Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1, pp.

15-24 (2007)

NUMERICAL ANALYSIS OF BRAIDED RIVERS

AND ALLUVIAL FAN DELTAS
Chang-Lae Jang * and Yasuyuki Shimizu **

* Korea Institute of Water and Environment, KOWCO, 462-1, Jeonmin-dong,

Yuseong-gu, Daejeon, 305-730, Korea
Email: cljang@kowaco.or.kr (Corresponding Author)
** Division of Environmental and Resource Engineering, Hokkaido University, North 13, West 8, Kitaku,
Sapporo, Hokkaido,060-8628, Japan
Email: yasu@eng.hokudai.ac.jp
ABSTRACT: Understanding the processes of morphological behavior in braided rivers and alluvial fan deltas is very
important for river engineering purposes to manage hydraulic structures and prevent disasters from flood, and
environmental purposes to maintain river ecosystem. A two-dimensional numerical model was developed to simulate
braided rivers and alluvial fan deltas with erodible bed and banks. A moving boundary-fitted coordinate system was
employed to calculate water flow, bed change, and bank erosion. CIP (Cubic Interpolated Pseudo-particle) method was
used to calculate flow, which introduced little numerical diffusion. Sediment transport equation in the streamline and
transverse wise, considering the secondary flow, was used to estimate bed and bank evolution in time. Bank erosion was
simulated by following the procedure proposed by Shimizu (2002). Braided river in laboratory was reproduced for
verifying the numerical model in the channel filled with nearly uniform sandy materials. Comparison of numerical
results and experimental data has shown relatively good agreements. The evolution processes of the alluvial fan deltas
were investigated numerically by controlling the sediment inflow, and the model reproduced the features of alluvial fan
deltas, i.e., the successive aggradation, avulsion and then migration of a river.
Keywords: Braided river; Alluvial fan delta; Bank erosion; Numerical model.

1. INTRODUCTION

The morphological changes of rivers are deeply
interrelated to the bed deformation and bank
erosion because of the mutual relationship between
water flow and sediment transport. Better
understanding of these process and mechanism is
very important for river engineering purposes to
manage hydraulic structures and prevent disaster
from flood, and environmental engineering
purposes to maintain river ecosystem.
In the process of channel development, bars emerge
under certain hydraulic conditions as the channel
widens from an initially straight channel, with
erodible bed and banks. Previous investigations
examined the mechanical processes of channels
with erodible banks theoretically (Ikeda et al., 1981;
Parker et al., 1982), and have provided a method to
reproduce lateral changes in the channel.
Recently several numerical models have been
developed to reproduce braided rivers with fixed
banks (Murray and Paola 1994). Numerical models
to reproduce the evolution of meandering channels,
taking bank erosion into consideration, have been
developed for beds and banks made of uniform
sediment (Shimizu et al., 1996; Nagata et al., 2000).
However, these models have some limitations to
calculate braided rivers with unconstrained banks.
Jang and Shimizu (2005a) proposed a numerical
model to simulate braided rivers with erodible
banks, and showed the possibility of simulating
braided rivers considering bank erosion.
Alluvial fan deltas are fan-shaped deposits of
sediment ending in standing water and formed by
aggrading and then migrating or avulsing river
channels due to sediment inflow from upstream
(Parker, 2000; Sun. et al., 2002). The repeated
processes cause the channels to develop a large-
scale morphology in the shape of a fan, a slope that
declines toward standing water and a prograding
front (Sun et al., 2002). A numerical model was
proposed to simulated these processes of the
alluvial fan delta by combining elements of two
approaches i.e. diffusional models of depositional

Received: 12 June 2006; Revised: Nov. 2006; Accepted: 19 Dec. 2006

15
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

basin filling and cellular models of channel

evolution by Sun et al.(2002). However, the model ( )
= g x2 + y2
H
( x x + y y ) H

was too simple to simulate the evolution processes
of the submerged bars in channels.
Cu
d
hJ
( ) (
2
y u y u n + x u x u )
2

The numerical model we developed is capable of

simulating two-dimensional bed elevation changes u u
+ ( t r2 )+ ( t r2 ) (2)
and width variations in braided rivers and alluvial
fan deltas with erodible bed and banks composed of
u u u
non-cohesive materials. A moving boundary-fitted
coordinate system was employed to describe a
+ (t + u )

