Fig. 20 Incremental oil recovery versus low salinity slug size for
physical dispersion). The results show both that numerical different approximations to a 5-spot compared with one-
dispersion is a good approximation to physical dispersion, and dimensional results.
that the level of transverse dispersion appears to be a second In summary, it appears numerical dispersion is an adequate
order effect. In addition, the slug size dependence in one- approximation to physical dispersion given the uncertainty in
physical dispersion. This means that ultra-high resolution
SPE 102239
simulations are not necessary. However, it appears that the
slug size dependence in multiple dimensions is different than
one dimension and, therefore, simulations must be run in
multi-dimensions to access slug size dependence. If the model
adequately describes the physics of the process, then one-
dimensional physical tests may not directly tell us the optimal
slug size to use in the field. Results also appear to be sensitive
to the particular relative permeability curves used so it is
difficult to extrapolate conclusions from one situation to the
next.
Fig. 21 Incremental oil recovery as a function of pore-volumes of
water injected (low salinity + chase high salinity) for a high
resolution two-dimensional quarter-five spot model for different
low salinity slug sizes compared to a one-dimensional model with
the same grid resolution.
Lithology Dependence
As described earlier, the impact of low salinity flooding
depends on the mineralogy of the rock and, perhaps more than
standard waterflooding, this dependence should be included in
simulation of low salinity processes. Along with the coreflood
and SWCTT to assess slugsize dependence, SWCTT were
done to examine the impact of lithology. The sandstone under
consideration, consisted of principly quartz and Kaolinite clay.
From a combination of 4 logs (desity, gamma ray, neutron and
resistivity) it is possible to find the content of quartz,
Kaolinite, porosity, and water saturation which reproduces the
log response. Doing this over the perforated intervals where
single well tracer tests were done gives an estimate of the
Kaolinite content of the rock and over all the wells gives the
input to maping the Kaolinite content across the field. The
Kaolinte content of the core can be estimated from the logged
interval around it and also from point-count and x-ray
defraction information from the end-piece of the coreplug, but
estimates are expected to be different as a result of the
different scales of averaging. Fig. 22 shows the correlation
between incremental oil recovery and Kaolinite content is
reasonably strong. This relationship coupled with mapped
Kaolinite contents can be used to distribute low salinity
impacts by assigining different relative permeability curves to
different Kaolinite classes. An example of this is shown in
11
Fig. 23, which shows relative permeability curves that give the
incremental recoveries consistent with the trend in the
SWCTT and the timing consistent with them as well. Clay
content was grouped into equally abundant bins with the
average levels indicated by the squares in Fig. 22.
Fig. 22 Incremental oil recovery from low salinity flooding
as a function of Kaolinite content from single well tracer
tests and a coreflood. Correlation and bins also indicated.
1
Increasing
Clay content
0.8
0.6
Krw, Kro
0.4
0.2
HiSal LoSal
0
0 0.2 0.4 0.6 0.8 1
Sw
Fig. 23 Relative permeability curves used to characterize
different Kaolinite content. The curves correspond to
different levels indicated by the squares in Fig. 22.
Pseudoization
Often the recommended grid resolution to model physical
dispersion with numerical is significantly larger than can be
simulated effectively, requiring many hours or days to run.
This is inadequate for most studies because optimization is
necessary to understand how to best use a limited supply of
low salinity water and because aspects of the description are
uncertain so that multiple scanrios must be assessed to
understand the best design subject to uncertainty and give
appropriate ranges of outcomes. A conventional way to handle
this is to pseudoize by finding parameters that can make a
 12
fine-grid simulation match a coarse grid simulation. Fig. 24
shows an example of psudoization where a resolution of 10
gridblocks gives results similar to the 30 gridblock case. This
was done by varying the salinities at which relative
permeability began to be impacted by salinity and that at
which the salinity reached its low salinity limit, as well as the
curve shapes. The rock relative permeability curves at low
salinity thresholds of 1440 ppm and 6400 ppm were changed
to 1780 ppm and 17161 to approximate finegrid behavior with
the coarser grid model. Only minor changes to the relative
permeability curves were required to achieve a match, thus the
salinity thresholds appear to be an effective way to pseudoize.
