River Flow Processes Reader2021

River Flow Processes
201900188
Lecture Notes
Academic Year: 2021 - 2022
2
3
River Flow Processes

201900188
Lecture Notes
Taught by: Vasileios Kitsikoudis and Fredrik Huthoff

Academic Year: 2021 - 2022
Water Engineering and Management (WEM)

Department of Civil Engineering
Faculty of Engineering Technology
UNIVERSITY OF TWENTE.
4
PREFACE
The present reader is for the course “River Flow Processes” (Academic year: 2021-
2022), which is taught as part of the Water Engineering and Management Master
Program at the Civil Engineering Department of the University of Twente. The
reader comprises modified parts from the readers that were put together by B. Ver-
meulen and J.S. Ribberink in previous years for the courses “River Flow Processes”
and “River Dynamics” (I: Shallow-water flows and II: Transport Processes and Mor-
phology), respectively, which were taught within the same Master Program. These
readers, and subsequently also the present one, were based on:
de Vriend. H.J. (2005). River Engineering. Reader Ct3340, TU Delft, Faculty of

Civil Engineering and Geosciences, Section Hydraulic and Offshore Engineering.
5
6
CONTENTS
List of symbols 11
1 Introduction 15
2 Turbulence and roughness 17
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 General basic equations . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2 Conservation of momentum . . . . . . . . . . . . . . . . . . . 19
2.3 Reynolds averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Steady uniform flow with a free water surface . . . . . . . . . . . . . 26
2.4.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.2 Continuity equation and equation of motion . . . . . . . . . . 27
2.4.3 Hydrostatic pressure distribution . . . . . . . . . . . . . . . . 28
2.5 Flow resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6 Roughness and the velocity profile . . . . . . . . . . . . . . . . . . . 31
2.7 Composite cross-sections . . . . . . . . . . . . . . . . . . . . . . . . . 35
3 Sediment transport and bed forms 39
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7
8 CONTENTS
3.2 Sediment characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.1 Size and shape . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.2 Chemical composition and density . . . . . . . . . . . . . . . 42
3.2.3 Fall velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Sediment transport processes . . . . . . . . . . . . . . . . . . . . . . 45
3.3.1 The transport mechanism . . . . . . . . . . . . . . . . . . . . 45
3.3.2 Incipient motion . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.3 Bed forms and alluvial roughness . . . . . . . . . . . . . . . . 49
3.4 Sediment transport formulas . . . . . . . . . . . . . . . . . . . . . . . 51
3.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.4.2 Bed load formula of Meyer-Peter and Müller (1948) . . . . . 53
3.4.3 Bed-material load formula of Engelund and Hansen (1967) . 53
3.4.4 Bed and suspended load formulae of van Rijn (1984a,b) . . . 54
3.5 Remarks on the use of sediment transport formulas . . . . . . . . . . 55
4 Suspended sediment transport 57
4.1 Local equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1.1 Settling and turbulence mixing . . . . . . . . . . . . . . . . . 57
4.1.2 Concentration profiles . . . . . . . . . . . . . . . . . . . . . . 58
4.1.3 Bottom concentration . . . . . . . . . . . . . . . . . . . . . . 60
4.1.4 Alternative boundary conditions . . . . . . . . . . . . . . . . 60
4.2 Non-equilibrium suspended sediment transport . . . . . . . . . . . . 61
4.2.1 Concentration equation . . . . . . . . . . . . . . . . . . . . . 61
4.2.2 Initial and boundary conditions . . . . . . . . . . . . . . . . . 64
4.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 Depth-averaged suspended load equation . . . . . . . . . . . . . . . . 68

CONTENTS 9
5 Flow and morphology in river bends 73
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2 Spiral flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.3 Bed topography (axially-symmetrical) . . . . . . . . . . . . . . . . . 77
6 Flow-cylinder interaction and local scour 81
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2 Estimation of maximum scour depth at bridge piers . . . . . . . . . 83
6.3 Estimation of scour depth below horizontal pipes . . . . . . . . . . . 87
References 90
10 CONTENTS
LIST OF SYMBOLS
A area [L2 ]
B channel width [L]
C Chézy coefficient [L1/2 T−1 ]
C0 Chézy coefficient related to grain roughness [L1/2 T−1 ]
C 00 Chézy coefficient related to bed forms [L1/2 T−1 ]
C90 Chézy coefficient related to D90 [L1/2 T−1 ]
CD drag coefficient [-]
Cf bed friction coefficient [-]
D sediment diameter [L]
D50 median sediment diameter [L]
Ds representative grain diameter of sediment in suspension [L]
Dx sediment diameter from which x% of the sediment mixture by weight is finer [L]
D∗ particle parameter [-]
F force [M L T−2 ]
FD drag force [M L T−2 ]
FDz drag force on a spherical particle [M L T−2 ]
Fgz submerged weight of an immersed spherical particle [M L T−2 ]
Fsus integration factor in the formula of van Rijn [-]
G gravity force on an immersed sediment particle [M L T−2 ]
KI K-factor expressing the effect of flow intensity on scour depth [-]
Kd K-factor expressing the effect of b/D50 on scour depth [-]
Kh K-factor expressing the effect of pier width on scour depth [-]
Ks K-factor expressing the effect of pier shape on scour depth [-]
Kt K-factor expressing the effect of time on scour depth [-]
Kα K-factor expressing the effect of pier alignment with the flow direction on scour
depth [-]
O wet perimeter of a cross-section [L]
P pressure [M L−1 T−2 ]
11
12 CONTENTS
Q flow discharge [L3 T−1 ]

R submerged specific gravity of the sediment [-]
Rh hydraulic radius [L]
Re Reynolds number [-]
Re∗ shear Reynolds number [-]
Reg Reynolds number for flow below a horizontal pipe [-]
Rep settling Reynolds number [-]
T bed shear stress parameter [-]
U time- and depth-averaged flow velocity along the longitudinal direction [L T−1 ]
Ucr critical time- and depth-averaged flow velocity along the longitudinal direction
[L T−1 ]
Ug averaged flow velocity in a gap below a horizontal pipe [L T−1 ]
Z Rouse parameter [-]
Z0 modified suspension number [-]
αi flow acceleration in the i direction [L T−2 ]
b pier width [L]
bp projected width of an oblique pier in the normal to the approaching flow direction
[L]
c volumetric sediment concentration [-]
c0 vertical distribution function of sediment concentration [-]
ds maximum scour depth around a bridge pier or a horizontal pipe [L]
fDW Darcy-Weisbach coefficient [-]
fD vertical flux of sediment due to turbulent mixing per unit bed area [L T−1 ]
fs vertical flux of sediment due to settling per unit bed area [L T−1 ]
g gravitational acceleration [L T−2 ]
h flow depth [L]
i energy slope [-]
ib bed slope [-]
iw water surface slope [-]
ks roughness height [L]
m mass [M]
ṁ mass flow rate [M T−1 ]
n Manning coefficient [T L−1/3 ]
q flow discharge per unit width [L2 T−1 ]
qbl volumetric bed load transport rate per unit width for the formula of van Rijn
(1984a) [L3 T−1 L−1 ]
qg flow discharge per unit width in a gap below a horizontal pipe [L2 T−1 ]
qs volumetric sediment transport rate per unit width [L3 T−1 L−1 ]
CONTENTS 13
qsus volumetric suspended load transport rate per unit width for the formula of van
Rijn (1984b) [L3 T−1 L−1 ]
r radial coordinate in a polar coordinate system [L]
t time [T]
te time needed to reach the maximum scour depth at equilibrium conditions around
a pier [T]
u flow velocity along the longitudinal direction [L T−1 ]
u time-averaged flow velocity along the longitudinal direction [L T−1 ]
u0 turbulent flow velocity fluctuation along the longitudinal direction [L T−1 ]
u∗ shear velocity [L T−1 ]
u∗cr critical shear velocity [L T−1 ]
v flow velocity along the transverse direction [L T−1 ]
v time-averaged flow velocity along the transverse direction [L T−1 ]
v0 turbulent flow velocity fluctuation along the transverse direction [L T−1 ]
w flow velocity along the vertical direction [L T−1 ]
ws settling or fall velocity of a sediment particle [L T−1 ]
w0 turbulent flow velocity fluctuation along the vertical direction [L T−1 ]
w time-averaged flow velocity along the vertical direction [L T−1 ]
x longitudinal or streamwise coordinate [L]
y transverse coordinate [L]
z vertical coordinate [L]
z0 level above the bed where the flow velocity is zero [L]
za reference level [L]
zb bed level [L]
zw water level [L]
Φ sedimentological grain size scale [-]
εs mixing coefficient or turbulent diffusion coefficient [L2 T−1 ]
porosity [-]
ζ water level [L]
κ von Karman constant [-]
µ molecular dynamic viscosity [M L−1 T−1 ]
µr ripple factor [-]
µt eddy dynamic viscosity [M L−1 T−1 ]
ν molecular kinematic viscosity [L2 T−1 ]
νt eddy kinematic viscosity [L2 T−1 ]
ρ density of water [M L−3 ]
ρb bulk density of sediment [M L−3 ]
ρs density of sediment [M L−3 ]
14 CONTENTS
σ normal stress [M L−1 T−2 ]

τ shear stress [M L−1 T−2 ]
τb bed shear stress [M L−1 T−2 ]
τbcr critical bed shear stress according to Shields criterion [M L−1 T−2 ]
τb0 bed shear stress related to the grains [M L−1 T−2 ]
τb00 bed shear stress related to the bed forms [M L−1 T−2 ]
τ∗ dimensionless shear stress or Shields parameter [-]
τc∗ critical dimensionless shear stress or critical Shields parameter [-]
φ sediment transport parameter [-]
φbl sediment transport parameter for bed load [-]
ψ flow parameter [-]
1 INTRODUCTION
River flows are characterized by the dynamic interaction between the water flow
and the deformable river bed. The river bed consists of sediment, which may be
entrained by the flow and transported downstream. The flow shapes the bed but
this change will subsequently alter the bed roughness and the flow resistance, thus
affecting the flow pattern. This intertwined relationship between the flow and the
deformable bed constitutes a major difference between river flows and flows over
solid boundaries with constant hydraulic roughness. In addition, in river flows there
is a need to quantify the amount of sediment that is carried downstream by the flow
and is deposited in low-energy environments, such as reservoirs. To address properly
these challenges, Hydraulic Engineers must have a good understanding of turbulent
flow and boundary layer theory and how these relate to semi-empirical equations that
are used for sediment transport. The course of “’River Flow Processes’ intends to
introduce these topics to Civil Engineering students with a combination of advanced
fluid mechanics, turbulent flows, and sediment transport, specifically tailored for
river flows.
This reader consists of five chapters. After this short Introduction, the second chap-
ter presents the continuity equation and the Navier-Stokes equations, which are the
fundamental equations for turbulent flows. Their applicability to river flows through
the so-called Reynolds decomposition is elaborated as well as the flow resistance and
the induced velocity profiles. The third chapter introduces sediment transport by
presenting the sediment properties and the forces that are exerted on the grains rest-
ing on the bed. Semi-empirical equations for the incipience of sediment motion and
the quantification of sediment load are presented in detail as well as the induced bed
forms. The fourth chapter presents the role of advection and diffusion in suspended
sediment transport and the fifth chapter presents the flow and morphology in river
bends.
15
16 CHAPTER 1. INTRODUCTION
2 TURBULENCE AND ROUGHNESS
2.1 Introduction
To describe water flows quantitatively, the momentum and continuity equation for
fluids are used. For our purposes in this course, the fluid water can be considered
incompressible, i.e., the density of water is constant (ρ ≈ 1000 kg/m3 ). Since the
kinematic viscosity of water (ν ≈ 10−6 m2 /s) is relatively small, the water flow in
rivers, estuaries, and seas is generally turbulent. Considering the Reynolds number:
uL
Re = (2.1)
ν
where L is a typical length scale (e.g., the depth) and u a typical flow velocity. This
leads to a typical Reynolds number for river flows of Re ≈ 104 − 106 . This means, in
contrast to laminar flows, that velocity and pressure fluctuations are present, which
lead to additional exchanges of mass and momentum. This is important for fluid
forces on objects in the river flow (e.g., banks, bed, vegetation, bridge piers, ships)
and for the mixing of dissolved substances (e.g., stirring up sediment).
In this reader, only a brief presentation of the general flow equations is given (Sec-
tion 2.2). The Bachelor module parts “Fluid Mechanics 1” and “Fluid Mechanics
2” provide basic background knowledge for fluid mechanics. The book of Fox and
McDonald (1992) that was used in these courses offers additional background infor-
mation for those who are interested.
This section treats the simplest flow with a free surface: the steady uniform flow
over a sloping bed. This type of flow constitutes the simplest model of a river flow.
The discussed phenomena and terminology are also used in later chapters to describe
more complex flows.
17
18 CHAPTER 2. TURBULENCE AND ROUGHNESS
2.2 General basic equations
2.2.1 Conservation of mass
The continuity equation, or conservation of mass, is derived from the mass balance
in a fluid element with infinitesimally small volume, such as the cube in Figure 2.1
with dimensions ∆x, ∆y, and ∆z. Such a fluid element is a control volume, in which
a net inflow of mass should result in a mass increase in the element, or alternatively
a net outflow of mass should lead to a mass decrease in the element, according to
mass conservation. This can be expressed as:
dm
ṁin − ṁout = (2.2)
dt
where ṁin is the mass flow rate (i.e, passing mass per unit time) that goes into the
control volume, ṁout is the mass flow rate that gets out of the control volume, and
m is the mass of fluid within the control volume.
In Figure 2.1, the incoming mass flow rate can be expressed with the aid of the fluid
density, ρ, and the volumetric flow rate through the sides ∆y∆z, ∆x∆z, and ∆x∆y,
where the flow velocities are ux , vy , and wz , respectively, according to:
ṁin = ρux ∆y∆z + ρvy ∆x∆z + ρwz ∆x∆y (2.3)
Similarly, the outgoing mass flow rate from the cube can be also expressed with
the aid of the volumetric flow rate through the sides of the cube, but with the flow
Figure 2.1: Mass flow rate in and out of an elementary control volume
2.2. GENERAL BASIC EQUATIONS 19
velocities changed to ux+∆x , vy+∆y , and wz+∆z :
ṁout = ρux+∆x ∆y∆z + ρvy+∆y ∆x∆z + ρwz+∆z ∆x∆y (2.4)
The mass within the control volume is equal to:
m = ρ∆x∆y∆z (2.5)
and its time derivative is:

∂m ∂ρ
= ∆x∆y∆z (2.6)
∂t ∂t
Substituting Equations (2.3), (2.4), and (2.6) into Equation (2.2) and dividing both
sides of the resultant equation with the volume of the cube, ∆x∆y∆z, leads to (after
rearranging some of the terms):
ρux+∆x − ρux ρvy+∆y − ρvy ρwz+∆z − ρwz ∂ρ

