SPE 87016

Effects of Vertical Heterogeneity on Waterflood Performance in Stratified Reservoirs: A

Case Study in Bangko Field, Indonesia
A. K. Permadi, SPE, Institut Teknologi Bandung, I. P. Yuwono, SPE, PT. Caltex Pacific Indonesia, A. J. S. Simanjuntak,
SPE, Institut Teknologi Bandung

Copyright 2004, Society of Petroleum Engineers Inc.

by comparison to actual production from a simple injection
This paper was prepared for presentation at the SPE Asia Pacific Conference on Integrated scheme (one injector and one producer) in Bangko Field,
Modelling for Asset Management held in Kuala Lumpur, Malaysia, 29-30 March 2004.
Indonesia, indicate that oil production rates obtained from the
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
correlation show good agreement with those of
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to production history.
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 Introduction
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 Heterogeneity plays an important role in predicting waterflood
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 performance of a stratified reservoir. Heterogeneity may take
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. place in both horizontal and vertical directions. In this study,
Box 833836, Richardson, TX 75083-3836 U.S.A., fax 01-972-952-9435.
we will consider the problem of vertical heterogeneity. One
aspect of vertical heterogeneity is permeability variation. This
Abstract situation may result from various geologic processes that took
Stratified reservoirs may have different types of heterogeneity place during the sedimentation of the reservoir.1 The
in terms of grain size distribution in vertical direction. sedimentation of a certain reservoir rock may vary and take
Geological surveys (i.e. using well logs) have long recognized place at different geological time from that of another leading
the existence of fining upward and coarsening upward to a somehow unique classification of grain size distribution
formations. In this study, such formations refer to as systems for the corresponding reservoir. This is the reason why a
with decreasing upward and increasing upward permeability certain formation may exhibit a fining upward or coarsening
trends, respectively. Many waterflood candidate reservoirs upward grain size distribution classification. This phenomenon
have been found to follow either classification. However, the may be observed from the gamma ray log results of the
awareness of including this distribution classification as one of reservoir, which is usually reported as the stratigraphic
the screening criteria prior to waterflooding has not been property of the reservoir. A fining upward grain size
established in the oil industry. distribution indicates that the grain size of the rock becomes
A simulation study using a number of conceptual stratified finer in an upward sequence, which consequently results in a
reservoir models has been conducted. The results show that decreasing upward trend of the permeability values. A
grain size distribution classification should have significant coarsening upward grain size distribution, on the other hand,
impact on waterflood performance. Each of the two indicates a situation in which the grain size of the rock
classifications yields different effects on vertical sweep becomes coarser along the upward direction thus, an
efficiency resulting from different crossflow mechanisms, increasing upward permeability trend is observed. This will
which consequently gives different waterflood performance. It lead the formation to becoming a uniquely stratified reservoir.
has been found that the oil recovery from waterflooding a The effects of such types of heterogeneity on waterflood
reservoir with coarsening upward formation will always be performance in reservoirs are investigated in the present study
higher than that from waterflooding exactly the same reservoir as well as the effect of the level of permeability variation. The
but with the opposite classification (i.e. fining upward effect of varied magnitude of crossflow is also studied by
formation) even though the two reservoirs are of the same varying the value of vertical-to-horizontal permeability ratio
level of heterogeneity (i.e. similar values of coefficient of (kv/kh). A simulation approach is applied in this study by using
permeability variation). In this study, the degree of a numerical three-dimensional streamline simulator.
heterogeneity effects on recovery were investigated as well as From the above description, the following objectives
the magnitude of vertical-to-horizontal permeability ratio are derived:
effects. Also, permeability noise was addressed as the 1. To observe and determine the effects of the two
reservoir may contain contrast permeability streaks in between classification of grain size distribution on waterflood
the adjacent layers. performance in stratified reservoirs.
A correlation has been derived based on the simulation 2. To observe and determine the effects of varied level of
results and has been proven to be able to predict simulation
results with relatively good accuracy. Validations performed
2 SPE 87016

