1

HELSINKI UNIVERSITY OF TECHNOLOGY

LABORATORY OF WATER RESOURCES
YHD-12.2010
HYDRAULICS
Computational aspect of flow in
open channels
JAN-2002
Tuomo Karvonen
Prof.
2
CONTENTS
1 INTRODUCTION
2 BASIC DEFINITIONS
2.1 Introduction
2.2 Classification of flow types
2.3 Energy, Bernoulli equation, specific energy
2.4 Friction losses, friction slope
2.5 Continuity equation in steady flow
3 COMPUTATION OF STEADY UNIFORM FLOW
3.1 Introduction
3.2 Trial-and-error method (algebraic method)
3.3 Newton-Raphson method
3.4 Graphical method
3.5 Exercises
4 GRADUALLY VARYING NON-UNIFORM FLOW
4.1 Introduction
4.2 Computation of non-uniform flow in one river reach
4.3 Computation of non-uniform flow in longer river reaches
4.4 Rating curves for non-uniform flow calculations
4.5 Exercises
5 UNSTEADY FLOW IN OPEN CHANNELS
5.1 Introduction
5.2 Derivation of the Saint Venant equations
5.2.1 Conservation of mass (continuity equation)
5.2.2 Conservation of momentum
5.2.3 Exercises
5.3 Simplifications of the full Saint Venant equations
5.3.1 Kinematic approximation
5.3.2 Diffusion analogy
5.3.3 Exercises
5.4 Numerical solution of the full Saint Venant equations
5.4.1 History
5.4.2 Introduction to finite difference methods
6 BED ROUGHNESS, FRICTION AND VELOCITY DISTRIBUTION IN UNVEGETATED AND
VEGETATED CHANNELS
6.1 Introduction
6.2 Laminar, turbulent and bottom shear stresses
6.2.1 Shear stress in laminar flow
6.2.2 Shear stress in turbulent flow
6.2.3 Total shear stress
6.2.4 Bottom shear stress
6.2.5 Friction velocity
6.3 Velocity distribution, bed roughness and friction factors in unvegetated channels
6.3.1 Viscous sublayer
6.3.2 Scientific classification of flow layers in unvegetated channels
6.3.3 Bed roughness k
s
in unvegetated channels
6.3.4 Characterisation of smooth and rough flow in unvegetated channel
6.3.5 Logarithmic velocity distribution in turbulent layer
6.3.6 Velocity profile in the viscous sublayer
6.3.7 Drag force
6.3.8 Chzy, Manning and Darcy-Weissbach friction coefficients
6.3.9 Velocity distribution in unvegetated channels
3
6.3.10 Exercise
6.4 Velocity distribution, roughness height and friction factors in vegetated channels
6.4.1 Unsubmerged-submerged flow, deflected-undeflected vegetation
6.4.2 Classification of flow layers in vegetated channel
6.4.3 Roughness height k
p
in vegetated channels
6.4.4 Velocity distribution in vegetated channels
6.4.5 Influence of vegetation density on friction factors in vegetated channels
6.4.6 Velocity distribution in the case of submerged vegetation
6.4.7 Exercise
6.5 Interaction between main channel flow and floodplain flow
6.5.1 Momentum transfer and shear stress between main channel and flood plain
6.5.2 Exercise
6.6 Methods used to evaluate the influence of vegetation on conveyance capacity of the main channel
and the floodplains
6.6.1 Calculation of composite roughness using iterative method
6.6.2 Comparison of two different methods to evaluate the influence of effect of floodplain on main
channel conveyance
REFERENCES
4
1 INTRODUCTION
What is hydraulics? Hydraulics can be defined as applied hydromechanics.
The material introduced in this course deals with:
flow and flow-related phenomena in artificial and natural channels
flow in pipes and pipe systems
The course is concentrating on computational aspects of flow in channels and pipes. The
teaching material of this course (theory and exercises) is available in Web-pages.
The theory is presented in very condensed form and therefore it does not does not replace
the standard hydraulic text books.
Hydrostatics is included in another course and it is not treated in this material. Moreover,
hydraulic structures, pumps and turbines are not included.
List of symbols for "Open channel hydraulics" course material
A: cross-sectional area of the flow (m
2
) or projectional area in drag force calculations
A
1
: parameter in Eq. (6-55)
A
3
: parameter in Eq. (6-54) (1.84 used)
A
M
: cross-sectional area of the main channel (m
2
)
A
v
: horizontal projected area of vegetation (perpendicular to flow
direction)
a
x
: distance of canopy elements in the direction of flow (m)
a
y
: distance of canopy elements in the direction perpendicular to flow (m)
a
z
: distance of branches in vertical direction (see Eq. (6-73))
A': the projected area of the grain to the horizontal plane (m
2
)
B: auxiliary parameter in Mertens' method, Eq. (6-86)
B
1
: parameter in Eq. (6-55) (1.3..2.0)
B
3
: parameter in Eq. (6-54) (0.35 used)
5
b
B
: auxiliary variable in Nuding's method to calculate b
EFF
(see Eq. (6-84))
b
c
: auxiliary variable in Nuding's method to calculate b
EFF
(see Eq. (6-84))
b
EFF
: effective width of the floodplain in Nuding's method (m)
b
EFF,M
: effective width in Mertens' method (m)
b
M
: surface width of the main channel (m)
b
N
: auxiliary variable in Nuding's method to calculate b
EFF
(see Eq. (6-84))
b
M,N
: test parameter for surface width of the main channel (m) in Nuding's method
b
W,F,max
: width of floodplain when maximum value of the f
I,C
is obtained
C: Chezy coefficient
c: parameter c in Mertens' method, see Eq. (6-87)
C
1
: empirical coefficient (1.52 < C
1
< 2.69) in Kouwen et al. /Eq. (6-45)
C
2
: empirical coefficient in Kouwen et al. /Eq. (6-47)
C
D
: drag coefficient (-)
C
L
: lift coefficient (-)
C
W
: drag coefficient of vegetation (-)
d: grain diameter (m) ; unless specified, assume that d=d
50
D : pipe diameter (=4R) (m)
d
50
: grain diameter d
50
, which is the 50 % point in the grain size distribution curve (m)
D
eff
: effective hydraulic diameter
d
g
: characteristic grain diameter (m); unless specified, assume that d=d
50
d
p
: diameter of the canopy element (m)
d
p,e
: equivalent diameter of the canopy element (m), see Eq. (6-74)
d
z
: thickness of the branches of plants (see Eq. (6-73))
f : bottom friction coefficient
f
BF,C
: friction factor of the bottom of the floodplain in the combination method
f
BF,N
: friction factor of the bottom of the floodplain in Nuding's method
f
DW
: Darcy-Weissbach friction coefficient
F
D
: drag force (N)
f
I,C
: influence of the interaction on the imaginary wall friction factor/combination method
f
I,C,max
: maximum value of the influence of the interaction on the imaginary wall friction
6
factor (obtained when true width b
W,F,max
)
f
IW
: friction coefficient of the imaginary wall between main channel and floodplain (-)
f
IW,C
: the total friction factor of the imaginary wall/the combination method
f
IW,N
: the total friction factor of the imaginary wall/Nuding's method
f
IW,M
: the total friction factor of the imaginary wall/Mertens' method
F
L
: lift force (N)
f
PL
: empirical correction factor, D
eff
= f
PL
D
f
S,0
: vegetation induced shape friction in the combination method
f
V
: vegetation friction coefficient calculated from Eq. (6-61) (-)
f
V,C
: vegetation friction coefficient in the combination method
f
V,N
: vegetation friction coefficient in Nuding's method
f
V,C
: vegetation friction coefficient in Mertens' method
h: water depth (m)
h
F
: water depth (m) at the location of the interaction zone: Nuding's method, Eq. (6-83)
h
I,L
: water depth (m) at the interface between main channel and left-side floodplain
h
I,R
: water depth (m) at the interface between main channel and right-side floodplain
h
p
: deflected height of vegetation (m)
h
p,m
: characteristic height in Eq. (6-60); can be taken as h
p
-h'
H
r
: ripple height (m); unless specified assume that H
r
=100d
50
h' : shift of the zero point from h
p
in logarithmic velocity distribution (see Fig. 6-10)
h'': zero-plane displacement in submerged vegetation h'' = h
p
-h'
I
hp
: mixing length calculated from Eq. (6-58)
k
IW
: the roughness height of the imaginary wall between main channel and floodplain
k
JW,M
: the total roughness height of the imaginary wall, see Eq. (6-88)
k
p
: roughness height of submerged vegetation (m)
k
s
: bed roughness (m); unless specified: for flat bed ks=2.5d
50
and for rippled bed k
s
=H
r
k
St
: Strickler coefficient (=1/n)
L: length of the river section (m)
l
m
: Prandtl's mixing length (m)
7
L
H
: Prandtl's mixing length in Eq. (6-53) (m)
L
r
: length of ripples (m), unless specified, assume L
r
=1000d
50
.
n: Manning coefficient
P: wetted perimeter (m)
P
I,T
: total length of the interfaces between main channel and the floodplains
P
M
: wetted perimeter of the main channel (bottom)
p: porosity of the bed material (-)
q: discharge in cubic meters per meter width unit width (m
2
s
-1
)
R: hydraulic radius (m)
Re: Reynolds number (-)
R
F
: hydraulic radius of the floodplain (=A
M
/P
M
)
s: relative density is denoted as s=
s
/
S: slope of the energy line (-)
S
o
: bottom slope (-)
S
f
: friction slope (-)
t: time (s)
V: the instantaneous velocity component in horizontal (x) direction (m s
-1
)
v: the symbol used for time averaged horizontal velocity of water is v (m s
-1
)
v(z): velocity profile, i.e. z versus v (m s
-1
)
v
A
: depth averaged velocity calculated using Eq. (6-34)
v
ABOVE
: average velocity above the canopy in the case of submerged vegetation (m s
-1
)
v
A,S
: depth averaged velocity in the case of submerged vegetation, Eq. (6-66) (m s
-1
)
v
F,C
: velocity in the floodplain/combination method
v
F,N
: velocity in the floodplain/Nuding's method
v
F,M
: velocity in the floodplain/Mertens' method
v
hp
: velocity just at the top of the deflected height h
p
(m s
-1
)
v
hp,T
: velocity at the top of the deflected canopy as calculated by (6-58)
v
IV
: average velocity inside the canopy (m s
-1
)
v
M,0
: velocity in the main channel when friction in the interaction zone is neglected
v
p
(z): velocity profile inside the canopy (m s
-1
)
8
v
p
': characteristic velocity inside the canopy calculated from Eq. (6-58) (m s
-1
)
v
s
: settling velocity of a single grain (m s
-1
)
v
V
: velocity in unsubmerged vegetation calculated from Eq. (6-62) (m s
-1
)
v
V,S
: velocity in submerged vegetation calculated from Eq. (6-63) (m s
-1
)
v': instantaneous velocity fluctuation in horizontal direction (m s
-1
)
v
*
: friction velocity (m s
-1
) calculated from Eq. (S-15)
W: the instantaneous velocity component in vertical (z) direction (m s
-1
)
w: the symbol used for time averaged vertical vel. of water is w (m s
-1
)
w': instantaneous velocity fluctuation in vertical direction (m s
-1
)
W': downward submerged weight (N)
x: distance in the horizontal direction (m)
y: distance in the direction perpendicular to flow direction (m)
z: vertical co-ordinate (m)
z
0:
the point of zero velocity (m)
: empirical coefficient when calculating velocity near bottom (<1.0) (-);
empirical parameter in the method of Tsujimoto et al./Eq. (6-58)

1
: empirical parameter in (6-53)
: angle between channel bottom and horizontal line (rad); empirical
parameter in the method of Tsujimoto et al./Eq. (6-58).