+ t + u

( )
natural shaped boundary assuming that erosion and
sediment deposition taken place in the cross
+ 4u u + 5u u + 6u u

sectional direction transformed the plane shape of
the water channel into an arbitrary shape. Flow field
(
= g x2 + y2) H
( x x + y y ) H

was calculated by using a high-order Godunov C u
scheme, that is, CIP method (Yabe et al. 1990). A d
hJ
( ) (
2
y u y u + x u x u )
2

partially drying or wetting boundary in the shallow

regions around emerged bars and near side banks
with a moving boundary was dealt. Bank erosion
was simulated by following the procedure proposed
by Shimizu (2002). The model results were
compared with laboratory experiments to
demonstrate its applicability, and the model
simulated well the characteristics of alluvial fan
deltas. As the sediment inflow was increased, the
fan delta height and areas were increased, which
accelerated to the development of the fan delta with
strongly lateral migration and aggradational
avulsion of channels.
2. GOVERNING EQUATIONS
2.1 Water flow
The governing equations, i.e. continuity and
momentum equations, for water flow are
transformed from the Cartesian coordinate
system to a moving boundary-fitted coordinate
system due to the deformation of side banks.
The equations in the moving boundary-fitted
coordinate system are as follows.
Continuity equation:
h
u u
+ ( t r2 )+ ( t r2 ) (3)
In the above equations, and = spatial
coordinates in the boundary-fitted coordinate
system; = time coordinate (= t in the present
model) in the coordinate system, x and y = spatial
coordinate components in the Cartesian coordinate
system; H = water surface elevation(= h + z b ); h =
water depth; z b =bed elevation referred to a
horizontal plane; g = gravitational acceleration;
C d = bed friction coefficient, which is written using
13
Mannings roughness coefficient as gn 2 h ; n =
Mannings roughness coefficient; J = Jacobian of
the coordinate transformation given as
J = t x y + t x y + t x y
( t x y + t x y + t y x ) ; u and u =
contravariant components of flow velocity in the
and directions defined as u = x u + y v and
u = x u + y v ; u and v = depth-averaged
velocity components in x and y directions,

+

(
+ u )
h
+
J
( + u) h
J
=0 respectively; t = depth-averaged diffusion
J
t t
coefficient (= u* h 6 ); = Von Karman
(1)
constant(=0.4); u* = shear velocity
Momentum equations in and directions:
u u u [
(= n g (u 2 + v 2 ) ]1/ 2
h1 / 6 ). The coefficients,

+ (t + u )

(
+ t + u

) 1 ~ 6 , are given in Jang and Shimizu (2005).
+ 1u u + 2u u + 3u u

16
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

2.2 Sediment transport 3.2 Bank erosion and channel migration

To simulate channel change, the two-dimensional The flow and bed changes are calculated using the
sediment continuity equation in a moving governing equations described above. Considering
boundary-fitted coordinate system is used. the bed change near the banks, the deformation in
Two-dimensional sediment continuity equation in a the shape of the channel is calculated in the
moving boundary-fitted coordinate system reads as following way. To calculate bank erosion, when the
zb 1 qb qb bed near the banks is scoured, and the bank slopes
+ + = 0 (4) becomes steeper than the submerged angle of
t J 1 J J repose, we assumed the sediment beyond the
in which z b = bed elevation; = porosity of the bed submerged angle of repose is instantly eroded to
attain the submerged angle of repose (Hasegawa,
material; qb and qb = contravariant components of 1984). Furthermore, we assumed that a sediment
the bed load transport rate per unit width in the load equal to the amount of sediment exceeding the
and directions, which are given by defining s in submerged angle of repose is deposited at the toe of
the bank slope. The bank failure shape is calculated
the direction of the stream line and n in the using a slump model (Pizzuto, 1990), in which the
perpendicular to the streamline: eroded rate is the same as the deposited rate at the
s n surface face, and the surface is reached at the top of
qb = q + q
s n the bank. At this time, if the erosion of the bank
(5) slope face advances toward the top of the slope, the
x y s x y n
= x + y q + x + y q computational range is enlarged in the cross-
s s n n sectional direction of the channel. In addition, the
s n amount of sediment beyond the critical angle of the
qb = q + q
s n submerged bank slope was included in the
(6) computation of bed evolution. Conversely, the inner
x y s x y n
= x + y q + x + y q bank and other parts in the channel, which were
s s n n transformed into land, were excluded from the
s n
where q and q = the bed load transport rate computations. Readers are referred to Jang and
components in the s and n directions, respectively. Shimizu (2005) for more details.

