Proc.

Proc.Schl.
Schl.Eng.
Eng.Tokai Univ.,
Tokai Ser. E
Univ., Ser. E
38 (2013）■-■
38 (2013) 9-14

Numerical Simulation of the Buckley-Leverett Problem

by

Abidullah ARABZAI *1 and Shigeo HONMA *2

(Received on Mar. 30, 2013 and accepted on May 16, 2013)

Abstract
The displacement of oil by water in a petroleum reservoir is analyzed via the famous Buckley-Leverett
displacement theory. A quantitative demonstration is performed for a typical relative permeability data and
reservoir condition in reference to the literature by Aziz and Settari. The immiscible displacement process is
then analyzed numerically by the finite difference method and the finite element method, in which the flow
equations of the two-phase liquid are transformed into a single–valued pressure equation, i.e. capillary pressure
is neglected, and water saturation equation. The pressure equation is solved implicitly, and the saturation
equation is solved explicitly with a special attention to the time step level on the pressure-saturation evaluation.
Numerical simulations developed in this paper show good correlation with the results by the Buckley-Leverett
analysis.

Keywords: Petroleum Reservoir Engineering, Waterflooding, Two-phase Flow, Immiscible Displacement,

Buckley-Leverett Theory, Finite Difference Method, Finite Element Method

Qo
1. Introduction
Production
well
In petroleum reservoir engineering, a technique of
injecting water into reservoir has been used in order to
Oil
maintain oil production rates during the pumping operation.
Petroleum
The method is known as the waterflooding technique, which reservoir
Qw L
provides high oil production rates and high degree of Injection
well
petroleum recovery when oil production rates deteriorate1). x Water
Transition
When water is injected into reservoir, oil is displaced zone

toward the production well in the situation of two-phase flow A

as illustrated in Fig.1. Oil and water are immiscible in each
other, so that this phenomenon is referred to the immiscible
displacement in porous media. The mechanism of immiscible
displacements of two-phase fluids was studied extensively
and a number of technical papers have been published. The
special case of one-dimensional, incompressible, two-phase
flow was theoretically investigated first by Buckley and
Leverett2) in 1942.
The Buckley and Leverett frontal displacement theory
described a method for calculating saturation profiles when
the effects of capillary pressure and gravity are neglected.
However, the theory gained popularity in the petroleum
engineering because of its simplicity and accuracy 1).
Since appearance of the theory, numerical analysis was
eagerly attempted to simulate the Buckley-Leverett problem.
*1 Research Student, School of Engineering
*2 Professor, Department of Civil Engineering
Fig.1 Waterflooding method for oil production.
The early numerical methods were simultaneous solution
method (SS) and implicit pressure-explicit saturation solution
(IMPES) by Finite Difference Method 3). The analysis was
then extended to the case which includes the effect of gravity,
capillary pressure, fluid compressibility and aquifer
compressibility. Through the numerical analysises by FDM, it
was found that, as time progress, the saturation becomes a
multiple-valued function of the grid spacing4). The Finite
Element Method was also developed to the analysis of flow
through porous media, and applied to the two-phase flow
problems. There are common numerical difficulties in FDM
and FEM that reverse changes in mobility factors in water
and oil phases appear in the numerical computation process,
which results in a convection-dominated situation 5). The

―1―
XXXVIII,
Vol. Ⅹ 2013
ⅩⅩⅧ, 2013
−9−
 Abidulla ARABZAI and Shigeo HONMA
Abidullah ARABZAI and Shigeo HONMA

upstream weighting for mobility factors is required to avoid 1   p

these difficulties in both FDM and FEM 6,7).  ( Qw w  Qo o )   o / w ( 4)
kx A k rw k ro x
In this paper, Buckley-Leverett analysis is performed first
based on the relative permeability data employed by Azis and From the relation of Qo+ Q w = QT, substitution of Qo= QT
4)
Settari , and the finite difference and the finite element Q
- w into Eq.(4) yields
analysis are attempted using the same physical data adopted
k ro  w k k A  po / w
in the Buckley-Leverett analysis. The numerical analysis is QT  Qw (1  )  x ro (5)
k rw o o  x
basically based on the method of IMPES, but special
attention has been paid to the time step level in the pressure from which the fraction of pore water flow, i.e., fractional
change in the numerical formulation. flow rate fw, is expressed as