:
theoretical thickness of the viscous sublayer (m)
: kinematic eddy viscosity (m
2
s
-1
)

p
: Kaiser's volume ratio in Eq. (6-76)
: von Karman constant (=0.40)
: density of water; usually around 1000 kg m
-3

s
: density of natural sediments ; usually a value 2650 kg m
-3
is used
: kinematic viscosity of water (10
-6
m
2
s
-1
in 20 C)

b
: bottom shear stress (N m
-2
)

b,I
: shear stress in the interface between main channel and floodplains (N m
-2
)

b,L
: shear stress in the interface between main channel and floodplain/left side (N m
-2
)
9

b,R
: shear stress in the interface between main channel and floodplain/right side (N m
-2
)

b,m
: bottom shear stress in the main channel (N m
-2
)

t
: shear stress in turbulent flow (N m
-2
)

: shear stress in laminar flow (viscous shear stress) (N m

-2
)

z
: shear stress in vertical direction (N m
-2
)

1
: vegetation parameter (usually is around 0.40-1.56) in Eq. (6-52)

v
: vegetation density in Eq. (6-58) and (6-61)

v,C
: vegetation density in the combination method

v,C
: vegetation density in Nuding's method, Eq. (6-73)
10
2 BASIC DEFINITIONS
2.1 Introduction
Rivers/channels can be either natural or man-made (artificial). The cross-section of the
natural channel is almost always irregular, i.e. cross-sectional area of flow changes along
the length of the channel. The cross-section of man-made channel is usually regular with
trapezoidal cross section with constant bottom width B
B
, and constant side slopes m (e.g.
1:2). Cross-section and longitudinal section of man-made channels and for natural
channels are shown in Fig. 2-1.
Fig. 2-1. Cross-section and longitudinal section of man-made channels and cross-section
for natural channels.
The definitions and symbols used throughout this material are:
h = water depth (flow depth) (m)
A= area of the flow cross-section (m
2
)
B = top (surface) width of the channel (m)
B
B
= bottom width of trapezoidal channel
P = wetted perimeter (m), length of line in contact with water
11
R = hydraulic radius =A/P (m)
D = hydraulic depth (characteristic depth) = A/B
z = elevation above a reference datum (m)
L = length of a river section (m)
S
o
= bottom slope (m m
-1
); z
B
= LS
o
h
f
= friction loss (m)
S
f
= friction slope (slope of the energy line) (m m
-1
); S
f
= h
f
/L
Q = discharge (m
3
s
-1
)
v = velocity (average) =Q/A (m s
-1
)

b
= shear stress (N m
-2
)
Channel cross-section is said to be prismatic if the cross-section does not vary and the
bottom slope and friction coefficient remain constant.
B / A D
P / A R
vA Q

(2-1)
where Q is discharge (m
3
s
-1
), v is average velocity (m s
-1
), A is cross-sectional area of the
flow (m
2
), R is hydraulic radius (m), P is wetted perimeter (m), D is hydraulic depth (m)
and B is the surface width of the cross-section. For very wide channels the wetted
perimeter is close to bottom width and in this case R can be assumed to be equal to water
depth h. Bottom slope S
o
= z
B
/L (m m
-1
) is bottom drop z
B
(m) divided by the distance L
(m). Friction slope S
f
= h
f
/L (m m
-1
) is friction loss h
f
(m) divided by the distance L.
Density of water, dynamic and kinematic viscosity
Density of water (kg m
-3
) and specific weight W (N m
-3
) are functions of temperature
and the values are given in Table 2-1. Viscosity is a measure of the internal friction of a
12
fluid, or its resistance to flow and movement (Hamill, 2001). Dynamic viscosity (Ns m
-2
or kg m
-1
s
-1
) and kinematic viscosity =/ (m
2
s
-1
) are also shown in Table 2-1.
________________________________________________________________________
Table 2-1. Density, specific weight, dynamic and kinematic viscosity of water as a
function of temperature.
__________________________________________________________________
Temp. Density Specific Viscosity
weight Dynamic Kinematic
C kg m
-3
N m
-3
Ns m
-2
m
2
s
-1
____________________________________________________________________
0 999.8 9804.7 1.7810
-3
1.7810
-6
4 1000.0 9806.7 1.5010
-3
1.5010
-6
10 999.6 9802.7 1.3010
-3
1.3010
-6
20 998.2 9789.0 1.0010
-3
1.0010
-6
30 995.6 9763.5 8.0210
-4
8.0610
-7
100
*
958.3 9397.7 2.8210
-4
2.9410
-7
____________________________________________________________________
* as liquid
2.2 Classification of flow types
Laminar and turbulent flow, Reynolds number
Laminar flow is associated with slow velocity and the paths of individual water particles
do not intersect as shown in Fig. 2-2. Laminar flow is observed very seldom in open
channels. In turbulent flow the paths of the water particles intersect; the flow is chaotic.
There is a so called transitional flow between laminar and turbulent flow.
13
The classification of flow to laminar or turbulent is based on a dimensionless parameter
called the Reynolds number Re:

vD
Re (2-2)
where hydraulic depth D is sometimes replaced by hydraulic radius R or flow depth h. In
open channels the flow is laminar if Re < 500 and turbulent if Re > 2000. Flow is
transitional if 500 Re 2000.
Fig. 2-2. Laminar and turbulent flow.
Subcritical, critical and supercritical flow
The Froude number Fr is used to classify if flow is subcritical, critical or supercritical:
gD
v
Fr (2-3)
where v is average velocity in the channel (m s
-1
) and D is the hydraulic depth (=A/B) of
the cross-section. Flow is subcritical if Fr <1.0, supercritical if Fr>1.0 and critical if
Fr=1.0.
Flow transition from subcritical to supercritical is possible only if average velocity
increases much enough. This can be caused by an increase in the bed slope or a reduction
14
in the area of flow. Change from supercritical to subcritical usually happens very
smoothly.
Flow transition from supercritical to subcritical includes e.g. cases when bottom slope
changes from a steep to mild one. Supercritical flow in the steep section has to change to a
slow, subcritical flow. This change cannot happen smoothly and gradually. Instead there is
a sudden increase in depth in the form of a hydraulic jump (Fig. 2-3).
Fig. 2-3. Hydraulic jump when flow changes from supercritical to subcritical.
Subcritical and supercritical flows differ from each other in the respect that in subcritical
flow disturbance travels both in upstream and downstream direction, but in supercritical
flow the disturbance travel only downstream. In other words, in subcritical case flow is
controlled by a downstream point (overflow dam, weir, hydraulic drop) and the
computations must be done from downstream towards the upstream end. In supercritical
flow the computations must be done towards the downstream.
Uniform and non-uniform flow
In uniform flow the cross-section does not vary along the distance and the bottom slope
equals the friction slope, i.e. S
o
= S
f
. This implies that the slope of the water surface equals
the slope of the channel bottom and water depth h, velocity v and discharge Q are constant
throughout the river reach. Uniform flow is possible only in man made channels. The
water depth related to uniform flow is called normal depth h
N
.
15
Fig. 2-4. Uniform and non-uniform flow (the shape of the cross-section is the same in both
cases).
Flow is non-uniform if the depth h (as well as A, R and P) vary along the length of the
channel. Flow is always non-uniform in natural channels. In man made channels flow is
non-uniform if flow depth changes in the longitudinal section even if the shape of the
cross-section does not change.
Steady and unsteady flow
Flow is steady (stationary) if discharge Q does not change with respect to time. In steady
flow the average velocity of flow, v, and flow depth of a certain cross-section do not
change with time. However, if the channel cross-section varies along the distance, flow
depth and flow velocity change with respect to distance. In unsteady flow discharge Q and
other flow factors (h and v) vary with time. In natural channels the flow is almost all the
time unsteady, although e.g. during low flow conditions the change in Q happens very
slowly.
Gradually and rapidly varied flow
If flow is gradually varied, the depth h and other flow factors vary smoothly from section
to another. In rapidly varied flow, depth h and other flow factors change abruptly over a
16
very short distance. Discontinuity of flow is also possible. Rapidly varied flow can be seen
e.g. at a weir, at a change in channel width, at a hydraulic jump or a hydraulic drop.
Classification of water surface profiles
Water surface longitudinal profiles can be characterised according to bottom slope S
o
.
According e.g. to Graf and Altinakar (1998) five different cases can be distinguished:
M : Channels on mild slope; S
o
< S
C
S : Channels on steep slope; S
o
> S
C
C : Channels on critical slope; S
o
= S
C
H : Channels on horizontal slope S
o
= 0
A : Channels on adverse slope S
o
< 0
In mild slopes (M), the bottom slope is smaller than the critical slope S
C
. Type M1 flows
(see Fig. 2-5).
Fig. 2-5. Different flow types.
17
2.3 Energy, Bernoulli equation, specific energy
Energy is defined as the capacity for doing work and work is defined as the force
multiplied by the distance moved in the direction of force. The unit of work is Nm. In
hydraulics four different forms of energy are usually considered:
potential energy
kinetic energy
pressure energy
energy losses due to friction
In physics potential and kinetic energy of an object with mass m (kg) are calculated as
2
K
p P
mv
2
1
E
mgz E

