Proceedings of the 19th IAHR-APD Congress 2014, Hanoi, Vietnam

ISBN 978604821338-1

COMPUTATIONAL MODELLING OF BANK EROSION BY A 2-D DEPTH-AVERAGED MODEL

UNDER HOMOGENEOUS AND HETEROGENEOUS CONDITIONS

SUPAPAP PATSINGHASANEE(1), ICHIRO KIMURA(2), YASUYUKI SHIMIZU(3) & KAZUYOSHI HASEGAWA(4)

(1)
Doctoral Student, Graduate school of engineering, Hokkaido University, Sapporo, Japan, supapap@ec.hokudai.ac.jp
(2)
Associate Professor, Graduate school of engineering, Hokkaido University, Sapporo, Japan, i-kimu2@eng.hokudai.ac.jp
(3)
Professor, Graduate school of engineering, Hokkaido University, Sapporo, Japan, yasu@eng.hokudai.ac.jp
(4)
Professor, Hokkaisuiko Consultant Co., Ltd., Sapporo, Japan, k-hasegawa@suiko.jp

ABSTRACT
Many river engineering problems were caused by bed deformation and bank erosion in all alluvial channels. Therefore,
the effective prediction of bed deformation and bank erosion is an urgent issue to understand the complicated alluvial
channel mechanisms. This paper focuses on computational modelling of bed deformation and bank erosion under
unsteady flow by using a two-dimensional shallow-water model on a curvilinear boundary-fitted coordinate system, an
equilibrium sediment transport model, and a simplified bank failure model for homogeneous and heterogeneous
conditions. The computational conditions are similar with those used in the existing experimental flume conditions in
the both grain size conditions. The computational models under homogeneous and heterogeneous conditions can
reproduce the experimental results by using an appropriate angle of repose of 35 o and a suitable transversal grid size of
1 cm. The temporal changes in cross-sectional profile averaged over longitudinal direction, resulted from the
homogeneous condition, are in a relatively good agreement with the experimental data and can be seen that one point
divides the region into deposition and scour zone, which is relatively common behavior for channel in a homogeneous
condition. However, dunes observed in the exiting homogeneous experiment data are not reproduced and alternate bars
are not well generated in computational results, respectively. In heterogeneous condition, the comparison results
between the computational results and the experimental data are satisfactory. However, the agreement of the bed
deformation are not satisfactory. The difference in bed deformations caused an overprediction in the dimensionless
shear stress exceeding the critical values over the whole range of bed channel and the static armoring layer is fully
developed to protect bank erosion on the top of bank toe, with an average grain size of 3-5 mm.

Keywords: Bank erosion, Two-dimensional depth-averaged model, Homogeneous condition, Heterogeneous condition.

1. INTRODUCTION Moreover, many computational models have been

Many river engineering problems have been caused by
bed deformation and bank erosion in all alluvial
channels. It is an important mechanism by which a
channel width adjustment, bank geometry change, and
slope convey water and sediment supplied to it.
Therefore, an effective prediction of bed deformation and
bank erosion is an urgent issue to understand the
complicated alluvial channel mechanism.
In the process of channel development (bed deformation
and bank erosion), several researchers are concerned with
theoretical study such as analytical stability analysis of
steep banks, the concept of critical shear stress of bank
erosion, and mechanical bank retreat caused by flow
erosion and mass failure under gravity. For instance,
Osman and Colin (1988) analyzed slope stability analysis
in steep slopes for calculation of the lateral erosion
distance and predicted bank stability by lateral erosion or
bed degradation. ASCE Task Committee (1998)
considered the mechanics of bank retreat caused by flow
erosion and mass failure under gravity, and bank
development caused by sedimentation. Moreover, they
motioned that the temporal and spatial averaged
boundary shear stress is an important parameter for
prediction equilibrium in width adjustment.
employed to predict the bed deformation and bank
erosion by a simplified failure model under
homogeneous condition and they validated their results
with experiment of flume data in homogeneous
conditions. Nagata et al. (2000) and Onda et al. (2010)
showed bank erosion results in term of channel process
and longitudinal distribution of erosion volume by
governing equations of flow in generalized curvilinear
coordinate system, non-equilibrium sediment transport
model, and intermittent bank-collapse model. Arimitsu et
al. (2010) investigated the characteristics of sediment
transport and bank erosion process in steep slope curve
channel by a numerical modelling with governing
equations of flow in generalized curvilinear coordinate
system, an equilibrium sediment transport model, and a
simplified bank failure model. In addition, Jang and
Shimizu (2005) developed
a two-dimensional numerical model to simulate
relatively wide, shallow rivers with an erodible bed and
banks. They employed a moving boundary-fitted
coordinate system with cubic interpolated pseudoparticle
method (CIP) to calculate water flow. Moreover, bank
erosion was simulated when the gradient in the cross-
sectional direction of banks was steeper than the
submerged angle of repose.