− − − = (2.7)
∆x ∆y ∆z ∂t
Since the dimensions of the cube are infinitely small, Equation (2.7) can also be
written with partial derivatives:
∂ (ρu) ∂ (ρv) ∂ (ρw) ∂ρ

+ + + =0 (2.8)
∂x ∂y ∂z ∂t
The fluid we have to deal with in rivers is water, which is incompressible, thus
the fluid density does not vary with time nor in space. This means that since the
density in the element cannot change and the volume of the element is also constant,
the mass in the element is constant and the net fluxes through the element must
equal zero. By taking this into account, the water density can be removed from
Equation (2.8) and the continuity equation is simplified to:
∂u ∂v ∂w
+ + =0 (2.9)
∂x ∂y ∂z
2.2.2 Conservation of momentum
For the derivation of the equations of motion (or conservation of momentum) in

river flows we generally consider the flow from an Eulerian perspective. This means
that instead of following a water parcel along a streamline (Lagrangian approach),
we define a control volume and keep track of the influx and outflux of momentum
and the net forces acting on the control volume (for a derivation of the Eulerian
laws from the Lagrangian laws see Fox and McDonald, 1992).
The equations of motion can be derived from Newton’s second law for an infinites-
imally small element of fluid, which can be written separately for the x, y, and z
directions:
ΣFx = mαx (2.10a)

ΣFy = mαy (2.10b)
ΣFz = mαz (2.10c)
where ΣF denotes the sum of the exerted forces on the fluid element, m is the mass
in the fluid element, and α is the flow acceleration. When the fluid is incompress-
ible, as is the case with water, the mass within the fluid element is constant (see
Section 2.2.1).
By applying the chain rule in the velocity derivatives and considering that u =
∂x/∂t, v = ∂y/∂t, and w = ∂z/∂t, we get:
du (x, y, z, t) ∂u ∂u ∂u ∂u
= +u +v +w (2.11a)
dt ∂t ∂x ∂y ∂z
dv (x, y, z, t) ∂v ∂v ∂v ∂v
= +u +v +w (2.11b)
dw (x, y, z, t) ∂w ∂w ∂w ∂w
= +u +v +w (2.11c)
with the time derivative of the velocity (i.e., the first term at the right hand side
of the equations) being known as the local acceleration and the rest of the terms at
the right hand side of the equations being known as the convective acceleration.
The product of mass and acceleration can alternatively be written as the rate of
change of momentum, since acceleration is the rate of change of velocity. Hence, by
also dividing with the volume of the fluid element, Equation (2.10) becomes:
ΣFx dρu
= (2.12a)
∆x∆y∆z dt
ΣFy dρv
= (2.12b)
∆x∆y∆z dt
ΣFz dρw
= (2.12c)
∆x∆y∆z dt
The forces that are exerted on the fluid element can be distinguished into external
and internal forces. In river flows, the gravitational force is the external force and the
fluid induced stresses are the internal forces. The internal forces along the x direction
are shown in Figure 2.2 and consist of shear and normal stresses. The internal forces
along the y and z directions can be conceptualized in a similar way. Considering
that the size of the fluid element is infinitesimally small (with dimensions ∆x, ∆y,
and ∆z), the opposing Fxx , Fyx , and Fzx forces (i.e., the pairs with indices 1 and 2
in Figure 2.2) can be written in differential form as stresses, after they get divided
by the volume of the fluid element, according to:
Fxx2 − Fxx1 ∂σxx
= (2.13a)
∆x∆y∆z ∂x
Fyx2 − Fyx1 ∂τyx
= (2.13b)
∆x∆y∆z ∂y
Fzx2 − Fzx1 ∂τzx
= (2.13c)
∆x∆y∆z ∂z
where σ denotes normal stress and τ denotes shear stress. The above transformation
is obviously valid also for the forces along the y and z directions with the appropriate
changes.
Thus, considering that the gravitational force is positive as the driving force of the
Figure 2.2: Fluid induced normal and shear forces exerted on a fluid element in the
x-direction.
flow, the force balance for each direction becomes:
ΣFx ∂σxx ∂τyx ∂τzx

= ρgx + + + (2.14a)
∆x∆y∆z ∂x ∂y ∂z
ΣFy ∂τxy ∂σyy ∂τzy
= ρgy + + + (2.14b)
ΣFz ∂τxz ∂τyz ∂σzz
= ρgz + + + (2.14c)
where gx , gy , and gz are the gravitational accelerations along the x, y, and z direc-
tions, respectively.
To express the stresses based on the fluid molecular viscosity and velocities, for
Newtonian fluids, we use the following formulas (the derivation of these formulas
can be found in fluid mechanics textbooks) for the shear stresses:
!
∂u ∂v
τxy = τyx = µ + (2.15a)
∂y ∂x
!
∂v ∂w
τyz = τzy = µ + (2.15b)
∂z ∂y
!
∂u ∂w
τxz = τzx = µ + (2.15c)
∂z ∂x
and for the normal stresses, which also include the pressure of the flow, P :
∂u
σxx = −P + 2µ (2.16a)
∂x
∂v
σyy = −P + 2µ (2.16b)
∂y
∂w
σzz = −P + 2µ (2.16c)
∂z
with µ being the dynamic viscosity of the fluid, which is equal to ρν, where ν is
the kinematic viscosity of water. These equations express the interaction between
neighboring water parcels that move at different velocities. Due to random molecular
motions and interactions, the parcels of water will interact and the faster moving
fluid will accelerate the slower moving fluid, and vice-versa: the slower moving fluid
will slow the faster moving fluid. This is experienced by the flow as a viscosity
(dutch: stroperigheid).
By substituting Equations (2.11), (2.14), (2.15), (2.16) into Equation (2.12) and
after the necessary term rearrangements with the aid of the continuity equation for
incompressible fluid (Equation 2.9), and by dividing with the water density, we get:
!
∂u ∂u ∂u ∂u 1 ∂P ∂2u ∂2u ∂2u
+u +v +w = gx − +ν + + (2.17a)
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
!
∂v ∂v ∂v ∂v 1 ∂P ∂2v ∂2v ∂2v
+u +v +w = gy − +ν + + (2.17b)
∂t ∂x ∂y ∂z ρ ∂y ∂x2 ∂y 2 ∂z 2
!
∂w ∂w ∂w ∂w 1 ∂P ∂2w ∂2w ∂2w
+u +v +w = gz − +ν + + (2.17c)
∂t ∂x ∂y ∂z ρ ∂z ∂x2 ∂y 2 ∂z 2
|{z} | {z } |{z} | {z } | {z }
1 2 3 4 5
To summarize, the meaning of the terms in Equation (2.17) is:
1. Temporal change of momentum in fluid element (local acceleration)
2. Net flux of momentum through fluid element (convective acceleration)
3. Gravity force acting on fluid element
4. Net pressure force on fluid element
5. Net viscous forces on fluid element (caused by molecular interactions)
The Equations (2.17) are called the Navier-Stokes equations and describe any kind
of fluid flow. They can also be expressed in vector form:
∂~v 1
+ ~v · ∇~v = ~g − ∇P + ν∇2~v (2.18)
∂t ρ
Although the Navier-Stokes equation describe any kind of hydrodynamic problem,

these equations have some important practical drawbacks. The terms on the left
hand side are governed by the geometry of the problem we try to solve, which for
rivers is of the order of hundreds of meters. The viscous forces on the right hand
side, however, become important only at very small scales. Therefore, in order to
solve a flow problem, all scales must be solved. The ratio of these scales is somehow
quantified by the Reynolds number, which for rivers is very large. A numerical
model based on the Navier-Stokes equations (called a direct numerical simulation)
would require a computational mesh with a mesh size of the order of a hundredth of a
millimeter, extending over the hundreds of meters we wish to solve. This imposes too
large computational times. Analytical solutions, on the other hand, only exist under
strong assumptions, due to the non-linearity of the equations. Another problem is

that for large Reynolds numbers, the flow becomes chaotic due to the strong non-
linearity of the equations. To be useful in practice, the Navier-Stokes equations call
for some further elaboration.
2.3 Reynolds averaging
Next to the large range of spatial and temporal scales present in typical high
Reynolds number flows, such flows are often also chaotic. Osborne Reynolds was
the first one to show that for high Reynolds numbers, flows become chaotic. If
we would measure flow characteristics at some point in the flow, the chaotic be-
havior could be recognized in the fluctuations of hydrodynamic variables. Velocity
and pressure continuously change and fluctuate. Despite these fluctuations, one can
think of averaging these variables over a certain period of time, such that some more
coherent behavior can be observed. Reynolds proposed this averaging of the bulk
flow variables. Therefore such averaging is called Reynolds averaging.
Considering an averaging period T , the time average of a flow variable (here u) is

equal to:
1 t0 +T
Z
u= u dt (2.19)
T t0
This also allows to write a flow variable as a function of its mean, u, and fluctuating,
u0 , part:
u = u + u0 (2.20)
Applying this decomposition to the Navier-Stokes equations leads to:

!
∂u ∂u ∂u ∂u 1 ∂P ∂2u ∂2u ∂2u
+u +v +w = gx − +ν + +
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
! (2.21a)
∂u0 u0 ∂u0 v 0 ∂u0 w0
− + +
∂x ∂y ∂z
!
∂v ∂v ∂v ∂v 1 ∂P ∂2v ∂2v ∂2v
+u +v +w = gy − +ν + +
∂t ∂x ∂y ∂z ρ ∂y ∂x2 ∂y 2 ∂z 2
! (2.21b)
∂v 0 u0 ∂v 0 v 0 ∂v 0 w0
− + +
∂x ∂y ∂z
!
∂w ∂w ∂w ∂w 1 ∂P ∂2w ∂2w ∂2w
+u +v +w = gz − +ν + +
∂t ∂x ∂y ∂z ρ ∂z ∂x2 ∂y 2 ∂z 2
! (2.21c)
∂w0 u0 ∂w0 v 0 ∂w0 w0
− + +
∂x ∂y ∂z
2.3. REYNOLDS AVERAGING 25
These equations are similar to the Navier-Stokes equations except that the velocities
have been replaced with the corresponding Reynolds averaged velocities and three
new terms appeared on the right hand side of each equation. These terms contain
the fluctuating parts of the instantaneous velocity components. If we consider one of
them: u0 w0 , we see that this is the average of the product of the fluctuations in the x
and z components of velocity. This is exactly the definition of the covariance of u and
w. Therefore, these terms are often also called eddy-covariances, i.e., the covariance
caused by the turbulent fluctuations, or eddies. Put differently, these terms describe
the effect that turbulent fluctuations have on the flow. To understand better what
these terms are, it is worth noting that these terms are very similar to the molecular
stress terms we derived before. In fact, these terms are often also referred to as
Reynolds stresses, which in summation convention can be written as:
τij = −ρu0i u0j (2.22)
Writing these equations in vector notation we obtain:
∂~v 1 1
+ ~v · ∇~v = ~g − ∇P + ν∇2~v + ∇ · R (2.23)
∂t ρ ρ
where R is the Reynolds stress tensor defined as:
 0 0
u u u0 v 0 u0 w0

R = −ρ  v 0 u0 v 0 v 0 v 0 w0  (2.24)
w0 u0 w0 v 0 w0 w0
We have seen now how turbulence acts on the fluid as a stress, in a similar way as
molecular stresses do. We can take this analogy even further by assuming that much
like molecular stresses also turbulent stresses will relate to the shear in mean flow:
!
∂ui ∂uj
τij = µt + = −ρu0i u0j (2.25)
∂xj ∂xi
with µt the dynamic turbulent viscosity. We can substitute this in Equation (2.23)
to obtain:
∂~v 1
+ ~v · ∇~v = ~g − ∇P + (ν + νt ) ∇2~v (2.26)
∂t ρ
As we can see from this last result, the turbulence resulted in additional viscosity of
the fluid. When we apply the Reynolds averaging, the fluctuating part of velocity
will mix momentum much like molecular motions do. When we have a shear in
the mean velocity, turbulent fluctuations will tend to accelerate the slower moving
fluid and decelerate the faster moving fluid. This is felt by the flow as additional
viscosity. In the major part of a river channel the turbulent viscosity dominates over
the molecular viscosity. Only very near the wall, molecular viscosity becomes more
important and finally dissipates energy from the mean flow, through turbulence, into
heat. Please note that in the rest of these lecture notes we will omit the overline for
Reynolds averaged quantities.
2.4 Steady uniform flow with a free water surface
2.4.1 General
The most simple flow with free surface is the flow in a straight channel with constant
bed slope and with a constant cross-section (prismatic channel) in which we further
assume that:
- the flow, i.e. the water depth and flow velocity do not vary in time (steady
flow)
- the flow also does not change in space, i.e. in the direction of the flow (uniform
flow)
In this case the water depth, h, is constant in the direction of the flow and the water
surface is parallel to the bed. This means that the water surface slope equals the
bed slope (Figure 2.3).
The bed slope, ib , is defined as:
∂zb
ib = − = tan β (2.27)
∂x
Figure 2.3: Steady and uniform flow

2.4. STEADY UNIFORM FLOW WITH A FREE WATER SURFACE 27
In most rivers (exception: mountain rivers) the bed slopes are small (order of mag-
nitude: 10−2 − 10−5 ) and approximately the following holds:
ib = tan(β) ≈ sin(β) ≈ β (2.28)
2.4.2 Continuity equation and equation of motion
Continuity equation
The continuity equation can be derived by integrating the basic equation (Equa-
tion 2.9 in Section 2.2.1) over the water depth. As an alternative, a mass balance
approach can be derived for a small element of the fluid with length ∆x and height
h; for a steady flow this simply means that the amount of inflowing water per unit
time and width, q, has to be equal to the amount of outflowing water per unit time
and width:
q = U h = constant (2.29)
where U is the time- and depth-averaged flow velocity.
For a channel with finite width B and a rectangular cross-section (depth h) the total
discharge, Q, is:
Q = U hB = qB = constant (2.30)
and for a channel with arbitrary cross-section with area A we can write:
Q = U A = constant (2.31)
in this case U is the average flow velocity over the cross-section.
Equation of motion
For the same small element ∆x, the momentum equation can also be applied. In
this case, the acceleration of the flow equals zero (steady uniform flow) and there is
equilibrium between two external forces: (1) the component of gravitation force in
the direction parallel to the bed, and (2) the bottom friction. This is explained in
more detail in the following.
The forces per unit width on a fluid element ∆x of a steady uniform flow are given
in Figure 2.4. The hydrostatic pressure forces F1 and F2 are equal. The gravity
component in the x-direction is Fg = ρg∆x∆y(h − z) sin β. The equilibrium of
forces in the x-direction yields:
τz ∆x∆y = ρg∆x∆y(h − z) sin β ⇒ τz = ρg(h − z) sin β (2.32)
Since sin β ≈ ib it follows that:
τz = ρg(h − z)ib (2.33)
Equation (2.33) expresses a linear shear stress distribution over the depth. For z = 0
in Figure 2.4, it follows that τz=0 = τb :
τb = ρghib (2.34)
Equation (2.34) is valid for a wide channel (B >> h). In case of an arbitrary cross-
section with cross-sectional area A and wet perimeter O (e.g., Figure 2.5), the force
balance for an element ∆x yields:
A
τb O∆x = ρgA∆x sin β ⇒ τb = ρg ib (2.35)
O
The hydraulic radius, Rh , is defined as the ratio of the cross-sectional area to the
wet perimeter, Rh = A/O, which for a simple rectangular cross-section (width B
and depth h) is:
A Bh
Rh = = (2.36)
O B + 2h
The combination of Equations (2.35) and (2.36) leads to:

τb = ρgRh ib (2.37)
For a wide rectangular cross-section where B >> h, we can reasonably approximate

Rh ≈ h.
Equations (2.33), (2.34), (2.35), and (2.37) are valid for laminar and turbulent flows.
2.4.3 Hydrostatic pressure distribution
The vertical distribution of the water pressure p(z) is, in case of a hydrostatic pres-
sure distribution, determined by the mass of the water column above each level z
Figure 2.4: Fluid forces on a fluid element ∆x in steady and uniform flow conditions
2.5. FLOW RESISTANCE 29
and gravity. Since there are no acceleration terms in a steady uniform flow, it can
be assumed that the pressure acts as in a static water column, where the water
pressure at each level z is linearly proportional to the distance to the water surface
(= zw − z), with the water surface being denoted by zw . This hydrostatic pressure
distribution can be written as:
P (z) = ρg (zw − z) (2.38)
2.5 Flow resistance
So we have seen from the previous that turbulence plays a crucial role in flows
with high Reynolds numbers. Flow in rivers is always turbulent and turbulence is
also responsible for dissipating energy of the mean (Reynolds averaged) flow. If we
consider the flow in rivers from a Reynolds averaged perspective, we look at the
mean flow. The mean flow is not fluctuating anymore, but the turbulence removes
energy from the flow. This energy is then transferred from the largest turbulent
scales to smaller scales. This process is called the “energy cascade” phenomenon.
You can think of it as shear in the mean flow that induces turbulent fluctuations, or
eddies. These eddies initiate smaller eddies. This process continues until very small
scales are reached where the viscous forces will remove energy from the turbulent
fluctuations and dissipate it into heat through molecular motions.
To summarize the previous we can say that turbulence dissipates energy of the flow.
This process is so complex that it is often hard to take it into account explicitly.
In practice, turbulence is often parameterized as roughness, or a resistance term in
the equations. To obtain this parameterization, we will try to simplify the Reynolds
averaged equations and find a way to relate the turbulent dissipation to one of the
main hydraulic variables, in this case the mean flow velocity.
In the simplest kind of flow in a river, we will have a uniform, steady flow, where
no accelerations are present and the water flows only in the downstream direction.
Figure 2.5: Wet perimeter, O, and cross-sectional area, A, of a cross-section with

arbitrary shape.
Under these conditions the Reynolds averaged flow equations (Equation 2.21) reduce
to: !
∂2u ∂u0 w0
0=µ + ρgib − ρ (2.39)
∂z 2 ∂z
Note that Equation (2.39) is derived from Equation (2.21a), where the temporal and
horizontal derivatives and all velocities except the streamwise velocity are zero. The
gravitational term was elaborated to ρgib , with ib denoting the bed slope, since we
consider the x-axis along the river bed.
Molecular viscosity is small compared to the turbulent effects, therefore we can also
neglect the molecular viscosity term (first term on the right hand side of Equa-
tion 2.39). Thus, we obtain:
∂u0 w0
ρgib = ρ (2.40)
∂z
Simplifying and integrating over depth we have:
τb
− u0 w0 = ghib = (2.41)
ρ
This equation can be interpreted as a balance between gravity that drives the flow
in the river and bed friction that dissipates energy, similarly to what was mentioned
in Section 2.4.2. These turbulent stresses at the bed are another way to express the
bed shear stress, τb . The units of measurement of the bed shear stress divided by
the density is that of velocity squared. Therefore we often refer to the bed shear
stress through the concept of shear velocity, u∗ , which is defined as follows:
s
τb
u∗ = (2.42)
ρ
This shear velocity can therefore also be interpreted as a measure for the efficiency
of turbulence to mix momentum (or sediment, as we will see later). To have an easy
to use parameterization of bed roughness, we try to relate it to one of the hydraulic
variables, such as the depth-averaged flow velocity with a friction coefficient (Equa-
tion 2.43). Considering all the above, a friction coefficient parameterizes the bed
shear stress, which in turn is an approximation of the effect of turbulence on the
flow. By assuming a quadratic relationship between the bed shear stress and the
depth-averaged flow velocity using a bed friction coefficient, Cf , (turbulent flow) we
obtain:
τb = ρCf U 2 (2.43)
There are several relationships to specify the friction coefficient. In the Nether-
lands, the most commonly used relationship is the one using a Chézy coefficient, C
2.6. ROUGHNESS AND THE VELOCITY PROFILE 31
(m1/2 s−1 ):
U2 g
τb = ρg 2
⇒ Cf = 2 (2.44)
C C
Note that a large Chézy-value means a small roughness. Another type of friction
coefficient is that of Darcy-Weisbach, fDW :
1 1
τb = fDW ρU 2 ⇒ Cf = fDW (2.45)
8 8
For a wide channel with depth h and width B >> h, a combination of Equa-
tions (2.34) and (2.44) leads to the equation of motion for steady uniform flow,
namely the Chézy law: p
U = C hib (2.46)
For a channel with an arbitrary cross-section with hydraulic radius Rh , a similar

relationship can be obtained for the flow velocity averaged over the entire cross
section: p
U = C Rh ib (2.47)
Another widely used relationship that links the flow depth to flow velocity in rivers
is the equation of Manning:
1 2/3 1/2
U = Rh ib (2.48)
n
where n in the roughness coefficient of Manning. The Chézy and Manning coeffi-
cients can be estimated with methods presented in Section 2.6.
2.6 Roughness and the velocity profile
We can look more closely how turbulence affects the vertical profile of velocity. To
this end we need to introduce a parameterization that describes how the turbulent
shear varies over depth. Ludwig Prandtl proposed an expression for the principle
shear stress: !2
∂u
u0 w0 = l (2.49)
∂z
where l is the mixing length. You could think of this mixing length as the distance
over which a turbulent eddy is able to mix momentum. For a river we assume that
this mixing length is proportional to the distance from the bed, i.e., an eddy cannot
be larger than its distance from the bed:
l = κz (2.50)
where κ is the von Karman constant and is equal to 0.41. We can combine Equa-
tions (2.41), (2.42), (2.49), and (2.50) to obtain:
∂u
u∗ = κz (2.51)
∂z
This differential equation can be solved by integrating over the river depth, h. In-
stead of integrating from z = 0, the integration is done from z = z0 , which is the
level at which the velocity is equal to zero. Integration leads to the logarithmic
velocity profile: !
u∗ z
u (z) = ln (2.52)
κ z0
Important to note here is that the velocity in a river is affected by the boundary (i.e.,
the bed) over the entire depth. In other words, the flow in rivers is fully immersed
in the boundary layer.
To fully describe the flow over the depth, we need an expression for z0 . To this end,
we make a distinction between:
1. hydraulic smooth boundaries, where viscous effects are still important very
close to the boundary
2. hydraulic rough boundaries, where the geometry of the boundary disturbances

(wall roughness) is dominant
3. a transition regime
A general expression for z0 is based on measurements of Nikuradse and reads:
ν
z0 = 0.11 + 0.03ks (2.53)
u∗
where ks is the roughness height of the wall. Often, for ks , the so-called equivalent
sand roughness is used for flat beds:
ks ≈ grain diameter (2.54)
The magnitude of the wall Reynolds number (or shear Reynolds number)
u∗ ks
Re∗ = (2.55)
ν
controls the wall roughness regime and leads to the following classification:
• Re∗ > 70: hydraulic rough wall, where wall roughness dominates
2.6. ROUGHNESS AND THE VELOCITY PROFILE 33
• Re∗ < 5: hydraulic smooth boundary, where viscous effects dominate
• 5 ≤ Re∗ ≤ 70: transitional regime
In general, in rivers we have hydraulic rough walls, for which Equation (2.53) be-
comes:
ks
z0 = (2.56)
33
Combining this with Equation (2.52), the following relationship can be obtained:
!
u∗ 33z
u (z) = ln (2.57)
κ ks
By integrating this equation vertically over the water column (between z = z0 and
z = h; with z0 h) the depth-averaged flow velocity U can be obtained:
1
Z h
U= u (z) dz (2.58)
h z0
Combining Equations (2.57) and (2.58), we obtain:

!
u∗ 12h
U= ln (2.59)
κ ks
Combining Equations (2.57) and (2.59), the term u∗ /κ can be eliminated to obtain
the following dimensionless expression for the shape of the velocity profile u (z):
! !
33z h
ln + ln
u (z) h ks
= ! (2.60)
U 12h
ln
ks
The ratio between the water depth and the roughness height, h/ks , controls the
shape of the velocity profile in turbulent flows. Figure 2.6 shows a comparison
between velocity profiles for turbulent flows and for viscous laminar flows. The
larger mixing caused by turbulence leads to more uniform velocity profiles (higher
velocities near the bottom).
The combination of Equations (2.42) and (2.44) leads to:

√
g
u∗ = U (2.61)
C
Figure 2.6: Velocity profiles for turbulent and laminar flows (van Rijn, 1990)
Substitution of Equation (2.61) in Equation (2.59) gives:

√
g 12h
C= ln (2.62)
κ ks
Using g = 9.81 m/s2 , κ = 0.41 and replacing ln by a log10 results in the White-
Colebrook relationship for the Chézy value of a hydraulic rough bed (wide channel
B h), which is often used in practice:
!
12h
C = 18 log10 (2.63)
ks
This expression is also used for channels with an arbitrary cross-section by using the
hydraulic radius, Rh , instead of the water depth:
!
12Rh
C = 18 log10 (2.64)
ks
An approximation for this formula is the Strickler relationship:

!1/6
Rh
C = 25 (2.65)
ks
2.7. COMPOSITE CROSS-SECTIONS 35
which gives reasonable estimates of the velocity profile for 40 < C < 70 m1/2 /s.
Instead of the Chézy coefficient, the so-called Manning coefficient, n, can be also
used:
1/6
R
n= h (2.66)
C
Combination of Equations (2.64) and (2.66) gives an expression for Manning n in

terms of roughness height ks . An approximation can be found using the Strickler
formula (Equation 2.65):
n = 0.04ks1/6 (2.67)
For both the Chézy coefficient and the Manning coefficient we need information
about the roughness height ks . Figure 2.7 below shows ks values for different mate-
rials like soil types, concrete, etc.
2.7 Composite cross-sections
For more complicated cross-sections, e.g., a river with floodplains (Figure 2.8) or
a cross-section with different roughness heights for the banks and the river bed
(Figure 2.9), the Chézy equation can be applied separately for each part of the
cross-section. In the case of composite cross-sections with different ks values, as
shown in Figures 2.8 and 2.9, the flow resistance can be determined by assuming
that the Chézy-formula is valid in each part of the cross-section and that the water
surface gradient is equal in each part. The shear stresses in the plane between each
part of the cross-section are assumed to be zero.
Method 1
Figure 2.8 clearly shows two separate parts A1 (= B1 h1 ) and A2 (= B2 h2 ). The

water surface gradient, iw , is constant for each part:
U12 U22 U2
iw = = = (2.68)
Rh1 C12 Rh2 C22 Rh C 2
The continuity equation yields:

Q = Q1 + Q2 ⇒ U A = U1 A1 + U2 A2 (2.69)
Substitution of Equation (2.68) in (2.69) gives:

!0.5 !0.5
Rh1 C12 Rh2 C22
U A = A1 U + A2 U (2.70)
Rh C 2 Rh C 2
Figure 2.7: Values for ks (Nikuradse roughness height)
The overall C value can then be computed as:
!0.5 !0.5
A1 Rh1 A2 Rh2
C = C1 + C2 (2.71)
A Rh A Rh
2.7. COMPOSITE CROSS-SECTIONS 37
Figure 2.8: Composite cross-section with a main channel and a floodplain
Figure 2.9: Composite cross-section with different roughness heights for the river
bed and the river banks
The other parameters are:
A1 = B1 h1
A2 = B2 h2
Rh1 = A1 /O1
Rh2 = A2 /O2
Rh = A/O
(2.72)
O 1 = B 1 + h1
O2 = B2 + 2h2 − h1
O = B1 + B2 + 2h2
C1 = 18 log10 (12Rh1 /ks,1 )
C2 = 18 log10 (12Rh2 /ks,2 )
assuming rough flow.
Method 2
In case of Figure 2.9, the area of the cross-sections A1 and A2 cannot be estimated
accurately. Therefore, another assumption is necessary being the hypothesis of Ein-
stein. He assumed U1 = U2 = U , so that Equation (2.68) yields:
1 1 1
2 = 2 = (2.73)
Rh1 C1 Rh2 C2 Rh C 2
Equation (2.73) yields two expressions:
1.
Rh2 C12
= (2.74)
Rh1 C22
Applying the Strickler formula (Equation 2.65), this results in:
" #1/4
Rh2 ks,2
= (2.75)
Rh1 ks,1
2.
Q2
R1 C12 = RC 2 = (2.76)
A2 i
and
Q2
R2 C22 = RC 2 = (2.77)
A2 i
3 SEDIMENT TRANSPORT AND BED FORMS
3.1 Introduction
In this chapter the basic theories on sediment transport are discussed. The dis-
cussion is restricted to non-cohesive material; the theory on sludge transport will
not be discussed. Furthermore, the discussion is restricted to sediment transport
under the influence of flow alone, the influence of short waves will not be treated in
this chapter. The transport of sediment is both theoretically and experimentally a
difficult problem. This is caused by the nature of the flow, which is a two-phase flow
(liquid and solid phase), and the interaction between the two phases: flow causes
transport of bed material, which induces development of bed forms (ripples, dunes)
leading to changes in bottom roughness that eventually affects water motion and
sediment transport.
3.2 Sediment characteristics
The material that is transported by rivers has its origin in the river basin. The
following sediment characteristics are important for the erosion, transport, and de-
position of sediment:
• dimension
• shape
• density
• fall velocity
• chemical composition
• pore content
39
40 CHAPTER 3. SEDIMENT TRANSPORT AND BED FORMS
3.2.1 Size and shape
Sediments can be classified based on their diameter (Table 3.1). In nature, sediment
grains are not perfectly round and the grain size cannot be defined explicitly by the
diameter. Depending on how the grain size is determined, the following distinctions
can be made:
• sieve diameter; size of sieve opening through which a grain will just pass
• sedimentation diameter; diameter of a sphere with the same specific weight
and fall velocity as the given grain in the same fluid
• nominal diameter; diameter of a sphere with the same volume as the grain
Table 3.1: Classification of sediment grains based on their diameter according to the
American Geophysical Union.
Class name Grain size in mm Grain size in Φ units

