water

Article
Simulating Laboratory Braided Rivers with Bed-Load
Sediment Transport
Haiyan Yang 1 , Binliang Lin 2, *, Jian Sun 2 and Guoxian Huang 2
1 Department of Hydraulic Engineering, South China Agricultural University, Guangzhou 510642, China;
haiyan.yang2@gmail.com
2 State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, China;
jsun@mail.tsinghua.edu.cn (J.S.); hgx01@mails.tsinghua.edu.cn (G.H.)
* Correspondence: LinBL@tsinghua.edu.cn; Tel.: +86-10-62796544

Received: 12 June 2017; Accepted: 5 September 2017; Published: 8 September 2017

Abstract: Numerical models provide considerable assistance in the investigation of complicated

processes in natural rivers. In the present study, a physics-based two-dimensional model has been
developed to simulate the braiding processes and morphodynamic changes in braided rivers. The
model applies the basic hydrodynamic and sediment transport principles with bed morphology
deformation and a TVD (Total Variation Diminishing) scheme to predict trans-critical flows and
bed morphology deformation. The non-equilibrium transport process of graded bed load sediment
is simulated, with non-uniform sediments, secondary flows, and sheltering effects being included.
A multiple bed layer technique is adopted to represent the vertical sediment sorting process. The
model has been applied to simulate the bed evolution process in an experimental river with bed load
transport. Comparisons between the experimental river and predicted river are analysed, including
their pattern evolution processes, important braiding phenomena, and statistical characteristics.
Avulsion activities have been found in the braiding evolution process, representing the primary ways
in which channels form and disappear in braided rivers. The increases in the active braiding intensity
and total braiding intensity show similar trends to those observed in the experimental river. Statistical
methods are applied to assess the scale-invariant topographic properties of the simulated river and
real rivers. The model demonstrated its potential to predict the morphodynamics in natural rivers.

Keywords: physics-based numerical model; non-uniform sediment; braided river; statistical

characteristics

1. Introduction
Braided rivers are a distinctive river morphology with high flow energy and frequent channel
changes. Braiding is one kind of fluvial pattern in these rivers characterized by complex dynamics,
in which water and sediment flow are divided into multiple branches, joining and splitting at the nodes
and shifting continuously in the floodplain [1]. The understanding of morphodynamics in braided
rivers is especially important for managing engineering problems such as flood management, riverbank
erosion, sedimentation in reservoirs and navigable waterways, and the design and construction of
artificial channels and bridges.
Braiding usually occurs with a highly variable discharge, an abundant sediment load, erodible
banks, steep valley slopes, and a high width/depth ratio [2]. A change in one of these parameters
might cause an exchange between meandering and braided patterns such as a large flood, an increased
sediment supply, a steeper riverbed, or reduced bank erodibility. Ham and Church studied the lower
reach of Chilliwack River in Canada, which is partially braided, and found that a large flood with
a return period of about five years can be considered a significant channel-forming event [3]. The
lower William River in Canada adjusted its channel from a relatively narrow and deep single-channel

Water 2017, 9, 686; doi:10.3390/w9090686 www.mdpi.com/journal/water

Water 2017, 9, 686 2 of 18

stream to a thoroughly braided pattern after picking up a 40-fold increase of bed load sediment [4]. The
Squamish River experienced a downstream sequence of braided-wandering-meandering in a 20 km
reach with a slope decrease from 0.0058 to 0.0015 [5]. Brierley and Hickin studied flow patterns
from minimum flow energy and found that meandering rivers may occur even as flow energy has
reached the level for braiding, since scouring has been confined by relatively cohesive or vegetated
banks [6]. Generally, although one of the factors mentioned above can initiate braided rivers, in nature
frequently two or more of them occur simultaneously. Moreover, it is very important to study rivers
over an extended time span because their appearance is controlled by the history of changing flow
stages, and some geometries arise specially as a result of changing discharge [7].
Numerical models provide useful tools for simulating fluvial processes and morphology in rivers,
and, in recent decades, efforts have been made to investigate advanced numerical models to better
represent the complex morphodynamic processes in rivers. A few models have been developed to
simulate the processes of sediment transport, bed deformation, and sediment downstream fining
in rivers (e.g., [8–11]). Compared with reduced-complexity approaches (e.g., [12,13]), physics-based
models provide more detailed process information to better understand natural rivers due to their
better representation of hydraulic and morphodynamic processes. Physics-based models have
produced braided patterns, bar formation, and avulsion activities observed in natural braided
rivers and have made significant progress (e.g., [14,15]). The evolution of wide shallow channels
and the behaviour of alternate bars in laboratories have been reproduced by a 2-D model [16,17].
Morphodynamic models such as HSTAR (Hydrodynamics and Sediment Transport in Alluvial
Rivers) [18], depth-integrated Delft3D [14,19], and 2-D morphodynamic models [20,21] have been
applied to simulate large sand-bed braided rivers and have produced morphological units with
evolution mechanisms, the characteristic morphology of compound bars, and channels and statistical
characteristics that are similar to those of natural sand-bed rivers. However, their development
and application are still in an early stage [22], and quantitative analyses and comparisons between
simulated and real rivers are still insufficient [14] especially for braided rivers dominated by bed-load
sediment transport.
In braided river simulation, previous models have incorporated basic sediment transport theories
with essential effects and have shown considerable potential (e.g., [23,24]). The effect of secondary
flow has been integrated into the shallow water equations [18,25] or in the sediment transport rate
equation [16,17]. A TVD (Total Variation Diminishing) scheme, which is a modified Ultimate Quickest
scheme, has been employed to solve the flow conservation equations (e.g., [23]), since the scheme is
more efficient than the implicit ADI (Alternating Direction Implicit) scheme in predicting trans-critical
flows that usually exist in shallow areas of braided rivers. Bank erosion, which is important for the
emergence of bars and the initiation of meandering or braiding, has been examined theoretically
through the mechanical processes of channels [26,27] and has been proposed to reproduce lateral
changes by the existing braided river models (e.g., [17,18]).
However, most of the existing physics-based models for braided river simulation adopted
a uniform sediment transport approach (e.g., [14,19,28]), except for the model of Nicholas [18] in
which two grain fractions of sand and silt were considered. Nevertheless, non-uniform sediments
exist in natural rivers, and different size fractions of sediment particles are transported at different
rates [29]. Sediment gradation plays an essential role in the bed deformation process, which is
especially important for the evolution of local units in natural [30] and laboratory braided rivers [31].
In addition, laboratory experiments [32] also show the influence of vertical grain sorting in the
active layer, indicating that it is necessary to divide the riverbed into multiple bed layers in river
simulation. This mixing layer concept has been adopted by many researchers such as Wu [29] and Van
Niekerk et al. [33]. Therefore, in order to represent a more realistic morphodynamic process in braided
rivers, it is necessary for a numerical model to include the effect of non-uniform sediment transport.
In the current study, a 2-D physics-based model has been developed to simulate an experimental
river to study the morphodynamics of braided rivers dominated by bed-load sediment. The model
Water 2017, 9, 686 3 of 18

incorporates the basic hydrodynamic and sediment transport theories, considers non-uniform
sediments in a fractional way, and adopts multiple bed layers to represent the erosion-deposition
process and the sorting effects among fractional sediments in beds. The study aims to investigate the
river morphodynamic processes and activities, analyse the statistical characteristics, compare them
with the experimental river and natural rivers, and discuss the potential of the model for representing
the morphodynamic processes in real rivers.

2. Numerical Model and Test

Flow, sediment transport, and bed deformation are normally the main processes considered in
river morphodynamic simulations. In the present model, flow and sediment transport are calculated
by solving the depth-integrated 2D shallow water equations and the non-equilibrium sediment
transport equations.

