Computational modeling of ow and sediment transport

and deposition in meandering rivers

Mehrzad Shams
a
, Goodarz Ahmadi
a,
*
, Duane H. Smith
b
a
Department of Mechanical and Aeronautical Engineering, Clarkson University, Box 5725 Camp Building, Potsdam, NY 13699-5725, USA
b
National Energy Technology Laboratory, U.S. Department of Energy, Morgantown, WV 26507-0880, USA
Abstract
A computational modeling analysis of the ow and sediment transport, and deposition in meandering-river models was per-
formed. The Reynolds stress transport model of the FLUENT
TM
code was used for evaluating the river ow characteristics, in-
cluding the mean velocity eld and the Reynolds stress components. The simulation results were compared with the available
experimental data of the river model and discussed. The Lagrangian tracking of individual particles was performed, and the
transport and deposition of particles of various sizes in the meandering river were analyzed. Particular attention was given to the
sedimentation patterns of dierent size particles in the river-bend model. The ow patterns in a physical river were also studied. A
Froude number based scale ratio of 1:100 was used, and the ow patterns in the physical and river models are compared. The result
shows that the mean-ow quantities exhibit dynamic similarity, but the turbulence parameters of the physical river are dierent from
the model. More strikingly, the particle sedimentation features in the physical and river models do not obey the expected similarity
scaling.
Published by Elsevier Science Ltd.
1. Introduction
Understanding the mechanisms that control sediment
transport in rivers is of fundamental importance to the
elds of hydraulics, hydrology, and water resources.
Flows in river are, generally, three-dimensional, un-
steady, in a state of turbulent motion, and involve
interactions of dierent phases. Therefore, accurate
analysis of ow and sediment transport in a meandering
river is a rather dicult task. While numerous books
and papers on the subject have been published, details of
the micro-mechanics of sediment resuspension, trans-
port, and deposition are not fully understood.
The traditional approaches for studying river ows
are based on eld measurements, laboratory exper-
iments, and simplied depth-averaged computer models.
Field measurements are rather tedious and expensive.
There are also problems associated with the laboratory
model studies. The laboratory models hardly ever satisfy
the principal of dynamic similarity with the original
physical systems, due to disparity of one or more of the
dominant nondimensional parameters. Therefore, direct
extrapolation of the ow condition from laboratory
data to eld applications is not always possible. At-
tempts to relate sediment transport in rivers and corre-
sponding model studies are even more questionable.
Furthermore, it is rather dicult from eld measure-
ments and laboratory studies to provide a detailed
physical understanding of the micro-mechanics of indi-
vidual particle transport and deposition processes.
River hydraulics and sediment transport are widely
studied subjects. Extensive reviews of earlier works were
provided by Chow [1], Vanoni [2], French [3], Chang [4],
Yalin [5,6] and Nezu and Nakagawa [7], among others.
Here, recent advances in numerical modeling of river
hydraulics are summarized; and this is followed by a
review of the recent computational modeling of sedi-
ment transport. Recent developments in related elds of
particle resuspension, transport, and deposition in tur-
bulent ows are also briey described.
In most earlier studies, a depth-averaging technique
was used to reduce the river ow to a two-dimensional
problem. Here, only the works that treat the full three-
dimensional ow (not based on depth-averaging) are
outlined. Flow in straight, open channels was considered
Advances in Water Resources 25 (2002) 689699
www.elsevier.com/locate/advwatres
*
Corresponding author. Tel.: +1-315-268-6536; fax: +1-315-268-
6438.
E-mail address: ahmadi@clarkson.edu (G. Ahmadi).
0309-1708/02/$ - see front matter Published by Elsevier Science Ltd.
PII: S0309- 1708( 02) 00034- 9
by a number of investigators. Rastogi and Rodi [8],
Alfrink and Rijin [9], and Gibson and Rodi [10] used the
two-equation and Reynolds stress transport turbulence
models and computed three-dimensional ows in open
channels. Leschziner and Rodi [11] and Demuren and
Rodi [12] employed a partially parabolic version of the
averaged NavierStokes equations and the ke turbu-
lence model (while neglecting streamwise diusion) to
study, respectively, the ow through a 180 bend and a
meandering channel.
Meselhe et al. [13] employed the fully elliptic form of
the governing equations formulated in generalized cur-
vilinear coordinates to simulate the ow through a
meandering channel with a trapezoidal cross-section.
Demuren [14] reported his computational study for ow
through a natural-like cross-section river. He used a -
nite volume numerical method for solving the full
Reynolds averaged NavierStokes (RANS) equations
and the ke turbulence model. Ambrosi et al. [15] de-
scribed their numerical simulation of the ow at the
delta of the Po River. They used a nite element ap-
proach for their modeling, but restricted their study to
a two-dimensional ow analysis. Recently Sinha et al.
[16] presented a three-dimensional numerical model for
simulating ow through a natural river. They included
the large-scale bed roughness and islands using a
boundary-tted mesh and also used a two-point wall
function approach. Their model predictions for a 4 km