3. NUMERICAL SIMULATIONS 3.3 Boundary conditions

3.1 Numerical method
The sediment transport rate in the stream line is
calculated using the formula of Ashida and Michiue
(1972) as modified by Hasegawa (2000), to explain
the effect of the bank slope, taking into accounting
the gravitational effect on the side bank proposed
by Kovacs and Parker (1994). As a numerical
scheme, the cubic interpolated psuedoparticle(CIP)
method proposed by Yabe et al.(1990) is used in the
advection terms of the momentum equation and the
central difference method is used in diffusion terms.
The numerical method solves boundary problems
while introducing little numerical diffusion, and
algorithm implementation is more straightforward
than for other high-order upwind schemes (Yabe
and Aoki, 1991; Yabe et al., 1991; Jang and
Shimizu, 2005).
Shear stress at the side walls was calculated by
estimating friction velocity using the parallel
velocity component (= u) considering the
logarithmic velocity distribution with the walls at
grids of the near wall.
Dealing with a partially drying or wetting boundary
is important for simulating the inundation of flood
plains and wetlands, shallow flows over and near
islands or emerged bars in rivers, and wave run-up
in a coastal line (Bradford and Sanders, 2002; Zhao
et al., 1994). Referring to Fig. 1, the boundary in
the shallow regions around emerged bars and near
side banks with a moving boundary is dealt with as
follows;

17
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

and 0.0035 m 3 / s under the initial conditions. The

Fig. 1. Scheme of boundary near the emerged bars and
banks in a staggered grid system: (a) dh/d < 0 and u > 0,
or dh/d < 0 and v > 0 in the longitudinal or transverse
direction; (b) dh/d > 0 and u < 0, or dh/d > 0 and v < 0
in the longitudinal or transverse direction.
In the flow direction, near the emerged bars:
dh
If < 0 and u i , j > 0 , then u i , j = 0 (7)
d flow velocity was measured by taking a photograph
dh
If > 0 and u i , j < 0 , then u i , j = 0 (8)
d which were put into the flow at the upstream before
In the transverse direction, near emerged bars and
side banks:
dh
If < 0 and vi , j > 0 , then vi , j = 0 (9)
d
dh
If > 0 and vi , j < 0 , then vi , j = 0 (10)
d outlet effect at the downstream of the channel. Bed
Since hmin has a very small value, but there is flow
velocity in the stream-wise or transverse directions
in these situations, the momentum equations are not
solved and the flow velocity is set to zero, which
probably leads to some numerical error, and must
be studied further in the future. In these equations,
hmin is set to 0.02mm to minimize the error. In the
fully wetted regions, the continuity and momentum
equations are solved using the CIP method in the
advection terms.
non-dimensional channel (width/depth ratio) was
59.1 and 85.3, respectively. Experimental and
calculation conditions are summarized at Table 1.
Sediments were given by hand regularly at the
upstream end of the channel to maintain the channel
shape without local disturbance during the
experiments. The bed was graded to the intended
slope by using a wide scraper and the initial channel
was cut by a wooden prototype attached to the
lower end of the scraper at the same time. Before
starting the experiments, a small discharge was set
to flow over the bed to saturate the bed surface.
Water depth was checked at some places using a
point gage after 15 minutes of water flow. Surface
using a high-speed video camera at the places.
Light particles were used as photographic objects,
collecting the data by the camera. Sediment
discharge was collected using a sand trapper at the
downstream end of the channel. Plastic screen was
set to maintain nearly steady state inflow into the
channel at the front of the channel inlet, and water
depth was regulated by sluice gate to minimize the
configuration was measured using a laser bed
profiler without water flow.