2.1. Numerical Model

2.1.1. Governing Equations for Flow and Sediment Transport

The flow in rivers is governed by the 2-D shallow water equations, given as follows:

∂ξ ∂p ∂q
+ + =0 (1)
∂t ∂x ∂y
p  2
gp p2 + q2 ∂ p ∂2 p ∂2 q

∂p ∂βpU ∂βpV ∂ξ
+ + = f q − gH − +ε 2 2 + 2 + (2)
∂t ∂x ∂y ∂x H 2 C2 ∂x ∂y ∂xy
p
gq p2 + q2 ∂2 q ∂2 q ∂2 p
 
∂q ∂βqU ∂βqV ∂ξ
+ + = − f p − gH − + ε 2 + + (3)
∂t ∂x ∂y ∂y H 2 C2 ∂x2 ∂y2 ∂xy
where ξ is the water surface elevation above the datum; p and q are the discharges per unit of width in
the x and y directions, respectively; β is the momentum correction factor for a non-uniform vertical
velocity profile; U and V are the depth-averaged velocity components in the x and y directions,
respectively; f is the Coriolis parameter due to the Earth’s rotation; H is the water depth, which is
equal to h + ξ, with h as the water depth below the datum; C is the Chézy coefficient, which equals
to 18log10 (12H/k s ), with k s denoting the hydraulic roughness, given according to van Rijn [34]; and
ε is the depth-averaged turbulent eddy viscosity. The hydrodynamic model, which is solved by the
TVD-MacCormack Scheme, has been validated in previous studies [35,36], showing its efficiency in
simulating trans-critical flows.
Assuming that the graded bed material can be divided into N fractions, the transport of the kth
fraction can be described by the following equation:
 
∂(qbk /ub ) ∂(αbx qbk ) ∂ αby qbk 1
+ + = (q − qbk ) (4)
∂t ∂x ∂y Ls b∗k

where qbk is the bed load transport rate of the kth size fraction; ub is the bed load transport velocity,
determined according to the equation of van Rijn [37]; k is the sediment fraction; qb∗k is the equilibrium
bed load transport rate of the kth size fraction; αbx , αby are the directions of the bed load movement;
and Ls is the non-equilibrium adaption length, denoting a characteristic distance for the sediment in
transport to adjust from a non-equilibrium to an equilibrium state.
The equilibrium bed load transport rate qbk (in m2 /s) for particles in the range of 0.2 to 2 mm can
be calculated by the following equations [38]:
Water 2017, 9, 686 4 of 18


Tk2.1
 pbk · 0.053[(s − 1) g]0.5 d1.5

k D0.3 , f or Tk < 3
qb∗k = ∗k (5)
Tk1.5
 pbk · 0.1[(s − 1) g]0.5 d1.5

k D0.3 , f or Tk ≥ 3
∗k

where pbk is the percentage of the kth size fraction in the mixing layer of the bed; s is the sediment
specific density; and dk is the representative particle diameter of the kth size fraction. The dimensionless
transport stage parameter Tk and the dimensionless particle parameter D∗k are given according to van
Rijn [37].

2.1.2. Influence of Bed Slope and Secondary Flow

In natural rivers, when the channel slopes are gentle, the effect of gravity on sediment transport
is usually ignored. However, when the riverbed slope is steep, gravity may change the sediment
transport capacity and influence bed deformation significantly. According to van Rijn [38], the shear
stress with the consideration of the slope effects can be expressed by:

τc,s = k δ,s k δ,n τc (6)

where k δ,s is the coefficient of longitudinal slope, with its value equal to sin(θ − δs )/ sin θ for downslope
flow (k δ,s < 1) and sin(θ + δs )/ sin θ for upslope flow (k δ,s > 1), with θ as the sediment’s angle of
repose and δs as the longitudinal slope angle; and k δ,n is the coefficient of the transverse slope, which

0.5
equals to cos δn 1 − tan2 δn /tan2 θ , with δn as the transverse slope angle.
The sediment transport rate for a sloping bed is derived as [38]:

qb,s = αs qb (7)

where αs is the Bagnold slope factor, with its value equal to tan θ/[cos δs (tan θ ± tan δs )], where ‘+’
represents upslope flow and ‘−’ represents downslope flow; and qb is the bed load transport rate for
a horizontal bed.
The bed load transport direction is influenced by the secondary flow and the transverse slope.
To consider this effect, the deviation between the bed flow and the sediment transport rate is integrated
into the bed load transport rate. Referring to the equations of Jang and Shimizu [16] and van Rijn [38],
the sediment transport rate for a transverse slope can be presented as:
"  0.5 #
H τc
qb,n = N∗ + e tanδn qb (8)
rs τb

where, on the right side, the two terms denote the effects of secondary flow and transverse slope,
respectively; rs is the radius of curvature of the stream line; N∗ is the coefficient of the strength of
secondary flow, with an value of 7.0 being adopted [39]; and e is the mixing coefficient with a value of
1.5 [38].

2.1.3. Areal Fraction and Sheltering Effect for Non-uniform Sediments

To consider the non-uniform sediments in a bed, the forces exerted on graded particles are
supposed to be proportional to the projected areas of the particles exposed to flow, and the volumetric
percentage in the mixing layer of bed, pbk , can be converted into the areal percentage p ak by the
following equation [40]:
pbk /dk
p ak = (9)
∑k=1 ( pbk /dk )
0 , is considered by a correction of the critical shear
The sheltering effect in a particle mixture, τck
stress for particle movement initiation of the kth size fraction, giving [41]
Water 2017, 9, 686 5 of 18

0
τck = τck (dk /d50 )−1 (10)

where d50 is the median grain size of the sediment mixture.

2.1.4. Bed Deformation and Multiple Bed Layers

During the river evolution process, bed erosion or deposition is determined by the amount of
sediment in transport and the sediment transport capacity of the flow. Conversely, the changes of bed
morphology and composition will influence the flow velocity and sediment transport rate. The bed
deformation can be described as follows:
∂zb 1
(1 − p0)
∂t
= ∑ Ls (qbk − qb∗k ) (11)

where p0 is the porosity of the bed material and ∂zb /∂t is the bed elevation change rate. The
model supposed that when the bank slope becomes steeper than the sediment’s angle of repose,
the sediments will immediately slide into neighbouring areas. The amount of sediments deposited
on the neighbouring areas is supposed to be equal to what is eroded from the bank. The sediments
deposited on the riverbed will be mixed with the upper layer of the bed, and, consequently, the
formation and composition of the river bed are renewed.
To consider the influence of bed sediment composition, it is important for numerical models to be
capable of resolving the spatial and temporal variation of sediment gradations of the loose layers in the
riverbed. This is usually approached by dividing the loose bed into several vertical layers. The present
model adopts the three bed layer method of Wei [42], with sediment deposition and erosion calculated
in the dynamic layers. More details on the methods of multiple bed layers can be found in Zhou and
Lin [43]. The sediment transport equations are solved with the Ultimate QUIKEST Scheme, which
was developed to simulate 2-D solute transport in coastal and estuarine waters [44]. The sediment
sorting effect has been investigated [45]. More detailed information of the numerical model and model
application can be found in Yang [20].

2.2. Model Morphodynamic Test

The model for bed load transport was first tested with a laboratory experiment, undertaken by
Seal et al. [46] and Paola et al. [47]. The experiment was performed in a flume. The test reach was 45 m
long and 0.305 m wide, and the initial channel slope was 0.002. A pond located at the downstream end
was blocked by a weir, forming a free overfall. The water elevation at the tailgate was set to 0.4 m. The
grain size ranged from 0.125 mm to over 64 mm, with two main modes at 0.35 mm and 16 mm. The
flow discharge was set to 0.049 m3 /s, with a sediment feed rate of 11.3 kg/min. The sediment was
fed into the channel manually at 1 m downstream of the headgate for 16 h and 50 min. A model was
set up with the same boundary condition as the experimental river. The time steps were set to 0.005 s
and 0.01 s for the hydrodynamic and sediment computations, respectively. The minimum water depth
used to resolve the drying and flooding of the floodplain was 1 mm.
The predicted result together with the experimental data is shown in Figure 1. Generally, the
model predicts the bed elevation and the water surface well. As the water and sediment mixture
flowed into the channel, sediment particles were deposited onto the bed and caused it to aggrade
quickly. The riverbed aggraded vertically with an average speed of 0.0518 m/h, and the downstream
head of the newly formed bed moved forward with an average speed of about 2.5 m/h. The aggrading
bed was relatively flat with a slope of about 0.2%, which agreed closely with the experimental data. For
the longitudinal water surface curve, some oscillations were observed near the upstream inlet when the
flow was very shallow, but the oscillations did not develop. The water elevation at 16.8 h was generally
simulated well. A hydraulic jump could be observed at the downstream end of the main bed deposit,
which was also observed in the experiments. Near the bar, there was a rapid water level change,
Water 2017, 9, 686 6 of 18

where the Froude number increased rapidly from 0.2 to 0.9. It illustrates that the hydrodynamic
model is capable of coping with sudden hydraulic changes. Figure 2 shows the model-predicted time
series of flow velocity, water surface elevation, and bed load transport rate at sequential downstream
locations of9,a686
Water 2017, 10 m step. It illustrates the variations in local hydraulic conditions as the bed aggrades. 6 of 17
Water 2017, 9, 686 6 of 17
Due to a steeper bed slope on the newly formed riverbed, the flow sped up, with a high sediment
concentration
original
originaldepthandofaof0.4
depth shallower
0.4m). water
m).Similar
Similar depth of
changes
changes 0.12 m sequentially
occurred
occurred (compared
sequentiallyto the original
along
along themodel
the depth
model ofas0.4
river
river asthem).
the Similar
newly
newly
formed
changes bed
occurredgrew downstream.
sequentially
formed bed grew downstream. along the model river as the newly formed bed grew downstream.

Figure 1. Predicted bed elevationsand

and water surface
surface of
ofthe
theexperiment ofof
Seal et al. [46].
Figure1.1.Predicted
Figure Predictedbed
bedelevations
elevations and water
water surface of the experiment
experiment Seal
of Seal etetal.al.[46].
[46].