1
Recently, several researchers simulated the bed u u u
deformation and bank erosion in real rivers. For instance, u u 1u u 2u u 3u u
Iwasaki et al. (2012) applied two-dimensional depth t
average model in a general curvilinear coordinate system
with simplified bank erosion model to simulate river


g x2 y2
H


x x y y
H



flow and bed deformation of the Otofuke river,
C f u
Hokkaido, Japan. Moreover, Li et al. (2013) introduced a
new norm related to stable slope to conduct alluvial
hJ
u y

y u u
2
x

xu 2
D
riverbank stability assessment. The new model was [2]
incorporated into one and two dimensional flow and
sediment equations in orthogonal boundary-fitted u u u
coordinates for simulation river evolution with bank u u 4 u u 5 u u 6 u u
t
erosion in Yangtze river, China.
In addition, the exiting flume experiments in straight


g x x y y
H


x2 y2
H


channels were conducted to reveal the effect of bank
erosion on the temporal change in cross-sectional profile C f u
averaged on longitudinal direction under homogeneous
hJ
u
y

y u u
2
x

x u 2
D
and heterogeneous conditions by Nakanishi et al. (2005)
[3]
and Abe et al. (2006), respectively.
where t is time, H is water surface elevation (=h+Zb), Zb is
In this regard, the main objectives of this paper are 1) to
bed elevation, g is gravitation acceleration, Cf is the
develop computational bank erosion model under
coefficient of riverbed shearing force (=gnm2/h1/3), nm is the
heterogeneous condition because a large number of bank
Manning's roughness coefficient by using the Manning-
erosion models are already available on homogeneous
Stricker equation (=ks1/6/7.66g1/2), ks is the relative
condition and 2) to compute channel processes with bed
roughness height which is defines as 1-3d, d is a sediment
deformation and bank erosion in a straight channel under
diameter, and D and D are the diffusion terms in and
homogeneous and heterogeneous conditions examined
directions.
by applying a two-dimensions depth averaged flow
model in a general curvilinear coordinate system, an The diffusion terms which express the turbulence
equilibrium sediment transport model, and a simplified diffusion were calculated as follows :
bank failure model. The simplified bank failure model is
assumed that bank erode when the gradient in the cross- u u
D t r2



t r2


sectional direction of the banks is steeper than the angle
of repose.
[4]

2. GOVERNING EQUATIONS u u
D t r2

t r2



This study applied a plane two-dimensional based
morphodynamic model to simulate the bed deformation [5]
and bank erosion. This section describes the main
concepts of numerical models and governing equations where r and r are the rations of local grid size in general
used into 3 parts. curvilinear coordinate system to the full-scale length of
the grid, t is the eddy viscosity coefficient which is
2.1 Flow equations estimated by using a zero equation turbulent model
(=u*h/6), is Von Karman constant (0.4), and u* is shear
A depth-averaged flow model in a general curvilinear velocity.
coordinate system was used to calculate the flow field.
The equations were described as follows (Jang and 2.2 Equilibrium sediment transport equation
Shimizu, 2005).
2.2.1 Homogeneous condition
2.1.1 Continuity equation
The two-dimension sediment continuity equation in a
h hu hu moving boundary-fitted coordinate system is
0
t J J J
zb 1 qb q b
[1] 0
t J 1 J J
where and are the spatial coordinate components in
general coordinate system, h is water depth, u and u are [6]
depth averaged flow velocity component in and where is porosity of bed material, qb and qb are the
directions, and J is the Jacobian of coordinate contravariant components of the bedload transport rate
transformation (=1/(x y.- x y)). per unit width in and directions, respectively.