The match between the 10 gridblock model run with pseudos
and the 30 gridblock model is shown in Fig. 24 for low
salinity water slug sizes of 0.25, 1.1 and infinity (continuous
injection). While the match is not precise it appears to capture
the major impacts. Often it is simplest to decide on the final
grid resolution and match the coreflood or SWCTT directly
with that resolution. This example demonstrates that
pseudoization can work but it is best practice to use
multidimensional simulations to scaleup because flows are
sufficiently different that multi-dimensional results may be
different than one-dimensional results.
0.12
Incremental Oil Recovery
0.1
0.08
0.06
0.04
0.02
0 recovery by Waterflooding, SPE Reservoir Engineering,
0 2 4 6 8 10
PV Throughput
Fig. 24 Match to the one-dimensional slug size dependence for
pseudos run on 10 gridblock models and results for 30 gridblocks
models. Low Salinity Slug size is indicated in the caption.
Discussion
The physics underlying the low salinity process is clearly
richer than in the simple model presented in this paper. While
we have been able to describe many observations in low
salinity flooding with this model it is likely that as richer sets
of experimental data become available aspects may be found
lacking. In partular, the ion-exchange/surfactant desorption
mechanisms may lead to changes in wave speeds and self-
sharpening and spreading of concentration fronts in some
portions of low salinity floods that are not modeled by the
simple (indifferent) dispersion used in this work. An example
of this is the cation-exchange process which has been shown
to be self-sharpening under dilution and spreading under
concentrating have been reported in the literature35-37. That
low salinity can be sustained at interwell distances has been
reported by Strange and Clould38 in the context of preflushes
for surfactant flooding.
SPE 102239
Conclusions
A model of wettability change with relative permeability
and capillary pressure dependent on wettability can be used to
describe the benefits of low salinity flooding.
The model produces oil banks in tertiary flooding consistent
with experimental observations.
In one dimension, the dispersion-free limit of the model can
be analyszed using standard methods. This analysis indicates
that higher recovery at breakthrough is not evidence that
connate water banking is not occurring and that accounting for
connate water baking is important in interpreting laboratory
relative permeability curves used in simulation.
Both at the laboratory and field scales, mixing is important
and influences the interpretation and predictions of flood
performance. Numerical dispersion appears to approximate
physical dispersion well for this model.
The model is capable of producing the slug-size dependence
observed in single well tracer tests as well as the behaviour of
corefloods.
Pseudo relative permeability curves and modified salinity
dependence can be used to appromae finer grid results with
cont
coarse grid simulations.
0.25
0.38 References
1. Yildiz, H.O. and Morrow, N.R. 1996 Effect of brine composition
0.51 on recovery waterflooding of Moutray crude oil by. Petroleum
1.1 Science and Engineering,. 14: p. 159-168.
2. Tang, G.Q. and Morrow, N.R. 1997, Effect of Temperature,
pseudo cont Salinity, and Oil Composition on Wetting behaviour and Oil
pseudo 1.1
November, pp. 269-276.
pseudo 0.25 3. Tang, G.-Q. and N.R. Morrow, 1999 Influence of brine
composition and fines migration on crude oil brine rock
interactions and oil recovery. Journal of Petroleum Science and
Engineering,. 24: p. 99-111.
4. Morrow, N.R., G. Tang, M. Valat, and X. Xie, 1998, Prospects
of Improved Oil Recovery Related to Wettability and Brine
Composition J. Pet. Sci. Eng., vol. 20, June, pp. 267-276.
5. McGuire, P. L., Chatam, J.R. Paskvan, F. K., Sommer, D.M. and
Carini, F. H. 2005 Low salinity oil recovery: an exciting
opportunity for Alaska's North Slope SPE 93903 presented at the
Western Regional Meeting held in Irvin, CA, 30 March 1 April.