Figure 2. Model predicted (a) flow velocity; (b) water depth; and (c) bed load transport rate.
Figure 2. Model predicted (a) flow velocity; (b) water depth; and (c) bed load transport rate.
Figure 2. Model predicted (a) flow velocity; (b) water depth; and (c) bed load transport rate.
3. Braided River Simulation
3. Braided River Simulation
3. Braided
3.1. Model River
SetupSimulation
3.1. Model Setup
3.1. Model The model was applied to a laboratory experiment conducted by Egozi and Ashmore [31], in
Setup
which
The amodelbraided was river pattern
applied to was generallyexperiment
a laboratory developed. conducted
The river was by originally
Egozi anddesignedAshmoreto[31], be ain
The model
generic was applied
Froude-scaled modelto a laboratory
with experiment conducted byinEgozi and Ashmore [31],
which a braided river pattern was the prototype
generally being the The
developed. Sunwapta
river wasRiver Canada.
originally Froude-scaled
designed to be a
in generic
which
models aFroude-scaled
braided
have often river
been pattern
applied
model withwas
in
the generally
prototypedeveloped.
geomorphological
being studies The river
of gravel
the Sunwapta wasinoriginally
braided
River designed
rivers (e.g.,
Canada. [48–51]).
Froude-scaled to be
a generic
They Froude-scaled
models satisfy the same
have often beenmodel
Froude with
applied inthe
number prototype
geomorphological being
(Fr) as natural the yet
rivers
studiesSunwapta
relax
of theRiver
gravel in Canada.
Reynolds
braided similarity
rivers Froude-scaled
criterion
(e.g., [48–51]).
modelsfor have
flows often
that are been applied
hydraulically in geomorphological
rough and turbulent. studies
Such a modelof gravel
does
They satisfy the same Froude number (Fr) as natural rivers yet relax the Reynolds similarity criterion braided
not produce rivers
exact (e.g., [48–51]).
plan-form
They satisfy
similarity the
for same
a Froude
particular number
river but (Fr)
yields as natural
sampling rivers
results yet relax
applicable the
for flows that are hydraulically rough and turbulent. Such a model does not produce exact plan-form to Reynolds
full scale similarity
rivers in criterion
general
[52]. that
forsimilarity
flows forare hydraulically
a particular riverrough and turbulent.
but yields sampling Such resultsa model
applicabledoestonot fullproduce
scale riversexactinplan-form
general
similarity
[52]. Thefor experiments
a particular were
river performed
but yields in an 18
sampling m long
results and 3 m wide
applicable flume,
to full with
scale a constant
rivers in slope
general of[52].
0.015.
The The numerical model was set up by 18adopting the experimental conditions (Figure 3). The
Theexperiments
experimentswere were performed
performed in an
in an 18 mmlong
longand and 3m 3m widewide flume,
flume, withwith a constant
a constant slopeslope
of
sediments
of 0.015.
0.015.The were
Thenumerical divided
numericalmodel into
modelwas nine
wasset groups,
setup
upby with the
byadopting size
adoptingthe fraction boundaries
theexperimental
experimentalconditions listed in Table
conditions(Figure 1. Three
(Figure3).3).TheThe
bed layers were adopted. Initially the bed was assumed to be flattened, with a trapezoidal straight
sediments
sedimentswere weredivided
dividedinto intonine
ninegroups,
groups, with
with the size fraction
fraction boundaries
boundarieslisted listedininTable
Table1.1.Three
Three
channel in the middle (Figure 3b). The spatial step was set to 2 cm, while an adaptive time step was
bed layers were adopted. Initially the bed was assumed to be flattened,
bed layers were adopted. Initially the bed was assumed to be flattened, with a trapezoidal straight with a trapezoidal straight
used to enhance computational efficiency [35], with the minimum and maximum time steps being
channel
channel ininthethemiddle
middle(Figure
(Figure3b).
3b).The
Thespatial
spatial step was set set to
to 22 cm,
cm,while
whilean anadaptive
adaptivetime timestepstepwas
was
0.001 s and 0.01 s, respectively. The minimum water depth used to resolve the drying and flooding
used
used to to enhancecomputational
enhance computationalefficiency
efficiency [35],
[35], with
with the minimum
minimum and
and maximum
maximum time
time steps
steps being
being
of the floodplain was 1 mm. The model simulation time was approximately 133 h, during which the
0.001
0.001 s and0.01
s and
discharge 0.01s,s,respectively.
increased respectively.
from 1.4 L/s The
The minimum
tominimum water
water
2.1 L/s at time depth
71.5 h.used
t =depth used to
to resolve
resolvewere
The sediments the
thedrying
drying
collected and
and flooding
from flooding
the
of
of the the floodplain
floodplain
outlet and then waswas 1 mm.
1 mm.
re-fed to the The
The model simulation
model boundary
upstream simulation time
of time was
inlet,was approximately
approximately
just the same as in the133 133 h,h, during whichthe
during
experiment. which the
discharge increased from 1.4 L/s to 2.1 L/s at time t = 71.5 h. The sediments were collected from the
outlet and then re-fed to the upstream boundary of inlet, just the same as in the experiment.
Water 2017, 9, 686 7 of 18

discharge increased from 1.4 L/s to 2.1 L/s at time t = 71.5 h. The sediments were collected from the
outlet and then re-fed to the upstream boundary of inlet, just the same as in the experiment.
Water 2017, 9, 686 7 of 17

Figure 3. Schematic diagrams of the river flume: (a) schematic flume; (b) horizontal plan; and (c) cross-
Figure 3. Schematic diagrams of the river flume: (a) schematic flume; (b) horizontal plan; and
section A–A’.
(c) cross-section A–A’.
Table 1. Sediment fractions in the experimental river and the numerical model.
Table 1. Sediment fractions in the experimental river and the numerical model.
Sand Groups 1 2 3 4 5 6 7 8 9
Grain Size (mm) 0.25 0.5 0.7 1.0 1.5 2.0 3.0 4.0 6.0 8.0
Sand Groups 1 2 3 4 5 6 7 8 9
Percentage (%) 4.95 14.05 9.20 16.16 13.64 12.00 11.78 11.22 4.42 2.58
Grain Size (mm)
Finer 0.25
than (%) 0.5
4.95 0.7 28.20 1.0 44.36 1.5
19.00 58.00 2.0
70.00 3.0 93.00 4.0 97.42 6.0
81.78 100.0 8.0
Percentage (%) 4.95 14.05 9.20 16.16 13.64 12.00 11.78 11.22 4.42 2.58
3.2.than
Finer Morphodynamic
(%) 4.95 Properties
19.00 28.20 44.36 58.00 70.00 81.78 93.00 97.42 100.0

3.2.1. River Evolution Processes and Phenomena

3.2. Morphodynamic Properties
The time serial pictures denoting the evolution process of the model-predicted river are shown
3.2.1. River Evolution
in Figure Processes
4. It can be observedand
that Phenomena
the development of a braided pattern in the predicted river shares
many common properties with the experimental river (Figure 5, activity number identified with
The time4).serial
Figure Mostpictures denoting the
of the morphologic evolution
elements process observed
and activities of the model-predicted
in the experimentalriver are[31]
river shown in
Figurehave
4. Itbeen
can found
be observed that the
in the predicted river.development of a braided pattern in the predicted river shares
many common properties with the experimental river (Figure 5, activity number identified with
In the first few hours of model simulation, alternate bars and channels were formed in the
Figureupstream
4). Mostreach,
of the mainly distributedelements
morphologic in the initially
andcut channel (hour
activities 8). Next,
observed in one
the main channel began
experimental river [31]
to take shape under the
have been found in the predicted river.strengthening connection of the alternate channels by incision, bank erosion,
and channel migration (hour 12). In this channel, the water depth increased further along with
In the first few hours of model simulation, alternate bars and channels were formed in the
increasing sinuosity (hour 19, point 1). Intense erosion occurred along the bank of the initially cut
upstream reach, mainly distributed in the initially cut channel (hour 8). Next, one main channel
channel, and the channel migrated outward. As the channel became more sinuous, the water
began overflowed
to take shape fromunder the strengthening
the banks of certain local connection of the across
bends and spread alternatethe channels
unchannelledby incision,
margins bank
erosion, and channel
downstream migration
of the bends (hour (hour 12). 2),
24, point Inwhich
this channel,
were called the water depth increased
‘pseudoanabranches’ further along
and sometimes
grew into new
with increasing channels.
sinuosity At certain
(hour local 1).
19, point bends, the channel
Intense erosion wasoccurred
dissectedalongby the upstream
the banktributary,
of the initially
and oneand
cut channel, bifurcation was formed
the channel (hour 24,
migrated point 3). As
outward. At the
thebifurcation
channel of the newly
became avulsed
more areas,the
sinuous, the water
flow rate decreased in the original main channel and increased
overflowed from the banks of certain local bends and spread across the unchannelled margins in the new channel, and the new
channel consequently became more dominant, whilst the old channel died out gradually. This
downstream of the bends (hour 24, point 2), which were called ‘pseudoanabranches’ and sometimes
activity can be called avulsion by incision [15], through which the main channel tends to find a
grew into new channels.
straighter pathway byAt certain
incising localcourse.
a new bends, the channel was dissected by the upstream tributary,
and one bifurcation was formed (hour
The braiding configuration was developed 24, pointby3).hour
At 27.
theSome
bifurcation of the reconnected
active channels newly avulsed with areas,
the flow rate decreased in the original main channel and increased in
each other again by overflow or new channel erosion (hour 27, point 5), while some secondary the new channel, and the new
channels were gradually abandoned (hour 27, point 4). As some channels
channel consequently became more dominant, whilst the old channel died out gradually. This activity became more sinuous,
can bewater
called overflowed
avulsionatbytheincision
bends, forming new channels
[15], through which(hour 24, point
the main 6). Flow
channel concentrated
tends to find ain straighter
the
main channel, and accordingly it formed another channel bend (hour 36, point 7). Certain channels
pathway by incising a new course.
were choked by sediment deposition (hour 38, point 8), which can be classified as avulsion by
The braiding configuration was developed by hour 27. Some active channels reconnected with
progradation [15]. By hour 40, the upstream reach had widened significantly by extending
each other again by
downstream andoverflow or new
into the original channel
margin areas,erosion
evolving(hour
into a27, newpoint
stage.5),Onewhile
of thesome secondary
bifurcated
channels were gradually abandoned (hour 27, point 4). As some channels became more sinuous,
water overflowed at the bends, forming new channels (hour 24, point 6). Flow concentrated in
Water 2017, 9, 686 8 of 18

