Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 170 (2017) 202 – 209

Engineering Physics International Conference, EPIC 2016

Numerical Simulation of Sediment Transport along a Channel with

Underwater Sill
Nurul Azizaha,*, Radianta Triatmadjaa, Bambang Agus Kironotoa
a
Department of Civil and Environmental Engineering, Universitas Gadjah Mada, Indonesia

Abstract

Sediment prevention structures such as underwater sill (UWS) may be used to divert the sediment movement, and prevent it to enter the port
basin or navigation channel. In this study, a quasi 3D-numerical model called multilayered model based on the Method of Characteristics was
utilized to simulate the flow, and a particle tracking method was developed to simulate the sediment transport around an UWS. A straight channel
with an UWS approximately a half way between upstream and downstream boundaries was simulated. The height of the UWS was varied to
observe the effectiveness of such structure against sedimentation. The influence of UWS on flow and sedimentation pattern, vertical distribution
of suspended sediment concentration, and the implication of effectiveness of underwater sill structures to control the flow and reduce
sedimentation were analyzed. The numerical simulations indicated that the UWS significantly altered the flow/current pattern. It presence
contributes in reducing the sediment transport that cause siltation problem in port basin.
© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2016 The Authors. Published by Elsvier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the Engineering Physics International Conference 2016
Keywords: Sediment transport, Underwater sill (UWS), Method of Characteristics, Particle tracking method, Numerical Simulation;

1. Introduction

The siltation caused by sediment process is commonly found in many harbor and port over the world. Due to siltation, the
harbor basins and navigational channels required frequent maintenance dredging to guarantee safe or minimum depth. The costs
associated with these maintaining activities are quite high and are qualified as the most expensive item in running costs of harbors.
In order to prevent and to reduce such problem, many attempt or solution have been made and proposed. One of them is hard
structure prevention such as underwater sill or submerged dike.
Unlike the breakwater structure that was fully hindered the sediment from entering the port basin or navigational channel,
underwater sill is an underwater structure that is used to deflect the flow of sediment in bottom section of water depth, whilst the
sediment that are transported in the upper section are passing through, and will be deposited or not in the basin depend on the
capacity of water flow in carrying sediment. Such structure has been used in Kumamoto port, Japan, and in Port of PT. Semen
Gresik (Persero) Tbk Tuban, located at North Coast of Java Island, Indonesia.
The most relevant processes in the deposition and erosion of the channel or basin zone of a port are advection of sediment
particles by the longitudinal, transversal and vertical fluid velocities, mixing of sediment particles by turbulent and orbital motions,
settling of the particles due to gravity (fall velocity), erosion of particle from bed by current and wave-induced bed-shear stresses
[1]. The deposition and erosion in port area where the existence of structure play an important role can be predicted by numerically
simulate those relevant process. Various model have been purpose and developed using different type of approaching method.
Tsuruya et al. [2] developed a multi-layered model to predict the mud transport in the area of Kumamoto port by using the
finite different method. Multi-layered model is adopted, in aim to reproduce vertical distribution of suspended mud concentration,
and to observe the effect of submerged dikes. Based on this research, calculation for the Kumamoto port design were conducted,

* Corresponding author. Tel.:+62-821-3621-1052.

E-mail address: nururu.azizah@gmail.com

1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the Engineering Physics International Conference 2016
doi:10.1016/j.proeng.2017.03.048
 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209 203