k x k A  po / w
2. Buckley-Leverett Analysis 1  Q ro
Qw T o x
f w  ( 6)
QT k 
In the first, the Buckley-Liverett frontal displacement 1  ro w
krw o
theory is reviewed, and an example of the analysis is
presented. The flow rate of oil and water through The capillary pressure po/w is actually very small compared
completely saturated porous medium in horizontal direction to the water pressure when water displaces oil in the reservoir.
is given by the Darcy’s law as follows. If the effect of capillary pressure is neglected, Eq.(6)
becomes the following simple expression.
k x kro A  po k x krw A  pw
Qo   , Qw   (1)(2)
o  x w  x 1
fw (7 )
k ro  w
where k x is the intrinsic permeability of the medium, A is the 1
k rw  o
cross-sectional area for permeation,  o and  w are the
dynamic viscosity of oil and water, p o and pw are the pore Next, continuity equation of pore water is expressed as
pressure of oil and water, and kro and krw are the relative
 Qw S w
permeability of oil and water respectively. The relative   A (8)
x t
permeability kro and k rw are generally given as a function of
water saturation Sw as shown in Fig.2, where Swr and S or are where φ is the porosity of the porous medium. Using the
the residual water and oil saturation. The pressure difference relations of Qw = fw QT and fw (Sw), Eq.(8) is rewritten as
at the contact surface between oil and water, i.e. capillary
pressure, is denoted by po/w, then pw of Eq.(2) is rewritten as S w Q d f S
( )   T ( w w )t (9 )
t x A d S w  x
k x krw A  ( po  po/w )
Qw   (3) Equation (9) is a first-order nonlinear hyperbolic equation.
w x
The main and very fruitful idea of Buckley and Leverett was
and subtracting Eq.(1) from Eq.(3) yields to transform Eq.(9) into the following form:

So x Q df
1 0.5 0 ( )S  T ( w )t (10 )
1.0 1.0 t w A d S w

Relative permeability
Buckley-Leverett
stating that the rate of advance of a plane of fixed saturation
0.8 0.8
datacapillary pressure
and po/w Sw is proportional to the rate of change in composition of the
k ro , k rw Approximated
(kPa)
0.6 0.6
flowing stream with saturation. As f w is not an explicit
function of t, Eq.(10) can be integrated to give the position of
0.4 p 0.4 a particular saturation as a function of time:
o/w
k rw
S wr S or QT t d f w
0.2 k ro 0.2
xS w  ( )  x0 (11)
A d S w

0 0 where x0 is the position of the water saturation at time t=0.

0 0.2 0.4 0.6 0.8 1
Sw According to Eq.(11), each saturation advances into the
Fig.2 Relative permeabilities and capillary pressure system at a rate in direct proportion to f wʼ= d fw /dS w. The
with saturation4). shape of the saturation profile calculated by Eq.(11) is exp-

―2― Proceedings
Proceedingsofofthe
theSchool
School of Engineering
Engineering,
Tokai
TokaiUniversity,
University, Series
Series E
− 10 −
 Numerical Simulation of the Buckley-Leverett Problem
Numerical Simulation of the Buckley-Leverett Problem