2.1.2 Momentum equations The sediment-transport rate in streamwise direction was

calculated using Ashida and Michiue's formula (1972),
and it is given as :

2
 nk
qbs 17 *3 2 1 *c
*
1 *c
*
S g gd 3

qb P q k bk
k 1
[7] [12]

where qbs is the sediment-transport rate in streamwise
direction, * is the non-dimensional bed shear stress
(=nm2V2/Sgdh1/3), V is the composite velocity, Sg is the
specific weight of sediment in fluid equal to 2.65, and *c
is the non-dimensional critical bed shear stress which is
obtained from Iwakagi's formula (1956).
The sediment-transport rate in transversal direction was
calculated using Hasegawa (1984) formula, given as:
where qb is the bedload transport rate per unit width, qbk is
the potential transport rate for a given size friction k, Pk is
the concentration of sediment size fraction k, and the
subscripts k and nk are the number and the total number
of size friction, respectively.
The sediment-transport rate in streamwise direction
under heterogeneous condition can be express by Ashida
and Michiue (1972) formula as follows :

u *c z b
qbn qbs nb qbk 17 *3 2 1 *ck 1 *ck S g gd k3

u
sb s k * n *k *k
[8] [13]

where qbn is the sediment-transport rate in transversal where *k is the dimensionless shear stress acting on grain
direction, unb and usb are the velocity components in the of the size in layer k (=u*2/Sggdk), u* is the shear velocity,
streamwise and transversal directions near the bottom, *ck is the dimensionless critical shear stress for grain size
respectively, and s and k are the static and kinetic in layer k, and dk is the representative grain size of layer k.
friction coefficients of bed material, respectively, which The shielding effect has to be considered for calculating
are assume to be 0.7 for the both of them. the dimensionless critical shear stress. Therefore, modify
When the streamline is curved, secondary flow is version of Asada's formula (1972) was employed as
generated because of the variation in the centrifugal follow:
forces. The near-bed velocity perpendicular to 2

streamwise direction is written as :
*ck log 23
h 10
unb usb N* *cm dk
rs log10 21 2
dm
[9] [14]
where rs is the radius of curvature of the streamwise, and where *cm is the dimensionless critical shear stress of
N* is the coefficient of the strength of the secondary flow median diameter dm.
which is assumed to be 7.0, as propose by Engelund
(1974). Then, the bed deformation was computed by using the
sediment continuity equation which is
The sediment-transport rate in and directions, qb and
qb, are given as (Watanabe et al., 2001)
zb



1


q bk

q bk

0
u z z t J 1 J J

qb qbs b b cos b
Vb [15]

and
[10] where qbk qbk are the bedload transport in and
u directions of layer k. Equations [10] and [11] is applied
z z
qb qbs b b cos b to calculate q and q for each grain size layer.
Vb bk bk

[11]
2.3 Bank erosion model

where ub and ub are the flow velocity near the bed in A simplified bank failure model is used, consistent with
the and directions, respectively, Vb is the resultant several researchers e.g. Hasegawa (1984), Nagata et al.
velocity near bed, is the intersection angle between (2000), Jang and Shimizu (2005), Onda et al. (2010),
and axes, is a correction coefficient of sediment Arimitsu et al. (2010), Iwasaki et al. (2012).
transport for slope gradient, and =(*c/s k *)1/2 as
The bank erosion of this study occurs when bed slope ()
proposed by Hasegawa (1984).
among four neighbor cells becomes steeper than the
2.2.2 Heterogeneous condition angle of repose () of the bed material after bed scouring.
Figure 1, a simplified bank failure model in which bank
The concept of size faction transport divides the bed
slope adjustment was set to be milder than the angle of
material into size fractions and it considers each size
repose and the volume of bank erosion was set equal to
fractions as a uniform material. The bed material
the volume of deposition. A simplified bank failure
transport rate can be calculated by multiplying the
equations can be expressed as follows:
potential transport rate corresponding to the given size
fraction with the percentage of material which can be
expressed as follows :