permeability variation on waterflood performance in analysis on crossflow caused by each of these forces.8,9
stratified reservoirs. Gaucher and Lindley10 presented a paper contained results
3. To observe and determine the effects of varied magnitude from an experiment conducted on scaled reservoir models
of crossflow on waterflood performance in with a five-spot waterflood pattern. They used several models
stratified reservoirs. of two-layered reservoirs of uniform thickness to describe the
Several assumptions are subject to this study as effects of injection rate and permeability stratification on
stated below: waterflood recovery efficiency involving the effects of the
1. All the models are conceptual with reservoir being three types of crossflow (i.e. viscous, gravitational, and
strongly water-wet. capillary). The results showed that the degree and distribution
2. The injection of water starts as the reservoir is put on of permeability stratification was an extremely important
production, therefore no history matching is necessary. factor to the oil recovery efficiency. It was also shown that for
3. The waterflooding system consists of one injector and a two-layered reservoir model, the position of the more
one producer. permeable layer had a greater effect on oil recovery than did
4. The wells are perforated along the interval of the the changes in injection rate.
productive zone. Goddin et al.6 also used several varied parameters to
5. Each layer is equal in thickness with one another. investigate their contributions on waterflood performance in
6. Mobility ratio between the displacing and displaced fluids such systems. Their results showed that capillary and viscous
is less than unity (M < 1). crossflows did have significant effects on waterflood
performance. Maximum crossflow due to capillary and
Waterflooding Stratified Reservoirs viscous forces was found to occur in the vicinity of the flood
Normally, the methods for predicting waterflood performance front in the more permeable layer. They found that, in the case
in stratified reservoirs are classified into two different types. of capillary crossflow, the initial capillary pressure in the
The first type is for a reservoir consisting of non- reservoir and that prevailing in the water-invaded region
communicating layers in which no vertical flow or crossflow behind the flood front was a very determining factor. Viscous
occurs between layers. In this case, it is assumed that each crossflow is such as to equalize the horizontal pressure
layer is separated by a continuous impermeable shale barrier gradients in adjacent layers. The effect of gravity to crossflow
from another. Dykstra and Parsons2 provided a very basic was also investigated. In a two-layered system, when the more
calculation method for predicting waterflood performance in permeable layer was on the bottom, gravity forces caused the
such reservoirs. In addition, they also introduced the denser fluid (i.e. water) to drain into the more permeable layer
coefficient of permeability variation, which can be statistically with unfavorable effect on the oil displacement efficiency, and
derived from a set of varied permeability values. This when the more permeable layer was on the top, water tended
coefficient is commonly known as the Dykstra-Parsons’ to flow from the loose (more permeable) into the tight (less
coefficient of permeability variation, VDP. The coefficient of permeable) layer, with favorable effect on the displacement
permeability variation is simply the ratio between the efficiency. It was also mentioned that for mobility ratios less
difference of P86 to P50 divided by P50 taken from a set of than unity, crossflow in such system was favorable.
permeability values. Several other authors have also presented Zapata and Lake8 attempted to provide an analytical theory
their methods of predicting waterflood performance in non- considering the contribution of viscous crossflow to fluid flow
communicating stratified systems.3,4,5 in a stratified reservoir. In their work, several simplifying
It is in fact that most stratified reservoirs do exhibit some assumptions were made. The effects of gravity and capillary
conditions in which crossflow is very likely to take place forces on crossflow were neglected as well as viscous
during the displacement process. Therefore the second type of fingering and several other aspects. The only determining
methods for predicting waterflood performance that appeared factor on which crossflow solely depend is the pressure
in literature is for a reservoir in which crossflow between gradient formed by the displacing and displaced fluids having
layers occurs. In this case, no impermeable barrier is present unequal mobilities. The direction of viscous crossflow is
between layers, and consequently calculations have to account governed by the mobility ratio of the interacting fluids (i.e.
for crossflow that may occur during the displacement process. water and oil). For a favorable displacement (M < 1), the
Goddin et al.6 presented a paper discussing waterflood direction of crossflow is from the low to the high velocity
performance in a stratified system with crossflow. Their study layer at the leading front and in the reverse direction at the
made use of numerical simulation and was performed on a trailing front as indicated by Fig. 1. This results in receding
field-scale model of a two-layer, water-wet sandstone and advancing of the leading and trailing fronts respectively,
reservoir. They showed that the computed oil recovery for which consequently improve the vertical sweep efficiency of
such system was always intermediate between that predicted such displacement compared to that where no crossflow is
for a uniform system and that for a stratified reservoir with no encountered since the flood front becomes more sharpening
crossflow. However, their study did not account for cases in (piston-like). For cases of reservoir models with three or more
which the models are comprised of more than two layers as layers, it was indicated that, different permeability orderings
well as the effect of permeability ordering in vertical direction. of the layered models produced different vertical sweep
Crossflow is a vertical flow from one layer to another in a efficiency for the corresponding water-oil displacements.
stratified reservoir. There are three contributing forces
initiating crossflow namely, viscous force, gravity force, and
capillary force.7 Several authors have attempted to perform
SPE 87016 3