Sw Table1 Relative permeabilities and fractional flow function

for Buckley-Leverett problem.(after Aziz and Settari4) )
S w =1
d S or k ro 1 '
1-S or Sw k rw k ro fw  1   f w  d fw / d S w
fwʼ k rw
SBL c 0.00 0.000 0.900 0.000
0.000
A b 0.16 0.000 0.900 0.000
0.425
0.20 0.012 0.675 0.017
A=B 0.680
B 0.25 0.025 0.465 0.051
a
f S wr 1.260
x 0.30 0.040 0.310 0.114
1.880
0.35 0.055 0.210 0.208
Fig.3 Tentative Saturation profile. 3.020
0.40 0.070 0.125 0.359
3.780
0.45 0.085 0.070 0.548
ressed by the curve abcd in Fig.3, but it does not display 4.040
0.50 0.105 0.035 0.750
1.740
sharp leading edge for the saturation front. 0.55 0.128 0.025 0.837
1.020
Morel-Seytoux8) applied the conservation of mass over 0.60 0.158 0.020 0.888
0.780
0.65 0.190 0.015 0.927
the front position to be A=B, and explained the saturation 0.660
0.70 0.240 0.010 0.960
established in the flowing system immediately behind the 0.480
0.75 0.310 0.005 0.984
front SBL can be evaluated from the tangent point c on the 0.320
0.80 0.410 0.000 1.000
1.00 0.410 0.000 1.000 0.000
fractional flow curve shown in Fig.4. The abrupt front in the
saturation profile is then expressed by cf in Fig.3. Residual water saturation S wr = 0.16
Residual oil saturation S or = 0.20
So
1.0 0.8 0.6 0.4 0.2 0
1.0
d
1.0 qTT = 1 m3 /day 0
fw φ = 0.2 , Q
c A = 100 m22, L = 100 m
0.8
1- Sor Sor
krw,k ro
f w 0.6 Oil
kro b at fw
ʼ max
max S BL
S w 0.5 0.5 So

800 days
200

300
t =100

400

500

600

700
0.4
S wr krw S or Water
0.2
S wr 1- Swr

0 a
0 0.2 0.4 S BL 0.6
Sw
Fig.4 Saturation at the sharp front on f w curve .
As a quantitative demonstration for the Backley-Leverett
analysis, relative permeability data shown in Table1 was used.
These data are exactly same as the curves illustrated in Fig.2
and Fig.4 in which the residual water saturation is S wr=0.16
and the residual oil saturation Sor=0.20. The corresponding
fractional flow rates fw and their derivatives fw’= dfw /dSw are
also shown in Table1. It was assumed that the viscosity of
wataer and oil are identical for simplicity. The value of
frontal saturation S BL was obtained by the graphic method,
and found to be SBL=0.53.
In a situation of reservoir porosity φ=0.2, cross-sectional
area for permeation A=100m2, reservoir length L=100m and
amount of water injected into reservoir QT =Qw=1m /day, the
calculated results of saturation profile by Buckley-Leverett
analysis is shown in Fig.5. It is seen from the figure that the
saturation front progresses with a constant speed toward the
outlet, however a considerable amount of oils still remains in
t he reservoir. This is attributed from the characteristics of the
0 1.0
0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
x/L
8) Fig.5 Calculated results of saturation profile by
Buckley-Leverett analysis.
relative permeabilities employed in the analysis. When
injected water front reached to the outlet, i.e. at breakthrough,
total amount of undisplaced oil is estimated to be 27% to the
entire mobile oil in the reservoir. There is a method9) to
calculate the oil production during displacement and the oil
recovery after breakthrough in the Backley-Leverett analysis,
but the method is omitted here.
3. Numerical Simulation
Next, the Buckley-Leverett oil displacement problem is
3
solved via numerical methods. For convenience, as in the case
of Buckley-Leverett analysis, the effects of capillary pressure,
and gravity forces have been neglected. The same reservoir
condition and physical properties are employed in both Finite
Difference Method and Finite Element Method.