stretch of the Columbia River showed good agreement
with the laboratory and eld measurements.
Evaluation of bed-load transport has a long history
starting with DuBoys [17], Shields [18], and Einstein
[19]. Reviews of more recent works were provided by
Chow [1], French [3], Chang [4], Yalin [5,6] and Nezu
and Nakagawa [7], among others. A state-of-the-art
review of river mixing and sediment dispersion was
published by Elhadi et al. [20]. Most of the earlier eort
was focused on developing semi-empirical expression
for bed-load transport. A recent application of this ap-
proach was presented by Yang et al. [21], who studied
sediment transport in the Yellow River in China.
While the semi-empirical models have been used ex-
tensively in engineering design, the details of processes
that control the sediment transport are not fully under-
stood. Clearly, sediment transport involves two-phase,
liquidsolid ows and their interactions. McTigue [22]
noted the need for using a two-phase ow model for
sediment transport analysis. Accordingly, suspended
particles are treated as a continuous second phase that
interacts with the uid phase. Kobayashi and Seo [23]
and Cao et al. [24] presented their analyses of sediment
transport based on a two-phase ow model that in-
cluded a bed-load layer. Cao and Ahmadi [25] per-
formed an investigation of sediment-laden ow in an
open channel using a two-phase mixture model, and
presented an approach for evaluating the sediment
concentration. Recently Lopez and Garcia [26] studied
the ow through an open channel with vegetation. They
also used a two-uid approach and the ke turbulence
model.
Extensive reviews of the literature on earlier work
related to two-phase ow models were provided by Soo
[27], Ishii [28], Hetsroni [29], and Ahmadi [30]. Recently,
more-advanced models for turbulent two-phase ows
were reported by Ahmadi and co-workers [3133].
Computational modeling of particulate and two-phase
ows were reported by Cao et al. [34] and Cao and
Ahmadi [35,36], among others.
The presented survey shows that the current ad-
vanced procedure for analyzing sediment transport uses
a dispersion model that treats the particles as a trans-
ferable scalar. Even the recently proposed two-uid
models cannot account for the full interactions of par-
ticles with the turbulence uctuations in rivers or the
detailed physics of sediment deposition processes. Fur-
thermore, the micro-mechanics of resuspension of sedi-
mentary particles is not fully understood and has not
been included in the available models. On the other
hand, there has been considerable recent progress in
computational modeling of particle deposition, trans-
port, and resuspension processes in related elds. Li and
Ahmadi and coworkers [3739] developed a computa-
tional model for simulating the turbulent deposition of
particles in complex passages using a Lagrangian par-
ticle trajectory analysis procedure. The approach was
extended and used by Ahmadi and Smith [40,41] for
analyzing particle transport, and deposition during hot-
gas ltration. The particle resuspension processes in
turbulent ow were recently studied by Soltani and
Ahmadi [4245], among others.
In this work, the ow and the sediment transport and
deposition in meandering-river models are studied. A
river model that is identical to the one used in the ex-
perimental investigation of Shiono and Muto [46] is
studied rst. This is followed by the analysis of a
physical-scale meandering-river model with a Froude
number the same as the laboratory-scale model. Using
the FLUENT
TM
code, the mean-ow properties, tur-
bulence intensities, and sedimentation patterns for both
the laboratory and the physical rivers are evaluated. The
computational results for the laboratory-scale model are
compared with the experimental data of Shiono and
Muto [46] and discussed. Motions of individual particles
and their deposition patterns in the laboratory and
physical-scale rivers are analyzed using a Lagran-
gian particle trajectory analysis procedure. In particu-
lar, the sedimentation patterns of particles of dierent
sizes under various conditions in the river-bends are
analyzed. These simulation results show that the
mean velocities satisfy the principle of dynamic simi-
larity. The secondary ow pattern, the turbulence in-
tensity and, more importantly, the particle deposition
690 M. Shams et al. / Advances in Water Resources 25 (2002) 689699
patterns, however, seem to defy the expected similarity
scaling.
2. Flow simulation
Since the ow in the river is in a state of turbulent
motion, it is important to use an appropriate turbulence
model for evaluating the mean-ow eld. The FLU-
ENT
TM
code provides options for using either the ke or
the Reynolds stress transport model (RSTM), which is a
simplied version of the one developed by Launder et al.
[47]. While the ke model is widely used in industrial
applications, it suers from several shortcomings. Its use
of an isotropic eddy viscosity limits its applicability and
causes the model to be incapable of handling the tur-
bulence normal-stress eects. The RSTM, however, ac-
counts for the evolution of individual turbulence stress
components, and is well suited for handling anisotropic
turbulence stresses. In the present study, the RSTM of
the FLUENT
TM
code was used in the simulations.
(Additional details of the ow simulation features
and the computational schemes may be found in the
FLUENT
TM
Users Guide [48].)
3. Mean-ow river model
For an incompressible uid ow, the equations of
continuity and balance of momentum for the mean
motion are given as
ouu
i
ox
i
0 1
ouu
i
ot
uu
j
ouu
i
ox
j