(2-4)
From (2-4) it is possible to calculate energy in terms of height, i.e. z
P
and a so called
velocity head z
K
.
g 2
v
mg
E
z
mg
E
z
2
K
K
P
p

(2-5)
Eq. (2-5B) is important because it converts velocity to unit of height (m) and z
P
and z
K
can
be easily compared and used in the same equations.
Bernoulli equation
Bernoulli equation can be used to study the conservation of energy between two cross-
sections 1 and 2. Water depths and velocities are h
1
, v
1
, h
2
and v
2
, respectively. Energy is
conserved and hence
18
L
h
S ; L S z
h
g 2
v
h L S
g 2
v
h
f
f o B
f
2
1
1 o
2
2
2

+ + + +
(2-6)
The coefficient in (2-6) is the so called velocity distribution coefficient, which takes
into account the fact that in channel cross-section the distribution of velocity is not
uniform. The coefficient always has a value 1.00 or greater and in many cases it is omitted
in the calculations. Basically it is possible to calculate by measuring velocity v
i
within
smaller subsections dA
i
of the flow and calculating by (Hamill, 2001):
A v
) dA v (
3
N
1 i
i
3
i

(2-7)
In this material it is usually assumed that =1.0.
Fig. 2-6. Principle of the Bernoulli equation.
19
Specific energy, critical flow, critical depth
Specific energy E calculated with respect to bottom of the channel is the sum of the water
depth h and velocity head v
2
/(2g) (=1.0), i.e.
2
2 2
gA 2
Q
h
g 2
v
h E + + (2-8)
The flow is said to be critical, when the specific energy E is at minimum:
gD
v
1
D gA
Q
1
gA
B Q
1
dh
dA
gA
Q
1
dh
dE
2
2
2
3
2
3
2
(2-9)
Note that B=dA/dh (Fig. 2-7) and hydraulic depth D=A/B.
Fig. 2-7. Surface width B defined as dA/dh.
Specific energy is at minimum when dE/dh = 0, i.e.
2
D
g 2
v
or
gD v
or
0
gD
v
1
C
2
C
C
C
C
2
C

(2-10)
20
The depth when specific energy is at minimum in a given cross-section is called the
critical depth h
C
and correspondingly the velocity v
C
is called the critical velocity, and D
C
the critical hydraulic depth. Moreover, critical slope S
C
is defined as the slope that results
in uniform flow at the critical depth.
2.4 Friction losses, friction slope
General equation for flow velocity v in an open channel can be expressed as:
b
f
a
S CR v (2-11)
where C is the resistance coefficient, R is hydraulic radius, S
f
is the energy (friction) slope,
and a and b are empirical constants.
Consider now the channel longitudinal section shown in Fig. (2-8). The force causing the
flow is the component Gsin(), where G is the weight of the whole water mass of the
section and sin() is the component in the direction of flow ( is the angle between
horizontal line and channel bottom). The force resisting is the friction acting on the whole
boundary between water body and bottom and banks of the section. These forces must
balance each other. The angle is usually so small that sin() can be replaced by
tan()=(z
B
)/L=S
o
. If energy slope S
f
is close to bottom slope S
o
, then tan() can be
replaced by S
f
.
PL Kv ALS ) sin( G
2
f
(2-12)
where is the specific weight of water, K is empirical coefficient, P is wetted perimeter
and L is the length of the river section.
21
Fig. 2-8. Forces acting on channel section.
From Eq. (2-12) velocity v can be solved
f f
RS C RS
K
v

(2-13)
Eq. (2-13) is known as the Chezy equation and C is the empirical Chezy coefficient.
Manning equation, Manning coefficient
The most widely used equation for calculating flow velocity was developed by Manning
2 / 1
f
3 / 2
S R
n
1
v (2-14)
where v is average velocity (m s
-1
), n is the Manning coefficient, (M=1/n is the so called
Strickler coefficient), R is hydraulic radius (m) and S
f
is friction (energy) slope (-). In
uniform flow S
f
can be replaced by S
o
.
22
Friction loss h
f
Friction loss h
f
is calculated by solving S
f
from Eq. (2-14) and by remembering that
S
f
=h
f
/L:
3 / 4 2
2 2
3 / 4
2 2
f
R A
Q n
L
R
v n
L h (2-15)
where L is the length of the channel section.
Darcy-Weisbach equation
The Darcy-Weissbach equation was originally developed for pipe flows but it can be also
used in open channel flow. The equation is physically sound and better than Mannings
equation. Darcy-Weissbach equation is discussed more thoroughly later on.
R gA 8
LQ
f
gR 8
Lv
f h
2
2 2
f
(2-16)
where f is the Darcy-Weissbach friction coefficient.
2.5 Continuity equation in steady flow
In steady flow the discharge Q is constant and does not vary as a function of time or along
the channel. Therefore at two different cross-sections along the channel Q
1
=Q
2
, or
v
1
A1=v
2
A
2
.
23
3 COMPUTATION OF STEADY UNIFORM FLOW
3.1 Introduction
In steady uniform flow discharge Q is does not vary with respect to distance or time (Q is
constant) and the bottom slope S
o
equals the friction slope S
f
, implying that water depth h
remains constant in the channel. The water depth in uniform flow is called normal depth
h
N
. Uniform flow is usually possible only in man made channels. By equating bottom
slope and friction slope, and taking into account the Manning equation and relation Q=vA
one gets an equation that can be used to calculate discharge in uniform and steady flow.
o
3 / 2
f
3 / 2
o f
S AR
n
1
Q
S R
n
1
v ; vA Q ; S S

(3-1)
Eq. (3-1) can be used to calculate Q if h is known. In many cases discharge Q and shape of
the cross-section are known and it is necessary to calculate what is the normal depth h
N
.
By examining Eq. (3-1) it can be seen that it is not possible to calculate directly h
N
from
(3-1) because both A and R are non-linear functions of water depth. Three different
methods are described below for calculating h
N
when Q, S
o
, n and shape of the cross-
section are known:
trial-and-error method (algebraic method)
Newton-Raphson iteration
Graphical method
24
3.2 Trial-and-error method (algebraic method)
Example 3-1. Calculate normal depth h
N
in a trapezoidal cross-section with bottom width
B
B
=20 m and side slopes 1:1.5. Discharge Q = 250 m
3
s
-1
, bottom slope S
o
is 0.0015 and
Manning coefficient n=0.035.
Solution: Rearrange Eq. (3-1) in such a way that the known factors are moved to one side
and the unknown variables to the other side, i.e.
o
3 / 2
S
nQ
AR (3-2)
Now the left hand side of Eq. (3-2) is a function of water depth h and the value of the right
hand side can be calculated. In this method the specific value of h that results in the same
value for left and right hand sides is obtained by trial-and-error procedure: choose h, and
calculate A, R and AR2/3 and change h until Eq. (3-2) is fulfilled. The results are shown
in Table 3-1 below.
____________________________________________________________________
Table 3-1. Calculation of normal depth h
N
by trial-and-error-method.
____________________________________________________________________
_____________________________________________________________________
Example 3-1. Uniform, steady flow: estimate normal depth h
QQ 250
nn 0.035
So 0.0015
BB 20 Right hand side = 225.924
Side 1.5 (1:Side)
h B A P R AR^(2/3)
3 29 73.5 30.8167 2.38507 131.207
4 32 104 34.4222 3.02131 217.353
4.2 32.6 110.46 35.1433 3.14313 237.018
4.1 32.3 107.215 34.7828 3.08242 227.083
4.09 32.27 106.892 34.7467 3.07632 226.101
4.088 32.264 106.828 34.7395 3.07511 225.905
o
3 / 2
S
nQ
AR
25
3.3 Newton-Raphson method
Assume that it is necessary to find the root of a function F(y), i.e. find the y-value that
gives F(y) =0. Newton-Raphson method is a very efficient way of solving this kind of
problem. The method is based on giving an initial estimate y
0
for the unknown variable. A
new value for the unknown y is then computed from
) y ( ' F
) y ( F
y y
k
k
k 1 k

+
(3-3)
In the case that successive iterations y
k+1
and y
k
are close to each other, the iteration can be
stopped. Example 3-2 illustrates the Newton-Raphson method. Find y-value that gives the
function F(y) = y
3
-3y
2
+6y-30 a value of zero. The solution is shown for two cases: a)
derivative F'(y) is computed analytically and b) derivative F'(y) is calculated numerically
using Eq. (3-4).
y
) y ( F ) y y ( F
) y ( ' F