―3―
Vol.Ⅹ
Vol. XXXVIII, 2013
ⅩⅩⅧ, 2013
− 11 −
 Abidulla ARABZAI and Shigeo HONMA
Abidullah ARABZAI and Shigeo HONMA

3.1 Finite Difference Method where

The finite difference method has a long history and has
Ai   ( wi  oi ) , Bi  Ai  Ci
been applied to various problems of flow through porous
C i   ( wi 1  oi 1 ) , Di  0 (21a, b, c, d )
media. The equations of flow for water and oil in an
isotropic-homogenous incompressible medium are written as in which   k x t /  x 2. The mobility factorsλo/w are norma-

 pw S lly evaluated at the upstream side nodes as shown in Fig.6, i.e.


( kx  w )  w (12) upstream weighting, to obtain stable solutions. If λo /w are
x x t
 p
 o S taken at the intermediate points, i.e. midstream weighting, the
( kx o )  o (13)
x x t coefficients in Eq.(20) are written as

Ai   ( wi 1/ 2  oi  1/ 2 ) , Bi  Ai  Ci
where  w  krw /  w and  o  kro /  o are mobility factors, and
pw and po are fluid pressures in each phase. There exists the C i   ( wi 1/ 2  oi 1/ 2 ) , Di  0 (22a, b, c, d )
following axially relationships for saturation and pressure;
Flow direction
S w  So  1 , po  pw  po/w (14 a , b ) o/ w i 1/ 2  o/ w i 1/ 2 Midstream
Upstream o/ w i 1  o/ w i
If the capillary pressure p o/ w is neglected, i.e. po= pw , the
water and oil pressures are replaced by a single variable p , i-1 i i+1
Δx Δx
and Eqs.(12)(13) are rewritten as
Fig.6 Evaluation points of the mobility factor.
 p S
( kx  w )  w (15)
x x t
Equation (20) can be solved implicitly associated with the
 p S
( kx o )   w (16) boundary conditions of (18a,b) using well-known tridiagonal
x x t
matrix solver 6). Once pressure is computed, water saturation
By adding the water and oil phase equations of Eq.(15) and can be calculated explicitly as follows.
Eq.(16), the equation of two-phase flow is described as 
S wik 1   p k 1 / 2  ( wi   wi 1 ) pik 1 / 2
 wi i 1
 p  p
(k x  w )  (k x  o ) 0 (17)   wi 1 pik11 / 2   S wik ( 23)
x x x x
Owing to high nonlinearities in the pressure and saturation
Now, time derivative term of saturation has been eliminated,
equations, calculations of Eqs.(20)(23) must be iterated until
Eq.(17) is called the pressure equation. It should be
successive changes of pi and S wi are settled in prescribed
k 1/2 k 1
emphasized that Eq.(17) is not a steady state equation but a
tolerances. Picard method and/or Newton-Raphson method
transient equation with respect to pressure change, because
can be employed to accelerate the convergence5).
the mobility factors undergo the change of water saturation.
The boundary conditions may be given by
t = 100 (a)
^
 p QT p p ^ 400
 kx  w  at x  0 , at x  L (18a, b) 800 days
x A Δ x =2m
Δ t = 1day
where p^ is the prescribed pressure at the outlet. Hereupon,
the pressure equation (17) is transformed to the finite
difference equation, and the resulting algebraic equation can
be written in general form as
1- S or = 0.8 φ = 0.2 , QT = 1m3 /day (b)
A = 100m2 , L = 100m
 Ai pi1  Bi pi
k  k 
 Ci pi1  Di
k 
(19)

Ifθ is chosen to be 1, Eq.(19) becomes fully implicit scheme. t = 100 200 300 400 500 600 700 800days

In this study, the time step level for the computation of p was
S wr = 0.16
selected at k +1/2 because pressure change in flow equations
(15)(16) should be evaluated between k and k +1. Thus,

 Ai pik11/ 2  Bi pik 1/ 2  Ci pik11/ 2  Di (20) Fig.7 Calculated results of the Buckley-Leverett problem by
FDM; (a) Pressure profile, (b) Saturation profile.

