You are on page 1of 21

HYDROLOGICAL PROCESSES

Hydrol. Process. 21, 534– 554 (2007)


Published online 23 August 2006 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/hyp.6264

Determination of Forchheimer equation


coefficients a and b
Melina G. Sidiropoulou, Konstadinos N. Moutsopoulos and Vassilios A. Tsihrintzis*
Laboratory of Ecological Engineering and Technology, Department of Environmental Engineering, School of Engineering, Democritus University of
Thrace, 67100 Xanthi, Greece

Abstract:
This study focuses on the determination of the Forchheimer equation coefficients a and b for non-Darcian flow in porous
media. Original theoretical equations are evaluated and empirical relations are proposed based on an investigation of available
data in the literature. The validity of these equations is checked using existing experimental data, and their accuracy versus
existing approaches is studied. On the basis of this analysis, some insight into the physical background of the phenomenon
is also provided. The dependence of the coefficients a and b on the Reynolds number is also detected, and potential future
research areas, e.g. investigation of inertial effects for consolidated porous media, are pointed out. Copyright  2006 John
Wiley & Sons, Ltd.
KEY WORDS Forchheimer equation; experimental data analysis; phenomenological coefficients; inertial effects; non-Darcian
flow

Received 16 May 2005; Accepted 22 November 2005

INTRODUCTION Re D qD/ > 10, with D [L] the porous medium particle
diameter and  [L2 T1 ] the kinematic viscosity of the
The classical assumption for the description of a large
fluid), the macroscopic hydraulic behaviour is described
number of problems related to flow in porous media is
by the Forchheimer law:
that, at the microscopic scale, a creeping flow takes place,
which, at the macroscopic scale, is equivalent to a linear
relationship between the flow rate Q and the piezometric rh D aq C bjqjq 2
head h, expressed as
dh The coefficient a [TL1 ] of the linear term in the Forch-
Q D KA 1a
dx heimer equation (Equation (2)) depends on the properties
or equivalently as of both the porous medium and the fluid. It represents
energy losses due to viscous forces (viscous friction)
dh at the fluid–solid interface and is equal to 1/K, where
q D K 1b
dx K is the hydraulic conductivity. Coefficient b [T2 L2 ]
Equation (1a) is the well-known Darcy law, where depends on the properties of the porous medium only. It
A [L2 ] is the cross section of the porous medium, is related to inertial forces, which are irrelevant to vis-
K [LT1 ] is the hydraulic conductivity (which depends cous forces. Although, theoretically, Equation (2) is more
on porous medium and fluid properties), q [LT1 ] is the appropriate to simulate the flow processes in any porous
Darcy velocity (defined as the mean velocity in a rep- medium, for simplicity, in practice, its use is limited to
resentative elementary volume), h [L] is the piezometric coarse granular porous media (for illustrative examples,
head, and x is the flow direction (for unidirectional flow). see Moutsopoulos and Tsihrintzis (2005)), fractured or
For non-unidirectional flow, and for isotropic and karstified aquifers.
homogeneous porous media, the following general Numerous analytical solutions, numerical methods and
expression can be used: software packages are available for the simulation of
Darcy flows. Similar tools are also available for the sim-
q D Krh 1c ulation of non-linear inertial flows, although restricted in
number (e.g. Volker, 1975; Zissis and Terzidis, 1991; Wu,
For situations where the inertial effects in the pore
2002a,b; Terzidis, 2003; Moutsopoulos and Tsihrintzis,
scale are not negligible (i.e. in practice, Reynolds number
2005). Their use, however, requires knowledge of the
phenomenological coefficients a and b of Equation (2).
* Correspondence to: Vassilios A. Tsihrintzis, Laboratory of Ecological Various studies have suggested expressions for a and
Engineering and Technology, Department of Environmental Engineering,
School of Engineering, Democritus University of Thrace, 67100 Xanthi, b. For example, Ward (1964) analysed experimental data
Greece. E-mail: tsihrin@otenet.gr of 20 different porous media and suggested the following

Copyright  2006 John Wiley & Sons, Ltd.


FORCHHEIMER EQUATION COEFFICIENTS 535

equations for the estimation of a and b: The above are typical examples of the equations avail-
able in the literature for evaluating the Forchheimer coef-

aD 3a ficients a and b. They are based on assumptions and
gk simplifications of the geometry of the pore space. Conse-
360 quently, these equations have varying degrees of accuracy
aD 3b in their application, depending also on the number and
gD2
quality of data used to derive them.
10Ð44
bD 3c The purpose of this article is to compare various
gD existing equations predicting Forchheimer coefficients a
and b with experimental data available in the literature,
where D [L] is the particle diameter, g [LT2 ] is the
and to propose alternative ways of estimating a and b. An
acceleration due to gravity, and k [L2 ] is the permeability
investigation of the physics of the phenomena examined
of the porous medium given by the following equation
is also performed in an attempt to derive a theoretical
k D D2 /360. equation. Finally, descriptions of several features of the
Blick (1966) considered a mixed model of a bundle of phenomenon, not previously referred to, are presented.
parallel capillary tubes with orifice plates spaced along
each tube, and proposed the following relations:
32 METHODS AND MATERIALS
aD 3d
gnD2 Theoretical background and proposed relations
CD The relations presented above (Ergun’s approach,
bD 3e
2gn2 D Equations (4a) and (4b), or a similar relation suggested
by Ahmed and Sunada (1969)) assumed that the energy
where CD is an appropriate phenomenological coefficient. losses depend solely on the size of the pore gaps (or
Ergun (1952), referred to in Bear (1979), extended equivalently on the grain diameter). The shape of the
the Kozeny–Carman model, originally developed for pore space is not taken into account. The assumption
creeping flows, and suggested the following expressions: behind the development of Ergun’s equations (Equations
(4a) and (4b)), i.e. that the pore space can be simulated by
1501  n2 circular pipes, is not compatible with the energy balance
aD 4a
gn3 D2 of the flow, for which the characteristics for the case of
1Ð751  n circular conduits and porous media are as follows:
bD 4b
gn3 D ž For high Reynolds number flows in conduits, turbu-
where n is the porosity of the medium. Similar expres- lence is produced near the walls and is transferred to
sions to Ergun’s (1952) were derived by Kovács (1981), the interior of the pipe, where it is transformed to heat
who analysed a set of 300 data in the range of 10 < Re < (Rodi, 1984).
100, and derived the following formulae for the case of ž Inertial flows in porous media are characterized by
homodisperse spherical particles: recirculation zones, which are delimited from the main
area of flow by closed streamlines. In these areas,
144 1  n2 no macroscopic transfer of the fluid particles takes
aD 4c place. As demonstrated by Panfilov et al. (2003), the
gD2 n3
energy for the eddies in these zones is provided by jet
2Ð4 1  n bunches, issuing from the main flow area. By arguing
bD 4d
gD n3 that the energy of these bunches is proportional to
the kinetic energy of the mean flow, Panfilov et al.
The basic assumption of the original Kozeny– (2003) associated the energy losses induced by the
Carman approach and Ergun’s (1952) extensions is that above-mentioned procedure to the quadratic terms of
flow in porous media can be simulated by a bunch of con- the Forchheimer equation.
duits. The computation of the non-linear term is based
on the hypothesis that turbulent flow takes place (Bird Owing to the complexity of the flow, it is obvious
et al., 1960). A similar approach has also been suggested that a description of the hydrodynamic characteristics by
by Ahmed and Sunada (1969). means of numerical simulation can give some insight
Kadlec and Knight (1996) suggested the following into these phenomena. Flow computations in porous
equations for the estimation of coefficients a and b: media, performed by means of conventional numerical
1 2551  n schemes (methods of finite differences and finite ele-
aD D 5a ments), were associated solely with a simple geometry of
K gn3Ð7 D2 the pore space (Latinopoulos, 1980; Coulaud et al., 1988;
21  n Ganoulis et al., 1989; Panfilov and Fourar, 2006). It is
bD 5b
gn3 D obvious that the flow behaviour in the above-mentioned

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
536 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

‘theoretical’ porous media is not identical with that in For Re > 80:
real-world media. However, the simulations provided 83161  n
‘theoretical verification’ of the Forchheimer law and gave aD 6c
useful information concerning inertial flows: gD2
88Ð651  n
bD 6d
ž As is depicted in Coulaud et al. (1988: Figure 6), the gD
influence of the porosity on non-linear head losses is
significant (a result also compatible with the findings 2. Using Equations (6a)–(6d), and considering that the
by Koch and Ladd (1997)). porosity in the porous medium examined is n D 0Ð26,
ž The same simulations demonstrated that the head losses the following equations result.
do not depend solely on the porosity of the medium and For 10 < Re  80:
the Reynolds number, but also on the shape of the pore 4861Ð8
space. aD 7a
gD2
72Ð594
More realistic flow simulations in three dimensions bD 7b
were performed by Hill and Koch (2002), by means gD
of the lattice-Boltzmann method (which makes use of For Re > 80:
the relation between fluid flow and kinetic gas theory),
and also by Fourar et al. (2004). Fourar et al. (2004), in 6153Ð84
aD 7c
their numerical study of high-velocity effects in periodic gD2
porous media, state that viscous dissipation in the recir- 65Ð60
culation area is not preponderant. They state that inertial bD 7d
gD
effects in porous media are mainly caused by deviation of
the streamlines induced by aforementioned recirculation Since a unique configuration of spheres was taken
eddies. Since it is generally accepted that the behaviour into consideration, the accuracy of Equations (6a)–
of real-world, random, porous media can be quite dif- (6d) for different porosity values has to be examined.
ferent from that of artificial ones, their conclusions may 3. An alternative way to estimate the values of the coef-
not be definitive. Other mechanisms related to non-linear ficients a and b is also proposed, by considering, in
energy dissipation cannot be excluded. In their theoreti- addition to the equations developed by Hill and Koch
cal analysis, Skjetne and Auriault (1999) state that inertial (2002), a relation linking the force F (acting on a rigid
energy losses are strongly localized around the boundary object) and the hydraulic head losses h induced by
layer, which induces the flow separation. Anyway, all it (Naudascher, 1987):
three inertia-related mechanisms cited above by Skjetne
and Auriault (1999), Panfilov et al. (2003) and Fourar F D qA? h
et al. (2004) are related to the formation of boundary
where  is the fluid density and A? is the cross-section
layer separation and recirculation eddies; thus, the basic
of flow in which no obstacles are present (Naudascher,
statements of the present study persist:
1987: equation (4Ð1)). In the present work, A? cannot
be defined exactly; therefore, two extreme cases are
1. The hydraulic behaviour in granular porous media is
considered (Figure 1):
essentially different from that of closed pipes.
2. Separation mechanisms of the boundary layer are
(a) By assuming that A? is equal to 2D2 (Figure 1a) and
important and, subsequently, the shape of the particle
that ∂h/∂x D h/l, where l is the distance between
may be crucial for the inertial losses.
two spheres, one obtains the following relations.
Hill and Koch (2002) investigated numerically the
flow processes in a closely packed, face-centred array z z
of spheres. In the present study, their theory was used to
develop relations for both coefficients a and b, as follows:
y y
1. Using their equations (3), (4), (5), (10) and (11),
the following relations for coefficients a and b were
obtained.
For 10 < Re  80:
65701  n (a) (b)
aD 6a
gD2 Figure 1. A projection of the close-packed face-centred cubic unit cell,
with the flow directed along the x-axis (perpendicular to the page) and the
98Ð11  n cross-section of flow (shaded area) A? in which no obstacles are present,
bD 6b
gD assumed to be: (a) equal to 2D2 ; (b) equal to D2 /4