Boulders > 256 < -8
Cobbles 64 – 256 -6 – -8
Very coarse gravel 32 – 64 -5 – -6
Coarse gravel 16 – 32 -4 – -5
Medium gravel 8 – 16 -3 – -4
Fine gravel 4–8 -2 – -3
Very fine gravel 2–4 -1 – -2
Very coarse sand 1–2 0 – -1
Coarse sand 0.5 – 1 1–0
Medium sand 0.25 – 0.5 2–1
Fine sand 0.125 – 0.25 3–2
Very fine sand 0.062 – 0.125 4–3
Coarse silt 0.031 – 0.062 5 – 4
Medium silt 0.016 – 0.031 6 – 5
Fine silt 0.008 – 0.016 7 – 6
Very fine silt 0.004 – 0.008 8 – 7
Coarse clay 0.002 – 0.004 9–8
Medium clay 0.001 – 0.002 10 – 9
Fine clay 0.0005 – 0.001 11 – 10
Very fine clay 0.00025 – 0.0005 12 – 11
Colloids < 0.00025 > 12
In general, the size of coarse material (D > 0.062 mm) is determined by sieving,
while the size of fine material is determined by measuring the fall velocity.
3.2. SEDIMENT CHARACTERISTICS 41
When the cumulative distribution of the sediment diameter is plotted against the
corresponding weight percentage, the so-called sieve curve is obtained, as shown for
example in Figure 3.1.
Figure 3.1: Example of a sieve curve
In sediment transport, typical diameters such as the following are often used:
• median grain diameter D50 : the grain size from which 50% of the sediment
mixture by weight is finer or the 50th percentile of the grain size distribution,
• mean grain diameter Dm ,
• other grain diameters such as D35 , D65 , or D90 : indicate the 35th, 65th or
90th percentiles of the grain size distribution, respectively.
The Φ-scale is also used for the classification of sediment grain sizes (see also Ta-
ble 3.1):
Φ = − log2 (D) (3.1)
For the calculation of Φ, the grain diameter D is in millimeter. Note: the higher
the value of Φ, the finer the material.
It must be noted that sediment in a natural river consists of a mixture of grain

sizes, which vary both spatially and temporally. To determine the sieve curve and
the typical grain diameters, field samples need to be collected and subsequently
analyzed (by sieving or by sedimentation tests). The resulting sieve curve depends
on the sampling method and sieving technique. Hence, specialized equipment and
standardized procedures have been developed for sampling and sieving.
Sometimes, the shape of sediment grains is characterized by the so-called shape

factor (SF ):
Dc
SF = √ (3.2)
Da Db
in which Da , Db , and Dc are respectively the longest, intermediate, and shortest of
the three mutually perpendicular axes of the grain. Most natural sand and gravel
grains have a shape factor of approximately 0.7.
3.2.2 Chemical composition and density
The mineral composition affects several properties of the sediment such as the di-
mension, the shape, and above all the density. The sediment density can be assumed
equal to ρs = 2650 kg/m3 (i.e., quartz density). In addition, the density of the de-
posited material (bulk density) may also be of importance. This is the density of
the material including the pores (porosity ). For quartz grains in the river bed, the
value of is typically 0.4. The bulk density, ρb , then becomes
ρb = ρs (1 − ) = 1590 kg/m3 (3.3)
These values can strongly differ for sludge and clay.
3.2.3 Fall velocity
The fall velocity of grains is especially important for suspended sediment transport
and is defined as the terminal fall velocity of a particle in non-moving water. If
a sediment grain is dropped into still water, it will initially accelerate under the
action of gravity. While the particle accelerates, the drag on the particle (which is
a function of the particle velocity) will increase. After a while, the exerted drag will
become equal to the submerged weight of the particle and the particle will reach a
terminal fall velocity, or settling velocity.
3.2. SEDIMENT CHARACTERISTICS 43
The submerged weight, Fgz , of an immersed spherical particle is given by:
1
Fgz = π (ρs − ρ) gD3 (3.4)
6
The drag force exerted by water on a spherical particle with cross-sectional area A,
moving vertically with fall velocity ws is approximated by:
1 1 1
FDz = CD ρAws2 = CD ρ πD2 ws2 (3.5)
2 2 4
with CD , the drag coefficient for sphere. When the particle reaches the final velocity,
the sum of all forces on the particle will be equal to zero:
1 1 1
Fgz = FDz ⇒ π (ρs − ρ) gD3 = CD ρ πD2 ws2 (3.6)
6 2 4
From Equation (3.6), we can get an expression for the settling velocity:
s
4 RgD
ws = (3.7)
3 CD
with R the submerged specific gravity of the sediment:
ρs − ρ
R= (3.8)
ρ
With Equation (3.7) we can compute the settling velocity of sediments, but for that
we need an estimate of the drag coefficient CD . The drag coefficient depends on the
settling Reynolds number, Rep :
ws D
Rep = (3.9)
ν
For small settling Reynolds numbers (Rep < 1, the Stokes area) the flow around the
sediment particle is laminar and a simple relation exists for the drag coefficient:
24
CD = (3.10)
Rep
Combining Equations (3.7), (3.9), and (3.10) we obtain
Rg 2
ws = D (3.11)
18ν
We can compute the settling velocity with Equation (3.11) but only within the
Stokes regime. The grain sizes for which the Stokes regime is valid can be obtained
Figure 3.2: Relation between drag coefficient for spheres CD and settling Reynolds
number for various shape factors (Albertson, 1953). In the text, the settling
Reynolds number is referred to as Rep .
by combining Equations (3.9) and (3.11) and considering that Rep < 1, which for
water leads approximately to D < 0.1 mm. Outside the Stokes regime the relation
between CD and Rep is more complex (Figure 3.2) and no simple expression exists
to describe it.
For natural sediments with D > 0.1 mm, the fall velocity was determined empirically.
To summarize, the following relations for settling velocity can be used (van Rijn,
1993):
Rg 2
ws = D , 0.001 < D ≤ 0.1 mm (3.12a)
18ν s 
10ν 0.01RgD3
ws =  1+ − 1 , 0.1 < D ≤ 1 mm (3.12b)
D ν2
p
ws = 1.1 RgD, D > 1 mm (3.12c)
The fall velocity does not depend only on the grain properties. The temperature,
for instance, affects settling velocity through the kinematic viscosity. The vicinity
of the walls and the presence of other grains can influence the fall velocity as well.
In the presence of other grains, the fall velocity depends on sediment concentra-
tion (hindered settling). Finally, cohesive sediment flocculation (the aggregation of
several small cohesive particles into larger ones) can increase significantly the fall
velocity.
In the rest of these lecture notes it will become apparent that sediment transport
3.3. SEDIMENT TRANSPORT PROCESSES 45
studies have a strong empirical character. This results in different authors using
different variables and notations in empirical formulas. Due to the great complexity
of sediment transport, estimates are often very rough and approximate while large
discrepancies among the different methodologies may be observed.
3.3 Sediment transport processes
3.3.1 The transport mechanism
A fluid flowing over a loose granular bed exerts forces on the grains. When these
forces exceed a certain critical value, some grains will begin to move. With increas-
ing water velocity more grains will move and sediment transport will take place.
Sediment transport occurs either as bed load or as suspended load (Figure 3.3).
Bed load is defined as the transport of bed material by rolling, sliding, and hopping.
Suspended load is defined as the transport of sediment that is suspended in the
water column for some time.
Figure 3.3: Classification of sediment transport (Jansen et al., 1979, p. 90)
Another distinction can be made between wash load and bed material load (Fig-
ure 3.3). Wash load consists of very fine sediments that are always in suspension
and do not settle on the river bed. Therefore, this sediment fraction is scarcely found
on the bed of the river. In other words, wash load does not exchange with the river
bed. The amount of wash load is determined exclusively by the upstream supply
and not by the transport capacity of the flow. Wash load is part of the suspended
load. The part of the sediment transport that interacts with the bed is called bed
material load and is composed of both bed load and part of the suspended load.
When solving morphological problems, the wash load, which does not interact with
the bed, should not be considered. In some situations, such as settling behind a
dam, wash load will eventually settle on the bed and should be considered.
3.3.2 Incipient motion
Sediment grains on the bed of a river move under the action of the flowing water.
When the velocity of water is small, the grains will remain on the bed. Once the
velocity exceeds a certain threshold, grains will start moving. This threshold is
called the threshold of motion, or incipient motion. Whether the sediment moves or
not is a result of the forces acting on a grain.
The following forces are exerted on a grain (Figure 3.4):
• flow forces; these consist of drag forces and lift forces,

• gravity forces (submerged weight),
• resultant reaction forces from the surrounding grains.
Figure 3.4: Forces on a grain resting on the river bed
The interaction of these forces is very complex and difficult to take into account
explicitly. Despite this complexity, it is still reasonable to assume that the initiation
of motion will be somehow related to the balance between gravity and the flow
induced drag acting on the particle. The ratio of the magnitude of these two forces
generates a dimensionless number called the Shields parameter or dimensionless
shear stress:
~
FD
τ ∗ = (3.13)
~
G
The magnitude of the drag force can be approximated with the bed shear stress, τb ,
acting on the surface of a particle:

~
FD ∝ τb D2 = ρu2∗ D2 (3.14)
The magnitude of the gravity force can be approximated as:

~
G ∝ (ρs − ρ) gD3 (3.15)
Combining Equation (3.13) with Equations (3.14) and (3.15) we get an expression
for the Shields parameter, τ ∗ :
u2∗
τ∗ = (3.16)
gRD
The Shields parameter is a measure for the mobility of sediments. The higher this
number, the more likely the sediment will move. When the flow is at initiation of
sediment motion, the value of the Shields parameter is called the critical Shields
parameter, τc∗ . When the Shields parameter is smaller than this critical value,
no sediment transport will take place. The critical value, which determines the
initiation of sediment motion, depends on the sediment properties and the flow
pattern near the bed.
Shields (1936) determined experimentally the critical Shields parameter τc∗ as a

function of the shear Reynolds number, Re∗ (Equation 2.55), for ks equal to a char-
acteristic grain diameter. Usually, D50 is used as the characteristic grain diameter.
The obtained experimental points formed the so-called Shields curve, as shown in
Figure 3.5.
Figure 3.5: Shields diagram for the initiation of motion
For flows that are slightly above the Shields curve, the sediment will be moving as
bed load by rolling, sliding, or hopping in close contact with the bed. As the flow
strength increases, after a certain point the sediment will be suspended in the water
column and be transported as suspended load. The threshold for particle suspension
is u∗ = ws , with suspension occurring when the shear velocity is greater than the
fall velocity (Garcia, 2008).
Since both τ ∗ and Re∗ depend on u∗ , it is difficult to determine a critical value for
u∗ directly from Figure 3.5. For this reason, several researchers used the particle
parameter, D∗ , as an alternative to Re∗ for the determination of τc∗ :
!1/3
Rg
D∗ = D (3.17)
ν2
van Rijn (1993) estimated explicitly the Shields curve based on D∗ with the following
relations:

0.24
if 1 < D∗ ≤ 4


D∗










 0.14
if 4 < D∗ ≤ 10


 D∗0.64






τc∗ = 0.04 (3.18)

 if 10 < D∗ ≤ 20
D∗0.1









0.013D∗0.29 if 20 < D∗ ≤ 150









0.055 if D∗ > 150

Comments:
• The flow conditions for which initiation of sediment motion occurs are diffi-
cult to be determined: Shields conducted his experiments at small sediment
transport rates and subsequently extrapolated to zero.
• The influence of gradation: it can be inferred from the Shields curve (Fig-
ure 3.5) that small non-cohesive grains are entrained easier than larger ones.
This phenomenon is less pronounced in mixtures, because the smaller grains
are sheltered by the larger grains, while larger grains protrude more into the
flow when they are surrounded by smaller grains. In practice this is only im-
portant for mixtures with D90 /D10 > 5. The effect of armoring also occurs
in mixtures and mostly in gravel-bed rivers. That is when the finer grains
have already been transported, so the top layer of the bed consists mostly
of coarse grains. These coarse particles will form an armor layer above the
remaining finer particles and further erosion is prevented. This phenomenon

is especially important downstream of a weir, but also occurs in for instance
the Grensmaas.
• Cohesive material: the erosion resistance of clay is strongly increased by the

cohesion of the material.
3.3.3 Bed forms and alluvial roughness
When calculating the bed shear stress or the shear velocity for the Shields parameter,
it is important to consider that the total bed roughness in a river is determined
both by skin friction drag, induced by the sediment of the bed surface, and pressure
drag, caused by pressure differences induced by bed forms. For the transport of
sediments only the former should be taken into consideration. Bed forms will, in
general, develop whenever sediment transport takes place. Figure 3.6 and Table 3.2
give an overview of the various bed forms and the accompanying flow patterns. Bed
forms cause additional resistance against the flow. With an alluvial soil the sediment
transport and the bed form depend on the flow. This means that a simple (quadratic)
relation between the bed shear stress and the flow velocity is not necessarily valid
anymore (Figure 3.7), as was the case with a flat non-moving bed.
Figure 3.6: Bed forms in alluvial channels (Simons and Richardson, 1966)
Table 3.2: Classification of bed forms
Flow Bed forms Bed material Mode of Type of Roughness

√
Regime concentration sediment roughness C/ g
(ppm) transport
ripples 10 - 200 7.8 - 12.4
ripples on
lower 100 - 1200 pulsed form
dunes
dunes 200 - 2000 7.0 - 13.2
washed out
transition 1000 - 3000 variable 7.0 - 20
dunes
plane bed 2000 - 6000 16.3 - 20
upper anti-dunes >2000 continuous grain 10.8 - 20
chutes and
>2000 9.4 - 10.7
pools
Since the bed forms and thus the alluvial roughness depend on the flow conditions,
it is no longer possible to determine the bottom roughness in advance. When the
roughness is unknown, the estimation of the flow characteristics and the sediment
transport must be performed iteratively:
Bed form → Chézy coefficient → Bed shear velocity (u∗ ) → Sediment transport →
Bed form
In practice, the roughness (Chézy) can be determined based on measured gradient

data. This makes the calculation of the sediment transport more accurate. However,
there are cases where the roughness is unknown, for example in the design of a new
canal. In that case, the alluvial roughness has to be predicted. At present, no reliable
methods are available for this. Examples of prediction methods of alluvial roughness
are those of Einstein and Barbarossa (1952); Engelund and Hansen (1967); van Rijn
(1993), etc. In all these methods a distinction is made between the bed shear stress
τb0 related to the grains and the bed shear stress τb00 related to the bed forms, which
is expressed with the following:
τb = τb0 + τb00 (3.19)
This can also be expressed with the aid of Chézy coefficients:

1 1 1
2
= 02 + 002 (3.20)
C C C
where C 0 denotes the Chézy coefficient related to the grain roughness and C 00 denotes
the Chézy coefficient related to the bed form.
3.4. SEDIMENT TRANSPORT FORMULAS 51
Figure 3.7: Relation between bed shear stress and velocity and how it changes in
the presence of bed forms (van Rijn, 1993).
3.4 Sediment transport formulas
3.4.1 Introduction
Sediment transport formulas are used to calculate the volume or mass of sediment
being transported under certain hydraulic conditions. There are a lot of sediment
transport formulas, all with their specific attributes. Most of these formulas have
been based on laboratory observations. Before a certain formula is applied on a
river reach, it is necessary to validate a number of formulas on the basis of field
measurements. Subsequently, the most suitable formula is selected. Sometimes the
constants in the formula have to be slightly adjusted to suit the circumstances.
The existing sediment transport formulas can be divided into different types:
• formulas for bed load

• formulas for suspended load
• formulas for bed-material load (bed load and suspended load)
In the literature, sediment transport formulas are presented in various forms. Some
formulas compute the mass transport rate (kg m−1 s−1 ) and others the specific
volume transport rate (m3 s−1 m−1 ). Even the same sediment transport formula
may be presented differently by various authors. Sometimes the volume includes
the pore volume. All sediment transport formulas state that the sediment transport
rate is a function of the gravitation acceleration (g), the fluid properties (ρ, ν), the
sediment properties (ρs , D) and one or more parameters regarding the flow (e.g.,
τb ). For example:
qs = f (g, ρ, ν, ρs , D, τb ) (3.21)
where qs is volumetric sediment transport rate
By means of dimensional analysis, these parameters can be combined into differ-

ent dimensionless variables. In many sediment transport formulas, the following
dimensionless variables are included:
the sediment transport parameter:
qs
φ= p (3.22)
gRD3
and the flow parameter:
µr τb hi
ψ= = µr τ ∗ = µ r (3.23)
ρgRD RD
where µr denotes the ripple factor, which represents the influence of the bed form
on shear stress partitioning: !
C
µr = f (3.24)
C90
where the Chézy coefficient C90 is related to D90 :

!
12h
C90 = 18 log (3.25)
D90
In addition, it must be noted that some formulas use D50 as the grain diameter D
and others use the mean diameter. The slope i represents the slope of the energy
head. This slope is only equal to the bed slope ib in uniform flow. In the next
sections, a number of frequently used sediment transport formulas will be given as
an example. Note that each formula has a specific area of application.
3.4. SEDIMENT TRANSPORT FORMULAS 53
3.4.2 Bed load formula of Meyer-Peter and Müller (1948)
The formula of Meyer-Peter and Müller (1948) is experimental and computes only
the bed load transport. The formula is developed for the following conditions:
• µr τ ∗ < 0.2
• D > 0.4 mm
• ws /u∗ > 1
According to this methodology, the bed load transport can be computed as:
φ = 8 (ψ − 0.047)3/2 (3.26)
The formula computes the volume transport of solid material per unit width. Some-
times the formula is presented with a constant value of 13.3 instead of 8. The formula
then applies to the transport of bulk volume (that is including the pores) with a
porosity = 0.4. It is more accustomed to use Equation (3.26). The formula uses
the mean grain diameter and the ripple factor is defined as:
!3/2
C
µr = (3.27)
C90
3.4.3 Bed-material load formula of Engelund and Hansen (1967)
Although originally developed to compute bed load, the formula of Engelund and
Hansen (1967) was proven to be particularly suited for the computation of bed-
material load of relatively fine sandy material, where a major part of it is transported
as suspended load. The formula can be applied under the following conditions:
• ws /u∗ < 1
• 0.07 < τ ∗ < 6
• 0.19 < D50 < 0.93 mm
The calculated sediment transport includes both bed load and suspended load of
the bed-material (thus excluding wash load). The formula can be written as:
φ = 0.05ψ 5/2 (3.28)
with !2/5
C2
µr = (3.29)
g
This formula computes the volume transport of solid material, excluding the pores.
The grain diameter used in this formula is the median diameter D50 . An alternative
formulation is: !3 !2
qs u∗ U
p
3
= 0.05 √ √ (3.30)
gRD50 gRD50 gRD50
3.4.4 Bed and suspended load formulae of van Rijn (1984a,b)
In the formula of van Rijn (1984a,b) a distinction is made between bed load, qbl ,
and suspended load, qsus . The bed-material load is then calculated as:
qs = qbl + qsus (3.31)
The formula uses the following two parameters instead of φ:
1. the bed shear stress parameter, T :

τb0 − τbcr
T = (3.32)
τbcr
with τb0 the bed shear stress related to the grain roughness and τbcr the critical
bed shear stress for the initiation of sediment motion according to Shields
criterion (Equation 3.18).
2. the dimensionless particle parameter D∗ (Equation 3.17), with D50 as the

characteristic grain diameter.
The calculation of τb0 is done with the following formula:

!2
C
τb0 = τb (3.33)
C0
where C 0 is equal to C90 , which can be estimated with Equation (3.25).
The bed load is computed with the following:


 T 2.1

0.053 for T < 3
D∗0.3



φbl = (3.34)
T 1.5



0.1 0.3 for T ≥ 3


D∗
where φbl is the sediment transport parameter for bed load:
qbl
φbl = p 3
(3.35)
gRD50
3.5. REMARKS ON THE USE OF SEDIMENT TRANSPORT FORMULAS 55
in which qbl represents the bed load excluding the pores with a porosity of 0.4.
The suspended load is computed with the following:
qsus = Fsus U hca (3.36)
The concentration ca (excluding the pores) can be computed as:
D50 T 1.5
ca = 0.015 (3.37)
α D∗0.3
with α being a reference level, which is typically assumed to be half of the bed form
height, or equal to the Nikuradse roughness height ks . The reference level should
always be larger than 0.01h for reasons of accuracy (van Rijn, 1984b).
Fsus is an integration factor, which can be calculated as:

!Z 0 !1.2
α α
−
h h
Fsus = !Z 0 (3.38)
α
1− (1.2 − Z 0 )
h
where !0.8 !0.4

ws ws ca
Z0 =  !2  + 2.5 (3.39)
ws u∗ 1−
1 + 2  κu∗
u∗
ws
is the modified suspension number, with Equation (3.39) being valid for 0.01 ≤ ≤
u∗
1.
The fall velocity, ws , is calculated with the aid of a representative grain diameter of
the sediment in suspension, Ds , given by:
! !
Ds 1 D84 D50
= 1 + 0.011 + − 1 (T − 25) (3.40)
D50 2 D50 D16
Because of the explicit distinction between bed load and suspended load, the area of
application of this formula is quite large. For more details see (van Rijn, 1984a,b).
3.5 Remarks on the use of sediment transport formulas
• For applications in non-stationary, uniform flows careful attention must be

paid to the meaning of the variables in the formula. Some parameters are
related to the bed shear velocity, others are related to the mean velocity. In
the formula of Engelund and Hansen (1967) and in the formula of van Rijn
(1984a,b) both are present. It is not practical to work with the energy slope i,
except for stationary, uniform flows. It is certainly not allowed to replace the
energy slope i with the bed slope ib . In tidal areas where sediment transport
can occur and ib < 0, it would lead to absurd results.
• When the suspended load plays a vital role, the sediment transport formula
actually only calculates the transport capacity. The real transport depends on
the situation upstream and the situation in the past, due to the adaptation
length and the adaptation time (see Chapter 4). Only if the adaptation length
and the adaptation time are relatively small compared to the scales of the con-
sidered problem, the transport and the transport capacity may be considered
equal.
4 SUSPENDED SEDIMENT TRANSPORT
In the previous chapter, we saw how suspended sediment transport can be deter-
mined with the use of sediment transport formulas. Such formulas compute the
amount of sediment being transported in suspension, but do not describe the verti-
cal distribution of sediments in the water column. In this section, we will estimate
the distribution of sediment in the water column, both for local equilibrium and
non-equilibrium. Local equilibrium occurs when the processes that tend to lift the
sediment upward in the water column are balanced by the processes that tend to
move the suspended sediment towards the bed.
4.1 Local equilibrium
4.1.1 Settling and turbulence mixing
Settling of sediment particles in the water column occurs due to the difference in
specific density between sand and water. The particles have the tendency to accel-
erate as they fall, but they are hindered by the water. In still water this leads to a
constant settling velocity ws (Figure 4.1), which is among others a function of the
particle size.
The vertical sediment flux per unit bed area, fs , corresponding to this settling is
equal to:
fs = −ws c (4.1)
Turbulent mixing is a result of the random fluctuations in the motion of the flowing
water that are caused by turbulence. When these fluctuations happen at a place
where the sediment concentration changes, they tend to mix and therefore smooth
the sediment concentration differences. This happens because a package of high
sediment concentration is moved into a region of lower sediment concentration and
vice-versa. In rivers, near the bed the concentration of sediment is generally very
high. The turbulence will therefore tend to move sediment from the near-bed region
with high concentration upwards into the water column where the concentration of
sediment is lower. This process is very similar to diffusion.
57
58 CHAPTER 4. SUSPENDED SEDIMENT TRANSPORT
Figure 4.1: Settling and turbulent mixing of sediment
The upward sediment flux per unit bed area due to turbulent mixing, fD , is given
by the following:
∂c
fD = −εs (4.2)
∂z
The parameter εs is called the mixing coefficient, or turbulent diffusion coefficient,
analogous to the turbulent viscosity, or eddy viscosity for transporting momentum
in turbulent flow.
At the equilibrium situation, for steady and uniform flow, the vertical flux due to
turbulent mixing, fD , is exactly compensated by the vertical flux due to settling,
fs , so:
∂c
fD + fs = 0 ⇒ ws c + εs =0 (4.3)
∂z
From this differential equation the vertical profile of sediment concentration can be
estimated, provided that a boundary condition is defined. The solution will also
depend on the vertical profile of the diffusivity εs . We will try to find a solution for
a constant diffusivity in depth and for a parabolic profile.
4.1.2 Concentration profiles
If we consider the sediment diffusivity to be constant over depth and given a con-
centration c(za ) at level za (boundary condition), we can solve Equation (4.3):
ws (z − za )
−
c (z) = c (za ) e εs (4.4)
4.1. LOCAL EQUILIBRIUM 59
If we consider a parabolic profile for the sediment diffusivity

!
z z
εs = κu∗ h 1− (4.5)
h h
The depth averaged value (which can be used in Equation 4.4) of the diffusivity is:
κ 1
ε̂s = u∗ h ≈ u∗ h (4.6)
6 15
We can combine Equation (4.5) with Equation (4.3) and separate variables in the
resulting differential equation:
!
dc ws dz ws dz dz
=− !=− + (4.7)
c κu∗ z κu∗ z h−z
z 1−
h
By defining the Rouse parameter, Z:

ws
Z= (4.8)
κu∗
and integrating Equation (4.7), we obtain:

ln c = −Z [ln z − ln (h − z)] + constant (4.9)
If c(za ) is given, this yields:

!Z
h − z za
c (z) = c (za ) (4.10)
z h − za
Note that for this profile c(z = h) = 0, which is not the case for the profile in the
case of a constant diffusivity (Equation 4.4). Furthermore, dc/dz|z=0 = ∞, which
means that the level za should be chosen at a small distance above the bed.
To determine the sediment concentration c(za ) at reference level za , it is possible

to use an empirical formula that relates this sediment concentration to hydraulic
conditions (such as Equation 3.37).
As a final remark, note that the Rouse parameter and the Shields parameter are
closely related. Combining Equations (3.7), (3.16), and (4.8) we obtain:
4
τ∗ = (4.11)
3CD κ2 Z 2
For coarse sediments, CD is no longer a function of grain size (Figure 3.2) and
τ ∗ ∝ 1/Z 2 .
4.1.3 Bottom concentration
The concentration at the bottom, c(za ), can be determined in two ways:
(a) via an empirical model (i.e., based on observations) that expresses this con-
centration as function of the hydraulic conditions and the sediment properties;
the total suspended transport can then be derived from
Z Z
qsuspended = uc dz = c (za ) uc0 dz (4.12)
h h
where c0 (z) = c(z)/c(za ) is the vertical distribution function of the concentra-

tion.
(b) via a sediment transport formula, which gives qsuspended as function of the
hydraulic conditions and the sediment properties (one can take for example
a total transport formula and assume what fraction of the total transport is
suspended load); the concentration at the bottom can then be determined from
qsuspended
c (za ) = R 0
(4.13)
h uc dz
4.1.4 Alternative boundary conditions
In principle it is not necessary to prescribe the concentration at the boundary or at

another point in the vertical to solve Equation (4.3). lt is for example also possible
to prescribe the concentration gradient at the bottom, or a linear combination of
the concentration and the concentration gradient that differs from Equation (4.3).
Suppose that we select as boundary condition:

∂c
= −A (4.14)
∂z za
in which A is a known constant. This in fact prescribes the sediment flux from the
bottom, since

∂c
fD za = −ε = εza A (4.15)
∂z za
In the equilibrium situation, it follows from Equation (4.3) that

ε ∂c A
c za = −
= ε (4.16)
ws ∂z za
ws za
4.2. NON-EQUILIBRIUM SUSPENDED SEDIMENT TRANSPORT 61
so that also the concentration at the bottom is known. In other words: the gradient-
type boundary condition (Equation 4.14) is in this case the same as prescribing the
concentration at the bottom.
As will be discussed in next section, the last conclusion does not apply for non-
equilibrium situations when the upward flux due to turbulent mixing does not bal-
ance the downward flux due to settling.
4.2 Non-equilibrium suspended sediment transport
4.2.1 Concentration equation
Suspended transport is not always in equilibrium. It is very well possible that a net
sediment flux occurs, so that the vertical concentration is filled up or emptied. To
model this phenomenon we consider the sediment balance for a small fluid element
(dx, dz) on the vertical plane (Figure 4.2):
Figure 4.2: Sediment balance
The net sediment import into the small fluid element of Figure 4.2 is derived as
follows:

∂c
left in: ul cl − εl dz (4.17a)
∂x l

∂c
right out: ur cr − εr dz (4.17b)
∂x r

∂c
bottom in: wb cb − εb dx (4.17c)
∂z b

∂c
top out: wt ct − εt dx (4.17d)
∂z t
which leads to net import, Inet :
Inet = (ul cl − ur cr ) dz + (wb cb − wt ct ) dx

! !
(4.18)

∂c ∂c ∂c ∂c
− εl − εr dz − εb − εt dx
∂x l ∂x r ∂z b ∂z t
The variation of the amount of sediment in the fluid element per unit time is:
∂c
∆sed = dzdx (4.19)
∂t
The increase of the amount of sediment must be equal to the net import (because
within the element no sediment is produced nor destroyed), thus:
∂c Inet
∆sed = Inet ⇒ = (4.20)
∂t dxdz
If we develop Inet into a Taylor series around the point (x, z):

∂ (uc)
+ O dx2

ur cr = ul cl + dx (4.21)
∂x l
!
∂c ∂c ∂ ∂c
+ O dx2

εr = εl + dx ε (4.22)
∂x r ∂x l ∂x ∂x
l

∂ (wc)
+ O dz 2

wt ct = wb cb + dz (4.23)
∂z b
!
∂c ∂c ∂ ∂c
+ O dz 2

εt = εb + dz ε (4.24)

∂z t ∂z b ∂z ∂z
b
Thus, the balance Equation (4.20) can then be written as:

! !
∂c ∂ (uc) ∂ ∂c ∂ (wc) ∂ ∂c
=− + ε − + ε + O (dx) + O (dz) (4.25)
∂t ∂x ∂x ∂x ∂z ∂z ∂z
We assume that the settling velocity in flowing water is equal to that in stagnant
water, ws , so:
u = uf (4.26)
and
w = wf − ws (4.27)
in which the subscript f refers to the flow (water motion). Substitution into Equa-
tion (4.25) yields:
! !
∂c ∂ (uf c) ∂ ∂c ∂ (wf c) ∂ ∂c ∂c
=− + ε − + ε + ws (4.28)
∂t ∂x ∂x ∂x ∂z ∂z ∂z ∂z
With the continuity equation for water:
∂uf ∂wf
+ =0 (4.29)
∂x ∂z
this leads to the sediment concentration equation:
! !
∂c ∂c ∂c ∂c ∂ ∂c ∂ ∂c
+ uf + wf = ws + ε + ε (4.30)
∂t ∂x ∂z ∂z ∂x ∂x ∂z ∂z
This equation includes Equation (4.3) for equilibrium conditions.
For steady, uniform flow and transport in the x-direction, we get:
∂ ∂
= 0; = 0; wf = 0 (4.31)
∂t ∂x
so that Equation (4.30) reduces to:

! !
∂c ∂ ∂c ∂ ∂c
0 = ws + ε ⇒ 0= ws c + ε (4.32)
∂z ∂z ∂z ∂z ∂z
The net vertical flux at the water surface must be zero, since no sediment can go
through the water surface:
∂c
ws c + ε =0 at z = ζ = zb + h (4.33)
∂z
where ζ is the water level and zb is the bed level.
From Equation (4.32) it follows that this is valid everywhere in the vertical:
∂c
ws c + ε =0 (4.34)
∂z
with this equation being equal to Equation (4.3).

4.2.2 Initial and boundary conditions
The concentration equation (4.30) can only be solved in combination with the correct
number of initial and boundary conditions. The equation only contains the first order
derivative of t, which means that only one initial condition is required. In general,
this is satisfied by prescribing the concentration field at t = 0: c(x, z, 0).
The equation contains the first and the second derivative to x, which means that
two boundary conditions are required at the upstream and/or downstream edge. It
is important to realize that two physical processes are involved: advection (trans-
portation by the flow) and diffusion (mixing due to turbulence). Advection is usually
dominant in rivers, so that certainly one boundary condition is required at the point
where the flow enters the model area, thus at the inflow edge. There the concentra-
tion is usually given: c(x0 , z, 0).
If only advection would occur, this would be sufficient. However, the second deriva-
tive in the diffusion-term makes a second boundary condition necessary. Contrary to
advection, diffusion has no preferential direction (“works in all directions”). It is log-
ical then to apply the second boundary condition at the downstream edge. Because
of the dominance of the advection, this boundary condition is usually formulated in
such a way that it affects the solution as less as possible, (e.g., dc/dx = 0).
The equation also contains the first and second order derivatives in the vertical. Here
the settling and the diffusion are dominant (see the consideration of the equilibrium
situation). Therefore, one boundary condition is applied at the water surface and one
at the bottom. No sediment goes through the water surface, thus, if this is horizontal
and remains at the same level, Equation (4.33) applies as boundary condition.
In other cases the sediment flux at the surface z = h(x, t), must exactly balance the
variation of the amount of sediment between this level and the water surface, thus:
∂c ∂ζ ∂ζ
(wf − ws ) c − ε = c + uc (4.35)
∂z ∂t ∂x
However, at the water surface the kinetic boundary condition for the water motion
applies:
∂ζ ∂ζ
wf = +u (4.36)
∂t ∂x
Thus, Equation (4.33) remains the boundary condition at the water surface for all
cases.
The boundary condition at the bottom can have various forms, like:
(a) the concentration is always equal to the equilibrium concentration (or, alter-
natively, the sediment concentration at the bottom adapts instantaneously to
the hydraulic conditions):

cza = ce za (4.37)
where ce (za ) is determined in the same way as in case of equilibrium, i.e., from
Equation (4.12) or (4.13).
(b) the upward flux is always equal to that at the equilibrium situation (or, alter-
natively, the flux, which is determined by the local hydrodynamic condition,
instantaneously adjusts to the hydraulic conditions):

∂c ∂ce
ε =ε = −ws ce za (4.38)
∂z ∂z
za za
4.2.3 Analysis
We now consider the physical meaning of each of the terms in the sediment concen-
tration equation (Equation 4.30):
! !
∂c ∂c ∂c ∂c ∂ ∂c ∂ ∂c
+ uf + wf = ws + ε + ε (4.30 revisited)
∂t ∂x ∂z ∂z ∂x ∂x ∂z ∂z
|{z} | {z } | {z } | {z } | {z } | {z }
1 2 3 4 5 6
1. storage in the water column (resulting in an increase or decrease of the sedi-

ment concentration)
2. horizontal advection term, which expresses the transport of suspended sedi-

ment by the horizontal water motion
3. vertical advection term, which expresses the transport of suspended sediment

by the vertical water motion
4. settling term
5. horizontal diffusion term, which expresses the gradual distribution of the sus-
pended sediment in the x-direction (leveling off of peaks and dips in the hori-
zontal concentration distribution)
6. vertical diffusion term, which expresses the gradual distribution of the sus-
pended sediment in the z-direction (leveling off of peaks and dips in the hori-
zontal concentration distribution)
Omitting terms (4), (5) and (6) yields pure advection of dissolved matter:
∂c ∂c ∂c dc
+ uf + wf =0 ⇒ =0 (4.39)
∂t ∂x ∂z dt
which means that, moving with the water the concentration remains constant.
Omitting terms (2), (3) and (4) yields pure diffusion of dissolved matter. For a
constant diffusion coefficient, ε, we have:
!
∂c ∂2c ∂2c ∂c
=ε 2
+ 2 ⇒ = ε∇2 c (4.40)
∂t ∂x ∂z ∂t
If we neglect the terms (2), (3), (4), and (6), we obtain free settling of sediment:
∂c ∂c dc dz
+ ws =0 ⇒ =0 along lines = −ws (4.41)
∂t ∂z dt dt
Example: Sand trap
A sand trap is an enlarged (deepened) area of a watercourse, which is meant to cause

decelerated flow, so that the sediment in suspension can settle and be removed from
the system. We assume that the sediment transport in the sand trap can be described
by:
∂c ∂c
uf − ws =0 (4.42)
∂x ∂z
This model contains thus horizontal advection and settling. The solution of this
equation can be simplified by introducing the following coordinate-transformation:
ws
y=z+ x and ζ = z (4.43)
uf
as a result of which:
∂c ∂c ∂y ∂c ∂ζ ws ∂c
= + = (4.44a)
∂x ∂y ∂x ∂ζ ∂x uf ∂y
∂c ∂c ∂y ∂c ∂ζ ∂c ∂c
= + = + (4.44b)
∂z ∂y ∂z ∂ζ ∂z ∂y ∂ζ
and after substitution in Equation (4.42):
∂c
=0 ⇒ c = c (y) (4.45)
∂ζ
Suppose that the concentration is given at the inflow edge, x = 0, where y = z:
c (y) = c (0, z) for y ≤ h (4.46a)

c (y) = 0 for y>h (4.46b)
Figure 4.3: Settling of suspended sediment in sand trap.
then the concentration variation along the sand trap can be graphically represented
as shown in Figure 4.3.
Take for example c(0, z) = c0 = constant ⇒ qs0 = uhc0 , then it follows for x > 0:
ws
c (x, z) = c0 for z ≤ h − x (y ≤ h) (4.47a)
uf
ws
c (x, z) = 0 for z >h− x (y > h) (4.47b)
uf
so that
!
ws qs (x) ws x
Z
qs (x) = uc dz = uf c0 h− x ⇒ =1− (4.48)
h uf qs,in uf h
So in this simple case the transport decreases linearly with x/h and the decrease
rate is determined by the ratio ws /uf .
If we apply a more extensive suspended transport model, with horizontal advec-

tion, settling and vertical diffusion, then we obtain a different pattern, as shown in
Figure 4.4.
So it is seen that:
1. for small values of x/h, the transport decreases faster than according to the
previous model with only horizontal advection and settling
2. for large values of x/h, the transport decreases slower than according to the
previous model
Figure 4.4: Sand trap including vertical diffusion
3. the concentration in the sand trap does not go to zero
These phenomena can be explained as follows:
1. With sudden deepening at the beginning of the sand trap, the vertical sediment
concentration gradient will be positive (Figure 4.5). Consequently the diffusion
works in the same direction as the settling, and accelerates the concentration
decrease. The settling on its own of the lower part of the vertical does not
result in a sign change of the sediment concentration gradient (Figure 4.3).
2. For larger values of x/h, the shape of the concentration vertical is adjusted such
that the vertical gradient is negative again. So the diffusion works against the
settling and the adaptation process of the transport is slower than according
to the model without vertical diffusion.
3. If there is vertical diffusion, the concentration and the transport for large
values of x/h will approach the equilibrium value. The concentration and the
transport will only be zero if the bed shear stress in the sand trap is below the
critical value of movement for the sediment (τb < τbcr ).
4.3 Depth-averaged suspended load equation
Many mathematical models at present still work in depth-averaged mode. Therefore

there is a demand for a depth-averaged concentration model for suspended sediment
transport. To derive this we start with the simplified concentration equation:
!
∂c ∂c ∂c ∂c ∂ ∂c
+ uf + wf = ws + ε (4.49)
∂t ∂x ∂z ∂z ∂z ∂z
4.3. DEPTH-AVERAGED SUSPENDED LOAD EQUATION 69
Figure 4.5: Concentration at the beginning of the sand trap
in which the horizontal diffusion is neglected. Using the continuity equation for
water this can be written in the conservative form:
!
∂c ∂ (uf c) ∂ (wf c) ∂ ∂c
+ + = ws c + ε (4.50)
∂t ∂x ∂z ∂z ∂z
Formal integration over the water depth yields:
ζ ζ

∂c ∂ (uf c)
Z Z

dz + dz + (wf c) − (wf c)

za ∂t za ∂x ζ za
! ! (4.51)
∂c ∂c
= ws c + ε − ws c + ε

∂z ∂z

ζ za
With the rule of Leibnitz:

Z b(t) ∂f (z, t) ∂
Z b(t) db da
dz = f (z, t) dz − f (b, t) + f (a, t) (4.52)
a(t) ∂t ∂t a(t) dt dt
Equation (4.51) can be written as
ζ Z ζ

∂ ∂ζ ∂za ∂ ∂ζ ∂za
Z
c dz − c + c + uf c dz − (uf c) + (uf c)
∂t za ζ ∂t za ∂t ∂x za ζ ∂x za ∂x
! ! (4.53)
∂c ∂c
+ (wf c) − (wf c) = ws c + ε − ws c + ε
∂z ∂z

ζ za ζ za
Using the boundary condition at the water surface and introducing the definition of
depth-averaged value:
1 ζ
Z
f= f dz (4.54)
h za
this yields: !
∂ (hc) ∂ (huf c) ∂c
+ = − ws c + ε (4.55)

∂t ∂x ∂z

za
Herein, the first term is the total storage of the suspended sediment in the water
column, the second term is the variation of the total suspended sediment transport
and the term on the right hand side is the net sediment flux from the bottom.
Equation (4.55) does not form a model for the depth-averaged concentration yet,
because there is still a term with c(za ) in it. To express this in terms of the depth-
averaged concentration we assume:
1. the upward, diffusive flux at the bottom is always equal to that at the equi-
librium situation, which corresponds to the hydraulic conditions (see Equa-
tions 4.37 and 4.38)
2. all concentration verticals are uniform, i.e.:
c(x, za , t) ce (z) c(x, za , t) ce (za )