and the result showed that the amount of decrease in volume deposition caused by the existence of submerged dike was around
30%. It is considered that the submerged dike has a considerable effect on reducing the deposition of sediment.
The research about effect of submerged dike was also done by Chai at. Al [3] by using a 2D numerical modeling of mud
transport. The numerical result indicate that submerged dike can reduce the concentration of mud passing the interesting point
significantly. The effect of flow velocity, water depth, and dike height on mud transport ware also investigated.
Other the finite different method, particle-based modelling is one of the method that can be used to simulate sediment transport,
especially suspended type. In other to solve the equation of suspended particles transport, a number of research has been made
using probabilistic or stochastic approaches. Indeed, the suspended particle movement has been regarded as an intrinsically
random process. Dimou and Adams[4] describe a two dimensional random walk particle-based model to simulate transport
(advection-dispersion) in vertically well-mixes estuaries and coastal waters. Applied to a channel of uniform cross section and
spatially variable dispersion coefficient and simulated average longitudinal concentration distribution.
Also using the random walk particle-based model, Argall et.al [5] simulate transport of non-cohesive sediment in rivers and
coastal zone where vertical distribution of mean velocity can be inferred under the assumption of a fully developed boundary
layer. The prediction of vertical sediment distribution follow the trend of Rouse profile, using roughly 100 particles per unit
volume of interest. The model then was applied to predict the settling process in Ballona Creek, a flood protection channels that
drains the Los Angeles basin. Using a range of grain size diameter in same time of simulation as an input, the result of this research
showed well predicted of the sediment settling. Oh [6] purposed a stochastic modelling using a Wiener process to express the
particle diffusion. Latter then he developed the stochastic diffusion jump model of suspended particle for extreme flow by
including the effect of buoyancy-type force due to flow acceleration, drag force, and the added mass.
This paper describes a multilayered model based on a particle-based model called particle tracking method for simulation of
sediment transport around underwater sill structure. In this study, the effect of UWS on sedimentation pattern, vertical distribution
of suspended sediment concentration, as well as the effectiveness of underwater sill structures to control the flow and reduce
sedimentation were analyzed.

2. Theoretical approach of suspended sediment transport model

There are a few factors that bring about variations in the motion of sediment particles. To begin with, the variety of particle
properties leads to the variations of the particle movement. Thus, particles that are transported in surface flows can vary from i.e.
wash load, suspended load, and bed load. Sediment transport concerns the movement of particles by flowing waters. Instead of
describing the concentration of the substance at various fixed points it is theoretically possible to track the motion of all the
particles of the substance as they are carried out by the fluid. All the particles are tracked continuously, into which the suspended
can be discretized, in order to compute the local concentration

2.1. The settling velocity

The settling velocity of sediment (ws) is the most fundamental property and considered as the primary contributor of particle
movement. It determine whether the particle was transported as suspended load or bed load. The settling velocity can be calculated
if the characteristic of the particle and fluid are known. It is a function of both density of fluid and sediment, the fluid viscosity,
volume and shape of sediment particle.
A convenient equation of estimating ws is given by Julien [7],

8Q ª
ws
d «¬
1  0.0139d   1º»¼
*
3
(1)

Where d* is a dimensionless of particle size given by

1/3
ª ( SG  1) g º
d* « Q2 » d (2)
¬ ¼

2.2. Vertical Sediment Distribution

In a steady, uniform turbulent flow the vertical distribution of suspended sediment is describe by the following equation,

dc
Hz  ws c 0 (3)
dz

Where the vertical diffusivity H z can be derived from logarithmic velocity distribution, as follow
204 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209

§ z·
Hz EN u* z ¨1  ¸ (4)
© h ¹

E is a constant ratio, usually the value is equal to 1, N is the von Kármán constant being typically 0.41, and u* is the shear stress
velocity.

Rouse then solved the Eq. (3) using value of vertical diffusivity as describe in Eq. (4) to give the better known formula of
vertical sediment distribution profile (the Rouse profile)

Ro
§ a (h  z ) ·
c( z ) ca ¨ ¸ (5)
© z (h  a ) ¹

Where c( z ) is the sediment concentration in z (the interesting point), ca represent the sediment concentration in a distance a
above the bed, h is the water depth, and Ro is the Rouse number, expressed as follow

ws
Ro (6)
EN u*

Figure 1 shows distribution profile of sediment based on Eq.14 for a range of Rouse Number.