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp
FORCHHEIMER EQUATION COEFFICIENTS 537

For 10 < Re  80: that consolidated porous media might exhibit specific
features, which deserve a special research effort.
1215Ð62
aD 8a A potential guideline for the adequate estimation of
gD2 the non-linear energy losses in fractured or karst aquifers
18Ð15 might be the simulation approach by Lao et al. (2004).
bD 8b
gD They used the pore diameter distribution proposed by
Yanuka et al. (1986) to create artificial porous media,
For Re > 80: composed of straight pipes of cylindrical cross-section
1538Ð60 and random orientation in space. The hydraulic behaviour
aD 8c
gD2 of this network was simulated assuming that in each
pipe the flow is described by the Poiseuille law, and
16Ð39
bD 8d additional head losses were taken into account due to
gD sudden contraction or expansion of the conduit diameter
(b) By assuming that A? is equal to D2 /4 (Figure 1b), and pipe splitting or bending. For these minor losses, the
one obtains the following relations. coefficients proposed by Bird et al. (1960) were used.
For 10 < Re  80: The computation of the flow field of the above-mentioned
network demonstrated that, on a macroscopic scale, the
3097Ð12 flow is described by the Forchheimer equation, where the
aD 9a
gD2 coefficient a is related to energy losses in the straight pipe
46Ð24 sections and the Poiseuille law, and where the inertial
bD 9b coefficient b depends on minor losses that are induced
gD
at pipe junctions. A serious drawback of the approach
For Re > 80: by Lao et al. (2004) is that the use of the relations by
3920Ð20 Bird et al. (1960), which are valid for fully developed
aD 9c turbulence and large Reynolds numbers, is not compatible
gD2
with the use of the Poiseuille law, which is valid for
41Ð79 creeping flow; Lao et al. (2004) could not reproduce the
bD 9d
gD experimental data by Jones (1987).

Owing to the fact that the term A? cannot be esti-


Experimental data from previous studies
mated exactly, one may assume that Equations (7a)–(7d)
are more accurate, although this issue remains open to In this study, the literature was searched and experi-
discussion. Equations (6a)–(9d), however, suggest that mental data were collected from previous studies on the
Forchheimer coefficients a and b are not constants, but Forchheimer coefficients a and b. A total of 115 data
depend on the bulk velocity, i.e. the Reynolds num- points were collected, which are presented in Table I.
ber. The influence of the flow field may not be crucial This table contains information on the medium type,
for a large number of practical problems (deviations are particle size D, porosity n, permeability k, Forchheimer
approximately 10% for coefficient b), but may be of some coefficients a and b, researcher providing the data, and
importance for simulation approaches for which a high the reference.
accuracy level is required. A plausible explanation for Ward (1964) evaluated experimentally the permeability
this dependence of the coefficients a and b is that the posi- k for various spherical and granular porous media. He
tion for which the boundary layer separation occurs, and did not present porosity n or coefficient a or b values,
subsequently the characteristics of the recirculation zone, but provided k values. Thus, parameter a in Table I was
depends on the Reynolds number (Batchelor, 1990). The computed using Equation (3a).
influence of the recirculation zone on the inertial losses Arbhabhirama and Dinoy (1973) presented two sets of
has already been presented. Eventually, the above-cited data: one set with porosity n and permeability k, and
dependence might be less important for porous media for another set with porosity n and coefficients a and b. In
which the solid phase has sharp edges. It is known that the first set, parameter a (Table I) was computed using
for such media the separation point of the boundary layer Equation (3a).
is not dependent on the Reynolds number, i.e. for isolated Ranganadha Rao and Suresh (1970) plotted their exper-
rigid bodies (Batchelor, 1990; Kotsovinos, 2003), and for imental data in the form of a graphical plot of ∂h/∂x/q
porous media as well (Latinopoulos, 1980). versus q. If the Forchheimer law holds, then the data fall
For the case of granular unconsolidated porous media, on a straight line, and the coefficients a and b can be
a large amount of information concerning experimental evaluated by the relation: ∂h/∂x/q D a C bq. They pre-
data and simulation results is available. Such information sented data of porosity n, permeability k and coefficients
is rather sparse for consolidated porous media. Although a and b.
to a certain extent the analysis of the mechanisms that Tyagi and Todd (1970), using data by Dudgeon (1966),
induce energy losses in granular media might also be presented values of k, a and b. The porosity n of the
relevant for consolidated porous media, it is believed porous medium was not available.

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
538

Table I. Experimental data for Forchheimer coefficients a and b and permeability k

No. Medium type Particle size D (m) Porosity n a (s m1 ) b s2 m2  k m2  Data by Reference

1 Glass beadsa 0.000273 —b 1 766Ð67c —b 5Ð77 ð 1011 Ward Ward (1964)


2 Glass beadsa 0.000322 —b 1 600Ð26c —b 6Ð3777 ð 1011 Ward Ward (1964)
3 Glass beadsa 0.000322 —b 1 219Ð34c —b 8Ð3677 ð 1011 Ward Ward (1964)

Copyright  2006 John Wiley & Sons, Ltd.


4 Glass beadsa 0.000322 —b 1 131Ð37c —b 9Ð0177 ð 1011 Ward Ward (1964)
5 Glass beadsa 0.00038 —b 886Ð41c —b 1Ð15 ð 1010 Ward Ward (1964)
6 Glass beadsa 0.000458 —b 545Ð12c —b 1Ð87 ð 1010 Ward Ward (1964)
7 Glass beadsa 0.000545 —b 389Ð07c —b 2Ð62 ð 1010 Ward Ward (1964)
8 Granular activated 0.00061 —b 312Ð69c —b 3Ð26 ð 1010 Ward Ward (1964)
carbon
9 Sand 0.000625 —b 342Ð07c —b 2Ð98 ð 1010 Ward Ward (1964)
10 Glass beadsa 0.00065 —b 293Ð77c —b 3Ð47 ð 1010 Ward Ward (1964)
11 Ion exchange resin 0.00073 —b 426Ð51c —b 2Ð39 ð 1010 Ward Ward (1964)
12 Granular activated 0.00114 —b 50Ð46c —b 2Ð02 ð 109 Ward Ward (1964)
carbon
13 Sand 0.00126 —b 74Ð95c —b 1Ð36 ð 109 Ward Ward (1964)
14 Gravel 0.00188 —b 34Ð21c —b 2Ð98 ð 109 Ward Ward (1964)
15 Anthracite coal 0.00236 —b 30Ð25c —b 3Ð37 ð 109 Ward Ward (1964)
16 Anthracite coal 0.00442 —b 8Ð03c —b 1Ð27 ð 108 Ward Ward (1964)
17 Gravel 0.00504 —b 6Ð03c —b 1Ð69 ð 108 Ward Ward (1964)
18 Anthracite coal 0.00882 —b 2Ð88c —b 3Ð54 ð 108 Ward Ward (1964)
19 Gravel 0.00921 —b 1Ð94c —b 5Ð26 ð 108 Ward Ward (1964)
20 Gravel 0.0161 —b 0Ð57c —b 1Ð8 ð 107 Ward Ward (1964)
21 Sand 0.0016 0.399 85Ð23c —b 1Ð196 ð 109 Arbhabhirama Arbhabhirama and Dinoy (1973)
22 Sand 0.0016 0.391 95Ð18c —b 1Ð071 ð 109 Arbhabhirama Arbhabhirama and Dinoy (1973)
23 Angular gravel 0.0064 0.467 8Ð80c —b 1Ð1581 ð 108 Arbhabhirama Arbhabhirama and Dinoy (1973)
M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

24 Angular gravel 0.0064 0.47 11Ð98c —b 8Ð51 ð 109 Arbhabhirama Arbhabhirama and Dinoy (1973)
25 Angular gravel 0.0283 0.465 1Ð16c —b 8Ð8255 ð 108 Arbhabhirama Arbhabhirama and Dinoy (1973)
26 Angular gravel 0.013 0.461 2Ð96c —b 3Ð4425 ð 108 Arbhabhirama Arbhabhirama and Dinoy (1973)

DOI: 10.1002/hyp
Hydrol. Process. 21, 534–554 (2007)
Table I. (Continued )

No. Medium type Particle size D (m) Porosity n a (s m1 ) b s2 m2  k m2  Data by Reference

27 Sand 0.00101 0Ð4 99Ð00 2 630 7Ð3 ð 1010 Subba Arbhabhirama and Dinoy (1973)
28 Sand 0.00101 0Ð381 115Ð00 3 450 6Ð53 ð 1010 Subba Arbhabhirama and Dinoy (1973)
29 Sand 0.0017 0Ð436 32Ð50 1 100 2Ð254 ð 109 Subba Arbhabhirama and Dinoy (1973)
30 Sand 0.0017 0Ð417 47Ð50 1 990 1Ð59 ð 109 Subba Arbhabhirama and Dinoy (1973)

Copyright  2006 John Wiley & Sons, Ltd.