1
q
opp
ox
i
m
o
2
uu
i
ox
j
ox
j

o
ox
j
R
ij
2
where uu
i
is the mean velocity, x
i
is the position, t is the
time, pp is the mean pressure, q is the constant mass
density, m is the kinematic viscosity, and R
ij
u
0
i
u
0
j
is the
Reynolds stress tensor. Here, u
0
i
u
i
uu
i
is the ith uid
uctuation velocity component.
The RSTM provides for dierential transport equa-
tions for evaluation of the turbulence stress components.
i.e.,
o
ot
R
ij
uu
k
o
ox
k
R
ij

o
ox
k
m
t
r
k
o
ox
k
R
ij

R
ik
ouu
j
ox
k

R
jk
ouu
i
ox
k

C
1
e
k
R
ij

2
3
d
ij
k

C
2
P
ij

2
3
d
ij
P

2
3
d
ij
e 3
where the turbulence production terms are dened as
P
ij
R
ik
ouu
j
ox
k
R
jk
ouu
i
ox
k
; P
1
2
P
ij
4
with P being the uctuation kinetic energy production.
Here m
t
is the turbulent (eddy) viscosity; and r
k
1:0,
C
1
1:8, C
2
0:6 are empirical constants [47].
The transport equation for the turbulence dissipation
rate, e, is given as
oe
ot
uu
j
oe
ox
j

o
ox
j
m

m
t
r
e

oe
ox
j

C
e1
e
k
R
ij
ouu
i
ox
j
C
e2
e
2
k
5
In Eq. (5), k
1
2
u
0
i
u
0
i
is the uctuation kinetic energy, and
e is the turbulence dissipation. The values of constants
are
r
e
1:3; C
e1
1:44; C
e2
1:92: 6
The RSTM of the FLUENT
TM
code and the standard
wall function boundary condition were used for evalu-
ating the mean velocity eld and the Reynolds stress
components in the river.
4. Fluctuating velocities simulation
The dispersion of small particles is strongly aected
by the instantaneous uctuation uid velocity. The
turbulence uctuations are random functions of space
and time. Here, a discrete random walk (DRW) model is
used for evaluating the instantaneous velocity uctua-
tions. The values of u
0
, v
0
and w
0
that prevail during the
lifetime of the turbulent eddy, T
e
, are sampled by as-
suming that they obey a Gaussian probability distribu-
tion. That is, the instantaneous velocity in the ith
direction is given as
u
0
i
f

u
0
i
u
0
i

7
In Eq. (7) f is a zero-mean, unit-variance, normally
distributed, random number;

u
0
i
u
0
i

is the local root-

mean-square (RMS) uctuation velocity in the ith direc-
tion; and the summation convention on i is suspended.
The characteristic lifetime of the eddy is dened as a
constant given by
T
e
2T
L
8
where T
L
is the eddy turnover time given as T
L

0:15k=e. The other option allows for a lognormal
random variation of eddy lifetime that is given by
T
e
T
L
logr 9
where r is a uniform random number between 0 and 1.
The particle is assumed to interact with the uid uc-
tuation eld, which stays xed over the eddy lifetime.
When the eddy lifetime is reached, a new value of the
instantaneous velocity is obtained by introducing a new
value of f in Eq. (7).
M. Shams et al. / Advances in Water Resources 25 (2002) 689699 691
5. Particle equation of motion
The equation of motion of a small particle, including
the eects of nonlinear drag and gravitational forces, is
given by
du
p
i
dt

3mC
D
Re
p
4d
2
S
u
i
u
p
i
g
i
10
and
dx
i
dt
u
p
i
: 11
Here, u
p
i
is the velocity of the particle and x
i
is its po-
sition, d is the particle diameter, S is the ratio of particle
density to uid density, g
i
is the acceleration of gravity,
and m is the mass of the particle.
The rst term on the right-hand side (RHS) of Eq.
(10) is the drag force due to the relative slip between the
particle and the uid. The drag force is, generally, the
dominating force. According to Hinds [49], the drag
coecient, C
D
, is given as
C
D