3
 L(tan tan ) 3. RESULTS AND DISCUSSIONS
0
A [16] Here we separate the results and discussions into 4 parts,
1 0 namely: 1. the existing experiment flume data, 2. the
A1 optimization results for the angle of repose, 3. the results
of transversal grid sensitivity analysis and the simulation
L(tan tan )
1 time, and 4. the computational results for the bank
A [17] erosion.
1 1
A0
3.1 The existing experiment flume data
where L is gird cell width, 0 and 1 are bank and bed
This study used the existing experiment flume data
elevation, 0 and 1 are bank and bed elevation change,
which revealed the effects of bank erosion on the
and A0 and A1 are bank and bed grid surface area,
temporal change in cross-sectional profile, averaged in
respectively.
longitudinal direction, for homogeneous condition by
Nakanishi et al. (2005) and heterogeneous condition by
Abe et al. (2006) for numerical model setup and
calibration. The flume experiments were conducted in a
straight trapezoidal channel with 11 m. length and 0.45
m. width. The channel erosion occurred only on the left
bank, illustrated as experiment flume characteristics in
Figure 2, grain size distribution of heterogeneous
condition in Figure 3, and summarized experiment flume
conditions in Table 1.
(A) Top view of a simplified bank failure model
The gradation coefficient of heterogeneous condition
(g=[d84/ d16]1/2) with d16= 0.029 mm. and d84 = 3.627 mm.,
shows a high gradation coefficient, equal to 11.18, which
describes a well-graded mixture in the existing grain size
condition.

(B) Side view of a simplified bank failure model

Figure 1 A simplified bank failure model

(A.) The experiment flume in plain view

2.4 Computational procedures
Figure 2. Diagram of experiment flume characteristics
The computational model applies the following processes
to compute channel processes with bed deformation and
bank erosion in a straight channel under homogeneous
and heterogeneous conditions with time at infinitesimal
interval up to the designated time for the given initial
conditions :
Compute the 2D depth-average flow in a general
curvilinear system of water channel. The governing
equation, i.e., the continuity and momentum
equations for water flow are numerically calculated
using finite difference method with computational (B.) The initial cross-sectional profile
grid in general coordinate systems (,). In order to
solve the governing equations, the cubic interpolated Figure 2. Diagram of experiment flume characteristics (cont'd)
pseudoparticle method (CIP) is used;
Compute the secondary flow perpendicular to the
streamwise direction of the 2D depth-average flow;
Compute sediment transport rate and river bed
deformation;
Determine how bank erosion and sediment
deposition alter the shape of channel;
Set a coordinate system using the new boundary and
update the computational data set;
Update the computational time.

4
 3.3 The results of transversal grid sensitivity analysis
and simulation time
The transversal grid sensitivity analysis was used to find
the appropriate size of transversal grids by various
transversal grid size between 5 cm., 2.5 cm., 1.25 cm., 1.0
cm., 0.5 cm., 0.25 cm., 0.125 cm., and 0.1 cm. with a fixed
grid size in the streamwise direction (x=1 m.), and a
simulation time step (t=0.0005 s.).
The transversal grid sensitivity analysis indicated that
when the transversal grid size was finer than 1 cm., the
accuracy of the numerical model was not dramatically
different than that with a transversal grid size of 1 cm..
Hence, the suitable transversal grid size for this case was
Figure 3. The grain size distribution of heterogeneous condition
set to 1 cm.. The results are shown in Figure 5.
In addition, the CourantFriedrichsLewy condition
Table 1. The existing experiment flume conditions (CFL) is a function of grid spacing, the cases with finer
case discharge bank bed mean simulation grid size (finer than 1 cm.) yield a relatively small time
(l/s) height slope diameter time step, which in its turn it increases the simulation time.
(cm.) (mm.) (hrs.) The required simulation time for the cases are shown in
Homogeneous Condition Figure 6.
1. 6 10.5 1/200 0.5 1
2. 6 10.5 1/500 0.5 1
3. 6 12.5 1/200 0.5 1
Heterogeneous Condition
4. 6.7 8.5 1/100 1.56 2
5. 2.6 7.5 1/100 1.56 4
6. 2.6 8.5 1/100 1.56 2