the main channel, and accordingly it formed another channel bend (hour 36, point 7). Certain
channels were choked by sediment deposition (hour 38, point 8), which can be classified as avulsion by
progradation [15]. By hour 40, the upstream reach had widened significantly by extending downstream
and into
Waterthe
2017,original
9, 686 margin areas, evolving into a new stage. One of the bifurcated 8 ofchannels
17
captured a large amount of flow from another and finally evolved into a deep channel, followed
by thechannels capturedofa the
disappearance largeoriginal
amountdominant
of flow from anotherofand
channel finally evolved
a bifurcation into44,
(hour a deep
pointchannel,
9). The main
followed by the disappearance of the original dominant channel of a bifurcation (hour 44, point 9).
channel shifted to one of its tributaries as the upstream flow conditions changed (hour 44, point 10).
The main channel shifted to one of its tributaries as the upstream flow conditions changed (hour 44,
Certainpoint
previous abandoned
10). Certain channels
previous (hourchannels
abandoned 38, point 8) were
(hour regenerated
38, point 8) were by channel annexation
regenerated by channel (hour
47, point 11). In the predicted river, a cycle of straight channels transitioning to sinuous
annexation (hour 47, point 11). In the predicted river, a cycle of straight channels transitioning channels
to and
then tosinuous
avulsions is anand
channels important
then to morphodynamic process morphodynamic
avulsions is an important during the evolutionary process
process during the of the
braidedevolutionary
river. These process of theplay
activities braided
an river. These role
important activities play an important
in maintaining role in maintaining
the dynamic the
braided form.
dynamic braided form.

FigureFigure 4. Morphologic
4. Morphologic changes
changes andand activities
activities during
during the
the evolutionprocess
evolution process of
of the model
model predicted
predicted river
(waterriver (water depth/mm).
depth/mm). 1: increasing
1: increasing sinuosity
sinuosity of the
of the mainmain channel;2:2: the
channel; the generation
generation and obliteration
and obliteration of
of pseudo-anabranches; 3: avulsion by incision; 4: abandonment of the secondary channel; 5:
pseudo-anabranches; 3: avulsion by incision; 4: abandonment of the secondary channel; 5: reconnecting
reconnecting of two active channels; 6: avulsion by overflow; 7: concentration of the flow in the main
of two active channels; 6: avulsion by overflow; 7: concentration of the flow in the main channel;
channel; 8: avulsion by progradation; 9: water diversion from one secondary channel to an existing or
8: avulsion by progradation; 9: water diversion from one secondary channel to an existing or new
new channel; 10: shifting of the main channel; and 11: regeneration of an abandoned channel.
channel; 10: shifting of the main channel; and 11: regeneration of an abandoned channel.
Water 2017, 9, 686 9 of 18
Water 2017, 9, 686 9 of 17

Figure 5. Morphologic changes and activities during the evolution process of the experimental river
5. Morphologic
Figure with changes and activities during the evolution process of the experimental river
the activity numbers identical to those in Figure 4.
with the activity numbers identical to those in Figure 4.
3.2.2. Properties of a Typical Braided River
3.2.2. Properties of a Typical
The model riverbed Braided River
was initially flat with a straight channel in the middle of the flume, and it
Thegradually
model evolved
riverbed into
wasa braided
initially river.
flat Awith
typical braided channel
a straight river wasin generated after of
the middle 41 the
h, with the and it
flume,
distributions of water depth, Froude number, bed shear stress, bed load concentration, and sediment
gradually evolved into a braided river. A typical braided river was generated after 41 h, with the
distribution shown in Figure 6. Key channel nodes, including bars, confluences, bifurcations, and a
distributions
series ofofsequential
water depth, Froude
pool-bar units number, bed shear
with repeated stress,
division bed load
and joining concentration,
of channels and sediment
can be observed
distribution shown in Figure 6. Key channel nodes, including bars, confluences,
(Figure 6a). These features share many similarities with the experimental river and natural rivers bifurcations, and
a series(e.g.,
of sequential
[7,31,53]). pool-bar units with repeated division and joining of channels can be observed
(Figure 6a).For most of
These the active
features channels,
share many thesimilarities
Froude number withwasthe
between 0.6 and 0.7 river
experimental (Figureand
6b). natural
In some rivers
local areas,
(e.g., [7,31,53]). especially in the areas where confluence and bifurcation occurred, the Froude number
varied in a range of 0.7 to 1.0. The distribution of bed shear stress has been shown to be relative to
For most of the active channels, the Froude number was between 0.6 and 0.7 (Figure 6b). In some
bed grain size distribution and water depth (Figure 6c). Higher shear stress values occurred mainly
local areas, especially in the areas where confluence and bifurcation occurred, the Froude number
in the main channel and bifurcation areas, and at times they also existed along channel edges.
varied in a range of 0.7
An active to 1.0.isThe
channel distribution
defined of bed
as a channel shear
with stress
visible bedhas been movement
material shown to be relative
[54]. In the to bed
grain size distribution
simulated river, ifand water
the bed depth
load (Figure 6c).
concentration Higher
is higher thanshear stress
a certain values
value (e.g.,occurred
6 g/m ), itmainly
3 can be in the
considered
main channel and‘active’; otherwise,
bifurcation areas,it and
is considered ‘non-active’.
at times they In thatalong
also existed sense,channel
a main edges.
active channel
An active channel is defined as a channel with visible bed material movement [54]. In the
simulated river, if the bed load concentration is higher than a certain value (e.g., 6 g/m3 ), it can be
considered ‘active’; otherwise, it is considered ‘non-active’. In that sense, a main active channel formed
in the river with the highest sediment concentration, and other active channels joined and bifurcated
from it (Figure 6d), which was similar to the experimental river. Due to the special sediment load
Water 2017, 9, 686 10 of 17
Water 2017, 9, 686 10 of 18
formed in the river with the highest sediment concentration, and other active channels joined and
bifurcated from it (Figure 6d), which was similar to the experimental river. Due to the special
method in
sediment themethod
load experiment in which
in the ‘the sand
experiment in leaving
which the
‘theend of the
sand flumethe
leaving wasend
recirculated from the
of the flume wastail
recirculated from the tail tank and fed back into the flume’ [31], finer sediments were more involved it
tank and fed back into the flume’ [31], finer sediments were more involved in the circulation because
inwas easier for them
the circulation to be
because it transported;
was easier fortherefore,
them to bethetransported;
sediments were often the
therefore, coarser on thewere
sediments two sides
often of
a channel than in the thalweg (Figure 6e).
coarser on the two sides of a channel than in the thalweg (Figure 6e).

