Professional Documents
Culture Documents
by
May 2013
1 ACKNOWLEDGEMENTS
I would like to thank all the individuals and organizations who have helped or
provided the guidance during the study. Among the organisations, firstly I would like
to thanks the Universiti Sains Malaysia who provided the opportunity to study in this
university under fellowship scheme. I am also thankful to USM for all efforts from
granting the study leave to providing all reports and drawings required for this thesis
work.
supervisor, Prof. Dr. Hj Aminuddin Ab. Ghani for his supervision, advice and
guidance. I really was honoured to have the opportunity to work under his
I would also like to thank River Engineering and Urban Drainage Research
Centre (REDAC) and its staffs especially Mrs. Nor Mawati Mohamad, Mr. Mohd
Sufian Osman, Mr. Rahim Ghazali and Mr. Khairul Nizam Abu for helping me in
I don’t have words to express my thanks to Dr. Farshid Bateni, in fact no words
can express his generosity. All through this work he has provided his guidance and
seniors, colleges and friends who helped one or another way to make possible this
study.
Last but not least I would extend my word of thanks to my family specially my
lovely sister Nasim who helped me during all this duration of my study.
ii
2 TABLE OF CONTENTS
1 ACKNOWLEDGEMENTS ................................................................................................ ii
iii
2.6.1 Genetic Programming (GP)..................................................................... 26
2.6.2 Artificial Neural Network (ANN) ........................................................... 28
2.7 Application of Soft Computing Modelling in Prediction of Bedload Transport .... 30
2.8 River Channel Confluence...................................................................................... 36
2.9 Sediment Transport Modelling ............................................................................... 41
2.9.1 SSIIM ...................................................................................................... 48
2.9.1.1 SIMPLE Algorithm ................................................................. 49
2.9.1.2 Control Volume Scheme.......................................................... 49
2.9.1.3 SSIIM Application ................................................................... 50
2.10 Summary................................................................................................................. 55
iv
4.2.3 Parameter Affecting Bedload Transport ................................................. 90
4.3 Particle Size Distribution ........................................................................................ 93
4.4 Evaluation of Bedload Size Distribution with Increasing Shear Stress .................. 97
4.5 Fractional Transport Rate ..................................................................................... 102
4.6 Performance of Bedload Transport Equation ....................................................... 107
4.6.1 Assessment of Existing Equation for Kurau River ............................... 107
4.6.2 Prediction of Bedload Transport in Kurau River with Nonlinear
Regression Method ............................................................................... 109
4.6.3 Prediction of Bedload Transport in Kurau River by Genetic Programming
............................................................................................................... 112
4.6.4 Combination of ANN and GP ............................................................... 117
4.6.5 Comparison of Bedload Equations for Kurau River ............................. 122
4.7 Development of Bedload Equation for Small Rivers (Kurau, Lui, Semenyih) .... 126
4.7.1 Assessment of Existing Equations for Small Rivers (Kurau, Luie and
Semenyih) ............................................................................................. 127
4.7.2 Nonlinear Regression Result for Small Rivers (Kurau, Lui and Semenyih)
............................................................................................................... 129
4.7.3 Artificial Neural Network Results ........................................................ 131
4.8 Sensitivity Analysis .............................................................................................. 134
4.9 Genetic Programming Result................................................................................ 136
4.9.1 Comparison of Bedload Equations for Small Streams .......................... 139
v
5.9 Sediment Flow Simulation in Confluence of Kurau and Ara River ..................... 160
5.9.1 Characteristics of Kurau -Ara Confluence ............................................ 161
5.9.2 Input Data .............................................................................................. 163
5.9.3 Input Files ............................................................................................. 165
5.9.3.1 Control File ............................................................................ 165
5.9.3.2 Timei File .............................................................................. 166
5.9.4 Numerical Algorithms ........................................................................... 169
5.9.5 Sensitivity Analysis............................................................................... 170
5.9.6 Calibration and Validation .................................................................... 171
5.9.6.1 Model Calibration .................................................................. 171
5.9.6.2 Model Validation ................................................................... 183
5.9.7 Short Term Changes in Bedload Transport, Bed Morphology and Bed
Material Characteristics ........................................................................ 186
5.9.7.1 Morphological Changes ......................................................... 188
5.9.7.2 Lateral bar .............................................................................. 209
5.9.7.3 Bedload Transport Rates........................................................ 211
5.9.7.4 Sediment Pattern .................................................................... 220
5.9.8 High Flow Modelling ............................................................................ 229
8 APPENDIX A
9 APPENDIX B
10 APPENDIX C
vi
3 LIST OF TABLES
Table 2.8 Comparison of bedload equations and the ANN model (Sasal et 32
al., 2009)
Table 3.2 Summary of flood frequency analysis for Kurau River at Pondok 65
Tanjung
Table 3.4 Typical cross sections along Kurau River (19 June 2010) 70
vii
Table 4.1 Range of field data 85
Table 4.2 Summary of large and medium rivers (Monalis and Wu, 2001) 86
Table 4.5 Parameter estimates of experimental data based on equation (3-14) 110
Table 4.6 Statistical analysis of experimental data based on equation (3-14 110
Table 4.7 Parameter estimates of experimental data base on equation (4-1) 111
Table 4.8 Statistical analysis of experimental data base on equation (4-1) 111
Table 4.13 Parameter estimates of experimental data based on equation (4-7) 129
Table 4.14 Parameter estimates of experimental data based on equation (4-8) 130
Table 4.15 Statistical analysis of experimental data base on equation (4-7) 130
Table 5.1 Comparison of Cont value for one and two block grid 159
viii
Table 5.4 Parameter calibrated in SSIIM 174
Table 5.5 Comparisons of water and bed level for Q=15 m3/s (19 July 2012) 183
Table 5.6 Comparisons of water and bed level for Q=43 m3/s (27 Sept 2012) 183
Table 5.7 Comparisons of water and bed level for Q=11 m3/s (8 Oct 2012) 185
Table 5.8 Hydraulic condition during an event at Kurau _Ara confluence 187
ix
4 LIST OF FIGURES
Figure 2.3 The ANFIS model for bed load sediment (Azamathulla et al., 33
2009)
Figure 2.4 Predicted bed load against measured bed load using ANFIS 33
(Azamathulla et al., 2009)
Figure 2.7 Observed versus predicted sediment load by GEP for Langat, 35
Kurau and Muda rivers (Ab. Ghani and Azamathulla, 2012;
Azamathulla et al., 2010a; Chang et al., 2012; Zakaria et al.,
2010)
Figure 2.8 (a) Measured bed levels after the flushing (b) Simulated bed 52
levels after the flushing (Haun and Olsen, 2012)
Figure 2.10 Comparison between measured values and simulation results at: 53
(a) cross section 80; (b) cross section 60; and (c) cross section 20
Figure 2.11 Measured water depths before (a) and after (b) the flood, 54
together with measured (c) and computed (d) bed elevation
changes (Fischer-Antze et al., 2008).
x
Figure 3.3 Ara -Kurau river 60
Figure 3.6 Water level chart for Kurau River at Pondok Tanjung 62
Figure 3.9 Langat River basin and data collection sites by Ariffin (2004) 67
Figure 3.12 River surveying at Ara River with river surveyor (ADP) 72
Figure 3.14 Hand held Helley-Smith sampler for bed load sampling 75
xi
Figure 4.10 Scatter plot of bedload transport rate against water depth 91
Figure 4.11 Scatter plot of bedload transport rate against B/Y ratio 91
Figure 4.12 Scatter plot of bedload transport rate against hydraulic radius 92
Figure 4.17 Particle size distributions of bedload and bed material samples 96
for Kurau River.
Figure 4.19 Mean bed load grain size distributions for shear stress bands 98
arranged in order of increasing shear stress (upstream of Kurau
River KRU5).
Figure 4.20 Mean bed load grain size distributions for shear stress bands 100
arranged in order of increasing shear stress (downstream of
Kurau River KRU1).
Figure 4.21 Variation in grain size at the10th, 16th, 30th, 50th, 84th and 90th 101
percentiles of the bedload size distribution with increasing shear
stress.
Figure 4.22 Transport ratio as a function of grain size at upstream (a) the 103
transport ratio Pi/fi where pi is the proportion of each size
fraction i present in transported material and fi is the proportion
of each size fraction in the bed material (b) the scaled fractional
transport rate computed as qbpi/fi, where qb is the sediment
transport rate.
Figure 4.23 Transport ratio as a function of grain size at downstream (a) the 104
transport ratio Pi/fi (b) the scaled fractional transport rate qbpi/fi.
Figure 4.24 Comparison of predicted and measured bedload rates for Kurau 108
River
xii
Figure 4.25 Bedload rating curve along Kurau River 109
Figure 4.28 Measured versus predicted values of Tb for the training data set. 115
Figure 4.29 Measured versus predicted values of Tbfor testing data set. 116
Figure 4.30 Measured versus predicted values of Tb for validation data set. 116
Figure 4.31 Measured versus predicted values of Tb for all data set. 117
Figure 4.33 Measured versus predicted values of Tb by ANN for training data 119
set
Figure 4.34 Measured versus predicted values of Tb by ANN for testing data 120
set
Figure 4.35 Measured versus predicted values of Tb by ANN for validation 121
data set
Figure 4.36 Measured versus predicted values of Tb by ANN for total data set 121
Figure 4.38 Comparison of bedload rating curve for Kurau River 125
Figure 4.39 Comparisons of predicted and measured bedload rates for Kurau 125
River
Figure 4.41 Performance of existing bedload transport formula in Kurau, Lui 128
and Semenyih rivers.
Figure 4.42 Measured versus predicted values of Tb for total data set 131
modelled by NLR
Figure 4.43 Measured versus predicted values of Tb by ANN for the training 133
data set
Figure 4.44 Measured versus predicted values of Tb by ANN for testing data 133
xiii
set
Figure 4.45 Measured versus predicted values of Tb by ANN for validation 134
data set
Figure 4.46 Measured versus predicted values of Tb by ANN with for total 134
data set.
Figure 4.47 Measured versus predicted values of Tb for the training data set. 137
Figure 4.48 Measured versus predicted values of Tb for testing data set 138
Figure 4.49 Measured versus predicted values of Tb for total dataset 138
Figure 4.50 Measured versus predicted values of Tb for validation dataset 139
Figure 4.51 Comparison of bedload rating curve for small streams 141
Figure 4.52 Comparisons of predicted and measured bedload rates for small 141
streams by different models
Figure 5.10 View of the confluence of the Kurau and Ara rivers 162
Figure 5.12 Sediment distribution size of bedload in Kurau River branch 163
xiv
Figure 5.13 Sediment distribution size of bedload in Ara River 164
Figure 5.14 Sediment distribution size of bedload in main Kurau River 164
Figure 5.18 Measured and simulated average velocity in Ara mouth 175
Figure 5.19 Measured and simulated average velocity in Kurau mouth 175
Figure 5.20 Comparison cross-sectional bed level and average velocity a) 176
simulated b) Measured, April 2012 at Ara River
Figure 5.21 Comparison cross-sectional bed level and average velocity a) 177
simulated b) Measured, April 2012 at Kurau River
Figure 5.24 Comparison cross sectional bed level in different condition of 179
Ara and Kurau confluence (Measured BL, April 2012)
Figure 5.25 Comparison of measured and simulated Longitudinal bed level 180
at downstream of confluence (AA') (Measured BL, April 2012)
Figure 5.26 Scatter plot of measured bed level against simulated bed level 180
(April 2012)
Figure 5.28 Scatter plot of measured water level against simulated water 181
level (April 2012)
Figure 5.31 Comparisons of water and bed level (AA') for Q=15 m3/s (19 184
July 2012)
xv
Figure 5.32 Comparisons of water and bed level (AA') for Q=15 m3/s (20 184
July 2012)
Figure 5.33 Comparisons of water and bed level (AA') for Q=11 m3/s (8 Oct 185
2012)
Figure 5.35 Longitudinal bed change profile of Ara and downstream of 189
confluence after Q=15m3/s
Figure 5.36 Longitudinal bed change profile of Kurau and downstream of 189
confluence after Q=15m3/s
Figure 5.38 Change in bed morphology after Q=15m3/s. Zone of erosion and 190
deposition during each period are illustrated with colour change
from white as deposition to black as erosion.
Figure 5.39 Channel cross section profiles, Q=15m3/s (Measured bed level 191
April 2012)
Figure 5.42 Longitudinal bed change profile of Ara and downstream of 194
confluence between Q=15m3/s and Q=31m3/s (Measured bed
level April 2012)
Figure 5.43 Longitudinal bed change profile of Kurau and downstream of 194
confluence between Q=15m3/s and Q=31m3/s (Measured bed
level April 2012)
Figure 5.44 Change in bed morphology between Q=15m3/s and Q=31m3/s. 195
Zone of erosion and deposition during each period are illustrated
with colour change from white as deposition to black as erosion.
xvi
Figure 5.49 Longitudinal bed change profile of Ara and downstream of 200
confluence between Q=31m3/s and Q=43m3/s (Measured bed
level April 2012)
Figure 5.50 Longitudinal bed change profile of Kurau and downstream of 200
confluence between Q=31m3/s and Q=43m3/s (Measured bed
level April 2012)
Figure 5.51 Change in bed morphology between Q=31m3/s and Q=43m3/s. 201
Zone of erosion and deposition during each period are illustrated
with colour change from white as deposition to black as erosion.
Figure 5.52 Longitudinal bed change profile of Ara and downstream of 202
confluence between Q=43m3/s and Q=35m3/s (Measured bed
level April2012)
Figure 5.53 Longitudinal bed change profile of Kurau and downstream of 202
confluence between Q=43m3/s and Q=35m3/s (Measured bed
level April2012)
Figure 5.55 Change in bed morphology between Q=43m3/s and Q=35m3/s. 203
Zone of erosion and deposition during each period are illustrated
with colour change from white as deposition to black as erosion.
Figure 5.59 Longitudinal bed change profile of Ara and downstream of 206
confluence between Q=35m3/s and Q=13m3/s (Measured bed
level April 2012)
Figure 5.60 Longitudinal bed change profile of Kurau and downstream of 207
confluence between Q=35m3/s and Q=13m3/s (Measured bed
level April 2012)
Figure 5.61 Change in bed morphology between Q=35m3/s and Q=13m3/s. 207
Zone of erosion and deposition during each period are illustrated
with colour change from white as deposition to black as erosion.
Figure 5.62 Channel cross section profiles, Q=13 m3/s 208
xvii
Figure 5.64 Cross sectional lateral change in different flow momentum 210
Figure 5.65 Bed load transport rating curve in Ara and Kurau River branch 211
Figure 5.66 Bed load transport rate value by SSIIM against the calculated 212
bedload transport rate with Eq. 4.11
Figure 5.67 Bed morphology and spatial distribution of bedload transport 213
rate Mr=0.9.
Figure 5.68 Bed morphology and spatial distribution of bedload transport 214
rate Mr=1.3.
Figure 5.69 : Bed morphology and spatial distribution of bedload transport 215
rate Mr=2.6.
Figure 5.70 Bed morphology and spatial distribution of bedload transport 216
rate Mr=0.7.
Figure 5.71 Shear layer and distinct vortices about vertical axes at RSK1 218
Figure 5.72 Shear layer in the confluence of Ara and Kurau 218
Figure 5.74 Distribution of bed median size, D50 Q=15 m3/s, Mr<1 222
Figure 5.76 Distribution of bed median size at high flow, D50 Q=43 m3/s, 224
Mr>1
Figure 5.78 Distribution of bed median size at low flow, D50 Q=13 m3/s, 226
Mr<1
xviii
Figure 5.83 Change in bed morphology after Q=191.32m3/s. Zone of erosion 232
and deposition during each period are illustrated with colour
change from white as deposition to black as erosion.
Figure 5.85 Modelled cross section changes before and after flood 2007 234
Figure 5.86 Bed morphology and spatial distribution of bedload transport 235
rate (Q=191.32m3/s)
xix
5 LIST OF ABREVIATIONS
Abbreviation Description
BL Bed Level
DR Discrepancy Ratio
GA Genetic Algorithm
GP Genetic Programming
Mr Momentum ratio
WL Water Level
WS Water Surface
xx
6 LIST OF SYMBOLS
Symbol Description
Flow area ( )
E East
f Friction factor
Fr Froude number
Gradation coefficient
xxi
Width of Helley-Smith sampler nozzle
N North
R2 Coefficient of determination
Re Reynolds number
Sf Channel slope
U Inequality coefficient
xxii
Mean weighted bed load sample of vertical for n section
, Flow depth
Standardized coefficient
Γ Diffusion coefficient
v kinematic viscosity
xxiii
PENGANGKUTAN BEBAN ENDAPAN DASAR UNTUK SUNGAI KECIL
DI MALAYSIA
7 ABSTRAK
dinamik sungai dan pengganggaran kadar pengangkutan beban endapan dasar adalah
penting untuk pengiraan variasi morfologi sungai untuk tujuan keselamatan umum,
pengurusan sumber air dan alam sekitar yang mampan. Pelbagai persamaan beban
endapan yang terkenal adalah terhad kepada kajian eksperimen saluran dalam
makmal atau kajian tapak. Persamaan ini yang dipengaruhi oleh kebolehpercayaan
dan perwakilan data yang digunakan dalam menentukan pembolehubah dan pemalar
endapan. Oleh itu, satu persamaan baru yang mudah dan tepat adalah perlu untuk
kegunaan di sungai-sungai kecil. Dalam kajian ini, data yang mudah diperolehi
seperti kadar alir, kedalaman sungai, kecerunan sungai dan saiz diameter zarah
endapan permukaan d50 daripada tiga sungai kecil di Malaysia digunakan untuk
artificial neural network (ANN) adalah berguna dalam menafsir data tanpa sebarang
had untuk pangkalan data yang luas digunakan sebagai alat untuk pemodelan
untuk meramal data hujan adalah memuaskan. Model yang diperolehi menunjukkan
kejituan yang tinggi dengan ketepatan keseluruhan sebanyak 97% untuk ANN dan
Satu model numerikal tiga dimensi telah digunakan untuk mengkaji morfologi
dasar dan pengangkutan beban endapan dasar sungai di pertemuan Sungai Ara dan
xxiv
Kurau untuk jangka masa pendek dengan kadar alir tinggi pada 100 ARI. Model tiga
dimensi SSIIM2 dengan k-epsilon aliran gelora yang merupakan model pengiraan
bendalir dinamik dengan grid adaptif, bukan ortogon dan tidak berstruktur telah
telah diuji dengan data dari kajian tapak di pertemuan Ara-Kurau. Ketepatan yang
memuaskan telah didapati di antara data endapan dasar dan aras dasar yang dianggar
merupakan alat yang berguna dalam meramal kadar pengangkutan beban dasar di
berhampiran pinggir lapisan ricih dan juga lapisan ricih yang menyebabkan aliran
menunjukkan taburan saiz zarah beting pasir di tepi hilir pertemuan sungai adalah
tidak berubah dimana saiz median tidak berubah sepanjang tempoh kajian manakala
saiz zarah di hulu beting pasir adalah lebih dipengaruhi oleh keadaan aliran.
xxv
BEDLOAD TRANSPORT OF SMALL RIVERS IN MALAYSIA
8 ABSTRACT
has major effects on public safety, water resources management and environmental
the relationship between the reliability and representativeness of the data utilized in
transport. Thus, a new simple equation based on a balance between simplicity and
accuracy is necessary for using in small rivers. In this study the easily accessible data
including flow discharge, water depth, slope, and surface grain diameter d50 from the
programming (GP) and artificial neural network (ANN) models that are particularly
presented as complementary tools for modelling bed load transport in small streams.
97% for ANN and 93% for GP compared with other traditional methods and
empirical equations.
morphology and bedload transport of the junction of Ara and Kurau rivers for short
term event and for high flow with 100 ARI. SSIIM2 a 3D, k-epsilon turbulence
xxvi
computational fluid dynamics model with an adaptive, non-orthogonal and
unstructured grid has been used for modelling the hydrodynamic of confluence. The
numerical model was tested against field data from Ara-Kurau confluence.
Satisfactory agreement was found between computed and measured bedload and bed
elevation in the field. The study indicates that numerical models became a useful
tool for predicting the bedload transport rate in such complex dynamic environment.
The results have demonstrated that the short term hydrologic variability can
for the different flow conditions the bedload transported near to edge of shear layer.
The coincidence of the shear layer that was generated the considerable turbulence
indicated that the increasing turbulence levels contribute substantially to the required
increase in bedload transport capacity. The simulation results showed the grain size
distribution on the bar at the downstream junction corner is remarkably constant and
the particle size in the upstream part of the bar is more affected by the changes in
flow conditions than the downstream end where the median diameters not varied
xxvii
1 CHAPTER 1-
INTRODUCTION
1.1 Background
water flow, river morphology and response of sediment particles to applied stress and
sediment through river channels has major effects on public safety, water resources
sediment transport, the interaction between sediment supply and bed surface
equations for the prediction of bedload transport. Numerous well-known bed load
equations were derived from limited flume experiments or field conditions (Bagnold,
1980; Camenen and Larson, 2005; Yang, 1996). Although morphologist and
engineers have gained profound insight into the mechanics of bedload transport ever
since the development of the duBoys equation (du Boys, 1879) (the first physically
based bedload transport equation) a simple question still cannot be answered: for
given sedimentary and hydraulic characteristics, what is the rate of bedload transport
in an alluvial channel? In other words, there is no single bedload equation that can be
1
applicable guidelines exist to facilitate the selection of an appropriate formula as the
bedload transport function (Almedeij and Diplas, 2003; Gomez and Church, 1989;
Simons and Şentürk, 1992; Yang and Huang, 2001). To overcome the difficulties of
developing the equations based on a balance between simplicity and accuracy, new
River flow, sediment transport and morphological processes are among the
confluence has always been a challenging subject for river hydrodynamics and
occurring in both the confluence and the downstream of confluence channel. The
complexity of the phenomena and processes arises from the strong three dimensional
flow effects resulting from several principal factors, including a) the discharge or
momentum ratio between tributary and main stream b) the planform shape of
upstream and post confluence channel and angle of the confluence c) the difference
between the levels of tributary and main stream (Best, 1986; Leite Ribeiro et al.,
In the last decade, the development of hydrodynamic existing methods and new
methods and tools for investigation of complex flows especially in three dimensions
has greatly improved the understanding of the dynamics of confluences (Biron et al.,
2004; Bradbrook et al., 2000; Weerakoon and Tamai, 1989) Therefore, laboratory
studies combined with field observations are needed to link a global quantitative
2
model of channel confluences for better understanding of complex hydrodynamic
Despite more than six decades of research, sedimentation is still probably the most
serious technical problem faces by water resource manager and engineers. Such
problems include accelerated soil erosion, reservoir sedimentation and the wider
exploitation.
Sediment transport in small streams is diverse and highly variable due to the
equations were derived from limited flume experiments or field conditions (Bagnold,
1980; Camenen and Larson, 2005; Yang, 1996). In such conditions, equations based
on the relationship between the reliability and representativeness of the data utilized
in defining reference values, constants, and relevant coefficients are time consuming
and required complex parameter to estimate bed load transport. Although a known
particular stream reach at a particular time, the same equation usually overpredict or
underpredict the observed bed load transport by a different order of magnitude when
applied to a different river or even to the same river at a different time. Therefore,
there is a real need to consider and derive a simple equation to predict bedload
3
Kurau River is selected as the case study due to its importance as a main
domestic water supply and Kerian irrigation scheme areas in the state of Perak. Bukit
Merah reservoir and the dam that was constructed approximately at the mid section
of the Kurau River system requires the river management such as controlling the
Human activity includes the recently railway construction, changes in land use
from 2004 to 2015 according to the Taiping Town Council on Larut Matang Local
Plan 2015 (Hamidun, 2010), and increasing river sand mining makes change to river
hydrology and increase in sediment load along the river. The loss of river capacity
reducing the supply of irrigation water, water supply, and the effectiveness of flood
control schemes. Kurau River sedimentation becomes the main cause of frequent
higher sediment yield and overflowing water cause serious damages to the
environment, infrastructures and also has an effect on the social activity. Therefore,
Upstream of Kurau River as a selected case study consisting of two main river
tributaries namely Kurau River and Ara River. The river condition and morphology
can be different in each section of river. One of the complex and effective place of
the river due to sediment transport behaviour is the confluence of two river channels.
4
condition. Evaluation of the bedload transport in confluence requires the use of
transport
artificial neural network and genetic programming and evaluate the prediction
methods.
To evaluate the changes in bed load sediment transport, bed morphology and
This study was carried out on Kurau River, a natural stream in Perak, Malaysia.
Herein, the genetic programming, artificial neural network and nonlinear regression
models which are particularly useful in modelling processes with data interpretation
Hydraulic and sediment data were taken at six locations along Kurau River
and combine with the Lui and Semenyih Rivers data (Ariffin, 2004) for development
5
The performance of the genetic programming, artificial neural network and
statistical (nonlinear regression) models were evaluated and compared with six
bedload transport equations such as Meyer-Peter and Müller (1948), based on energy
slope method and Rottner (Yang, 1996), Chang (Cheng, 2002), Julien (2002) and
vanRijn (1993) based on regression method and Wong and Parker (2006) based on
this study for modelling the Ara-Kurau confluence. It solves the Navier-Stokes
diffusion equation for sediments. SSIIM uses the "k-epsilon" model for turbulence,
The field site for the modelling is the junction of the Kurau and Ara rivers in
Pondok Tanjung at the upstream of the Bukit Merah reservoir in Perak. The study
was carried at confluence limited in areas with approximately 141.5 m in length and
111.5 m in width.
Chapter 2 has a brief review about the headworks and different types of
6
the models and summary of model application relevant to this study was briefed in
this section.
Chapter 3 states some facts about the study for which this study has been done.
Data collection, data analysis and some soft computing method for predicting
bedload transport.
Chapter 5 illustrates the theory behind the SSIIM. It is not possible to go into
further detail due to dearth of space and time. Maximum reference has been made to
user manual for SSIIM. Manual in itself is quite explanatory. It is readily available
over the net. One of the nicety of this program or the liberality of the developer is
that this program is freely available over net with manual. This chapter also provides
the information the way the program is used herby. It includes the bedload transport
future study. Bibliography and appendices are enclosed at the end of this thesis.
7
2 CHAPTER 2 -
LITERATURE REVIEW
2.1 Introduction
variations because the transport of sediment through river channels has a major
Sediment transport in small streams is greatly variable and different due to the
ways unique because it is influenced by its own particular history of flow conditions,
material is called the stream load, and it is divided into bedload, suspended load, and
dissolved load (Figure 2.1). Bedload transport refers to the movement of bed
sediments along the stream bed by rolling, sliding, or jumping (Wang et al., 2011),
8
Bedload transpoort as a funndamental physical
p process in alluuvial rivers provides
p
manage riiver channeel morpholoogy. To claarify the caauses and effect of chaanges in
channel morphology
m and also too make info
ormed manaagement deccisions thatt affect a
2004).
distributioons found in
i sedimenttary deposits and to study
s the size sorting process
(Niekerk et
e al., 1992)).