6. Webb, K.J., C.J.J. Black, and H. Al-Jeel, 2004 Low salinity oil
recovery - log inject log SPE,. 89379.
7. Webb, K.J., C.J.J. Black, and I.J. Edmonds. Low salinity oil
recovery - the role of reservoir condition corefloods. in EAGE
conference. 2005. Budapest, Hungary.
8. Lager, A., Webb, K.J., C.J.J. Black 2006 Low Salinity Oil
Recovery - An Experimental Investigation presented at the
Society of Core Analysts meeting in Trondhiem, Norway.
9. Sincock, K.J. and Black, C.J.J. 1988 "Validation of Water/Oil
Displacement Scaling Criteria Using Microvisualization
Techniques," SPE 18294 presented at the 63 Annual Technical
Conference and Exhibition of SPE, Houston, TX Oct. 2-5.
10. Sutanto, E., Davis, H.T. and Scriven, L.E. 1990 "Liquid
Distributions in Porous Rock Examined by Cryo-Scanning
Electron Microscopy," SPE 20515 presented at the 1990 SPE
Annual Technical Conference and Exhibition on September 23-
26, 1990 in New Orleans, Louisiana.
SPE 102239 13

11. Fassi-Fuhri, O., Robin, M. and Rosenberg, E., 1991 "Wettability 31. Mahadevan, J., Lake, L.W., Johns, R. T., 2003 Estimation of True
Studies at the Pore-Level: A New Approach by the Use of Cryo- Dispersivity in Field-Scale Permeable Media, SPE 86303.
Scanning Electron Microscopy," SPE 22596, SPE Annual Tech. 32. Moulds, T.P., McGuire, P.L., Jerauld, G.R., Lee, S.-T., and
Conf. Exhib., Dallas, Oct.. Solano, R. 2003 Pt. McIntyre: A Case Study of Gas Enrichment
12. Jerauld, G.R. and Rathmell, J.J. 1997 "Wettability and Relative Above MME, presented at the SPE Annual Technical Meeting
Permeability of Prudhoe Bay: A Case Study In Mixed-Wet and Exhibilition, Denver, 5-8 Oct.
Reservoirs," SPERE February, 12 no 1,58-65. 33. McGuire, P.L., Spence, A.P., and Redman, R.S. 2000
13. Sharma, M.M. and Filoco, P.R. 2000 Effect of Brine Salinity and Performance Evaluation of a Mature Miscible Gas Flood at
Crude-Oil Properties on Oil Recovery and Residual Saturations, Prudhoe Bay, paper SPE 59326 presented at the SPE/DOE
SPEJ, Sept. pp 293-300. Improved Oil Recovery Symposium, Tulsa, 35 April.
14. Salter, S. J., and Mohanty, K.K. 1982 Multiphase Flow in Porous 34. McGuire, P.L. et al. 1995 Core Acquisition and Analysis for
Media: I. Macroscopic Observations and Modelling, SPE 11017 Optimization of the Prudhoe Bay Miscible Gas Project, SPERE
presented at the 57th Annual Fall Technical Conf. of the SPE, (May) 94.
New Orleans, LA , Sept. 26-29. 35. Pope, G. A., Lake, L.W. and Helfferich, F. G. 1977 Cation
15. Wang, F.H.L 1986 The effect of Wettability Alteration on Exchange in Chemical Flooding: Part 1 Basic Theory Without
Water-oil Relative Permeability, Dispersion, and Flowable Dispersion SPE 6771 presented at the SPE Annual Technical
Saturation, SPE 15019 given at the Permian Basin Oil &Gas Meeting and Exhibilition, Denver, Oct. 9-12.