24
Re
p
for Re
p
< 1 12
and
C
D

24
Re
p
1

1
6
Re
2=3
p

for 1 < Re
p
< 400; 13
where Re
p
is the particle Reynolds number dened as
Re
p

d u
j
u
p
j

m
14
The relaxation time of the particles is dened as
s
qd
2
18l
15
The relaxation time is nondimensionalized on the basis
of the river length-scale and velocity scale:
s

sU
0
H
16
where U
0
is the average inlet velocity and H is the river
height.
Eq. (10) includes all the relevant forces and forms the
basis for the discrete second-phase analysis of the
FLUENT
TM
code that was used in the present compu-
tation. The particle equation of motion used requires
knowledge of the instantaneous turbulent uid velocity
at the location of each particle at every instance of time.
As noted before, the mean liquid velocity was evaluated
by the use of the Reynolds stress transport turbulence
model (RSTM) and the uctuation velocity components
were calculated form Eq. (7).
6. Results
Shiono and Muto [46] performed a series of detailed
experimental measurements of the ow conditions in a
laboratory-scale meandering river model using a laser-
Doppler anemometer. They reported the cases of main-
channel and over-bank ows. In the present study, the
river model geometry studied is the same as that in
the experimental measurements of [46] for the ow in
the main channel. The river is assumed to have a lon-
gitudinal slope of 0.001. The meandering shape of the
river is modeled by connecting 60 circles with internal
radius of 35 cm and straight segment of 37.6 cm long as
shown in Fig. 1. Sections 13 at the river-bend are
identied in Fig. 1a, for which details of the secondary
ow patterns are studied. At the third bend (Section 3),
the riverbed is divided into 10 equal-area zones for
particle deposition analysis. The details of the zones
used for sedimentation analysis in the river-bend are
shown in Fig. 1b. The cross-section of the laboratory
river model, which is 15 cm wide rectangle with a height
of 5 cm, is shown in Fig. 1c.
The ow patterns for the physical (natural-scale) river
are also evaluated. The Froude number scaling is used
to relate the parameters of the laboratory model and the
physical river. The Froude number is identied as
F
V
2
gL
17
where V is the mean velocity, L is a length-scale (water
depth) and g is the acceleration of gravity. The physical
(natural) river is assumed to be 100 times larger than the
laboratory model. That is, the width and depth of
the physical river are 15 and 5 m, respectively. To keep
the Froude number for the laboratory and the physical
river the same, the velocity ratio then becomes 1:10.
A 40 20 400 staggered rectangular grid is gener-
ated with the Gambit code for analyzing the ow in the
river (FLUENT
TM
, 1998). Fig. 2 shows the details of
the computational grid for the meandering-river model.
The grid is mostly uniform; the distance between the
rst node and the wall is 1.2 mm, and the longitudinal
distance between the nodes is 3 mm.
The free surface was treated as a symmetry boundary
condition. (Thus, the shear stress and all uxes becomes
zero at the free surface.) The SIMPLE algorithm of the
FLUENT
TM
Code was also used for solving the dis-
cretized equations. As noted before, the RSTM was used
in the simulation, and the computation was continued
until the solution converged with a total relative error of
less than 0.0005.
6.1. Laboratory river model
Laboratory-scale model results for an inlet water
velocity of 0.2 m/s are presented in this section. This
692 M. Shams et al. / Advances in Water Resources 25 (2002) 689699
corresponds to one of the cases studied experimentally
by Shiono and Muto [46]. The ow Reynolds number,
based on the water depth, is about 10
4
. The velocity
magnitude contours on the free surface are shown in
Fig. 3. Here the ow is from right to left. This gure
shows that the contour patterns are rather complex. The
details of the velocity pattern are shown in Fig. 3b. It is
observed that the peak velocity occurs near the inner
part of the bend on the approaching ow section. The
velocity then decreases away from these regions.
To study the nature of the secondary ows in the
river-bend, in-plane ows at Sections 13 and the outlet
section of the laboratory-scale model were carefully
examined, and the results are shown in Fig. 4. These
sections are at dierent bends as illustrated in Fig. 1a.
The inner and outer parts of the bend are also identied
in the gure for clarity. As noted before, the mean
velocity is 0.2 m/s.