1
0.9
0.8
0.7
Ro =1/16
(z-a)/(h-a)

0.6 1/8
0.5
1/4
0.4
1/2
0.3
1
0.2
2
0.1 4
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
C/Ca

Fig. 1. Rouse profile, a = 0.05h.

2.3. Particle tracking model

The sediment transport is controlled by the physical and hydrodynamic processes of advection and dispersion. The
hydrodynamic dispersion is a result of turbulent diffusion combined with motion or transport associated with the variation of fluid
velocity distribution. The particle tracking model adopting the Lagrangian approach is capable to simulate those processes
(Hydrodynamic advection and dispersion). Advection is accomplished by the local fluid velocity. From the coordinates of the
tracked particle, local fluid velocity is computed (possibly interpolated from neighboring locations). This fluid velocity was
simulated using the hydrodynamic model, therefore the effect of dispersion caused by variation of fluid velocity was included.
Meanwhile diffusion was simulated adopting stochastic approach where movement of particle sediments caused by turbulent
diffusion are describe with random motion using Brownian concept.
The governing equations of stochastic diffusion in a form of differential equations can be written as follow

dX t u (t , X t )dt  V (t , Xt )dBt (7)

Where, X t is the trajectory of particle, u (t , X t ) is the drift velocity, V (t , X t ) is diffusion coefficient tensor, and Bt is Brownian
motion or sometimes can be called as a Wiener process mathematically. Using equivalence between the Fokker-Planck equation
and the stochastic Langevin equation, Eq. (7) can be rewritten as [8]

§ wH i ·
dX i ¨ U i  wx ¸ dt  2H i dBt (8)
© i ¹
 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209 205

The relationship between the diffusion coefficient and turbulent diffusivities is shown as

1 2
Hi Vi (9)
2
And

wH i
u (t , X t ) U i  (10)
wxi

Where U i is the mean fluid velocity, and H i is turbulent diffusivities. Both in i-direction (longitudinal, transverse, and vertical,
respectively).

The numerical discretization of particle tracking equation (Eq. (8)) using an explicit scheme for particle position, can be written
as follows,

§ wH x ·
X p (t  't ) Xp (t )  ¨ U ( X p , Yp , Z p , t ) 
wx ¸ 't  2H x 'B(t ) (11)
© ¹

§ wH y ·
Yp (t  't ) Xp (t )  ¨ V ( X p , Yp , Z p , t )  ¸ 't  2H y 'B(t ) (12)
© wy ¹

§ wH z ·
Z p (t  't ) Yp (t )  ¨ W ( X p , Yp , Z p , t )  ws p 
wz ¸ 't  2H z 'B(t ) (13)
© ¹

With 'B (t ) N (0,1) 't

Where Xp, Yp, Zp represent the position of the pth particle, 't is the time step, 'B (t ) is the Gaussian distributed random variable
that present a Wiener process, and N(0.1) represent a random number with a mean zero and a standard deviation of one.
In order to simulate the particle sediment transport, this particle tracking equation require both initial condition and boundary
condition. The initial condition are particle coordinate at time t = 0, and the fluid velocity (computed previously using the
hydrodynamic model). Meanwhile the boundary condition such as water surface and wall are enforced using a reflection
procedure. Any particle predicted to move across the boundary are re-positioned in equidistant from the boundary (inside domain).
For the bed, particles are reflected off the bed when the shear stress exceeds the critical value for settling defined by The Shields
parameter and sticks to bed otherwise.

2.4. Sediment Concentration

The Sediment concentration can be estimated from the number of particles in a finite volume of interest (finite fixed grid point),
as well as particles properties (particle diameter).
3
3 S (d p
)n U
4
C (14)
WLD

Where n is the number of particles within a finite grid, and W,L,D is the width, length, and height of volume grid.