9
Sediment size moves as bedload in rivers is important in sediment load
and older fluvial systems, boundary roughness and heavy mineral advancement
Bedload size distribution and bed material particle size specifications are
required to determine the sediment transport process (Ghoshal et al., 2010). The
basis for the prediction of sediment transport rates. Bedload size and bed material
demonstrate the size of material transported downstream and the size of material
resistance of the armoring layer and the ability of the stream to move surface
particles. The individual sediment size and the characteristic of the bed sediment
(Buffington and Montgomery, 1997; Church, 2006) and patterns, such as sheltering,
imbrications, armoring, and variations in sorting, can also affect the stabilities and in
turn the critical shear stress required to carry the sediment (Charlton, 2007; Clayton,
2010).
transport theory, the development of channel morphology, and are the basis for a
10
method of measuring the bed load transport rate (Pyrce and Ashmore, 2003).
is required when riverbed have different particle sizes, particularly in gravel bed
rivers due to the wide range of particle size. The movement of individual particles
depend on their relative as well as absolute size was shown by many researches that
using the field and laboratory sediment transport data. The overall transport rate of
mixed-sized sediments and the effects of changing sands and gravel contents were
studied in a laboratory flume (Curran and Wilcock, 2005; Wilcock and Crowe, 2003;
Wilcock et al., 2001). In an attempt to assess the evolution of bedload grain size,
Kuhnle (1989) worked on a stream with sand and gravel mixture. He discovered that
sediment size had a bimodal distribution and that sand fraction was entrained at
Fractional bedload transport has been studied in the field (Bond, 2004; Diplas,
1992; Kuhnle, 1989; Kuhnle, 1992; Powell et al., 2001; Wathen et al., 1995) and in
the laboratory (Wilcock and McArdell, 1993; Wilcock and Southard, 1989). A
determine the critical shear stress of each size fraction from five different sediment
beds (Kuhnle, 1993). All grain sizes of sand and gravel beds start to move at a nearly
identical shear stress. However, a constant relationship between critical shear stress
and grain sizes was observed in sand size sediments for the beds composed of sand–
gravel mixture, but for the gravel fraction, the critical shear stress increased with the
increase in size. Further studies show that most sand sizes may have nearly equal
entrainment mobility in both laboratory and field studies (Church et al., 1991; Parker
et al., 1982; Wilcock and Southard, 1989). The experiments were conducted in a
11
flume with mixed-sized sediments (Lanzoni and Tubino, 1999). Results show that
the capacity of the sediment transport be modified by the different mobility of the
diverse grain-size fractions in the mixture and induce a longitudinal and transverse
pattern in sorting.
in the Nahal Eshtemo River. Below this threshold, size selective occurs and above it,
a condition approaching equal mobility occurs. This range of threshold is about twice
principles and attempts to relate the level of bedload transport to several parameters
such as water discharge, shear stress or stream power (Martin, 2003; Yang, 1972).
One of the main problems in measuring bed material transport is that, under
more than 50 percent may be expected (Dietrich and Gallinati, 1991). Because of
formulae have been developed for a wide range of sediment sizes and hydraulic
the reliability and representativeness of the data utilized in defining reference values,
12
complete correlation. Therefore it is really difficult, if not possible, to recommend a
global equation for engineers to use in the field under all conditions (Camenen and
Numerous bed load transport equations have been formulated under limited
laboratory or field conditions that are available in the literature (Habersack and
Laronne, 2002). Table 2.1 to Table 2.7 are summary of bedload equations based on
derivation approach with their name and years and cited references.
Range of Cited
No Name Equation
applicability references
0.173
0.125≤ d50≤
qb 0
1 Du Boys (Yang,
d 34 0
50
(1879) 0.0125 0.019 d 50 4.0 (mm) 1996)
Sf > 0.00005
f cr
2 Kalinske 0.088≤ d35 ≤ (Yang,
u* s d50 0
qb
(1947) 45.3(mm) 1996)
f 0 c r
Sf > 0.00005
0.088≤ d50≤
u* s d 50
3 Grand and qb (Yang,
Albertson 45.3(mm) 1996)
u*3 u*2
(1961)
20 ≤ Re ≤1000 (Garde
F f ( n)
4 Sato,
Gs g u*c 2
0.088 ≤ d50 ≤
3
Kikkawa qb Gs gd50 and Raju,
and 5.66(mm) 2000)
u 2
F *2
Ashida 1
u*c u 2
1 8 *c
(1958)
u * 2
4
n 0 .0 2 5 : f ( n ) 0 .6 2 3
n 0 .0 2 5 : f ( n ) 0 .6 2 3 ( 4 0 n ) 3 .5
10qs f ( 0 cr )
qb
5 Shields 1.56 ≤ d50 ≤ (Yang,
(Gs 1) d50
(1936) 2.47(mm) 1996)
1.06 < Gs<
b 11( c r )1.65
4.20
6 Ribberink 0.088 ≤ d50 ≤ (Ribberink
(1998) 2.83(mm) , 1998)
b 12( cr )
7 Wilson 3
2
0.088 ≤ d50 ≤ (Wilson,
(1996) 2.83(mm) 1966)
13
Table 2.1: Continue
Range of Cited
No Name Equation
applicability references
4.93( 0.047)1.6
b
0.088 ≤ d50 ≤
3
8 Wong and (Wong and
3.97( 0.0495) 2
Parker 4(mm) Parker,
(2006) 2006)
10 Wiberg
b s ( c r ) 0.088≤ d35 ≤ (Wiberg
3
2
s 9.64( 0.166 )
and Smith 5.66 and Smith,
(1989) (mm) 1989)
b 5.7( cr )
13 Femandez 3
2
0.9 ≤ d50 ≤ 3.3 (Fernandez
-Luque (mm) Luque and
and Van 0.05 < θcr < Van Beek,
Beek 0.058 1976)
(1976)
Range of Cited
No Name Equation
applicability references
s gGs
2007)
qb s tan 0 vav eb
3 Bagnold 0.088≤ d50 ≤ 1.41 (Bagnold
(1966) (mm) , 1977)
14
Range of Cited
No Name Equation
applicability references
8( cr ) 3 2 cr
b
0.4≤ d50 ≤ 30(mm)
2 Meyer - (van Rijn,
0 cr
Peter and 0.25≤ Gs ≤3.2 1993)
Muller 1≤ Rh ≤<120 (cm)
0.0004≤ Sf ≤0.02
d90 0.06 vav 0.5
(1948)
b 4 s f ( cr )
3 Smart and 0.088≤ d50 ≤ 2.83 (Smart
d30 u*
Jaeggi (mm) and
(1983) 0.03≤ Sf ≤ 0.2 Jaeggi,
qb 10.217d500.594 s f 1.681q0.237
1983)
4 Pica 0.088≤ d50 ≤ 45.3 (Pica,
(1972) (mm) 1972)
Range of Cited
No Name Equation
applicability references
1 Abrahams 0.088≤ d50 ≤ (Abraham
b 1.5 (1 c r )3.4 ( a v )
and Gao v 5.66 (mm) s and Gao,
(2006) u* 2006)
d 0.088≤ d50 ≤
0.667 50 0.14
3
4 Rottner 2 3 (Yang,
Rh
(1959) 45.3 (mm) 1996)
qb s Rh va v
d 50
0.778
2 3
Rh
0.3 ≤ d50 ≤ 7
b 18.74( cr ) 0.7(cr ) (mm)
5 England (Engelund
and 12 12 and
Fredsoe θcr= 0.05 Fredsoe,
(1976) 1982)
15
Table 2.4: Continue
Range of Cited
No Name Equation
applicability references
0.2 ≤ d50 ≤ 2
0.053 1.5 cr
(mm)
b ( 1) 2.1
van Rijn Fr <0.9 (van
6 (1984,19 0.3 0.31<vav <1.29 Rijn,
87,1993) D* m/s 1993)
0.001≤ Sf ≤ 0.01
0.1≤ Rh ≤1 (m)
7 England 0.088≤ d50≤ 45.3 (Engelun
v
b 0.05 av 5 2
2
and (mm) d and
u*
Hansen Hansen,
(1967) 1967)
qb ( cr ) 1 2 (cr )1 2
0.088≤ d50 ≤
8 Fredsoe 30 (Fredsøe
and 45.3(mm) and
Deigaard Deigaard,
qb 17( c r ) 1 2 ( c r )1 2
(1992) 1992)
9 Ashida 0.088≤ d50 ≤ 45.3 (Ashida,
and (mm) 1972)
Michiue θcr= 0.05
18 g d50 2
(1972)
b
10 Julien 32 0.088≤ d50 ≤ 2.83 (Julien,
g (Gs 1)d503
(2002) (mm) 2002)
Sf > 0.0001
0.1<Θ< 1.0
d30 s
Qs
Q
(1991)
qb 4 s gd 50
13 Smart d50 < 29 (mm) (van
3
(1983) sf <0.2 Rijn,
1993)
d 0.2
50 s f 0.6 av ( c r )
v
d 90
u*
d =0.23
Garcia (mm) and
(1998) Garcia,
1998)
16
Table 2.4: Continue
Range of Cited
No Name Equation
applicability references
c r
b 12 0.5 exp 4.5
17 Camenen 0.088≤ d50 ≤ 5.66 (Camenen
and Larson (mm) and
(2005) Larson,
2005)
0.088 ≤ d50 ≤ 5.66
0.072078 0.353 T* 0.04 (mm)
18 Bhattachar T*0.898 (Bhattacha
ya, Price D* rya et al.,
b 0.000182 *0.0673 T* 0.04 and D* 181.3
and T0.13
2007)
Solomatine
D*
(2007)
0.0000124 *0.673 T* 0.04 and D* 181.3
T0.13
D*
Table 2.5: Bedload transport equations, Deterministic Discharge and velocity method
Range of Cited
No Name Equation
applicability references
d 1.8
(1935) (mm) 1935)
q c r 6.5 10 6 500.5
sf
0.305≤ d50 ≤ 7.02
qb (q qcr )
2 Sckoklitsch (Yang,
s
2 .5
3 2
S
(1934) f (mm) 1996)
0.24< vav≤ 0.0876
d 1 .5
q c r 0 .2 6 G s 1 1 0 6 5 07 6
5 3
sf
Sf >0.003
v v
0.088≤ d50 ≤ 45.3
qb 0.187qS f s av cr (mm)
3 Barekyan (Barekyan,
s vcr
(1962) 1962)
17
Table 2.6: Bedload transport equations, Deterministic Equal mobility method
Range of Cited
No Name Equation
applicability references
w*u*3s
qb , 50 ,
1 Pitlick et 2.0≤ d50≤ 45.3 (Pitlick et
al., (Gs 1) g cr (mm) al., 2009)
=
(1990a,b) u*2 0.79≤vav ≤ 1.13
(Gs 1) gd50sub (m/s)
0.853
11.9 1
2.9×10-4≤ Sf ≤
50 1.59
4.5
50
3.3×10-3
W * 0.00218exp 14.2 50 1 9.2850* 12
1.0 50 1.59
0.002550 50 1.0
14.2
w * u*3 s
qb , 50
2 Parker and 2≤ d50 ≤ 45.3 (Pitlick et
(Gs 1) g cr
,
Klingeman (mm) al., 2009)
=
and Mclem u*2 Sf >0.003
(1982) (Gs 1) gd50 sub θcr=0.0876
0.853
11.2 1 50 1.65
4.5
50
W 0.0025exp 14.2 50 1 50 12
*
w * u *3 s
4 Wilcock qb 2.0≤ d50 ≤ 45.3 (Pitlick et
(2001) ( G s 1) g (mm) al., 2009)
1 1 .2 1 0 .8 4 6 c r 0 cr
4 .5
0
W g*
0
0 .0 0 2 5 0 cr
cr
= , 50
u*2
( G s 1) g d 5 0 s u b cr
w * u *3 s
qb , 50
5 Wilcock 2.0≤ d50 ≤ 45.3 (Pitlick et
( G s 1) g cr
,
and Crowe (mm) al., 2009)
=
(2003) u*2
( G s 1) g d 5 0 s u b
0 .8 5 3
1 4 1 5 0 1 .3 5
4 .5
5 0 0 .5
*
W
0 .0 0 2 5 0 1 .3 5
7 .5
18
Table 2.7: Bedload transport equations, Deterministic Probabilistic method
Range of Cited
No Name Equation
applicability references
1 Einstein 0.315≤ d50 ≤ 28.6 (van Rijn,
b
(1942 (mm) 1993)
s s
qb 1
and 1.25≤ Gs ≤ 4.25
gd503
1950)
b 40 cr 1
3 Gill 3 0.088≤ d50 ≤ 2.83 (Gill,
0
(1972) (mm) 1972)
4 Parker
b 11.20
0.03 cr 2.83≤ d50 ≤ 5.66 (Pitlick et
4.5
(1979) (mm) al., 2009)
3
0.00035≤Sf ≤
0.0108
cr
r 1, =2.45
cr ( s )0.4
natural gravel bed rivers and 45 bedload transport observation from 3 flumes to
analyse some bedload transport equations. The authors conclude that there is no
equation to be tested performed consistently well, due to limited data used and the
19
The performance of 13 sediment transport formula in terms of their ability to
describe sediment transport was tested by Yang and Huang (2001) . They achieved
that the sediment transportation formulae based on the level of energy dissipation or
the concept of power flow, more accurately describe transported observed data. Also
the rate formulae complexity does not always translate into increased model
accuracy.
Prior to the extensive work of Yang and Huang (2001), Barry et al. (2004)
bed rivers with 2104 bedload transport observation in Idaho to evaluate the fitness of
concluded that there was no reliable relationship between formulae performance and
degree of calibration or complication. They found that transport data were best
transport equation and identify the channel and watershed characteristics effect on
The ability of the deterministic empirical equations of van Rijn (1984, 1993)
and Meyer-Peter and Muller (1948) was assessed by Claude et al. (2012) for a large
sand–gravel bed river to determine the unit and total bedload transport rates by
predictions. The authors concluded that the tested equations were unable to predict
the daily temporal variations of the total bedload transport at low and medium flow
conditions. The formulas described the bedload hysteresis but underestimated its
magnitude. For high flow conditions, the best agreement was observed for the total
20
bedload discharges computed by the van Rijn equation. The obtained results
indicated that the empirical equations only able to predict the temporal variations of
Schoklitsch (1950), Frijlink (1952), Yalin (1963), Bagnold (1980), Engelund and
Hansen (1967), Bijker (1971), Ackers and White (1973), Parker et al. (1982), van
Rijn (1984, 1987) and Cheng were evaluated with measured bedload by a Helley-
Smith sampler in the Node River, a gravel bed river in the northeast part of Iran
(Haddadchi et al., 2012). The results indicated that the statistic equation such as van
Rijn- Stochastic, Einstein and Bijker were not able to predict bed load in that gravel
bed river. Van Rijn, Frijlink and Myer-Peter and Mueller equations based on shear
stress achieved good results while some of them like Yalin and Cheng’s gave very
poor results. Equations based on the energy concept including Bagnold and Engelund
and Hansen equations tended to overestimate the real state in that river. Generally the
equations presented by van Rijn, Meyer-Peter and Mueller, and Ackers and White
might tolerably predict bedload transport in the range of field data of Node River.
variety of problems ranging from simple to complex physical and industrial systems.
If nothing is known a function may be assumed and fitted to experimental data on the
21
system. In other cases where the result of linear regression is unacceptable other
single explanatory variable X. In the simplest case the proposed functional relation
is:
Y 0 1 X (2-1)
In this model ε is a random error (or residual) which is the amount of variation
regression coefficients, are unknown and to be estimated. It will be assumed that the
error ε is independent and have a normal distribution with mean zero and variance σ2,
regardless of what fixed value of X is being considered. Then the value of 0 and 1
can be estimated by the method of the last squares (Bethea et al., 1995).
The multiple linear regression is similar to simple linear regression except that
dependent variable Y (Bethea et al., 1995). The general form of the multiple
Y 0 1 X 1 2 X 2 ... p X p (2-2)
where the ε is random error (or residual). The general form of multiple linear
or
22
Y 0 ( X 1 ) 1 ( X 2 ) 2 ...( X p )
p
(2-4)
The regression coefficients ( i ) are same to simple regression and can be obtained
The method has been adopted to find the best-fit line or curve from a given set of
data. In the standard formulation, a set of N pairs of observations {Yi , Xi} is used to
find a function relating the value of the dependent variable Y to the values of an
independent variable X . Assume that the set of data points are (x1,y1), (x2,y2), …,
(xp,yp) where x is the independent variable and y is dependent variable. The fitting
curve f(x) has the deviation (error) of ε from each data point, i.e., ε1=y1-f(x1), ε2=y2-
f(x2),..., εp=yp-f(xp). According to the method of least squares, the best fitting curve
SS E yi 0 1 X i minimum
p
2
(2-6)
i 1
To determine the minimum sum of square due to error (SSE), the partial
derivative of SSE which respect to each constant ( 0 , 1 ) is set equal to zero to yield:
yi 0 1 X i 2 0
( SS E ) p
0 0 i 1
(2-7)
23
yi 0 1 X i 2 0
( SS E ) p
1 1 i 1
(2-8)
0 Y 1 X (2-9)
1
( X X )(Y Y )
(X X )
i i i
2 (2-10)
i i
linear regression this method can be used to determine the regression coefficients of
i .
Y 0 1 X 2 X 2 ... p X p (2-11)
there is only one independent variable (X). Therefore the power of X can be
relationship between the dependent variable and a set of independent variables. The
(Bethea et al., 1995). This procedure estimates the parameter value that minimizes
24
the error sum of squares in a nonlinear least-squares routine. Because the model is
between independent and dependent variables. For example Equation (2.12) is a non-
The river flow condition and river environment have most effect on the
bedload transport rate in different river, and the computed results from various
equations often differ from each other and even from the measured data set.
Consequently the recent proposed equations need to be adopted for the new
Various kinds of soft computing techniques have been introduced and applied
in water engineering problems since the last two decades (Nagy et al., 2002). Soft
programming have been successfully applied. The regression method also has been
25
regression technique such as Yang (1996), Ariffin (2004) , Karim and Kennedy
nonlinear relationships between parameters (Liu et al., 2003; Nasseri et al., 2011;
GA, individuals are linear strings of fixed length (as chromosomes), whereas in GP,
individuals are nonlinear entities of different sizes and shapes (as parse trees). The
major advantages of GP are used in works where (i) the interrelationships among the
relevant variables are poorly understood (or where it is suspected that the current
understanding may well be less than satisfactory), (ii) finding the ultimate solution is
difficult, (iii) small improvements in the performance are routinely measured (or
easily measurable), (v) an approximate solution is acceptable (or is the only result
that is ever likely to be obtained), and (vi) conventional mathematical analysis does
presentations of GP can be found in Babovic and Abbott (1997) and Babovic and
Keijzer (2000).
26
applications, artificial intelligence, industrial, and mechanical models such as water
resources ,hydraulic processes and electricity demand, etc (Ashour et al., 2003;
Azamathulla et al., 2011; Babovic and Bojkov, 2001; Harris et al., 2003; Khu et al.,
2001; Muttil and Lee, 2005; Sivapragasam et al., 2006; Sreekanth and Datta, 2011;
Keijzer, 2003), sedimentary particle settling velocity equations (Babovic and Bojkov,
2001), emulating the rainfall runoff process (Liong et al., 2007; Whigham and
(Kizhisseri et al., 2005), modelling the bed material load for rivers (Zakaria et al.,
(1998) to enhance the GP accuracy. The amount of trees that can be employed is the
main difference between multigene GP and traditional GP. Several trees may
describe the model in multigene GP, whereas a single tree expresses the model in
traditional GP. All of the genes have specific optimal weights, and a summation of
weighted genes plus a bias term forms the final formula as the best obtained
where d0 is the bias term and di is the weight of the ith gene. Multigene GP is actually
27
GPTIPS was employed in this study to perform a multigene GP for accurate
Regression” code based on multigene GP for use with MATLAB (Searson, 2009b).
that variables do not need to be stationary and normally distributed for analysing
compare to classical stochastic model. ANN's structure can control the non stationary
effort (Ariffin, 2004). The application of ANN models is the topic of a large number
ANN is an algorithm designed after the function of the human brain, which
obtains knowledge through a learning process that involves finding an optimal set of
weights for the connections and threshold values for the nodes. A neural network
Each neuron multiplies every input by its interconnection weight, which is usually
determined through training the system, sums the product, and then transmits the
sum through an activation (or transfer) function to reach its result. This type of
network.
28
The network solves the problem by using the information giving from weights.
The net usually has two or more layers of processing units, where each unit in each
layer is connected to all of the processing units in the side layers. The desired output
is achieved by adjusting the weights on the links between the neurons, calculating the
value of error function for a particular input, and then back-propagating the error
The neural networks have been used for many branches of science. It is
becoming a strong tool for providing hydraulic and environmental engineers with
sufficient details for design purposes and management practices. The technique has a
growing body of applications for river engineering and water resources such as
resources concluded that the advantage of ANNs is their ability to extract the
relationship between the inputs and outputs of a process without explicitly providing
the physics to the user and have them reveal it back after training.
show that the model is capable of describing flow and sediment transport processes
in a river system. In addition, the ANN can be successfully applied for sediment
transport when other approaches cannot succeed due to the uncertainty and the
stochastic nature of the sediment movement (Chang et al., 2012; Kumar, 2012; Nagy
29
Among the numerous ANN structures, the multilayer, feed-forward network is
the most widely used in the area of sediment transport (Rumelhart et al., 1985). The
squares technique based on the backpropagation process was used in this study to
Transport
measurements of the Boise River to derive the bed load sediment transport formula.
They used four inputs to give the best balance between input variables and prediction
of sediment transport, namely: the grain Froude number (Fg), the grain Reynolds
number (Rg), the characteristics of the particle size distribution of the transported
sediment as the grain size standard deviation-mean sediment diameter ratio ( g/d50)
and the relative roughness (h/d50). For the application of sediment transport a 2 layer
1 output. They derived a pure advection equation by the linear ANN that able to
C 3.1
5.62(6.61Rg Fg )
34 4
1 31570e d50
(2-14)
2.52(7.38Rg )
1 3.11e *
h
predict sediment transport rates from field observations For the Boise River.
30
The performancce of approoaching ANN
N results coompared wiith 2 otherss derived
equations..
Marquardtt algorithm
m) Artificial Neural Neetwork (AN
NN) architeccture from bedload
measurem
ments in 16 different
d rivvers. The AN
NN with thee two inputss, two hiddeen layers
superior performance
p e for all staatistical parrameters exxcept for thhe discrepan
ncy ratio
31
1
(Table 2.8). The authors suggested that bedload transport in a variety of sediment
Table 2.8: Comparison of bedload equations and the ANN model (Sasal et al., 2009)
predicting bed material load transport, based on measured field data of several
Malaysian rivers. Figure 2.3 shows the scenarios of building the ANFIS model with
the inputs and output in the network. From the 346 collected data sets, around 80%
of these patterns were used for training (chosen randomly until the best training
performance was obtained), while the remaining patterns (20%) were used for testing
or validation. From the analysis, the ANFIS model obtained an accuracy of 90.4% in
predicting bed-load transport for all the measured data with an average discrepancy
32
Input Inputmf Rule Outputmf Output
d50 Cv
Ss
Figure 2.3: The ANFIS model for bed load sediment (Azamathulla et al., 2009)
Figure 2.4: Predicted bed load against measured bed load using ANFIS (Azamathulla et al.,
2009)
Programming (GEP) (Ab. Ghani and Azamathulla, 2012; Azamathulla et al., 2010;
33
Chang et al., 2012; Zakaria et al., 2010), Feed Forward Neural Networks (FFNN)
(Ab. Ghani et al., 2011), and Support Vector Machine (SVM) (Azamathulla et al.,
2010b) were evaluated in the prediction of total bed material load for three
Malaysian rivers namely Kurau, Langat and Muda. The results of evaluation
comparisons with traditional method were very good: FFNN (R2 = 0.958, RMSE =
0.0698), SVM (R2 = 0.958, RMSE = 0.0698), GEP (R2 = 0.97, RMSE = 0.057),
RMSE= 6.654), which supported the use of the soft computing techniques in the
demonstrate the predicted total bed material load against the measured total bed
Figure 2.5: Observed versus predicted sediment load by SVM for Langat, Kurau and
Muda rivers (Azamathulla et al., 2010b)
34
Figure 2.6: Observed versus predicted sediment load by FFNN for Langat, Kurau and
Muda rivers (Ab. Ghani et al., 2011)
25
Langat
Predicted Total bed material load (kg/s)
Kurau
20 Ideal fit
Muda
15
10
0
0 5 10 15 20 25
Obseved Total bed material load (kg/s)
Figure 2.7: Observed versus predicted sediment load by GEP for Langat, Kurau and
Muda rivers (Ab. Ghani and Azamathulla, 2012; Azamathulla et al., 2010a; Chang et al.,
2012; Zakaria et al., 2010)
35
2.8 River Channel Confluence
environments in which the combination of material (water and sediment) and energy
(power flow) of two different channels occur. The interaction of these components
provided insight into the complex flow structure and distinct geomorphic features at
natural confluences (Ashmore, 1993; Biron et al., 1993; Boyer et al., 2006;
Kenworthy and Rhoads, 1995; Lane et al., 1999; Rhoads and Sukhodolov, 2001;
Rhoads and Sukhodolov, 2004) and at laboratory junctions (Best, 1986, 1987; Best,
1988; Best and Roy, 1991; Mosley, 1976). This experiential research has been
numerical modelling (Baranya and Masa, 2007; Bradbrook et al., 1998; Bradbrook et
al., 2000; Bradbrook et al., 2001; Đorđević, 2012; Weerakoon and Tamai, 1989).
(discharge and momentum ratios of the combining flows), channel geometry (cross
material (grain-size distribution of the sediment load) (Đorđević, 2012; Leite Ribeiro
et al., 2012).
36
Flow at river confluences is three-dimensional and often characterized by the
presence of helical flow cells. The number of these cells, their presence or absence,
these cells have been studied in a long-standing contest (Ashmore and Parker, 1983;
Biron and Lane, 2008; Bradbrook et al., 1998; Fujita and Komura, 1988; Mosley,
1976; Parsons et al., 2007; Rhoads and Kenworthy, 1995; Rhoads and Sukhodolov,
2001).
The principal factors controlling the flow structure and channel morphology
are (1) the confluence angle and plan view (asymmetrical or symmetrical ) (Ashmore
and Gardner, 2008; Best, 1987; Best, 1988; Leite Ribeiro et al., 2012; Mosley, 1976),
(2) the discharge, and/or momentum ratios of flow and sediment between the two
confluent channels (Rhoads, 1996) , and (3) the bed elevation discordance between
the two confluent rivers (Best and Roy, 1991; Biron et al., 1993).
confluences (Szupiany et al., 2009) such as (1) a scour hole that is normally adjusted
along the region of maximum velocity where both flows begin to converge and mix;
(2) avalanche faces at the mouth of both river channels, which dip into a central
scour hole; (3) sediment deposition within the stagnation zone at the upstream
junction corner; and (4) bars formed within the flow separation zone at the
37
Table 2.9: Summary of the major foregoing studies considering the morphodynamics of channel confluences (Leite Ribeiro et al., 2012)
Confluence Reference Bed Discordance (1) Discharge Ratio Sediment Measurements Comments
(Mild/Moderate/ (Qt/Qm); Supply
Pronounced) (2) Momentum Flux
Ratio (Mt/Mm)
Bayonne-Berthier Biron et al. (1993a) Moderate (ratio between the (1) No information Natural Bed morphology Two investigated
confluence. height of the step and the (2) 0.68–2.02 regime and shear layer situations:
Angle of 65° and width ratio flow depth is around 0.35) turbulence 1) Bed load transport
Bt/Bm = 1 (Low gradient in both rivers and
rivers) 2) bed load transport only
in the tributary
Biron et al. (1993b) Moderate (ratio between the (1) 0.38–1.33 Natural Bed morphology and Measurements during the
height of the step and the (2) 0.18–2.04 regime shear layer dry season. Bed load
flow depth is around 0.35) turbulence transport only in the
tributary
38
Leclair and Roy (1997) Moderate (ratio between the (1) 0.29–1.87 Bed morphology Transport-effective flow
height of the step and the (2) 0.20–10.4 conditions
flow depth is around 0.35)
De Serres et al. (1999) Moderate (ratio between the (1) 0.38–1.33 Bed morphology,
height of the step and the (2) 0.18–2.02 flow velocity and
flow depth is around 0.35) turbulence
Roy et al. (1999) Moderate (ratio between the (1) No information Turbulence and bed
height of the step and the (2) No information load transport
flow depth is around 0.35)
t
Biron et al. (2002) Moderate (ratio between the (1) 0.57–1.48 Bed morphology,
height of the step and the (2) 0.71–2.22 turbulence and water
flow depth is around 0.35) surface topography
Boyer et al. (2006) Moderate (ratio between (1) 0.38–1.33 Bed morphology,
the height of the step and the (2) 0.18–2.03 3D velocity,
flow depth is around 0.35) turbulence and bed
load transport
Table 2.9: Continue
Confluence Reference Bed Discordance (1) Discharge Ratio Sedimen Measurements Comments
(Mild/Moderate/ (Qt/Qm); t Supply
(2) Momentum
Pronounced) Flux Ratio
Ruisseau du Sud Roy and Bergeron Mild (1) 0.45–0.65 (2) Natural Bed morphology, Tracking of
confluence. (1990) 0.28–0.50 regime flow velocity and different gravel
Angle of 60° and particle tracking size particles
width ratio Bt/Bm = 1
Kaskaskia–Copper Kenworthy and Mild (1) 0.64–6.64 Bed morphology
Slough confluence. Rhoads (1995) (2) 0.46–42 and sediment
Angle of 60° and width concentration
ratio Bt/Bm = 1
(Low gradient rivers)
39
(1976)
Angles between 15 and (2) No information in both confluents zone were free to adjust to the
imposed hydro-sedimentary
180° and width ratio
conditions and 2) only the
(Bt/Bm =1) confluence and the downstream
channel were adjustable
Birkbeck College, Best (1988) Mild (1) 0.5–1.6 Bed load transport Bed morphology Low gradient channels with
University of London (UK). (2) 0.25–2.42 of uniform material and particle Sub critical flow conditions.