Figure
Figure 3-Dimages
6. 6.3-D imagesofof the model
modelpredicted
predictedriver at at
river hour 41: 41:
hour (a) water depth
(a) water (mm);(mm);
depth (b) Froude number;
(b) Froude
(c) bed shear stress (N/m 2 ); (d) bed load concentration (g/(m · s)); and (e) bed median grain size (mm).
number; (c) bed shear stress (N/m ); (d) bed load concentration (g/(m·s)); and (e) bed median grain
2

size (mm).
3.2.3. Sensitivity Analysis
3.2.3. Sensitivity Analysis
A sensitivity analysis was performed to investigate the effects of secondary flow and grid size
on A thesensitivity analysis was
model performance. performed
Without to investigate
considering theof
the effect effects of secondary
secondary flow 7a),
flow (Figure andagrid
riversize
with
onalternating
the model performance.
pool-bar unitsWithout considering
also formed, the effectpattern
and a braided of secondary flow (Figure
was generated 7a), a41.
at hour river
Thiswith
river
alternating pool-bar
shared a similar units pattern
channel also formed, and
with the a braided
river pattern
considering thewas
effectgenerated
of secondaryat hour
flow41. This 7b),
(Figure riveryet
shared a similar
it generated channel
fewer patternchannels
but wider with thewith
rivera considering the effect rate,
slower development of secondary
illustratingflow
the(Figure 7b),the
role that
yetsecondary
it generatedflowfewer
playsbut wider channels
in transverse withtransport
sediment a slower and
development rate, illustrating the role that
channel migration.
the secondary
Figure 8flow
showsplays
thein transverse
braiding sediment transport
configurations with gridsand
of channel
(a) 0.03 × 0.03 m2 , (b) 0.025 × 0.025 m2 ,
migration.
andFigure 8 shows
(c) 0.02 m2 at
× 0.02 the braiding
hour 27. configurations
Compared with with
thegrids
riverofwith
(a) 0.03
a cell× size
0.03of m0.02 × 0.02 m2 , them
2, (b) 0.025 × 0.025 2,
river
and
with(c) a0.02 × 0.02
lower m at hour
2
resolution 27. Compared
of 0.03 2 with the river
× 0.03 m produced fewer with
andawider
cell size of 0.02 ×In0.02
channels. m , thethe
2
contrast, river
river
with
with a lower
a middleresolution of 0.03of
cell resolution × 0.025
0.03 m×2 0.025 m2 generated
produced fewer anda similar
wider channels. In contrast,
braided pattern to thethe river
river with
with a middle
a cell size of cell
0.02resolution
× 0.02 m of2 0.025
. The × 0.025numbers
channel m generated
2 a similar
and widths of braided
these two pattern
riversto the river
were with
comparable
a cell
withsizeeachofother.
0.02 × 0.02 m2. The channel numbers and widths of these two rivers were comparable
with each other.
Water 2017, 9, 686 11 of 18
Water 2017,
Water 2017, 9,
9, 686
686 11 of
11 of 17
17

Figure 7.
Figure 7. Braiding
Braiding configurations (a)
Braiding configurations (a) without
without and
and (b)
(b) with
with the
the effect
effect of
of secondary
secondary flow.
flow.
flow.

Figure 8.
Figure 8. Braiding
Braiding configurations
configurations at
at hour
hour 27
27 with
with cell
cell sizes
sizes of
of (a)
(a) 0.03
0.03 ×× 0.03
0.03 m
m222;; (b)
(b) 0.025
0.025 ×× 0.025
0.025 m
m222;;
Figure 8. Braiding configurations at hour 27 with cell sizes of (a) 0.03 × 0.03 m ; (b) 0.025 × 0.025 m ;
and (c)
and (c) 0.02
0.02 ×× 0.02
0.02 m222..
and (c) 0.02 × 0.02m m .

3.3. Statistical
3.3. Statistical Characteristics
Characteristics
3.3. Statistical Characteristics
Due to
Due to the
the nature
nature of of braided
braided rivers
rivers with
with frequent
frequent channel
channel migrations,
migrations, statistical
statistical methods
methods areare
Due to the nature of braided rivers with frequent channel migrations, statistical methods are
usually employed
usually employed to to assess
assess the
the characteristics
characteristics of
of braided
braided rivers
rivers [55].
[55]. Given
Given that
that braided
braided rivers
rivers in
in
usually employed to assess the characteristics of braided rivers [55]. Given that braided rivers in
nature exhibit
nature exhibit some
exhibit some intrinsic
some intrinsic characteristics
intrinsic characteristics common
characteristics common
common to to all
to all braided
all braided rivers,
braided rivers, several
rivers, several methods
several methods
methods havehave been
have been
been
nature
proposed to
proposed to evaluate
evaluate the
the simulation
simulation capability
capability of
of physical
physical and
and numerical
numerical models
models (e.g.,
(e.g., [55,56]).
[55,56]). In
In the
the
proposed to evaluate the simulation capability of physical and numerical models (e.g., [55,56]). In the
present study,
present study, three
study, three statistical
three statistical methods,
statistical methods, including
methods, including braiding
including braiding intensity,
braiding intensity, state-space
intensity, state-space plots,
state-space plots, and
plots, and transect
and transect
transect
present
topography and
topography and slope
slope frequency,
frequency, were
were applied
applied to
to evaluate
evaluate the
the model-predicted
model-predicted river.
river.
topography and slope frequency, were applied to evaluate the model-predicted river.
3.3.1. Braiding
3.3.1. Braiding Intensity
Intensity
3.3.1. Braiding Intensity
Braiding intensity
Braiding intensity has
has been
been proposed
proposed to to measure
measure thethe channel intensity
intensity inin braided
braided rivers
rivers
Braiding intensity has been proposed to measure the channelchannel
intensity in braided rivers [54,57,58].
[54,57,58].
[54,57,58]. Total
Total braiding intensity
braiding(BI
intensity (BI T) (defined as the number of total wetted channels counted and
(BIT) (defined as the number of total wetted channels counted and
Total braiding intensity T ) (defined as the number of total wetted channels counted and averaged
averaged over
averaged over aa number
number of of cross
cross sections
sections of of aa river),
river), active
active braiding
braiding intensity
intensity (BI
(BIAA)) (defined
(defined asas the
the
over a number of cross sections of a river), active braiding intensity (BIA ) (defined as the number
number of
number of channels with
with visible
visible bed
bed load
load sediment
sediment transport
transport counted andand averaged over over aa number
number
of channelschannels
with visible bed load sediment transport counted counted
and averagedaveraged
over a number of cross
of cross sections of a river), and their ratio (BI T/BIA) were calculated (Figure 9) and compared with
of cross sections of a river), and their ratio (BIT/BIA) were calculated (Figure 9) and compared with
the experimental
the experimental results
results (Table
(Table 2).
2). In
In general,
general, the
the model
model successfully
successfully predicted
predicted the
the response
response ofof the
the
Water 2017, 9, 686 12 of 18

Water 2017, 9, 686 12 of 17

sections of a river), and their ratio (BIT /BIA ) were calculated (Figure 9) and compared with the
experimental results (Table 2). In general, the model successfully predicted the response of the channel
channel pattern to increasing discharge, with the total and active braiding intensities increased to
pattern to increasing discharge, with the total and active braiding intensities increased to new stages.
new stages. The maximum BIT values were greater than twice the maximum BIA values, which were
The maximum BIT values were greater than twice the maximum BIA values, which were similar to
similar to those observed in the experiments [31]. Compared with the experimental river, the model-
those observed in the experiments [31]. Compared with the experimental river, the model-predicted
predicted river developed nearly the same BIA and a slightly lower BIT in Experiment 7. However, in
river developed nearly the same BIA and a slightly lower BIT in Experiment 7. However, in Experiment
Experiment 8, the differences between the model-predicted and the measured BIA and BIT values
8, the differences between the model-predicted and the measured BIA and BIT values increased after
increased after hour 100. This is because several channels co-existed and the difference between the
hour 100. This is because several channels co-existed and the difference between the main channel and
main channel and other channels became less obvious.
other channels became less obvious.

Figure 9. Variations in total braiding intensity (BIT), active braiding intensity (BIA), and BIA/BIT during
Figure 9. Variations in total braiding intensity (BIT ), active braiding intensity (BIA ), and BIA /BIT
sequential flow stages.
during sequential flow stages.