4. COMPARIONS BETWEEN
EXPERIMENTAL AND NUMERICAL
RESULTS

4.1 Experimental procedure

Laboratory experiments were carried out in a flume
with 12 m in length and 2 m in width, and sidewalls
of the flume were wooden plates. The flume was
filled with well sorted-sand with 1.25 mm of mean
diameter. An initial channel was set to trapezoidal
shape with 80 cm at the bottom, 3 cm at the height,
and 40 degree at the bank slope as shown in Fig.2.
Water discharge was controlled to 0.0045 m 3 / s
Fig. 2. Sketch of experimental setup
Although the channel length was 12 m, the
upstream 2 m of the channel was excluded from the
experimental reach since the bed and side banks
were not changed due to the effects of inlet of the
channel. The longitudinal bed profile was measured

18
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

Table 1. Experimental and numerical conditions

Run Mean diameter of Initial Water Initial Width/ Reynolds Froude Friction
the bed material Slope discharge water depth Depth Number number Velocity
(mm) (%) ( m3 / s ) (cm) (m/s)
1 1.25 1.0 4.5 1.41 59.1 6852 1.3 0.037

2 1.25 1.5 3.5 0.93 89.9 4194 1.5 0.030

from 1 m at the upstream to 8m at every 20 cm in banks are eroded. And also the mechanism of
the experimental reach since the upstream 1 m was braided rivers under large aspect ratio explained by
not considered to avoid the effects of inlet and Ashmore (1991) was illustrated for Run-2. Flow
disturbance of water profile due to the sediment was concentrated into lower channels, and some of
supply. The transverse bed profile was surveyed them were scoured deeply (Fig. 4(b)). Bars
between 0.05 m and 1.95 m at every 1 cm due to the appeared to the downstream of the lower channels
moving limitation of the profiler. and complicated braided bars were developed due
to flow division around the bars, leading to bank
4.2 Numerical results erosion (Fig. 4(c)).
The suggested numerical model is applied to verify
Above -13.75 -25.00 - -21.25 -36.25 - -32.50
the applicability at the actual experimental reach, -17.50 - -13.75 -28.75 - -25.00 -40.00 - -36.25
-21.25 - -17.50 -32.50 - -28.75 Below -40.00 Unit: mm
10m, with erodible banks. Experiment Run-1 was
carried out under the initial condition that the
channel width to depth ratio corresponding to the 1

transition regime (aspect ratio = 59.1) between

y(m)

0 Cal.

alternate bars and braided bars, and Run-2 was done -1

0 2 4 6 8 10
x(m)
corresponding to the condition for braided bras 1

(aspect ratio = 89.9), according to the Kuroki and

y(m)

0 Exp.
Kishi (1984)s regime criteria on bars and braids in -1
0 2 4 6 8 10
alluvial straight channels in the space given by x(m)

(a) T=65min
dimensionless tractive force and the channel width 1

to depth ratio considering the channel slope. Cal.

y(m)

Figs. 3 and 4 show the channel development -1

numerically and experimentally in Run-1 and Run-2, 1
0 2 4
x(m)
6 8 10

respectively. Fig. 3 shows that braided river is

y(m)

0 Exp.
developed from meandering channel because aspect
-1
ratio is increased more and more while bank erosion 0 2 4
x(m)
6 8 10

is advanced on left and right alternately as time is (b) T=95min

progressed. The process and mechanism of braided
1
y(m)

Cal.
river from an initial straight channel with erodible 0

banks investigated by Ashmore (1982) were shown -1

0 2 4 6 8 10
for Run-1. Alternate bars were grown up in the 1
x(m)

straight channel as shown in Fig. 3(a), and then a

y(m)

0 Exp.

meandering channel was developed due to the side -1

0 2 4 6 8 10
banks erosion by the flow deflection of the bars (see x(m)

(c) T=138min
Fig. 3(b)). After that, braided river was developed
since the alternate bars were nearly fixed and
Fig. 3. Comparison of the temporal channel deformation
broken down as flow was diverged as shown in Fig. between calculation and experiment for Run-1
3(c). Fig. 4 displays that a braided river is shown up
from the initial stage and is remarkably developed
while the channel width is enlarged since the side