+
(3-4)
According to Eq. (3-4), the function F is computed using two arguments: y and y+y and
the derivative F'(y) is approximated by the finite difference equation shown in (3-4).
Usually a suitable value for y is around 0.005-0.01 m for hydraulic computations.
26
The application of the Newton-Raphson method for estimation of the normal depth is
based on reformatting Eq. (3-2) by moving both terms to the same side and requiring that
0
S
nQ
AR ) h ( F
o
3 / 2
(3-5)
The method can be summarised as follows:
0) set k=0 and choose an initial guess for depth h
k
=h
0
1) compute function F(h
k
) from Eq. (3-5)
2) compute the derivative F'(h) numerically as follows:
a) compute F(h
k
+ h), i.e. deviate h by h (use e.g. h=0.01 m)
b) compute numerical approximation for derivative F(h
k
) = (F(h
k
+ h) - F(h
k
))/ h
3) h
k+1
= h
k
- F(h
k
)/F(h
k
)
Example 3-2.
Newton-Raphson iteration using EXCEL.
Solve F(y) = 0, i.e. what is the y-value that gives F(y)=0.
a) Compute the derivative analytically as shown above
Iter. no y F(y) F'(y) y-new
1 0 -30 6 5
2 5 50 51 4.019608
3 4.019608 10.5917 30.35409 3.67067
4 3.67067 1.060497 24.39743 3.627202
5 3.627202 0.015056 23.70657 3.626567
6 3.626567 3.18E-06 23.69656 3.626567
7 3.626567 1.42E-13 23.69656 3.626567
b) Compute the derivative F'(y) numerically
Iter. no y F(y) F(y+dy) F'(y) y-new
1 0 -30 -29.9403 5.9701 5.025041
2 5.025041 51.28465 51.80189 51.72373 4.03353
3 4.03353 11.01606 11.32304 30.69802 3.674678
4 3.674678 1.158412 1.403833 24.54204 3.627477
5 3.627477 0.021564 0.259462 23.78982 3.62657
6 3.62657 7.8E-05 0.237833 23.77551 3.626567
7 3.626567 2.59E-07 0.237755 23.77546 3.626567
6 y 6 y 3 ) y ( ' F
0 30 y 6 y 3 y ) y ( F
2
2 3
+
+
27
If the successive iterations are close enough to each other, i.e. Abs(h
k+1
-h
k
)< , then stop
the iteration; otherwise increase iteration counter and go back to 1). Iteration stopping
criteria must be small enough, e.g. =0.0005-0.001 m.
Example 3-3. Newton-Raphson method was used to solve the same problem than in the
Example 3-1. The results are shown in Table 3-2.
____________________________________________________________________
Table 3-2. Calculation of normal depth h
N
by Newton-Raphson method. Derivative of
function F(h
k
) was calculated numerically using h=0.01 m and h
0
=1.0 as the initial guess.
____________________________________________________________________
Example 3-4. The normal depth problem discussed in previous examples can also be
solved using the Visual Basic modules of Excel. The idea is to write a small Visual Basic
function to calculate the values of the function F(h
k
) in Excel-cells. The own function can
be called directly from Excel-cells in the same way than other Excel-functions. The
solution is shown in Table 3-3 and the Visual Basic function in Fig. 3-1. Your own
function can be written by selecting in Excel "Tools - Visual basic Editor - Insert -
Exampl e 3- 3.
QQQ 250 Tr apezoi dal cr oss- sect i on; bot t om wi dt h 20 m, si de sl opes 1: 1. 5
nnn 0. 035 Lef t hand si de= A* R^( 2/ 3)
I I I 0. 0015 = Bot t om sl ope Ri ght hand si de = n* Q/ sqr t ( I ) = 225. 924
Bl ev 20 = Bot t om wi dt h Fi nd h t hat gi ves F( h) = Lef t - Ri ght =0
Si deS 1. 5 Newt on- Raphson: hnew=hol d - F( h) / F' ( h)
del h 0. 01 I ni t i al guess: h=1. 0 m.
Der i vat i ve F' ( h) cal cul at ed numer i cal l y.
h B A P R F( h) F' ( h) h_new I t er .
h 1. 000 23. 000 21. 500 23. 606 0. 911 - 205. 72
h- dev 1. 010 23. 030 21. 730 23. 642 0. 919 - 205. 38 34. 080 7. 037 1
h 7. 037 41. 110 214. 999 45. 370 4. 739 380. 672
h- dev 7. 047 41. 140 215. 410 45. 407 4. 744 382. 285 161. 288 4. 676 2
h 4. 676 34. 029 126. 328 36. 861 3. 427 61. 247
h- dev 4. 686 34. 059 126. 668 36. 897 3. 433 62. 350 110. 304 4. 121 3
h 4. 121 32. 363 107. 896 34. 859 3. 095 3. 241
h- dev 4. 131 32. 393 108. 219 34. 895 3. 101 4. 230 98. 865 4. 088 4
h 4. 088 32. 265 106. 836 34. 740 3. 075 0. 014
h- dev 4. 098 32. 295 107. 159 34. 777 3. 081 0. 996 98. 194 4. 088 5
h 4. 088 32. 264 106. 831 34. 740 3. 075 0. 000
h- dev 4. 098 32. 294 107. 154 34. 776 3. 081 0. 982 98. 191 4. 088 6
28
Module" and then writing the code as shown in Fig. 3-1. The function is called from Excel
sheets as follows:
=NR(B10;QQQQ;nnnn;Bottom;Soo;SideSlope)
or
=NR(B10;$A$5; $A$6; $A$7; $A$8; $A$9)
where B10 includes the water depth h. The other parameters can be taken as QQQQ is the
discharge in cell A5, etc.
________________________________________________________________________
Table 3-3. Visual Basic function used to calculate the function F(h
k
) and F(h
k
+h) in the
Newton-Raphson iteration method (alternative solution compared to Table 3-2).
_______________________________________________________________________
________________________________________________________________________
Fig. 3-1. The Visual Basic function used to calculate F(h
k
) in the Newton-Raphson
method.
Exampl e 3- 4. Vi sual Basi c modul e used t o Cal cul at e F( h)
QQQQ 250
nnnn 0. 035
Soo 0. 0015 = Bot t om sl ope
Bot t om 20 = Bot t om wi dt h
Si deSl op 1. 5
del t ah 0. 01
Vi sual Basi c modude cal l ed i n cel l s F( h) and F( h+dh) :
=NR( B10; QQQQ; nnnn; Bot t om; Soo; Si deSl op)
where h is in cell B10
Iter h F(h) F(h+dh) F'
0 3 -94.7176 -93.9561 76.14757
1 4.243868 15.51615 16.5299 101.3753
2 4.090812 0.255673 1.238084 98.24108
3 4.088209 0.000335 0.982214 98.18788
4 4.088206 3.49E-07 0.981878 98.18781
5 4.088206 3.63E-10 0.981878 98.18781
6 4.088206 3.98E-13 0.981878 98.18781
Function NR(h, Q, n, BB, Slope, Side)
Dim A, R, P As Double
B = BB + 2 * Side * h
A = h * 0.5 * (B + BB)
P = BB + 2 * h * (1 + Side * Side) ^ 0.5
R = A / P
NR = A * R ^ (2 / 3) - n * Q / Slope ^ 0.5
End Function
29
3.4 Graphical method
The graphical method described below is suitable for irregular cross-sections when A and
R cannot easily be formulated as a function of h. The method is based on plotting AR
2/3
as
a function of h and selecting from graph the h-value which gives AR
2/3
=nQ/S
o
1/2
as shown
in Fig. 3-2.
Fig. 3-2. Determination of normal depth using the graphical method.
3.5 Exercises
Exercise 3-1. Assume that flow in a channel is steady and uniform. Calculate the normal
depth h
N
in a rectangular channel with dimensions: Discharge Q is 30 m
3
s
-1
, bottom slope
is S
o
= 0.0015, bottom width B
B
= 10 m, and Manning coefficient n = 0.035. Solve the
problem by trial-and-error method.
Exercise 3-2. Assume that flow in a channel is steady and uniform. Calculate the normal
depth h
N
in a channel with trapezoidal cross-section: Discharge Q is 300 m
3
s
-1
, bottom
Graphical method to determine normal depth
100
130
160
190
220
250
2.5 2.7 2.9 3.1 3.3 3.5 3.7 3.9 4.1 4.3 4.5
h (m)
A
R
^
(
2
/
3
)
30
slope is S
o
= 0.0015, bottom width B
B
= 30 m, side slopes are inclined to ratio 1:2 and
Manning coefficient n = 0.035. Solve the problem by Newton-Raphson method.
Exercise 3-3. Consider an irregular cross-section as shown in the graph below.
a) Compute the hydraulic parameters h, B, A, P and R from the graph for heights
65,66,68,70 and 72 m.
b) Use the graphical method to estimate normal depth in steady flow (bottom slope equals
friction slope) when discharge Q is 100 m
3
s
-1
, Manning coefficient is n=0.04 and bottom
slope is 0.005.
c) What is the maximum discharge Q
MAX
that can flow in the cross-section without
overflowing the banks (i.e. y cannot be higher than 72.0 m)
Exercise 3-4. Derive the equation for critical depth h
C
if (see Eqs. (2-8)..(2-10)).
a) Cross section is rectangular and discharge is Q and bottom width is B
B
. Calculate h
C
when Q = 20 m
3
s
-1
and B
B
=5.0 m.
b) Cross section is triangular where, discharge is Q, and the angle of the side slopes is 1:Z.
Calculate h
C
when Q = 20 m
3
s
-1
and Z=2. Critical depth h
C
is water depth at midpoint of
31
the cross section.
c) Cross section is trapezoidal; discharge is Q, bottom width is B
B
and side slope is 1:Z.
Calculate h
C
when Q = 20 m
3
s
-1
and B
B
=5.0 m and Z=2.
Exercise 3-5. Take the rectangular cross section of exercise 2-4c (Q = 20 m
3
s
-1
, B
B
=5.0 m
and side slope is 1:2).
a) Calculate a specific energy versus water depth graph for the cross-section. An example
of the specific energy versus water depth graph is shown below for a cross section with Q
= 1 m
3
s
-1
, B
B
=5.0 m and side slope is 1:2. According to the graph two different water
depths h
1
(supercritical) and h
2
(subcritical) water produce the same specific energy.
b) The cross section in a) is considered: When flow changes from supercritical to
subcritical (e.g. steep slope changes to a mild one), water depth changes from a value that
is smaller than the critical depth h
C
, to a value that is greater than the critical depth and a
hydraulic jump is formed. Assume that water depth at supercritical flow is h
1
=0.6 m and
after the hydraulic jump the flow is subcritical with water depth h
2
=1.6 m. How big is the
total energy loss (unit m) in the hydraulic jump? Try to figure out why it is necessary that
the hydraulic jump is formed? I.e. why is it not possible that the water depth would change
smoothly from supercritical to subcritical?
32
4 GRADUALLY VARYING NON-UNIFORM FLOW
4.1 Introduction
In non-uniform flow the energy (friction) slope S
f
is not the same than the bottom slope S
o
.
Although the discharge Q may be the same in all cross-sections, flow depth, top width,
cross-sectional area A and flow velocity v vary gradually from section to section. Non-
uniform flow is the prevailing flow condition in natural rivers. In non-uniform flow
velocity cannot be computed by knowing bottom slope and the friction slope has to be
used.
In supercritical flow conditions, the disturbance does not travel upstream and therefore
non-uniform flow conditions must proceed downstream, i.e. in the same direction as the
flow. In subcritical conditions the influence of a disturbance travels both down- and
upstream. Therefore, in subcritical condition the non-uniform flow conditions have to be
started from the downstream end of the river and calculation proceeds upstream.
A very usual case where gradually varied non-uniform flow needs to be calculated is a
dam or a weir that raises water level upstream the dam. The aim is to calculate how far
upstream the dam influences the water levels. An example of the non-uniform flow
conditions is shown in Fig. 4-1. Due to the influence of the dam, water level is raised at
the location of the dam and this influences water depths h and water levels z
W
=z
B
+ h
upstream the dam. In the special case shown in Fig. 4-1, the channel is prismatic (the
shape of the cross-section does not vary along the distance) and at certain point the
influence of the dam stops. In Fig. 4-1 this point is around 1300 m upstream from the dam
(see also Table 4-2, Section 4-3, where the results of the calculation of the example are
shown). The normal depth in this case is 1.85 m and this depth is prevailing in the river
when distance from the dam is more than 1300 m. The engineering point is to be able to
calculate the backwater curve, i.e. how far the influence of different constructions reaches.
33
0
1
2
3
4
5
6
7
8
9
10
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
6
0
0
7
0
0
8
0
0
9
0
0
1
0
0
0
1
1
0
0
1
2
0
0
1
3
0
0
1
4
0
0
1
5
0
0
1
6
0
0
1
7
0
0
1
8
0
0
1
9
0
0
2
0
0
0
Ypohja
Y
Fig. 4-1. The longitudinal water profile. (Water depth h immediately upstream the dam is
4.0 m, QQQ = 150 m3/s, bbb = 30 m, nnn = 0.035, III = 0.004, Side slope 1:1, Alfa = 0.0,
i.e. flow velocity is decreasing in the direction of flow. Y = z
W
= water level elevation
above the reference level, Ypohja = z
B
is bottom elevation above the reference level.)
Control point (section), stage-discharge relationship
In non-uniform flow conditions the calculation is possible if water level or a so called
stage-discharge curve is known at a control point (section). In subcritical flow this point
has to be at the downstream end of the river and in supercritical flow at the upstream end.
The following type of control points can be used as the starting point of calculation:
water level in a lake or sea in the case of subcritical flow
control structures such as weirs and sluice gates have rating curves (stage-discharge
relationships)
where there is an increase in the bed slope causing flow to pass through critical depth
near the change in gradient
34
4.2 Computation of non-uniform flow in one river reach
Consider the energy balance in the river reach shown in Fig. 4-2. The flow is assumed to
be subcritical and therefore the calculation proceeds upstream and water depth at section 1,
h
1
, is known. What is the water depth h
2
and water level z
W
at section 2?
Bernoullis equation can be used in the calculation of the gradually varying non-uniform
flow.
L
h
S ; L S z
h
g 2
v
h L S
g 2
v
h
f
f o B
f
2
1
1 o
2
2
2