Methodology cumulative injected water as fractions of the total pore volume

Cases Studied. Several cases were created through conceptual of the reservoir.
reservoir modeling. The first step considered in creating such
cases is to prepare four sets of varied permeability values Simulator Used. In the present study, the use of a three-
ranging from 0.6 to 0.9 in terms of their Dykstra-Parsons dimensional streamline simulator is fully encouraged. All the
coefficients of permeability variation, VDP. This range of VDP models built for this study have been set in such a manner as
is chosen since most heterogeneous reservoirs exhibit such to be able to describe the effect of spatial placement of vertical
variation of permeability values. This preparation is heterogeneity on waterflood performance in the simplest way.
accomplished by making use of a computer program to The simulator first solve the pressure solutions for a
generate permeability values that falls within the expected certain time-step by applying an IMPES (Implicit Pressure
range of variation expressed by their VDP. Therefore, each set Explicit Saturation)-like method and proceed with solving for
contains as many as twenty-five values of permeability since the saturation by using a front-tracking method. This front-
each model consists of twenty-five layers. These values of tracking method does not depend on grid size and model
permeability are an arbitrary function of porosity such that low geometry since it tracks the saturation fronts (discontinuities
values of permeability result from low values of porosity and in saturation) along the streamlines/tubes established from the
vice versa. pressure solution, which results in a typical Bukley-Leverett11
These sets of permeability values were then assigned to the saturation profile. This is somehow a quite distinguishing
simulator with two different trends, namely decreasing upward technique compared to the methods commonly used by other
and increasing upward. A decreasing upward permeability reservoir simulators.
trend represents a fining upward grain size distribution (most Streamtube models was first proposed by Muskat and
permeable layer on bottom of the reservoir) and an increasing advanced by several other authors.7 Streamtubes for the
upward permeability ordering represents a coarsening upward reservoir model are drawn from a set of estimated pressure
grain size distribution (most permeable layer on top of the values where it is assumed that there is no crossflow between
reservoir). Thus, we have eight sets of permeability data the tubes. The purpose of this method was originally to reduce
coming from two different grain size distributions in which a set of two-dimensional displacement problems to a set of
each consists of four sets of permeability values. one-dimensional problems. In this simulator, the streamtubes
Another set of permeability values is also considered here. are drawn from isobar lines (obtained from the implicitly
This set of data consists of permeability values that exhibit solved pressure equations) and then the saturation equations
very low variation (very small VDP), which is used to represent are solved along the streamlines resulting in the reduction of a
the performance of a reservoir with no permeability set of three-dimensional saturation equations to a set of one-
stratification (uniform system). In this case, a set of dimensional saturation equations.
permeability values with a coefficient of variation of 0.02
is prepared. Model Building
Several modified data sets are also prepared so that the Pertinent reservoir and fluid data are presented in Table 1
output of the simulator can adequately present the effects of while relative permeability data is shown in the following
some varied variables that are expected to be seen from the Table 2. From these relative permeability values, the end-point
calculation results. These modifications consist of two types as mobility ratio, M, was calculated as follows:
described as follows:
1. Variation on the values of vertical-to-horizontal ⎛ k rw ⎞
⎜ ⎟ 0.318
permeability ratio, kv/kh, in order to see the effect of ⎜µ ⎟
crossflow magnitude. A value of kv/kh being very small λ ⎝ w ⎠ @ Sor
M= w = = 0.96 = 0.3224 ………. (1)
(close to zero) indicates a very minimum crossflow and is λ o ⎛ k ro ⎞ 0.97
⎜ ⎟
used in this simulation study to represent the case in which ⎜µ ⎟ 0.95
no vertical communication takes place between layers. ⎝ o ⎠ @ Swi
2. An alteration in several permeability values within the
layers so that the model does not exhibit an absolute fining
upward or coarsening upward grain size distribution, but is
Table 2–Relative Permeability Data
still considered to generally exhibit either one of the two.
This is accomplished by generating a set of random
Sw krw kro
numbers to be employed as multiplying factors so that the
effect of realization, in which no absolute fining or 0.2 0 0.97
coarsening upward of grain size distribution can be found 0.25 0.003 0.94
in the field, can also be represented by the 0.35 0.028 0.83
simulator results. 0.4 0.069 0.171
0.55 0.143 0.05
A generalization of the simulator results is obtained by
0.7 0.318 0
normalizing the producing oil rate and time as to be able to
1 0.318 0
describe the performance of different reservoir models in a
general way. This is accomplished by analyzing the
The permeability and porosity values for the reservoir layers
performance of the reservoirs in terms of oil recovery as
are shown in Tables 3 through 6. All the permeability values
fractions of the original oil in place and the amount of
4 SPE 87016

are set in a decreasing upward manner (fining upward grain coefficient of variation to avoid the models from having
size sequence). Four other sets of permeability data have different levels of permeability variation. In this specific case,
exactly the same values (with the same VDP) as those shown in verification is performed on systems with VDP = 0.8. As many
Table 3 through 6 but are set in an increasing upward manner as five additional models were run, in which each model
(coarsening upward grain size sequence). represents the effect of realization due to actual field
For our simulation study, only the effects of the two types condition. Figs. 10 through 14 shows the effects of realization
of grain size distribution are considered (i.e. fining upward in which each figure represents different type of noise. From
and coarsening upward), no mixed of grain size distribution is these charts, it is clear that realization does not change the
considered. We assumed that there will be no cases in which trend of the performance of such stratified reservoirs due to
the grain size distribution of a particular reservoir would permeability trend (grain size distribution). The waterflood
follow the combination of the two (fining and coarsening or performance of a reservoir consisting of layers with increasing
vice versa) since there should be a fine permeability barrier upward permeability still lies above that of a reservoir with
between the two types of grain size sequence due to different decreasing upward permeability. Thus, all the results are valid
geological time in which the sedimentation of such reservoir and consequently should be able to represent the actual
takes place.1 performance of such reservoirs.
However, reservoirs do exhibit certain conditions in which
grain size is not distributed in an absolutely fining or Regression Analysis. It was promptly observed from the
coarsening upward manner. Therefore, in order to encounter above results that the waterflood performance of the two
the effect of such phenomena, several models containing non- distributions is different from one to another. Therefore, the
absolute fining upward or coarsening upward grain size next step taken in this study is to perform regression analysis
sequence are also investigated. The non-absolute condition is on the simulator results to relate such specific distributions to
attained through generating several other models of which the oil recovery. In other words, this analysis relates oil recovery
permeability values of some of the layers are allowed to (dependent variable) to three independent variables namely,
exhibit non-absolute trend. Thus the results obtained from the cumulative injected water, coefficient of permeability
simulation should be able to describe the general effects of variation, and another independent variable called the median
permeability trend on waterflood performance in of “horizontal permeability ratio of sequential layers.” A
such reservoirs. correlation is presented as the product of data analysis. This
In this study, it is assumed that the level of water-oil correlation succeeds in distinguishing the performance of a
contact is so far below the reservoir that the distribution of reservoir with increasing upward permeability trend and that
water saturation is uniform throughout the reservoir (before of a reservoir with decreasing upward permeability trend.
displacement takes place) and thus the values are equal to the The term horizontal permeability ratio of sequential layers
minimum water saturation (i.e. irreducible water saturation). refers to a value that can be obtained by following the
Consequently, there is no need to define capillary pressure- procedure as taken in the following example:
water saturation relationship for the model. a. Consider a reservoir consisting of layers of five varied
horizontal permeability values. The permeability values
Results And Observation are 15, 20, 100, 125, and 500 mD, respectively. These
Results. Figs. 2 through 5 shows the effects of permeability layers are ordered in two different manners i.e. decreasing
trend on waterflood performance of the stratified models in upward (least permeable layer on top) and increasing
terms of oil recovery versus cumulative injected water. Figs. 6 upward (most permeable layer on top).
and 7 show the effects of different level of permeability b. For those values ordered in a decreasing upward manner,
variation, which is represented by varied values of VDP on the values of “horizontal permeability ratio of sequential
waterflood performance in stratified reservoir in terms of oil layers” are 0.75, 0.2, 0.8, and 0.25, respectively. So, each
recovery versus cumulative injected water. The effects of value is expressed as their fraction relative to the next layer
crossflow magnitude may be implied by varying the values of in the sequence. The last value (500 mD) does not have
vertical-to-horizontal permeability ratio, kv/kh. These effects any ratio since there is no other layer beneath the layer.
are indicated by Figs. 8 and 9. As mentioned, it is quite c. For those values ordered in an increasing upward manner,
impossible for a certain formation to have neither an the values of “horizontal permeability ratio of sequential
absolutely fining upward nor coarsening upward grain size layers” are 4, 1.25, 5, and 1.33, respectively. In this case,
sequence. Normally, there will be some layers that exhibit less the last value (15 mD) has no ratio since there is no other
fine or less coarse grain size contrary to their distribution, but layer beneath the layer.
still can be generally considered as to have a fining upward d. Therefore the median (50 percentile) for those values of a
grain size sequence or coarsening upward grain size sequence. decreasing upward system is 0.5 and for those of an
In this study, the term noise is used to indicate such increasing upward system is 2.665.
phenomenon. Therefore verification of the results is done by A convention that has to be made here is that the calculation
allowing the models to have certain layers in which the starts from the uppermost layer towards the lowermost layer.
permeability values does not solely exhibit a decreasing The analysis was performed on 160-observation data. The
upward or increasing upward trend. A computer program is comparison between the simulator results and correlation
used to generate a certain range of random numbers that are results can be seen in Fig. 15. This comparison is based on a
assigned to the models as multiplying factors of the stratified system with VDP equals 0.75 and the only difference
permeability values. A check should be made on their is in the type of permeability trend. This can also be accounted
SPE 87016 5