19
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

Above -13.75 -25.00 - -21.25 -36.25 - -32.50

-17.50 - -13.75 -28.75 - -25.00 -40.00 - -36.25 The temporal change of the channel width in the
-21.25 - -17.50 -32.50 - -28.75 Below -40.00 Unit: mm
calculation for Run-1 is overall in good agreement
with it of the experiment. Fig. 5 displays the
1
satisfactory agreement between the predicted results
y(m)

0
Cal. and the observed data for the bed change in the
-1
0 2 4 6 8 10
cross section for Run-1. However, the predicted
1
x(m)
results for Run-1 in Fig. 5(a) and (c) show that the
deeply eroded bed is taken place in the vicinity of
y(m)

0
Exp.
-1 the left bank, while the bed in the laboratory
0 2 4
x(m)
6 8 10
experiment is eroded deeply near the right bank.
(a) T=64min
1
The channel width for Run-2 is relatively simulated
well with the experiment in Fig. 6(a). As time
y(m)

0
Cal.
increase, however, the channel width for Run-2 is
-1
0 2 4
x(m)
6 8 10 slightly underestimated, and the transverse bed
1
change is overestimated in Fig. 6(b) and (c).
The difference between the numerical and
y(m)

0
Exp.
-1 experiment results is probably due to the fact that
0 2 4
x(m)
6 8 10
the numerical model does not simulate the three-
(b) T=95min
dimensional flow structure at the confluence of the
1
flow, the numerical method used to treat partially
y(m)

0 Cal.
wetting or drying boundaries cannot exactly
-1
0 2 4 6 8 10
account for the experimental flume, and the
1
x(m)
numerical model does not exactly reflect
unavoidable experimental error, e.g., water pump
y(m)

0 Exp.

-1
0 2 4
x(m)
6 8
(c) T=125min
Fig. 4. Comparison of the temporal channel deformation
between calculation and experiment for Run-2
The calculation and experiment results also
reproduce that meandering and braided channel
patterns, which are dependent on the channel width
to depth ratio, namely, braiding pattern is favored if
the bed slope and channel width to depth ratio are
increased, and meandering pattern is favored if the
bed slope and the width to depth ratio are decreased
(Parker 1976; Kuroki and Kishi 1984). The
calculation results are in overall agreement with the
experimental results, although the longitudinal
wavelength and thalweg of the bed in the
calculation are a little difference from those of the
experiment.
Figs. 5 and 6 show the more quantitative
comparison of cross sectional changes at 6m from
the source of the flow for Run-1 and Run-2,
respectively. Since the water depth and flow
velocity were gained at 6 m from the upstream, the
location was selected to compare the cross sectional
changes between computation and experiment for
each run.
conditions and flow perturbation at the entrance of
10
flow when the sand was fed, etc. The assumed
parameters and initial conditions of the numerical
model may be one cause of the differences.
Nevertheless, the numerical results are in relatively
good agreement with the experimental data.
To investigate the applicability of the numerical
model and the evolution processes of alluvial fan
delta, the model was applied to a fan-shaped
channel that was 20 m long, 0.22 m width at the
entrance of upstream and 40.22 m width at the end
of downstream. The evolution processes of alluvial
fan delta were investigated by controlling the
sediment inflow upstream. The hydraulic conditions
for the numerical experiments of the alluvial fan
deltas were explained in detail in Table 2.
The computational time step was 0.01 s and there
were 120 grid points in the streamwise direction
and 30 in the transverse direction. A rectangular
bump, which was 2.45 m wide, 1.3 m long and 1.15
mm high, was located at 4.5 m of the upstream in
the channel bed to encourage the development of
the fan delta.