+ + + +
(4-1)
Fig. 4-2. Energy principle in non-uniform flow (L is the length of the river reach).
Water level at the upstream end h
2
is unknown and can be solved from Bernoullis
equation:
f o
2
2
2
1
1 2
h L S
g 2
v
g 2
v
h h +

,
_

+ (4-2)
Water depth and energy slope do not remain constant during the river section and therefore
velocity v
2
and energy losses h
f
in equation (4-2) depend on the unknown h
2
value and the
problem cannot be solved without iteration. Friction losses have to be calculated by
35
solving friction slope from Manning's equation and using average values for area and
hydraulic radius
2
R R
R ;
2
A A
A
R A
Q n
L h
2 1
m
2 1
m
3 / 4
m
2
m
2 2
f
+

(4-3)
In the case that flow velocity is decreasing in the direction of flow (flow depth increasing),
the velocity distribution coefficient is usually assumed to be zero, but if the flow is
accelerating, a value =1.1 is usually used.
The iteration procedure is shown in Fig. 4-3 for one river reach.
0) give an initial guess for the unknown water depth h
2
=h
1
.
1) Compute average values for A
m
and R
m
.
2) Compute friction losses h
f
from Eq. (4-3)
3) Compute new estimate for the unknown water depth at cross-section 2 =h
2
'
4) Compare if the successive iterations h
2
' and h
2
are close enough to each other,
e.g. Abs(h
2
' - h
2
)<e, where e is small enough: usually around 0.001-0.005 m.
5) In the case that the conditions shown in 4) is fulfilled, then the iteration is stopped, and
otherwise make a substitution h
2
= h
2
' and go back to 1).
Example 4-1. Solve one river reach of non-uniform flow in subcritical case. Cross-
section is rectangular with bottom width B
B
=20 m, Manning coefficient n=0.04, bottom
slope S
o
=0.002, discharge running in the river section is 100 m
3
s
-1
, length of the section is
L=100 m and water depth at downstream end is h
1
=2.0 m. The results of the iteration are
shown in Table 4-1 and the difference between the third and the fourth iteration (as
absolute value) is smaller than the iteration stopping criteria 0.001 m.
36
Fig. 4-3. Iteration procedure for non-uniform flow calculations (subcritical flow,
calculation proceeds upstream, h
1
known and h
2
unknown).
________________________________________________________
Table 4-1. Results of the Example 4-1.
________________________________________________________
________________________________________________________
37
4.3 Computation of non-uniform flow in longer river reaches
The next step is to apply the calculation procedure described in the previous section to a
longer river section. The idea is simply to divide the total river reach to smaller section
denoted by L or x. The length of the x should not be too big, because the iteration may
fail. On the other hand, too small x increases the computational burden. Usually in
practical cases x is between 50 and 200 m. The idea is basically simple: apply the
Bernoulli equation to a single section starting from the downstream end in subcritical flow.
Solution will be water depth h and water level z
W
at a point located x m from the control
point. Then apply the equation to the next section and so on.
Example 4-2. A dam is built on the downstream end of a river reach and water depth
above the dam will be 4.0 m (this dam has gates that can be used to control the water level
above the dam freely). Discharge in the river is 100 m
3
s
-1
, Manning coefficient n = 0.035,
bottom width of the trapezoidal cross-section B
B
=30 m, side slope of the bank is 1:1 and
bottom slope S
o
=0.004. How far upstream does the dam raise water depths compared to
uniform flow depths (normal depth). Based on the methods described in Chapter 3 it is
possible to calculate that normal depth h
N
=1.85 m in this case. The results of the
calculations are shown in Table 4-2. The computations can be most conveniently done
using the spreadsheets (Excel in this case).
In the solution of the problem iteration is needed as shown in the Example 4-1. In Excel
iteration can be switched on by selecting TOOLS - OPTIONS - CALCULATION and
clicking the appropriate sign. Choose also the iteration stopping criteria. Friction loss h
f
is
computed using the average A and R from two successive cross-sections. In the solution
shown in Table 4-2, additional columns have been reserved for distance X from the dam,
bottom level z
B
and water surface level z
W
=z
B
+h. Eq. (4-2) is used between two
consecutive cross-sections and the friction losses h
f
are computed using Eq. (4-3). In the
solution iteration is needed due to the fact that when h
f
is calculated, A and R from two
38
successive cross-sections are needed but the water level of the upstream section is
unknown.
Table 4-2. Non-uniform flow calculation in a prismatic channel (shape of the cross-
section does not vary along the distance).
According to Table 4-2, water depth is very close to the normal depth at a distance of 1200
m (difference less than 1 cm). If X is greater than 1200 m, the flow can be considered
uniform. The water surface profile was already shown in Fig. 4-1 indicating clearly that
flow velocity is decreasing in the direction of flow.
39
Fig. 4-4 shows the results of a calculation when the downstream water depth is assumed to
be 1.3. In this case flow is still subcritical but accelerating and velocity distribution
coefficient =1.1.
Fig. 4-4. Water surface profile z
W
and bottom profile z
B
in the case that downstream water
depth is 1.3 m.
4.4 Rating curves for non-uniform flow calculations
In non-uniform flow conditions and later on unsteady flow models (e.g. solution of the
Saint Venant equations) a so called boundary condition is needed at the downstream end if
flow is subcritical. In some cases, as mentioned before, it is possible to use a known water
level (e.g. water level in a lake or in a sea) as the starting point of the non-uniform flow
calculations (or as a boundary condition of the unsteady flow calculations). However, in
many cases a rating curve is needed. This curve is also known as stage-discharge
relationship and it gives a unique relationship with water level z
W
(=z
B
+h) and discharge Q
flowing at the control section.
Discharge over a crested weir (overflow dam)
The water level at the location of the dam depends on top level of the weir and on
40
discharge flowing over the dam. Discharge Q
W
over a crested weir (overflow dam) can be
calculated as follows:
2 / 3
W
2 / 1
W W W
H ) g 2 ( B C
3
2
Q (4-4)
where B
W
is the width of the weir, C
W
is a coefficient of discharge and H
W
is the distance
from water level to the top of the weir. In Eq. (4-4) it is assumed the water level below the
dam does not influence of the discharge capacity of the weir.
Eq. (4-4) implies that there exists a rating curve (stage-discharge ratio) which depends on
the top elevation of the dam and discharge flowing in the river (and over the weir). The
greater the discharge, the higher will be the water level upstream the dam. This water level
is the starting point of the non-uniform flow conditions.
4.5 Exercises
Exercise 4-1. Calculate one channel section of non-uniform flow by solving the Bernoulli
equation.
- Q = 20 m
3
/s
- B
B
= 10 m, rectangular channel
- n = 0.030
- S
o
= 0.0015 (bottom slope)
- L = 100 m (length of the channel section)
- Alfa = 1.1, if flow is accelerating and 0.0, if flow velocity is decreasing
41
- h
1
= 2.2 m (water depth at downstream cross-section)
- flow is subcritical
a. What is the water depth h
2
? Use h
2
= h
1
as the initial guess. Repeat the iteration
procedure until the successive iterations differ from each other less than 0.001 m. How
many iterations are needed?
b. Assume that h
1
= 1.3 m. Repeat the calculations.
Ecercise 4-2. A weir dam (overflow dam) is aimed to be built at downstream end of a
river reach which ends up to a rapid (i.e. cross-section where flow is supercritical).
Upstream of the rapid the flow is subcritical and the dam is aimed to raise water level
during low flow conditions in such a way that there is always enough water depth in the
river upstream the dam. Low flow conditions pose no special problem, but during flood
events it is possible that the dam raises water levels too much and the embankments of the
river might not be high enough. The critical section where the embankments might fail is
1000 m upstream from the dam. The elevation of the embankment is there 36.00 m (above
sea level).
42
a) The task is to design the top level of the overflow dam in such a way that during flood
conditions water level z
W
1000 m upstream the dam is not higher than 35.50 m. Find the
highest allowed top level of the weir dam. Note: elevation of the water level z
W
=z
B
+h.
The following input data are available: discharge Q during extreme flood conditions is
assumed to be 400 m
3
s
-1
, bottom elevation z
B
is 30.0 m at the location of the dam, bottom
slope in the river is S
o
=0.001, Manning coefficient n=0.035, the cross-section is
trapezoidal with bottom width B
B
= 50 m and side slopes 1:2. Calculate the non-uniform
flow in the river as x =100 m intervals starting from the location of the dam and
proceeding upstream. In computations the velocity distribution coefficient = 0.0 if flow
velocity in decreasing in the direction of flow. If velocity is increasing, then = 1.1.
The water level at the location of the dam depends on top level of the weir and on
discharge flowing over the dam. Discharge Q
W
over a crested weir (overflow dam) can be
calculated as follows:
(B-1)
where B
W
= 50 m is the width of the weir, C
W
=0.60 is a coefficient of discharge and H
W
is
the distance from water level to the top of the weir. Eq. (B-1) implies that there exist a
rating curve (stage-discharge ratio) which depends on the top elevation of the dam and
discharge flowing in the river (and over the weir). The greater the discharge, the higher
will be the water level upstream the dam. This water level is the starting point of the non-
uniform flow conditions.
43
b) How far upstream does the dam influence water depths in flood conditions? The criteria
is 3 cm. Calculate water level profile (both h and z
W
= z
B
+h) in low flow conditions when
Q = 10 m
3
s
-1
.
Output from the exercise: a) weir top level, water surface profile (X versus z
W
) as a table
and as a graph, b) distance from the dam, c) water level profile distance X versus z
W
as a
table.
The results shown in Example 4-2 can be used to check your own Excel-solution.
Note 1! In overflow dam water level above (behind) the dam cannot be controlled. Water
level depends on the rating curve of the dam and on discharge Q running in the river.
Note 2! Friction loss h
f
is computed using the average A and R from two successive cross-
sections. Iteration is needed since in the calculation the water depth of the upstream
section is unknown. In Excel the iteration can be switched on by TOOLS - OPTIONS -
CALCULATION. The problem cannot be accurately solved without iteration.
Exercise 4-3 Fig. below shows a river that is divided to two branches that flow around the
island located in the middle of the river. Calculate how much water is flowing in branch 1
and branch 2. What is the water depth at point A?
The solution principle is that friction losses must be the same when calculated through
branch 1 and 2. The total discharge Q is 300 m
3
s
-1
, bottom slope S
o
=0.003 (the same slope
in sections 1 and 2). Water depth h at point B is assumed to known (subcritical flow),
h
B
=2.8 m. Both branches have a rectangular cross-section and the bottom width and
Manning coefficient are shown below. The distance between points A and B is 300 m.
The parameters of the branches:
Branch 1 Branch 2
Bottom width B
B
30 25
Manning n 0.045 0.030
44
Discharge Q in the point A is Q = 300 m
3
/s
5 UNSTEADY FLOW IN OPEN CHANNELS
5.1 Introduction
In unsteady flow discharge Q(t) is varying as a function of time and therefore all the
hydraulic factors of a cross section change as a function of time, i.e. water depth h(t) and
A(t) and in natural rivers also surface width B(t) change over time. Flow is also non-
uniform. Throughout this material channel flow is assumed to be strictly one-dimensional,
even though truly one-dimensional flow does not exist in nature.
The equations to be solved are called the Saint Venant equations which were derived as
early as 1871. The equations for unsteady flow are based upon the following series of
assumptions:
The flow is one-dimensional i.e. the velocity is uniform over the cross section and the
water level across the section is horizontal
45
The streamline curvature is small and vertical accelerations are negligible, hence the
pressure is hydrostatic
The effects of boundary friction and turbulence can be accounted for through resistance
laws analogous to those used for steady state flow
The average channel bed slope is small so that the cosine of the angle it makes with the
horizontal may be replaced by unity
The hydrostatic pressure P can be calculated as
h y ); y h ( g P (5-1)
where is density of water, h is water depth and y is the vertical co-ordinate.
The cross sections of the channel can basically be of arbitrary shape and may vary along
the channel axis, although the variation is limited by the condition of small streamline
curvature. One-dimensional unsteady flow in channels, assuming that the density is
constant, can be described by two dependent variables. E.g. the following combinations
are possible:
water depth h and discharge Q
water level z (z
W
) and discharge Q
water depth h and velocity v
These dependent variables define the state of the fluid motion along the water course and
in time, i.e. as a function of two independent variables (x for space and t for time). Two
dependent variables are sufficient to describe one-dimensional flow, and therefore only
two equations are needed, each of which must represent a physical law. Three physical
laws are available: conservation of mass, momentum and energy. When the flow variables
are not continuous (hydraulic jump, bore), two representations are possible: conservation
of mass and momentum, or conservation of mass and energy. Since the mass-momentum
couple of conservation laws is applicable to both discontinuous and continuous situations
46
while the mass-energy couple is not, the Saint Venant equations are derived from mass
and momentum conservation principles.
5.2 Derivation of the Saint Venant equations
5.2.1 Conservation of mass (continuity equation)
Consider a river reach with length x and cross sectional area A shown in Fig. 5-1. The
water balance components of the section are as follows:
x
t
A
storage in Change
x
x
Q
Q Outgoing
x q Q g min Inco