3.2 The optimization results of angle of repose

This process is very important to adjust the angle of
repose of channel material which is generally reported to
be 30o to 32o for sand (0.25 mm. d < 1 mm.) (Julien, 2010)
because the angle of repose with bank erosion mainly
depends on the hydraulic conditions and the
characteristics of channel materials. Therefore, the
(A.) The comparison results of Case 2 (homogeneous condition) at
appropriate angle of repose can be found from the 15 minutes in each angle of repose
computational model by varying the angle of repose
form 20o to 45o in the homogeneous condition and from
30o to 50o in the heterogeneous condition. As a result, the
optimized angles of repose, from the numerical model,
were found to be 35 o in both grain size conditions,
providing a suitable computational results compared
with the existing experiment flume data. In addition, the
result for the angle of repose by trial and error is in good
agreement with the empirical formula by Zhang et al.
(1989).

32.5 1.27d m
[18]

where is the angle of repose. This empirical formula

(B.) The comparison results of Case 2 (homogeneous condition) at
yields angles of repose equal to 33.1 o and 34.5 o for the 30 minutes in each angle of repose
homogeneous and heterogeneous conditions,
respectively. Figure 4. The optimization results of angle of repose in
homogeneous (Case 2) and heterogeneous (Case 5) conditions
Moreover, it is found that the banks with a larger angle of
repose show less erosion rate than those with smaller
angle of repose for the both grain size conditions. In
addition, The sediment caused by the bank erosion is not
supplied from the river banks with a large angle of
repose, but it is supplied from the river banks with a
small angle of repose. The computational results for the
optimization of angle of repose in homogeneous
condition (Case 2) and heterogeneous condition (Case 5)
at 15 and 30 minutes of simulation time are shown in
Figure 4.

5

(C.) The comparison results of Case 5 (heterogeneous condition) at
15 minutes in each angle of repose
(D.) The comparison results of Case 5 (heterogeneous condition) at
30 minutes in each angle of repose
Figure 4. The optimization results of angle of repose in
homogeneous (Case 2) and heterogeneous (Case 5) conditions
(cont'd)
Figure 5. The results of transversal grid sensitivity analysis
Figure 6. The results of simulation time in difference transversal
grid size
3.4 The computational results of bank erosion
Numerical computations were setup and compute under
the existing experiment flume conditions. The
computational angle of repose () and computation time
step (t) were 35o and 0.0005 s., respectively.
We set 12 grid points in the streamwise direction and 47
grid points in transversal direction which led to grid sizes
of 1 m. and 1 cm., respectively. A periodic boundary
condition was set in streamwise direction for the bed
deformation, channel adjustment and bank erosion.
Moreover, the bottom friction parameter was estimated
by the Manning-Strickler equation which is 0.014 in the
homogeneous condition and 0.016 in the heterogeneous
condition. In addition, the void rations (porosity) of
channel material were 0.4 and 0.6 for the homogeneous
and the heterogeneous condition, respectively.
3.4.1 The results of homogeneous condition
Using the computational model, we replicated the
existing experiment conducted by Nakanishi et al. (2005),
as shown in Table 1 for the homogeneous condition for
Case 1 to Case 3. The difference between Case 1 and Case
2 is the initial bed slope, and Case 2 and Case 3 include
different bank heights. These cases are conducted to
investigate the influence of shear stress and bank height
on the bank erosion rate, respectively.
The results of homogeneous condition show the temporal
change in cross-sectional profile averaged over
longitudinal direction as shown in Figure 7. The
computational results are in a relatively good agreement
with the experimental data. However, Figure 7 shows
that the computational results exceed the experimental
data and this error tends to increase with time. In
addition, it is found that the results of Case 1 and Case 3
(steep slope condition, 1/200) after 60 minutes of
computational time show an overprediction in the bank
erosion when the bank erosion reaches to channel
boundary condition.
The compared computational results of Case 1 and Case
2 with difference initial bed slope show that the steeper
initial slope (Case 1) performs larger shear stress than
initial mild bed slope (Case 2) (Figure are not shown.). In
addition, comparison results between Case 1 and Case 3
with different bank height were shown in Figure 7. It can
be seen that the erosion rate in Case 3 has greater rate
than that other homogeneous cases.
Moreover, it can be seen that there is a point dividing the
region into deposition and scour zone, which is relatively
common behavior for channels with homogeneous
condition. On the other hand, dunes in Case 1, observed
in the exiting homogenous experiment data, were not
reproduced in computational results. It is due to the fact
that the limitation of two-dimensional depth averaged
flow model is applied to calculate the flow fields.
However, alternate bars in Case 2 were not well
simulated in computational results.
3.4.2 The results of heterogeneous condition
Using the computational model, we replicated the
existing experiments conducted by Abe et al. (2006) for
heterogeneous condition as described in Table 1 for Case
4 to Case 6.
6
Figure 8 show the temporal change in cross-sectional (A.) The comparison results in Case 4 (heterogeneous Condition)
profile averaged over longitudinal direction for the case
of heterogeneous condition. The comparison results Figure 8. The comparison results between computational results
and experimental data in heterogeneous condition
between the computational results and experimental data
show a good agreement for the bank erosion rate but
unsatisfactory agreement for the bed deformation.