as a validation to the correlation since the simulator results of favorable effect of gravity forces. On the other hand, when the
this system (VDP = 0.75) were not used as the observation data more permeable layers are on the lower part of the reservoir;
for deriving the correlation. since viscous crossflow still moves toward the same direction
Another validation is performed by using actual field data. (oil flows from the less to the more permeable layers at the
In this case, predictions are made on oil production rates leading front and water flows counter-currently at the trailing
resulting from a certain amount of injected water taken from front) and gravity forces still forces water to drain into the
simulated field injection rates data. These predicted values are lower part of the reservoir, the effect of gravity crossflow is no
then compared with those obtained from field oil production longer favorable. This phenomenon had also been observed by
rates data. The result, as can be seen in Fig. 16, shows very Berruin and Morse12 when conducting their investigation.
good agreement between each other. However, they mentioned that there had been certain limits to
which this condition still holds due to injection rate. In our
Discussion study, the effect of injection rate is already encountered by
From Figs. 2 to 5, we can see the effects of different normalizing the cumulative injected water by dividing it to the
permeability trends on waterflood performance provided that reservoir pore volume. Fig. 17 illustrates this phenomenon in a
all other properties of the models are similar. It is very clear stratified system with three layers where in our case the
that, for a certain stratified reservoir, an increasing upward relatively similar mechanism is assumed to occur repeatedly.
permeability trend should result in a better waterflood The black arrows pointing towards the lower part of the
performance of the reservoir compared to that of a decreasing reservoir indicate the contribution of gravity forces to
upward permeability trend. The computed amount of oil that crossflow while the other bended arrows indicate the direction
can be recovered (oil recovery) from a waterflooded reservoir of viscous crossflow in adjacent layers.
with increasing upward permeability trend is larger than that It should be noted here that one way to better understand
with an opposite trend of permeability. The two different and extend such mechanism to a more complex case is
trends of permeability variation cause crossflow to yield through water saturation profiles at different time stages
different effects on displacement process within the reservoir. during the displacement process. It was found in our study that
For a reservoir with increasing upward permeability trend, the in systems with no-crossflow, after water has broken through
effects of crossflow are more favorable compared to that for a within the most permeable layer, most of the injected water
reservoir with the opposite permeability trend. In addition, as will continue flowing toward that layer (most permeable),
Goddin et al.6 noted earlier, the oil recovery for systems with leaving other layers remain poorly swept. On the other hand,
crossflow (both with increasing and decreasing upward in systems with crossflow, the movement of fluid (i.e. oil and
permeability trends) are always intermediate between that water) in vertical direction enhances vertical sweep efficiency,
predicted for a uniform system and that for a stratified with different contribution depending upon permeability trend.
reservoir with no crossflow. As can also be seen in Figs. 2 to 5, the difference in
As stated in the previous section, maximum crossflow due performance between reservoirs with increasing upward
to viscous forces occurs at the vicinity of the front in the more permeability trend and that with an opposite permeability
permeable layer. However, according to Zapata and Lake, trend will reduce as the reservoir achieves homogeneity. For
viscous crossflow does not only take place at the leading front models with VDP being equal to 0.6, the difference in
in the more permeable layer, but also at the trailing front in the performance due to different trend of permeability variation is
less permeable layer.8 For a displacement with mobility ratio less than those of which the Dykstra-Parsons’ coefficients are
of the interacting fluids being less than unity (M < 1), the equal to 0.7, 0.8, and 0.9. Therefore the level of heterogeneity
direction of viscous crossflow is from the less to the more has significant impact on the difference in performance due to
permeable layer at the leading front and from the more to the different trends of permeability variation. As the reservoir
less permeable layer at the trailing front as illustrated by Fig. achieves homogeneity, the effect of permeability trend
1. In our case, mobility ratio is less than unity and water is becomes less significant on waterflood performance. This
displacing oil, therefore oil flows from the less to the more phenomenon can be recognized more clearly through the next
permeable layer at the leading front and water flows counter- Figs. 6 and 7. It is shown that in both cases, i.e. systems with
currently at the trailing front from the more to the less increasing upward and decreasing upward permeability trend;
permeable layer. Since the number of layer is more than two, a as the level of heterogeneity decrease (decreasing VDP), oil
relatively similar mechanism is considered to occur repeatedly recovery from such reservoirs will increase. This is due to the
in the adjacent layers throughout the reservoir. These should tendency of the reservoir to act more like a uniform system as
result in receding and advancing of the fronts within the more discussed in the preceding section.
and the less permeable layers resulting in a better performance Now we will consider the effect of crossflow magnitude.
of a stratified reservoir with crossflow compared to those with From Figs. 8 and 9, we can see that a relatively small change
no-crossflow. In other words, the existence of crossflow in kv/kh does not have any significant effect on the models
enhances the vertical sweep efficiency of the displacement in performance. However a change up to one hundred folds may
stratified reservoirs. have significant impact on the performance as indicated by the
When the more permeable layers are on the upper part of lines representing the performance of such reservoirs with
the reservoir, crossflow due to gravity forces favors the flow kv/kh being 0.001. It is shown that a change of kv/kh from 0.1
of water to drain toward the less permeable layers on the lower to 0.001 has greater impact on the performance of a reservoir
part of the reservoir. Thus, vertical sweep efficiency is even with increasing upward permeability trend. When compared to
better than that when only viscous crossflow occurs due to the the performance in which kv/kh equals 0.1; for a system with
6 SPE 87016