Table 2. The braiding intensities of the model predicted river and the experimental river.
Table 2. The braiding intensities of the model predicted river and the experimental river.
Predicted River Experimental River
Parameters
Parameters
Predicted River Experimental River
BIA BIT BIA/BIT BIA BIT BIA/BIT
BIA BIT BIA /BIT BIA BIT BIA /BIT
Exp. 7 Exp.
1.1 7 1.1
2.6 2.6 0.35–0.5
0.35–0.5 1.1 1.1 2.8 0.3–0.5
2.8 0.3–0.5
Exp. 8 1.8 8
Exp. 4.5
1.8 4.5 0.35–0.5
0.35–0.5 1.3 1.3 3.8 3.8
0.3–0.5 0.3–0.5

In both flow stages, the active braiding intensity, BIA, developed quickly to a stable value (Figure
In both flow stages, the active braiding intensity, BI , developed quickly to a stable value (Figure 9).
9). In the first stage, it took slightly longer to stabilize, Abut, in the second stage, it took only a couple
In the first stage, it took slightly longer to stabilize, but, in the second stage, it took only a couple of
of hours. In contrast, BIT took a much longer time to increase progressively to a stable state
hours. In contrast, BI took a much longer time to increase progressively to a stable state (approximately
(approximately 40 h).T The quick stabilization of BIA was explained by Egozi and Ashmore [31] as
40 h). The quick stabilization of BIA was explained by Egozi and Ashmore [31] as follows: the increase
follows: the increase in BIA requires only a sufficiently large flow within an existing channel to
in BIA requires only a sufficiently large flow within an existing channel to mobilize the bed material,
mobilize the bed material, and this can be accomplished very quickly; however, the increase in BIT
and this can be accomplished very quickly; however, the increase in BIT needs time for the river to
needs time for the river to erode new channels, resulting in a much longer time for its stabilization.
erode new channels, resulting in a much longer time for its stabilization. The longer duration for
The longer duration for the stabilization of BIA and BIT during the first flow stage resulted from the
the stabilization of BI and BI during the first flow stage resulted from the fact that it took some
fact that it took someAtime to Tdevelop a braided pattern from the initially straight channel. The
time to develop a braided pattern from the initially straight channel. The variation in BIA /BIT shows
variation in BIA/BIT shows a consistent trend progressively approaching nearly 0.4 (Figure 9), which
a consistent trend progressively approaching nearly 0.4 (Figure 9), which shared similar stable values
shared similar stable values of BIA / BIT with the experimental river but with a slightly narrower range
of BI / BI with the experimental river but with a slightly narrower range (Table 2).
(TableA 2). T

3.3.2. State Space Plots

Natural braided rivers have shown an intertwining effect of braiding, represented by the
repeated division and joining of channels with bars between them [7]. Flow at one location affects the
downstream morphology, and braided rivers show recurring geometric and topographic
Water 2017, 9, 686 13 of 18

3.3.2. State Space Plots

Natural braided rivers have shown an intertwining effect of braiding, represented by the repeated
division
Water 2017, and
9, 686 joining of channels with bars between them [7]. Flow at one location affects 13 ofthe
17
downstream morphology, and braided rivers show recurring geometric and topographic characteristics.
characteristics.
A method named A ‘state-space
method named plot’ ‘state-space
has been employedplot’ has been employed
to evaluate the spatial to recurring
evaluate the spatial
patterns of
recurring patterns
braided rivers (e.g.,of[56]).
braided
The rivers (e.g., plot
state-space [56]).isThe state-space
constructed usingplota is constructed
series of valuesusing
of onea essential
series of
values
variableofby
one essential
plotting eachvariable by plotting
value versus each value
the previous versus
value at itsthe previous value
neighbouring at its neighbouring
upstream cross section,
upstream
and whencross section,
it moves fromand onewhen
pointittomoves from onepoints
the successive point and
to the successive
eventually points
forms and eventually
a closed loop, the
forms a closed
information onloop, theorder
spatial information on spatial
is recorded. order
The total is recorded.
channel widths, The total channel
sequential scourwidths,
depths sequential
at channel
scour depthsand
confluences, at channel
heights of confluences,
successive barsand heights
have been of proposed
successivetobars havethe
analyse been proposed
spatial repeatingto analyse
pattern
the spatial repeating
of braided pattern of braided rivers [21,55,59,60].
rivers [21,55,59,60].
In the present study, the maximum scour depth in each cross-section was chosen for state-space
plotting. With
With thethe data
data of of aa fully
fully developed
developed braided
braided river,
river, state-space
state-space plots plots were
were constructed
constructed by by
plotting the
the scour
scourdepth
depthversus
versus thethe preceding
preceding scourscour depth,
depth, bothboth normalised
normalised to thetooverall
the overall scour
scour depth.
depth.
Figure Figure
10 shows 10 shows
the statethespace
state space
plots plots
of theofpredicted
the predictedriverriver
and and a laboratory
a laboratory river
river [60].
[60]. Clusters
Clusters of
of loops
loops cancan
be be
seenseen in the
in the plotsplots of predicted
of the the predictedriverriver
and and the laboratory
the laboratory river,river, indicating
indicating thatshare
that they they
share a common
a common statistically
statistically spatialspatial
repeatingrepeating
pattern. pattern.
However, However, the predicted
the predicted river hasriver has a thinner
a thinner pattern
pattern
than thethan the laboratory
laboratory river, indicating
river, indicating that, in the that, in the laboratory
laboratory river, the
river, the channel channel
scour depthscour
changesdepth in
changes in a more
a more abrupt abrupt
way than waypredicted
in the than in the predicted
river. river.crossing
Nonetheless, Nonetheless, crossing
lines were morelines
commonwerefor morethe
common
laboratory forriver
the laboratory
than for the river than forriver.
predicted the predicted
They were river. They were
considered considered
to originate to originate
from from
some statistic
some statistic
behaviour behaviour
of the rivers [59].of the rivers [59].

Figure
Figure 10.
10. State-space
State-space plots
plots for
for (a)
(a) the
the model-predicted
model-predicted river
river and
and (b) a laboratory river.

3.3.3.
3.3.3. Transect
Transect Topography and Slope
Topography and Slope Frequency
Frequency
The
The riverbed
riverbed topography
topography structure
structure cancan be
be described
described byby cross-section
cross-section surveys
surveys such
such as
as the
the transect
transect
topography
topography and the cumulative frequency of lateral slopes [55]. The former provides direct transect
and the cumulative frequency of lateral slopes [55]. The former provides direct transect
distributions
distributions of of bed
bed topographic
topographic highshighs and
and lows.
lows. The
The latter
latter estimates
estimates the
the frequency
frequency of of bed
bed lateral
lateral
slopes
slopes relative to a reference level; for example, the median elevation. They can be employed to assess
relative to a reference level; for example, the median elevation. They can be employed to assess
the
the similarities
similarities between
between the
the simulated
simulated and and natural rivers.
natural rivers.
The
The predicted
predicted river
river with
with aa fully developed braided
fully developed braided pattern
pattern was
was analysed
analysed toto determine
determine the the
structure
structure of the transect topography. The results of two representative cross-sections are
of the transect topography. The results of two representative cross-sections are shown
shown in in
Figure 11,together
Figure 11, togetherwith with those
those from from the Sunwapta
the Sunwapta River River and a laboratory
and a laboratory river
river [55], with[55], with the
the Sunwapta
Sunwapta
River being River being the prototype
the prototype of the experimental
of the experimental river simulated
river simulated in the study.
in the present presentItstudy.
can beItseen
can
be seen that, in the cross-sections shown in Figure 11, the channels are relatively narrow
that, in the cross-sections shown in Figure 11, the channels are relatively narrow and deep, while the and deep,
while
bars aretherelatively
bars arewide relatively wide and
and gentle, with gentle, with thescour
the maximum maximum
depth ofscour depth of exceeding
the channels the channels the
exceeding the maximum deposition height of the bars. These are common
maximum deposition height of the bars. These are common characteristics for the model-predicted characteristics for the
model-predicted river and the prototype natural and laboratory rivers. However, compared to the
Sunwapta River and laboratory river, the predicted river presents a smoother cross-sectional bed
elevation. This indicates that the riverbed of the simulated river is more ideal than those of real rivers.
Water 2017, 9, 686 14 of 18

river and the prototype natural and laboratory rivers. However, compared to the Sunwapta River and
laboratory river, the predicted river presents a smoother cross-sectional bed elevation. This indicates
Water
that 2017,riverbed
the 9, 686 of the simulated river is more ideal than those of real rivers. 14 of 17
Water 2017, 9, 686 14 of 17

Figure
Figure 11. Deviations from
from the elevation
elevation median of cross-sections for
for (a)
(a) the
the model
model predicted
predicted river;
Figure 11. Deviations
11. Deviations from the
the elevation median
median of of cross-sections
cross-sections for (a) the model predicted river;
river;
(b)
(b) the
the Sunwapta
Sunwapta River;
River; and
and (c)
(c) a
a laboratory
laboratory river.
river. The
The dotted
dotted line
line represents
represents the
the elevation
elevation median.
median.
(b) the Sunwapta River; and (c) a laboratory river. The dotted line represents the elevation median.