Angles between 15 and 105° and in both confluents tracking Include a case study of a small
confluence in UK
width ratio Bt/Bm = 1
Ecole Polytechnique Fédérale Leite Pronounced (1) 0.11 Bed load transport Bed morphology, Small tributary with steep
de Lausanne (Lausanne, Ribeiro (2) 0.21 of poorly sorted water levels, 3D slope and transcritical flow
Switzerland). Angle of 90° and width sediments. Tributary velocities, (Fr ≈ 1), larger main
(2012) turbulence and bed channel with subcritical flow
ratio Bt/Bm = 0.30 Qst = 0.3 kg/min; Main
constitution
channel Qsm = 0
Ecole Polytechnique Fédérale Leite Pronounced (1) 0.11–0.23 Bed morphology, Small tributary with steep
de Lausanne (Lausanne, Ribeiro (2) 0.21–0.45 water slope and transcritical
Switzerland). Angle of 90° levels and bed flow (Fr ≈ 1), larger main
(2011) constitution channel with subcritical flow
and width ratio Bt/Bm = 0.30
2.9 Sediment Transport Modelling
River engineering studies typically are needed for analyse some level of spatial
dimensional sediment transport models are valuable tools for river engineering
investigations.
long-term channel response is used to predict future project operations and needs
required for planning and design purposes. A more specific approach is required for
both short and long term channel response evaluating in complex alluvial channels
that exhibit widely varying channel planform, morphology, and bed composition.
fluid, continuous, and momentum along with the differential equation for sediment
different physical fields easier than physical models, which are typically constructed
is that they are not subject to deformation effects of physical models, when a solution
can be obtained for the same flow condition that are present in the field
41
The following capabilities are required for the ideal hydrodynamic/sediment
transport model:
flow
Capable of performing multiple grain size analysis for both cohesive and non-
cohesive sediments
hydrodynamic simulation
visualization of results.
become very popular and developed over the past three decades, mostly due to the
1988; Rodi, 2006). Many computer models are now available for users to purchase
(FLOW-3D, FLUENT). Some of the models are in the public domain and can be
comfort to further advance the use of numerical models as a popular tool for solving
42
sediment transport models can be found in Onishi (1994), Blazejewski et al. (1995),
Spasojevic and Holly (2000), ASCE (2008) Sedimentation Engineering Manual no.
Table 2.10 provides the information on the model formulation, the spatial and
hydrodynamics, and the model’s predictive capabilities. This table provides useful
information about the model capabilities to handle unsteady flows, bed load and
and examples of the different model applications are summarized in Table 2.11.
43
Table 2.10: Summary of Some 3D hydrodynamic/sediment transport Models (Papanicolaou et al, 2008)
Bed Suspended Sediment
Last Sediment Cohesive Source
Model and references Flow sediment sediment exchange Executable Language
update mixtures sediment code
transport transport processes
ECOMSED: Estuarine,
Coastal, and Ocean Entrainment
V.1. 3
Model—Sediment Unsteady Yes Yes No Yes and PD PD F77
transport; Blumberg and (2002) deposition
Mellor (1987)
FAST3D: Flow
Analysis Simulation Entrainment
V.Beta-1.1
Tool; Landsberg et al. Unsteady Yes Yes No No and LD P F90
(1998) (1998)
deposition
Entrainment
Delft 3D; Delft V.3.25.00
Unsteady Yes Yes No Yes and C LD F77
Hydraulics (1999) (2005)
deposition
Entrainment
TELEMAC; Hervouet
--- Unsteady Yes Yes No Yes and C P F90
and Bates (2000)
deposition
Entrainment
Zeng et al. (2005) --- Unsteady Yes Yes No No and P P F90
deposition
Note: V= version; C= copyrighted; LD= Limited distribution; P= proprietary; PD= public domain; F = FORTRAN
Table 2.11: Applications for selected 3D models (Papanicolaou et al, 2008)
ECOMSED: Simulation of the flow and sediment transport processes of Lavaca Bay,
(Blumberg and Mellor, 1987) Texas (HydroQual 1998)
Simulation of the flow and sediment transport processes of the Klarälven River east and west channels at the
RMA-10:
bifurcation, Sweden (Admass 2005) Modelling of the Nisqually River Delta to evaluate habitat restoration
Associates; (King, 1988)
alternatives, Washington
GBTOXE: Modelling the hydrodynamics of flow and sediment of the Los Angeles and Long Beach harbors California
(Bierman, 1992) (Tetra Tech 2004) Simulation of fate and transport of PCBs in Green Bay, Wisconsin
46
EFDC3D:
Modelling of the hydrodynamic and sediment processes in Moro Bay, California
(Hamrick, 1992)
Simulation of flow and sediment transport of Lake Hartwell reservoir on the Savannah River between South
ROMS:
Carolina and Georgia Modelling of sediment transport and estuary turbidity maximum of the Hudson River
(Song and Haidvogel, 1994)
Estuary, New York
Simulation of flow and sediment quality of the Southern California Bight, California
CH3D-SED: Evaluation of the relative impact of different sediment sources on the shore areas of the western basin of Lake
(Spasojevic and Holly, 1994) Erie, Ohio (Velissariou et al. 1999) Simulation of sedimentation on bends, crossings, and distributaries on
the lower Mississippi River and Atchafalaya River, Lousiana
Table 2.11: Continue
SSIIM:
Tested against experimental data from Colorado State University (Olsen 2003)
(Olsen, 1994)
MIKE 3:
(Jacobsen and Rasmussen, Simulation of the flow, sediment transport processes, and water quality of Upper Klamath Lake, Oregon
1997)
FAST3D: Simulation of the flow, sediment transport processes, and water quality of Tampa Bay, Florida Tested against the
(Landsberg et al., 1998) experimental data of Odgaard and Bergs (1988)
47
Simulation of contaminated regions resulting from hypothetical airborne agent releases in major urban areas at
Washington D.C., Maryland, and Chicago, Illinois (Pullenet al. 2005)
Delft 3D;
Simulation of the flow, sediment transport processes and water quality of Tolo Harbor and Mirs Bay, Hong Kong
(Delft3D, 1999)
(Delft Hydraulics 1999)
Morphodynamic modelling of the German Wadden Sea and Duck, North Carolina (Delft Hydraulics 1999)
Development of a mesoscale hydrodynamic and sediment transport model for the Peru Basin in the Southeast
TELEMAC;
Pacific Ocean (Zielke et al. 1995)
Hervouet and Bates (2000)
Simulation of transport and Fate of Toxic Chemicals in Shasta Reservoir, California (Gu and Chung 2003)
(Zeng et al., 2005) Tested against the experimental data of Odgaard and Bergs _1988)
2.9.1 SSIIM
programs, the first type such as PHOENICS, STAR-CD, CFX, FLUENT and
FLOW-3D are general purpose programs and the second type include TELEMAC,
MIKE3, DELFT-3D, CH3D, TABS and SSIIM are absolutely developed for river
engineering.
and complete software is freely available over the net with user manual (Olsen,
2011).
SSIIM solves the Navier-Stokes equations using the control volume method
with the SIMPLE algorithm and the k-epsilon turbulence model that is based on an
sediment transport, using van Rijn's formula for the bed boundary (Olsen, 2011).
river bed in complex geometry. It also includes bed load and suspended load
transport modelling with multiple sediment sizes, bed forms and associated sorting
The program has an interactive graphical grid editor creating a structured grid.
The post-processor includes vector graphics, contour plots, profiles etc. which can
48
run simultaneously with the solver, enabling viewing of intermediate result. A post-
The model has been extended to other hydraulic engineering applications such
currents. The model has also been used for water quality and habitat studies in rivers.
The User's Manual (Olsen, 2011) gives more information about the SSIIM.
Equations. The SIMPLE algorithm was developed in the early 1970s. Since then it
has been extensively used by many researchers to solve different kinds of fluid flow
and heat transfer problems (Ghia et al., 1982; Karp et al., 2003).
Several well-established numerical schemes have been employed in the past for
solving flow and sediment transport model governing equations. The streamline-
element method (C-FEM), the fully upwind finite element method (FU-FEM), and
problem are more useful techniques available to solve numerically the fluid flow and
Helmig (1997).
49
The control volume method is in substance a finite volume formulation that
uses the integral forms of the governing equations. The domain of calculation is
divided into a random number of control volume and the equations are discretized by
calculating the number of streams that crosses the volume control boundaries
(Chung, 2002).
The main advantage of the control volume method is the flexibility of the
method that can be employed in both structured and unstructured grid systems.
Momentum, mass and energy can automatically conversed by the numerical scheme
Interior, 2011).
models. Each sediment transport model that is used as engineering tools has some
limitation for solving fluvial hydraulic problem. For that reason, selection of the
model. In this section some of the SSIIM applications for sediment transport
against field measurements for the Angostura reservoir in Costa Rica (Haun and
Olsen, 2012). The applied numerical model (SSIIM 2) solved the Reynolds-averaged
Navier-Stokes (RANS) equations in three dimensions and used to discretize the finite
volume method together with a second-order upwind scheme. The used grid was
50
adaptive and unstructured grid, which moved with the time-dependent changes for
both water and bed levels. Results from the numerical simulation of the deposition
and the flushing were compared with bathymetry data of the bed level from the
prototype.
The computations demonstrated that the deposition was easier to model than
the flushing. The amount of flushing out sediments shows reasonable agreement
compared with the measured data from the prototype. Therefore the simulation of a
of three-dimensional SSIIM model (Haun and Olsen, 2012). Figure 2.8 shows the
SSIIM was applied to compute uniform and nonuniform sediment transport and
cross section, and a slope of S = 0.005 (Feurich and Olsen, 2011). The sediment size
of 4.2 mm (gravel) was used as movable bed material. Significant good agreement
was found between the measured and computed bed elevations for Wu’s formula and
Vain Rijn's formula. Several parameters were tested in modelling such as grid
distribution in vertical, lateral, and longitudinal direction, time step, number of inner
iterations/time step, active sediment layer thickness, and the Shields coefficient. The
overall pattern variation of parameters tested gave some differences in the results, but
the total bed elevation changes gained the same value. The comparison of simulation
and measured longitudinal bed level changes are shown in Figure 2.9 and Figure 2.10
51
Figure 2.8: (a) Measured bed levels after the flushing (b) Simulated bed levels after the
flushing (Haun and Olsen, 2012)
52
Figure 2.9: Comparison of bed level changes: (a) measurements; (b) numerical simulation
with uniform sediment; and (c) nonuniform sediment (Feurich and Olsen, 2011)
Figure 2.10: Comparison between measured values and simulation results at: (a) cross
section 80; (b) cross section 60; and (c) cross section 20 (Feurich and Olsen, 2011)
located between Vienna and the Austrian-Slovakian border were computed using a
SSIIM model (Fischer-Antze et al., 2008). A time series of discharges during the
flood in 2002 was used for modelling. The Wu et al. (2000) formula was used to
53
compute the nonunniform sediiment transsport with hiding exxposure alg
gorithms
indicated the
t model is
i able to reepresent the relevant morphodynam
m mic processses, such
opposite side
s due to the related erosion pro
ocesses. Figgure 2.11 shows the measured
m
the discrettisation of thhe dendriticc flow domaains such ass the flow inn river confl
fluences.
54
4
SIMPLE algorithm is used to achieve the coupling of the continuity and
momentum equations in SSIIM. Due to high pressure and velocity gradients in the
confluence, the second-order upwind scheme is used for discrete convective terms in
the momentum equations. Đorđević (2012) validated the model SSIIM2 with using
both the experimental and field data and concluded that transfer of the momentum
from the tributary to the main river can be described satisfactorily using the 3D
model with the k-ε type turbulence model closure. Therefore, variations of the
recirculation zone width throughout the flow depth were predicted correctly.
2.10 Summary
sediment inflow on average balances the sediment outflow. A river, in effect, can be
the causes and consequences of changes in fluvial form and also to make informed
management decisions that affect a river’s function, it will require a good knowledge
geometry.
particles. The individual sediment size and the characteristic of the bed sediment
influence sediment transport. Bedload size distribution and bed material particle size
specifications are required to determine the sediment transport process. The extracted
parameter from affective factors on sediment transport can be used as a basis for the
55
Various bed load transport equations have been formulated under limited
laboratory or field conditions as mentioned in section 2.4. The river flow condition
and river environment have most effect on the bedload transport rate in different
rivers, and the computed results from various equations often differ from each other
and even from the measured data set. Consequently the recent proposed equations
genetic programming (GP) have been successfully applied for water engineering
problems since the last two decades. The good performance of ANN and GP
regression method also has been widely used to analyse and develop relationship
was mentioned in section 2.6 indicate these models which are particularly useful in
database predict well the bedload transport in different locations with different
circumstances.
A more specific approach is required for short and long term channel response
evaluating in complex alluvial channels such river channel confluences that exhibit
widely varying channel plan form, morphology, and bed composition. Multi-
this level of analysis. Each sediment transport model has its limitations for solving
the fluvial hydraulic problem. Therefore, the selection of right model under certain
56
constraints requires a satisfactory knowledge of the capabilities and features of
available models. The sediment transport model applications illustrate the capability
bedload transport in Ara -Kurau river channel confluence has been implemented. The
data collection and analysis will be discussed in Chapter 4 and result of sediment
57
3 CHAPTER 3
METHODOLOGY
3.1 Introduction
Study area, the methodology includes the study flowchart, data collection such
as hydrologic data, geometry data and sediment data were described in this chapter.
This chapter also provides information on the setting up of different methods that
include nonlinear regression, artificial neural network and genetic programming for
predicting the bedload transport rate. Figure 3.1 illustrates the research framework as
a study guide.
Literature review
Data collection
1. Geometry
2. Sediment
3. Hydrology
Define boundary
condition
NLR Modification of
ANN model equation
GP
Simulation
Calibration and
validation model
Conclusions
58
3.2 Study Area
Kurau River sub-basin lies between latitude 530,000 (N) and 570,000 (N),
longitude 683,300 (E) and 723,300 (E) in Zone 47 in UTM coordinate system. The
catchment area is approximately 1600 Km2, consisting of two main river tributaries
namely Kurau River and Ara River. The river starts partly in the Bintang Range and
partly in the Main Range where the territory in the upper reaches is steep and
mountainous. Mid valleys of the river are characterized by low to undulating terrain,
which give way to broad and flat floodplains. Ground elevations at the river
headwaters are moderately high, being 1,200 m and 900 m. The slopes in the upper
6.5 km of the river averaged 12.5% whilst those lower down the valleys are much
lower, of the order of 0.25% to 5%. Kurau River sub basin and data collection sites
included the Kurau- Ara confluence are shown in Figure 3.2 and Figure 3.3.
Land use distribution in the year 2004 shows that primary forests contribute
most at about 50% of the total area of Kurau River sub-basin, followed by rubber
dominate at about 24% of the total area, oil palm at about 11.28% of the total area. It
can be concluded that Kurau River sub basin is an undeveloped area with the
Rapid development in the Kurau River basin has led to an increased demand
for river sand as a source of construction material, which has resulted in the increase
in river sand mining activities that have rise to various problems. Kurau River is one
of selected river based on previous studies (DID, 2009) that sand mining activities
occurred in the river basin. The bed material sizes are in the sand-gravel range.
59
Figuree 3.2: Kurau River sub-basin and dataa collection ssites
60
0
3.3 River Hydrology and Hydraulic
Pondok Tanjung (Ara- Kurau confluence) and it has been operating since 1960 as
shown in Figure 3.4. The station is a well hydrometric site for low and high flow
measurement.
provided by the DID Hydrology Division from year 1960 to the year 2008. The
hydrographs for year 1970, 1972, 1986, 2006 and 2007 which present the highest
discharge at the Pondok Tanjung streamflow station are shown in Figure 3.5.
The historical water level record at the Pondok Tanjung streamflow station is
provided by the DID Hydrology Division from year 1960 to the year 2008. Figure
3.6 shows the water level chart for year 1970, 1972, 1986, 2006 and 2007 which
61
Figure 3.5: Discharge hydrograph for Kurau River at Pondok Tanjung
Figure 3.6: Water level chart for Kurau River at Pondok Tanjung
Figure 3.7 shows the flow rating curve for years 1996 to 2007. The shift in the
flow rating curve association reflects the variability of flow at the Pondok Tanjung
62
station at Kurau River. The rating curve for the year 2007 and 2002 are defined for
higher flow.
20
19
18
Stage (m)
17
16
1996 1997
15 1998 1999
2000 2001
2002 2003
2004 2005
14
0 20 40 60 80 100 120 140 160 180 200
Discharge (m³/s)
The ranking of flood over 48 years are given in Table 3.1. The review indicates
that the 2007 flood at the Pondok Tanjung streamflow station had the highest
discharge measured in 48 year period. Flood frequency analysis was carried out for
48 years of stream flow data using Normal distribution, Generalized extreme value, 3
Parameter Pearson, 3 Parameter lognormal, Gumbel Min, Log Pearson type III and 2
63
Table 3.1: Flood ranking for Kurau River at Pondok Tanjung
Figure 3.8 shows the measured stream flow data for the different type of
distributions. The most valid model was determined with the goodness of fit tests.
Chi-Squared test the most popular goodness of fit tests was used to compare the
fitted distributions. Since the goodness of fit test statistical indicates the distance
between the data and the fitted distributions, it is obvious that the distribution with
the lowest statistic value is the best fitting model. The generalized extreme value has
the lowest Chi- squared statistic value with better agreement with the measured
64
The goodness of fit tests can be used to compare the fitted distributions. Table
3.3 shows the value of Chi Squared indicating the best distribution for flood analysis.
The Generalized extreme value distribution with the lowest Chi squared was used for
considered as the design peak discharge and sediment transport study for Kurau
River.
Table 3.2: Summary of flood frequency analysis for Kurau River at Pondok Tanjung
Discharge (m3/s)
Return Generalized Pearson Log Normal Gumbel Log Log
period Extreme 3 normal 3 Distribution Min Pearson normal 2
Value Parameter Parameter Type III Parameter
65
Normal Distribution Lognormal 3 Parameter
220 220
200 actual data 200 actual data
prediction
discharge Q (m³/s)
180 prediction 180
discharge Q (m³/s)
160 160
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Weilbull Probability Weilbull Probability
220
Lognormal 2 Parameter 220
Pearson 3 Parameter
200 200 actual data
actual data prediction
180 180
discharge Q (m³/s)
prediction
discharge Q (m³/s)
160 160
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Weilbull Probability Weilbull Probability
Gumbel Min Log-Pearson 3
220 220
200 actual data 200
180 180 actual data
discharge Q (m³/s)
discharge Q (m³/s)
prediction
160 160 prediction
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Weilbull Probability Weilbull Probability
160
140
120
100
80
60
40
20
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Weilbull Probability
66
3.4 Field Data Measurement
The current study was conducted at six cross sections of the Kurau River from
January 2010 to January 2013 because of the difficulty in sampling and possibility of
wading in the water in these areas. Owing to bank erosion and severe bed
degradation, other locations were either inaccessible or impossible to wade into the
water. These sites and the data collected by Ariffin (2004) from two other sites in Lui
The confluence zone of Kurau and Ara Rivers was also selected for the
modelling part of this study. Hydraulic and sediment measurements were made along
a series of cross sections in April 2012, with each reach being separated by
approximately 15 to 20 meters.
Figure 3.9: Langat River basin and data collection sites by Ariffin (2004)
67
3.4.1 Flow Measurement
1976). Measurement taken includes flow depth (y0), velocity (V) and river width (B).
The flow velocities on the confluence zone of Kurau and Ara River at
approximately the same time as the river depth were also measured. A Hydroboard
Acoustic Doppler Profiler, ADP (SonTek River Surveyor core system; S5), and a
and Communications Module) connected to a laptop computer were used for this
68
Transect coordinates were received by the PCM at 10-Hz and transferred to the
ADP internal memory for integration and processing. It is then transmitted along
with the ADP data from the PCM to the laptop for navigation to transect start and
end points. As much of the wetted width of each transect was sampled as possible.
Due to the blanking distance (0.2 m) and mounting depth (0.2 m) of the transducer,
measured velocities include all but the top 0.5m of the water column. Survey data
was processed using SonTek’s River Surveyor (v3.10) software. River Surveyor was
used to create discharge summaries, export transects positional data (profile number,
velocity profiles for each transect. Data exported from the program for each transect
69
3.4.2 Geometry Data
The six cross sections and confluence zone of Kurau and Ara rivers were
surveyed using Electronic Distance Meter (EDM) (Table 3.4). Water surface and
bed elevation during different flow were also observed. The geometry data along a
series of cross sections were collected by using Sontek River Surveyor for sediment
Table 3.4: Typical cross sections along Kurau River (19 June 2010)
KRU1 28.0
27.0
26.0
Elevation(m)
25.0
24.0
23.0
0 5 10 15 20 25 30 35 40
Distance(m)
KRU2 20
19
Elevation(m)
18
17
16
0 2 4 6 8 10 12 14 16 18 20
Distance(m)
40
KRU3
39
38
Elevation(m)
37
36
35
0 5 10 15 20 25 30
Distance(m)
70
Table 3.4: Continue
40
KRU4
39
38
Elevation(m)
37
36
35
0 5 10 15 20 25 30
Distance (m)
36
KRU5
35
34
Elevation(m) 33
32
31
30
0 4 8 12 16 20 24 28 32 36
Distance(m)
55
ARA1
54
53
Elevation(m)
52
51
50
0 10 20 30 40 50
Distance(m)
71
Figure 3.12: River surveying at Ara River with river surveyor (ADP)
Bedload and bed material particle size distributions were selected to represent
River bed materials were collected by Van Veen grab sampler (Figure 3.13).
The width of the river was divided into seven spaced measuring points from left bank
to the right bank. The spacing between measuring points differs for one cross section
to the other and depends on the river width at different water levels.
72
Figure 3.13: Van Veen grab for bed material sampling
3.4.3.2 Bedload
In the last decade there has been an increase in interest in the measurement and
visualization of bed load movement in streams. As a result, there have been a series
repeated bed load sampling using Helley-Smith samplers (Bunte, 1990; Claude et al.,
2012; Helley and Smith, 1971; Ryan and Emmett, 2002; Sear, 2003), Brikbeck-type
slot samplers with pressure pillows (Garcia et al., 2000; Laronne et al., 2003; Sear et
al., 2000), magnetic induction devices (Bunte, 1996; Ergenzinger et al., 1994)
acoustic Doppler velocity devices (Claude et al., 2012; Ramooz, 2007; Rennie et al.,
2002) and hydrophones and impact sensing devices (Banzinger and Burch, 1990;
73
These devices could sample the pattern of movement across the stream width,
or the pattern and quantity of movement through time. Direct and indirect methods
sampling technologies and their applications can be found in Ryan et al. (2005)
The type of sampler was used based on the ease of handling, the sampling
efficiency and its operating cost (Ariffin, 2004). The hand Helley- Smith sampler
was selected in this research due to its ability to capture a wide sample material range
(0.5 to 16 mm), the high sediment trapping efficiency (Helley and Smith, 1971;
Yuqian, 1989) easy handling, suitable for short term measurement and low operating
cost.
In the field study, each section was sampled eight times. At the beginning of
each sampling event, water surface height was surveyed. The channel cross section
was then divided into eight equal spaced increments based on flow width at sampling
time. At each increment, flow depth and velocity were measured. Bedload was
sampled immediately after velocity at each increment. Bedload was collected with a
Helley-Smith bedload sampler made up of a square 7.6 cm orifice and 0.25 mm mesh
bag with frame and sampling durations ranging from 3-10 minutes, depending on the
74
Figure 3.14: Hand held Helley-Smith sampler for bed load sampling
developing such equations based on a balance between simplicity and accuracy. The
simple formula can estimate the bedload transport of small streams. Genetic
75
programming (GP) and artificial neural network (ANN) are powerful tools for
pattern recognition and data interpretation. They were employed and compared with
the nonlinear regression (NLR) method to present an explicit predictive equation for
Most of the equations depend on a lack of field data, a limited database, and
limited to special conditions developed; only a few are generally accepted for
practical use.
The Meyer-Peter and Muller (1948) Rottner (1959), Chang (2002), Julien
(2002), Wong and Parker (2006) and vanRijn (1993) are some of the most common
and popular equations used to estimate the bedload transport rate in rivers and are
sandy bed rivers and that the boundary conditions suit those of the Kurau River.
76
Table 3.5: The common bedload transport equations
Name Equation No
s g n
0.047 s d 50
3
qb 8
3 2
2
Meyer-Peter ( s ) nt
nt
and Muller 1 2
3
s 2R (3-1)
( 1948)
V
n 90
1
d 6
26
d d 3
3
0.05 s 2 R
b 13 exp 1.5
1 2
V
Chang 1.5
3
(3-3)
(Cheng, 2002)
18 g d 50 2 2
b
3
g Gs 1d 50
Julien
(3-4)
( Julien, 2002) 3
Wong and
ϕb=4.93(θ-0.047)1.6 0.088≤ d50 ≤ 4 (mm) (3-5)
0.053 1.5 c r
Parker (2006)
b ( 1) 2.1
vanRijn (1993) (3-6)
D*0.3
developing a general bedload equation (Khorram and Ergil, 2010). Based on the
theorem, the proposed influential parameter is the general form of the intensity of the
bed-load rate, b :
b qb
G s 1gd 50 3 (3-8)
G 1 g
Dgr d50 s 2
1
3
(3-10)
77
where qb (m2 ⁄s) is the volumetric bed-load sediment rate per unit width, Dgr is
dimensionless grain size. In this study the median grain-size diameter, d50, assumed
as the grain diameter ds. The terms Rh ⁄ds and Gs are embedded in the Shields’
parameter θ, and taking θ = f (Re), one can generate a rather simple relationship:
b = f (θ) (3-11)
b = αθ (3-12)
The bedload transport rate at Kurau River sites found to be similar to Barry et al.
(2004) and generally well described in log10 space (0.50 < R2<0.9) by a simple power
Tb= αQ β (3-13)
the fact that the prediction equation depends nonlinearly on one or more unknown
parameters. This method can be employed when there is the relationship between the
NLR can estimate models with random relationships between independent and
linear models.
Based on the fundamental data and the relationship between the variables, the
78
b m.qn . i .Dgr h s
0 Gs 1 .g.d503 (3-14)
where the following hydraulic parameters were used in the regression analysis: S0,
water surface slope, θ, Shields parameter, q stream discharge per unit width (m2/s),
dimensionless grain size Dgr, g, acceleration gravity, Gs, sediment specific gravity
A neural network toolbox contained within the MATLAB package was used in
this study. Bedload transport equations were integrated into a multilayer feed-
network with sigmoid hidden neurons and linear output neurons (fitnet) can fit
sufficient neurons in its hidden layer. Field data were provided and an appropriate
neural network structure was selected for training purposes. Training was performed
using the Levenberg–Marquardt backpropagation, where the input and output were
presented to the neural network as a series of learning. The network was set up with
the four parameters as the input pattern of discharge (Q), water surface slope (S0),
mean grain size (d50), and Shields parameter for the initiation of motion (θ) as most
influential parameters were widely used in bedload transport equations, and the bed
load transport rate Tb as the output pattern. In other words, the input layer contains
four neurons while the output layer contains one. Between the two layers, there is
another hidden layer that contains a suitable number of neurons under investigation
(Figure 3.15).