Recovery Conf. of the SPE held in Midland TX, March 13-14. 36. Lake, L.W. and Helfferich, F. G. 1977 Cation Exchange in
16. Lantz, R.B. 1971 Quantitative Evaluation of Numerical Chemical Flooding: Part 2 The Effect of Dispersion, Cation
Diffusion, SPEJ Sept. p 315-320. Exchange, and Polymer/Surfactant Adsorption on Chemical
17. Russel, R.G. Morgan, F., Muskat, M., 1946 Some Experiments Flood Environment SPE 6769 presented at the SPE Annual
on the Mobility of Interstitial Waters, Trans AIME. 170, p 51- Technical Meeting and Exhibilition, Denver, Oct. 9-12.
61. 37. Appelo, C.A. 1994 Catio and proton exchange, pH variations,
18. Brown, W.O., 1957 The Mobility of Connate Water During a and carbonate reactions in a freshening aquifer, Water Resources
Water Flood, Trans AIME 210, p 190-195. Research, Vol 30, No. 10, pp 2793-2805, Oct.
19. Kelly, D.L. and Caudle, B.H.. 1966 The Effect of Connate Water 38. Strange, L.K. and Cloud, W.B. 1976 Displacement of Reservoir
on the Efficiency of High-Viscosity Waterfloods, Trans AIME Brine by Fresh Water-Four Field Case Histories SPE 5834
Nov. p 1481-1486 (SPE 1615). presented at the SPE IOR conference Tulsa, O.K., March 22-24.
20. Jones, S.C. 1985 Some Surprises in the Transport of Miscible
Fluids in the Presence of a Second Immiscible Phase, SPEJ Feb.,
101-112.
21. Nielsen, C.M., Olsen, D., Bech, N. 2000 Imbibition Processes in
Fractured Chalk Core Plugs with Connate Water Mobilization
SPE 63226 presentation at the 2000 SPE Annual Technical
Conference and Exhibition, Dallas, TX, 14 Oct.
22. Graue, A., Moe, R.W. and Bogn, T. 2000 Oil Recovery in
Fractured Reservoirs
http://www.ift.uib.no/SAFT/reservoarfysikk/Filer/PDFfiler/Nordi
sk2001%20graue%20moe%20 bogno.pdf
23. Ovens, J.E.V., Larsen, F.P. , and Cowie, D.R. 1998 Making
Sense of Water Injection Fractures in the Dan Field SPE
Reservoir Evaluation & Engineering, December , p 556-566.
24. Kralik, J.G., Manak, L.J. Jerauld, G.R. and Spence, A.P. 2000
"Effect of Trapped Gas on Relative Permeability and Residual Oil
Saturation in an Oil-Wet Sandstone" SPE 62997 Annual Tech.
Conf. and Exh. Dallas, TX, 1-4 Oct. 2000.
25. Sorbie, K.S. and Walker, D.J. 1988 A Study of the Mechanism
of Oil Displacement Using Water and Polymer in Stratified
Laboratory Core Systems, SPE/DOE 17397, p 823-834.
26. Sorbie, K.S., Wat, R.M.S. and Rowe, T.C. 1987 Oil
Displacement Experiments in Heterogeneous Cores: Analysis of
Recovery Mechanisms, SPE 16706 presented at the 62nd Annual
Technical Conference and Exhibition of SPE, Dallas, TX
September 27-30.
27. Pope, G. A. 1980 The Application of Fractional Flow Theory to
Enhanced Oil Recovery SPEJ June, p 191-205.
28. Lake, L.W., 1989. Enhanced Oil Recovery. Prentice-Hall,
London, pp. 175181. ISBN 0-13-281601-66.
29. Element, D.J., Goodyear, S.G., Sargent, N.C. and Jayasekera, A.J.
2001 Comparison of Polymer and Waterflood Residual Oil
Saturations 11th European Symposium on Improved Oil
Recovery, Amsterdam, The Netherlands, 11-12 June.
30. Osterloh, W. T., and. Law, E. J 1998 Polymer Transport and
Theological Properties for Polymer Flooding in the North Sea
Captain Field presentation at the 1998 SPE\DOE Improved Oil
Recovery Symposium held in Tulsa, Oklahoma, 19-22 April.