The slope frequency distribution curves were constructed based on the cross-section slopes in
The slope frequency distribution curves were constructed based on the cross-section slopes in
the model predicted reach of 4 to 17 m, containing the data of 650 cross-sections. Figure 12 shows the
the model predicted reach of 4 to 17 m, containing the data of 650 cross-sections. Figure 12 shows the
results of the predicted river at hour 41, the Sunwapta River, and a laboratory river [55], all of which
results of the predicted river at hour 41, the Sunwapta River, and a laboratory river [55], all of which
present similar varying trends. The curves corresponding to the channels are significantly wider than
present similar varying trends. The curves corresponding to the channels are significantly wider than
the bars, illustrating that the bars exhibit statistically gentler elevation changes than do the channels.
the bars, illustrating that the bars exhibit statistically gentler elevation
elevation changes
changes than
than do
do the
the channels.
channels.
This is consistent with the findings from the bed-transect topographic distributions (Figure 11). The
This is consistent with the findings from the the bed-transect
bed-transect topographic
topographic distributions (Figure 11). The
steep sides in the channels were produced by fast flow and energetic erosion activity, while an
steep sides
sidesininthe
the channels
channels were
were produced
produced by flow
by fast fast and
flowenergetic
and energetic
erosionerosion
activity,activity,
while anwhile an
opposite
opposite situation occurred on the bars. In addition, the wider slope frequency domains of both the
opposite situation
situation occurred occurred on In
on the bars. theaddition,
bars. In addition,
the widerthe wider
slope slope frequency
frequency domains of domains
both theof channels
both the
channels and the bars for the predicted river indicate that statistically the predicted river channels
channels
and andfor
the bars thethe
bars for the river
predicted predicted river
indicate indicate
that that statistically
statistically the predictedtheriver
predicted river
channels arechannels
slightly
are slightly steeper than those of the Sunwapta River and the laboratory river.
are slightly steeper than those of the Sunwapta River and
steeper than those of the Sunwapta River and the laboratory river. the laboratory river.

Figure 12. Cumulative distributions of the lateral slopes of the bars and channels for (a) the model
Figure 12. Cumulative
Cumulative distributions
distributions of
of the
the lateral slopes
slopes of the
the bars
bars and
and channels
channels for (a) the
the model
model
Figure 12.
predicted river; (b) the Sunwapta River; andlateral
(c) a laboratory of river. for (a)
predicted river;
predicted river; (b)
(b) the
the Sunwapta
Sunwapta River;
River; and
and (c)
(c) aa laboratory
laboratory river.
river.

4. Discussion and Conclusions

4. Discussion and Conclusions
A physics-based numerical model was developed to simulate the morphodynamic processes in
A physics-based numerical model was developed to simulate the morphodynamic processes in
laboratory braided rivers with non-uniform bed load transport. The model applied the basic
laboratory braided rivers with non-uniform bed load transport. The model applied the basic
hydrodynamic and sediment transport principles with a fractional method and multiple bed layers.
hydrodynamic and sediment transport principles with a fractional method and multiple bed layers.
It employed the TVD-MacCormack Scheme for solving the hydrodynamic equations of trans-critical
It employed the TVD-MacCormack Scheme for solving the hydrodynamic equations of trans-critical
flows, and was tested with a sediment aggradation case and found to be capable of effectively
flows, and was tested with a sediment aggradation case and found to be capable of effectively
Water 2017, 9, 686 15 of 18

4. Discussion and Conclusions

A physics-based numerical model was developed to simulate the morphodynamic processes
in laboratory braided rivers with non-uniform bed load transport. The model applied the basic
hydrodynamic and sediment transport principles with a fractional method and multiple bed layers.
It employed the TVD-MacCormack Scheme for solving the hydrodynamic equations of trans-critical
flows, and was tested with a sediment aggradation case and found to be capable of effectively
predicting the flow and bed deformation.
With the same boundary conditions as an experimental river, the model successfully reproduced
the evolution processes and phenomena of braiding, avulsion activities, and the responses of the
channel pattern to increased discharge. The incorporation of the effect of secondary flow played
an essential role in producing a more realistic braided pattern. Certain geomorphologic phenomena
and processes observed in the experimental river and natural rivers have been found in the evolutionary
process of the predicted river. A cycle of straight channels transitioning to sinuous channels and then
to avulsions was an important morphodynamic process for the river to maintain a dynamic braided
form. Avulsion activities with similar evolution processes to natural rivers have been found in both
the experimental and predicted rivers. A high shear stress usually occurred in areas where the flow
was energetic, erosion was possible, and the riverbed sediment was mainly composed of coarse bed
material. The changes in the braiding intensities illustrated trends similar to those in the experimental
river in the process of braiding generation and in response to increased discharge. Active braiding
intensity developed quickly to a stable value, whereas total braiding intensity increased to a stable state
gradually due to the time consumed by cutting new channels. The predicted river shared common
statistical topographic characteristics with natural and laboratory rivers, showing a recurring spatial
pattern of shallow and deep topography and statistical transect gentle bars and steep channels.
Generally, the results presented above demonstrate the potential for the present model to represent
the morphodynamic processes of natural braided rivers. Nevertheless, it is necessary to gradually
improve the understanding and capability of the model in conjunction with field research and
laboratory experiments.

Acknowledgments: We are grateful for the financial support from the State Key Laboratory of Hydroscience and
Engineering, Tsinghua University (2015-ky-2), the Guangzhou Science Technology and Innovation Commission
(201605030009), and the South China Agricultural University (7600-217244). We thank Roey Egozi and Ana Maria
A.F. da Silva for their comments and suggestions on the present research.
Author Contributions: Haiyan Yang and Binliang Lin conceived of the braided river simulation research and
developed the physics-based model; Haiyan, Jian Sun and Guoxian Huang set up and debugged the programmes;
Haiyan Yang analysed the results and wrote the paper; Haiyan Yang and Binliang Lin provided editorial
improvements to the paper.
Conflicts of Interest: The authors declare no conflicts of interest. The founding sponsors had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; nor in the
decision to publish the results.

References
1. Bertoldi, W.; Gurnell, A.; Surian, N.; Tockner, K.; Zanoni, L.; Ziliani, L.; Zolezzi, G. Understanding reference
processes: Linkages between river flows, sediment dynamics and vegetated landforms along the Tagliamento
River, Italy. River Res. Appl. 2009, 25, 501–516. [CrossRef]
2. Knighton, D.A. Fluvial Forms and Processes: A New Perspective; Arnold, Hodder Headline, PLC: London, UK,
1998; p. 383.
3. Ham, D.G.; Church, M. Bed-material transport estimated from channel morphodynamics: Chilliwack River,
British Columbia. Earth Surf. Process. Landf. 2000, 25, 1123–1142. [CrossRef]
4. Smith, N.D.; Smith, D.G. William River: An outstanding example of channel widening and braiding caused
by bed-load addition. Geology 1984, 12, 78. [CrossRef]
5. Brierley, G.J.; Hickin, E.J. Channel planform as a non-controlling factor in fluvial sedimentology: The case of
the Squamish River floodplain, British Columbia. Sediment. Geol. 1991, 75, 67–83. [CrossRef]
Water 2017, 9, 686 16 of 18