79
Inputs Hidden layer
Output
10
Q
S0
Tb
d50
A GPTIPS run with the following settings was performed: Population size =
pressure), Dmax = 3, Gmax = 4, Elitism = 0.01 % of the population, function node set =
(plus, minus, times, protected). The default GPTIPS multigene symbolic regression
function was used in order to minimize the root mean squared prediction error on the
Crossover events = 0.85, mutation events = 0.1, direct reproduction = 0.05. The
following sub-event probabilities were used: high level crossover = 0.2, low level
crossover = 0.8, subtree mutation = 0.9, replace input terminal with another random
standard deviation of Gaussian = 0.1) (Table 3.6). These settings are not considered
‘optimal’ in any sense but were based on experience with modelling different data
80
The selection of appropriate model input variables in GP, as with any data-
generally based on a previous knowledge for most influential variables and physical
insight into the problem (Khorram and Ergil, 2010). Four input parameters including
discharge (Q), water surface slope (S0), mean grain size (d50) and Shields’ parameter
bedload transport equations as variable data and Tb (bedload rate) as invariable data
Parameter Range
Population size 500
Function set +, -, *, /
Number of generations 25
Maximum number of genes 4
Maximum number of nodes per tree 13
Maximum depth of trees 3
Probability of GP tree mutation 0.1
Probability of GP tree cross over 0.85
Probability of GP tree direct copy 0.05
81
4 CHAPTER 4
BEDLOAD TRANSPORT CHARACTERISTICS
4.1 Introduction
the bedload transport rate is of high importance for detailed study and improvement
of models for prediction of bedload transport rate, prediction of channel change, and
Millar, 2004).
In this section detailed analyses for bedload and bed material characteristics
were performed for upstream and downstream of Kurau River. Size gradation of
bedload and bed material was analysed in relation to shear stress, and flow
discharges. Differences in bed load size distributions depending on the type of the
fractional transport rate was determined for each location as a function of the particle
size to assess the relative mobility of various size classes in the upstream and
In this section NLR, ANN and GP river system models were used to simulate
and predict bedload transport in Kurau River. These models were employed for other
small streams. Data from six sediment stations on Kurau River in Perak and two
sediment stations in Lui and Semenyih River in Selangor (Ariffin, 2004), were
compiled to obtain the formula as well as for comparison with other existing bedload
82
transport formulas. The performances of the GP, ANN, and statistical (NLR) models
for small streams were evaluated and compared with five bedload transport formulas
such as Meyer-Peter and Muller (1948) based on the energy slope method; Rottner
(1959), Yang (1996), Chang (Cheng, 2002), van Rijn (1993), and Julien (2002) based
on the regression method; and Wong and Parker (2006) based on the shear stress
method.
Data of the six channel criteria ranged from 20000 km long in the drainage
area and included a variety sand–gravel bed channels. Discharges ranged from 0.55
m3/s to 12.79 m3/s. All cross sections in the Kurau River have a single thread channel
width. The top width ranged from 7 m to 19 m; the stream gradients ranged from
0.0007 to 0.001; median particle size of bed material (d50) ranged from 0.65 mm to
1.84 mm. Most channels were bounded by flood plains or alluvial terraces and were
able to adjust freely to discharge sediment inputs. The dynamics of the river are
relatively natural because the structures (i.e., bridges and some bank protection
hydraulics parameter and sediment data (grain size distribution, d50), were gathered
from the Lui River, Semenyih River (Ariffin, 2004) and Kurau River as small
streams. The range of measured data is shown in Table 4.1. Figure 4.2 to Figure 4.6
show cross sectional changes at different flow discharges. Lower bed elevations
83
Molinas and Wu (2001) categorized the rivers in large and medium by flow
depth, flow discharge and flow velocities. They pointed out that large rivers refer to
those with yearly average flow depths greater than 4 m, and medium rivers refer to
those with yearly average flow depths between 2 m and 4 m. They mentioned that
large rivers have flow discharges more than 134 m3/s for, flow velocities bigger than
0.21 m/s, flow depths more than 3 m, water surface slopes in the range of 0.02×l04 to
1.8×104 and median bed material diameters in the range of 0.09 mm to 0.99 mm. The
flow discharges for medium rivers are between 13 m3/s to 4791 m3/s, flow velocities
in the range of 0.20 m/s to 2.30 m/s, flow depths in the range of 1.50 m to 9.29 m,
water surface slopes in the range of 0.06×104 to 25×104, median bed material
diameters in the range of 0.02 mm to 2.60 mm. The summary of some rivers (large
and medium) data is shown in Table 4.2. The range of data in Kurau, Semenyih and
Lui rivers such as flow discharge (0.55-17.2), flow depth (0.23-1.15) and etc... are
not the in the range of large and medium rivers as mentioned above. Consequently
these three rivers are considered as small rivers. Details of the present data for Kurau
84
Table 4.1: Range of field data for Kurau , Lui and Semeneyih
No
Q
River Location of V (m/s) So×10-2 B (m) Y0 (m) A (m2) R (m) d50 (mm) Tb (kg/s)
(m3/s)
data
KRU1 8 3.18-12.8 0.53-0.82 0. 05-07 17-19 0.47-1.15 6-15.51 0.412-0.885 0.65-1.044 0.23-2.10
KRU2 8 1.6-6.1 0.5-0.73 0. 07-1.85 9-10.3 0.42-1.15 2.87-8.37 0.313-0.76 0.699-1.084 0.17-0.86
Kurau KRU3 8 0.55-1.52 0.31-0.52 0. 06-0. 96 7-9.2 0.28-0.38 1.39-2.89 0.166-0.303 0.99-1.404 0.03-0.26
(present data)
KRU4 8 0.56-4.7 0.15-1.22 0. 1-0. 62 13-Dec 0.27-0.52 1.99-6.03 0.161-0.286 1.02-1.83 0.01-0.50
85
KRU5 8 2.32-6.6 0.49-1.56 0. 03-0. 51 13-Dec 0.37-1.03 3.46-9.78 0.224-0.699 0.74-1.51 0.13-1.52
ARA1 8 0.77-5.25 0.4-0.69 0. 03-3.12 11.3-13 0.27-0.86 1.94-7.57 0.167-0.567 1.29-1.84 0.12-1.04
Lui Kg Lui 92 0.7 – 17.2 0.2 – 1.0 0.03 - 0.93 15-15.5 0.23– 0.99 3.42-16.84 0.221-0.887 0.50 – 1.74 0.04-1.55
(Ariffin, 2004)
Semenyih Kg.
(Ariffin , 2004) Rinching
50 2.6 – 8.0 0.4 – 0.9 0.23 – 1.5 13-15 0.36 – 0.82 5.42-11.49 0.345-0.735 0.88 – 2.29 0.65-3.15
Table 4.2: Summary of large and medium rivers (Monalis and Wu, 2001)
Flow Flow Flow Water Median Bed-Material
Data source Discharge Velocity Depth Surface Diameter Concentration
3 4
(m /s) (m/s) (m) Slope×l10 (mm) (ppm)
(a) Large Rivers (dyr (i)> 4.0 m)
Amazon and Orinoco River Systems (Posada 1995) 134-235000 0.37-2.42 3.56-62.33 0.14-1.8 0.093-0.90 0.1-2360
Mississippi River System (Posada 1995) 332-4100 0.37-1.77 3.17-21.80 0.03-1.8 0.18-0.99 0.2-370
Atchafalaya River at Simmesport (Toffaleti 1968) 382-14188 0.21-2.03 6.10-14.75 0.02-0.51 0.091-0.31 0.6-570
Mississippi River at Tarbert Landing (Toffaleti 1968) 4228-48830 0.62-1.61 6.74-16.40 0.18-0.43 0.18-0.33 12-260
Mississippi River at St. Louis (Toffaleti 1968) 1512-21608 0.62-2.42 4.66-17.28 0.25-1.34 0.18-1.15 7-510
Red River at Alexandria (Toffaleti 1968) 190-1538 0.37-1.14 3.00-7.38 0.66-0.82 0.10-0.22 8-500
Total of Large Rivers 134-235000 0.21-2.42 3.00-62.33 0.02-1 .8 0.091-0.99 0.1-2360
86
Kurau River surveyed cross sections are shown in Figure 4.1 to Figure 4.6.
The figures indicate the changes in bed morphology during the data collection time.
The maximum degradation occurred in KRU1 around 0.8 m during the minimum
28
27
26
25
Elevation(m)
24
23
22
0 5 10 15 20 25 30 35 40
distance (m)
28/04/2010 Q=7.21 mᵌ/s 11/11/2010 Q=5.21 mᵌ/s
29/12/2010 Q=5.58 mᵌ/s 19/01/2011 Q=3.99 mᵌ/s
24/02/2011 Q=12.79 mᵌ/s 9/3/2011 Q=4.91 mᵌ/s
20
19
18
Elevation(m)
17
16
15
0 2 4 6 8 10 12 14 16 18 20
Distance(m)
19/05/2010 Q=1.6(mᵌ/s) 12/10/2010 Q=2.1(mᵌ/s)
1/12/2010 Q=6.1(mᵌ/s) 19/01/2011 Q=2.25(mᵌ/s)
16/02/2011 Q=1.65(mᵌ/s) 3/03/2011 Q=1.95(mᵌ/s)
87
20
19
18
Elevation(m)
17
16
15
0 2 4 6 8 10 12 14 16 18 20
Distance(m)
5/7/2010 Q=0.79 mᵌ/s 5/10/2010 Q=0.55 mᵌ/s
29/12/2010 Q=1.03 mᵌ/s 6/1/2011 Q=0.66 mᵌ/s
16/2/2011 Q=1.32 mᵌ/s 9/3/2011 Q=0.62 mᵌ/s
11/05/2011 Q=1.52 mᵌ/s 2/06/2011 Q=0.72 mᵌ/s
40
39
38
Elevation(m)
37
36
35
0 5 10 15 20 25 30
Distance(m)
19/05/2010 Q=0.73 mᵌ/s 5/10/2010 Q=1.33 mᵌ/s
20/12/2010 Q=0.56 mᵌ/s 26/01/2011 Q=1.18 mᵌ/s
8/2/2011 Q=2.59 mᵌ/s 16/02/2011 Q=1.41 mᵌ/s
5/5/2011 Q=4.7 mᵌ/s 9/6/2011 Q=2.21 mᵌ/s
88
35
34
33
Elevation(m)
32
31
30
0 2 4 6 8 1012 14 16 18 20 22 24 26 28 30
Distance(m)
28/06/2010 Q=6.44mᵌ/s 26/05/2011 Q=4.6 mᵌ/s
12/10/2010 Q=2.32 mᵌ/s 20/12/2010 Q=4.06 mᵌ/s
6/1/2011Q=5.68 mᵌ/s 8/2/2011 Q=5.39 mᵌ/s
24/02/2011 Q=6.6 mᵌ/s 21/06/2011 Q=2.23 mᵌ/s
55
54
53
Elevation(m)
52
51
50
0 5 10 15 20 25 30
Distance (m)
12/5/2010 Q=1.27 mᵌ/s 27/10/2010 Q=0.776 mᵌ/s
1/12/2010 Q=5.25 mᵌ/s 6/1/2011 Q=2.29 mᵌ/s
1/2/2011 Q=1.19mᵌ/s 3/3/2011 Q=1.02mᵌ/s
5/5/2011 Q=1.68 mᵌ/s 2/6/2011 Q=2.29 mᵌ/s
89
4.2.3 Parameter Affecting Bedload Transport
All measured variables were plotted against the bedload transport to indicate
the correlations of different parameters and to be used as basis for developing new
bedload transport equation. The scatter plots of this variable against bedload
10
kurau River
Bedload transport rate Tb(kg/s)
Lui River
Semenyih River
1
0.1
0.01
0.1 1 10
Discharge Q (m3/s)
10
Kurau
Bedload transport rate Tb(kg/s)
Lui
Semenyih
1
0.1
0.01
0.01 0.1 1 10
Velocity V ( m/s )
90
10
Kurau
0.1
0.01
1 10 100
Width B (m)
10
Kurau
Bedload transport rate Tb(kg/s)
Lui
Semenyih
1
0.1
0.01
0.01 0.1 1 10
Water depth Yo (m)
Figure 4.10: Scatter plot of bedload transport rate against water depth
10
Kurau
Bedload transport rate Tb(kg/s)
Lui
Semenyih
1
0.1
0.01
1 10 100
B/Y ratio
Figure 4.11: Scatter plot of bedload transport rate against B/Y ratio
91
10
Kurau
0.1
0.01
0.01 0.1 1 10
Hydraulic radus R (m)
Figure 4.12: Scatter plot of bedload transport rate against hydraulic radius
10
Kurau
Bedload transport rate Tb(kg/s)
Lui
Semenyih
1
0.1
0.01
0.1 1 10 100
Area A ( m2 )
10
Bedload transport rate Tb(kg/s)
0.1
Kurau
Lui
Semenyih
0.01
0.0001 0.001 0.01 0.1 1
Slope S0
92
10
0.1
Kurau
Lui
Semenyih
0.01
0.1 1 10 100
Median grain size d50 (mm)
Figure 4.15: Scatter plot of bedload transport rate against median grain size
Analysis of particle size distributions of bedload and bed material for different
The particle size distributions of bedload and bed material are illustrated in
Figure 4.17. The results show that the bedload material is finer than the surface bed
material for all analyzed sites. The median bedload particle size and median particle
of bed material are less than unity in the upstream of the Kurau River. This finding
demonstrates the size selectivity of bedload transport during the different water
In most of the samples analysed, sand and fine gravel were the main fractions
of the bedload transport rate for the measured range of discharges, and the size
93
fractions enlarged with the increase in discharge. The source of fine material could
be from external sources or material from the bed surfaces (i.e., fine material
The bedload frequency curves obtained from the upstream of the river were
mainly bimodal and were unimodal only in a few cases. The unimodal bedload
frequency curve indicates that uniform fine material is present in the bedload sample,
whereas the bimodal curve shows sand and gravel modes with some concentration in
special sizes because of the mobilization of coarser bed particles during higher
unimodal, and the size of fractions was enlarged, following approximately the same
range.
The presence of sand and fine gravel in most of the bedload particle size
distributions in the downstream was caused by the fine sediment coming from the
upstream network during flood events. The coarser fractions of bedload particles
were transported by higher discharges, but they were generally smaller than those
formed on the bed surface. The median particle size of the bedload sample, even for
the highest measured values of the bedload transport rate, was equal or smaller than
the median particle size of bed material. Comparison of the distribution size of the
downstream (KRU1) of the Kurau River (Figure 4.18) indicates that the amount of
sediment particles of each fraction size in the upstream is greater than that in the
94
30
Upstream
25
Percentage retained (%)
20
15
6.44 (m³/s)
10 5.39 (m³/s)
4.6 (m³/s)
5 2.23 (m³/s)
0
0.01 0.1 1 10
Particle size (mm)
30
Downstream
25
Percentage retained (%)
20
15 7.21 (m³/s)
5.58 (m³/s)
10 12.79 (m³/s)
3.18 (m³/s)
5
4.91 (m³/s)
0
0.01 0.1 1 10
Particle size (mm)
Figure 4.16: Bedload frequency distribution size of upstream (KRU5) and downstream
(KRU1) of Kurau River
95
100
90 Upstream
80
Percentage passing (%)
70
60
50 3.18 (m³/s)
40 4.91 (m³/s)
30 5.58 (m³/s)
7.21 (m³/s)
20
12.79 (m³/s)
10
bed material
0
0.001 0.01 0.1 1 10
Particle size (mm)
100
90 Downstream
80
70
Percentage passing (%)
60
50
2.23 (m³/s)
40
4.6 (m³/s)
30
5.39 (m³/s)
20
6.44 (m³/s)
10
bed material
0
0.001 0.01 0.1 1 10
Particle size (mm)
Figure 4.17: Particle size distributions of bedload and bed material samples for Kurau River.
96
100
90
80
60
50
40
30
20
Downstream
10
Upstream
0
0.01 0.1 1 10
Particle size (mm)
Figure 4.18: Comparison of particle size distributions of bedload samples for upstream and
downstream of Kurau River in same discharge.
Figure 4.19 shows the size distribution of bedload sampled at different shear
stresses and discharges. Unlike in the following analyses, all material in the upstream
and downstream was included, and no upper size truncation was applied. To clarify
the emerging pattern, the grain size distributions for all discharges and all samples
were demonstrated in each shear stress, and an average size distribution was derived
for each shear stress. Individual size classes were classified into five groups to
represent fine sand (<0.3 mm), medium sand (0.3 mm to 0.71 mm), very coarse sand
97
Table 4.3: The classifiication of seddiments by particle
p size according
a to the Wentworth scale
100%
90%
80%
Percentage passing (%)
70%
60%
50%
40% 5.33-25 mm
30% 2-5.3 mm
20% 0.771-2 mm
0.33-0.71 mm
10%
<00.3 mm
0%
Figure 4.119: Mean bedd load grain size distributtions for shear stress bannds arranged in order
of incrreasing shearr stress (upstrream of Kuraau River KR
RU5).
98
8
At the lowest flow in the upstream of the Kurau River ( = 5.4 N/m2), around
50% of the load was medium sand. The remainder was dominated by coarse sand
(30%) and granules (18%); only 2% accounted for sediment coarser than 5.3 mm.
With increasing shear stress, the grain size distribution became coarser. The
proportion of bedload in fine sand and medium sand reduced, and the proportion of
granules and fine pebbles increased. Interestingly, the proportion of coarse sand did
not change significantly over the range of monitoring flows. However, from the point
at which shear stress was 13.5 N/m2 and 14.2 N/m2 by increasing the flow, the
granule size increased and the transience of fine pebbles was observed. In the
upstream, the movement of fine size sand depended largely on its availability within
the channel, leading to the decrease in transport rates after the peak discharge.
Similar to the upstream at the lowest flow, in the downstream of the Kurau
River (Figure 4.20), most of the material load was medium sand and coarse sand,
with only 30% granules and approximately 5% fine pebbles. As shear stress
increased, the grain size distribution became coarser. The proportion of bedload in
fine sand and medium sand decreased, and the proportion of granules and fine
pebbles increased, but the amount of coarse sand did not change in all shear stresses.
The increase in the movement of sand at high flow and high shear stress in the Kurau
River was caused by the unavailability of coarse granules and fine pebbles in the
river. As mentioned previously, the Kurau River is a sand–gravel bed river with a
99
100%
90%
80%
g ((%))
70%
g ppassing
60%
5.3-25 mm
Percentage
50%
40% 2-5.3 mm
30% 0.71-2 mm
0.3-0.71 mm
m
20%
<0.3 mm
10%
0%
34.87
60.77
2.82
3 07
3.07
3.70
Shear stress (N
N/m²)
Figure 4.220: Mean bedd load grain size distributtions for shear stress bannds arranged in order
of increaasing shear stress
s (downsstream of Kuurau River K
KRU1).
The changes in
i distributtion size frrom fine sand
s to coaarser fractiion with
from 1.5 mm
m at low flow
f to 3 mm
m at high fllow.
in the upsttream of thee river. To describe beetter the deppendence off grain size on shear
stress for each dx, thee trend line is illustrateed in this study. As shoown in Figu
ure 4.21,
the gradieent of trend line from d10 to d90 off each sampple increasees by increaasing the
100
0
upstream, the trend of fine size (i.e., d10, d16, and d30) is mostly straight, and the trend
of coarse size is soft. However, in the downstream, the increasing grain size is
steady. The gradients are steeper than those upstream because of the existence of
10
Upstream
Particle size (mm)
d10
1 d16
d30
d50
d60
d84
d90
0.1
5 7 9 11 13 15
10
Downstream
Particle size (mm)
d10
1 d16
d30
d50
d60
d84
d90
0.1
0 20 40 60 80
Figure 4.21: Variation in grain size at the10th, 16th, 30th, 50th, 84th and 90th percentiles of
the bedload size distribution with increasing shear stress.
101
4.5 Fractional Transport Rate
The previous analysis demonstrates that bedload grain size varies with
referring to the size distribution of bed material available for transport in the same
section. Wilcock and Southard (1989) normalized the fractional transport rates by
dividing all the various fractions by the corresponding proportion f in the sediment
i
bed to obtain equal mobility. Therefore, the ratio of the fractional transport rate of a
given size fraction to the proportion of the given size fraction in the bed sediment is
The fractional transport rate piqb, was determined for each location as a
function of the particle size to assess the relative mobility of various size classes for
all samples in the locations. The results describe the comparison between bedload
and bed material grain size. The fraction of bedload particle size in the ith size range
is pi, and qb is the total transport, calculated as the mean for the sampling period.
Results of pi/fi for different values of Q were plotted, as the overall transport
rate varies with the applied discharges (Figure 4.22 and Figure 4.23). Figure 4.22a
and Figure 4.23a show the range of sediment sizes present in all recorded discharges.
The degree to which the curves revolve from the horizontal indicates how much the
particle size distribution of the bedload departs from that of the bed material
Based on Figure 4.22a, the middle range sediment size has an equal rate of
bedload transport and bed material. Fine sediment was loaded less in the upstream
and downstream compared with the bed material. However, the fine sediment size
102
loaded more at low flow and less at high flow compared with the bed material
distributions because the flow moves finer sediment more easily than coarser
sediment from the bed. In contrast, at high flow, the fine sediment loaded less than
the coarser sediment. In this condition, the increase in flow increased in shear stress
10.00
a
1.00
Pi/Fi
2.23 (m³/s)
0.10
4.6 (m³/s)
5.39 (m³/s)
6.44 (m³/s)
0.01
0.01 0.1 1 10
Particle size (mm)
b
Scaled Fractional transport rate (kg/ms)
10
2.23 (m³/s)
0.1 4.6 (m³/s)
5.39 (m³/s)
A B 6.44 (m³/s)
0.01
0.01 0.1 1 10
Particle size (mm)
Figure 4.22: Transport ratio as a function of grain size at upstream (a) the transport ratio Pi/fi
where pi is the proportion of each size fraction i present in transported material and fi is the
proportion of each size fraction in the bed material (b) the scaled fractional transport rate
computed as qbpi/fi, where qb is the sediment transport rate.
103
10.00
a
1.00
Pi/Fi
3.18 (m³/s)
0.10
4.91 (m³/s)
5.58 (m³/s)
7.21 (m³/s)
12.79 (m³/s)
0.01
0.01 0.1 1 10
10
b
Scaled Fractional transport rate (kg/ms)
3.18 (m³/s)
0.1
5.58 (m³/s)
4.91 (m³/s)
7.21 (m³/s)
A B 12.79 (m³/s)
0.01
0.01 0.1 1 10
Particle size (mm)
Figure 4.23: Transport ratio as a function of grain size at downstream (a) the transport ratio
Pi/fi (b) the scaled fractional transport rate qbpi/fi.
The range of 0.7< pi/fi <2 was selected to define the balance condition
approximately. This range indicates that the transport proportion of the bedload is the
same as that of bed material of this size. The departure from the value pi/fi range was
104
Figure 4.22b and 4.23b demonstrate that the fractional sediment transport ratios
for the given flows have a similar range, except the fine sediment side and the coarse
sediment side (i.e., outside of lines A and B). Fractions finer than line A on the fine
sediment side are relatively rare in the bedload than in the bed material. The
condition of equal mobility may seem confusing because transporting the coarser
fractions is more difficult than transporting the finer fractions for a flow. Larger
particles are more difficult to move because they are heavier. This condition is
important countervailing effects tend to offset the particle–weight effect. First is the
hiding–sheltering effect, in which larger particles are more exposed to the flow, and
thus a greater fluid force is exerted on them. However, smaller particles tend to be
sheltered from the forces of the flow by the larger particles (Einstein, 1950 ). Second
is the rollability effect, in which larger particles can roll easily over a bed of smaller
particles but not the other way around. The relative importance of the particle–
weight's effect and the combination of the hiding–sheltering' s effect and the roll
2006). The hiding -sheltering and rollability effect usually occurred on the armored
bed surface. Armoring is a small-scale sorting process that results in a thin layer of
coarse grains at the bed surface (Sutherland, 1987). The presence of an armor layer
on the bed surface is a common phenomenon in rivers. Two types of armor layers
can be distinguished: stable armor layers and dynamic armor layers. Dynamic armor
layers develop if the bed shear stress is large enough to transport both the fine and
upstream) and the innate difference in mobility between coarse and fine grains
causes the fine grains to be winnowed from the bed surface, overexposing the coarse
105
grains on the bed surface. Dynamic armor layers may disappear at high bed shear
stresses, but this is not necessarily the case (Gomez, 1995; Wilcock and DeTemple,
2005).
The dynamic armored bed condition at the downstream of Kurau River can be
over a mixed sand and gravel bed. It is described by the distribution size of bed
downstream of Kurau River. Therefore, because of this dynamic armored bed with
coarse sediment the fine sediment is relatively rare in the bedload than in the bed
material.
size to assess the relative mobility of various sizes of classes, shows that fractional
transport rates decreases with the decrease in sediment loading in the first part (i.e.,
left side of line A). The reason is that the fine fraction is present in the bedload
because of the overpass in the suspension at high flows, not at low flows. In the
second part (i.e., between lines A and B), the fractional transport rates are
approximately equal. Sediments are present in the load in proportions similar to those
present in the bed. In the third part (i.e., right side of line B) in the coarse material,
the fractional transport ratio decreases with the increase in particle size and moves in
The horizontal part of the data points (i.e., between lines A and B) indicates
that the transport of variously sized sediment particles approaches equal mobility.
106
The transported bedload is composed approximately of the same value of sediment
as bed material. The fractional transport ratio depends regularly on the proportion in
the bed in the Kurau River and the transport rate of fraction independent of the
particle size. In comparison, the range of fractions in equal mobility in the upstream
within 0.4<dx<4.5 was found to be greater than that in the downstream equal
mobility range (0.55< dx < 3). This finding demonstrates the stable condition of
sediment transport of the upstream compared with that in the downstream. The low
amount of Pi/fi in the downstream shows the sedimentation in the section caused by
(1959), Chang (Cheng, 2002), Julien (2002), Wong and Parker (2006) and van Rijn
(1993) were compared with observed values from Kurau River. The performances of
the equations were measured using the discrepancy ratio (DR), which is the ratio of
discrepancy ratio of 0.5-2.0 (0.5<DR <2.0) was used as a criterion in the evaluation
of the selected equations. Based on the relationship within measured and predicted
values the formulas in most cases performed disconcert and they over predicted and
under predicted of the measured values (Table 4.4). All equations produced an
average discrepancy ratio out of range 0.5-2. Figure 4.24 depicts comparisons of
bedload transport predictions and measurements from Kurau River study sites.