31 Sand 0.0017 0Ð403 40Ð00 1 640 1Ð878 ð 109 Subba Arbhabhirama and Dinoy (1973)
32 Sand 0.00286 0Ð43 13Ð50 720 5Ð665 ð 109 Subba Arbhabhirama and Dinoy (1973)
33 Sand 0.00286 0Ð423 22Ð50 880 3Ð46 ð 109 Subba Arbhabhirama and Dinoy (1973)
34 Sand 0.00404 0Ð384 7Ð50 530 9Ð79 ð 1010 Subba Arbhabhirama and Dinoy (1973)
35 Sand 0.00404 0Ð367 10Ð50 780 6Ð847 ð 109 Subba Arbhabhirama and Dinoy (1973)
36 Gravel 0.0055 0Ð372 4Ð30 430 1Ð658 ð 108 Subba Arbhabhirama and Dinoy (1973)
37 Gravel 0.0055 0Ð356 7Ð50 550 1Ð028 ð 108 Subba Arbhabhirama and Dinoy (1973)
38 Gravel 0.0055 0Ð346 10Ð50 780 7Ð338 ð 109 Subba Arbhabhirama and Dinoy (1973)
39 Round river gravel 0.00101 0Ð4 99Ð00 2 630 7Ð3 ð 1010 Ranganadha Ranganadha et al. (1970)
40 Round river gravel 0.00101 0Ð381 115Ð00 3 450 6Ð53 ð 1010 Ranganadha Ranganadha et al. (1970)
41 Round river gravel 0.0017 0Ð436 32Ð50 1 100 2Ð254 ð 109 Ranganadha Ranganadha et al. (1970)
42 Round river gravel 0.0017 0Ð417 47Ð50 1 990 1Ð59 ð 109 Ranganadha Ranganadha et al. (1970)
43 Round river gravel 0.0017 0Ð403 40Ð00 1 640 1Ð878 ð 109 Ranganadha Ranganadha et al. (1970)
44 Round river gravel 0.0017 0Ð392 51Ð50 3 330 1Ð488 ð 109 Ranganadha Ranganadha et al. (1970)
45 Round river gravel 0.00286 0Ð43 13Ð50 720 5Ð665 ð 109 Ranganadha Ranganadha et al. (1970)
46 Round river gravel 0.00286 0Ð423 22Ð50 880 3Ð46 ð 109 Ranganadha Ranganadha et al. (1970)
47 Round river gravel 0.00286 0Ð403 34Ð00 400 2Ð21 ð 109 Ranganadha Ranganadha et al. (1970)
FORCHHEIMER EQUATION COEFFICIENTS

48 Round river gravel 0.00404 0Ð384 7Ð50 530 9Ð79 ð 1010 Ranganadha Ranganadha et al. (1970)
49 Round river gravel 0.00404 0Ð367 10Ð50 780 6Ð847 ð 109 Ranganadha Ranganadha et al. (1970)
50 Round river gravel 0.0055 0Ð372 4Ð30 430 1Ð658 ð 108 Ranganadha Ranganadha et al. (1970)
51 Round river gravel 0.0055 0Ð356 7Ð50 550 1Ð028 ð 108 Ranganadha Ranganadha et al. (1970)
52 Round river gravel 0.0055 0Ð346 10Ð50 780 7Ð338 ð 109 Ranganadha Ranganadha et al. (1970)
53 Blue metal 0.0019 —b 16Ð61 959 8Ð05 ð 109 Dudgeon Tyagi and Todd (1970)
54 River gravel 0.002 —b 19Ð04 2 174 7Ð07 ð 109 Dudgeon Tyagi and Todd (1970)
55 Nepean sand 0.00027 —b 811Ð61 961 1Ð66 ð 1010 Dudgeon Tyagi and Todd (1970)
56 Blue metal 0.0047 —b 7Ð79 573 1Ð72 ð 108 Dudgeon Tyagi and Todd (1970)
57 River gravel 0.00095 —b 78Ð91 2 232 1Ð69 ð 109 Dudgeon Tyagi and Todd (1970)

DOI: 10.1002/hyp
Hydrol. Process. 21, 534– 554 (2007)
539
540

Table I. (Continued )

No. Medium type Particle size D (m) Porosity n a (s m1 ) b s2 m2  k m2  Data by Reference

58 Blue metal 0.0105 —b 1Ð43 220 9Ð2 ð 108 Dudgeon Tyagi and Todd (1970)
59 Blue metal 0.0105 —b 0Ð51 97 2Ð3 ð 107 Dudgeon Tyagi and Todd (1970)
60 Blue metal 0.011 —b 1Ð15 162 1Ð03 ð 107 Dudgeon Tyagi and Todd (1970)
61 River gravel 0.012 —b 1Ð89 262 7 ð 108 Dudgeon Tyagi and Todd (1970)

Copyright  2006 John Wiley & Sons, Ltd.


62 Marbles 0.0156 —b 1Ð10 103 1Ð2 ð 107 Dudgeon Tyagi and Todd (1970)
63 Marbles 0.0156 —b 0Ð50 63 2Ð47 ð 107 Dudgeon Tyagi and Todd (1970)
64 Marbles 0.0156 —b 0Ð76 95 1Ð56 ð 107 Dudgeon Tyagi and Todd (1970)
65 Marble mixture 0.0158 —b 0Ð73 77 1Ð75 ð 107 Dudgeon Tyagi and Todd (1970)
66 River gravel 0.019 —b 0Ð82 145 1Ð62 ð 107 Dudgeon Tyagi and Todd (1970)
67 Blue metal 0.019 —b 0Ð61 117 2Ð2 ð 107 Dudgeon Tyagi and Todd (1970)
68 Marbles 0.0246 —b 0Ð58 66 2Ð24 ð 107 Dudgeon Tyagi and Todd (1970)
69 Blue metal 0.025 —b 0Ð33 121 4 ð 107 Dudgeon Tyagi and Todd (1970)
70 Marbles 0.0285 —b 0Ð36 49 3Ð18 ð 107 Dudgeon Tyagi and Todd (1970)
71 River gravel 0.04 —b 0Ð24 51 5Ð4 ð 107 Dudgeon Tyagi and Todd (1970)
72 River gravel 0.084 —b 0Ð06 15 2Ð04 ð 106 Dudgeon Tyagi and Todd (1970)
73 Sand 0.00107 —b 230Ð00 3 080 6Ð91 ð 1010 Ahmed Ahmed and Sunada (1969)
74 Sand 0.000764 —b 380Ð00 4 540 3Ð96 ð 1010 Ahmed Ahmed and Sunada (1969)
75 Sand 0.0014 —b 149Ð00 2 400 1 ð 109 Ahmed Ahmed and Sunada (1969)
76 Sand 0.00054 —b 739Ð00 7 450 2Ð1 ð 1010 Ahmed Ahmed and Sunada (1969)
77 Sand 0.00199 —b 93Ð80 1 790 1Ð69 ð 109 Ahmed Ahmed and Sunada (1969)
78 Sand 0.00258 —b 69Ð40 1 650 2Ð21 ð 109 Ahmed Ahmed and Sunada (1969)
79 Sand 0.00105 —b 116Ð40 2 920 8 ð 1010 Lindquist Ahmed and Sunada (1969)
80 Sand 0.00492 —b 6Ð74 368 1Ð38 ð 108 Lindquist Ahmed and Sunada (1969)
81 Ottawa sand 0.0007 —b 1 660Ð00 79 600 8Ð277 ð 1011 Fancher Ahmed and Sunada (1969)
82 Sand 0.003 —b 1Ð23 9Ð2 7Ð6 ð 108 Forchheimer Ahmed and Sunada (1969)
83 Sand 0.005 —b 0Ð41 5 2Ð3 ð 107 Forchheimer Ahmed and Sunada (1969)
M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

84 Glass spheresa 0.003 —b 14Ð50 648 6Ð45 ð 109 Sunada Ahmed and Sunada (1969)
85 Glass beads 0.0032 —b 14Ð90 623 6Ð7 ð 109 Blake Ahmed and Sunada (1969)
86 Nickel saddles 0.00334 —b 8Ð90 210 1Ð12 ð 108 Brownell Ahmed and Sunada (1969)
87 Glass beads 0.0053 —b 6Ð47 183 1Ð5 ð 108 Brownell Ahmed and Sunada (1969)
88 Granular absorbent 0.000855 —b 147Ð00 1 420 8Ð6 ð 1010 Allen Ahmed and Sunada (1969)

DOI: 10.1002/hyp
Hydrol. Process. 21, 534–554 (2007)
Table I. (Continued )

No. Medium type Particle size D (m) Porosity n a (s m1 ) b s2 m2  k m2  Data by Reference

89 Sand 0.005 —b 18Ð90 1 370 4Ð94 ð 109 Mobasheri Ahmed and Sunada (1969)
90 Marble 0.016 —b 0Ð9 117 1Ð19 ð 107 Kirkham Ahmed and Sunada (1969)
91 Gravel 0.012 —b 1Ð26 35Ð1 8Ð09022 ð 108c Bordier–Zimmer Bordier and Zimmer (2000)
92 Gravel 0.03 —b 0Ð63 30Ð8 1Ð60784 ð 107c Bordier–Zimmer Bordier and Zimmer (2000)

Copyright  2006 John Wiley & Sons, Ltd.