Fig. 4 shows that there are noticeable vortical mo-
tions at the river-bend. At Sections 1 and 3, the vortex at
the outer edge is clockwise, while the one near the bot-
tom of the inner edge is counter clockwise. The direc-
tion of the vortices reverses at Section 2 and outlet
section. As a result, at the bottom of the channel there is
always a movement from the outer edge toward the
inner edge. The existence of such motion was conjec-
tured by a number of researchers, although many pos-
tulated the presence of a single vortical motion at the
river-bends. In their laboratory experiments, Shiono
and Muto [46] observed similar multi-vortical mo-
tions. Close comparison shows that the general fea-
tures of the simulated secondary ow are similar to
those reported [46], but there are some quantitative
Fig. 1. (a) Schematics of the meandering-river model Layout. (b) Details of the zones 110 in the third bend. (c) River cross-section.
Fig. 2. (a) Computational grid on the free surface for a segment of the
river. (b) Zoomed grid near the bend.
M. Shams et al. / Advances in Water Resources 25 (2002) 689699 693
dierences on the locations and strength of the vortical
motions.
The velocity magnitude contour plots at Sections 13
and the outlet section are shown in Fig. 5. It is observed
that the velocity magnitude contours at dierent bends
of the river model are comparable. The peak velocity
reaches about 0.3 m/s, and its location is near the
channel central part but closer to the inner part of the
bend. The signicant dierence between the peak and
Fig. 3. (a) Velocity magnitude contours on the free surface of the river model. (b) Enlarged region at the third river bend.
Fig. 4. Secondary ow velocity vector plot at dierent sections of the
river model for an inlet velocity of 0.2 m/s.
Fig. 5. Velocity magnitude contours at dierent sections of the river
model for an inlet velocity of 0.2 m/s.
694 M. Shams et al. / Advances in Water Resources 25 (2002) 689699
the mean velocities indicates the ow does not resemble
a fully developed turbulent ow condition.
The contours of excess pressure (total pressure be-
yond hydrostatics) at Sections 13 and the outlet section
of the river model are shown in Fig. 6. It should be
emphasized that the hydrostatic pressure is excluded
from the pressure shown in this gure. The value of the
excess pressure is decreasing in the streamwise direction.
The maximum value of the excess pressure is near the
mid-section of the channel but toward inner walls.
Fig. 7 shows the turbulence intensity contours at
Sections 13 and the outlet section. Here the turbulence
intensity is dened as the ratio of the RMS of the
velocity uctuations to the mean-ow velocity. The tur-
bulence intensity is relatively low of the order of 0.02
0.04, and remains roughly the same at dierent bend
sections. This trend is in general agreement with the data
reported by Shiono and Muto [46].
6.2. Physical-scale river
Computer simulations also are performed for a rect-
angular cross-section natural river. As noted before, a
length-scale ratio of 1:100 is used. Base on the Froude
number scaling, the inlet velocity of the physical river is
assumed to be 2 m/s to maintain dynamic similarity with
the laboratory model with a velocity of 0.2 m/s. The
Reynolds number of the physical model, however, is 10
7
compared with 10
4
for the laboratory model.
The secondary ow patterns for the physical river at
Sections 13 and the outlet section are shown in Fig. 8.
While the general features of the secondary ow are
qualitatively similar to that of laboratory model, there
are quantitative dierences. The directions of the vor-
tical motions in Fig. 8 are the same as those for the
model shown in Fig. 4, but the size and strength of the
vortices are quite dierent. Also, the monotonic motion
from the outer to inner bend near the riverbed is now
distorted with the appearance of small segments for
which the direction of velocity is reversed.
The corresponding velocity magnitude contours are
shown in Fig. 9. This gure shows that the maximum
velocity is 2.3 m/s and occurs near the inner part of the
bend. At high Reynolds numbers, the axial velocity
prole is atter and resembles fully developed turbulent
ows, when compared with those shown in Fig. 5 for the
river model. The maximum value of the velocity also
markedly shifts toward the inner wall. Comparing Figs.
5 and 9 shows that the maximum velocity is shifted more
toward the inner wall in the physical river compared
with the laboratory-scale river model.