(B.) The comparison results in Case 5 (heterogeneous Condition)

(A.) The comparison results in Case 1 (homogeneous condition)

(C.) The comparison results in Case 6 (heterogeneous Condition)

(B.) The comparison results in Case 2 (homogeneous condition) Figure 8. The comparison results between computational results
and experimental data in heterogeneous condition (cont'd)

The difference in the bed deformation between

computational results and experimental data causes an
exceeding in the dimensionless shear stress over the
critical shear stress (based on Shields diagram as it is
mentioned by Iwakagi's formula.) over the whole range
of bed channel and the distribution of the dimensionless
shear stress around riverbank is complicated for
computation in case of bank erosion. Figure 9 shows the
dimensionless shear stress profile for heterogeneous
condition as a representative result at 60 minutes of the
physical time.

(C.) The comparison results in Case 3 (homogeneous condition)

Figure 7. The comparison results between computational results

and experimental data in homogeneous condition

(A.) The dimensionless shear stress in Case 4

Figure 9. The dimensionless shear stress profile of heterogeneous

condition

7
 (B) The dimensionless shear stress in Case 5
(c) The dimensionless shear stress in Case 6
Figure 9. The dimensionless shear stress profile of heterogeneous
condition (cont'd)
In addition, the influence of the amount of sediment
deposited on a static armor layer on the sediment
transport is presented in our computational results.
Hence, armoring effect of bed layer which is the result of
erosion of fine particles on the bed (leaves the coarse
fractions of the mixture on the bed) induced coarsening
of the bed material and it is fully developed to protect
bank erosion after 60 minutes of physical time. Therefore
the development of static armor layer, as a well mixed
sediment materials (0.3 mm. < d < 13 mm., dm=1.54 mm.),
was placed on the top of bank channel toe with an
average grain size of 3-5 mm. as shown in Figure 10.
(A.) The armoring effect of bed layer in Case 4
Figure 10. The mean diameter profile of armoring effect in
heterogeneous condition
(B.) The armoring effect of bed layer in Case 5
(C.) The armoring effect of bed layer in Case 6
Figure 10. The mean diameter profile of armoring effect in
heterogeneous condition (cont'd)
4. CONCLUSIONS
A computational model for bed deformation and bed
erosion under homogenous and heterogeneous
conditions were employed to predict channel
morphology processes by using a two-dimension
boundary-fitted curvilinear coordinate system, an
equilibrium sediment transport, and a simplified bank
failure model. It is found that both grain size conditions
can reproduce the experimental results by using an
appropriate values of the angle of repose. It should be
noted that a good agreement can be observed in
homogeneous condition in the case of mild slope channel
(Case 2, 1/500). On the other hand, dunes in Case 1,
observed in the exiting homogenous experiment data, are
not reproduced in computational results. It is due to the
fact that the limitation of 2D depth averaged flow
model is applied to calculate the flow fields. However,
alternate bars in Case 2 are not well simulated in
computational results. In the heterogeneous condition,
the results of bank erosion rate showed a satisfactory
agreement with the experimental data but unsatisfactory
agreement for the bed deformation. The difference in bed