93 Cr. Rock 0.00288 0Ð42 35Ð15c 940 2Ð9 ð 109 Niranjan Vankataraman and Rao (1998)
94 Cr. Rock 0.00925 0Ð43 1Ð90c 258 5Ð36 ð 108 Niranjan Vankataraman and Rao (1998)
95 Cr. Rock 0.0144 0Ð415 2Ð25c 115 4Ð53 ð 108 Niranjan Vankataraman and Rao (1998)
96 Gl. Sps.a 0.02 0Ð383 2Ð03c 44Ð8 5Ð02 ð 108 Niranjan Vankataraman and Rao (1998)
97 Gl. Sps.a 0.025 0Ð392 1Ð53c 27Ð5 6Ð67 ð 108 Niranjan Vankataraman and Rao (1998)
98 Cr. Rock 0.0299 0Ð466 0Ð38c 29 2Ð69 ð 107 Niranjan Vankataraman and Rao (1998)
99 Cr. Rock 0.0069 0Ð472 2Ð40c 304 4Ð24 ð 108 Nasser Vankataraman and Rao (1998)
100 Cr. Rock 0.0168 0Ð445 0Ð64c 1 280 1Ð59 ð 107 Nasser Vankataraman and Rao (1998)
101 Cr. Rock 0.01065 0Ð441 1 1Ð2c 210 8Ð5 ð 108 Jayachandra Vankataraman and Rao (1998)
102 Cr. Rock 0.014 0Ð444 1Ð13c 122 9 ð 108 Jayachandra Vankataraman and Rao (1998)
103 Cr. Rock 0.019 0Ð440 8 0Ð72c 108 1Ð42 ð 107 Jayachandra Vankataraman and Rao (1998)
104 Round gravel 0.012 0Ð373 6Ð03c 207 1Ð69 ð 108 Arbhabhirama Vankataraman and Rao (1998)
105 Round gravel 0.012 0Ð357 5Ð63c 187 1Ð81 ð 108 Arbhabhirama Vankataraman and Rao (1998)
106 Angular gravel 0.013 0Ð479 2Ð71c 130 3Ð76 ð 108 Arbhabhirama Vankataraman and Rao (1998)
107 Cr. Rock 0.0131 0Ð47 4Ð45c 67 2Ð29 ð 108 Pradip Kumar Vankataraman and Rao (1998)
108 Gl. Sps.a 0.0156 0Ð395 1Ð20c 66Ð7 8Ð49 ð 108 Pradip Kumar Vankataraman and Rao (1998)
109 Gl. Sps.a 0.0184 0Ð382 7 1Ð25c 56 8Ð14 ð 108 Pradip Kumar Vankataraman and Rao (1998)
FORCHHEIMER EQUATION COEFFICIENTS

110 Cr. Rock 0.0201 0Ð458 8 1Ð55c 56Ð7 6Ð59 ð 108 Pradip Kumar Vankataraman and Rao (1998)
111 Gl. Sps.a 0.0289 0Ð413 1 0Ð60c 26Ð9 1Ð69 ð 107 Pradip Kumar Vankataraman and Rao (1998)
112 Cr. Rock 0.0289 0Ð487 3 1Ð1c 31 9Ð28 ð 108 Pradip Kumar Vankataraman and Rao (1998)
113 Gl. Sps.a 0.0156 0Ð355 3 1Ð42c 97Ð9 7Ð2 ð 108 Sharma Vankataraman and Rao (1998)
114 Gl. Sps.a 0.0156 0Ð355 8 0Ð76c 145 1Ð35 ð 107 Sharma Vankataraman and Rao (1998)
115 Gl. Sps.a 0.0289 0Ð398 2 0Ð31c 38Ð6 3Ð31 ð 107 Sharma Vankataraman and Rao (1998)

a Spherical grains.
b Data not available.
c Data computed from Equation (3a).

DOI: 10.1002/hyp
Hydrol. Process. 21, 534– 554 (2007)
541
542 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

Ahmed and Sunada (1969) presented data of parame- spherical and non-spherical porous media. Where values
ters k, a and b without reference to the porosity n. The of a were not available, they were computed using
data were based on their own studies and on studies by Equation (3a).
Forchheimer (1901), Blake (1922), Fancher and Lewis
(1933), Lindquist (1933), Allen (1944), Brownell and
Katz (1947), Mobasheri and Todd (1963) and Kirkham
(1966). They evaluated experimentally the values of a RESULTS OF EXPERIMENTAL DATA ANALYSIS
and b based on the graphical plot of ∂h/∂x/q versus q.
Empirical equations from experimental values of a and b
Bordier and Zimmer (2000) obtained data for a and b
on the basis of a graphical plot of macroscopic velocity The experimental data in Table I were used to derive
q versus the experimentally measured hydraulic gradient empirical equations, of various forms, relating coeffi-
∂h/∂x. Porosity n values were not available. cients a and b to D and/or n. In an effort to derive
Venkataraman and Rao (1998) presented experimental equations similar to Equations (3b) and (3c), graphs of
data provided by Nasser (1970), Arbhabhirama and the experimental values of a or b as a function of par-
Dinoy (1973), Niranjan (1973), Pradip Kumar (1994), ticle size D are presented in Figure 2. In Figure 2a and
Jayachandra (1995) and Sharma (1995). They studied b, all the data are used and the following two regression

10000

1000
a = 0.000859D-1.726567
R2 = 0.924634
100
aexp.(s/m)

10

0.1

0.01
0.0001 0.001 0.01 0.1
D (m) (a)

10000

b = 0.546692D-1.253135
1000
R2 = 0.915710
bexp.(s2/m2)

100

10

1
0.0001 0.001 0.01 0.1
D (m) (b)

10000
b exp (Non Sph.)
b exp.(Sph.)
1000
bn-sph. = 0.679715D-1.218168
bexp.(s2/m2)

R2 = 0.909655
100

10
bsph. = 0.237922D-1.385925
R2 = 0.905440

1
0.0001 0.001 0.01 0.1
D (m) (c)
Figure 2. (a) Correlation of aexp and D, (b) correlation of bexp and D, and (c) correlation of bexp and D for spherical and non-spherical porous media

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp
FORCHHEIMER EQUATION COEFFICIENTS 543

equations are derived: b D 1Ð107 768 ð 1010 D1Ð301 82 n13Ð836 369


a D 0Ð000 859D1Ð726 567 R2 D 0Ð9246; N D 115 1  n18Ð365 290 R2 D 0Ð8806; N D 49
10a 13b
1Ð253 135 2
b D 0Ð546 692D R D 0Ð9157; N D 89 In the above empirical Equations (10a)–(13b) the units
10b are metres for D, seconds per meter for a and seconds
squared per meter squared for b.
where N is the total number of experimental data points The different number of data N used in deriving the
used in the derivation of the empirical equations and R2 previous equations is a result of the lack of n and/or b
is the coefficient of determination. values in some data sets.
In Figure 2c, the following two regression lines, con-
Validity testing of various equations used to predict a
cerning b values, are shown separately for non-spherical
or b
and spherical porous media:
The experimental data of Table I were used in testing
b D 0Ð679 715D1Ð218 168 R2 D 0Ð9097; N D 80; the validity and in comparing the various theoretical and
non-spherical 10c empirical relations presented. The following two methods
were used to test the validity of various equations:
1Ð385 925 2
b D 0Ð237 922ÐD R D 0Ð9054; N D 9;
spherical 10d 1. The root-mean-square error (RMSE) and the normal-
ized objective function (NOF) of the theoretical and
The difference between Equation (10c) and Equation experimental values of a and b were computed. RMSE
(10d) is not significant, and the number of data points is defined as

used to derive Equation (10d) is limited; therefore, most  N

interest is given to the general equation, Equation (10b).  xi  yi 2

In an attempt to derive more accurate empirical  iD1
equations, n was also introduced in the regression RMSE D 14
N
analysis. By multiple regression analysis of experimental
data of coefficients a and b versus particle size D and where xi are the experimental values of a or b (Table I),
porosity n, the following equations were obtained: yi are the values computed by Equations (3a)–(13b), and
N is the total number of points. The parameter RMSE
a D 0Ð003 333D1Ð500 403 n0Ð060 350 R2 D 0Ð9108; has to be as close to 0Ð0 as possible for good prediction.
N D 55 11a NOF is the ratio of the RMSE to the overall mean X of
1Ð265 175 1Ð141 417 2
the experimental data, defined as:
b D 0Ð194 325D n R D 0Ð8715;
RMSE
N D 49 11b NOF D 15
X
In an attempt to derive equations similar to Equations 
where X D 1/N N iD1 xi is the average value of the
(6a)–(6d), multiple regression analysis was performed experimental data (for a or b). NOF has to be as close to
between the experimental data of a and b versus par- 0Ð0 as possible. However, when parameter NOF is less
ticle size D and parameter (1  n), and the following than 1Ð0, then the theoretical method is reliable and can
equations were obtained: be used with sufficient accuracy (Hession et al., 1994;
Kornecki et al., 1999).
a D 0Ð002 789D1Ð502 361 1  n0Ð216 014 R2 D 0Ð9142;
2. The validity is also tested through scattergrams of
N D 55 12a computed (Equations (3a)–(13b)) versus experimental
1Ð263 314 1Ð532 475 2 values of a and b. The best match occurs when all
b D 1Ð228 873D 1  n R D 0Ð8762;
points fall on a 1 : 1 slope line. Deviation from that
N D 49 12b line is measured by fitting through the points a straight
regression line of equation
Finally, in an attempt to derive equations similar to
Equations (4a)–(4d) or (5a) and (5b), multiple regression y D x 16
analysis was performed between the experimental data of
a and b versus particle size D, porosity n and parameter where y implies computed values and x experimental
(1  n), and the following equations were obtained: values. The slope  of this straight line should be equal
to 1Ð0 for a perfect match. If this slope  is less than 1Ð0,
a D 6Ð527 953 ð 1015 D1Ð547 45 n16Ð068 711 then the theoretical equation underestimates the experi-
mental data. If the slope  is greater than 1Ð0, then the
1  n23Ð157 232 R2 D 0Ð9188; N D 55
theoretical equation overestimates the experimental val-
13a ues. Another parameter that evaluates the accuracy of the