= ⇒ = =β
ˆ (4.56)
c(x, t) ce c(x, t) ce
In that case
uf c = αuf c (4.57)
in which α is constant. The depth-averaged concentration equation then becomes
∂ (hc) ∂ (huf c)
+α = −ws β (c − ce ) (4.58)
∂t ∂x
With the depth-averaged continuity equation for water flow,
∂h ∂ (huf )
+ =0 (4.59)
∂t ∂x
this can be written as
∂c ∂c ws β
+ αuf =− (c − ce ) (4.60)
∂t ∂x h
This equation is entirely expressed in depth-averaged variables and therefore fits

perfectly in a depth-averaged model. However, the strict uniformity of the vertical
concentration profiles is often a too rigorous assumption. The exchange processes in
the vertical direction are often much stronger than those in the horizontal direction,
4.3. DEPTH-AVERAGED SUSPENDED LOAD EQUATION 71
so it can be expected that the vertical concentration profile is most sensitive to

the deviations from the equilibrium situation. It appears to be possible, using a
somewhat more complicated derivation, to find an equation for the depth-averaged
concentration in which this effect is taken into account (see e.g., Galappatti and
Vreugdenhil, 1985). The result can be written as:
∂c ∂c
Ta + La = −c + ce (4.61)
∂t ∂x
In which the adaptation time Ta and the adaptation length La can be approximated
by:
h uf h
Ta ≈ and La ≈ (4.62)
ws ws
Thus Ta is approximately equal to the time that a sediment grain needs to settle
in stagnant water over a distance equal to the water depth h, and La is approx-
imately equal to the distance traveled by the flow in this time. Mathematically,
Equation (4.62) is the same as Equation (4.60) with a = 1 and β = 1.
Note that according to Equation (4.57), a can be much smaller than 1, especially
if the vertical concentration is “poorly filled’. So Equation (4.62) is not always an
approximation of Equation (4.60).
5 FLOW AND MORPHOLOGY IN RIVER BENDS
5.1 Introduction
Lowland rivers are seldom straight and often exhibit a meandering behavior, i.e., the
river has a single main channel with a pattern that consists of a series of alternating
bends. This has various effects, among which:
• A river length that is considerably larger than the valley length. This also
means a longer navigation distance.
• Reduction of the river slope as compared with the valley slope.
• Non-uniform cross-section. The water depth in the outer bend is considerably

larger than in the inner bend.
• Reduction of the effective navigable width, due to the depth variation and the
channel curvature. The latter is important for long ships and barges.
• Tendency of the outer bend to erode and of the inner bend to accrete. As a
consequence, meanders tend to grow and migrate, or (in the case of a protected
outer bank) to increase their depth in the outer bend.
• Spatial segregation of coarse and fine sediments, usually with the most coarse
material in the outer bend.
• Formation of bars and shoals around the transition between two consecutive
bends.
• Complex 3-D flow, giving rise to extra dispersion of transported matter, hence
to an increased “carrying capacity” of the river.
Therefore, there are many reasons to investigate what is actually happening in a

river bend.
Let us first consider the water motion in an infinitely coiling channel with a fixed
bed and a rectangular cross-section. After some distance from the entrance, we
73
74 CHAPTER 5. FLOW AND MORPHOLOGY IN RIVER BENDS
may assume that the water motion no longer changes in the downstream direction
(axially symmetric flow). Next, we will consider the effects of this water motion
on the equilibrium profile in the case of a sandy bed. Finally, we will see what
happens in a bend with a finite angle of rotation and straight reaches upstream and
downstream.
5.2 Spiral flow
Consider a circular bend with axially symmetrical flow, but now taking the vertical
structure of the velocity field into account. The radial equation of motion reads:
!
u2 ∂ζ ∂ ∂v
− = −g + νt (5.1)
r ∂r ∂z ∂z
in which r denotes the radial coordinate in a polar coordinate system with its pole
in the center of the curvature of the bend, ζ is the water level, z is the vertical
coordinate, and νt is the eddy viscosity. The left-hand side of this equation represents
the centripetal acceleration needed to make the water follow the curvature of the
bend.
Averaging this equation over the water depth yields:
U2 ∂ζ τbr
− = −g − (5.2)
r ∂r ρh
in which U denotes depth averaged velocity and τbr is an inward directed secondary
bed shear stress. ζ can be eliminated from Equations (5.1) and (5.2) by simply
subtracting them from each other:
!
u2 − U 2 ∂ ∂v τbr
− = νt + (5.3)
r ∂z ∂z ρh
This can be considered as an equation in v(z), with the left-hand part as a forcing
term.
As far as the vertical distribution of this forcing term (Figure 5.1) is concerned, and
taking into account the fact the u(z) increases from the bed to the water surface,
we must have:
• u2 − U 2 > 0 in the upper part of the water column
• u2 − U 2 < 0 in the lower part of the water column
Solving v(z) from Equation (5.3) leads to the conclusion that this radial velocity
component is directed towards the outer bank in the upper part of the water column
5.2. SPIRAL FLOW 75
Figure 5.1: Forcing of the curvature-induced secondary flow
(z > 0.4h) and towards the inner bank in the lower part (z < 0.4h), such that the
depth-integrated value of v equals zero. This is qualitatively shown in Figure 5.2
(see Jansen et al., 1979, p. 62, for the exact vertical distribution). This circulation
is usually weak as compared to the downstream velocity.
Figure 5.2: Vertical distribution of the curvature-induced secondary flow
Thus the centripetal acceleration, together with the vertical structure of the down-
stream velocity u(z), gives rise to a cross-stream circulation, often named secondary
flow. When combined with the main (downstream) flow u(z) this leads to a spiraling
flow field, often called spiral flow (Figure 5.3).
Figure 5.3: Spiral flow in a river
Although the secondary flow is relatively weak, it has major consequences:
• Matter transported by the flow is dispersed much faster over the cross-section
than in a straight channel
• The secondary flow gradually redistributes the downstream flow momentum,

and therewith the downstream velocity, over the cross-section. As a conse-
quence, even if the river bed is horizontal, the velocity maximum gradually
shifts from the inner to the outer bend
• The secondary flow is attended by an inward directed secondary bed shear

stress, which can be described by:
√ !
g h2 g
τbr = −ρ 2 u2 2 1− (5.4)
C rκ κC
5.3. BED TOPOGRAPHY (AXIALLY-SYMMETRICAL) 77
where κ is the von Karman constant. If we combine this with the expression
for the downstream bed shear stress
g 2
τbφ = ρ u (5.5)
C2
we find for the bed shear stress direction relative to the channel axis:
√ !
h2 g
tgδ = − 2 1 − (5.6)
rκ κC
Apart from the constants, this angle is determined by the ratio of the water
depth and the bend radius of curvature. Usually, in natural bends δ is not
more than a few degrees.
5.3 Bed topography (axially-symmetrical)
One may readily assume that sand grains on a horizontal river bed tend to be
transported in the direction of the bed shear stress, rather than in the direction of
the depth-averaged flow. The grains “feel” the flow velocity and the shear stress
close to the bed.
If, for simplicity, we assume that the magnitude of the transport can be described
by a power law formula, i.e.
q p
qs = qsφ2 + q 2 = mun u2 + v 2
sr tot with utot = (5.7)
The transport components in the φ- and r-directions of a polar coordinate system

are:
qsφ = qs cos δ ≈ muntot (5.8)
√ !
n h 2 g
qsr = qs tgδ ≈ −mutot 2 1 − (5.9)
rκ κC
Thus the secondary flow tends to transport sediment from the outer to the inner
bend. This causes erosion in the outer bend and deposition in the inner bend, so
ultimately a cross-slope of the river bed. This change of bed topography influences
the water motion, but not to the extent that the inward directed transport ultimately
goes to zero. The cross-slope will therefore keep on steepening until it is steep enough
to make another transport mechanism reduce the cross-stream transport.
Sediment that is transported over a sloping bed will tend to roll downhill under
the influence of gravity. In addition to the above cross-stream transport induced
by the secondary bed shear stress, there is another component that is directed
downhill. In the axially symmetrical case considered herein, this concerns the radial
transport component, which will be proportional to the magnitude of the shear-

induced transport (the more grains are in motion, the greater the downhill sediment
flux) and the magnitude of the cross-slope (the steeper the slope, the greater the
downhill transport). In formula:
√ !
∗ n h 2 g ∂zb
qsr = qs tgδ + qsr ≈ −mutot 2 1 − − αmuntot (5.10)
rκ κC ∂r
in which α is a constant and zb is the bed level.
We now have two opposite transport components, of which at least one depends
strongly on the bed topography. If these two components balance each other through-
out the cross-section, there is no net transport and the profile is in equilibrium. This
is the case if:
∂zb h
= −β (5.11)
∂r r
where β is a dimensionless constant with a value of about 10.
By definition, the water level equals the bed level, zb , plus the water depth, h.
Hence:
∂zb ∂ζ ∂h
= − (5.12)
∂r ∂r ∂r
The cross-slope of the water surface is relatively small and can be neglected with
respect to the bed slope. In that case, Equation (5.11) leads to:
∂h h
≈β hence h ∼ rβ (5.13)
∂r r
In Figure 5.4 the resulting profile is outlined. The cross-stream variation of the
water depth, hence the bed level, is indeed much stronger than that of the water
level.
Figure 5.4: Equilibrium bed profile in a bend (axially symmetrical)
The formation of such a profile has a number of hydraulic engineering consequences,

such as:
5.3. BED TOPOGRAPHY (AXIALLY-SYMMETRICAL) 79
• reduction of the effective navigable width as compared with that in a straight

channel
• in the case of erodible banks: erosion of the outer bend, leading to bank erosion
and deposition of the erosion products in the outer bend (Figure 5.5)
• in the case of vertical bank protection: increased erosion in the outer bend
(h ∝ rβ extends through to the protection structure), which requires a deeper
foundation of the protection structure (Figure 5.5)
• near bridges: erosion around the bridge piers in the outer bend
Figure 5.5: Natural vs. protected outer bank

6 FLOW-CYLINDER INTERACTION AND LOCAL
SCOUR
6.1 Introduction
In the previous chapters, the river flow was considered to be unobstructed. Hence,
the only source of flow resistance was the bed drag. However, rivers have become
heavily regulated and numerous man-made structures disrupt the flow. Some exam-
ples of such structures are bridge piers and abutments, weirs, bed sills, and horizontal
pipes for the transportation of oil, gas, water, etc. The majority of the equations
that we have seen so far in this reader for the quantification of flow velocity and sed-
iment transport were developed for fully developed boundary layers in unobstructed
flows. Thus, they are not always valid in the vicinity of a structure that disrupts
the flow. Structures in a river alter the nearby flow field and may induce flow accel-
erations that lead to enhanced exerted bed shear stresses. This leads to an increase
in sediment transport and the development of local scour. In this course, only local
scour around bridge piers and horizontal pipes will be treated.
Local scour can be distinguished into clear-water scour and live-bed scour. The
difference between these two conditions is the occurrence of sediment transport
upstream of the structure. In clear-water scour, the approaching flow from the
upstream is not able to entrain sediment and local scouring occurs due to local flow
accelerations due to the presence of the structure. Equilibrium conditions in clear-
water conditions are reached when the scour hole obtains such a shape where the
bed shear stresses can no longer entrain sediment and local scouring seizes. In live-
bed scour, the incoming flow induces sediment transport and the scour hole around
the structure gets continuously supplied with sediment from upstream. At the same
time, sediment is removed from the scour hole due to the effect of the structure.
Equilibrium conditions in live-bed scour are attained when the rate of the upstream
sediment supply becomes equal to the rate of the sediment that is transported out
of the scour hole.
Since only scour around perpendicular cylinders (i.e., bridge piers) and horizontal
cylinders (i.e., pipes on the bed) will be treated in this course, it is important to
briefly analyze how the flow interacts with a cylindrical body. Figure 6.1 shows an
81
82 CHAPTER 6. FLOW-CYLINDER INTERACTION AND LOCAL SCOUR
upright cylinder mounted on the bed with incoming steady and fully developed flow.
As the flow approaches the cylinder, it gets decelerated and a portion of the flow is
accelerated around the cylinder, while another portion heads downward forming a
downflow. The latter is induced by a vertical pressure gradient along the cylinder,
since the stagnation pressures are greater higher in the water column. There are
two areas of interest in flow-cylinder interaction: high in the water column where
the effect of the bed is not so prominent and near the bed where the flow is affected
both by the cylinder and the bed. Higher in the water column, the portion of the
flow that gets diverted around the cylinder, initially gets accelerated due to the
contraction of the streamlines and gets separated at the rear side of the cylinder due
to adverse pressure gradient (i.e., pressure increases along the flow path) This leads
to the formation of lee-wake vortices, usually with the occurrence of vortex shedding
(i.e., the interchangeable flow separation at the two sides of the cylinder creates an
oscillating flow pattern that gets convected downstream), and the significant increase
of turbulence. Near the bed, the so-called horseshoe vortex is formed in front of the
cylinder due to the incoming flow that gets separated from the bed in front of the
cylinder due to adverse pressure gradient owed to the presence of the cylinder. The
combination of the horseshoe vortex and the downflow that meets the bed leads to
a vortical structure around the upstream face of the cylinder that gets convected
downstream increasing substantially the bed shear stress in front and around the
cylinder. The vortical structures will eventually deform downstream, but until this
happens there is increased scour potential in the vicinity of the cylinder.
Figure 6.1: Conceptual sketch of flow-cylinder interaction
It needs to be mentioned that local scour is still not well understood and constitutes
an active field of research, despite the fact that it has been studied by numerous
6.2. ESTIMATION OF MAXIMUM SCOUR DEPTH AT BRIDGE PIERS 83
researchers for decades. Thus, the predictive equations that are used in practice,
some of which will be presented here, are only able to provide rough estimates, which
Engineers must use cautiously.
6.2 Estimation of maximum scour depth at bridge piers
The accurate prediction of local erosion around bridge piers is of utmost importance
for the safe design and protection of bridges. Excessive scouring around the bridge
piers may lead to exposure of the foundations, which could compromise the stability
of the structure. Breusers et al. (1977) and Dey (2014) summarized the variables
that affect scour at piers and listed them as follows:
- Variables associated with the approaching flow from upstream: Flow velocity
and depth, shear velocity, roughness.
- Variables associated with the pier: Size and shape of the pier, orientation with
respect to the incoming flow direction, number of piers and distance between
them, existence or not of protective systems.
- Variables associated with the bed sediment: Median grain diameter and grain
size distribution, density of the sediment, angle of repose, cohesiveness or not
of the sediment.
- Variables associated with the fluid properties: Fluid density and viscosity,
acceleration due to gravity, temperature in case the free surface freezes.
- Variables associated with time: Time after scour has commenced for a scour
case that is evolving.
There are many empirical formulas in the literature for the computation of maximum
scour depth, ds , around a bridge pier. These equations were developed based on ex-
perimental data, mostly from laboratory flumes, and they are valid for conditions
similar to those they were derived. The lack of a wide range of field measurements
makes it hard to deduce which formulas are reliable and which perform better than
the rest. Nevertheless, there have been some formulas that have performed consis-
tently well, such as the K-factors methodology (Melville, 1997) and the method of
Sheppard et al. (2014).
Herein, only the K-factors methodology will be presented, based on Melville (1997)
and Dey (2014). According to this methodology, the maximum scour depth, ds ,
around a bridge pier can be formulated as the product of six different empirical
factors, named K-factors, which express the effect of most of the variables that were
listed above.
ds = Kh KI Kd Ks Ka Kt (6.1)
The K-factors are analyzed below.
• The factor Kh expresses the effect of pier width, b, on scour depth for a specific
flow depth, h, according to:


2.4b if b/h < 0.7



 √

Kh = 2 hb if 0.7 ≤ b/h < 5 (6.2)






4.5h if b/h ≥ 5

• The factor KI expresses the effect of flow intensity on scour, as the ratio of
the scour depth that will occur for the given depth-averaged flow velocity,
U , to the scour depth that occurs for the critical depth-averaged flow veloc-
ity, Ucr , which corresponds to the initiation
p of sediment motion. The factor
KpI can be estimated both for uniform ( D84 /D16 < 1.3) and non-uniform
( D84 /D16 ≥ 1.3) sediment, but here only the methodology for uniform sed-
iment is presented. The square root of the ratio D84 to D16 expresses the
geometric standard deviation of the sediment particle size distribution. Ad-
ditional information and the methodology for non-uniform sediment can be
found in Dey (2014).