decreasing upward permeability trend, a change of kv/kh to A collection of simulator results was observed, and
0.001 may decrease the oil recovery up to around 2% whereas regression analysis was performed on 160 sets of data in order
for a system with increasing upward permeability trend the to derive a correlation relating oil recovery to MPR, pore
same change in kv/kh may decrease the oil recovery up to more volume of injected water, and coefficient of permeability
than 10%. In general, a decrease in crossflow magnitude variation. As a result, we can rely on this correlation to
would result in worse vertical sweep efficiency. However, in approach the simulator results to distinguish the effect of
most actual reservoirs the value of kv/kh usually falls not very permeability trends on waterflood performance in stratified
far from 0.1. reservoir. The correlation is given as follows:
As mentioned before, it is very likely to find a reservoir in
which the permeability does not exhibit an absolutely Oil recovery = 0.089406(MPR ) − 0.1485(VDP )
decreasing or increasing upward trend. There may be several
layers among all the layers constructing the reservoir that do − 12.2323((0.1 × PVI) −0.0069 ) + 12.80076
not necessarily exhibit an absolute trend, but still considered ………. (2)
to generally follow either one of the two trends. For such
cases, as shown in Figs. 10 to 14, the difference in where MPR = median of horizontal permeability ratio of
performance due to different permeability trends does not sequential layers and PVI = cumulative injected water as
follow the phenomena shown by Figs. 2 to 9 (decrease with fraction of reservoir pore volume.
decreasing level of heterogeneity) due to the existence of A validation to the correlation is performed by using it to
noise. This is also the reason why some of the models do not predict waterflood performance of such system with VDP of
exhibit increasing oil recovery although their heterogeneity 0.75, one case that has not been used in the observation
levels have been reduced. However, it is shown that a samples. The correlation results show good agreement with
reservoir with increasing upward permeability trend still yields simulator results, as can be seen in Fig. 15.
better performance compared to that with the opposite From Fig. 16, it can also be seen that, the correlation can
permeability trend provided that all other properties be used to predict oil production rate with relatively good
are similar. accuracy for a certain rate of injected water. The actual oil
The last part considered in this study is performing some production rate data were taken from Bangko Field, in Duri,
data analysis of the results. From all the results, we can see Riau, Indonesia for a system with one producer and one
that even though the only difference between a reservoir and injector. At early phase of the production period, the predicted
another is only in permeability trend, performance of either of oil rates show higher values than those of the actual data due
the model may vary from one to another. The level of to the assumption of allowing waterflood to start as the
heterogeneity (indicated by Dykstra-Parsons’ coefficient of reservoir model is put on production. It is also observed that,
permeability variation) is not the only factor influencing at the early stage of production, oil rates resulting from a
waterflood performance in stratified reservoirs. We also have system with increasing upward permeability trend, are higher
to account for the impact of permeability trend. Of course, the than those from a decreasing upward permeability trend. This
amount of injected water is also important; as more and more will consequently result in higher recovery for a system with
water is injected into the layers comprising the reservoir, more increasing upward permeability trend as compared to that with
and more oil can be recovered. In other words, as time the opposite permeability trend.
increases, more and more oil will be swept out of the stratified
reservoir resulting in higher cumulative oil recovery. Concluding Remarks
A question is what other parameter that may be able to From our results, keeping in mind the limitations and
affect the performance of such stratified reservoir as to assumptions applied to this study, we can derive the following
distinguish the effect of permeability trend on waterflood conclusions:
performance. In this study, we try to relate oil recovery not 1. The waterflood performance in stratified reservoirs
only to the amount of injected water and coefficient of depends upon the classification of grain size distribution
permeability variation but also to another parameter called along the vertical direction. A stratified reservoir with
“median of the horizontal permeability ratio of sequential increasing upward permeability trend will yield better
layers,” MPR. waterflood performance (higher vertical sweep efficiency)
It is shown that, for a stratified reservoir with increasing as compared to that of the opposite permeability trend
upward permeability trend, the median value of the horizontal “provided that all other properties are similar.”
permeability ratio of sequential layers will be more than unity 2. In a stratified reservoir the waterflooding oil recovery
whereas for that with decreasing upward permeability trend, decreases as the permeability variation in vertical direction
the median value will be less than unity. Hence now there is a increases. This decrease will be greater in a reservoir with
fine line separating the system of increasing upward decreasing upward permeability trends. However, for
permeability trend from that of decreasing upward systems exhibiting either absolute increase or decrease in
permeability trend. Furthermore, the use of this parameter upward permeability trends, as reservoirs achieve
should be helpful in order to distinguish stratified systems uniformity; the difference in performance “due to
with increasing upward permeability trend (coarsening upward permeability trend” will reduce.
grain size distribution) from stratified systems with decreasing 3. The presence of crossflow will generally improve vertical
upward permeability trend (fining upward grain sweep efficiency of waterflood in stratified reservoirs. This
size distribution). phenomenon is more notable in a reservoir with increasing
SPE 87016 7