6. Huang, H.Q.; Chang, H.H.; Nanson, G.C. Minimum energy as the general form of critical flow and maximum
flow efficiency and for explaining variations in river channel pattern. Water Resour. Res. 2004, 40, W04502.
[CrossRef]
7. Bridge, J.S. The interaction between channel geometry, water flow, sediment transport and deposition
in braided rivers. In Braided Rivers; Best, J.L., Bristow, C.S., Eds.; Geological Society: London, UK, 1993;
pp. 13–71.
8. Cui, Y.; Paola, C.; Parker, G. Numerical simulation of aggradation and downstream fining. J. Hydraul. Res.
1996, 34, 185–204. [CrossRef]
9. Duc, B.M.; Wenka, T.; Rodi, W. Numerical modeling of bed deformation in laboratory channels.
J. Hydraul. Eng. 2004, 130, 894–904. [CrossRef]
10. Meselhe, E.A.; Georgiou, I.; Allison, M.A.; McCorquodale, J.A. Numerical modeling of hydrodynamics and
sediment transport in lower Mississippi at a proposed delta building diversion. J. Hydrol. 2012, 472, 340–354.
[CrossRef]
11. Bai, Y.; Duan, J.G. Simulating unsteady flow and sediment transport in vegetated channel network. J. Hydrol.
2014, 515, 90–102. [CrossRef]
12. Murray, A.B.; Paola, C. A cellular model of braided rivers. Nature 1994, 371, 54–57. [CrossRef]
13. Coulthard, T.J.; Hicks, D.M.; Van De Wiel, M.J. Cellular modelling of river catchments and reaches:
Advantages, limitations and prospects. Geomorphology 2007, 90, 192–207. [CrossRef]
14. Schuurman, F.; Marra, W.A.; Kleinhans, M.G. Physics-based modeling of large braided sand-bed rivers: Bar
pattern formation, dynamics, and sensitivity. J. Geophys. Res. 2013, 118, 2509–2527. [CrossRef]
15. Slingerland, R.; Smith, N.D. River avulsions and their deposits. Annu. Rev. Earth Planet. Sci. 2004, 32,
257–285. [CrossRef]
16. Jang, C.L.; Shimizu, Y. Numerical simulations of the behavior of alternate bars with different bank strengths.
J. Hydraul. Res. 2005, 43, 596–612. [CrossRef]
17. Jang, C.L.; Shimizu, Y. Numerical simulation of relatively wide, shallow channels with erodible banks.
J. Hydraul. Eng. 2005, 131, 565–575. [CrossRef]
18. Nicholas, A.P. Modelling the continuum of river channel patterns. Earth Surf. Process. Landf. 2013, 38,
1187–1196. [CrossRef]
19. Schuurman, F.; Kleinhans, M.G. Bar dynamics and bifurcation evolution in a modelled braided sand-bed
river. Earth Surf. Process. Landf. 2015, 40, 1318–1333. [CrossRef]
20. Yang, H. Development of a Physics-Based Morphodynamic Model and Its Application to Braided Rivers.
Ph.D. Thesis, Cardiff University, Cardiff, UK, 2013.
21. Yang, H.; Lin, B.; Zhou, J. Physics-based numerical modelling of large braided rivers dominated by suspended
sediment. Hydrol. Process. 2015, 29, 1925–1941. [CrossRef]
22. Kleinhans, M.G. Sorting out river channel patterns. Prog. Phys. Geogr. 2010, 34, 287–326. [CrossRef]
23. Takebayashi, H.; Okabe, T. Numerical modelling of braided streams in unsteady flow. Proc. ICE Water Manag.
2009, 162, 189–198. [CrossRef]
24. Lotsari, E.; Wainwright, D.; Corner, G.D.; Alho, P.; Kayhko, J. Surveyed and modelled one-year
morphodynamics in the braided lower Tana River. Hydrol. Process. 2014, 28, 2685–2716. [CrossRef]
25. Schuurman, F.; Kleinhans, M.G. Self-Formed Braid Bars in a Numerical Model; American Geophysical Union
Fall Meeting Abstracts: Washington, DC, USA, 2011; p. 672.
26. Ikeda, S.; Parker, G.; Sawai, K. Bend theory of river meanders, part 1: Linear development. J. Fluid Mech.
1981, 112, 363–377. [CrossRef]
27. Parker, G.; Sawai, K.; Ikeda, S. Bend theory of river meanders, part 2: Nonlinear deformation of
Finite-amplitude bends. J. Fluid Mech. 1982, 115, 303–314. [CrossRef]
28. Kleinhans, M.G.; van den Berg, J.H. River channel and bar patterns explained and predicted by an empirical
and a physics-based method. Earth Surf. Process. Landf. 2011, 36, 721–738. [CrossRef]
29. Wu, W. Computational River Dynamics; Taylor & Francis Group: London, UK, 2007; p. 499.
30. Ferguson, R.I.; Ashmore, P.E.; Ashworth, P.J.; Paola, C.; Prestegaard, K.L. Measurements in a braided river
chute and lobe: 1. Flow pattern, sediment transport, and channel change. Water Resour. Res. 1992, 28,
1877–1886. [CrossRef]
31. Egozi, R.; Ashmore, P.E. Experimental analysis of braided channel pattern response to increased discharge.
J. Geophys. Res. 2009, 114. [CrossRef]
Water 2017, 9, 686 17 of 18

32. Leduc, P.; Ashmore, P.; Gardner, J.T. Grain sorting in the morphological active layer of a braided river
physical model. Earth Surf. Dyn. 2015, 3, 577.
33. Van Niekerk, A.; Vogel, K.R.; Slingerland, R.L.; Bridge, J.S. Routing of heterogeneous sediments over movable
bed: Model development. J. Hydraul. Eng. 1992, 118, 246–262. [CrossRef]
34. Van Rijn, L.C. Sediment transport, part III: Bed forms and alluvial roughness. J. Hydraul. Eng. 1984, 110,
1733–1754. [CrossRef]
35. Liang, D.; Falconer, R.A.; Lin, B. Comparison between TVD-MacCormack and ADI-type solvers of the
shallow water equations. Adv. Water Resour. 2006, 29, 1833–1845. [CrossRef]
36. Liang, D.; Lin, B.; Falconer, R.A. Simulation of rapidly varying flow using an efficient TVD-MacCormack
scheme. Int. J. Numer. Methods Fluids 2007, 53, 811–826. [CrossRef]
37. Van Rijn, L.C. Sediment transport, part I: Bed load transport. J. Hydraul. Eng. 1984, 110, 1431–1456. [CrossRef]
38. Van Rijn, L.C. Principles of Sediment Transport in Rivers, Estuaries and Coastal Seas; Aqua Publications:
Amsterdam, The Netherlands, 1993.
39. Engelund, F. Flow and bed topography in channel bends. J. Hydraul. Div 1974, 100, 1631–1648.
40. Karim, F. Bed material discharge prediction for nonuniform bed sediments. J. Hydraul. Eng. 1998, 124,
597–604. [CrossRef]
41. Van Rijn, L.C. Unified view of sediment transport by currents and waves III: Graded beds. J. Hydraul. Eng.
2007, 133, 761–775. [CrossRef]
42. Wei, Z.L. One Dimensional Sediment Model for the Yellow River; Report of Wuhan University of Hydro-electrical
Power (In Chinese). Wuhan University: Wuhan, China, 1993.
43. Zhou, J.; Lin, B. Modelling Bed Evolution Processes in Alluvial Rivers. In Proceedings of the International
Conference on Fluvial Hydraulics, Izmir, Turkey, 3–5 September 2008; pp. 1305–1311.
44. Lin, B.; Falconer, R.A. Tidal flow and transport modeling using Ultimate Quickest scheme. J. Hydraul. Eng.
1997, 123, 303–314. [CrossRef]
45. Sun, J.; Lin, B.; Yang, H. Development and application of a braided river model with non-uniform sediment
transport. Adv. Water Resour. 2015, 81, 62–74. [CrossRef]
46. Seal, R.; Paola, C.; Parker, G.; Southard, J.B.; Wilcock, P.R. Experiments on downstream fining of gravel: I.
Narrow-channel runs. J. Hydraul. Eng. 1997, 123, 874–884. [CrossRef]
47. Paola, C.; Parker, G.; Seal, R.; Sinha, S.K.; Southard, J.B.; Wilcock, P.R. Downstream fining by selective
deposition in a laboratory flume. Science 1992, 258, 1757–1760. [CrossRef] [PubMed]
48. Ashmore, P.E. Channel morphology and bed load pulses in braided, gravel-bed streams. Geogr. Ann. Ser. A
1991, 73, 37–52. [CrossRef]
49. Hoey, T.; Sutherland, A.J. Channel morphology and bedload pulses in braided rivers: A laboratory study.
Earth Surf. Process. Landf. 1991, 16, 447–462. [CrossRef]
50. Shvidchenko, A.B.; Kopaliani, Z.D. Hydraulic modeling bed load transport in gravel bed Laba River.
J. Hydraul. Eng. 1998, 124, 778–785. [CrossRef]
51. Warburton, J.; Davies, T.R.H. Use of hydraulic models in management of braided gravel-bed rivers.
In Gravel-Bed Rivers in the Environment; Klingeman, P.C., Beschta, R.L., Komar, P.D., Bradley, J.B., Eds.;
Water Resources Publications: Littleton, CO, USA, 1998; pp. 512–542.
52. Egozi, R.; Ashmore, P.E. Defining and measuring braiding intensity. Earth Surf. Process. Landf. 2008, 33,
2121–2138. [CrossRef]
53. Ashmore, P.E. Anabranch confluence kinetics and sedimentation processes in gravel-braided streams.
In Braided Rivers; Best, J.L., Bristow, C.S., Eds.; Geological Society, Special Publications: London, UK, 1993;
Volume 75, pp. 129–146.
54. Ashmore, P.; Bertoldi, W.; Tobias Gardner, J. Active width of gravel-bed braided rivers. Earth Surf. Process.
Landf. 2011, 36, 1510–1521. [CrossRef]
55. Doeschl, A.B.; Ashmore, P.E.; Davison, M. Methods for Assessing Exploratory Computational Models of
Braided Rivers. In Braided Rivers: Process, Deposits, Ecology and Management; Sambrook-Smith, G.H., Best, J.L.,
Bristow, C.S., Petts, G.E., Jarvis, I., Eds.; Wiley-Blackwell: Malden, MA, USA, 2006; pp. 177–197.
56. Murray, A.B.; Paola, C. A new quantitative test of geomorphic models, applied to a model of braided streams.
Water Resour. Res. 1996, 32, 2579–2587. [CrossRef]
57. Ashmore, P.E. How do gravel-bed rivers braid? Can. J. Earth Sci. 1991, 28, 326–341. [CrossRef]
Water 2017, 9, 686 18 of 18

58. Bertoldi, W.; Zanoni, L.; Tubino, M. Planform dynamics of braided streams. Earth Surf. Process. Landf. 2009,
34, 547–557. [CrossRef]
59. Sapozhnikov, V.B.; Murray, A.B.; Paola, C.; Foufoula-Georgiou, E. Validation of Braided-Stream Models:
Spatial state-space plots, self-affine scaling, and island shapes. Water Resour. Res. 1998, 34, 2353–2364.
[CrossRef]
60. Rosatti, G. Validation of the physical modeling approach for braided rivers. Water Resour. Res. 2002, 38, 1295.
[CrossRef]

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).