―4― Proceedings
Proceedingsofofthe
theSchool
School of Engineering
Engineering,
Tokai
TokaiUniversity,
University, Series
Series E
− 12 −
 Numerical Simulation of the Buckley-Leverett Problem
Numerical Simulation of the Buckley-Leverett Problem

In the numerical analysis, it is convenient to make a
functional approximation for relative permeabilities, and the
following forms of approximation were used in the simulation.
5 .5
krw 0.23Se  0.18Se
4.8
kro  0.86 (1 Se )  0.04 (1  Se )
In the FD analysis developed here, the upstream weighting
for mobility factors were introduced. If midstream weighting
of Eqs.(22a,c) are used, an unrealistic saturation profile
comes out as shown in Fig.9. Therefore, the upstream
( 24a ) weighting for mobility factors is inevitably required for the
( 24b) two-phase flow simulation by finite difference method.

where S e is the effective saturation defined by 3.2 Finite Element Method

S w  S wr
Se 
1  Swr  Sor
The calculated results by FDM under the same reservoir
condition and physical parameters are shown in Fig.7. The
grid spacing and time step size selected were ⊿ x=2m and ⊿ t
=1day. The numerical results showed good coincidence with
the results by the Buckley-Leverett analysis previously
shown in Fig.5, but a smearing of the displacement front
appears. This can be improved by the use of finer grid
spacing as illustrated in Fig.8 and Fig.9.
For the finite element formulation of the boundary value
( 24c ) problems, the weighted residual method is frequently
employed. Here, the finite element formulation for the
Buckley-Leverett problem is performed via Galerkin finite
element procedure.
Application of the Galerkin’s criterion to Eq.(17) leads to
 p  p
R N  x ( k x  w x )  x ( k x  o x )  dR  0
I ( 25 )
where NI is the linearly independent weighting function.
Application of the Green’s theorem to Eq.(25), it follows

NI p p
R x
( k x  w  k x  o ) dR 
x x
B N q dB  0
I ( 26 )
t = 100 (a) Pressure
200 profile where q is the outward normal flux on the boundary.
300
400 days Introducing the following form of a trial function

p ( x, t )  N J ( x) pJ (t ) (27)

and subdividing the region R into m finite elements, Eq.(26)

is written in matrix form as
Δ x =1m (b) Saturation m e m
 E  p   R
k 1 / 2 e
Δ t =1day profile (28)
i 1 i 1

t = 100
in which the elements E  and R  are given by
200 300 400days e e

e NI NJ
EIJ   e k x ( w  o ) dR (29a)
R x x

R I   e NI  (q k 1  q k ) / 2  dB
e
Fig.8 Improvement of the smearing of sharp front with the ( 29b)
B
use of finer grid spacing.
In the similar manner, finite element formulation for
Eq.(15) is performed via Galerkin method.
1.0
t = 400 days  p S w
1- Sor Δt = 1day
R N  x ( k x  w
I
x

t
)  dR  0 ( 30)
Buckley-Leverett
Midstream Δx = 2 m By Green’s theorem
S w 0.5 Upstream Δx = 2 m
Upstream Δx = 1 m NI p S w
Water
Oil R x
( k x  w ) dR 
x B N q dB  R N 
I I
t
dR  0
(31)
S wr
Introducing the following form of trial functions
0
0 0.2 0.4 0.6 0.8 1.0 p ( x, t )  N J ( x) pJ (t ) , S w ( x, t )  NJ ( x) S wJ (t )
x/L (32a, b)

Fig.9 Comparison of saturation front by weighting and subdividing the region R into m finite elements, Eq.(30)
of mobility factors and grid spacing.
is written in matrix form as