upward permeability trend. Moreover, for reservoirs with References

either increasing or decreasing upward permeability trends, 1. Van Wagoner, J.C., Mitchum, R.M., Campion, K.M., and
if crossflow occurs, the waterflood performance will be Rahmanian, V.D.: “Siliclastic Sequence Stratigraphy in Well
intermediate “between that of a uniform system and that of Logs, Cores, and Outcrops,” AAPG Methods in Exploration
a stratified system with no-crossflow.” Series, No. 7, AAPG (1990).
4. In stratified reservoirs with different grain size distribution, 2. Dykstra, H. and Parsons, R.L.: ”The Prediction of Oil Recovery
the median value of horizontal-permeability-ratio of by Waterflood,” Secondary Recovery of Oil in The U.S., API
sequential layers (MPR) is an alternative parameter to (1950) 160-174.
distinguish waterflood performance. 3. Stiles, W.E.: ”Use of Permeability Distribution in Waterflood
Calculations,” Trans., AIME (1949), 186, 9.
Considering the above conclusions, several new concerns 4. Reznik, A.A., Enik, R.M., and Panvelker, S.B.: ”An Analytical
come to light and therefore we are suggesting further studies Extension of the Dykstra-Parsons Vertical Stratification Discrete
in order to discover ways to: Solution to a Continuous, Real-Time Basis,” SPEJ (Dec. 1984)
1. Improve vertical sweep efficiency in reservoirs with 643-656.
decreasing upward permeability trend (fining upward grain 5. Tompang, R. and Kelkar, B.G.: “Prediction of Waterflood
size distribution). Performance in Stratified Reservoirs,” paper SPE 17289
2. Evaluate the effects of different thicknesses “of the layers presented at the SPE Permian Basin Oil and Gas Recovery
comprising the reservoir” on waterflood performance. Conference, Midland, TX, March 10-11, 1988.
6. Goddin, C.S., Jr., Craig, F.F., Jr., Wilkes, J.O., and Tek, M.R.:
Nomenclature “A Numerical Study of Waterflood Performance in a Stratified
k = permeability, mD System with Crossflow,” JPT (June 1966), 765-771.
kr = relative permeability 7. Willhite, G.P.: Waterflooding, SPE Textbook Series, Vol. 9, SPE,
kh = horizontal permeability, mD Richardson, TX (1986).
kv = vertical permeability, mD 8. Zapata, V.J. and Lake, L.W.: “A Theoretical Analysis of Viscous
kro = relative permeability to oil Crossflow,” paper SPE 10111 presented at the 56th Annual Fall
krw = relative permeability to water Technical Conference and Exhibition of the SPE of AIME, San
M = mobility ratio Antonio, TX, October 5-7, 1981.
Sor = residual oil saturation, fraction 9. Haq, S. and Reis, J.C.: “Predicting Capillary Crossflow in
Swi = initial water saturation, fraction Layered Reservoirs,” paper SPE 26651 presented at the 68th
VDP = Dykstra-Parsons’ coefficient of variation Annual Technical Conference and Exhibition of the SPE of
µo = oil viscosity, cp AIME, Houston, TX, October 3-6, 1993.
µw = water viscosity, cp 10. Gaucher, D.H. and Lindley, D.C.: “Waterflood Performance in a
φ = porosity, fraction Stratified Five-Spot Reservoir–A Scaled Model Study,” Trans.,
λo = oil mobility AIME (1960), 219, 208.
λw = water mobility 11. Buckley, S.E. and Leverett, M.C.: “Mechanism of Fluid
Superscript: Displacement in Sands,” Trans., AIME (1941), 142, 152.
o
= end-point 12. Berruin, N.A. and Morse, R.A.: “Effect of Flooding Rate and
Permeability Ordering on Waterflooding Stratified Reservoirs,”
paper SPE 7158 (Jan. 1978).

Table 1–Reservoir and Fluid Data

Reservoir length = 500 ft divided into 10 blocks of equal length (50 ft each)
Reservoir width = 500 ft divided into 10 blocks of equal length (50 ft each)
Total thickness of reservoir = 50 ft
Number of layers = 25
Thickness of each layer = 2 ft
Depth to top of reservoir = 1000 ft
VDP range = 0.6 – 0.9
Injection rate = 250 STB/day
Oil Formation Volume Factor = 1.100 RB/STB
Oil Density = 54.31 lb/ft3
Oil Viscosity = 0.95 cp
Water Formation Volume Factor = 1.0034 RB/STB
Water Density = 62.998 lb/ft3
Water Viscosity = 0.96 cp
Number of wells = 2; one injector and one producer
8 SPE 87016