20
 Engineering Applications of Computational Fluid Mechanics Vol.1, No. 1 (2007)

0 0
-10
-10
z(mm)

-20

z(mm)
Exp.
Cal. -20
-30 Exp.
W.L. Cal.
-40 -30
W.L.
-0.5 0.0 0.5 -40
y(m) -1.0 0.5
-0.5 0.0
y(m)
(a) T=65min
(a) T=64min
0
0
-10
z(mm)

-10

z(mm)
-20 Exp.
Cal. Exp.
-30
W.L. -20 Cal.
-40 W.L.
-30
-0.5 0.0 0.5
y(m) -1.0 -0.5 0.0 0.5
y(m)
(b) T=95min
(b) T=95min
0
0
-10
z(mm)

-20 -10
z(mm)

Exp.
-30 -20
Cal. Exp.
-40 W.L. -30 Cal.
W.L.
-50
-40
-0.5 0.0 0.5
y(m) -1.0 -0.5 0.0 0.5
y(m)
(c) T=138min
Fig. 5. Comparison of cross sectional change between (c) T=125min
calculation and experiment at 6m from the upstream for Fig. 6. Comparison of cross sectional change between
Run-1 calculation and experiment at 6m from the upstream for
Run-2
Table 2. Hydraulic conditions for the numerical
experiments of the alluvial fan deltas Fig. 7 shows the development of the alluvial fan
delta for Run-3. As time increased, the channel was
Run Discharge Slope Fan angle Sediment avulsed and bifurcated into several channels on the
(CMS) (%) (deg.) inflow by alluvial fan delta at the initial stage of the
submerged development (Fig. 7(a)). The depositional areas of
weight the fan were increased with time, and the fan delta
( kg / sec ) migrated toward downstream with forming new
1 0.035 0.2 90.0 7.95 depositional lobes. Channels on the fan migrated
2 0.035 0.2 90.0 39.75 laterally in the downstream, new channels were
3 0.035 0.2 90.0 79.50 generated, whereas old channels were degenerated
in the downstream, which led to re-networked
channels (Fig. 7(b) and (c)). As time progressed,
the channel avulsion and depositional lobes were
developed near the banks (Fig. 7(d)).

21
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)

20
15

Fan Delta Height(cm

)
10 15 R-3
5 R-2
10
y(m)

0
R-1
-5
5
-10

-15
0
0 50 100 150 200
5 10 15 5 10 15 Time(min)
(a) T=70min (b) T=100min Unit : cm (a)
Above 7.55
7.10- 7.55
15 6.65- 7.10 100
6.20- 6.65
5.75- 6.20

2 )
10
80

Fan Delta Area(m

5.30- 5.75
4.85- 5.30
5 4.40- 4.85
3.95- 4.40 R-3
60
y(m)

3.50- 3.95
0 3.05- 3.50
2.60- 3.05
R-2
-5 2.15- 2.60
1.70- 2.15
40
1.25- 1.70
-10 0.80- 1.25
0.35- 0.80 20 R-1
-0.10- 0.35
-15 -0.55- -0.10
-1.00- -0.55
Below -1.00 0
5 10 15 5 10 15 0 50 100 150 200
(c) T=150min (d) T=200min Time(min)
(b)
Fig. 7. Simulated alluvial fan delta for Run-3. Flow is Fig. 8. Characteristics of (a) fan delta height and (b) fan
from left to right. delta areas with time for each run.

Fig. 8 shows the change of the fan delta height and
areas with time. The delta height, the difference
between the maximum aggradated bed elevation
and the minimum scoured bed elevation, was
increased abruptly to 10 minutes, and the delta
height was kept in equilibrium state with a little
fluctuation for Run-1. However, the delta height
was increased in 110 minutes for Run-2, in 80
minutes for Run-3 (Fig. 8(a)). As the sediment
inflow at the upstream of the fan delta was
increased, the starting time of the increasing delta
height was fast.
The alluvial fan delta areas were increased sharply
to 10 minutes in the initial stage of the fan delta
development, kept in constant to 120 minutes with
little change of the fan delta height, and then
increased linearly for Run-1. For Run-2 and 3, the
fan delta areas were increased after the initial stage
of the fan delta development (Fig. 8(b)). The fan
delta areas ware increased with sediment inflow
upstream.
As the sediment inflow was increased, the fan delta
height and areas were increased, which accelerated
to the development of the fan delta with strongly
lateral migration and aggradational avulsion of
channels shown in Fig. 7. Moreover, the channel
gradient was increased (Fig. 8(a)).
Fig. 9 shows the cross-sectional changes at 6m from
upstream with time to investigate the avulsion
processes of channels in the fans with sediment
inflow. The channel bed was aggraded from the
initial channel in the fan and the channels were
migrated laterally as time increased for Run-1 (Fig.
9(a)). As the sediment inflow was increased, the
channel bed was aggraded rapidly and the channels
were migrated laterally. Moreover, the channel
number and the relief were increased for Run-2 (Fig.
9(b)). As the sediment inflow increased excessively,
the central channel in the fan was aggraded and
avulsed into two channels after 150 minutes with
migrating laterally as shown in Fig. 9(c). Therefore,
the channel bed aggradation in the fan due to
sediment inflow excessively is one of factors of
channels avulsion.