107
Table 4.4: Summary of bedload transport equations assessment
10
Rottner Meyer-Peter and Muller
Predicted Tb(kg/s)
1
Predicted Tb(kg/s)
0.1
0.1
0.01
0.01 0.001
0.01 0.1 1 10 0.01 0.1 1 10
Measured Tb(kg/s) Measured Tb(kg/s)
10
wong julien
1
Predicted Tb(kg/s)
1
Predicted Tb(kg/s)
0.1
0.1
0.01
0.001 0.01
0.01 0.1 1 10 0.01 0.1 1 10
Measured Tb(kg/s) Measured Tb(kg/s)
10 10
Chang Van Rijn
1 1
Predicted Tb(kg/s)
Predicted Tb(kg/s)
0.1 0.1
0.01 0.01
0.01 0.1 1 10 0.01 0.1 1 10
Measured Tb(kg/s) Measured Tb(kg/s)
Figure 4.24: Comparison of predicted and measured bedload rates for Kurau River
108
4.6.2 Prediction of Bedload Transport in Kurau River with Nonlinear
Regression Method
Blizard and Wohl (1998) reported the relationship between bedload transport
and hydraulic variables, thus, multiplication of the hydraulic variables in the form of
power law based on Equation (3-12) and (3-13) can better describe the behaviour of
the bedload transport rate (Tb). Figure 4.25 shows the bedload rating curve and fit of
this function at study sites. Power function of flow discharge, Shields’ parameter (θ),
median grain size (d50) and the channel gradient assumed as the variables of the
bedload transport rate function in the Kurau River sites. The average flow rate and
the sediment movement are strongly coupled in a highly non-linear manner (Wang et
al., 2011). Therefore new equation was extracted based on the relationship between
intensity bedload rate and hydraulic data where parameters α and in terms of
10
y = 0.144x1.1
R² = 0.88
1 KRU1
KRU2
0.1 KRU3
KRU4
0.01
KRU5
A1
0.001
0.001 0.01 0.1 1 10
Discharge Q (m3/s)
109
The nonlinear regression method was used to assess the relationship between
each independent variable and the bedload transport rate. The coefficients obtained
Partial R2 values were calculated for each variable included in the models. Based on
equation (3.7) several runs were performed with various initial settings and the
performance of the developed equation was analysed and calibrated for each run. The
best value estimated parameters are listed in Table 4.5. Table 4.6 shows the statistical
Based on the first accurate analysis the value of n was rounded to 0.2, h=1
where simplify the final approximation equation, the value i=1.00 applied to the
equation. Equation (4-1) was entered as a base for the next run of nonlinear
regression with the new adjusted invariables. Table 4.7 shows the briefly parameter
110
estimate for coefficient and Table 4.8 shows that this equation is significant with the
s Gs 1 .g.d503
R2 value of 0.948.
b m.q0.2 . .Dgr 0
(4-1)
The unit of parameters follow the SI unit. Therefore, the value of Tb was found
in the same scale of the other variables in SI unit (kg/s). The nonlinear equation was
The predictive abilities of the NLR equation (4-2) are assessed through
validation the model by the set of data of the Kurau River of present study and
previous study (DID, 2009). The assessment of Equation (4-2) is shown in Table 4.9
with acceptable average discrepancy ratio of 1 ,and Figure 4.26 shows the best fitting
111
Table 4.9: Assessment of NLR equation
Coefficient of Average
Data No Data determination, Discrepancy ratio
(R2 ) (0.5<DR <0.2)
Present study 48 0.98 0.85
DID 2009 20 0.82 1.16
Total 60 0.90 1.00
10
predicted Tb (kg/s)~[Eq 4-2]
0.1
Present study
DID 2009
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Programming
Multiple sets of training, testing, and validation data were randomly selected
and numerous runs were performed with various model setting such as number of
generation and genes and depth of trees by the trial and error. From 69 available data
50% were used for training (present study) and 25 % were used for testing and 25 %
112
(DID, 2009) for validation. Consequently, the models were selected according to
The best relationship was selected from the optimum R2, RMSE and MAE for
each training, test and validation to prevent from over fitting of the model by
selecting the high R2 of the training. The following relationship was selected to
where Tb is the bedload transport rate (kg/s), d50 median grain size (mm), S0 water
surface slope (m/m) and θ Shield's parameter. Figure 4.27 shows the expression of
113
Gene 1and base term -
* C
* C
Q S
Gene *
* C
+ Q
d50 θ
Gene
*
S C
Gene *
+ C
* *
Q C S C
versus predicted values of bedload rate for training, testing, and all data as shown in
Figure 4.28, Figure 4.29 and Figure 4.31, respectively. The values of R2, RMSE, and
MAE are equal to 0.96, 0.083 and 0.067, respectively, for training sets (Figure 4.28)
and 0.78, 0.159 and 0.099, respectively, for testing sets (Figure 4.29).
114
10
R²=0.96 RMSE=0.083
Predicted Tb (kg/s) 1
0.1
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.28: Measured versus predicted values of Tb for the training data set.
The predictive abilities of the GP equation (4-3) were assessed through modle
validation using Kurau River data sets of the present study and previous study (DID,
2009). The measured versus predicted values of bedload rate for validation set is
illustrated in Figure 4.31. The values of R2, RMSE, and MAE for this data set were
obtained equal to 0.89, 0.110, and 0.082, respectively. The R2 is equal to 0.90 while
RMSE and MAE are equal to 0.116 and 0.080 respectively for all data sets (Figure
4.31). In fact, the evolved model has achieved higher accuracy for both testing and
115
10
R²=0.78 RMSE=0.159
predicted Tb (kg/s) 1
0.1
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.29: Measured versus predicted values of Tbfor testing data set.
10
R²=0.89 RMSE=0.110
1
predicted Tb (kg/s)
0.1
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.30: Measured versus predicted values of Tb for validation data set.
116
10
R²=0.90 RMSE=0.116
0.1
Present study
DID 2009
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.31: Measured versus predicted values of Tb for all data set.
The combination of GP and ANN was suggested for best prediction result for
predicting the sediment transport (Singh et al. (2007). The combination of GP and
ANN was performed for the modelling of bedload transport rate in Kurau River.
First the bedload transport rate was calculated using GP Equation (4-3), and
then the outcome was given as input to the ANN, which consisted of one input node,
one output node and 10 hidden layers. Figure 4.32 shows the test result in the form
error measures are R2 = 0.92, RMSE = 0.11 kg/s. The results show the combination
117
of the input of four variables (Q, S, d50, θ) and one output Tb was trained and
validated. For this purpose, the data were shuffled and divided in two parts; one part
of them was used in the learning process by random, the other part was used for the
verification. Often this can be done in more than one way by changing the percentage
of data for training process and verification. Finally from 69 available data 50% were
used for training and 25 % were used for testing and validation. The number of
neurons in the hidden layer was determined by calibration using several computer
10
R²=0.92 RMSE=0.11
1
predicted Tb (kg/s)
0.1
Present study
DID 2009
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
The best fit of the obtained and given data for bedload transport rate is shown
in Figure 4.33 to Figure 4.36 where the number of neurons in the hidden layer is 15.
The correlation of determination, root mean square error and mean absolute error of
118
training, testing and validation of modelling shows in Table 4.10. The result shows
10
R²=0.90 RMSE=0.16
1
predicted Tb (kg/s)
0.1
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.33: Measured versus predicted values of Tb by ANN for training data set
The outcome of the ANN was calibrated with GP and the result shows an
expected success with improving the R2, and the indicating errors (R2= 0.94,
119
RMSE=0.1 and MAE= 0.075). The plot scatter of measured against the predicted
The combined ANN-GP model results thus appear to be more acceptable than
the single ANN or GP models. The combination shows that the ANN first carries out
solution easier and improve the accuracy of the single ANN and GP results.
10
R²=0.81 RMSE=0.16
Predicted Tb (kg/s)
0.1
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.34: Measured versus predicted values of Tb by ANN for testing data set
120
10
R²=0.90 RMSE=0.1
predicted Tb (kg/s) 1
0.1
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.35: Measured versus predicted values of Tb by ANN for validation data set
10
R²=0.86 RMSE=0.15
1
predicted Tb (kg/s)
0.1
Present study
DID 2009
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.36: Measured versus predicted values of Tb by ANN for total data set
121
10
R²=0.95 RMSE=0.10
predicted Tb (kg/s) 1
0.1
Present study
DID 2009
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Many measures for model evaluation have been documented in the literature of
coefficient (r or R2), index of agreement (d), root mean squared error RMSE, and so
on, were critically reviewed by (Legates and McCabe, 1999), and suggested that it is
inappropriate to use only correlation coefficient for model evaluation. The authors
‘goodness-of-fit’ or relative error measure like d and at least one absolute error
two conventional evaluation criteria, RMSE (root mean square error) and U
(inequality coefficient), are used in the present study to measure the performances of
122
RMSE provides a quantitative indication of the model absolute error in terms
of the units of the variable, with the characteristic that larger errors receive greater
attention than smaller ones. This characteristic can help eliminate approaches with
significant errors (Wu et al., 2008). The inequality coefficient (U) was used to
determine how accurate a bedload equation predicted the actual value of bedload
U
i 1 Tbp
rmse
n i 1 Tbo i
1 n 2 1 n 2
1 1
2 2
n i
(4-4)
n Tbo Tbp i
RSME i 1
1
2 2
n
(4-5)
MAE
Tboi Tbpi
n
i 1
n (4-6)
where Tbi is the measured bed load rate, Tbo is the predicted bedload rate, i
denotes a given flow, and n is the number of flows. The scaling of the denominator is
such that U always falls between 0 and 1. If U = 0, then Tbi= Tbo and there is a perfect
fit. If U = 1, then Tbo Tbp and the equation lacks a predictive value. For the purpose
of this study, the GP, NLR methods can represent the measured data when U is very
small and closed in 0. For the Meyer-Peter and Muller, Rottner, Wong, Chang, Julien
and vanRijn equations, U near to 1.This demonstrates that the predicted value does
123
The results in Table 4.11 show that the observed transport data are not very
well predicted by the existing bedload transport formulae. Furthermore, the observed
transport data are best fitted with GP, ANN based and nonlinear regression functions.
Figure 4.38 shows the comparison of the bedload rating curve with different
equation and methods. Figure 4.39 demonstrates observed versus predicted transport
rate from Kurau River study sites, and indicates that, predicted values by GP, GP-
ANN, ANN, ANN-GP and NLR methods are typically within an order of magnitude
of observed values. However the ANN-GP model shows a better performance with
0.95 as the correlation coefficient, but the function of NLR has the minimum errors
124
Present study(R² = 0.89)
NLR (R² = 0.98)
GP (R² = 0.90)
Rottner (R² = 0.76)
Calculated bedload Tb (kg/s) 10 MPM (R² = 0.35)
chang (R² = 0.22)
Julien (R² = 0.18)
Wong (R² = 0.19)
1 Van Rijn (R² =0.2)
0.1
0.01
0.01 0.1 1 10
Discharge Q ( mᵌ/s)
10
1
predicted Tb (kg/s)
GP
0.1
GP-ANN
ANN
ANN-GP
NLR
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.39: Comparisons of predicted and measured bedload rates for Kurau River
125
4.7 Development of Bedload Equation for Small Rivers (Kurau, Lui,
Semenyih)
Sediment transport in small streams is diverse and highly variable due to the
ways unique because it is influenced by its own particle history of flow conditions,
As shown in Figure 4.40, the bedload transport rate for these rivers has good
relation with discharge in power function. All these rivers can be represented by the
10
y = 0.17x1.31
Bedload transport rateTb(kg/s)
R² = 0.95
1
0.1
126
4.7.1 Assessment of Existing Equations for Small Rivers (Kurau, Lui and
Semenyih)
Rottner (1959), Chang (Cheng, 2002), Julien (2002), Wong and Parker (2006) and
vanRijn (1993) were compared with observed values with a log10 transformation
applied to all equations. A discrepancy ratio of 0.5-2.0 (0.5<DR <2.0) was used as a
criterion in the evaluation of the selected equations (Table 4.12). The results of the
study sites are shown in Figure 4.41. Based on the relationship between measured
and predicted values, the formula results were disconcerted, and the measured values
Table 4.12: Summary of bedload transport equations assessment for three rivers
127
10 1.000
Rottner Meyer-Peter and Muller
Predicted Tb(kg/s)
1
Predicted Tb (kg/s)
0.100
0.1
0.01 0.010
0.01 0.1 1 10 0.001 0.01 0.1 1 10
Measured Tb(kg/s) Measured Tb (kg/s)
wong julien
100
10
Predicted Tb (kg/s)
Predicted Tb (kg/s)
10
1
1
0.1
0.1
0.01 0.01
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
Measured Tb (kg/s) Measured Tb (kg/s)
10 10
Chang VanRijn
Predicted Tb(kg/s)
Predicted Tb (kg/s)
1 1
0.1 0.1
0.01 0.01
0.01 0.1 1 10 0.01 0.1 1 10
Measured Tb (kg/s) Measured Tb(kg/s)
Figure 4.41: Performance of existing bedload transport formula in Kurau, Lui and Semenyih
rivers.
128
4.7.2 Nonlinear Regression Result for Small Rivers (Kurau, Lui and
Semenyih)
Based on the section 4.6.2 and the relationship between the variables, the
following function as same as a function for Kurau River is suggested for data of
s Gs 1 .g.d503
small streams:
b m.qn . i .Dgr h 0
(4-7)
where the following hydraulic parameters were used in the regression analysis: S0,
water surface slope, θ, Shields parameter, q, stream discharge per unit width (water
discharge was calculated for each increment, and stream discharge was obtained by
Based on Equation (4-7), several runs were performed with various initial
settings for power and constant values, and the performance of the developed
equation was analysed for each run. The best values of estimated parameters are
Based on the first accurate analysis, the value of n was rounded to 0.1, h = 1,
where the final approximation equation was simplified by applying the value i = 1.00
129
to the equation. Equation (4-8) was entered as a base for the second run of NLR with
the new adjusted invariables. Table 4.14 shows the brief parameter estimate for the
coefficient. Table 4.15 shows that this prediction is significant with the R2 value of
s Gs 1 .g.d503
0.99.
b m.q0.1. .Dgr 0
(4-8)
The unit of the parameters follow the SI unit. Therefore, the value of Tb was
The nonlinear equation was derived from the analysis expressed in Equation
s Gs 1 .g.d503
(4-9).
130
The predictive abilities of the NLR, Equation (4-9) were assessed by the total
data gathered from the Kurau, Lui, and Semenyih rivers. Figure 4.42 shows the best
0.1
Figure 4.42: Measured versus predicted values of Tb for total data set modelled by NLR
It is well known that the bedload transport predictions are of higher accuracy
when more measured data are taken into consideration. The best configuration may
be determined by the trial and error method. For this purpose, the data were shuffled
and divided in two parts; one part of them was used in the learning process by
random, the other part was used for the verification. Also, the data from all rivers
were uniformly distributed among the training and test data sets. Often this can be
done in more than one way by changing the percentage of data for training process
and verification. Finally from 190 available data 60% were used for training and
131
40 % were used for testing and validation. The number of neurons in the hidden layer
was determined by calibration using several computer run tests on random data sets.
ANN model performance. The common procedure is the use of the coefficient of
determination R2, when evaluating the goodness of fit of models. The best fit of the
model obtained and given data for bedload transport rate is shown in Figure 4.43 to
Figure 4.46, where the number of neurons in the hidden layer is 15. Figure 4.46
shows the verification between the measured and estimated values for these new
patterns, which clearly show that the linear coefficient of correlation is very high
between the observed data and the values predicted through neural nets. The values
are 0.99 and 0.93 in training and testing. Overall, the linear coefficient of correlation
is 0.978.
10
R²=0.99 RMSE=0.047
Estimated bedload transport rate Tb(kg/s)
0.1
0.01
0.01 0.1 1 10
Measured bedload tranport rateTb(kg/s)
132
Figure 4.43: Measured versus predicted values of Tb by ANN for the training data set
10
R²=0.93 RMSE=0.114
0.1
0.01
0.01 0.1 1 10
Measured bedload tranport rateTb(kg/s)
Figure 4.44: Measured versus predicted values of Tb by ANN for testing data set
10
Estimated bedload transport rate Tb(kg/s)
0.1
R²=0.95 RMSE=0.118
0.01
0.01 0.1 1 10
Measured bedload tranport rateTb(kg/s)
133
Figure 4.45: Measured versus predicted values of Tb by ANN for validation data set
10
R²=0.97 RMSE=0.161
0.1
0.01
0.01 0.1 1 10
Measured bedload tranport rateTb(kg/s)
Figure 4.46: Measured versus predicted values of Tb by ANN with for total data set.
Table 4.16 shows the result of the sensitivity analysis for 5 parameters with
the reduced data set. In the process of the sensitivity analysis, the parameters were
excluded one by one from the list of input variables. Then, the parameter with the
least relative importance compared to all of the other parameters is extracted from
the model construction, based on the highest correlation coefficient. This procedure
is repeated for all parameters one by one. In the first stage, the least effective
obtained when P is excluded from the input list. Therefore, in the second stage, P
134
was omitted, and the analysis repeated for the rest of the variables. In Table 4.15, the
results also are given for different hidden nodes, and it can be seen for getting better
results, increasing the number of hidden nodes is not necessary. In summary, the
parameters can be listed from the most effective to the least effective as follows: Q,
d50, θ, S0 and P.
S, d50, θ 0.91 0.91 0.3 0.87 0.89 0.335 0.91 0.93 0.3
Q, S, θ 0.98 0.98 0.202 0.98 0.97 0.184 0.99 0.96 0.184
Q, S, d50 0.97 0.97 0.2 0.98 0.97 0.21 0.98 0.98 0.18
Q, d50, θ 0.98 0.98 0.196 0.99 0.98 0.176 0.99 0.98 0.174
Model with variables Q, d50, θ
Q, d50 0.98 0.98 0.205 0.98 0.98 0.194 0.98 0.97 0.194
Q, θ 0.98 0.97 0.206 0.99 0.97 0.185 0.97 0.97 0.194
d50, θ 0.88 0.94 0.3 0.92 0.87 0.31 0.93 0.89 0.29
Model with variables Q, d50
135
4.9 Genetic Programming Result
Multiple sets of training, testing, and validation data were randomly selected
and numerous runs were performed with various model setting such as number of
generation and genes and depth of trees by the trial and error. A GPTIPS run with the
4, Elitism = 0.01% of the population, function node set = (plus, minus, times,
protected). As same as Kurau River four input parameters as variable data, including
discharge (Q), water surface slope (s0), mean grain size (d50), and Shields parameter
for the initiation of motion (θ), as well as the bed load rate (Tb) as invariable data
The performance of the developed equation was analysed for each run.
Consequently, the best models were selected according to statistical criteria such
as R2, root mean square error (RMSE), and mean absolute error (MAE).
The best relationship was selected from the optimum R2 for each training, test,
and validation to prevent from over fitting of the model by selecting the high R2 of
the training. The following relationship was selected to model the bedload transport:
136
The precision of the developed equation is examined by plotting the measured
versus predicted values of bedload rate for training and testing. All of the data are
shown in Figure 4.47 to Figure 4.48. The values of R2, RMSE, and MAE are equal to
0.97, 0.079, and 0.063, respectively, for the training sets (Figure 4.47), and 0.90,
0.13, and 0.098, respectively, for the testing sets (Figure 4.48).
For all of the data sets, R2 = 0.93, RMSE = 0.11, and MAE = 0.085 (Figure
4.49). Figure 4.50 illustrates the measured versus predicted values of bedload rate
for the validation data set. The obtained values of R2, RMSE, and MAE for this data
set were equal to 0.92, 0.124, and 0.093, respectively. In fact, the evolved model
achieves high accuracy for both testing and validation sets in order to confirm that
10
R²=0.97 RMSE=0.08
Estimated bedload transport rate Tb (kg/s)
0.1
0.01
0.01 0.1 1 10
Measured bedload transport rateTb(kg/s
Figure 4.47: Measured versus predicted values of Tb for the training data set.
137
10
R²=0.90 RMSE=0.14
0.1
0.01
0.01 0.1 1 10
Measured bedload transport rate Tb(kg/s)
Figure 4.48: Measured versus predicted values of Tb for testing data set
10
R²=0.93 RMSE=0.11
Estimated bedload transport rate Tb (kg/s)
0.1
0.01
0.01 0.1 1 10
Measured bedload transport rateTb(kg/s
138
10
R²=0.92 RMSE=0.12
0.1
0.01
0.01 0.1 1 10
Measured bedload transport rateTb(kg/s)
The results in Table 4.17 show that the observed bedload transport data do not
formulas; otherwise, the observed bedload transport data best fitted by the GP and
ANN estimates and the NLR function result. The predicted bedload transport rates
from the GP method, ANN, and NLR were compared to the observed values.
Comparison of the bedload rating curve for different formula and methods is shown
in Figure 4.51.
139
Table 4.17: Bedload equations assessment
Figure 4.52 provides the observed versus predicted transport rate of the small
rivers, and indicates that the predicted values by GP, ANN, and NLR are typically
within an order of magnitude of observed values. The ANN model actually shows a
the GP and ANN approach to bedload transport modelling is the automatic ability of
the GP and ANN to select input variables that contribute beneficially to the model
and to ignore those that do not, and also the GP does not assume any a priori
functional form of the solution but in NLR, the model structure is specified in
advance (which is in general difficult to do) and the model coefficients are
determined.
140
Present Study (R=0.92)
GP (R² = 0.93)
NLR(R² = 0.88)
ANN(R² = 0.84)
Predicted Tb(kg/s)
10 Rottner (R² =0.83)
MPM (R² =0.48)
Chang (R² =0.18)
Julien(R² =0.14)
Wong (R² =0.13)
1 VanRijn(R² =0.06
0.1
0.01
0.01 0.1 1 10
Discharge Q( m3/s)
10
1
Predicted Tb(kg/s)
0.1
NLR
GP
ANN
0.01
0.01 0.1 1 10
Measured Tb(kg/s)
Figure 4.52: Comparisons of predicted and measured bedload rates for small streams by
different models
141
5 CHAPTER 5
RIVER CONFLUENCE SEDIMENT TRANSPORT MODELLING
5.1 Introduction
water supply and main sources for irrigating the paddy areas in some part of the state
of Perak. Human activity includes the recently railway construction, changes in land
use and river sand mining make changes to river morphology and perturbation in
river equilibrium. Changes in the sediment load affect the efficiency of reservoir as
most of the transported sediment in the river will be deposited in the reservoir and
this problem require the river management at the upstream of river such as
The different morphology and geometry in each section of the river provide the
The Kurau sub basin is consisting of two main river tributaries namely Kurau
River and Ara River. These two rivers join together at Pondok Tanjung. River
complex place due to rapid changes in sediment discharge, flow structure, and
142
Various alternative techniques have been developed to provide quantitative
predictions of the complexity of the flow movement and its interaction with its
different processes such as sediment transport, water quality, and water surface
profiles etc has been developed. These multi-dimensional programs may be two-
topography survey, bed roughness data and boundary conditions such as water level
and discharge.
In this part of the study the sediment flow in Ara and Kurau confluence was
5.2 SSIIM
option. The program is designed for use generally in research for hydraulic, river,
in a three-dimensional non-orthogonal grid, using the "k-ε" model for turbulence, the
control volume method with the SIMPLE algorithm and it solves the convection-
143
The advantage of using SS
SIIM, comp
pared to othher CFD proograms is th
hat it can
sizes and can handle wetting andd drying off cells resultting in a chhanging grid
d (Olsen,
unstructurred grid (ceells and nodes are nott arranged in rows annd columns (Figure
a unstructuured 3D gridd, each celll will have one index, it is not possible to
5.2)). In an
144
4
Figure 5.2: Unstructured grid
SSIIM 1 is easier to use, but cannot apply wetting and drying of cells (Olsen,
complex geometry and its algorithms for wetting and drying. For the simulations to
be done in this study, only SSIIM 2 will be used due to the complex geometry and
wetting and drying of cells. In the rest of this thesis, when the name SSIIM is used, it
program solves the Navier-Stokes equations with the k-epsilon model for velocity
volume method is used for the power-law scheme or the second order upwind
scheme together with the discretization. The SIMPLE method is used for the pressure
coupling. The velocity field in the geometry obtains with an implicit solver. The
convection-diffusion equations for different sediment sizes are solved by using the
velocities.
145
5.4.1 Water Flow Calculation
The Navier-Stokes equations for turbulent flow are solved to obtain the water
Ui
t
Ui
Ui 1
xi xi
Pij ui u j (5-1)
The left most term on the left side is transient term and the next term is
convective term. The first term on the right hand side is pressure term and the next
term on the right side of the equation is the Reynolds stress term. A turbulence model
is required for evaluating this term. The transient term is neglected in the default
algorithm in SSIIM. To include this term, time steps and a number of inner iterations
in the calculations, different data sets are used in the control file. For transient
calculations it is possible to give the water levels and discharges as input time series.
The eddy viscosity concept with the k-ε turbulence model is introduced with
U Ui 2
ui u j vT i
xi x j k ij
3 (5-2)
The first term on the right side of the equation forms the diffusive term in the
Navier- Stokes equation. The second term is often neglected, but can be included in
SSIIM 1 by adding some data set in the control file. The third term on the right side
vT c
2
k
(5-3)
146
Turbulent kinetic energy k, defined by:
k ui u j
1
2 (5-4)
k is modelled as
k k vT k
U j P
t x j x j k x j k (5-5)
U j U j Ui
Pk vT
xi xi x j (5-6)
vT 2
U j C 1 Pk C 2
t x j xi k x j k k (5-7)
In all above equations C's are different constants. The k- ε model is the default
The default wall law in SSIIM for rough as is given by Schlichting (1979)
defined as:
U 1 30 y
ln
u x k ks (5-8)
distance to the wall and the roughness, ks, is equivalent to a diameter of particles on
the bed.
147
5.4.2 Sediment Flow Calculation
suspended load can be calculated with the convection-diffusion equation for the
sediment concentration, c
c c c c
U j w
t xj z x j T x j (5-9)
where w is the fall velocity of the sediment particles and Г diffusion coefficient,
vT
Sc (5-10)
Where Sc is the Scmidth number, set to 1.0 as default in model, but different
specified in the control file, where the diameter and fall velocity is given. A vertical
then be used. The Rouse number (Whipple, 2004) commands the mode of sediment
transport. It is the ratio of particle settling velocity to the shear velocity (rate of fall
ws
Rouse number #= ; k = 0.4 (Von Karman’s constant)
ku*
2.5
ws
Bedload:
ku*
1.2 s 2.5
w
50% Suspended:
ku*
0.8 s 1.2
w
100% Suspended:
ku*
0.8
ws
Wash Load:
ku*
148
Van Rijn (1984) formula is used in SSIIM for calculating the suspended load
c
1.5
0.015 c
a s w g 0.1
0.3
d
Cbed
w
2
(5-11)
where d is the sediment particle diameter, a is a reference level set equal to the
roughness height, is the bed shear stress, c is the critical bed shear stress for
movement of sediment particles according to Shield’s curve, and s are water and
w
sediment density, is the viscosity of the water and g is the acceleration of gravity.
The empirical parameters in the equation (0.015, 1.5 and 0.3) may be changed by
The bed load, qb, also can be calculated by using van Rijn’s formula:
c
c
2.1
0.053
0.3 s w g
qb
s w g
D50
0.1
w
w
1.5
D50 2
(5-12)
discharges specified and simulations can be started for water flow or for water flow
with sediments. It is also possible to follow the simulations and to view the results
in the view option of the menu. The different views are Map graphics with contour
149
Editor. Some of the most important variables are velocity vectors, water level, bed
changes a sediment concentration. The results are shown as plots of the different
variables (Olsen, 2011). Figure 5.3 shows SSIIM's graphical interface. In this figure,
the interface is showing a map of Kurau- Ara influence, and the chosen variable is
bed level.
In general, a SSIM run starts by reading input files, or generating the grid using
the Grid Editor. After generation of the grid, the inflow and outflow should be
defined using the Discharge Editor. Then the data was saved in the Unstruc or
koordina files, before the computations are started and the results are viewed.
150
As an input for model four main things are needed as follow:
3. Sediment data
To run the program a file called Control is necessary. Control file controls all
parameters. Most of the parameters used in the simulations are in the Control file.
This file include physical parameters like water level, discharge and friction factors,
also the other parameters like time step, number of iterations and parameters that
decides what kind of formulae to be used. SSIIM have default values for most of the
parameters, so for simple situations, the program can be run without a complicated
Control file.
Transient calculations run with parameters in a file called Timei. This file
contains parameters which can vary over time such as water level, discharge, and
sediment concentrations. All inputs and outputs of the SSIIM model are given in SI
The Control file is made with data sets, all the data sets such as F, G, W, S, I
which can be used are explained in the SSIIM manual (Olsen, 2011).
results from the calculation. It shows parameters as average water velocity, water
151
depth and shear stress in the initialization. Trap efficiency and sediment grain size
distribution are also written in this file. If any error occurs during the run of program,
when the solution has converged, the results are written to 'result' file and 'bedres'
file. The result file stores the information about the water flow simulation. This
bedres file is written only after sediment simulation. It stores information about bed
roughness, grain size distribution, sediment thickness, and bedform height. SSIIM
can read the result and bedres files later to show all the graphical results from the
Figure 5.4 shows flowchart included the various files are used in each SSIIM
run. Most of the files are only used for special purposes and they are normally not
required. The program can also produce many of the input files. All the necessary
152
unstruc control
koordina result
SSIIM2
geodata timei
koomin timeo
boogie compres
interpol interres
process. The grid generation has done by using three softwares: Gridmeister, Matlab
The Gridmeister program usually applies to support the CFD and especially the
SSIIM user working in the field of hydraulic engineering with the structured grid
generation. The input data is the geometry data of Kurau and Ara confluence
boundary that represented by x-and y- coordinates. The geometry data was saved by
"DXF" format in a CAD program and used in gridmeister. The outputs are koordina
and a control file. The control file includes the correct grid information and koordina
file includes i and j that are the cells number in X and Y direction, X and Y the
153
Figure 5.5: Koordina file
software for making a 3D unstructured grid. The koordina file and the field geometry
cross section were used in MATLAB for making a 3D surface grid by using the
TriscatterdInterp function. The format of output file was change for using in SSIIM
that named koosurf to generate the 3D mesh (Figure 5.6). Figure 5.7 shows a koosurf
154
Ara
Kurau
155
5.8.1 Grid Editor
When the koosurf file is present, the grid for the xy- plane of Kurau and Ara
confluence can be viewed in the graphical interface of SSIIM by choosing the add
block from koosurf in blocks menu. The program generates the grid in the vertical
menu. The grid can either be multiblock or the simpler version with only one block.