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
544 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

agreement is the coefficient of determination R2 , which experimental data is seen for Equations (10a)–(10b) and
shows how well a straight regression line fits the data. in Figures 3b and 4b for Equations (11a) and (11b), (12a)
The closer R2 is to 1Ð0, the less the points are scattered and (12b) and (13a) and (13b).
around the straight line. Scattergrams of predicted versus experimental values
of a and b are presented in Figures 5 and 6 respectively.
A comparison between the various theoretical values A linear regression equation (Equation (16)) and the best
of a and b and the experimental values is presented fit line (1 : 1 slope) are also shown. Best agreement for a
in Figures 3 and 4, where experimental and computed value, based on Figure 5, is observed for Equation (11a),
data points are plotted as a function of particle diameter. with slope  D 0Ð8547 and R2 D 0Ð9160. Good agree-
Figure 3a uses 115 data points, Figure 3b uses 55 data ment is also observed for Equation (12a) and Equation
points, Figure 4a uses 89 data points and Figure 4b uses (13a). Based on Figure 6, for the b value, good agree-
49 data points. As mentioned, the different number of ment show the empirical Equations (11b), (12b) and (13b)
data points used in Figures 3 and 4 is a result of the with slopes  close to 0Ð98 and R2 close to 0Ð87. It is
lack of values of porosity n and/or b in some data also noted that the range of values of diameter D has
sets. In Figures 3a and 4a, the best agreement with the been also divided into smaller ranges to see whether the

10000
Exper.
Eq.3b
Eq.7a
Eq.7c
1000 Eq.8a
Eq.8c
Eq.9a
Eq.9c
Eq.10a
100

10
a (s/m)

0.1

0.01

0.001
0.0001 0.001 0.01 0.1
D (m) (a)

10000
Exper.
Eq.4a
Eq.5a
Eq.6a
1000 Eq.6c
Eq.11a
Eq.12a
Eq.13a
100

10
a (s/m)

1.0

0.01

0.001
0.0001 0.001 0.01 0.1
D (m) (b)
Figure 3. Comparison of experimental and computed a values as a function of particle size D

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp
FORCHHEIMER EQUATION COEFFICIENTS 545

100000 Exper.
Eq.3c
Eq.7b
Eq.7d
Eq.8b
Eq.8d
10000 Eq.9b
Eq.9d
Eq.10b

1000
b (s2/m2)

100

10

1
0.0001 0.001 0.01 0.1
D (m) (a)

10000 Exper.
Eq.4b
Eq.5b
Eq.6b
Eq.6d
Eq.11b
Eq.12b
Eq.13b
1000
b (s2/m2)

100

10

1
0.0001 0.001 0.01 0.1
D (m) (b)

Figure 4. Comparison of experimental and computed b values as a function of particle size D

slope  and the coefficient of determination R2 are better slope  D 0Ð9203 and R2 D 0Ð9765. Figure 10 indicates
for some of those ranges. No improvement was detected that the best method for the evaluation of coefficient b is
based on these tests, so only the general results are pre- Equation (8b) with slope  D 0Ð9768 and R2 D 0Ð9474. In
sented (Figures 5 and 6). Figure 9, owing to the small number of available experi-
For spherical porous media, a comparison between mental data, Equations (4a), (5a), (6a), (6c), (11a), (12a)
the theoretical values of a and b and the experimental and (13a) have small values of R2 . For the same reason,
values is presented in Figures 7 and 8 for various particle Equations (4b), (5b), (6b), (6d), (11b), (12b) and (13b)
diameters. Figure 7a uses 17 data points, Figure 7b uses in Figure 10 present small values of R2 . Owing to the
eight data points, Figure 8a uses nine data points and small number of data sets, further checks are necessary
Figure 8b uses eight data points. to ensure the reliability of the results.
For spherical grains, scattergrams of theoretical ver- Table II summarizes values of slope  and coef-
sus experimental values of a and b are presented in ficient of determination R2 from the scattergrams of
Figures 9 and 10 respectively. Again, a linear regression Figures 5, 6, 9 and 10, and of parameters RMSE
line (Equation (16)) and a best fit (1 : 1 slope) line are (Equation (14)) and NOF (Equation (15)) for all
also presented. Figure 9 indicates that the best method methods used for evaluating a and b (Equations
for evaluating the parameter a is Equation (8a) with (3a)–(13b)). As is shown in Table II, the best RMSE

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
546 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

10000 10000 10000

a(Eq.3b)(s/m)

a(Eq.5a)(s/m)
a(Eq.4a)(s/m)
100 100 100

1 1 1
1:1 y=0.2859x 1:1 y = 0.7075x 1:1 y = 3.8125x
R2=0.8928 R2= 0.8832 R2 = 0.8829
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp.(s/m) (a) aexp.(s/m) (b) aexp.(s/m) (c)

10000 10000 10000

a(Eq.7a)(s/m)
a(Eq.6c)(s/m)
a(Eq.6a)(s/m)

1:1
100 100 100

1 1 1
1:1 y = 3.1151x 1:1 y = 3.9430x y = 3.8350x
2 2
R = 0.8982 R = 0.8982 R2=0.8899
0 0 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp.(s/m) (d) aexp.(s/m) (e) aexp.(s/m) (f)

10000 10000 10000

a(Eq.8c)(s/m)
a(Eq.8a)(s/m)
a(Eq.7c)(s/m)

100 100 100

1 1 1
1:1 y = 4.8541x 1:1 y = 0.8055x
1:1
y = 1.0195x
2 2 2
R = 0.8899 R = 0.7526 R = 0.7526
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp.(s/m) (g) aexp.(s/m) (h) aexp.(s/m) (i)

10000 10000 10000


a(Eq.10a)(s/m)
a(Eq.9a)(s/m)

a(Eq.9c)(s/m)

100 100 100

1 1:1 1 1:1 y = 2.5977x


1 y = 0.6619x
y = 2.0523x 1:1
2 2 2
R = 0.7526 R =0.7526 R = 0.7685
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp.(s/m) (j) aexp.(s/m) (k) aexp.(s/m) (l)

10000 10000 10000


a(Eq.11a)(s/m)

a(Eq.13a)(s/m)
a(Eq.12a)(s/m)

100 100 100

1 1:1 y = 0.8547x 1 1
2
R = 0.9160 1:1 y = 0.8546x 1:1 y = 0.8473x
2
0.01 R2 = 0.9149 R = 0.9159
0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp.(s/m) (m) aexp.(s/m) (n) aexp.(s/m) (o)

Figure 5. Scattergrams of computed and experimental values of coefficient a: (a) Equation (3b); (b) Equation (4a); (c) Equation (5a); (d) Equation
(6a); (e) Equation (6c); (f) Equation (7a); (g) Equation (7c); (h) Equation (8a); (i) Equation (8c); (j) Equation (9a); (k) Equation (9c); (l) Equation
(10a); (m) Equation (11a); (n) Equation (12a); (o) Equation (13a)

and NOF values for coefficient a are for Equation (11a) (a) ratio of Equation (4a) to Equation (6a)
and for coefficient b for Equation (13b). A general com- 1501  n
parison of all the relations is presented on Table III, with f1 D 17a
the best results for a and b estimation being Equations 6570n3
(11a) and (11b), Equations (12a) and (12b) and Equations (b) ratio of Equation (4c) to Equation (6a)
(13a) and (13b). 1441  n
It is also interesting to compare the semi-empirical f2 D 17b
6570n3
relations of Ergun (1952) (Equations (4a) and (4b)) and
Kovács (1981) (Equations (4c) and (4d)) with those (c) ratio of Equation (4b) to Equation (6b)
proposed in this study, i.e. Equations (6a) and (6b). This 1Ð75
can be done by building the following ratios: f3 D 17c
98Ð1n3

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp
FORCHHEIMER EQUATION COEFFICIENTS 547

10000 10000 10000

b(Eq.4b) (s2/m2)

b(Eq.5b) (s2/m2)
b(Eq.3c) (s2/m2)
1000 1000 1000

100 100 100

10 y = 0.3174x 10 y = 0.5335x 10 y = 0.6097x


1:1 1:1 1:1 2
R = 0.8589
R2 = 0.8419 R2 = 0.8589
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (a) bexp.(s2/m2) (b) bexp.(s2/m2) (c)

10000
10000 10000
b(Eq.6b)(s2/m2)

1000

b(Eq.6d) (s2/m2)

b(Eq.7b) (s2/m2)
1000 1000
100
100 100
10 y = 1.8466x
10
1:1 R2 = 0.8406
10 y = 1.6688x y = 2.2076x
1:1 1:1 R2 = 0.8419
R2 = 0.8406
1 1
1
1 100 10000 1 100 10000 1 100 10000
bexp. (s2/m2) (d) bexp. (s2/m2) (e) bexp. (s2/m2) (f)

10000 10000 10000

b(Eq.8d) (s2/m2)
b(Eq.7d) (s2/m2)

b(Eq.8b) (s2/m2)

1000 1000 1000


100 100 100
10 y = 1.9949x 10 y = 0.5519x 10 y = 0.4984x
R2 = 0.8419 1:1 1:1 R2 = 0.8419
1:1 R2 = 0.8419
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (g) bexp.(s2/m2) (h) bexp.(s2/m2) (i)

10000 10000 10000


b(Eq.10b) (s2/m2)
b(Eq.9d) (s2/m2)
b(Eq.9b) (s2/m2)

1000 1000 1000

100 100 100

10 1:1 y = 1.4062x 10 y = 1.2709x 10 1:1 y = 0.9353x


1:1
R2 = 0.8419 R2 = 0.8419 R2 = 0.8784
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp. (s2/m2) (j) bexp. (s2/m2) (k) bexp. (s2/m2) (l)