22
 Engineering Applications of Computational Fluid Mechanics Vol.1, No. 1 (2007)

150 min.
10
100 min. streamline and the transverse wise, considering the
0
70 min.
34 min. secondary flow, was employed to estimate bed and
bank evolution in time. To simulate bank erosion, it
z(mm)

-10

-20
was assumed that bank was eroded when gradient
Initial bed of bank in the transverse direction was steeper than
-30
the angle of repose because the bed was scoured in
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 the vicinity of the banks induced by secondary flow,
x(m)
leading to channel with natural shaped boundary. At
(a) Run-1
the same time, the amount of bank material beyond
20 150 min. the angle of repose was included to the calculation
100 min.
10 70 min. of the bed evolution as sediment supply. However,
34 min. inner bank and other parts in the channel, which
0
z(mm)

-10
were changed into land, were not included in the
range of computation.
-20
The model was verified by experiments on braided
-30 Initial bed
channels with erodible banks. The calculation
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 results of the longitudinal change in time for Run-1
x(m) and Run-2 were relatively satisfied with the
(b) Run-2 experimental results, although the longitudinal
wavelength and thalweg of the bed in the
150 min. calculation are slightly different from those of the
20 100 min.
70 min. experiment. The channel width was in good
10 34 min.
agreement between calculation and experiment for
z(mm)

0
Run-1, while the width was underestimated for
-10 Run-2 as time increased. The comparison of cross
-20 sectional changes between calculation and
Initial bed
-30 experiment at 6m from upstream for Run-1 and
Run-2, respectively, showed a little difference,
-5 -4 -3 -2 -1 0
x(x)
1 2 3 4 5 6 because of the poorly treated results of the three-
(c) Run-3 dimensional flow structures at the confluence of the
flow, boundaries between the dried and wetted parts
Fig. 9. Cross-sectional changes at 6m from upstream of the bar in the channel, and unavoidable
experimental error.
The numerical model simulated well the features of The numerical model simulated well the features of
the alluvial fan delta, i.e., the successive the alluvial fan delta, i.e., the successive
aggradation, avulsion and then migration of a river aggradation, avulsion and then migration of a river.
(Parker et al., 1998). As the sediment inflow was increased, the fan delta
height and areas were increased, which accelerated
5. SUMMARY AND CONCLUSIONS to the development of the fan delta with strongly
lateral migration and aggradational avulsion of
In this paper, a model applicable to braided rivers channels.
with erodible banks and alluvial fan deltas has been
presented to estimate to the channel evolution REFERENCES
quantitatively.
A moving boundary-fitted coordinate system was 1. Ashida K, Michiue M (1972). Study on
used for natural shaped boundary because the hydraulic resistance and bed-load transport rate
erosion and deposition occurred laterally, and the in alluvial streams. Proc. JSCE 201:59-69 (in
channel shapes were transformed into arbitrary Japanese).
shapes. As a numerical scheme, Cubic Interpolated 2. Ashmore PE (1982). Laboratory modeling of
Pseudo-particle method was used in the flow field gravel braided stream morphology. Earth surf.
since the method introduced little numerical Proc. landforms 7: 201-225.
diffusion. Sediment transport equation in the