With the most recent wetting/drying algorithm, it may be more suitable to make a
which are glued together. For the simulations of sediment flow in Ara and Kurau
river confluence, experiments have been made with both the two blocks and the one
For making grid with 2 blocks, one block was added by choosing the Add
block from koosurf in the Block menu. The next step was to make the grid 3D, this is
done by choosing Generate 3D grid in the interface. After having the grid the content
has written to the Unstruc file. This is done in the File option of the main menu. The
next step was reading Unstruc file by the new SSIIM window then the second block
was added from the Block menu. Then the blocks were glued together. The water
surface was first covered with blocks, and then the boxes were connected. In the end
there will be an unstructured grid covering the entire water body. The program then
generates the grid in the vertical direction according to the bed levels given in the
koosurf file. A three-dimensional multiblock grid for the given water body has then
been generated.
156
Block 2
Block 1
Ara River
Kurau River
70.0 m
Level 2
Block 2
Block 1
70.0 m
Level 2
157
Outblocked
Ara River
Outblocked
Kurau River
70.0 m
Level 1
editor. There can be several groups of inflows and outflows in the grid, but for
making continuity, total inflow discharges and total outflow discharge should be
equal to each other. The information about the grid, including the discharges was
stored in the Unstruc file by choosing the write unstruc from File menu.
158
For the simulations of sediment flow in river confluence two type grids (two
blocks and one block) was tested for equal situations and the conclusion was that the
simulation gave better results and converged faster for the grid with only one block.
A disadvantage of using two block grids was that the time needed for the simulation
The sum of all the water inflow and outflow in the geometry is shown in
Boogie file that start with the word "Cont:". This should be a very low value,
typically under 10-7. This value for simulation with 2 blocks increased by time but
with one block the value is acceptable. The solutions were diverged with increasing
Table 5.1 shows some Cont value for one and 2 blocks simulation. Due to this,
only the one-block grid has been used for the simulations in this study.
Table 5.1: Comparison of Cont value for one and two block grid
159
During the grid generation, some considerations have been taken to ensure a
well functioning grid that will be given stable calculations. The grid cells are almost
false diffusion, the grid lines are aligned with the direction of the flow, especially
close to inflow and outflow areas. The distortion ratio (the dimension of a grid cell in
one direction divided by the dimension of the cell in the other direction) is not too
big. The size of a grid cell does not differ too much from the size of the neighbouring
For this purpose the grid sensitivity has done by choosing a different size of
the grid cell. The best results have achieved by the grid size of 0.75×0.75 m for each
cell. The grid for Ara-Kurau river confluence has about 64138 cells at the start of the
calculations. The grid has up to 11 cells in the vertical direction depending on the
depth of the specific location in the river confluence. The number of cells may
decrease during calculations due to wetting and drying algorithm. If the water level
goes down, or if the bed level goes up due to sedimentation, there might be a
decrease of cells in the vertical direction. As cells dry up there will also be a decrease
computation with sediments, moving surface, and moving bed. It also has to include
the wetting and drying of cells, as the flow is changed and many cells will dry up due
to sedimentation. The simulation carried out at Ara and Kurau river confluence that
160
is located between 691915.7559 and 691874.4946 North Latitude and 554178.7400
approximately 141.5 m in length and 111.5 m in width. The coordinate was changed
with deducting 691800 from latitude and 554000 from longitude coordinate for easy
using in software.
The field site for this study is the confluence of the Kurau and Ara rivers in
Pondok Tanjung at the upstream of the Bukit Merah reservoir in Perak (Figure 3.2).
The two confluent channels have different widths and different bed height. The Ara
width is around 28 m and Kurau around 23 meters and the bed of Ara is
approximately 0.45 m higher than the bed of the Kurau channel and goes through the
to 26 m in width further downstream. The confluence is a sand bed junction and this
sandy bed making the most of the likelihood of active sediment movement and
avalanche slopes, a central scour, and a bar formed below the downstream junction
corner (Figure 5.11). The deepest zone within the two confluent channels is an
extension of the Kurau channel thalweg in to the confluence. The finest bed
sediments (d50 = 0.5 mm) are found the left side of the Ara mouth while the coarsest
bed sediments (d50 = 1.5 mm) are located on downstream from the region of
maximum scour. The two confluent channels have same grain size distributions and
161
Kurau
Ara
Lateral bar
Figure 5.10: View of the confluence of the Kurau and Ara rivers
16.4
260 16.2
240 16
Lateral bar
Ara 15.8
220
15.6
200
Latitude N
15.4
Scour zone
180
15.2
160
15
140 14.8
120 14.6
Kurau
14.4
80 100 120 140 160 180 200 220 Elevation (m)
Longitude E
162
5.9.2 Input Data
SSIIM needs input data for sediment sizes, sediment fall velocities, and
sediment concentrations. For the simulation, eight sediment sizes of bedload samples
in Kurau and Ara River were used. Figure 5.12 to Figure 5.14 shows the distribution
grain size of bedload sample in Kurau and Ara River. The fall velocities (Vanoni,
The concentrations were calculated from the percentages of each sediment size
for the given water discharge and its sediment load. The calculation made and
concentrations results are given as cubic meters sediments per cubic meters water.
100
90
80
Percentage Passing (%)
70
60
50
40
30
20
10
D50=1 mm
0
0.01 0.1 1 10 100
Particle size (mm)
163
100
90
80
100
90
80
Percentage Passing (%)
70
60
50
40
30
20
10 D50=1.8 mm
0
0.01 0.1 1 10 100
Particle size (mm)
164
Table 5.2: Sediment characteristics
Fall velocity
No dx Size (mm)
(ωs) (m/s)
1 d90 3.67 0.29
2 d85 2.84 0.25
3 d75 2.22 0.21
4 d65 1.37 0.18
5 d50 1 0.14
6 d25 0.7 0.085
7 d15 0.47 0.068
8 d10 0.38 0.055
The most important input files were prepared for the sediment flow simulation
is Control and Timei file. The control file was made after several tests concluded in
what algorithms would give a good and stable solution. The timei file was prepared
by the chosen values for discharges, water levels and sediment concentrations.
In addition to these files, the Unstruc file included the grid information has
been used in the simulations. The information about cells outside of the grid has been
stored in a file called koordina. This file is also used in case new cells become wet.
Some parts of the data set in the control file are explained in this section. For
more details on the data sets, see the SSIIM manual (Olsen, 2011). The simulation
uses van Rijn's formula to calculate the concentrations at the bed. This is given in the
F 10 data set. The F 6 data set gives the coefficients for this formula. This data set
has been used to calibrate the model to give a total bed change as close to the
165
measured amount as possible. The roughness in the rivers were measured and used as
input for the simulations. The value is set to 0.063 metres in the F 16 data set. In the
F 33 data set the time step of the simulation is set to 30 seconds, with 10 inner
iterations per time step. This simulation is a transient sediment computation with free
water surface, specified on the F 36 and F 37 data sets. Since wetting and drying may
happen for this simulation, an algorithm that changes the shape of the grid cells close
to the boundaries is necessary, this is given in the F 102 data set. Algorithms that
help to stabilise triangle cells are also included in data sets F 113 and F 235. The
chosen sediment sizes and their fall velocities are given in the S data sets. The N data
set gives the percentage size fraction of the bed sediments which is taken from
G 13 data set was used for outblocking option in this study. Two parts of Ara
and Kurau confluence were blocked out for using 1 block to ensure getting better
results (Figure 5.9). The sediment flow simulation uses a 30 second time step. To
simulate 3 days, 86400 iterations are necessary. This is given in the K 1 data set in
the control file. The control file was used in this study is shown in Figure 5.15.
The timei file used in this simulation gives the concentrations of sediment
loading in Kurau and Ara River. The timei file is shown in Figure 5.16. The data
given in the file are upstream water levels, and downstream water levels, and
different water discharge. The file specifies the concentrations of the 8 sediment
groups for both Kurau and Ara rivers given in the control file.
166
Figure 5.15: Control file used in SSIIM modelling
167
Figure 5.16: Time File
168
5.9.4 Numerical Algorithms
flow of Kurau and Ara river confluence simulation. The combination of algorithms
in the control files leads to a successful simulation. The simulation may give
different result or may lead to crashing by choosing other algorithms. The algorithms
Data set F 36 7 was used for the computation of the vertical elevation of the
water surface. The data set reads one integer. If the integer is 7, as it is in this
simulation, the water surface is updated based on the pressure in only its
neighbouring cells.
Data set F 64 was used for the grid generation to generate the grid lines in the
11that is the most tested options for sediment transport computations in rivers. The
algorithm gives a body fitted grid with priority to close to the bed. While most of
sediments are transported close to the bed, the hexahedral cells will give better
The F 102 1 algorithm is also employed for the sediment flow simulation. This
algorithm is used to change the shape of the grid cells close to the boundary for the
The F 113 data set was implemented to stabilize the solution in the shallow
areas close to the side walls. The algorithm used in these simulations is the F 113 4.
169
The algorithm uses second-order interpolation instead of third-order interpolations
The F 222 data set invokes algorithms which prevents the downstream bed
The F 233 data set invokes an algorithm that, instead of using the pressure in
the surface cells to compute the water level, uses a depth-averaged pressure field.
The F 235 data set was used to improve the stability in triangular cells. F 235
10 which is used in this case, is the most successful of these algorithms. This option
The F 244 data set was implemented to reduce instabilities in triangular cells.
Two relaxation factors used in the algorithms, the first floating point is used for the
velocities in the cells, in the F235 10 algorithms. The second integer is used for the
predicting velocity, water level and bed elevation for the confluence. A sensitivity
analysis was conducted to provide parameter estimation guidance for the calibration.
Large numbers of parameters were tested, from grid size, turbulence models,
discretization schemes, values for critical bed shear stress parameters, parameters in
170
bed form and roughness prediction formulas, and empirical coefficients in the
sediment transport capacity formula etc. It is not possible to go into further detail due
analysis that was used to calibrate the model is shown in Table 5.4.
SSIIM model first is tested with sensitivity analysis and calibrated using field
data from one time and then validated with a different set of data in different time.
until a reasonable match was found between model predictions and field data.
Validation for the model was carried out by comparing measured, water level and
bed level with different discharges to the prediction average velocity, water level and
Calibration of SSIIM was performed using the flow and bed elevation data in
April 2012. In order to calibrate flow several roughness coefficient and relaxation
factor for the tree velocity equations, the pressure correction equation and k and e
equation are adjusted to gain the correct observe velocity and water level, meanwhile
response need to be adjusted to fit with the observed sediment data. Coefficients in
van Rijn's formula for bedload and suspended sediment transport were adjusted to
gain a good fit with the observed sediment transport. Also several algorithms as
mentioned in section 5.9.4 were used in a SSIIM program for obtaining the correct
171
The vanRijn's bedload transport parameter was calibrated for different
discharge flow and bedload transport rate with the value of calculating bedload
transport from genetic programming derived equation for Kurau River in section 4.9.
The estimate of bedload transport rate with GP formula and van Rijn formula are
The parameters of the van Rijn formula were determined for calibration
follows:
c
c
1.115
420
g
qb
s w g
D500.5 s w2
0.1
w
w
1.5
D50
(5-13)
100.000
Eq. 4.11
Van Rijn formula
Bed load transport rate Tb (kg/s)
10.000
1.000
0.100
0.1 1 10 100
Discharge Q(m³/s)
172
Table 5.3: Comparison of Bedload transport rate
The parameters after calibration are summarized in Table 5.4. The simulated
flow average velocity, water level and bed elevation were compared with observed
data and coefficient of determination was calculated for measured and simulated
water level and bed level as shown in Figure 5.18 to Figure 5.30.
173
Table 5.4: Parameter calibrated in SSIIM
No Description
2 Schumidt's coefficient 1
174
1.2
Measured
1 Simulated
0.8
Velocity (m/s)
0.6
0.4
0.2
0
0 5 10 15 20 25 30
Distance (m)
0.7
0.6 Measured
Simulated
0.5
Velocity (m/s)
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25
Distance (m)
175
0.8 (a)
0.6
Depth (m)
0.4
0.2
0
0 5 100 1
15 20 25 30
Distancce (m)
(b)
Figure 5.20:
5 Compaarison cross-ssectional bed
d level and avverage velocity a) simulaated b)
Measuured, April 2012 at Ara River
R
176
6
1.5 (aa)
Depth (m)
1
0.5
0
0 5 10 1
15 20
Distance (m)
(
(b)
Figure 5.21:
5 Compaarison cross-ssectional bed
d level and avverage velocity a) simulaated b)
Measurred, April 2012 at Kurau River
177
7
Ara
A'
Legend
A
16.466812
16.133136
15.799461
15.465785
15.132109
14.798434
14.464758
Kurau
90.0 m
Ara
RSK7
RSK6
RSK5
RSK4
RSK3
RSK2 RSA
RSK1
Legend
16.460000
16.130000
15.800000
15.450000
15.130000 Kurau
14.800000
14.450000
90.0 m
178
16.5 16.5
RSA RSK1
16.0 16.0
Elevation(m)
Elevation(m)
15.5 15.5
15.0 15.0
Measured BL Measured BL
Simulated BL Simulated BL
14.5 14.5
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5 17.0
RSK2 RSK3
16.5
16.0
Elevation(m)
Elevation(m)
16.0
15.5
15.5
15.0
15.0
14.5 14.5
Measured BL Measured BL
Simulated BL Simulated BL
14.0 14.0
0 10Distance20(m) 30 40 0 10 20 30
Distance (m)
16.5 16.5
RSK4 RSK5
16.0 16.0
Elevation(m)
Elevation(m)
15.5 15.5
15.0 15.0
14.5 14.5
Measured BL Measured BL
Simulated BL Simulated BL
14.0 14.0
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5
16.5
RSK6 RSK7
16.0
16.0
Elevation(m)
Elevation(m)
15.5
15.5
15.0 15.0
Figure 5.24: Comparison cross sectional bed level in different condition of Ara and Kurau
confluence (Measured BL, April 2012)
179
16.0
15.8
15.6
15.4
Elevation(m)
15.2
15.0
14.8
14.6
14.4
14.2 Measured BL
Simulated BL
14.0
0 10 20 30 40 50 60 70
Distance (m)
Figure 5.25: Comparison of measured and simulated Longitudinal bed level at downstream
of confluence (AA') (Measured BL, April 2012)
16.0
15.5 R² = 0.98
Simulated bed level(m)
15.0
14.5
14.0
14.0 14.5 15.0 15.5 16.0
Measured bed level (m)
Figure 5.26: Scatter plot of measured bed level against simulated bed level (April 2012)
180
16.44
16.42
16.40
16.38
16.36
Elevation(m)
16.34
16.32
16.30
16.28
16.26
16.24 Measured wl
16.22 Simulated wl
16.20
0 10 20 30 40 50 60 70
Distance (m)
16.5
R² = 0.97
Simulated water level (m)
16.4
16.3
16.2
16.2 16.3 16.4 16.5
Measured water level (m)
Figure 5.28: Scatter plot of measured water level against simulated water level (April 2012)
181
Legend
16.580581
16.528214
16.475847
16.423481
16.371114
16.318747
16.266380
80.0 m
Legend
16.563013
16.515460
16.467906
16.420353
16.372799
16.325246
16.277692
80.0 m
182
5.9.6.2 Model Validation
Sediment transport process was validated for the Ara- Kurau confluence. The
simulation SSIIM was carried using surveyed cross section and flow discharge
measurement for three events: low flow, moderate flow and high flow at the
confluence. As part of the validation, measured bed level profiles, water levels
during July to October were compared to the prediction bed level profiles and water
levels by SSIIM model (Table 5.5 to Table 5.7 and Figure 5.31to Figure 5.33).
Table 5.5: Comparisons of water and bed level for Q=15 m3/s (19 July 2012)
Table 5.6: Comparisons of water and bed level for Q=43 m3/s (27 Sept 2012)
183
18.00
16.00
15.00
14.00
Water level Observed water level
Bed level Observed bed level
13.00
0 10 20 30 40 50 60 70
Distance (m)
Figure 5.31: Comparisons of water and bed level (AA') for Q=15 m3/s (19 July 2012)
18.00
17.00
Elevation (m)
16.00
15.00
14.00
Water level Observed water level
Bed level Observed bed level
13.00
0 10 20 30 40 50 60 70
Distance (m)
Figure 5.32: Comparisons of water and bed level (AA') for Q=15 m3/s (20 July 2012)
184
Table 5.7: Comparisons of water and bed level for Q=11 m3/s (8 Oct 2012)
18.00
17.00
Elevation (m)
16.00
15.00
14.00
Water level Observed water level
Bed level Observed bed level
13.00
0 10 20 30 40 50 60 70
Distance (m)
Figure 5.33: Comparisons of water and bed level (AA') for Q=11 m3/s (8 Oct 2012)
185
5.9.7 Short Term Changes in Bedload Transport, Bed Morphology and Bed
Material Characteristics
The morphology and sedimentology of sand bed river channel confluences are
Short term variation in bed morphology and spatial patterns of bed material
was documented in detail over a period of 7 days. During this short event water level
varied from 1.7 m to 2.8 m of bankfull depth, and the momentum ratio Mr =
The simulation was started with the low flows with Mr < 1 during the first 2
days and then followed a high flow event when the confluence was characterized by
a high momentum ratio. For analysing the bed change and characteristic of sediment
patterns, result of flow and bedload transport from six of the flow stage with Mr>1
change during an event in Kurau and Ara confluence at the end of September and
first week of October data collection. The morphology changes will be shown in 8
cross sections, two at Ara and Kurau mouth (RSA, RSK1) and the other six cross
sections (RSK2, RSK3, RSK4, RSK5, RSK6, RSK7) along the main Kurau at the
186
Table 5.8: Hydraulic condition during an event at Kurau _Ara confluence
187
5.9.7.1 Morphological Changes
The confluence morphology during the first step (Q=9 to Q=15 m3/s) was
drastically modified. The bed morphology at Q=15 m3/s is shown in Figure 5.37,
cross sectionals change are shown in Figure 5.39 and longitudinal change in Ara and
Kurau are shown in Figure 5.35 and Figure 5.36. The valance face for both of Kurau
and Ara is illustrated in the longitudinal profile of confluence. Low flow condition
(Mr<1), privileged the expansion of the Ara mouth bar over 4 meters and extended
by ~ 2m into the confluence and constrained the principal flow at the confluence in
the middle of the channel, as indicated by expansion of the deepest zone of the bed
(Figure 5.38). The angel of the avalanche face changed from 50 o to 30o at the end of
the Ara mouth bar. During this period bed erosion occurred on the right hand of
main Kurau in lateral bar and along the channels in downstream of confluence
(RSK3, RSK4). This erosion was followed by a deposition phase at the downstream
of the lateral bar in the secondary bar (RSK5, RSK6, RSK7) and at the left hand of
confluence downstream.
188
16.2
16.0
15.8
15.6
15.4
Elevation(m)
15.2
15.0
14.8
14.6
14.4
Measured BL April 2012
14.2
Q= 15 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.35: Longitudinal bed change profile of Ara and downstream of confluence after
Q=15m3/s
15.8
15.6
15.4
15.2
Elevation(m)
15.0
14.8
14.6
14.4
Measured BL April 2012
14.2
Q= 15 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.36: Longitudinal bed change profile of Kurau and downstream of confluence after
Q=15m3/s
189
Ara
Legend
16
6.777551
16
6.364626
15
5.951700
15
5.538775
15
5.125850
14
4.712925
14
4.360000
Kurau
u
80.0 m
Ara
Kurau
Figure 5.38: Change in bed morpphology afterr Q=15m3/s. Zone of erossion and deposition
during eaach period arre illustrated with colour change from
m white as deeposition to black
b as
erosion.
190
0
16.5 16.5
RSA RSK1
16 16.0
Elevation(m)
Elevation(m)
15.5 15.5
15 15.0
Measured Measured
Q= 15 (mᵌ/s) Q= 15 (mᵌ/s)
14.5 14.5
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5 16.5
RSK2 RSK3
16 16
Elevation(m)
Elevation(m)
15.5 15.5
15 15
16.5 16.5
RSK4 RSK5
16 16
Elevation(m)
Elevation(m)
15.5 15.5
15 15
16.5 16.5
RSK6 RSK7
16 16
Elevation(m)
Elevation(m)
15.5 15.5
15 15
Figure 5.39: Channel cross section profiles, Q=15m3/s (Measured bed level April 2012)
191
The erosion in downstream of confluence continued with increasing flow and
changing the momentum ration from Mr<1 to Mr>1. The bed morphology after this
event is shown in Figure 5.40. During this period deposition occurred in the Ara
mouth bar and at the right hand of the post confluence entrance (Figure 5.42 and
Figure 4.41). The cross sectional changes in bed morphology during this period are
shown in Figure 5.41 and the plan of erosion and deposition is shown in Figure 5.44.
In this event the lateral bar and also the sediment deposition in the downstream of
main bar were scoured in the inner bank (RSK5, RSK6 and RSK7). Increasing the
discharge was not great enough to obviously change the shape of the downstream
channel bed of confluence (RSK2 and RSK3). The channel morphology changes and
sediment transport in this period largely reflected sustained next high flow
conditions.
Ara
Legend
16.590823
16.233427
15.876031
15.518635
15.161239
14.803843
14.446447 Kurau
80.0 m
192
16.5 16.5
RSA RSK1
16 16.0
Elevation (m)
Elevation (m)
15.5 15.5
15 15.0
Q=15 (mᵌ /s) Q=15 (mᵌ /s)
Q= 31 (mᵌ/s) Q= 31 (mᵌ/s)
14.5 14.5
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5 16.5
RSK3
RSK2 16
16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
16.5
16.5
RSK4 RSK5
16
16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
16.5 16.5
RSK6 RSK7
16 16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
193
16.2
16.0
15.8
15.6
15.4
Elevation(m)
15.2
15.0
14.8
14.6
14.4 Measured
Q= 15 (mᵌ/s)
14.2
Q= 31 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.42: Longitudinal bed change profile of Ara and downstream of confluence between
Q=15m3/s and Q=31m3/s (Measured bed level April 2012)
15.8
15.6
15.4
15.2
Elevation(m)
15.0
14.8
14.6
14.4
Measured
14.2 Q= 15 (mᵌ/s)
Q= 31 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.43: Longitudinal bed change profile of Kurau and downstream of confluence
between Q=15m3/s and Q=31m3/s (Measured bed level April 2012)
194
Ara
Kurau
Figure 5.44: Change in bed morphology between Q=15m3/s and Q=31m3/s. Zone of erosion
and deposition during each period are illustrated with colour change from white as
deposition to black as erosion.
High flows occurred in next step and discharge flow changed from 31 to 43
m3/s, and the water depth changed from 2.35 to 2.8 m. At this larger discharge ratio
(Mr=2. 6), the influence of the Ara River into the confluence and the curvature of
flows through tributary increase. This, aided by larger flow separation, generates
around the confluence rather than through the centre of the confluence. The
morphological effect of these changes is larger bed scour, larger bar formed within
the separation zone (Figure 5.45 and Figure 5.46) and the retreat of the tributary
195
Ara
1.7299 m/s
10.0 m
Level 11
Kurau
Ara
Kurau
196
During this high flow event the Ara mouth bar was eroded over and the scour
zone at the entrance the Kurau River increased in the direction of the Ara side. This
led to the retraction of the front of the Ara mouth bar and the development of a
steeper avalanche face. The steepest part of the face was located close to the edge of
the shear layer and appeared to be maintained by the presence of the shear layer
along the side of the Ara mouth bar and the lateral bar (Boyer et al., 2006). The bed
Ara
Legend
17.391625
16.864688
16.337750
15.810813
15.283875
14.756938
14.280000
Kurau
80.0 m
197
This high momentum flow was directed toward the main channel side of
downstream channel after the Ara mouth. The deepest part of the main channel
extended further downstream through the confluence and near to the bars due to
increase in bed shear stress over distance and high bed shear stresses in the centre of
the confluence both of which advance scour zone (Rhoads and Sukhodolov, 2008).
The cross sectional and longitudinal changes during this flood event is shown in
Figure 5.48 , Figure 5.49 and Figure 5.50. The central portion of the confluence was
scoured and cross sections profile reshaped through erosion of the inner channels
(RSK2, RSK3) and the scour hole align itself in the direction of Ara River (RSK4).
Excavation of bed material decreased bed elevation by 0.5 and shifted the face of this
part of the cross section toward the outer bank of the downstream confluence by 1
Over this period, a bar complex developed along the inner bank that include
the lateral bar and a secondary bar created at the downstream end of the lateral bar
due to deposition of suspended and eroded sediments in a separation zone in the lee
of the lateral bar (Figure 5.51). Deposition along the front of the protruding lateral
bar in downstream of confluence increased the bed elevation by about 0.5 m within
198
16.6 17.0
RSA RSK1
16.5
16.1
Elevation (m)
Elevation (m)
16.0
15.6
15.5
15.1
15.0
Q= 31 (mᵌ/s) Q= 31 (mᵌ/s)
Q= 43 (mᵌ/s) Q= 43 (mᵌ/s)
14.6 14.5
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5 17
RSK2 RSK3
16.5
16
Elevation (m)
Elevation (m)
16
15.5
15.5
15
15
14.5 Q= 31 (mᵌ/s) 14.5 Q= 31 (mᵌ/s)
Q= 43 (mᵌ/s) Q= 43 (mᵌ/s)
14 14
0 10 20 30 40 0 10 20 30
Distance (m) Distance (m)
17 17
RSK4 RSK5
16.5 16.5
Elevation (m)
Elevation (m)
16 16
15.5 15.5
15 15
Q= 31 (mᵌ/s) 14.5 Q= 31 (mᵌ/s)
14.5
Q= 43 (mᵌ/s) Q= 43 (mᵌ/s)
14 14
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
17.5 16.5
17 RSK6 RSK7
16
16.5
Elevation (m)
Elevation (m)
16 15.5
15.5
15
15
Q= 31 (mᵌ/s) 14.5 Q= 31 (mᵌ/s)
14.5
Q= 43 (mᵌ/s) Q= 43 (mᵌ/s)
14 14
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
199
16.2
16.0
15.8
15.6
15.4
Elevation(m)
15.2
15.0
14.8
14.6
14.4 Measured
Q= 31 (mᵌ/s)
14.2
Q= 43 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.49: Longitudinal bed change profile of Ara and downstream of confluence between
Q=31m3/s and Q=43m3/s (Measured bed level April 2012)
15.8
15.6
15.4
15.2
Elevation(m)
15.0
14.8
14.6
14.4 Measured
Q= 31 (mᵌ/s)
14.2
Q= 43 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.50: Longitudinal bed change profile of Kurau and downstream of confluence
between Q=31m3/s and Q=43m3/s (Measured bed level April 2012)
200
Ara
Kurau
Figure 5.51: Change in bed morphology between Q=31m3/s and Q=43m3/s. Zone of erosion
and deposition during each period are illustrated with colour change from white as
deposition to black as erosion.
From this temporal change in high flow and discharge momentum ratio Mr>1,
it was possible to demarcate the location on the Ara side were more responsive and
active in its morphological change and it responded to the migration of the shear
layer within the confluence as flow stage changes from Mr<1 to Mr>1.
The bed morphology of confluence after Q=35 m3/s is shown in Figure 5.54.