+
+
(5-2)
Fig. 5-1. Water balance of a small river reach.
where Q is the incoming discharge (m
3
s
-1
) and q is lateral flux (m
2
s
-1
) (additional flux) to
the section. The difference between incoming and outgoing mass equals the change of
storage (volume) over time. The continuity equation can thus be expressed as
0 q
x
Q
t
A

(5-3)
47
By remembering that small change in the area of the cross section A = Bh as shown in
Fig. 5-2, the continuity equation can also be given in the form
Fig. 5-2. The incremental area A defined as the area of a rectangle Bh.
0 q
x
Q
t
h
B

(5-4a)
In some cases the floodplains can be treated as storage areas only and in this formulation
floodplain width B
S
(storage width) is used in the continuity equation but estimated
"effective width" and "effective area" are used in the momentum equation (Fig. 5-3.)
0 q
x
Q
t
h
B
S

(5-4b)
Fig. 5-3. The concept of storage width B
S
(storage area A
S
) and effective width B
(effective area A).
48
5.2.2 Conservation of momentum
Momentum is defined as the mass multiplied by the velocity. Momentum is zero if
velocity is zero. The rate of change of momentum is equal to the forces acting on it. The
rate of change of momentum of a body is (Hamill, 2001):
t / ) mv mv ( ma F
momentum of change of rate Force
1 2

(5-5)
where F is the force, m is the mass and v
1
and v
2
are the initial and final velocity,
respectively and t is the time over which the change of velocity occurs, and a is
acceleration of the body. In hydraulics it is convenient to notice that density (kg m
-3
)
multiplied by discharge Q (m
3
s
-1
) gives the amount of mass flowing through a cross
section over time, i.e. m/t in Eq. (5-5) can be replaced by Q giving
) v v ( Q F
1 2
(5-6)
The conservation of momentum in the x-direction requires that the change of momentum
in the control volume during time step t must be equal to the sum of the net inflow of
momentum into the control volume and the integral of the external forces acting on it over
the same time interval. Momentum is the product of mass and velocity, and momentum
flux through the flow section is the product of the mass flow rate and velocity, and
therefore it can be written that
t
Q
t
) Q (
time over momentum of change
x
x
) A / Q (
A / Q outflow momentum
A / Q Qv low inf momentum
2
2
2

+

(5-7)
49
It is assumed that the only important external forces acting upon the control volume in the
x-direction are pressure, gravity, and frictional resistance.
x
h
x gA change pressure
xS gA friction
xS gA ) sin( x gA gravity
f
o

(5-8)
where the angle is so small that sin() is equal to tan() =S
o
, i.e. bottom slope. The
pressure term in (5-8) can be derived as follows. The total force F due to pressure over a
certain cross section can be integrated from

h
0
dy ) y ( B ) y h ( g F (5-9)
where B(y) is the width of the cross section. For a rectangular cross section with bottom
width B
B
, the force can be integrated:
2
B
h gB
2
1
F (5-10)
The change of pressure during the river reach is
x
h
x gA
x
h
x h gB x
x
F
B

(5-11)
which is valid for cross sectional area, but it can be shown (e.g. Cunge et al. 1981) that Eq.
(5-11) holds for irregular cross sections as well.
The momentum conservation equation can now be written in the form
50
0 ) S S ( gA
x
h
gA
x
) A / Q (
t
Q
f o
2

(5-12)
The friction slope S
f
can be calculated e.g. using the Manning's equation (see e.g. Eq. (2-
16)):
3 / 4 2
2 2
f
f
R A
Q n
x
h
S

(5-13)
Different type of formulations can be given for the Saint Venant equations depending on
the problem.
5.2.3 Exercises
Exercise 5-1. The full Saint Venant equations can be shown e.g. in the form
(C-1)
(C-2)
(C-3)
where t is time, x is distance (m), B is surface (top) width of the channel (m), h is water
depth (m), A is cross-sectional area (m
2
) , R is hydraulic radius (m), Q is discharge (m
3
s
-
1
), S
f
is friction (energy) slope (-), n is Manning's coefficient and S
o
is bottom slope (-) and
is the velocity distribution coefficient.
a) In kinematic approximation it is assumed that friction slope equals the bottom slope, i.e.
S
f
= S
o
. This simplification leads to one single dependent variable (water depth h). In
diffusion analogy the acceleration terms are excluded but water depth is not constant
51
throughout the river, i.e. S
f
and S
o
are not the same, but S
f
equals the slope of the water
surface. What kind of equations has to be solved for diffusion analogy? How many
dependent variables are needed in the equations?
b) Many often Saint Venant equations are shown in the form where dependent variables
are z (or z
W
) and Q. z is water level above a reference datum, e.g. sea level (z=h+z
B
).
Derive these equations from (C-1) and (C-3).
c) Special cases of the Saint Venant equations are uniform flow and non-uniform flow
equations Derive these equations starting from (C-1)..(C-3).
Exercise 5-2. a) It is necessary to develop a mathematical model for solving the full Saint
Venant equations for the Oulujoki river shown below. The river starts from the Lake
Oulujrvi and outflow is controlled by a dam and there is also a power plant. Discharge Q
through the power plant can be freely controlled using a segment gate. The river ends up
the Gulf of Bothnia (sea). There are altogether several dams/power plants in the river and
these dams form a discontinuous point where the Saint Venant equations are not valid.
Therefore the river has to be calculated in several successive solutions of the Saint Venant
equations. A model has to be solved first for the river reach between Oulujrvi and the
next power plant after Oulujrvi (river section A in the Fig. below) and so on.
What type of upper and lower boundary conditions can be used in this case?:
for the first river section A
for the last river section from Merikoski power plant located very close to the sea
52
c) Tuusulanjoki river is a tributary of the Vantaanjoki river. It is necessary to solve the
fullSaint Venant equations in the Tuusulanjoki river. What kind of lower boundary
condition can be used for the Tuusulanjoki model? What kind of difficulties can be
expected related to the boundary condition?
5.3 Simplifications of the full Saint Venant equations
The simplified solutions are aimed to be used in special cases when the full Saint Venant
equations are not needed. The methods introduced here are
kinematic approximation
diffusion analogy
numerical method of lines (NMOL)
5.3.1 Kinematic approximation
In the kinematic approximation it is assumed that friction slope equals the bottom slope,
i.e. S
f
= S
o
. This implies that the momentum conservation equation collapse to a simple
form and the equations can be written as:
53
2 / 1
o
3 / 2
S AR
n
1
Q
0 q
x
Q
t
h
B