Table 3–Permeability & Porosity Table 4–Permeability & Porosity

Values with VDP = 0.6 in Values with VDP = 0.7 in
Descending Order Descending Order
Layer k, mD φ, fraction Layer k, mD φ, fraction
1 0.72 0.008 1 0.29 0.009
2 2.02 0.03 2 0.98 0.01
3 4.01 0.06 3 2.20 0.04
4 6.88 0.08 4 4.12 0.06
5 10.84 0.09 5 7.00 0.08
6 16.20 0.11 6 11.18 0.10
7 23.36 0.12 7 17.14 0.11
8 32.86 0.13 8 25.52 0.13
9 45.38 0.15 9 37.18 0.14
10 61.83 0.16 10 53.35 0.15
11 83.48 0.17 11 75.73 0.16
12 112.07 0.18 12 106.76 0.18
13 150.00 0.19 13 150.00 0.19
14 200.92 0.20 14 210.97 0.20
15 269.85 0.21 15 297.63 0.21
16 364.46 0.22 16 422.64 0.23
17 496.64 0.23 17 606.42 0.24
18 685.56 0.24 18 883.28 0.25
19 963.58 0.26 19 1313.86 0.27
20 1389.96 0.27 20 2014.57 0.28
21 2080.51 0.28 21 3225.56 0.30
22 3284.55 0.30 22 5495.83 0.32
23 5640.43 0.32 23 10330.18 0.34
24 11261.37 0.35 24 23153.25 0.37
25 32038.90 0.38 25 78518.16 0.42

Table 5–Permeability & Porosity Table 6–Permeability & Porosity

Values with VDP = 0.8 in Values with VDP = 0.9 in
Descending Order Descending Order

Layer k, mD φ, fraction Layer k, mD φ, fraction

1 0.12 0.007 1 0.05 0.007
2 0.48 0.009 2 0.23 0.009
3 1.20 0.01 3 0.66 0.01
4 2.46 0.04 4 1.47 0.02
5 4.52 0.06 5 2.92 0.05
6 7.72 0.08 6 5.32 0.07
7 12.57 0.10 7 9.22 0.09
8 19.82 0.12 8 15.39 0.11
9 30.47 0.13 9 24.97 0.12
10 46.03 0.15 10 39.72 0.14
11 68.69 0.16 11 62.31 0.16
12 101.71 0.18 12 96.89 0.17
13 150.00 0.19 13 150.00 0.19
14 221.52 0.20 14 232.60 0.21
15 328.29 0.22 15 362.11 0.22
16 490.14 0.23 16 568.43 0.24
17 740.49 0.25 17 904.24 0.25
18 1138.06 0.26 18 1466.39 0.27
19 1791.50 0.28 19 2442.83 0.29
20 2919.90 0.30 20 4232.16 0.31
21 5001.03 0.32 21 7754.26 0.33
22 9196.66 0.34 22 15391.23 0.36
23 18922.21 0.36 23 34666.61 0.39
24 47616.10 0.40 24 97955.21 0.42
25 192565.58 0.45 25 472635.80 0.48
SPE 87016 9

Table 7–Permeability values with noise

No. Noise_1 Noise_2 Noise_3 Noise_4 Noise_5

1 0.11 0.12 0.10 0.13 0.10
2 0.47 0.46 0.50 0.45 0.49
3 1.18 1.20 1.19 1.22 1.19
4 2.33 2.67 2.29 2.64 2.32
5 4.06 4.10 4.01 5.01 5.15
6 9.32 8.46 7.59 6.41 8.98
7 11.49 15.47 10.65 14.23 14.55
8 17.00 16.32 25.13 23.52 17.06
9 34.13 27.54 29.59 33.78 37.94
10 31.51 55.06 54.16 66.15 60.26
11 58.36 51.34 72.67 104.66 52.50
12 163.13 69.26 132.97 98.64 68.76
13 96.52 101.60 100.23 168.99 224.01
14 147.99 372.36 253.53 375.17 177.46
15 239.54 368.16 268.46 391.03 565.66
16 886.14 530.49 827.45 902.73 332.85
17 1074.30 733.74 1036.31 955.58 1014.80
18 939.60 1974.45 1074.00 1305.15 1022.84
19 1214.02 1341.44 1092.56 2802.85 1197.53
20 2047.55 1916.31 5137.69 1818.01 5106.97
21 5036.97 5580.45 6207.28 5214.82 4223.04
22 4279.05 22866.76 22558.14 14231.46 22713.42
23 24343.54 11647.50 11039.24 33797.79 23396.36
24 35808.70 102088.32 62561.57 28821.71 21446.30
25 188663.14 84578.96 330776.33 61267.35 58894.97

Leading
o Front
M <1

Layer 1 o o o
(High Velocity) λrw1 λrw1 λro1

Layer 2
o o o
(Low Velocity) λrw 2 λro2 λro2

Trailing
Front Flow Direction

Fig. 1–Direction of viscous crossflow for a two-layer model

with end-point mobility ratio of less than unity (M<1).8
10 SPE 87016

0.7 0.7
Oil recovery, fraction of OOIP

Oil recovery, fraction of OOIP

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3
Vdp = 0.6, decreasing upward, with crossflow Vdp = 0.9, decreasing upward, with crossflow
Vdp = 0.6, increasing upward, with crossflow Vdp = 0.9, increasing upward, with crossflow
0.2 0.2
Vdp = 0.6, decreasing upward, no crossflow Vdp = 0.9, decreasing upward, no crossflow
Vdp = 0.6, increasing upward, no crossflow Vdp = 0.9, increasing upward, no crossflow
0.1 0.1
Vdp = 0.02, uniform system Vdp = 0, uniform system
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

PV Injected PV Injected

Fig. 2–Effects of Permeability Trend with VDP = 0.6 as Compared to Fig. 5–Effects of Permeability Trend with VDP = 0.9 as Compared to
Systems of Uniform Permeability and No Crossflow. Systems of Uniform Permeability and No Crossflow.