Erosive event during high flow was followed by a deposition phase as the discharge
decreased to 35 m3/s and water level decreased from 2.8 to 2.5 m (Figure 5.52 and
Figure 5.53). The bed elevation was increased at the entrance of Kurau (RSK1) by
~0.4 m on the right hand and erosion occurred along the outer of lateral bar and
201
deposition was concentrated along the inner bank of RSK4, RSK5 and RSK6 and in
the downstream of main Kurau , RSK7 (Figure 5.56). The scour hole migrated to the
centre of the channel. Bed change morphology which included the deposition of a
bar in the inner bank and weak scour at the outer bank during this period is shown in
Figure 5.55.
16.0
15.8
15.6
15.4
Elevation(m)
15.2
15.0
14.8
14.6
14.4 Measured
Q= 43 (mᵌ/s)
14.2
Q= 35 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.52: Longitudinal bed change profile of Ara and downstream of confluence between
Q=43m3/s and Q=35m3/s (Measured bed level April2012)
15.8
15.6
15.4
15.2
Elevation(m)
15.0
14.8
14.6
14.4 Measured
Q= 43 (mᵌ/s)
14.2
Q= 35 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.53: Longitudinal bed change profile of Kurau and downstream of confluence
between Q=43m3/s and Q=35m3/s (Measured bed level April2012)
202
Ara
Legend
17.500000
16.500000
16.000000
15.500000
15.000000
14.500000
14.310000
Kurau
80.0 m
Ara
Kurau
Figure 5.55: Change in bed morphology between Q=43m3/s and Q=35m3/s. Zone of erosion
and deposition during each period are illustrated with colour change from white as
deposition to black as erosion.
203
16.6 16.6
RSA RSK1
16.1 16.1
Elevation(m)
Elevation(m)
15.6 15.6
15.1 15.1
Q= 43 (mᵌ/s) Q= 43 (mᵌ/s)
Q= 35 (mᵌ/s) Q= 35 (mᵌ/s)
14.6 14.6
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5 17
RSK2 RSK3
16.5
16
Elevation(m)
Elevation(m)
16
15.5
15.5
15
15
14.5 Q= 43 (mᵌ/s) 14.5 Q= 43 (mᵌ/s)
Q= 35 (mᵌ/s) Q= 35 (mᵌ/s)
14 14
0 10 20 30 40 0 10 20 30
Distance (m) Distance (m)
17 17
RSK4 RSK5
16.5 16.5
Elevation(m)
Elevation(m)
16 16
15.5 15.5
15 15
17.5 17
17
RSK6 RSK7
16.5
16.5
Elevation(m)
Elevation(m)
16
16
15.5
15.5
15
15
Q= 43 (mᵌ/s) 14.5 Q= 43 (mᵌ/s)
14.5
Q= 35 (mᵌ/s) Q= 35 (mᵌ/s)
14 14
0 10 20 30 0 10 20 30 40
Distance (m) Distance (m)
204
Next step the total discharge was decreased (Q=13m3/s) and discharge ratio
dropped below 1, the Kurau flow discharge was the dominant flow. The bed
morphology after change of flow momentum is shown in Figure 5.57. The low
confluence (Figure 5.58). Comparisons of channel cross sections confirm that bed
morphology is similar to Q= 35 and Q=13 m3/s but that some minor changes can be
identified (Figure 5.59 and Figure 5.60). These minor changes and low discharge
ratio flow eroded the Kurau mouth at the cross section RSK1 and lateral and
secondary bars along the inner bank (RSK6 and RSK7). The deposition occurred in
left hand of the Ara mouth (RSA) and outer bank of the lateral bar in cross sections
Ara
Legend
16.615858
16.241548
15.867239
15.492929
15.118619
14.744310
14.360000
Kurau
80.0 m
205
Ara
1.6078 m/s
10.0 m
Level 11
Kurau
16.0
15.8
15.6
15.4
Elevation(m)
15.2
15.0
14.8
14.6
14.4
Measured
Q= 35 (mᵌ/s)
14.2
Q= 13 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.59: Longitudinal bed change profile of Ara and downstream of confluence between
Q=35m3/s and Q=13m3/s (Measured bed level April 2012)
206
15.8
15.6
15.4
15.2
Elevation(m)
15.0
14.8
14.6
14.4 Measured
Q= 35 (mᵌ/s)
14.2
Q= 13 (mᵌ/s)
14.0
0 20 40 60 80 100 120 140 160
Distance (m)
Figure 5.60: Longitudinal bed change profile of Kurau and downstream of confluence
between Q=35m3/s and Q=13m3/s (Measured bed level April 2012)
Ara
Kurau
Figure 5.61: Change in bed morphology between Q=35m3/s and Q=13m3/s. Zone of erosion
and deposition during each period are illustrated with colour change from white as
deposition to black as erosion.
207
16.6 16.5
RSA RSK1
16.1 16.0
Elevation (m)
Elevation (m)
15.6 15.5
15.1 15.0
Q= 35 (mᵌ/s) Q= 35 (mᵌ/s)
Q= 13 (mᵌ/s) Q= 13 (mᵌ/s)
14.6 14.5
0 10 20 30 0 5 10 15 20 25
Distance (m) Distance (m)
16.5 16.5
RSK2 RSK3
16 16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
16.5
16.5
RSK4 RSK5
16
16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
17 16.5
16.5 RSK6 RSK7
16
Elevation (m)
Elevation (m)
16
15.5
15.5
15 15
Q= 35 (mᵌ/s) Q= 35 (mᵌ/s)
14.5 14.5
Q= 13 (mᵌ/s) Q= 13 (mᵌ/s)
14 14
0 10 20 30 0 10 20 30 40
Distance (m) Distance (m)
208
5.9.7.2 Lateral bar
is that formed just below the downstream junction corner. The origin of this bar is
linked to the formation of a large zone of separated flow (Best and Reid, 1984). Flow
separation occurs at the downstream junction corner where fluid of the Ara channel
cannot remain attached to the channel wall. This creates a zone of low velocity,
pathway and is carried into this zone. Because of the low flow velocities present
reveals that this bar is composed of relatively fine grained sediment, another
indication of the low velocities within the region. The separation zone bar dips into
the central scour but grades into the general bed elevation downstream where the
effects of the flow separation zone diminish beyond the point at which the combined
flow reattaches itself to the stream bank. The size of this bar is therefore related to
the size of the separation zone which grows both at higher confluence angles and
higher discharge ratios (Best and Reid, 1984; McGuirk and Rodi, 1978). Erosion of
the far bank may cause channel widening opposite this bar because of the
constriction of the effective channel width through which the combined discharges
must flow (Best and Reid, 1984). Figure 5.63 and Figure 5.64 show the change of
lateral bar during the event longitudinally and cross sectional. The Figures indicate
that the morphological change in lateral bar completely depend on flow momentum.
209
16.6
16.5
16.3
16.2
16.1
16.0
15.9
0 5 10 15 20 25 30 35
Latral bar distance (m)
Q=15(mᵌ/s) Q=31(mᵌ/s) Q=43(mᵌ/s) Q=35(mᵌ/s)
16.5
RSK4
16.0
Elevation (m)
15.5
Q=12 (m3/s)
Q=15 (mᵌ /s)
15.0 Q= 31 (mᵌ/s)
Q= 43 (mᵌ/s)
Q= 35 (mᵌ/s)
14.5
Q= 25 (mᵌ/s)
Q= 13 (mᵌ/s)
14.0
0 5 10 15 20 25 30
Distance (m)
210
5.9.7.3 Bedload Transport Rates
The Bedload transport rating curve of Ara and Kurau at upstream of confluence
is obtained with SSIIM is shown in Figure 5.65. The figures depict the good
determination (R2= 0.98) of bedload transport rate by SSIIM shows the ability of
10
Ara
Bedload transport rate Tb (kg/s)
Eq 4.11
SSIIM
0.1
1 10 100
Discharge (m³/s)
10
Kurau
Bedload transport rate Tb (kg/s)
Eq 4.11
SSIIM
0.1
1 10 100
Discharge (m³/s)
Figure 5.65: Bed load transport rating curve in Ara and Kurau River branch
211
10
Kurau
Ara
0.1
0.1 1 10
Bedload transport rate (Eq.4.11)Tb (kg/s)
Figure 5.66: Bed load transport rate value by SSIIM against the calculated bedload transport
rate with Eq. 4.11
The relationship between sediment loads and the discharges in the main
The simulation of flow sediment in the confluence of Ara and Kurau shows
that for each flow discharge, bed load transport rates varied throughout the
confluence, laterally and longitudinally, and also from one event to the other. In the
first step, bed load transport rates were found to increase at the entrance of the
confluence (RSK2, RSK3) and then decrease in the downstream direction at cross
Bed load transport rates were higher in the upstream portion of the confluence
than in the Kurau and Ara channels, corresponding with erosion within the
212
confluencee. For this flow dischaarge and wiith the mom
mentum ratiio less than
n one the
Araa
Bed load
l transportt rate
Leg
gend 0.4 kg s-1m-11
16.777
7551 Flow directiion
16.364
4626
15.951700
15.538
8775 Shear layerr
15.125
5850
14.712
2925
14.360
0000
Kurrau
8
80.0 m
Bed
d levels, min=
m 14.360 m, max=
= 16.778 m
Figure 5.67:
5 Bed moorphology annd spatial disstribution of bedload
b trannsport rate Mr=0.9.
M
213
3
some trannsport of seediment from
m Kurau. At
A upstream
m maximum
m bedload transport
t
Araa
0.4 kg s-11m-1
Flow dirrection
Legend
16.777551
16.364626 Shear laayer
15.951700
15.538775
15.125850
14.712925
14.300000
Ku
urau
80.0 m
Bed leve
els, min= 14.300
1 m, max=
m 16.778 m
Figure 5.68:
5 Bed moorphology annd spatial disstribution of bedload
b trannsport rate Mr=1.3.
M
is shown in Figure 5.69. Thiis figure illlustrates thhe high beedload tran
nsport in
transport rate
r in Kurrau River iss very low, and may be
b explainedd by the fact that a
bundled.
214
4
Thesse relativelly high bedd load tran
nsport rates are continnuing, conssiderably
transport rates
r at eacch cross sections are the
t Ara sidde along the confluencce. The
distributioon of bed load transporrt rates in crross sections shows thaat the highesst values
at the coonfluence (ccross sectioons RSK2 and RSK33) are gennerally nearr to the
boundaries of the shhear layer, and these patterns arre changedd depending
g on the
momentum
m ratio. Thhe forms off bedload transport
t allso define tthat deposittion and
Ara
0.4 kg s-1m-1
Leg
gend Flow directioon
17.3916
625
16.8646
688 Shear layer
16.3377
750
15.8108
813
15.2838
875
14.7569
938
14.2800
000 Kurau
K
8
80.0 m
Bed
d levels, min
n= 14.280 m,
m max= 17..392 m
Figure 5.69
5 : Bed moorphology annd spatial disstribution of bedload trannsport rate Mr=2.6.
M
215
5
Nexxt event the flow dischaarge come down
d and also
a the dom
minant flow
w is from
the Kurauu River witth flow moomentum raatio less thaan one andd the shear layer is
transport in
i both Araa and Kurauu Rivers hap
ppened in the middle oof channelss (RSK1,
RSK4), higher
h sedim
ment transpport rates occurred
o onn the Ara side whilee further
Ara
Bed load
l transportt rate
0.025 kg s-1m-1
Lege
end Flow directiion
16.61585
58
16.23821
15
15.86057
72 Shear layerr
15.48292
29
15.10528
86
14.72764
43
14.35000
00
Kuraau
80
0.0 m
Figure 5.70:
5 Bed moorphology annd spatial disstribution of bedload
b trannsport rate Mr=0.7.
M
216
6
The shear layer that develops along the interface of two merging flows is a
characterized by increased turbulence levels at the junction of the flows coming from
the main channel and the tributary (Rhoads and Sukhodolov, 2008).
position of shear layer varies according to the discharge ratio and the height of the
bed discordance. As shown in Figure 5.67 to Figure 5.70 this position is critical for
the dynamics of the confluence as it influences flow mixing and sediment transport
pathways thus affecting the resulting bed morphology (Boyer et al., 2006).
either side of the mixing interface at the entrance to the downstream channel (Figure
5.71), which sweeps sediment laterally away from the centre of the confluence,
contributing to scour and separation of sediment loads. The locus of the zone of high
bed shear stress near the centre of the channel leads to degradation of accumulated
sediment over the inner part of the downstream channel (edge the shear layer).
217
Shear layer Shear layer
Figure 5.71: Shear layer and distinct vortices about vertical axes at RSK1
Ara
Kurau
218
Figure 5.73 shows the variation of bedload transport rate in the cross sections
at the downstream of confluence with changing the discharge and flow momentum.
This figure depicted the increasing of bedload capacity through the confluence zone
between RSK2 and RSK4. The increases in bedload transport capacity from the Ara
interactions (Rhoads and Sukhodolov, 2001). The existing of the lateral bar at the
inner bank of the downstream channel caused a reduction in the local flow depth, an
acceleration of the near-bed flow, and outward deflection of this flow by topographic
steering. The bed discordance between Ara and Kurau gave rise to a two-layer flow
cores of high velocity with increased bedload transport capacity. The coincidence of
the shear layer that was generated the considerable turbulence indicated that the
11
Q=15 m³/s Mr=0.9
10
Q=31 m³/s Mr=1.3
9 Q=43 m³/s Mr=2.6
Bedload transport rate Tb (kg/s)
219
The spatial distribution of bed load transport rates consequently seems to be
varied with the changes in bed morphology. For low flow condition, the extent of the
bed load transport pathway on the Kurau side responds to the migration of the Ara
mouth bar and the lateral bar into the confluence (Figure 5.67), this migration in turn
being controlled by the change in the position of the shear layer. As the shear layer
invades on the lateral bar and secondary bar (high momentum ratio), it causes their
regular erosion and may explain the high bed load transport rates occurred between
RSK3 and RSK6 along the Ara side (Figure 5.68). These forms are interrelated with
the boundaries of the shear layer and respond to the movement of the shear layer as
Patterns of sediment distribution within the confluence respond not only to the
flow dynamics but also control the bed morphology: this in turn affects both the flow
The bedload transport between the Ara and the Kurau River as mentioned in
previous section mainly occurs near the downstream junction corner of the
confluence due to the formation of depression in the upstream junction corner that
provided by the Ara to the Kurau channel is mainly transported by the near-bed flow
originating from the Kurau branch channel. This near-bed flow has an element that is
focused on the slope of the bar. The interaction between the near-bed flow and the
220
downward slope gravitational impinges on the sediment particles conditions the
slope of the bar and causes sediment sorting. Moreover, this near-bed flow prevents
deposition of fine sediment at the inner-bank and the formation of a zone of flow
The spatial distribution of bed material was evaluated for tree flow conditions
as Mr <1 with total discharge Q=15m3/s, at high flow Mr>1 and Q=43 m3/s and low
At the apex of the junction when the momentum flux ratio is less than 1
(Q=15m3/s), the distribution of bed material reveals separate zones of fine and coarse
sediment within the confluence (Figure 5.74). A zone of fine gravel from the Ara
extends laterally across the downstream confluence channel from the RSK2 and the
inner bank of the lateral bar into the base of a scour hole (RSK3). Figure 5.74
illustrates the lateral displacement of fine gravel by the position of limit of the 3 mm
curve moving from the Ara right bank toward the centre of the confluence. This
movement of coarse sediment seems to be related to the position of the mixing layer
which is controlled by the momentum flow and sediment ratio. The sediment near
the inner bank of the downstream channel consists mainly of sand (d50=1.5 mm) that
extends along the top of the bar. A track of coarse sand extends along the front of the
lateral bar near the outer bank. This track of sediment, presumably comprised of
material from the Kurau River, located within the downstream channel where it is
replaced by the fine gravel-dominated sediment from the Ara River. The median
grain size in upstream junction as a flow stagnation zone where the two incoming
flows diverge towards the outer banks is fine sand (d50=1.5 mm) and it could
221
explained by the low bed shear stresses in the stagnation zone (Best, 1988) (Figure
5.75).
Ara
Legend
0.003505
0.003012
0.002519
0.002025
0.001531
0.001038
0.000545
Kurau
80.0 m
Bed grain size distribution, D50, layer 2, min= 0.54 mm, max= 3.51 mm
Figure 5.74: Distribution of bed median size, D50 Q=15 m3/s, Mr<1
222
Ara
Legend
2.756875
2.326643
1.896410
1.466178
1.035946
0.605714
0.175482
Kurau
80.0 m
The high flow in the next step cussed re adjustment of the grain size pattern to
reflect the dominant flow of the Ara River. At high flow with Mr>1 the Ara flow
penetrates slightly into the main channel, allowing the Ara sediments to be deposited
on the Kurau side (Figure 5.76). As the momentum flow ratio increased, finer
particles from the Kurau are found on the Ara bed at the upstream junction.the
median grain size increased from 3 mm to 3.5 mm at the lateral moving from the Ara
to the left bank of downstream confluence included the maximum depth zone (Figure
5.76) .
223
The increase in the median grain size appears to be the result of the increasing
bed shear stress as flow in Ara and Kurau rises (Figure 5.77). For high momentum
ratio the coarser sediments are located near to the left bank while in low flow ratio
the coarser sediment is located in centre of the channel. A sudden transition from fine
gravel to coarse sand occurred along the inner bank, between the outer bank and the
top of the lateral bar. Mix gravel and sand cover the channel bed near the outer bank
of the lateral bar (cross section RSK4). The area along the inner bank of secondary
bar at the right hand of downstream main Kurau is underlain by fine and coarse sand
material.
Ara
Coarse
Fine
Legend
0.003506
0.003012
0.002519
0.002025
0.001532
0.001038
0.000545
Kurau
80.0 m
Bed grain size distribution, D50, layer 2, min= 0.54 mm, max= 3.51 mm
Figure 5.76: Distribution of bed median size at high flow, D50 Q=43 m3/s, Mr>1
224
Ara
Legend
3.100000
2.636577
2.173153
1.709730
1.246306
0.782883
0.319459
Kurau
80.0 m
<1 redistributed the surficial bed material through the confluence. The pattern of the
bed sediment differs from the high flow condition in that input of sediment from
Kurau and Ara River remain segregated well downstream of confluence (Figure
5.78). During this period a wedge of sediment advanced to the confluence of the
Kurau River. Medium to coarse sand cover the outer bank along the lateral bar and
fine gravel is shifted toward the inner bank of lateral bar. The coarse material is
confined to the mouth of Ara and a narrow track in outer bank of lateral bar and
continued to farther downstream. Fine sediment also is present along the inner bank
225
of secondary bar and the inner channel and extended far down stream. This
movement of fine sediment from Kurau and coarse sediment from Ara in low flow is
because of high bed shear stresses in the centre of the confluence and an increase in
bed shear stress over distance (Figure 5.79) due to acceleration of flow along the
Ara
Coarse
Fine
Legend
0.003505
0.003012
0.002519
0.002025
0.001531
0.001038
0.000545
Kurau
80.0 m
Bed grain size distribution, D50, layer 2, min= 0.54 mm, max= 3.51 mm
Figure 5.78: Distribution of bed median size at low flow, D50 Q=13 m3/s, Mr<1
226
For the three discharge flow, the grain size distribution on the bar at the
downstream junction corner was considerably constant compared to the other part of
confluence such as scour hole and upstream of the junction. The particle size of
lateral bar was usually finer than the average median grain size of the post
confluence channel. However, the particle size in the upstream part of the bar is more
affected by the changes in flow conditions than the downstream end where the
median diameters not varied during the period. During high discharge-ratio the flow
curvature from the Ara into the downstream channel produces two effects that
influence bed material patterns. First, the magnitude of bed shear stress along the
inner bank appears to decrease rapidly leading to deposition along the inner bank and
a downstream fining of bed material on the bar surface. On the other hand, the
magnitude of bed shear stress increases rapidly near the outer bank along the bar
edge, downstream coarsening of the bed material. Second, curvature of the flow from
the Ara leads to the development of helical motion of flow within the downstream
channel that is characterized by outward near-surface flow and inward near-bed flow
During low discharge-ratio events, the main flow of the Kurau River causes
penetration of a wedge of fine sediment in the downstream channel along the outer
bank. Fine material from the Kurau River and coarse material from the Ara are
confined within the confluence to opposing sides of the scour hole before combining
cells, which separate the sediment loads of the two incoming streams, into a single
dominant helical cell that extends over most of the downstream channel, which leads
227
to connect bedload transport from the Kurau River toward the inner bank of the
Ara
Legend
2.000000
1.695345
1.390691
1.086036
0.781382
0.476727
0.172072
Kurau
80.0 m
228
5.9.8 High Flow Modelling
The 2007 flood is the largest flood for Kurau River since 1960, where this
flood event is slightly lower than the 100 ARI. Therefore, the high discharge at
191.32 m3/s of the event occurred on 23 October 2007 was considered as the design
peak discharge. Figure 5.80 shows the hydrograph for the October 2007. The
morphology variation and bedload transport analysis was carried out in this flood
event. The grid was created with105437 cells and discharge specified for Ara and
Kurau rivers.
250
200
191.322 m3/s
Discharge (m³/s)
150
100
50
32.708
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Time (day)
229
Figure 5.81: The morphology of Kurau-Ara confluence before flood
Figure 5.82 depicts peak water surface and changes of the channel geometry
due to erosion and deposition by the simulated changes in the channel bed profile.
The original channel geometry was based on survey in April 2012. The result shows
the erosion of the bed occurred at all cross sections along the downstream of the
confluence (Figure 5.83) and flood level was higher at the downstream compare to
230
Ara
Legend
20.256214
19.226845
17.500000
16.500000
15.500000
14.500000
14.080000
Kurau
90.0 m
231
Ara
Kurau
Figure 5.83: Change in bed morphology after Q=191.32m3/s. Zone of erosion and deposition
during each period are illustrated with colour change from white as deposition to black as
erosion.
20.0
19.5
19.0
18.5
18.0
17.5
Elevation(m)
17.0
16.5
16.0
15.5
15.0
14.5
Befor flood
14.0
After flood
13.5 Peak W.s.
13.0
0 10 20 30 40 50 60 70
Distance (m)
232
Figure 5.85 shows the cross section changes for the location along the
confluence of Kurau and Ara rivers. In general erosion has occurred at most inner
bank of cross sections after flood 2007, while the high flow favoured the progression
of the Ara mouth bar at the left side of Ara River. The high junction angle between
Ara and Kurau rivers made deep scour hole and segregation of bedload from each
Bed load transport during this flood event as expected increased but not
surprising in amount and may be explained by the fact that a large proportion of the
bed sediment is transported in suspension rather than the bedload (Figure 5.86). The
distribution of bed load transport in the Kurau river mouth is more than Ara mouth
and the bed load transport rate slightly decreased further downstream particularly
along the edge of the shear layer (RSK5, RSK6). The location of maximum bed load
transport rates of each cross section is varied along the both sides of confluence.
233
19 20.0
RSA RSK1
18 19.0
Elevation(m)
Elevation(m)
18.0
17
17.0
16
16.0
15 15.0
Befor flood Befor flood
After flood After flood
14 14.0
0 10 20 30 40 0 10 20 30 40 50
Distance (m) Distance (m)
20 20
RSK2 19 RSK3
19
18 18
Elevation(m)
Elevation(m)
17 17
16 16
15 15
14 Befor flood 14 Befor flood
After flood After flood
13 13
0 10 20 30 40 0 10 20 30 40
Distance (m) Distance (m)
20 20
19 RSK4 19 RSK5
18 18
Elevation(m)
Elevation(m)
17 17
16 16
15 15
14 Befor flood 14 Befor flood
After flood After flood
13 13
0 10 20 30 40 0 10 20 30 40
Distance (m) Distance (m)
20 20
19 RSK6 19 RSK7
18 18
Elevation(m)
Elevation(m)
17 17
16 16
15 15
Befor flood Befor flood
14 14
After flood After flood
13 13
0 10 20 30 40 0 10 20 30 40
Distance (m) Distance (m)
Figure 5.85: Modelled cross section changes before and after flood 2007
234
Ara
Legend
d B load transsport rate
Bed
20.25621
14 1 kg s-1m-1
19.22684
45 Flow diirection
17.50000
00
16.50000
00 Shear llayer
15.50000
00
14.50000
00
14.08000
00
Ku
urau
90.0
0m
Bed lev
vels, min= 14.080
0 m, max=
= 20.256 m
Figuure 5.86: Bed morphologgy and spatiaal distributionn of bedloadd transport ratte
32m3/s)
(Q=191.3
235
5
6 CHAPTER 6
CONCLUSIONS AND RECOMMENDATIONS
6.1 Conclusion
The bedload transport in the Kurau River is low; movement of large sediments
(i.e., granules and pebbles) of bed material are rare, and it occurs at a relatively high
discharge. At a low discharge, sand was transported over the bed. By increasing
flow, the pattern of the mobile sediment changed, and distribution became more
discharge between the upstream and downstream of the Kurau River indicates that
the amount of sediment particles of each fraction size in the upstream is greater than
that in the downstream in the same fractions. This finding demonstrates the size
The equal mobility of the bedload and bed material in the Kurau River is
achieved at moderate flows in streams, even when the size distribution of the bedload
is finer than the size distribution of the bed material at high flow. The frequency of
the discharge shows that the mean size distribution of the bedload is similar to that of
236
6.1.2 Estimating Bedload Transport
Hydraulic and sediment data from Kurau River in Malaysia were used to
as Meyer-Peter and Muller, Wong and Parker, as well as Chang, Julien and van Rijn
were evaluated. No consistent relationship was observed between the predicted and
observed bedload at the sites. Moreover, based on the relationship between the
bedload transport (Tb) and discharge (Q) and the Shields parameter (θ), the power
function of the hydraulic variable best described the observed bedload transport at
the small streams. This power function was subsequently developed into a predictive
transport equation. The NLR, ANN and GP methods were used to predict bedload
transport at first for Kurau River and then were updated by feeding networks by
more data from the other small streams for obtaining simple equation.
were obtained using the iterative estimation algorithm procedure. The RSME (0.069)
observed and predicted bedload transport rates for Kurau River. The Eq 4.2 was
updated by adding the Semenyih and Lui river data. The bedload transport predicted
by Equation 4.9 shows a good result by RMSE equal to 0.00 kg/s and U= 0.00.
The results show that the ANN and GP model with four input nodes Q, S, θ,
and d50 can accurately not surprisingly estimate the bedload transport rate. The
observed bedload transport rate. The developed equations for small streams by GP
237
(Eq 4.11) and ANN after updating the GP and ANN by feeding the networks with the
Lui and Semeniyh data also show reasonable performance under field conditions.
The predicted bedload transport was compared with observed values, and the
minimum RMSE and inequality ratio (U) were used to select the best performing
model. In this case, the ANN and GP models performed better than the NLR-based
model and other equations. It should be considered that specific condition and
complex behaviour of small streams affect the bedload transport rate. So, the
correct the equation coefficients. In general, unlike the other transport equations that
were tested, the equation derived using the GP model mostly predicted the bedload
streams and had the lowest RMSE. However, it is not much simpler than the NLR
equation.
very complex and include various associations between flow structure, bed
morphology and sediment transport which will change over differing temporal and
spatial scales.
238
morphology, sediment pattern and bedload transport within and at the downstream of
the confluence.
SSIIM2 was calibrated and validated for average velocity, water surface and
bed elevation profile, bedload transport at upstream of confluence with using the
calibrated vanRijn equation for several times. Good agreement was obtained for
bedload transport rate, and bed level and water profiles between the measured data
The study has demonstrated that the short term hydrologic variability can
low flow with momentum ratio, Mr, less than 1, the shear layer is in the middle of
the confluence, the input sediment loads are separated around the scour hole before
mixing, the bed load mostly travels through the confluence near the edge of shear
layer in the left side of the post confluence channel (Kurau side), the Ara mouth bar
is created and lateral bar is eroded, the sediment deposited on the downstream of the
lateral bar (secondary bar). A zone of fine gravel from the Ara extends laterally
related to the position of the mixing layer which is controlled by the momentum flow
For flow condition when momentum ratio is greater than 1, the shear layer is
near the Kurau channel side, the bedload transport follow the flow coming from Ara
in to the confluence, moving from the Ara side toward the downstream of the
confluence. The erosion occurs in the Kurau mouth and right side of Ara at the
239
entrance of confluence and inner bank of lateral bar. Sediment deposition occurs at
For the different flow condition, the grain size distribution on the bar at the
confluence such as scour hole and upstream of the junction. The particle size of
lateral bar is usually finer than the average median grain size of the Kurau channel.