―5―
Vol.Ⅹ
Vol. XXXVIII, 2013
ⅩⅩⅧ, 2013
− 13 −
 Abidulla ARABZAI and Shigeo HONMA
Abidullah ARABZAI and Shigeo HONMA

m e m
 F  S w   Q
e
k 1
(33) was performed based on the Buckley-Leverett frontal
i 1 i 1
displacement theory and the numerical methods. The major
in which the elements F  and Q are given by
e e conclusions obtained through this study are as follows:

e 
FIJ   e NI NJ dR (34a) 1) The Buckley-Liverett equation is a first-order hyperbolic
R t
equation with respect to saturation change, and which
 NI NJ k 1/ 2
dR ) pJ exhibits highly convection-dominated situation. The advance
k
 NI NJ dR ) S wJ  (  e k x  w
e
QI  (
R
e
t R x x
of saturation front is controlled by the derivative of fractional
  e NI q k 1/ 2dB (34b) flow rates to water saturation. The quantitative demonstration
B
revealed a constant progress of abrupt leading edge for the
Element matrix of Eq.(34a) is the capacitance matrix, and saturation front.
integral on the second term of Eq.(34b) is the conductance 2) The simultaneous two-phase flow equations can be
matrix. Experience indicates that the capacitance matrix transformed to a pair of pressure equation and saturation
should be lumped by diagonalizing the element components equation, under the assumption of neglecting capillary
to eliminate numerical oscillation. The upstream weighting pressure. The equations can be solved successively with
for mobility factors was also introduced. The matrix IMPES (Implicit Pressure-Explicit Saturation) scheme by
equations (28) and (33) are solved using an appropriate finite difference method with an aid of upstream weighting
solver until successive changes of pik 1/2 and S wik 1 are settled in

for the mobility factors.
the prescribed tolerances. 3) The Buckley-Leverett problem can be solved by finite
Figure 10 illustrates the calculated results by FEM up to t = element method using Galerkin finite element formulation of
400days. Pressure and saturation profiles are the same as FD pressure equation and saturation equation subsequently.
solutions previously shown in Fig.7. Upstream weighting is also required, and the results showed
good agreement with the Backley-Leverett solution.
t = 100 (a)
200 References
300
Δ x =2m
400 days
Δ t = 1day
1) Craft, B.C. and Haukins, M., Revised by Ronald E.Terry
(1991), Applied Petroleum Reservoir Engineering,
Prentice Hall, pp.1-6.
2) Buckley, S.E. and Leverett, M.C. (1942), Mechanism of
(b) Fluid Displacement in Sands, Transactions AIME,
Vol.146, pp.107-116.
Δ x =2m , Δ t =1day
3) Richardson, J.G. and Stone, H.L. (1973), A Quarter
t = 100 200 300 400days Century of Progress in the Application of Reservoir
Engineering, Journal of Petroleum Technology,
pp.1371-1379.
4) Aziz, K. and Settari,A. (1979), Petroleum Reservoir
Simulation, Applied Science Publishers.
Fig.10 Calculated results of the Buckley-Leverett problem by
FEM; (a) Pressure profile, (b) Saturation profile. 5) Huyakorn, P.S. and Pinder, G.F. (1983), Computational
Methods in Subsurface Flow, Academic Press, pp.172-179.
6) Roache, P.J. (1980), Computational Fluid Dynamics,
Through the numerical simulations by finite difference and Hermosa Publishers, pp.345-349.
finite element methods for oil displacement phenomena, we 7) Honma, S. and Karadi, G.M. (1986), Determination of
can obtain not only the saturation profile but also the pressure Optimal Upstream Weighting Parameter for the Finite
profile in the reservoir, which provide us useful information Element Solution of Transient Transport Equation, Z.
for the waterflooding operation in the petroleum production angew. Math. Mech., Vol.66, No.10, pp.465-469.
technology. 8) Morel-Seytoux, H.J. (1969), “Flow through Porous Media”
(R.J.M. de Wiest ed.), Academic Press, pp.456-516.
3. Conclusion 9) Dullien, F.A.L. (1979), Porous Media, Fluid Transport
and Pore Structure, Academic Press, pp.303-309.
A quantitative analysis for the displacement of oil by water

―6― Proceedings
Proceedingsofofthe
theSchool
School of Engineering
Engineering,
Tokai
TokaiUniversity,
University, Series
Series E
− 14 −