3. Methods

In other to simulate the sediment transport, a quasi 3D multi-layered model based on particle tracking method were utilized
where the hydrodynamic of the flow is simulated separately based on characteristic method which was developed by Triatmadja
[7].
Two model of sediment transport simulation were done within this research. The first one was a 2D model with a uniform
turbulent flow. This simulation is designed to characterize how well the particle tracking model resolved the vertical distribution
of sediment concentration over the range of flow conditions. Meanwhile the second model is the quasi-3D multi-layered model.
The main purpose of this model is to observe the sediment transport affected by underwater sill structure.
For the 2D model, uniform flow at a depth of 1 m, which modeled using length of 50 m and periodic boundary condition. That
is, the particles transported past the downstream boundary were repositioned at the upstream boundary at the same vertical position
relative to bed level. Sediment particle was initially released with a uniform distribution over the length and depth of the channel,
206 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209

and the particle tracking equation was integrated until the particle distribution reach equilibrium condition. For the quasi 3D multi-
layered model, the flow was obtained from the computation of hydrodynamic model. The sediment particles were released in
upstream continuously, with a uniform distribution over the depth. The particle tracking model was used to estimate the fate of
sediment moving along the channel with an underwater sill structure.
The model layout for computational domain that are a straight channel with an UWS approximately a half way between
upstream and downstream boundaries, are shown in Figure 2. A total of 50x20 grid calculated point (per-layer) and 9 layer were
used in this simulation. The rectangular shape underwater sill was used for the height variation of 1/7 to 4/7 from water depth.
Meanwhile the length and width were fixed. For sediment model, the diameter of particle were varied and were presented as a
dimensionless parameter called Rouse Number (Eq. (6)). Variation running scenario were summarized in table 1.

A) 50 x Dx

20 x Dy
Luws

Buws
A A
B)
1st Layer

huws
k-th Layer

Fig. 2. (A) Model Layout for computational domain, (B) Section A-A.

Table 1. Running scenario

Cases UWS Height (huws) Rouse Number (Ro)

Scenario-1 - 0.8, 1.2, 1.8
Scenario-2 1/7 h, 2/7 h, 3/7 h, 4/7 h 0.8, 1.2, 1.8

4. Results and Discussion

4.1. Prediction of Vertical Sediment profiles using 2D model

For this 2D model, particles initially were distributed uniformly over length and depth of the channel, and the particle tracking
equations were integrated until the particle distribution reaches equilibrium. A reflecting boundary condition was used in the bed,
as if the transport and settling were modeled. Sediment concentration were obtained by counting the number of particle located
within the calculated grid point, and these are compared to the distribution described by Rouse profile in Eq.(6). Figure 3 shows
result of suspended sediment concentration distribution of time step (dt) equal to 1 second, for a range of Rouse number (Ro= 0.3,
0.5, 1.0, and 2.0), compare to Rouse profile. The sediment concentration is estimated using Eq. (15). Range of Rouse number
represent the variation of sediment and flow parameter. In this simulation, 50 particles were released in every grid point, and the
channel consist a total of 50x20 grid point, which mean the total particle used in this simulation are 50.000 particles. These
indicated that using 50 particles per grid point capture the basic trend in vertical sediment concentration profile.

Fig. 3. Comparison of Suspended Sediment Concentration prediction – Rouse and Particle Tracking (dt = 1 s).
 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209 207

However, for larger time step the trend in the particle tracking prediction start to change and diverge, but generally still follow
the trend of Rouse profile. Therefore, for accuracy purposes the time step needs to be constrained. By making predictions with the
range of time step and Rouse number, it was found that the trend in the vertical profile was captured within 35% when time step
is roughly constrained by ws 't / h  0.05 .