deformation caused the dimensionless shear stress
exceeded the critical values over the whole range of bed
channel and the distribution of the dimensionless shear
stress around riverbank is complicated for computation
in case of bank erosion. Finally, the static armor layer is
fully developed to protect bank erosion on the top of
bank toe with an average grain size of 3-5 mm after 60
minutes of simulation time.
8
ACKNOWLEDGMENTS Osman A.M. and Colin R.T. (1988). Riverbank Stability
Analysis. I: Theory. J. Hydraul. Eng., 114, 134-
We gratefully acknowledge Dr.Mohamed Nabi and
150.
Dr.Toshiki Iwasaki for many valuable suggestions and
Watanabe A., Fukuoka S., Yasutake Y., and Kawaguchi
recommendations.
H. (2001). Groin arrangements made of natural
willows for reducing bed deformation in curved
REFERENCES
channel. Advances in River Engineering, 7, 285-290. (in
Abe Y., Takahashi K., Hasegawa K., and Nakanishi S. Japanese)
(2007). Experiment on bed evolution with bank Zhang R.J., Xie J.H., Wang M.F., and Huang J.T. (1989).
erosion of straight channel in the case of Dynamic of fluvial sediment transport, China Water
sediment diameter of mixture. Proceeding of the Power Press, China. (in Chinese)
Hokkaido Branch, Japan Society of Civil
Engineering, Vol.63, B-29 (CD-ROM). (in Japanese)
Arimitsu T., Ooe K., Onda S., Hosoda T., and Shirai H.
(2010). Experimental and numerical
investigation of river channel processes with
bank erosion in steep curved channel. River, Coastal
and Estuarine Morphodynamic: RCEM 2009-
Vionnet et al. (eds.), 983- 988.
Asada H., and Ishikawa H. (1972). Study on selective
transportation of river material. Central Research
Institute of Electric Power Industry, No.71015.
ASCE Task Committee on Hydraulic (1998). River Width
Adjustment. I: Process and Mechanisms. J.
Hydraul. Eng., 124(9), 881-902.
Ashida K. and Michiue M. (1972). Hydraulic resistance
and bed transport rate in alluvial stream. Proceeding
of JSCE, 201, 59-69. (in Japanese)
Engelund F. (1974). Flow and bed topography in channel
beds. J.Hydraul.Div., Am. Soc. CivEng, 100 (11), 1631-
1648
Hasegawa K. (1984). Hydraulic research on planimetric
forms, bed to pographies and flow in alluvial river.
PhD Dissertation, Hokkaido University, Sapporo, Japan.
1-184. (in Japanese)
Iwagaki Y. (1956). Hydrodynamical study on critical
tractive force. Proceeding of JSCE, 44, 1-21. (in
Japanese)
Iwasaki T., Shimizu Y., and Kimura I. (2012). Numerical
simulation on bed evolution and channel
migration in rivers. River Flow 2012-Murillo (Ed.),
673-679.
Jang C. and ShimizuY. (2005). Numerical Simulation of
Relative Wide, Shallow Channels with Erodible
Banks. J. Hydraul. Eng, 131, 565-575.
Julien P.Y. (2002). River Mechanics. Cambridge University
Press, 9-30.
Li Y.T., Wang F., Tang J.W., and Zhu L.L. (2013).
Mechanism and prediction of bank failure.
Advances in River Sediment Research - Fukuoka et al.,
21-34.
Nagata N., Hosoda T., and Muramoto Y. (2000).
Numerical Analysis of River Channel Processes
with Bank Erosion. J. Hydraul. Eng., 126(4),
243-252.
Nakanishi S., Hasegawa K., and Takahashi K. (2006).
One-dimensional analysis of bed evolution
accompanying bank erosion. River, Coastal and
Estuarine Morphodynamic: RCEM2005-
Parker&Garcia (eds), 937-942.
Onda S., Shirai H., Hosoda T., Arimitsu T., and Ooe K.
(2010). Numerical simulation of river channel
processes with bank erosion in steep curved
channel. River Flow 2010-Dittrich, Koll,
Aberle&Geisenhainer (eds.), 993-1,000.