Fig. 6. Pressure contours at dierent sections of the river model for an
inlet velocity of 0.2 m/s.
Fig. 7. Turbulence intensity contours at dierent sections of the river
model for an inlet velocity of 0.2 m/s.
M. Shams et al. / Advances in Water Resources 25 (2002) 689699 695
The excess pressure contours at dierent sections of
the physical (natural-scale) river are shown in Fig. 10.
This gure indicates that the excess pressure contours at
Sections 13, and the outlet section have a much more
complex pattern compared with those for the river
model shown in Fig. 6. In particular, a sharp pressure
variation near the inner wall appears in Fig. 10. The
excess pressure also decreases along the ow direction
due the xed free-surface assumption used in the model.
The turbulence intensity contours at Sections 13 and
the outlet section of the physical river with an average
velocity of 2 m/s are shown in Fig. 11. As in the river
model of Fig. 7, the turbulence intensity has a similar
pattern at dierent section and is not changing along the
ow. The maximum intensity of turbulence here is 0.2,
compared to 0.03 in the river model with the same
Froude number.
6.3. Particle deposition patterns
To simulate the deposition rates of sediments of dif-
ferent sizes in a meandering river, ensembles of 30,000
particles with a given diameter are initially released from
Section 2 (as shown in Fig. 1) and their trajectories are
analyzed. The initial distribution of particles on the
plane is assumed to be uniform. To reduce the statistical
error, reecting-wall boundary conditions are used for
all walls, except for zones 110 in the third river-bend.
As shown in Fig. 1, zone 1 is in the outer part of the
bend and zone 10 is in its inner part. The capture e-
ciencies (the percentages of the particles that are de-
posited) for dierent size sediments for inlet velocities of
0.2 m/s for the laboratory model, and 2 m/s for the
physical (natural-scale) river are evaluated.
Fig. 12 shows the percentage of deposited particles in
the bend for an inlet velocity of 0.2 m/s in the river
model. For sake of clarity, the deposition patterns of
small particles (140 lm) are shown in Fig. 12a, and
those for large particles (120300 lm) are plotted in Fig.
12b. This gure shows that most particles are deposited
near the inner bend of the laboratory river model. Small
particles less than a few microns remain suspended in
the ow and leave through the outlet with little depo-
sition. This gure shows that the deposition rate of
particles larger than 20 lm has a peak at zone 7 (near
the inner part of the bend). The magnitude of the peak is
about 2024% for particles smaller than 40 lm. The
peak becomes quite sharp and reaches 4050% for par-
ticles between 160 and 300 lm. For 120 lm particles the
peak is about 20% and is rather broad covering both
zones 6 and 7. The peak deposition rate of 15% for 10
lm particles occurs at the inner wall of the river-bend.
Fig. 8. Secondary ow velocity vector plots at dierent sections of the
physical river for an inlet velocity of 2 m/s.
Fig. 9. Velocity magnitude contours at dierent sections of the phys-
ical river for an inlet velocity of 2 m/s.
696 M. Shams et al. / Advances in Water Resources 25 (2002) 689699
The sediment deposition pattern in the physical river
also is studied. Again 30,000 particles are released at
plane 2 of the natural-scale river, and their deposition
patterns are analyzed. Here particles in the range of
13000 lm are treated.
Percentages of the particles that are deposited on the
physical river-bend for the case that the inlet velocity is 2
m/s are shown in Fig. 13. Fig. 13a shows the deposition
pattern of particles in the size range of 140 lm, and
Fig. 13b shows the pattern for particles in the range of
1203000 lm. The pattern of the particle deposition is
roughly uniform for small particles. The deposited per-
centage is less than 10% for particles of less than 40 lm
in diameter, and is approximately constant in the cross-
section. Fig. 13b shows that the percentage of deposited
particles becomes independent of particle diameter, for
particles larger than 1000 lm. The majority of the par-
ticles do not deposit in the bend, but leave through the
outlet. Comparison of Figs. 12 and 13 show that the
deposition patterns in the physical (natural-scale) river
are quite dierent from those in the river model, even for
particles with the same Stokes number (nondimensional
relaxation time). That is, the sedimentation process