0.7 0.7

0.6 0.6

Oil recovery, fraction of OOIP

Oil recovery, fraction of OOIP

0.5 0.5

0.4 0.4

Vdp = 0.6, increasing upward

0.3
0.3 Vdp = 0.7, decreasing upward, with crossflow Vdp = 0.7, increasing upward
Vdp = 0.7, increasing upward, with crossflow Vdp = 0.8, increasing upward
0.2
0.2 Vdp = 0.7, decreasing upward, no crossflow Vdp = 0.9, increasing upward
Vdp = 0.7, increasing upward, no crossflow
0.1 Vdp = 0.02, uniform system
0.1 Vdp = 0 uniform system

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

PV Injected PV Injected

Fig. 3– Effects of Permeability Trend with VDP = 0.7 as Compared Fig. 6–Effects of Varied VDP with Increasing Upward Permeability
to Systems of Uniform Permeability and No Crossflow. Trend as Compared to a System of Uniform Permeability.

0.7 0.7
Oil recovery, fraction of OOIP

0.6 0.6

0.5 0.5
Oil recovery, fraction of OOIP

0.4 0.4

0.3 0.3
Vdp = 0.8, decreasing upward, with crossflow Vdp = 0.6, decreasing upward
0.2 Vdp = 0.8, increasing upward, with crossflow Vdp = 0.7, decreasing upward
0.2
Vdp = 0.8, decreasing upward, no crossflow Vdp = 0.8, decreasing upward
0.1 Vdp = 0.8, increasing upward, no crossflow Vdp = 0.9, decreasing upward
0.1
Vdp = 0, uniform system Vdp = 0.02, decreasing upward
0
0 1 2 3 4 5 6 7 8 0
0 1 2 3 4 5 6 7 8
PV Injected PV Injected

Fig. 4–Effects of Permeability Trend with VDP = 0.8 as Compared to Fig. 7–Effects of Varied VDP with Decreasing Upward Permeability
Systems of Uniform Permeability and No Crossflow. Trend as Compared to a System of Uniform Permeability.
SPE 87016 11

0.5 0.7

Oil recovery, fraction of OOIP

0.6
Oil recovery, fraction of OOIP

0.4
0.5

0.3 0.4

0.3
0.2
(kv/kh) = 0.1, Vdp = 0.8, decreasing upward 0.2 Vdp = 0.772, Noise_2, decreasing upward
(kv/kh) = 0.05, Vdp = 0.8, decreasing upward
Vdp = 0.772, noise_2, increasing upward
0.1 (kv/kh) = 0.15, Vdp = 0.8, decreasing upward 0.1
(kv/kh) = 0.001, Vdp = 0.8, decreasing upward
(kv/kh) = 0, Vdp = 0.8, decreasing upward 0
0 0 1 2 3 4 5 6 7 8

0 1 2 3 4 5 6 7 8 PV Injected
PV Injected

nd
Fig. 8–Effects of Crossflow Magnitude with Decreasing Upward Fig. 11–Effect of Realization, 2 Noise.
Permeability Trend.

0.7
0.6

Oil recovery, fraction of OOIP

0.6
0.5
0.5
Oil recovery, fraction of OOIP

0.4
0.4

0.3 0.3

0.2
0.2
Vdp = 0.8, kv/kh = 0.1, increasing upward. Vdp = 0.791, Noise_4, decreasing upward
Vdp = 0.8, kv/kh = 0.001, increasing upward Vdp = 0.791,Noise_4, increasing upward
0.1
Vdp = 0.8, kv/kh = 0.15, increasing upward
0.1
0
0 1 2 3 4 5 6 7 8
0
0 1 2 3 4 5 6 7 8 PV Injected
PV Injected

th
Fig. 13–Effect of Realization, 4 Noise.
Fig. 9–Effects of Crossflow Magnitude with Increasing Upward
Permeability Trend.

0.7 0.7
Oil recovery, fraction of OOIP

Oil recovery, fraction of OOIP

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2 Vdp = 0.796, Noise_1, decreasing upward 0.2

Vdp = 0.821, Noise_5, decreasing upward
Vdp = 0.796, Noise_1, increasing upward
0.1 0.1 Vdp = 0.821, Noise_5, increasing upward

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

PV Injected PV Injected

st th
Fig. 10–Effect of Realization, 1 Noise. Fig. 14–Effect of Realization, 5 Noise.
12 SPE 87016

0.7
Oil recovery, fraction of OOIP

0.6

0.5

0.4

0.3
Vdp = 0.75, coarsening upward, simulator
0.2 Vdp = 0.75, coarsening upward, correlation
Vdp = 0.75, fining upward, simulator
0.1
Vdp = 0.75, fining upward, correlation

0
0 1 2 3 4 5 6

PV Injected

Fig. 15–Comparison Between Simulator Results and

Correlation Results.

1000
Oil production rate, STB/day

900
Oil production rate, correlation, decreasing upward
800
Oil production rate, actual history
700
Oil production rate, correlation, increasing upward
600

500

400

300

200

100

0
9800 10000 10200 10400 10600 10800 11000

Cumulative time, days

Fig. 16–Comparison Between Correlation Results and Historical

Production Data.

k1 water oil

k2 water oil

k3 water oil

k1 water oil

k2 water oil

k3 water oil

Fig. 17–Crossflow in increasing upward (k1>k2>k3, top) and

decreasing upward (k1<k2<k3, bottom) permeability
trend formations.