10000 10000 10000


b(Eq.13b) (s2/m2)
b(Eq.12b) (s2/m2)
b(Eq.11b) (s2/m2)

1000 1000 1000

100 100 100


1:1
10 y = 0.9774x 10 1:1
y = 0.9776x 10 1:1 y = 0.9743x
2
R = 0.8653 R2 = 0.8643 R2 = 0.866
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (m) bexp.(s2/m2) (n) bexp.(s2/m2) (o)

Figure 6. Scattergrams of computed and experimental values of coefficient b: (a) Equation (3c); (b) Equation (4b); (c) Equation (5b); (d) Equation
(6b); (e) Equation (6d); (f) Equation (7b); (g) Equation (7d); (h) Equation (8b); (i) Equation (8d); (j) Equation (9b); (k) Equation (9d); (l) Equation
(10b); (m) Equation (11b); (n) Equation (12b); (o) Equation (13b)

(d) ratio of Equation (4d) to Equation (6b) (f) ratio of Equation (4c) to Equation (6c)
2Ð4 1441  n
f4 D 17d f6 D 17f
98Ð1n3 8316n3
(e) ratio of Equation (4a) to Equation (6c) (g) ratio of Equation (4b) to Equation (6d)
1501  n 1Ð75
f5 D 17e f7 D 17g
8316n3 88Ð65n3

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
548 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

10000
Exper.
Eq.3a
Eq.7a
Eq.7c
Eq.8a
1000 Eq.8c
Eq.9a
Eq.9c
Eq.10a

a (s/m) 100

10

0.1

0.01
0.0001 0.001 0.01 0.1
D (m) (a)

10000
Exper.
Eq.4a
Eq.5a
Eq.6a
Eq.6c
1000 Eq.11a
Eq.12a
Eq.13a

100
a (s/m)

10

0.1

0.01
0.0001 0.001 0.01 0.1
D (m) (b)
Figure 7. Comparison of experimental and computed values of coefficient a, as a function of particle size D, for spherical porous media

(h) ratio of Equation (4d) to Equation (6d) not clear, they give a reasonable to excellent comparison
with the numerical results obtained using the minimum
2Ð4
f8 D 17h value of porosity. Therefore, it would be interesting to
88Ð65n3 perform further numerical simulations for different values
As presented in Table IV, the values of the coeffi- of the porosity and compare them with the semi-empirical
cients f1 , f2 , f3 and f7 are close to 1Ð0 for n D 0Ð26 Equations (4a)–(4d) and (5a) and (5b).
(minimum porosity); for higher values of the porosity
the coefficients fi assume values smaller than 1Ð0. As
will be discussed in the following section, the coeffi- DISCUSSION AND CONCLUSIONS
cients a and b also decay with increasing porosity. The In this study, the application field of the Forchheimer
explanation for this is that, in the denominator of the coef- equation was presented, and the range of values and the
ficients fi is a theoretical coefficient that corresponds to physical significance of its parameters were analysed. The
the minimum porosity. This effect would be even more procedure included the analysis of existing experimental
pronounced if Equations (7a)–(7d) were used to compute data, and the exploitation of existing research, based on
the fi coefficients. Although the physical background of numerical simulation approaches (Hill and Koch, 2002).
Ergun’s (1952) and Kovács’s (1981) approximations is Original relations for the parameters a and b are presented

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp
FORCHHEIMER EQUATION COEFFICIENTS 549

10000 Exper.
Eq.3c
Eq.7b
Eq.7d
Eq.8b
Eq.8d
Eq.9b
Eq.9d
1000 Eq.10b

b (s2/m2)

100

10

1
0.001 0.01 0.1
D (m) (a)

10000 Exper.
Eq.4b
Eq.5b
Eq.6b
Eq.6d
Eq.11b
Eq.12b
Eq.13b
1000
b (s2/m2)

100

10

1
0.001 0.01 0.1
D (m) (b)
Figure 8. Comparison of experimental and computed values of coefficient b, as a function of particle size D, for spherical porous media

in this study based on analysis of experimental and experimental data evaluation. Thereafter, the agreement
numerical data available in the literature. with the data is of a statistical nature. Since in the present
The main conclusions of our experimental data analysis study a larger amount of data have been analysed, it is
are as follows. Concerning the estimation of a, by the believed that the empirical relations derived are more
use of the relations presented in the Introduction (i.e. the reliable. Concerning the estimation of b, by the use of
inverse of hydraulic conductivity), the Kozeny–Carman relations presented in the Introduction, the Kadlec and
approximation included in Ergun’s (1952) approach Knight (1996) approximation gives the best results.
(Equation (4a)) gives excellent results, a conclusion Equations (7a)–(7d) are based on Hill and Koch’s
compatible with previous findings. The reason for (2002) numerical simulations, who investigated flow phe-
the discrepancies between Ergun’s approach and the nomena in a closely packed bed of spheres. Since it is
reported experimental data is that, as already discussed, known that this type of porous formation exhibits the
its theoretical background is not consistent with the lowest possible porosity, and that the inertial resistance
physical processes taking place in porous media. Ward’s coefficient b is inversely proportional to the porosity
(1964) and Kadlec and Knight’s (1996) approaches, (Blick, 1966), the relation above is consequently ade-
which provide a better approximation, are based on an quate rather for the estimation of the upper bound of the

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
550 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

10000 10000 10000


1000 1000 1000

a(Eq.4a) (s/m)

a(Eq.5a) (s/m)
a(Eq.3b) (s/m)
100 100 100
10 10 10
1 1 y = 0.2348x
1
y= 0.2725x y = 1.2707x
0.1 R2 =0 .9765
0.1 1:1 R2 = -0.2151 0.1 1:1 R2 = -0.2135
1:1
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp. (s/m) (a) aexp. (s/m) (b) aexp.(s/m) (c)

10000 10000 10000


1000 1000 1000
a(Eq.6a) (s/m)

a(Eq.7a) (s/m)
a(Eq.6c) (s/m)
100 100 100
10 10 10
1 1 y = 1.1040x
1 1:1 y = 3.6806x
y = 0.8722x
0.1 1:1 R2 = -0.3493 0.1 1:1 R2 = -0.3493 0.1 R2 = 0.9765
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp. (s/m) (d) aexp. (s/m) (e) aexp. (s/m) (f)

10000 10000 10000


1000 1000 1000
a(Eq.8a) (s/m)

a(Eq.8c) (s/m)
a(Eq.7c) (s/m)

100 100 100


10 10 10
1 1:1 1 y = 0.9203x
1
y = 4.6587x 1:1 y = 1.1648x
0.1 R2 = 0.9765
0.1 1:1 R2 = 0.9765 0.1 R2 = 0.9765
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp. (s/m) (g) aexp. (s/m) (h) aexp. (s/m) (i)

10000 10000 10000


1000 1000 1000
a(Eq.10a) (s/m)
a(Eq.9a) (s/m)

a(Eq.9c) (s/m)

100 100 100


10 10 10
1 1 y = 2.9678x
1
y = 2.3447x 1:1 y = 0.7403x
0.1 1:1
1:1
R2 = 0.9765
0.1 R2 = 0.9765 0.1 R2 = 0.9683
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp. (s/m) (j) aexp. (s/m) (k) aexp. (s/m) (l)

10000 10000 10000


1000 1000 1000
a(Eq.12a) (s/m)

a(Eq.13a) (s/m)
a(Eq.11a) (s/m)

100 100 100


10 10 10
1 1:1 1 1:1 1 1:1
y = 0.8869x y = 0.8791x y = 0.8381x
0.1 0.1 R2 = -0.7661
0.1 R2 = -0.5049
R2 = -0.7561
0.01 0.01 0.01
0.01 1 100 10000 0.01 1 100 10000 0.01 1 100 10000
aexp. (s/m) (m) aexp. (s/m) (n) aexp. (s/m) (o)
Figure 9. Scattergrams of computed and experimental values of coefficient a for spherical porous media: (a) Equation (3b); (b) Equation (4a);
(c) Equation (5a); (d) Equation (6a); (e) Equation (6c); (f) Equation (7a); (g) Equation (7c); (h) Equation (8a); (i) Equation (8c); (j) Equation (9a);
(k) Equation (9c); (l) Equation (10a); (m) Equation (11a); (n) Equation (12a); (o) Equation (13a)

coefficient b than for estimation purposes. Porosity val- may be used by models dealing with uncertainty in
ues of 0Ð26 are seldom obtained in laboratory columns of aquifers, e.g. using Monte Carlo and fuzzy analysis
homogeneous spheres. A better approach is obtained if approaches (de Marsily, 1986). For the latter approach,
the influence of the porosity is taken into account (i.e. by Equations (11a) and (11b) can be used to estimate the
using Equations (6a)–(6d)). However, further numerical ‘most likely values’ of the parameters a and b.
checks are necessary to ensure the reliability of the above To our knowledge, a quantitative dependence of the
relation. resistance coefficient on the Reynolds number has not
The determination of the range of values of the Forch- been previously reported. However, further investigation
heimer law coefficients, presented in the previous section, is necessary to detect the influence of porosity, particle

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp
FORCHHEIMER EQUATION COEFFICIENTS 551

10000 10000 10000

b(Eq.4b) (s2/m2)
b(Eq.3c) (s2/m2)

b(Eq.5b) (s2/m2)
1000 1000 1000

100 100 100

10 y = 0.5616x 10 y = 1.4662x 10 y = 1.6756x


1:1 1:1 1:1
2
R = 0.9474 R2 = 0.5544 R2 = 0.5544
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (a) bexp.(s2/m2) (b) bexp.(s2/m2) (c)