(5-14)
i.e. discharge Q in any point can be calculated directly as a function of bottom slope S
o
.
This simplification leads to one single dependent variable (water depth h).
Example 5-1. Consider the river shown in Fig. 5-4. There is a time dependent inflow
Q
IN
(t) to the upstream end of the river and this flood wave (hydrograph) is flowing in the
river and the goal is to calculate when does the flood wave reach the downstream end and
also how much the peak discharge is reduced in the river. Outflow Q
OUT
(t) is time
dependent as well. The flow is thus non-uniform and unsteady. The problem can be solved
using the kinematic approximation by
dividing the river into N subreaches with equal length = x
dividing the total computational time to time steps t starting from known initial
condition at time t=0: water depth h
i
, i=1,.., N, is known for all subreaches
making mass balance equations for each subreach i
at upstream boundary the inflow Q
IN
(t) is known and lateral flow q=0
at downstream end it is assumed that outflow Q
OUT
(t) can be computed from (5-14)
(lower equation), i.e. as a function of water depth h
N
(or A
N
and R
N
) and bottom slope
S
o
54
Fig. 5-4. Simplification of the Saint Venant equations using the kinematic approximation.
In the solution of Eq. (5-14) it is necessary to use both subscript and superscript for the
water depth h. The notation h
i
n
refers to water depth at reach i and at time level n. The time
derivative in Eq. (5-14) is replaced by an approximation
t
h h
dt
dh
t
h
n
i
1 n
i i

+
(5-15)
where water depth values from time level n+1 are assumed to be unknown and values
from time level n are known values. The solution starts from the known initial condition
where n=0. The water balance equation for the first section and for section river section i
(i=2,..., N) can be written as follows
x
Q Q
B
1
dt
dh
1 N ,.., 2 i ;
x
Q Q
B
1
dt
dh
x
Q Q
B
1
dt
dh
OUT N , 1 N
i
N
1 i , i i , 1 i
i
i
2 , 1 IN
i
1

+
(5-16)
55
By combining (5-15) and (5-16) and assuming the discharges in (5-16) can be calculated
using the known water depth from time level n, it is possible to solve the equations using
Euler's method.
, |
, |
2 / 1
o
3 / 2
n
i
n
i
1 i , i
2 / 1
o
3 / 2
n
1 i
n
1 i
i , 1 i
n
1 i , i
n
i , 1 i
i
n
i
1 n
i
S R A
n
1
Q
S R A
n
1
Q
N ,.., 2 i ;
x
Q Q
B
t
h h

+
+

+
+
(5-17)
and correspondingly to the first and the last reach.
5.3.2 Diffusion analogy
The kinematic approximation cannot take into account the influence of backwater of
hydraulic computations. Therefore it is suited only to steep slopes where water level in
downstream section does not influence discharge in the river too much. An approximation
that eliminates one problem of the kinematic approximation is the diffusion analogy,
where it is assumed that acceleration terms in Eq. (5-12), i.e. the first two terms, are
neglected but surface elevation term is included. The equations to be solved in the
diffusion analogy are:
0 q
x
Q
t
h
B

(5-18a)
f o
S S
x
h

(5-18b)
Eq. (5-18 b) can be solved in terms of S
f
to get an equation for calculating discharge Q:
56
x
h
S AR
n
1
S AR
n
1
Q
x
h
S S
o
3 / 2
f
3 / 2
o f

(5-19)
According to Eq. (5-19) it is possible to calculate discharge as a function of bottom slope
S
o
and slope of the water depth
x
h

. It is relatively straightforward to modify the numerical

solution of the kinematic approximation to explicit numerical solution of the diffusion
analogy:
, |
, |
) R R (
2
1
R ); A A (
2
1
A
) R R (
2
1
R ); A A (
2
1
A
x
h h
S R A
n
1
Q
x
h h
S R A
n
1
Q
N ,.., 2 i ;
x
Q Q
B
t
h h
n
1 i
n
i
1 i , i
n
1 i
n
i
1 i , i
n
i
n
1 i
i , 1 i
n
i
n
1 i
i , 1 i
n
i
n
1 i
o
3 / 2
1 i , i 1 i , i
n
1 i , i
n
1 i
n
i
o
3 / 2
i , 1 i i , 1 i
n
i , 1 i
n
1 i , i
n
i , 1 i
i
n
i
1 n
i
+
+
+
+

+
+ +
+

+
+
+ +
+ +

+
(5-20)
The solution is straightforward except the boundary conditions. In the first node
(upstream) Q
i-1,i
is replaced by the known inflow hydrograph (Q
IN
(t)). At the downstream
end where water level outside the calculation area is not known, i.e. in node N+1 (N is the
last real node), and cannot be used to calculate outflow Q
N,N+1
from the last node. Two
possibilities exist: 1) assume that
x
h

=0 in the last node or 2) assume that

x
h

can be
computed using the slope between nodes N-1 and N. Moreover, when calculating outflow
from the last node (subreach) it is necessary to assume that A
N,N+1
=A
N
n
and R
N,N+1
=R
N
n
.
57
Example 5-2. Comparison of the results of kinematic approximation and diffusion
analogy was carried out using the following data: mild bottom slope, So=0.002, Manning
coefficient n=0.04, rectangular cross-section with bottom width B
B
=20 m and time step of
computation was t= 5 s. The results are shown in Fig. 5-5 for discharge Q and water
depth h at the downstream end (Q
N
and h
N
) indicating that for rivers with mild slope the
kinematic approximation gives very inaccurate results and it is suggested that it can be
used only for steep slopes (S
o
>0.02). Comparison of the result for steeper bottom slope,
S
o
=0.01 are shown in Fig. 5-6 indicating that in this case the approximations give results
very close to each other. Comparison to full Saint Venant solution is shown later.
Comparison of approximations of the St Venant
eqs.: h in the downstream end
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20 25
Time (min)
h

(
m
)
h1 hN/Diff. analogy hN/Kinematic appr.
58
Comparison of approximations of the St Venant
eqs.: Q in the downstream end
0
20
40
60
80
100
120
0 5 10 15 20 25
Time (min)
Q

(
m
3
/
s
)
QIN Qout/Diff. analogy Qout/Kinematic appr.
Fig. 5-5. Comparison of two different approximations of the Saint Venant equations:
kinematic approximation and diffusion analogy. In both graphs Q/h at upstream end is
shown for reference. Bottom slope is mild, S
o
=0.002.
Comparison of approximations of the St Venant
eqs.: h in the downstream end
0
0,5
1
1,5
2
0 5 10 15 20 25
Time (min)
h

(
m
)
h1 hN/Diff. analogy hN/Kinematic appr.
59
Fig. 5-6. Comparison of two different approximations of the Saint Venant equations:
kinematic approximation and diffusion analogy. In both graphs Q or h at upstream end is
shown for reference. Bottom slope is relatively steep, S
o
=0.01.
5.3.3 Exercises
Exercise 5-3. Consider the kinematic approximation (=simplification of the Saint Venant
equations) in the form that is given in Section 5.3 (Eqs. (5-14)..(5-17)). Inflow Q
IN
(t)
follows a sinusoidal pulse between time 0..10 min and after that it is constant
Note Q
IN
(t) is given in units m
3
s
-1
but time in previous eq. is minutes!
The total length of the river is 100 m and it is divided to 10 reaches. Calculate the flood
case using the kinematic approximation.
Bottom slope So=0.02, bottom width in a rectangular cross section is B
B
=20 m, Manning
coefficient n =0.04 and at initial state the water depth is the whole river is equal to the
Comparison of approximations of the St Venant
eqs.: Q in the downstream end
0
20
40
60
80
100
120
0 5 10 15 20 25
Time (min)
Q

(
m
3
/
s
)
QIN Qout/Diff. analogy Qout/Kinematic appr.
60
normal depth. Use time step t=0.25 min. The total computation period is 20 min.
Output: the Excel-sheet as an attachment file, draw also graph time versus Q
IN
and Q
OUT
.
b) Kinematic approximation is good approximation only in special conditions. Try to
figure out when the approximation is good and when does it fail for sure.
Exercise 5-4. Consider the diffusion analogy (=simplification of the Saint Venant
equations) in the form that is given in Section 5.3.2 (Eqs. (5-16) and (5-20)). Inflow Q
IN
(t)
follows a sinusoidal pulse between time 0..10 min and after that it is constant
Note Q
IN
(t) is given in units m
3
s
-1
but time in previous eq. is minutes!
The total length of the river is 1000 m and it is divided to 10 reaches. Calculate the flood
case using the diffusion analogy and assuming that the boundary condition at the
downstream end is
x
h

=0, i.e. S
f
=S
o
there.
Bottom slope So=0.005, bottom width in a rectangular cross section is B
B
=20 m, Manning
coefficient n =0.04 and at initial state the water depth is the whole river is equal to the
normal depth. Use time step t=0.1 min. The total computation period is 20 min.
Output: the Excel-sheet as an attachment file, draw also graph time versus Q
IN
and Q
OUT
.
b) Diffusion analogy is a very good approximation in many conditions. Try to figure out
cases (problems) when the approximation might not be accurate enough?
61
5.4 Numerical solution of the full Saint Venant equations
5.4.1 History
The Saint-Venant equations were derived more than one hundred years ago (1871) and
their analytical solution in general form is not possible. However, it is possible to solve
them numerically using e.g. finite difference (FD) methods, characteristic methods and
numerical method of lines (NMOL). Finite element (FE) solution is also possible but it
does not provide any real benefit in 1-D cases. The material shown here is mainly
concentrated on explicit and implicit finite difference methods.
The basic formulation of the FD-method was introduced by Courant, Friedrichs and Leavy
already in 1928. However, it was 20 years later that Isaacson, Troesch and Stoker
formulated a mathematical model from certain sections of the Ohio- and Mississippi-
rivers. Preissmann introduced his famous scheme in 1961 in 1970's the computers started
to develop very fast and hundreds of numerical solutions have been presented after that for
solving the full Saint Venant equations using mainly FD-methods.
The most famous models developed are:
CARIMA and CAREDAS developed in SOGREAH (Societe Grenobloise d`Etudes
etd`Applications Hydrauligues) (developed since 1961, ref. e.g. Liggett ja Cunge,
1975)
The models developed in the hydraulic laboratory of Delft (Vreugdenhill, 1973)
The models of the Danish Hydraulic Institute (DHI), System 11 SIVA (Verwey,
1971) and its successor MIKE11
DAMBRK -model by NWS (National Weather Service, USA) (Fread, 1978 ja 1982)
HEC-models (Hydrologic Engineering Center of the United States Army Corps of
Engineers) (Eichert, 1981).
62
5.4.2 Introduction to finite difference methods
Discretisation
The basic idea in the finite difference method is to replace partial derivatives using
approximation
x
Q
x
Q
x
Q
x