does not satisfy the similarity principal, even though the
Froude number for the river model and the physical
river are kept the same.
7. Conclusions
Fluid ow as well as particle deposition patterns in a
meandering-river model with a cross-section of 15 5
cm
2
, and a physical-size river with a cross-section of
15 5 m
2
were studied. The RSTM turbulence model of
the FLUENT
TM
code was used and the three-dimen-
sional ow elds in the river were evaluated. Particular
attention was given to secondary ows at the river-
bends. A Lagrangian scheme was used for evaluation of
sediment deposition. The Froude number based simi-
larity was used to maintain the dynamic similarity be-
tween the model and the physical river. On the basis
of the results presented, the following conclusions are
drawn:
1. Flows through meandering physical and laboratory
model rivers with a rectangular cross-section were
three-dimensional and involves secondary ows with
multiple vortices.
2. In the laboratory river model, the ow near the bed
was from the outer part of the bend toward the inner
part. The computed secondary ow patterns at the
bends of the river model were comparable to those
observed in the experimental study of Shiono and
Muto [46].
Fig. 10. Pressure contours at dierent sections of the physical river for
an inlet velocity of 2 m/s.
Fig. 11. Turbulence intensity contours at dierent sections of the
physical river for an inlet velocity of 2 m/s.
M. Shams et al. / Advances in Water Resources 25 (2002) 689699 697
3. In the physical (natural-scale) river, there are regions
in which the direction of the secondary ow near the
bed reversed.
4. Comparison of the ow patterns in the physical river
and the laboratory model river with the same Froude
number showed that the secondary ow and turbu-
lence intensity did not scaled properly and did not
obey the (dynamic) similarity rule. This was because
the ow Reynolds numbers in the laboratory model
and the physical river were dierent.
5. The secondary ow vortices formed on the river bend
aected the particle deposition patterns.
6. In the river model the deposition rate proles gener-
ally peaked near the inner part of the bend. The par-
ticle deposition for the physical river, however, was
roughly constant over the bend cross-section.
7. The particle sedimentation patterns did not show
similarity based on the Froude number scaling be-
tween the model and physical rivers.
8. The particle deposition rate was controlled by the
nondimensional relaxation and the ow Reynolds
number.
9. Sedimentation rates of particles with equal nondi-
mensional relaxation times in the laboratory river
model and the physical-scale river are dierent.
Acknowledgements
The support of the Department of Energy, National
Energy Technology Laboratory, is gratefully acknowl-
edged.
References
[1] Chow VT. Open channel hydraulics. New York: McGraw-Hill;
1959.
Fig. 12. Percentages of deposited particles on the bottom wall of the
bend for an inlet velocity of 0.2 m/s in the river model: (a) Small
particles. (b) Large particles.
Fig. 13. Percentages of deposited particles on the bottom wall of the
bend for an inlet velocity of 2 m/s in the physical river: (a) Small
particles. (b) Large particles.
698 M. Shams et al. / Advances in Water Resources 25 (2002) 689699
[2] Vanoni VA. ASCE sedimentation engineering, Manual and
Report on Engineering Practice, No. 54, 1975.
[3] French RH. Open channel hydraulics. New York: McGraw-Hill;
1985.
[4] Chang HH. Fluvial processes in river engineering. New York:
John Wiley; 1988.
[5] Yalin MS. Mechanics of sediment transport. Oxford: Pergamon
Press; 1977.
[6] Yalin MS. River mechanics. Oxford: Pergamon Press; 1992.
[7] Nezu I, Nakagawa H. Turbulence in open-channel ow. Rotter-
dam: A A Balkema; 1993.
[8] Rastogi A, Rodi W. Predictions of heat and mass transfer in open
channels. J Hydr Eng ASCE 1978;104(3):397420.
[9] Alfrink BJ, Rijn LC. Two equation turbulence model for ow in
trenches. J Hydr Eng ASCE 1983;109(7):94158.
[10] Gibson MM, Rodi W. Simulation of free surface eects on
turbulence with a Reynolds stress model. J Hydr Res Delft
1989;27:23344.
[11] Leschziner MA, Rodi W. Calculation of strongly curved open
channel ow. J Hydr Eng ASCE 1979;105(10):1297314.
[12] Demuren AO, Rodi W. Calculation of ow and pollutant
dispersion in meandering channels. J Fluid Mech 1986;172(1):
6392.
[13] Meselhe EA, Sotiropoulos F, Patel VC. Three dimensional
numerical model for open channels. In: Proc Int Conf on