4.2. The quasi-3D Particle Tracking model

For this model, the particle tracking equations describe in Eq. (11), (12), (13) were applied to characterize transport sediment
along a large channel that was discretized with 50 x 20 x 9 computational grid. An UWS was placed approximately a half way
between upstream and downstream boundaries, where the water level specified to be mean sea level. The sediment particles
entering the upstream boundary of the model were assumed as a continuous source that were distributed uniformly over depth and
width of the channel. A range of particle diameter that was used, presented in non-dimensional parameter called Rouse number
with the value of 0.8, 1.2, and 1.8. Each simulation was done for 10000 second, with consideration of particles that were released
in the upstream, already reached the downstream boundary.
The distribution of sediment particle concentration in both longitudinal and transversal (x-y plane) in each layer was calculated.
Figure 4 shows the calculated distribution of sediment particle concentration in certain layers for running scenario 1 of Rouse
number of 1.3, where no UWS was installed. In figure 5, the distributions are expressed in a dimensionless form of particle
concentration ( C * ), that were defined as:

1
C* C
CM

Where C is the particle concentration in the local grid calculated using Eq.(14), and CM is the average number of particle
concentration that were released per time step per unit volume of interest.

Fig. 4. Simulation result of distribution of sediment concentration, C*(x,y), for Rouse number of 1.3, without UWS (Scenario 1).

4.3. Vertical distribution of sediment

Vertical distribution profile of sediment was depend on the parameter of sediment itself such as diameter, as well as the flow
condition. This two term are accommodate in a dimensionless parameter called Rouse number. Different value of Rouse number
present the different type of vertical distribution profile. As explained before, the existence of UWS generally affect the proses of
sediment transport. Therefore in other to know how it affected the siltation proses in basin, the vertical distribution of sediment
need to be observed. The simulations were done by using a range of Rouse number, and various of UWS height.
The calculated vertical distribution obtained from simulation of scenario-1 (without UWS) for a Rouse number of 1.2, are
shown in figure 5. The simulation result shows that the concentration at the bottom layer in basin area are greater than in the
outside. This, due to the reduction of sediment transport capacity caused by the smaller current velocity that passing through the
basin area. In the area before port basin, sediment is distributed normally similar to the theoretical profile of Rouse. Meanwhile
208 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209

in the basin area, the concentration at bottom layer are significantly high, but decreasing as the distance get further from the
upstream of the port basin. The sediment was deposited in the upstream of the basin mainly due to the direction of the current. In
reality, sediment will be deposited according to dominant current direction.
Figure 6 shows the calculated vertical distribution from simulation of scenario-2, with variation of UWS height, for a Rouse
number of 1.2. It is showed that the existence of UWS caused the decreasing of the concentration at bottom layer inside the basin.
This due to some of the sediment were deposited before entering the port basin, and some were diverted to the side of UWS. The
concentration in port basin were decreasing as the increasing of UWS height.

Fig. 5. The calculated vertical distribution of scenario-1 (without UWS), Rouse number of 1.2.

Fig. 6. The calculated vertical distribution of scenario-2 (with variation of huws), Rouse number of 1.2.

4.4. Prediction of siltation’s reduction due to the underwater sill using the sediment transport model

The sediment disposition in port basin was calculated in the sediment transport model, for each variation of sediment parameter.
The existence of UWS causes a significant reduction of sediment disposition in port basin. Figure 7, shows the calculated
prediction of cumulative of sediment disposition’s reduction along with the ratio of UWS height to the water depth (h uws/h), for
variation of sediment parameter (Rouse Number). The result indicated that the capability of UWS in reducing the siltation
corresponds to its height. In general, the capability of UWS in protecting the basin from sediment disposition increased as the
height of UWS increased. Rouse number present the sediment parameter, and indicated how it transported (distribution profile).
The higher the Rouse number, the higher concentration of sediment particle in the bottom layer, which mean the sediment
transported more in layer near the bed.
 Nurul Azizah et al. / Procedia Engineering 170 (2017) 202 – 209 209

100
90 Ro = 0.8
Ro = 1.2

Cumulative Reduction (%)

80
Ro = 1.8
70
60
50
40
30
20
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6
huws/h

Fig. 7. Cumulative Reduction of Sediment disposition.