10000 10000 10000

b(Eq.6d) (s2/m2)
b(Eq.6b) (s2/m2)

b(Eq.7b) (s2/m2)
1000 1000 1000

100 100 100

10 1:1 y = 4.1463x 10 y = 3.7469x 10 y = 3.9071x


1:1 1:1
R2 = -0.8673 2
R = -0.8673 R2 = 0.9474
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (d) bexp.(s2/m2) (e) bexp.(s2/m2) (f)

10000 10000 10000

b(Eq.8d) (s2/m2)
b(Eq.7d) (s2/m2)

b(Eq.8b) (s2/m2)

1000 1000 1000

100 100 100

10 y = 3.5307x 10 1:1 y = 0.9768x 10 1:1 y = 0.8821x


1:1 2
R = 0.9474
2
R = 0.9474 R2 = 0.9474
1 1 1
1 100 10000 1 100 10000 1 100 10000
2 2 2 2 2 2
bexp.(s /m ) (g) bexp.(s /m ) (h) bexp.(s /m ) (i)

10000 10000 10000


b(Eq.10b) (s2/m2)
b(Eq.9b) (s2/m2)

b(Eq.9d) (s2/m2)

1000 1000 1000

100 100 100


1:1
10 y = 2.4887x 10 y = 2.2492x 10 y = 1.2034x
1:1 2
R = 0.9474
1:1 2
2
R = 0.9834
R = 0.9474
1 1 1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (j) bexp.(s2/m2) (k) bexp.(s2/m2) (l)

10000 10000 10000


b(Eq.11b) (s2/m2)

b(Eq.13b) (s2/m2)
b(Eq.12b) (s2/m2)

1000 1000 1000

100 100 100

10 y = 1.1702x 1:1 10 y = 1.1663x


1:1 10 y = 1.1608x
2
R = 0.1291 R2 = 0.0385
1:1 R2 = 0.6037
1 1
1
1 100 10000 1 100 10000 1 100 10000
bexp.(s2/m2) (m) bexp.(s2/m2) (n) bexp.(s2/m2) (o)
Figure 10. Scattergrams of computed and experimental values of coefficient b for spherical porous media: (a) Equation (3c); (b) Equation (4b);
(c) Equation (5b); (d) Equation (6b); (e) Equation (6d); (f) Equation (7b); (g) Equation (7d); (h) Equation (8b); (i) Equation (8d); (j) Equation (9b);
(k) Equation (9d); (l) Equation (10b); (m) Equation (11b); (n) Equation (12b); (o) Equation (13b)

shape and flow regime. A dependence of the coefficients losses due to direction changes, contractions of the con-
a and b on the Reynolds number will eventually be duits, etc., should also be taken into account.
important for consolidated porous media. In summary, existing relations evaluating Forchheimer
For the laminar flow regime, the simulation approach coefficients a and b have been presented. Equations based
by Lao et al. (2004) is adequate, but one has to consider on theoretical approaches and experimental analysis have
our remarks. If turbulence occurs, one should use the been compared and evaluated. By comparison of all
approach of Ahmed and Sunada (1969), where additional these relations, based on RMSE (Equation (14)), NOF

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
552

Copyright  2006 John Wiley & Sons, Ltd.


Table II. Results of validity tests for various equations used to estimate a and b

Coefficient a Equation (x)


Parameter testing 3b 4a 5a 6a 6c 7a 7c 8a 8c 9a 9c 10a 11a 12a 13a
validity
RMSE (s m1 ) 273Ð97 13Ð99 116Ð40 87Ð42 120Ð03 992Ð57 1332Ð55 165Ð75 190Ð85 535Ð66 746Ð88 167Ð14 9Ð60 9Ð64 9Ð72
NOF 2Ð112 0Ð648 5Ð390 4Ð048 5Ð558 7Ð653 10Ð275 1Ð278 1Ð472 4Ð130 5Ð759 1Ð289 0Ð444 0Ð446 0Ð450
 all data 0Ð2859 0Ð7075 3Ð8125 3Ð1151 3Ð9430 3Ð8350 4Ð8541 0Ð8055 1Ð0195 2Ð0523 2Ð5977 0Ð6619 0Ð8547 0Ð8546 0Ð8473
R2 all data 0Ð8928 0Ð8832 0Ð8829 0Ð8982 0Ð8982 0Ð8899 0Ð8899 0Ð7526 0Ð7526 0Ð7526 0Ð7526 0Ð7685 0Ð9160 0Ð9149 0Ð9159
 for spherical grains 0Ð2725 0Ð2348 1Ð2707 0Ð8722 1Ð1040 3Ð6806 4Ð6587 0Ð9203 1Ð1648 2Ð3447 2Ð9678 0Ð7403 0Ð8869 0Ð8791 0Ð8381
R2 for spherical grains 0Ð9785 0Ð2151 0Ð2135 0Ð3493 0Ð3493 0Ð9765 0Ð9765 0Ð9765 0Ð9765 0Ð9765 0Ð9765 0Ð9683 0Ð7561 0Ð7661 0Ð5049
Coefficient b Equation (x)
Parameter testing 3c 4b 5b 6b 6d 7b 7d 8b 8d 9b 9d 10b 11b 12b 13b
validity
RMSE (s2 m2 ) 8346Ð13 602Ð29 523Ð80 1237Ð57 1020Ð74 8107Ð22 8037Ð21 8213Ð31 8241Ð20 7971Ð68 7983Ð63 8092Ð68 358Ð57 360Ð08 356Ð82
NOF 4Ð755 0Ð752 0Ð654 1Ð545 1Ð275 4Ð618 4Ð579 4Ð679 4Ð695 4Ð541 4Ð548 4Ð610 0Ð448 0Ð450 0Ð446
 all data 0Ð3174 0Ð5335 0Ð6097 1Ð8466 1Ð6688 2Ð2076 1Ð9949 0Ð5519 0Ð4984 1Ð4062 1Ð2709 0Ð9353 0Ð9774 0Ð9776 0Ð9743
R2 all data 0Ð8419 0Ð8589 0Ð8589 0Ð8406 0Ð8406 0Ð8419 0Ð8419 0Ð8419 0Ð8419 0Ð8419 0Ð8419 0Ð8784 0Ð8653 0Ð8643 0Ð8660
 for spherical grains 0Ð5616 1Ð4662 1Ð6756 4Ð1463 3Ð7469 3Ð9071 3Ð5307 0Ð9768 0Ð8821 2Ð4887 2Ð2492 1Ð2034 1Ð1702 1Ð1608 1Ð1663
R2 for spherical grains 0Ð9474 0Ð5544 0Ð5544 0Ð8673 0Ð8673 0Ð9474 0Ð9474 0Ð9474 0Ð9474 0Ð9474 0Ð9474 0Ð9834 0Ð1291 0Ð0385 0Ð6037
M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

DOI: 10.1002/hyp
Hydrol. Process. 21, 534–554 (2007)
FORCHHEIMER EQUATION COEFFICIENTS 553

Table III. General ranking of various methods

Equation (x) Rank of parameter used for validity testing˛

 (Figures 9 and 10 R2 (Figures 9 and 10


RMSE NOF  (Figures 5 and 6) R2 (Figures 5 and 6) for spherical porous media) for spherical porous media)

Coefficient a
3b 11 8 8 6 10 1
4a 4 4 6 9 11 —b
5a 6 11 12 10 9 —b
6a 5 9 11 4, 5 5 —b
6c 7 12 14 4, 5 2 —b
7a 14 14 13 7, 8 14 2, 3, 4, 5, 6, 7
7c 15 15 15 7, 8 15 2, 3, 4, 5, 6, 7
8a 8 5 5 12, 13, 14, 15 1 2, 3, 4, 5, 6, 7
8c 10 7 1 12, 13, 14, 15 7 2, 3, 4, 5, 6, 7
9a 12 10 9 12, 13, 14, 15 12 2, 3, 4, 5, 6, 7
9c 13 13 10 12, 13, 14, 15 13 2, 3, 4, 5, 6, 7
10a 9 6 7 11 8 8
11a 1 1 2 1 3 —b
12a 2 2 3 3 4 —b
13a 3 3 4 2 6 —b
Coefficient b
3c 15 15 12 7, 8, 9, 10, 11, 12, 13 7 2, 3, 4, 5, 6, 7, 8
4b 5 5 9 5–6 8 —b
5b 4 4 6 5–6 9 —b
6b 7 7 13 14–15 15 —b
6d 6 6 11 14–15 13 —b
7b 12 12 15 7, 8, 9, 10, 11, 12, 13 14 2, 3, 4, 5, 6, 7, 8
7d 10 10 14 7, 8, 9, 10, 11, 12, 13 12 2, 3, 4, 5, 6, 7, 8
8b 13 13 8 7, 8, 9, 10, 11, 12, 13 1 2, 3, 4, 5, 6, 7, 8
8d 14 14 10 7, 8, 9, 10, 11, 12, 13 2 2, 3, 4, 5, 6, 7, 8
9b 8 8 7 7, 8, 9, 10, 11, 12, 13 11 2, 3, 4, 5, 6, 7, 8
9d 9 9 5 7, 8, 9, 10, 11, 12, 13 10 2, 3, 4, 5, 6, 7, 8
10b 11 11 4 1 6 1
11b 2 2 2 3 5 —b
12b 3 3 1 4 3 —b
13b 1 1 3 2 4 —b

a 1: best method; 15: worst method.


b Low values of R2 .