However, the particle size in the upstream part of the bar is more affected by the
changes in flow conditions than the downstream end where the median diameters not
SSIIM2 has been used to simulate the river channel confluence for flood event
with 100 ARI. Bed load transport during flood event as expected increased but not
surprisingly in amount and may be explained by the fact that a large proportion of the
bed sediment is transported in suspension rather than the bedload. The channel bed
degradation had occurred at most cross sections and deposition had occurred at the
6.2 Recommendations
confluence.
transport at confluence.
240
7 REFERENCES
Ab. Ghani, A., Azamathulla, H. M., Chang, C., Zakaria, N., and Hasan, Z. (2011)
Prediction of total bed material load for rivers in Malaysia: A case study of
Langat, Muda and Kurau Rivers. Environmental Fluid Mechanics 11, 307-
318.
Abrahams, A. D., and Gao, P. (2006) A bed-load transport model for rough turbulent
open-channel flows on plane beds. Earth Surface Processes and Landforms
31, 910-928.
Ackers, P., and White, W. R. (1973) Sediment transport: new approach and analysis.
Journal of the Hydraulics Division, ASCE 99, 2041-2060.
Almedeij, J. H., and Diplas, P. (2003) Bedload transport in gravel-bed streams with
unimodal sediment. Journal of Hydraulic Engineering 129, 896-904.
241
Ashmore, P., and Gardner, J. T. (2008). Unconfined Confluences in Braided Rivers.
In "River Confluences, Tributaries and the Fluvial Network"John Wiley &
Sons, Ltd, pp. 119-147.
Ashmore, P., and Parker, G. (1983) Confluence scour in coarse braided streams.
Water Resources Research 19, 392-402.
Ashworth, P. J., Ferguson, R. I., Ashmore, P. E., Paola, C., Powell, D. M., and
Prestegaards, K. L. (1992) Measurements in a Braided River chute and lobe:
2. Sorting of bed load during entrainment, transport, and deposition. Water
Resources Research 28, 1887-1896.
Azamathulla, H., Ghani, A., Leow, C., Chang, C., and Zakaria, N. (2011) Gene-
Expression programming for the development of a stage-discharge curve of
the Pahang River. Water Resources Management, 1-16.
Azamathulla, H. M., Ab. Ghani, A., Zakaria, N. A., Chang, C. K., and Abu Hasan, Z.
(2010a) Genetic programming approach to predict sediment concentration for
Malaysian rivers. International Journal of Ecological Economics &
Statistics 16, 53-64.
Azamathulla, H. M., Chang, C. K., Ab. Ghani, A., Ariffin, J., Zakaria, N. A., and
Abu Hasan, Z. (2009) An ANFIS-based approach for predicting the bed load
for moderately sized rivers. Journal of Hydro-environment Research 3, 35-44.
Azamathulla, H. M., Ghani, A. A., Chang, C. K., Hasan, Z. A., and Zakaria, N. A.
(2010b) Machine Learning Approach to Predict Sediment Load – A Case
Study. CLEAN – Soil, Air, Water 38, 969-976.
Babovic, V., and Abbott, M. B. (1997) The evolution of equations from hydraulic
data, Part II: Applications. Journal of Hydraulics Research 35, 411–430.
Babovic, V., and Keijzer, M. (2003) Rainfall Runoff Modelling Based on Genetic
Programming Nordic Hydrology 34, 331–346.
242
Bagnold, R. A. (1980) An Empirical Correlation of Bedload Transport Rates in
Flumes and Natural Rivers. Proceedings of the Royal Society of London. A.
Mathematical and Physical Sciences 372, 453-473.
Banzhaf, W., Nordin, P., Keller, R., and Francone, F. (1998). "Genetic programming
- An introduction," Morgan Kaufmann Publishers, Inc.
Baranya, S., and Masa, J. (2007) Numerical and laboratory investigation of the
hydrodynamic complexity of a river confluence. Periodica Polytechnica,
Civil Engineering 51, 3-8.
Barry, J. J., Buffington, J. M., Goodwin, P., King, J. G., and Emmett, W. W. (2008)
Performance of bed-load transport equations relative to geomorphic
significance: predicting effective discharge and Its transport rate. Journal of
Hydraulic Engineering 134, 601-615.
Barry, J. J., Buffington, J. M., and King, J. G. (2004) A general power equation for
predicting bed load transport rates in gravel bed rivers. Water Resources
Research 40, W10401(22p).
Best, J., and Reid, I. (1984) Separation Zone at Open Channel Junctions. Journal of
Hydraulic Engineering 110, 1588-1594.
243
Bethea, R. M., Duran, B. S., and Boullion, T. L. (1995). "Statistical methods for
engineers and scientists," New York, Marcel Dekker, Inc.
Bhattacharya, B., Price, R., and Solomatine, D. (2007) Machine learning approach to
modeling sediment transport. Journal of Hydraulic Engineering 133, 440-450.
Biron, P., Roy, A., Best, J. L., and Boyer, C. J. (1993) Bed morphology and
sedimentology at the confluence of unequal depth channels. Geomorphology
8, 115-129.
Biron, P. M., and Lane, S. N. (2008). Modelling hydraulics and sediment transport at
river confluences. In "River Confluences, Tributaries and the Fluvial
Network"John Wiley & Sons, Ltd, pp. 17-43.
Blizard, C. R., and Wohl, E. E. (1998) Relationships between hydraulic variables and
bedload transport in a subalpine channel, Colorado Rocky Mountains, U.S.A.
Geomorphology 22, 359-371.
Boyer, C., Roy, A. G., and Best, J. L. (2006) Dynamics of a river channel confluence
with discordant beds: Flow turbulence, bed load sediment transport, and bed
morphology. Journal of Geophysical Research 111, F04007.
Bradbrook, K., Biron, P., Lane, S., Richards, K., and Roy, A. (1998) Investigation of
controls on secondary circulation in a simple confluence geometry using a
three dimensional numerical model. Hydrological Processes 12, 1371-1396.
244
Bradbrook, K., Lane, S., and Richards, K. (2000) Numerical simulation of three-
dimensional, time-averaged flow structure at river channel confluences.
Water Resources Research 36, 2731-2746.
Bradbrook, K., Lane, S., Richards, K., Biron, P., and Roy, A. (2001) Role of bed
discordance at asymmetrical river confluences. Journal of Hydraulic
Engineering 127, 351-368.
Bunte, K. (1990). Experiences and results from using a big-frame bed load sampler
for coarse material bed load. In "Hydrology in Mountainous regions" (H.
Lang and A.Musy, eds.). IAHS publ.193, pp. 223-230.
Bunte, K. (1996). "Analysis of the temporal variation of coarse bed load transport
and its grain size distribution (squaw Creek, Montana, USA)." RM-GTR-
228.US Department of Agriculture, Forest Service, Rocky mountain Fortest
and Rang Experiment Station, Fort Collins, Colorado, USA.
Caamano, D., Goodwin, P., and Manic, M. (2006) Derivation of a bed-load sediment
transport formula using artificial neural networks. In: Seventh International
Conference on Hydroinformatics HIC, Nice, France.
Camenen, B., and Larson, M. (2005) A general formula for non-cohesive bed load
sediment transport. Estuarine, Coastal and Shelf Science 63, 249-260.
Chang, C., Azamathulla, H., Zakaria, N., and Ghani, A. (2012) Appraisal of soft
computing techniques in prediction of total bed material load in tropical
rivers. Journal of Earth System Science 121, 125-133.
Cheng, N.-S. (2002) Exponential formula for bedload transport. Journal of Hydraulic
Engineering 128, 942-946.
Church, M. (2006) Bed Material Transport and the Morphology of Alluvial River
Channels. Annu. Rev. Earth Planet. Sci. 34, 325-354.
245
Church, M., Wolcott, J. F., and Fletcher, W. K. (1991) A test of equal mobility in
fluvial sediment transport: behavior of the sand fraction. Water Resources
Research 27, 2941-2951.
Claude, N., Rodrigues, S., Bustillo, V., Bréhéret, J.-G., Macaire, J.-J., and Jugé, P.
(2012) Estimating bedload transport in a large sand–gravel bed river from
direct sampling, dune tracking and empirical formulas. Geomorphology 179,
40-57.
Curran, J., and Wilcock, P. (2005) Effect of Sand Supply on Transport Rates in a
Gravel-Bed Channel. Journal of Hydraulic Engineering 131, 961-967.
Delft3D (1999) Delft hydraulic users’ manual [Online], [Accessed 1999]. Available
from World Wide Web:
http://www.deltaressystems.com/hydro/product/621497/delft3d-suite
Engelund, F., and Fredsoe, J. (1982) Hydraulic theory of alluvial rivers. Advances in
Hydroscience 13, 187-215.
246
Ergenzinger, p. (1992). River bed adjustments in a steep step-pool system: Lainbach,
Upper Bavaria. In "Dynamics of Gravel Bed Rivers" (P. Billi, R. D. Hey, C.
R. Thorne and P. Tacconi, eds.). John Wiley, Chichester, UK, pp. 415-430.
Fan, S. S., ed. (1988) Twelve selected computer stream sedimentation models
developed in the United States. U.S. Interagency Subcommittee Rep. on
Sedimentation, Federal Energy Regulatory Commission, Washington, D.C.
Fernandez Luque, R., and Van Beek, R. (1976) Erosion And Transport Of Bed-Load
Sediment. Journal of Hydraulic Research 14, 127-144.
Force, E. R., Eidel, J. J., and Maynard, J. B. (1991). "Sedimentary and diagenetic
mineral deposits: a basin analysis approach to exploration," Society of
Economic Geologists.
Frey, P., and Church, M. (2011) Bedload: a granular phenomenon. Earth Surface
Processes and Landforms 36, 58-69.
Fujita, I., and Komura, S. (1988). Visualization of the flow at a confluence. In "Third
Internatinal Symposium on Refined Flow Modelling and Thrbulence
Measurements", International Association of Hydraulic Research, Tokyo,
Japan, pp. 611-618.
247
Garcia, C., Laronne, J. B., and Sala, M. (2000) Continues monitoring of bed load
flux in a mountain gravel bed river. Geomorphology 34, 23-31.
Ghia, U., Ghia, K., and Shin, C. (1982) High-Re solutions for incompressible flow
using the Navier-Stokes equations and a multigrid method. Journal of
Computational Physics 48, 387-411.
Ghoshal, K., Mazumder, B. S., and Purkait, B. (2010) Grain-size distributions of bed
load: Inferences from flume experiments using heterogeneous sediment beds.
Sedimentary Geology 223, 1-14.
Gill, M. A. (1972) Erosion of sand beds around spur dikes. Journal of the Hydraulics
Division, ASCE 98, 1587-1602.
Gomez, B. (2006) The potential rate of bed load transport PNAS 103, 17170-17173.
Gomez, B., and Church, M. (1989) An assessment of bed load sediment transport
formulae for gravel bed rivers. Water Resour Research 26, 1161-1186.
248
Harris, E. L., Babovic, V., and Falconer, R. A. (2003) Velocity predictions in
compound channels with vegetated floodplains using genetic programming.
River Basin Management 1, 117-123.
Hinchliffe, M. P., Willis, M. J., Hiden, H., Tham, M. T., McKay, B., and Barton, G.
W. (1996) Modelling chemical process systems using a multi-gene genetic
programming algorithm. Late Breaking Papers at the Genetic Programming
1996 Conference, 56-65.
Karp, R. M., Shenker, S., and Papadimitriou, C. H. (2003) A simple algorithm for
finding frequent elements in streams and bags. ACM Transactions on
Database Systems (TODS) 28, 51-55.
Khorram, S., and Ergil, M. (2010) Most Influential Parameters for the Bed-Load
Sediment Flux Equations Used in Alluvial Rivers1. JAWRA Journal of the
American Water Resources Association 46, 1065-1090.
Khu, S. T., Liong, S.-Y., Babovic, V., Madsen, H., and Muttil, N. (2001) Genetic
programming and its application in real-time runoff forecasting. JAWRA
Journal of the American Water Resources Association 37, 439-451.
249
King, I. P. (1988). "A finite element model for three dimensional hydrodynamic
systems." Waterways Experiment Station, U.S. Army Corps of
Engineers.Vicksburg, Miss.
Kisi, O., Dailr, A. H., Cimen, M., and Shiri, J. (2012) Suspended sediment modeling
using genetic programming and soft computing techniques. Journal of
Hydrology 450–451, 48-58.
Kuhnle, R. A. (1989) Bed surface size changes in gravel bed channel Journal of
Hydraulic Engineering 115, 731-741.
Kuhnle, R. A. (1992) Bed load transport during rising and falling stages on two small
streams. Earth Surface Processes and Landforms 17, 191-197.
Kuhnle, R. A. (1993) Fluvial transport of sand and gravel mixtures with bimodal size
distributions. Sedimentary Geology 85, 17-24.
Landsberg, A., Chtchelkanova, A., Lind, C., Boris, J. , and Young, T. (1998) Fast3D
user and programmer reference manual [Online], [Accessed 1998]. Available
from World Wide Web: http://www.lcp.nrl.navy.mil/csm-
cta/csm4_refman.html
Lane, S. N., Bradbrook, K. F., Richards, K. S., Biron, P. M., and Roy, A. G. (1999)
Time-averaged flow structure in the central region of a stream confluence: a
discussion. Earth Surface Processes and Landforms 24, 361-367.
Lanzoni, S., and Tubino, M. (1999) Grain sorting and bar instability. Journal of Fluid
Mechanics 393, 149-174.
Laronne, J. B., Alexandrov, Y., Bergman, N., Cohen, H., Garcia, C.,, Habersack, H.,
Powell, D. M., and Reid, I. (2003). The continuous monitoring of bedload
flux in various fluvial systems In "Erosion and Sediment Transport
Measurement in Rivers—Technological and Methodological Advances" (J.
Bogen, T. Fergus and D. Walling, eds.). IAHS Publication 283, pp. 134-145.
250
Lefort, P. (1991). "Transport solide dans le lit des cours d’eau: dynamique fluviale,"
ENSHMG-INPG.
Legates, D. R., and McCabe, G. J., Jr. (1999) Evaluating the use of "goodness-of-fit";
Measures in hydrologic and hydroclimatic model validation. Water Resources
Research 35, 233-241.
Leite Ribeiro, M., Blanckaert, K., Roy, A. G., and Schleiss, A. J. (2012) Flow and
sediment dynamics in channel confluences. Journal of Geophysical Research
117, F01035.
Lingireddy, S., and Brion, G. M. (2005). "Artificial Neural Networks In water supply
engineering," American Society of Civil Engineers.
Liong, S.-Y., Gautam, T. R., Khu, S. T., Babovic, V., Keijzer, M., and Muttil, N.
(2007) Genetic programming: a nemw paradigm in rainfall runoff modeling.
JAWRA Journal of the American Water Resources Association 38, 706-718.
Liu, J., Savenije, H. H. G., and Xu, J. (2003) Forecast of water demand in Weinan
City in China using WDF-ANN model. Physics and Chemistry of the Earth,
Parts A/B/C 28, 219-224.
Maier, H. R., and Dandy, G. C. (2000) Neural networks for the prediction and
forecasting of water resources variables: a review of modelling issues and
applications. Environmental Modelling & Software 15, 101-124.
Martin, Y. (2003) Evaluation of bed load transport formulae using field evidence
from the Vedder River, British Columbia. Geomorphology 53, 75-95.
McGuirk, J. J., and Rodi, W. (1978) A depth averaged mathematical model for the
near field of side discharges into open channel flow. Journal of Fluid
Mechanics 86, 761-81.
Meyer-Peter, E., and Müller, R. (1948) Formulas for bed-load transport. Proceedings
of the 2nd Meeting of the International Association for Hydraulic Structures
Research, 39-64.
Molinas, A., and Wu, B. (2001) Transport of sediment in large sand-bed rivers.
Journal of Hydraulic Research 39, 135-146.
251
Muskatirovic, J. (2008) Analysis of bedload transport characteristics of Idaho
streams and rivers. Earth Surface Processes and Landforms 33, 1757-1768.
Muttil, N., and Lee, J. H. W. (2005) Genetic programming for analysis and real-time
prediction of coastal algal blooms. Ecological Modelling 189, 363-376.
Nagy, H. M., Watanabe, K., and Hirano, M. (2002) Prediction of Sediment Load
Concentration in Rivers using Artificial Neural Network Model. Journal of
Hydraulic Engineering 128, 588-595.
Nasseri, M., Moeini, A., and Tabesh, M. (2011) Forecasting monthly urban water
demand using Extended Kalman Filter and Genetic Programming. Expert
Systems with Applications 38, 7387-7395.
Niekerk, A. v., Vogel, K., Slingerland, R., and Bridge, J. (1992) Routing of
heterogeneous sediments over movable bed: model development. Journal of
Hydraulic Engineering 118, 246-262.
Nino, Y., and Garcia, M. (1998) Experiments on saltation of sand in water. Journal
of Hydraulic Engineering-Asce 124, 1014-1025.
Papanicolaou, A., Elhakeem, M., Krallis, G., Prakash, S., and Edinger, J. (2008)
Sediment Transport Modeling Review—Current and Future Developments.
Journal of Hydraulic Engineering 134, 1-14.
Parker, G., Klingeman, P. C., and McLean, D. G. (1982) Bedload and Size
Distribution in Paved Gravel-Bed Streams. Journal of the Hydraulics
Division,ASCE 108, 544-571.
Parsons, D. R., Best, J. L., Lane, S. N., Orfeo, O., Hardy, R. J., and Kostaschuk, R.
(2007) Form roughness and the absence of secondary flow in a large
confluence–diffluence, Rio Paraná, Argentina. Earth Surface Processes and
Landforms 32, 155-162.
252
Pica, M. (1972) Su Alcuni Aspetti del Trasporto Solido in Alvei Torrentizi.
L’Energia Elettrica 8, 497-508.
Pitlick, J., Cui, Y., and Wilcock, P. (2009) Manual for computing bed load transport
using BAGS (Bedload Assessment for Gravel-bed Streams) Software. Gen.
Tech. Rep. RMRS-GTR-223. Fort Collins, CO: U.S. Department of
Agriculture, Forest Service, Rocky Mountain Research Station. 45.
Powell, D. M., Reid, I., and Laronne, J. B. (2001) Evolution of bed load grain size
distribution with increasing flow strength and the effect of flow duration on
the caliber of bed load sediment yield in ephemeral gravel bed rivers. Water
Resources Research 37, 1463-1474.
Raghuwanshi, N. S., Singh, R., and Reddy, L. S. (2006) Runoff and sediment yield
modeling using artificial neural networks: upper Siwane River, India. Journal
of Hydrologic Engineering 11, 71-79.
Rennie, C., Millar, R., and Church, M. (2002) Measurement of Bed Load Velocity
using an Acoustic Doppler Current Profiler. Journal of Hydraulic Engineering
128, 473-483.
Rhoads, B. L., Riley, J. D., and Mayer, D. R. (2009) Response of bed morphology
and bed material texture to hydrological conditions at an asymmetrical stream
confluence. Geomorphology 109, 161-173.
253
Rhoads, B. L., and Sukhodolov, A. N. (2001) Field investigation of three-
dimensional flow structure at stream confluences: 1. Thermal mixing and
time-averaged velocities. Water Resources Research 37, 2393-2410.
Rhoads, B. L., and Sukhodolov, A. N. (2004) Spatial and temporal structure of shear
layer turbulence at a stream confluence. Water Resources Research 40,
W06304.
Rhoads, B. L., and Sukhodolov, A. N. (2008) Lateral momentum flux and the spatial
evolution of flow within a confluence mixing interface. Water Resources
Research 44, W08440.
Ribberink, J. S. (1998) Bed-load transport for steady flows and unsteady oscillatory
flows. Coastal Engineering 34, 59-82.
Rodi, W. (2006) DNS and LES of some engineering flows. Fluid Dynamics Research
38, 145-173.
Ryan, S. E., Bunte, K., and Potyondy, J. P. (2005). Breakout session II, Bedload-
transport measurement: Data needs, uncertainty, and new technologies. In
"Proceedings of the Federal Interagency Sediment Monitoring Instrument and
Analysis Workshop", Circular Vol. 1276.
Ryan, S. E., and Emmett, W. W. (2002). "The nature of flow and sediment
movement in Little Granite Creek near Bondurant, Wyoming.." Gen. Tech.
Rep. RMRS-GTR-90. Ogden, UT: U.S. Department of Agriculture, Forest
Service, Rocky Mountain Research Station.
Sasal, M., Kashyap, S., Rennie, C. D., and Nistor, I. (2009) Artificial neural network
for bedload estimation in alluvial rivers. Journal of Hydraulic Research 47,
223-232.
254
Scott, S. H., and Jia, a. Y. (2005) Simulation of sediment transport and channel
morphology change in large river systems. US-CHINA Workshop on
advanced computational modelling in hydroscince & enginnerin. September
19-21, Oxford, Mississippi, USA.
Sear, D. A. (2003). Event bed load yield measurement with load cell bed load traps
and prediction of bed load yield from hydrograph shape. In "Erosion and
Sediment Transport Measurement in Rivers: technological and
methodological advances" (J. Bogen, T. Fergus and D. E. Walling, eds.).
Wallingford, UK, IAHS Press (International Association of Hydrological
Sciences Publication 283), pp. 146-153.
Sear, D. A., Damon, W., Booker, D. J., and Anderson, D. G. (2000) A Load cell
based continues recording bed load trap. Earth Surface Processes and
Landforms 25, 659-672.
Simons, D. B., and Şentürk, F. (1992). "Sediment Transport Technology: Water and
Sediment Dynamics," Water Resources Publ.
Song, Y., and Haidvogel, D. (1994) A Semi-implicit Ocean Circulation Model Using
a Generalized Topography-Following Coordinate System. Journal of
Computational Physics 115, 228-244.
255
Southard, J. (2006). An Introduction to Fluid Motions, Sediment Transport, and
Current-generated Sedimentary Structures. In "MIT Open Course Ware",
Massachussets Institute of Technology.
Spasojevic, M., and Holly, F. M. (2000). Field data and 3D mobile-bed modeling:
Help or hindrance? In "4th Int. Conf. on Hydroinformatics", Iowa.
Stevaux, J. C., paes, R. J., Franco, A. A., Etchebehere, M. L. d. C., and Fujita, R. H.
(2009) Morphodynamics in the confluence of large regulated rivers: the case
of Paraná and Paranapanema rivers. Latin american journal of sedimentology
and basin analysis 16 101-109.
Sutherland, A. J. (1987). "Static armour layers by selective erosion," Wiley & Sons.
Szupiany, R. N., Amsler, M. L., Parsons, D. R., and Best, J. L. (2009) Morphology,
flow structure, and suspended bed sediment transport at two large braid-bar
confluences. Water Resour. Res. 45, W05415.
Tabesh, M., and Dini, M. (2009) Fuzzy and neuro-fuzzy models for short-term water
demand forecasting in Tehran. Iranian Journal of Science and
Technology,Transaction B: Technology 33, 61-77.
Wang, X., Yang, Q., Lu, W., and Wang, X. (2011) Effects of bed load movement on
mean flow characteristics in mobile gravel beds. Water Resources
Management, 1-15.
Wathen, S. J., Ferguson, R. I., Hoey, T. B., and Werritty, A. (1995) Unequal
Mobility of Gravel and Sand in Weakly Bimodal River Sediments. Water
Resources Research 31, 2087-2096.
Wentworth, C. K. (1922) A scale of grade and class terms for clastic sediments.
Geology 30, 377-392.
256
Whigham, P. A., and Crapper, P. F. (2001) Modeling rainfall-runoff using genetic
programming. Mathematical and Computer Modelling 33, 707-721.
Wiberg, P., and Smith, D. J. (1989) Model for Calculating Bed Load Transport of
Sediment. Journal of Hydraulic Engineering 115, 101-123.
Wilcock, P., and Crowe, J. (2003) Surface-based Transport Model for Mixed-Size
Sediment. Journal of Hydraulic Engineering 129, 120-128.
Wilcock, P. R., and Kenworthy, S. T. (2002) A two-fraction model for the transport
of sand/gravel mixtures. Water Resources Research 38, 1194.
Wilcock, P. R., Kenworthy, S. T., and Crowe, J. C. (2001) Experimental study of the
transport of mixed sand and gravel. Water Resources Research 37, 3349-
3358.
Wilcock, P. R., and Southard, J. B. (1989) Bed load transport of mixed size
sediment: Fractional transport rates, bed forms, and the development of a
coarse bed surface layer. Water Resources Research 25, 1629-1641.
Wilson, K. (1966) Bed load transport at high shear stress. Journal of Hydraulic
Engineering 92 49-59.
Wong, M., and Parker, G. (2006) Reanalysis and correction of bed-load relation of
Meyer-Peter and Muller using theirown database. Journal of Hydraulic
Engineering, ASCE 132, 1159-1168.
Wu, C. L., Chau, K. W., and Li, Y. S. (2008) River stage prediction based on a
distributed support vector regression. Journal of Hydrology 358, 96-110.
Wu, W. (2007). "Computational river dynamics," CRC Press, Boca Raton, Florida.
Wu, W., Wang, S. S. Y., and Jia, Y. (2000) Nonuniform sediment transport in
alluvial rivers. Journal of Hydraulic Research 38, 427-434.
Yang, C. T. (1972) Unit stream power and sediment transport. Journal of the
Hydraulics Division, ASCE 98, 1805-26.
257
Yang, C. T. (1996) Sediment transport theory and practice. McGraw-Hill, New York.
Yeganeh-Bakhtiary, A., Shabani, B., Gotoh, H., and Wang, S. S. Y. (2009) A three-
dimensional distinct element model for bed-load transport. Journal of
Hydraulic Research 47, 203-212.
Yitian, L. I., and Gu, R. R. (2003) Modeling flow and sediment transport in a river
system using an artificial neural network. Environmental Management 31,
0122-0134.
Zakaria, N. A., Azamathulla, H. M., Chang, C. K., and Ghani, A. A. (2010) Gene
expression programming for total bed material load estimation—a case study.
Science of The Total Environment 408, 5078-5085.
258
8 APPENDIX A
CONFLUENCE
Figure B.1: Survoyed boundry and cross sections in Ara-Kurau confluence (9 April
2012)
Table B.1: The measured data with river surveyor of Ara River in Ara-Kurau confluence
Elevation (m)
15.6 15.5
15.1 15.0
Q= 35 (mᵌ/s) Q= 35 (mᵌ/s)
Q= 25 (mᵌ/s) Q= 25 (mᵌ/s)
14.6 14.5
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
16.5 16.5
RSK2 RSK3
16 16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
16.5
16.5
RSK4 RSK5
16
16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
17 16.5
16.5
RSK6 RSK7
16
Elevation (m)
Elevation (m)
16
15.5
15.5
15 15
Elevation (m)
15.6 15.5
15.1 15.0
Q= 25 (mᵌ/s)
Q= 13 (mᵌ/s) Q= 25 (mᵌ/s)
14.6 Q= 13 (mᵌ/s)
14.5
0 10 20 30
0 5 10 15 20 25
Distance (m) Distance (m)
16.5 16.5
RSK2 RSK3
16 16
Elevation (m)
Elevation (m)
15.5 15.5
15 15
Elevation (m)
16
15.5
15.5
15
15
14.5 Q= 25 (mᵌ/s) 14.5 Q= 25 (mᵌ/s)
Q= 13 (mᵌ/s) Q= 13 (mᵌ/s)
14 14
0 10 20 30 0 10 20 30
Distance (m) Distance (m)
17 17
RSK6 RSK7
16.5 16.5
Elevation (m)
Elevation (m)
16 16
15.5 15.5
15 15
14.5 Q= 25 (mᵌ/s)
14.5 Q= 25 (mᵌ/s)
Q= 13 (mᵌ/s)
Q= 13 (mᵌ/s) 14
14
0 10 20 30 40
0 10 20 30
Distance (m) Distance (m)