For Rouse number equal to 1.8, sediment are concentrated more in bottom layer. Therefore, deposition’s reduction due to the
existence of UWS are significantly high. Even just by UWS with height of 1/7 h, the reduction already reach 45% and almost 90%
for UWS with height of 4/7h. This indicated that UWS effective in protecting the basin from sediment that was mostly transported
at bottom layer, near the bed. In contrast to Rouse number equal to 0.8, UWS with height of 1/7h hardly produce significant effect
in reducing sediment deposition. This due to sediment concentrations were distributed more uniform over the depth as the smaller
value of rouse number. The sediment is easily transported above UWS and then entered the basin. Therefore, it almost gives no
effect in reducing the sediment disposition. For UWS higher than 1/7h, the reduction of deposition gradually increased as the
increasing of UWS height. Meanwhile for Rouse number of 1.2, the reduction of sediment deposition are almost linearly increase
as the increment of UWS height.

5. Conclusion

The numerical simulation that was present in this study are consisted of two main model which are: the Hydrodynamic model
that was based on the characteristic numerical scheme, and the sediment transport model that are based on particle-tracking model.
Both model are capable to simulate the current and sediment transport around underwater sill. The Result of particle tracking
model of vertical sediment distributions follow the trend of Rouse profile where the time step is roughly constrained by
ws 't / h  0.05 . It indicates that Particle tracking model is capable to characterize the sediment transport process, as long as the
time step requirement are fulfilled.
Based on the simulation result, the existence of underwater sill give significant effect on both current and sediment pattern
around the structure. The capability of UWS in deflecting the current increase along with the increment of UWS height. The
existence of underwater sill affect the pattern of sediment transport for h uws higher than 1/7 h but not significantly altered for the
smaller huws. It indicates UWS with height smaller than 1/7h does not have enough capability to divert the sediment.
Based on the Particle tracking model result, the amount of siltation’s reduction provide by UWS with the smallest height (huws
= 1/7 h) was 45% for sediment with rouse number of 1.8, 15 % for sediment with rouse number of 1.2, and only 1 % for sediment
with rouse number of 0.8. This indicates that UWS is most effective in reducing the siltation for sediment with higher value of
Rouse Number, which mean sediment was mostly transported near the bed. The siltation’s reduction is increase as the increasing
of UWS height.

References

[1] Van Rijn. L.C., (2013): Basic of Channel Deposition/Siltation. [Online] Available at: http://www.leovanrijn-sediment.com (Accessed June 13th 2015)
[2] Tsuruya. H., Murakami. K., and Irie. I., (1990): Mathemarical of Mud Transport in Ports with a Multi-Layered Model-Application to Kumamoto Port. Report
of The Port and Harbour Researh., vol 29, No 1, pp 3-51.
[3] Chai. J.C., Hayashi. S., and Yamanishi. H., (2002): Effect of Submerged Dike/Lifted Area on Seabed Mud Transport. Proceedings of the International
Symposium on Lowland Technology, September 2002, Saga University.
[4] Dimou. K.N. and Adams. E.E., (1993): A Random-Walk, Particle Tracking Model for Well-Mixed Estuaries and Coastal. Waters, Estuarine, Coastal and Shelf
Science, 37, pp. 99-110.
[5] Argall. R., Sanders. B. F., and Poon. Y., (2003): Random-Walk Suspended Sediment Transport and Settling Model. Proceedings of 8th International Conference
on Estuarine and Coastal Modeling, pp. 713-730
[6] Oh. J., (2011): Stochastic Particle Tracking Modeling For sediment Transport In Open Channel Flow. Doctoral Dissertation, Available from ProQuest
Dissertations and Theses database. (UMI No. 3460786)
[7] Julien. P.Y., (1995): Erosion and Sedimentation, Cambridge University Press, Cambridge.
[9] Triatmadja. R., (1990): Numerical and Physical Studies of Shallow Water Waves With Special References to Land Slide Generated Waves and The Method
of Characteristics. Doctoral Thesis, Department of Civil Engineering, University of Strathclyde, Glasgow, UK.