Table IV. Values of ratios fi comparing Ergun’s (1952) and Kovács’s (1981) expressions (Equations (4a)–(4d)) with the present
Equations (6a)–(6d) (based on numerical data of Hill and Koch (2002)), for: 10 < Re  80 (coefficients f1 to f4 ) for coefficients a
(coefficients f1 and f2 ) and b (coefficients f3 and f4 ); and Re > 80 (coefficients f5 to f8 ) for coefficients a (coefficients f5 and
f6 ) and b (coefficients f7 and f8 )

n f1 f2 f3 f4 f5 f6 f7 f8

0.26 0Ð96 0Ð92 1Ð01 1Ð39 0Ð76 0Ð73 1Ð12 1Ð54


0.30 0Ð59 0Ð57 0Ð66 0Ð91 0Ð47 0Ð45 0Ð73 1Ð00
0.40 0Ð21 0Ð21 0Ð28 0Ð38 0Ð17 0Ð16 0Ð31 0Ð42
0.50 0Ð09 0Ð09 0Ð14 0Ð19 0Ð01 0Ð01 0Ð16 0Ð22

(Equation (15)) and linear correlation of computed ver- ACKNOWLEDGEMENTS


sus experimental values (Equation (16)), the empirical
Equations (11a) and (11b) seem to be the best for the The first author acknowledges a fellowship by the Mpo-
estimation of coefficients a and b; but empirical equations dossakis Foundation, Greece, which was essential for
Equations (12a) and (12b) and (13a) and (13b) are also the completion of this study. The contributions of the
very good. However, one should be cautious when using anonymous referees are gratefully acknowledged. In par-
Equations (13a) and (13b) in accurately evaluating poros- ticular, we thank the second anonymous referee for his
ity n, because small differences in porosity n may cause suggestions concerning the incorporation of Blick’s and
large differences in values of coefficients a and b, because Kovács’s equations and the comparison of Ergun’s rela-
of the large exponents of n. tion to our equations.

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534– 554 (2007)
DOI: 10.1002/hyp
554 M. G. SIDIROPOULOU, K. N. MOUTSOPOULOS AND V. A. TSIHRINTZIS

REFERENCES Lao H-W, Neeman HJ, Papavassiliou V. 2004. A pore network model
for the calculation of non-Darcy flow coefficients in fluid flow
Ahmed N, Sunada DK. 1969. Nonlinear flow in porous media. Journal through porous media. Chemical Engineering Communications 191:
of Hydraulic Division, ASCE 95(6): 1847– 1857. 1285– 1322.
Allen HV. 1944. Pressure drop for flow through beds of granular Latinopoulos PD. 1980. Study of fluctuating flow in a pipe with periodical
absorbents. Petroleum Refiner 23(7): 247– 252. wall geometry. Doctoral dissertation, Department of Civil Engineering,
Arbhabhirama A, Dinoy AA. 1973. Friction factor and Reynolds number Aristotle University of Thessaloniki, Greece (in Greek).
in porous media flow. Journal of the Hydraulic Division, ASCE 99(6): Lindquist E. 1933. On the flow of water through porous soils. In
901– 911. Proceedings of the 1st Congress des Grands Barrages, Stockholm,
Batchelor GK. 1990. An Introduction to Fluid Dynamics. Cambridge Sweden, vol. 5; 81– 101.
University Press: Cambridge. Mobasheri F, Todd DK. 1963. Investigation of the hydraulics of flow near
Bear J. 1979. Hydraulics of Groundwater . Department of Civil recharge wells. Water Resources Center Contribution, 72, University
Engineering, Technion– Israel Institute of Technology: Haifa, Israel. of California, Berkeley, CA, USA.
Bird RB, Stewart WE, Lightfoot EN. 1960. Transport Phenomena. Moutsopoulos KN, Tsihrintzis VA. 2005. Approximate analytical solu-
Wiley: New York. tions for the Forchheimer equation. Journal of Hydrology, 309(1– 4):
Blake FC. 1922. The resistance of packing to fluid flow. Transactions of 93–103.
the American Institute of Chemical Engineers 14: 415–421. Nasser MSS. 1970. Radial non-Darcy flow through porous media.
Blick EF. 1966. Capillary orifice model for high speed flow through MApplSci thesis, University of Windsor, Canada.
porous media I & EC. Process Design and Development 1(5): 90–94. Naudascher E. 1987. Hydraulik der Gerinne und Gerinnebauwerke.
Bordier C, Zimmer D. 2000. Drainage equations and non-Darcian Springer-Verlag: Wien.
modeling in coarse porous media or geosynthetic materials. Journal Niranjan HS. 1973. Non-Darcy flow through porous media. MTech thesis,
of Hydrology 228: 174–187. Civil Engineering Department, IIT, Kanpur, India.
Brownell LE, Katz DL. 1947. Flow of fluids through porous media. Panfilov M, Fourar M. 2006. Physical splitting of nonlinear effects in
Chemical Engineering Progress 43: 537– 548. high-velocity stable flow through porous media. Advances in Water
Coulaud O, Morel P, Caltagirone JP. 1988. Numerical modeling of Resources 29(1): 30–41.
nonlinear effects in laminar flow through a porous medium. Journal of Panfilov M, Oltean C, Panfilova I, Bues M. 2003. Singular nature of
Fluid Mechanics 190: 393–407. nonlinear macroscale effects in high-rate flow through porous media.
De Marsily G. 1986. Quantitative Hydrogeology. Academic Press: New Comtes Rendus Mecanique 331(1): 41–48.
York. Pradip Kumar GN. 1994. Radial non-Darcy flow through coarse granular
Dudgeon CR. 1966. An experimental study of flow of water through media. PhD thesis, Sri Venkateswara University, Tirupati, India.
coarse granular media. La Houille Blanche 21(7): 785–801. Ranganadha Rao RP, Suresh C. 1970. Discussion of ‘Non-linear flow in
Ergun S. 1952. Fluid flow through packed columns. Chemical porous media’, by N Ahmed and DK Sunada. Journal of Hydraulic
Engineering Progress 48: 89–94. Division, ASCE 96(8): 1732– 1734.
Fancher GH, Lewis JA. 1933. Flow of simple fluids through porous Rodi W. 1984. Turbulenzmodelle und ihre Anwendungen mit Hilfe von
materials. Industrial and Engineering Chemistry 25: 1139– 1147. Differenzen Verfahren. Lecture notes, Fredericiana Univeristy of
Forchheimer PH. 1901. Wasserbewegung durch Boden. Zeitschrift des Karlsruhe.
Vereines Deutscher Ingenieure 50: 1781– 1788. Sharma PV. 1995. Non-Darcy parallel flow through porous media. MTech
Fourar M, Radilla G, Lenormand R, Moyne C. 2004. On the non-linear thesis, Sri Venkateswara University, Tirupati, India.
behavior of a laminar single-phase flow through two and three- Skjetne E, Auriault J-L. 1999. High-velocity laminar and turbulent flow
dimensional porous media. Advances in Water Resources 27: 669–677. in porous media. Transport in Porous Media 36: 131– 147.
Ganoulis J, Brunn P, Durst F, Holweg J, Wunderlich A. 1989. Laser Terzidis G. 2003. Steady non-Darcian groundwater flow in unconfined
measurements and computations of viscous flows through cylinders. aquifers. In Proceeding of the 9th Congress of the Hellenic
Journal of Hydraulic Engineering, ASCE 115(9): 1223– 1239. Hydrotechnical Society, 2–5 April, Thessaloniki, Greece; 369– 376 (in
Hession WC, Shanholtz VO, Mostaghimi S, Dillaha TA. 1994. Uncal- Greek).
ibrated performance of the finite element storm hydrograph model. Tyagi AK, Todd DK. 1970. Discussion of ‘Non-linear flow in porous
Transactions of ASAE 37(3): 777– 783. media’, by N Ahmed and DK Sunada. Journal of Hydraulic Division,
Hill RJ, Koch DL. 2002. The transition from steady to weakly turbulent ASCE 96(8): 1734– 1738.
flow in a close-packed ordered array of spheres. Journal of Fluid Venkataraman P, Rao PRM. 1998. Darcian, transitional and turbulent
Mechanics 465: 59–97. flow through porous media. Journal of Hydraulic Engineering, ASCE
Jayachandra K. 1995. An experimental study of steady non-linear parallel 124(8): 840–846.
flow through coarse granular media. MTech thesis, Sri Venkateswara Volker R. 1975. Solutions for unconfined non-Darcy seepage. Journal of
University, Tirupati, India. the Irrigation and Drainage Division, ASCE 101: 53– 65.
Jones SC. 1987. Using the inertial coefficient, ˇ, to characterize Ward JC. 1964. Turbulent flow in porous media. Journal of Hydraulic
heterogeneity in reservoir rocks. In 62nd Annual SPE Technology Division, ASCE 90(5): 1–12.
Conference, Dallas, Texas, 27–30 September; SPE 16949. Wu YS. 2002a. Numerical simulation of single-phase and multiphase
Kadlec HR, Knight LR. 1996. Treatment Wetlands. Lewis Publishers. non-Darcy flow in porous and fractured reservoirs. Transport in Porous
Kirkham CE. 1966. Turbulent flow in porous media—an analytical and Media 49: 209– 240.
experimental model study. PhD thesis, University of Melbourne, Wu YS. 2002b. An approximate analytical solution for non-Darcy flow
Melbourne, Australia. toward a well in fractured media. Water Resources Research 38(3):
Koch DL, Ladd AJC. 1997. Moderate Reynolds number flows through 5–7.
periodic and random arrays of cylinders. Journal of Fluid Mechanics Yanuka M, Dullien FAL, Elrick DE. 1986. Percolation processes and
349: 31–66. porous media. I. Geometrical and topological model of porous media
Kornecki TS, Sabbagh GJ, Storm DE. 1999. Evaluation of runoff, using a three-dimensional joint pore distribution. Journal of Colloidal
erosion and phosphorus modeling system–SIMPLE. Journal of the Interface Science 112: 24.
American Water Resources Association 35(4): 807–820. Zissis T, Terzidis G. 1991. Unsteady non-Darcy flow in fractured
Kotsovinos NE. 2003. Fluid Mechanics. Democritus University of aquifers. In Proceedings of the European Conference Advances in Water
Thrace: Xanthi (in Greek). Technology, 20–23 March, Athens, Greece; 185– 194.
Kovács G. 1981. Seepage Hydraulics, Development in Water Sciences.
Elsevier: New York.

Copyright  2006 John Wiley & Sons, Ltd. Hydrol. Process. 21, 534–554 (2007)
DOI: 10.1002/hyp

You might also like