23
 Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 1 (2007)
3. Ashmore PE (1991). How do gravel-bed rivers
braid?. Can. J. Sci. 28:326-341.
4. Bradford SF, Sanders BF (2002). Finite-volume
model for shallow-water flooding of arbitrary
topography. J. Hydraul. Engrg. ASCE
128(3):289-298.
5. Hasegawa K (1984). Hydraulic research on
planimetric forms, bed topographies and flow in
alluvial rivers. PhD Dissertation, Hokkaido
University, Sapporo, Japan (in Japanese).
6. Hasegawa K (2000). Hydraulic characteristics
of mountain streams and their practical
application. Lecture notes of the 33rd Summer
Seminar on Hydrau. Engrg., JSCE:A-9-1-20(in
Japanese).
7. Ikeda S, Parker G, Sawai K (1981). Bend
theory of river meanders. Part 1. Linear
development. J. Fluid Mech. 112:363-377.
8. Jang C-L, Shimizu Y (2005a). Numerical
simulation of relatively wide, shallow channels
with erodible banks. J. Hydr. Engrg. ASCE
131(7):565-575.
9. Jang C-L, Shimizu Y (2005b). Numerical
experiments of the behavior of alternate bars
with different bank strength. J. Hydr. Res.
IAHR 43(6):596-612.
10. Kovacs A, Parker G (1994). A new vectorial
bedload formulation and its application to the
time evolution of straight river channels. J.
Fluid Mech. 267:153-183.
11. Kuroki Kishi (1984). Regime criteria on bars
and braids in alluvial straight channels. Proc.
JSCE 342:87-96 (in Japanese).
12. Murray AB, Paola C (1994). A Cellular model
of braided rivers. Nature(London) 371:54-57.
13. Nagata N, Hosoda T, Muramoto Y (2000).
Numerical analysis of river channel processes
with bank erosion. J. Hydr. Engrg. ASCE
126(4):243-252.
14. Parker G (1976). On the cause and
characteristic scales of meandering and braiding
in rivers. J. Fluid Mech. 76(3):457-480.
15. Parker G, Paolar C, Whipple K, Mohring D
(1998). Alluvial fan formed by channelized
fluvial and sheet flow. I: theory. J. Hydr. Engrg.
ASCE 124(10):985-995.
16. Parker G (2000). Progress in the modeling of
alluvial fans. J. Hydraul. Res. 37(6):805-825.
17. Parker G, Sawai K, Ikeda S (1982). Bend
theory of river meanders. Part 2. Nonlinear
deformation of finite-amplitude bends. J. Fluid
Mech. 115:303-314.
24
18. Pizzuto JE (1990). Numerical simulation of
gravel river widening. Water Resour. Res.
26:1971-1980.
19. Shimizu Y (2002). Numerical model of braided
stream with bank erosion. River flow 2002, Intl
Conf. on Fluvial Hydraulics. Louvain-la-Neuve,
Belgium: 793-801.
20. Shimizu Y, Hirano N, Watanabe Y (1996).
Numerical calculation of bank erosion and free
meandering. Annu. J. Hydr. Engrg. JSCE
40:921-926 (in Japanese).
21. Shimizu Y, Itakura T (1991). Calculation of
flow and bed deformation with a general
nonorthogonal coordinate system. Proc. XXIV
IAHR Congr. Madrid, Spain:C-241-C-248.
22. Sun T, Paola C, Parker G (2002). Fluvial fan
deltas: linking channel processes with large-
scale morphodynamics. Water Resour. Res.
38(8):26.
23. Yabe T, Aoki T (1991). A universal solver for
hyperbolic equations by cubic-polynomial
interpolation I. One-dimensional solver. Comp.
Phys. Comm. 66:219-232.
24. Yabe T, Ishikawa T, Kadota Y, Ikeda F (1990).
A numerical cubic-interpolated pseudoparticle
(CIP) method without time splitting technique
for hyperbolic equations. J. Phy. Soc. Japan
59(7):2301-2304.
25. Yabe T, Ishikawa T, Wang PY, Aoki T, Kadota
Y, Ikeda F (1991). A universal solver for
hyperbolic equations by cubic-polynomial
interpolation I. Two and three-dimensional
solver. Comp. Phys. Comm. 66:233-242.
26. Zhao DH, Shen HW, Tabios III GQ, Lai JS,
Tan WY (1994). Finite-volume two-
dimensional unsteady-flow model for river
basins. J. Hydraul. Engrg. ASCE 120(7):863-
883.