U U
if <1


Ucr Ucr


KI = (6.3)


 U
1 if ≥1

Ucr
The critical depth-averaged velocity, Ucr , is estimated from:
!
Ucr h
= 5.75 log 5.53 (6.4)
u∗cr D50
where the critical shear velocity, u∗cr , is calculated with:


 1.4
0.0115 + 0.0125D50 for 0.1 ≤ D50 < 1 mm



u∗cr = (6.5)
 √ 1
0.0305 D50 − 0.0065 1 ≤ D50 < 100 mm

 for
D50
Note that in Equation (6.5) u∗cr is in m/s while D50 is in mm.
6.2. ESTIMATION OF MAXIMUM SCOUR DEPTH AT BRIDGE PIERS 85
• The factor Kd expresses the ratio of the scour depth that occurs for a value of
b/D50 to the maximum scour depth that is observed for values of b/D50 above
a certain threshold. Above this threshold, the ratio b/D50 does not affect the
maximum scour depth. The following equations are valid for uniform sediment.
Modifications to these equations for non-uniform sediment can be found in Dey
(2014).
For sand:
 !
 b
0.57 log 2.24 D50 for b/D50 ≤ 25



Kd = (6.6)




1 for b/D50 > 25
For gravel, there are additional equations by Raikar and Dey (2005):
 !
 b

0.6 log 3.88 for b/D50 ≤ 9
D50








 !
Kd = b (6.7)
0.184 log 14070 for 9 < b/D50 ≤ 25



 D50






1 for b/D50 > 25

• The factor Ks expresses the ratio of the maximum scour depth around a bridge
pier that has a specific shape with width b to the maximum scour depth around
a pier with circular cross-section and diameter also equal to b. The value of
Ks for the pier of interest can be selected from Table 6.1.
• The factor Ka expresses the ratio of the maximum scour depth around a bridge
pier with its longitudinal axis not aligned with the flow direction (i.e., not being
parallel to the flow direction) to the maximum scour depth around a bridge
pier of the same shape that is aligned with the flow direction. The factor Ka
is obtained from the following equation:


1 for piers with circular cross-section



Ka = !0.65 (6.8)
 bp
for piers with non-circular cross-section



 b
with bp the projected width of the oblique pier in the normal to the approaching
flow direction, as shown in Figure 6.2.
Table 6.1: Values of the shape factor Ks for pier scour for different pier shapes
(adapted from Dey, 2014)
Figure 6.2: Definition sketch for the parameter bp for the estimation of the alignment
factor Ka for pier scour (adapted from Dey, 2014)
• The factor Kt expresses the ratio of the maximum scour depth around a pier at
time t to the maximum scour depth at equilibrium conditions that are attained
after time te . For live-bed scour, this ratio is assumed to be equal to one as
the equilibrium conditions occur rather soon, while for clear-bed scour the
following equation can be used.



1 for live-bed scour



Kt = (6.9)
 !1.6 
U t
−0.03 cr ln

exp  for clear-water scour


U te



where te can be estimated from the following:

6.3. ESTIMATION OF SCOUR DEPTH BELOW HORIZONTAL PIPES 87
 !
 b U h U
48.26 − 0.4 for > 6 and > 0.4





 U Ucr b Ucr

te = (6.10)
 ! !0.25
b U h h U




30.89 − 0.4 for ≤ 6 and > 0.4
U Ucr b b Ucr


Note that the output of te from Equation (6.10) is in days.
6.3 Estimation of scour depth below horizontal pipes
Besides the interaction of flow with an upright cylinder (e.g., a bridge pier), the flow
may interact with cylinders that are laid horizontally on the bed. Such cylinders are
usually pipes that transfer oil, gas, water, etc. The estimation of scour depth in such
cases is needed to determine the sections of the pipe that will become unsupported
by the bed. This is particularly important because the pipe may not have adequate
support for its own weight and the weight of the material it transports. In addition,
if the bed gets scoured deep enough, there will be a manifestation of vortex shedding
on the pipe, which will cause the pipe to vibrate and enhance the pressure load on it.
This may lead to fatigue failure, which can potentially have devastating consequences
for the pipeline system and for the local ecosystem in case the pipe transfers oil.
To understand the mechanism that induces scour below a horizontal pipe with cir-
cular cross-section, consider steady flow conditions and the pipe laid on the bed
normal to the flow direction. Due to pressure difference between the contact points
of the pipe and the bed at the upstream and downstream sides of the pipe, there is
a subsurface flow below the pipe. As the flow velocity increases, the subsurface flow
below the pipe strengthens and at some point the sediment at the downstream side
of the pipe will be dislodged. This is the scour threshold and this phenomenon is
called piping. The small gap that has been developed below the pipe due to piping
becomes gradually larger, as there is flow acceleration in the gap with elevated bed
shear stresses that enhance erosion. This process is known as tunnel erosion and
it gradually becomes less intense as the gap below the pipe increases in size. The
widening of the gap stops when the exerted bed shear stress becomes equal to the
critical bed shear stress for clear-water scour or when the exerted bed shear stress
becomes equal to the upstream bed shear stress for live-bed scour. When the gap
becomes large enough, vortex shedding occurs from the flow-cylinder interaction
and the generated eddies entrain sediment at the downstream until an equilibrium
condition is reached.
Similar to the bridge pier case, there is an abundance of empirical formulas in the
literature for the estimation of maximum scour depth below the pipe. Herein, the
estimation of the maximum scour depth will be done with the iterative methodology
of Chiew (1991), which is presented as analyzed by Dey (2014).This methodology is
valid for clear-water conditions and is described in the following steps:
1. Estimate the gap discharge per unit width, qg , with the empirical relationship
of Chiew (1991) and Dey and Singh (2007):
!0.787
b
qg = 0.781q (6.11)
h
where q is the upstream discharge per unit width.
2. Assume a value for maximum scour depth, ds . Based on this value, calculate
the average flow velocity, Ug , in the gap below the pipe.
qg
Ug = (6.12)
ds
3. Calculate the corresponding bed shear stress, τb , with:
fDW 2
τb = ρUg (6.13)
8
where ρ is the water density and fDW is the Darcy-Weisbach coefficient, which
can be determined from the Moody diagram. To facilitate the iterative process,
we can alternatively use the formula of Haaland (1983), which was developed
for pipe flows. Assuming that the flow in the gap below the pipe is similar to
a flow in a pipe and that the scour depth ds corresponds to the pipe diameter
and the velocity Ug corresponds to the mean flow velocity across the pipe
cross-section, we get:
 !1.1 
1 ks 6
0.5 = −0.782 ln  +  (6.14)
fDW 3.7ds Reg
where ks can be considered equal to D50 for relatively uniform sediment and
Reg = Ug ds /ν, with ν the kinematic viscosity of water.
4. We aim to determine the maximum scour depth, ds , which occurs in equi-

librium conditions. In clear-water scour, equilibrium conditions are attained
when τb reaches τbcr , where τbcr is the critical bed shear stress, which is ob-
tained from the Shields diagram. In case the computed bed shear stress from
step 3 is not equal to the critical bed shear stress we repeat the process by
assuming a different value for maximum scour depth in step 2.
REFERENCES
Albertson, M. (1953). Effect of Shape on the Fall Velocity of Gravel Particles. State
University of Iowa.
Breusers, H. N. C., Nicollet, G., and Shen, H. W. (1977). Local scour around
cylindrical piers. Journal of Hydraulic Research, 15(3):211–252.
Chiew, Y.-M. (1991). Prediction of maximum scour depth at submarine pipelines.

Journal of Hydraulic Engineering, 117(4):452–466.
Dey, S. (2014). Fluvial Hydrodynamics – Hydrodynamic and Sediment Transport

Phenomena. Springer.
Dey, S. and Singh, N. P. (2007). Clear-water scour depth below underwater pipelines.
Journal of Hydro-environment Research, 1(2):157–162.
Einstein, H. A. and Barbarossa, N. L. (1952). River channel roughness. Transactions

of the American Society of Civil Engineers, 117(1):1121–1132.
Engelund, F. and Hansen, E. (1967). A monograph on sediment transport in alluvial

streams. Teknisk Forlag, Copenhagen, Denmark.
Fox, R. W. and McDonald, A. T. (1992). Introduction to Fluid Mechanics. John

Wiley and Sons, Inc., New York, USA.
Galappatti, G. and Vreugdenhil, C. B. (1985). A depth-integrated model for sus-

pended sediment transport. Journal of Hydraulic Research, 23(4):359–377.
Garcia, M. H. (2008). Sedimentation Engineering: Processes, Measurements, Model-

ing, and Practice, ASCE Manuals and Reports on Engineering Practice No. 110,
chapter Sediment Transport and Morphodynamics, pages 21–163. ASCE.
Haaland, S. E. (1983). Simple and explicit formulas for the friction factor in turbulent
pipe flow. Journal of Fluids Engineering, 105(1):89–90.
Jansen, P. P., van Bendegom, L., van den Berg, J., de Vries, M., and Zanen, A.
(1979). Principles of river engineering: the non-tidal alluvial river. Water Re-
sources Engineering Series. Pitman.
Melville, B. W. (1997). Pier and abutment scour: Integrated approach. Journal of

Hydraulic Engineering, 123(2):125–136.
Meyer-Peter, E. and Müller, R. (1948). Formulas for bed-load transport. In Proceed-

ings of the 2nd meeting of the International Association for Hydraulic Research,
pages 39–64, Stockholm, Sweden.
89
90 REFERENCES
Raikar, R. V. and Dey, S. (2005). Clear-water scour at bridge piers in fine and
medium gravel beds. Canadian Journal of Civil Engineering, 32(4):775–781.
Sheppard, D. M., Melville, B., and Demir, H. (2014). Evaluation of existing equa-
tions for local scour at bridge piers. Journal of Hydraulic Engineering, 140(1):14–
23.
Shields, A. F. (1936). Application of similarity principles and turbulence research to

bedload movement. Mitt. Preuss. Versuchsanst. Wasser., 26, Transl. into English
by Ott, W.P. and van Uchelen, J.C. California Institute of Technology.
Simons, D. B. and Richardson, E. V. (1966). Resistance to flow in alluvial channels.

US Government Printing Office.
van Rijn, L. C. (1984a). Sediment transport, part I: Bed load transport. Journal of
Hydraulic Engineering, 110(10):1431–1456.
van Rijn, L. C. (1984b). Sediment transport, part II: Suspended load transport.
Journal of Hydraulic Engineering, 110(11):1613–1641.
van Rijn, L. C. (1990). Principles of fluid flow and surface waves in rivers, estuaries,
seas and oceans. Aqua Publications, Amsterdam, The Netherlands.
van Rijn, L. C. (1993). Principles of sediment transport in rivers, estuaries and

coastal seas. Aqua Publications, The Netherlands.

River Flow Processes Reader2021

Uploaded by

Copyright:

Available Formats

River Flow Processes Reader2021

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

River Flow Processes Reader2021

Uploaded by

Copyright:

Available Formats

River Flow Processes

River Flow Processes

Taught by: Vasileios Kitsikoudis and Fredrik Huthoff

Water Engineering and Management (WEM)

de Vriend. H.J. (2005). River Engineering. Reader Ct3340, TU Delft, Faculty of

2 Turbulence and roughness 17

2.2 General basic equations . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.1 Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . 18

2.2.2 Conservation of momentum . . . . . . . . . . . . . . . . . . . 19

2.3 Reynolds averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.4 Steady uniform flow with a free water surface . . . . . . . . . . . . . 26

2.4.2 Continuity equation and equation of motion . . . . . . . . . . 27

2.4.3 Hydrostatic pressure distribution . . . . . . . . . . . . . . . . 28

2.5 Flow resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.6 Roughness and the velocity profile . . . . . . . . . . . . . . . . . . . 31

2.7 Composite cross-sections . . . . . . . . . . . . . . . . . . . . . . . . . 35

3 Sediment transport and bed forms 39

3.2 Sediment characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.2.1 Size and shape . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.2.2 Chemical composition and density . . . . . . . . . . . . . . . 42

3.2.3 Fall velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.3 Sediment transport processes . . . . . . . . . . . . . . . . . . . . . . 45

3.3.1 The transport mechanism . . . . . . . . . . . . . . . . . . . . 45

3.3.2 Incipient motion . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.3.3 Bed forms and alluvial roughness . . . . . . . . . . . . . . . . 49

3.4 Sediment transport formulas . . . . . . . . . . . . . . . . . . . . . . . 51

3.4.2 Bed load formula of Meyer-Peter and Müller (1948) . . . . . 53

3.4.3 Bed-material load formula of Engelund and Hansen (1967) . 53

3.4.4 Bed and suspended load formulae of van Rijn (1984a,b) . . . 54

3.5 Remarks on the use of sediment transport formulas . . . . . . . . . . 55

4 Suspended sediment transport 57

4.1 Local equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

4.1.1 Settling and turbulence mixing . . . . . . . . . . . . . . . . . 57

4.1.2 Concentration profiles . . . . . . . . . . . . . . . . . . . . . . 58

4.1.3 Bottom concentration . . . . . . . . . . . . . . . . . . . . . . 60

4.1.4 Alternative boundary conditions . . . . . . . . . . . . . . . . 60

4.2 Non-equilibrium suspended sediment transport . . . . . . . . . . . . 61

4.2.1 Concentration equation . . . . . . . . . . . . . . . . . . . . . 61

4.2.2 Initial and boundary conditions . . . . . . . . . . . . . . . . . 64

4.3 Depth-averaged suspended load equation . . . . . . . . . . . . . . . . 68

5 Flow and morphology in river bends 73

5.2 Spiral flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.3 Bed topography (axially-symmetrical) . . . . . . . . . . . . . . . . . 77

6 Flow-cylinder interaction and local scour 81

6.2 Estimation of maximum scour depth at bridge piers . . . . . . . . . 83

6.3 Estimation of scour depth below horizontal pipes . . . . . . . . . . . 87

Q flow discharge [L3 T−1 ]

σ normal stress [M L−1 T−2 ]

2.2 General basic equations

2.2.1 Conservation of mass

velocities changed to ux+∆x , vy+∆y , and wz+∆z :

ṁout = ρux+∆x ∆y∆z + ρvy+∆y ∆x∆z + ρwz+∆z ∆x∆y (2.4)

The mass within the control volume is equal to:

and its time derivative is:

ρux+∆x − ρux ρvy+∆y − ρvy ρwz+∆z − ρwz ∂ρ

∂ (ρu) ∂ (ρv) ∂ (ρw) ∂ρ

2.2.2 Conservation of momentum

For the derivation of the equations of motion (or conservation of momentum) in

ΣFx = mαx (2.10a)

ρb = ρs (1 − ) = 1590 kg/m3 (3.3)