Hydropower, Hydropower 1995. ASCE; 1995. p. 2102120.
[14] Demuren AO. A numerical model for ow in meandering
channels with natural bed topography. Water Resour Res
1993;l29(4):126977.
[15] Ambrosi D, Corti S, Pennati V, Saleri F. Numerical simulation of
unsteady ow at Pr River Delta. J Hydraul Eng 1996;122:73543.
[16] Sinha SK, Sotiropoulos F, Odgaarrd AJ. Three dimensional
numerical model for river ow through natural rivers. J Hydr Res
1998;124(1):1324.
[17] DuBoys H. Le Rhone et les Rivieres a Lit Aouillable. Annales
des Ponts et Chaussees 1879;18(series 5):14195.
[18] Shields A. Anwendung Aenlich Keitsmechanik und der Turbu-
lenzfor-schung auf Die Geschiebebewegung, Mitteilungen de
reussischen Versuchsastalt fuer Wasserbau und Schibau, Berlin,
Germany, 1936.
[19] Einstein HA. Formulas for the transportation of bed-load. Trans
ASCE 1942;107:56173.
[20] Elhadi N, Harrington A, Hill I, Lau YL, Krishnappan BG. River
mixinga state of the art report. Can J Civ Eng 1983;11:1984.
[21] Yang C, Molians A, Wu B. Sediment transport in Yellow River.
J Hydr Eng 1996;122(4):23744.
[22] McTigue DF. Mixture theory for suspended sediment transport.
J Hydr Div 1981;107(9):65973.
[23] Kobayashi N, Seo SN. Fluid and sediment interaction over a
plane bed. J Hydr Eng 1985;111(6):90321.
[24] Cao Z, Wei L, Xie J. Sediment laden ow in open channels from
two phase ow viewpoint. J Hydr Eng 1996;121(10):72535.
[25] Cao J, Ahmadi G. Two-phase solidliquid mixture ows down
inclined chutes, to appear.
[26] Lopez F, Garcia MH. Open channel ow through simulated
vegetation suspended sediment transport modeling. Water Resour
Res 1998;34(9):234152.
[27] Soo SL. Fluid dynamics of multiphase systems. Waltham, MA:
Blaisdell; 1980.
[28] Ishii M. Thermo-uid dynamics of two-phase ow. France:
Eyrolles; 1975.
[29] Hetsroni G. Handbook of multiphase systems. New York:
Hemisphere Publ; 1982.
[30] Ahmadi G. A continuum theory for two phase media. Acta Mech
1982;44:299317.
[31] Ahmadi G, Ma D. A thermodynamical formulation for dispersed
multiphase turbulent ows. Int J Multiphase Flows 1990;16:
32340.
[32] Ma D, Ahmadi G. A thermodynamical formulation for dispersed
multiphase turbulent ows, part II: simple shear ows for dense
mixtures. Int J Multiphase Flows 1990;16:34151.
[33] Abu-Zaid S, Ahmadi G. A thermodynamically consistent rate-
dependent model for turbulent two phase ows. Int J Nonlinear
Mech 1995;30:50929.
[34] Cao J, Ahmadi G, Massoudi M. Granular ows of slightly
frictional particles down an inclined bumpy chute. J Fluid Mech
1996;316:197221.
[35] Cao J, Ahmadi G. Gas-particle two phase turbulent ow in a
vertical duct. Int J Mutiphase Flow 1995;21:120328.
[36] Cao J, Ahmadi G. Gas-particle two-phase ow in horizontal and
inclined ducts. Int J Eng Sci 38, 196181.
[37] Li A, Ahmadi G. Deposition of aerosols on surfaces in a turbulent
channel ow. Int J Eng Sci 1993;31:43551.
[38] Li A, Ahmadi G, Bayer RG, Gaynes MA. Aerosol particle
deposition in an obstructed turbulent duct ow. J Aerosol Sci
1994;25:91112.
[39] Li A, Ahmadi G, Gaynes MA, Bayer RG. Aerosol particle
deposition in a recirculating region. J Adhes 1995;5:10523.
[40] Ahmadi G, Smith DH. Particle transport and deposition in a hot-
gas cleanup pilot plant. Aerosol Sci Technol 1998;29:183205.
[41] Ahmadi G, Smith DH. Gas ow and particle deposition in the
hot-gas lter vessel at the Tidd 70 MWE PFBC demonstration
power plant. Aerosol Sci Technol 1998;29:20623.
[42] Soltani M, Ahmadi G. On particle adhesion and removal
mechanisms in turbulent ows. J Adhes Sci Technol 1994;7:763
85.
[43] Soltani M, Ahmadi G. On particle removal mechanisms under
basic acceleration. J Adhes 1994b;44:16175.
[44] Soltani M, Ahmadi G, Bayer RG, Gaynes MA. Particle detach-
ment mechanisms from rough surfaces under substrate accelera-
tion. J Adhes Sci Technol 1995;9:45373.
[45] Soltani M, Ahmadi G. Direct numerical simulation of particle
entrainment in turbulent channel ows. Phys Fluid A 1995;7:647
57.
[46] Shiono K, Muto Y. Complex ow mechanisms in compound
meandering channels with overbank ow. J Fluid Mech 1998;
376:22161.
[47] Launder BE, Reece GJ, Rodi W. Progress in the development of a
Reynolds stress turbulent closure. J Fluid Mech 1975;68:5378.
[48] FLUENT User Guide, Fluent Incorporated, Centerra Resources
Park, 10 Cavendish Court, Lebanon, NH 03766, 1998.
[49] Hinds WC. Aerosol technology, properties behavior, and
measurement of airborne particles. New York: John Wiley and
Sons; 1982.
M. Shams et al. / Advances in Water Resources 25 (2002) 689699 699