0
lim (5-21)
and further an approximation will be used for the right hand side of Eq. (5-21) as derived
below. Assume that a function f(x) is continuous and its derivatives are available. Then
f(x) can be replaced by Taylor's series:
..
! 6
) (
! 2
) (
) ( ) (
3
3 3
2
2 2
+

+ + +
dx
f d x
dx
f d x
dx
df
x x f x x f
(5-22)
By solving Eq. (5-22) with respect to df/dx we get
); ( 0
) ( ) (
x
x
x f x x f
dx
df
+

+
(5-23)
where O(x) is the residual term which includes the terms with derivatives higher than 1
st
degree. The forward finite difference approximation can be obtained dropping the residual
term from (5-23) (see also Fig. 5-7).
x
x f x x f
dx
df

) ( ) (
(5-24)
The backward finite difference approximation can be derived in an analogous way:
..
! 6
) (
! 2
) (
) ( ) (
3
3 3
2
2 2
+

+
dx
f d x
dx
f d x
dx
df
x x f x x f
(5-25)
which leads to
x
x x f x f
dx
df

) ( ) (
(5-26)
By subtracting (5-25) from (5-22) we get
...
! 3
) ( 2
2 ) ( ) (
3
3 3
+

+ +
dx
f d x
dx
df
x x x f x x f
(5-27)
63
The central finite difference approximation can now be written as
x
x x f x x f
dx
df

2
) ( ) (
(5-28)
Fig. 5-7. Approximation of function derivative with backward, forward and central
approximation.
The central approximation is most accurate but in numerical solution the forward
approximation is usually used due to numerical reasons.
In the finite difference approximation the basic idea is to divide the (x, t)-space into
intervals in both directions and this means that space-time domain is overlaid by a 'grid' or
'network of nodes' as shown in Fig. (5-8).
Fig. 5-8. The computational grid used in the finite difference method.
64
In the numerical solution of the Saint Venant equations the dependent variables, discharge
Q (or velocity u) and water level y (or water depth h) will be calculated only at the nodal
points of the grid (not as continuous functions). The partial derivatives will be replaced by
approximation shown in (5-21) and (5-24), (5-26) or (5-28), most often (5-24). The
notation with respect to super- and subscripts is clarified in Fig. 5-9. Superscripts n and
n+1 refer to time step in such a way that values at time level n are assumed to be known
and values at time level n+1 are unknowns which have to be solved numerically from the
Saint Venant equations. Subscripts j-1, j and j+1 refer to location in the longitudinal
section of the river starting from upstream end. The total number of nodal points in x-
direction is denoted by N. As initial values the Q and y must be given as input data for all
nodes j=1...N from time level n=0 at time t=0.
Fig. 5-9. Part of the computational grid of the finite difference method (e.g. discharge at
point E is denoted by
n
j
Q and water level at point C is denoted by
1
1
+
+
n
j
y .
Now it is necessary to define the approximations for the partial derivatives of Q and y with
respect to x and t using the points of the computational grid shown in Fig. 5-9. Consider
first the spatial derivative with respect to x at point E: forward, backward and central finite
difference approximations can be given as follows:
j
n
j
n
j
E
x
f f
x
f

,
_

+1

(5-29)
65
1
1

,
_

j
n
j
n
j
E
x
f f
x
f

(5-30)
j j
n
j
n
j
E
x x
f f
x
f
+

,
_

+
1
1 1

(5-31)
where f denotes to some variable to be approximated (Q, u, y, h). The forward partial
derivative of f with respect to time at point E can be analogously written as
t
f f
t
f
n
j
n
j
E

,
_

+1

(5-32)
Other notations are according to Fig. 5-9. The variables (A, B, etc.) need to be
approximated in the solution. This can be done in several ways:
e.g. A(x, t) = 0.5 (A
j
n
+ A
n
J+1
) or A(x, t) = 0.5 (A
j
n
+ A
n
j+1
+ A
j
n +1
+ A
n+1
j+1
).
Discretisation is the process, where the partial derivatives of the Saint Venant equations
are replaced by the approximations described above. There exist several ways to do the
discretisation and this leads to either explicit or implicit solution methods.
6 BED ROUGHNESS, FRICTION AND VELOCITY DISTRIBUTION IN
UNVEGETATED AND VEGETATED CHANNELS
6.1 Introduction
The purpose of this Chapter is to discuss one of the most important topics of open channel
hydraulics: what is the friction coefficient Darcy-Weissbach, Manning or Chezy in
unvegetated channels and how is vegetation influencing the friction coefficients? The
problem is very crucial just nowadays due to the fact that Environmentally friendly river
engineering has received wider acceptance in the scientific community and among
practical engineers as well.
66
Estimation of the friction coefficients cannot be discussed thoroughly without taking into
account the following topics:
What is bed roughness k
s
in unvegetated and vegetated conditions?
What is velocity distribution in unvegetated and vegetated conditions?
What is the drag force exerted on vegetation?
What are the shear stresses in laminar and turbulent flow?
How can flow layers be classified in unvegetated and vegetated channels?
What are the shear stresses at the bottom and at the boundary between undeflected or
deflected vegetation in unsubmerged and submerged cases?
What is the shear stress in the imaginary wall between the main channel and the
floodplain? How is momentum transfer, which leads to a so called Reynolds stress,
influencing the conveyance capacity of the main channel in the case that there exists a
floodplain?
The purpose of this material is to cover only the basic phenomena related to evaluation of
bed roughness, friction coefficient, shear stresses, theory of turbulence and velocity
distribution in the channel. Most of the equations shown in this material are definitely
well-known, and referenced in numerous papers. In this case of unvegetated channels,
velocity distribution, shear stresses and friction coefficient follow the derivation shown by
Liu (2001). In vegetated cases Stephan (2001) has collected in her PhD-thesis a large
number of methods that have been utilised in the derivation of the methods shown in this
material. The material included is not intended to replace the text books. On the contrary,
this should encourage to get acquainted with the theory more thoroughly (ref. Graf and
Altinakar, 1998; Nezu and Nakagawa, 1993).
67
6.2 Laminar, turbulent and bottom shear stresses
6.2.1 Shear stress in laminar flow
In laminar flow the fluid particles move horizontally without macroscopic mixing, i.e. the
laminar flow can be visualised as layers which slide smoothly over each other as shown in
Fig. 6-1. The layers close to the bottom have slower velocity and therefore there is a shear
stress between the different "layers" since the upper layer moves faster than the layer
below. This shear stress can also be called viscous shear stress since it is caused by the
viscosity of water.
Fig. 6-1. Shear stress in laminar flow (u(z) refers to v(z)).
The shear stress can be given by Newtons law of viscosity
dz
dv

(6-1)
where is density of water and is kinematic viscosity, v is velocity and z is vertical co-
ordinate (positive upwards). The velocity difference dv/dz determines the magnitude of the
shear stress. It has to be remembered that also in the turbulent flow there is narrow
laminar flow layer near the bottom.
68
6.2.2 Shear stress in turbulent flow
In turbulent flow the particles move both horizontally and vertically and there is a
continuous "mixing of particles". Some move upwards and some downwards according to
a stochastic process. In turbulent flow the instantaneous velocity components in horizontal
(x) direction, V, and in vertical (z) direction, W can be defined as
' w w W
' v v V
+
+
(6-2)
where v and w are the time averaged velocities in x and z direction, respectively and v
and w are the instantaneous velocity fluctuations. Time averaged values v and w are
usually used in the computations.
In turbulent flow the particles move irregularly causing continuous exchange of
momentum (remember: momentum is mass multiplied by velocity) from one portion of
fluid to another and this momentum exchange is the reason for the turbulent shear stress,
which is also called the Reynolds stress. The turbulent shear stress, given by time-
averaging of the Navier-Stokes equation, is
' w ' v
t
(6-3)
Eq. (6-3) can be used if measurements of velocities in horizontal and vertical direction are
available over a period (e.g. few minutes) and the readings are stored e.g. several times
during one second. Then it is possible to calculate the time-averaged values v and w from
the measurement series and thereafter v' and w' can be calculated for each individual
measurement. The Reynolds stress can then be calculated by calculating the time-averaged
value of v'w' (2
nd
momentum, i.e. variance) and by multiplying the value with density as
shown in (6-3). Without real measurements equation (6-3) is not very useful in
calculations and therefore further development is needed for those cases.
69
Prandtls mixing length theory
Prandtl introduced the mixing length theory in order to calculate the turbulent shear stress.
As we remember from above, turbulent shear stress is caused by particle movement from
one layer to another causing momentum exchange. Prandtl's mixing length theory is based
on the assumption that a fluid parcel has to travel over a length l
m
before its momentum is
transferred. Basically it is easy to understand the concept of the mixing length: since
turbulent stress is caused by momentum transfer, it is useful to have a physical length that
defines when the momentum has been transferred.
Fig. 6-2. Prandtl's mixing length theory (adapted from Liu 2001) (u refers to v).
Fig. 6-2 shows the time-averaged profile. The fluid particle locating in layer 1 and having
the velocity u
1
, moves to layer 2 due to eddy motion. There is no momentum transfer
during the movement, i.e. the velocity of the fluid parcel is still u
1
when it just arrives at
layer 2, and decreases to u
2
some time later by momentum exchange with other fluid in
layer 2. This action will speed up the fluid in layer 2, which can be seen as turbulent shear
stress
t
acting on layer 2 trying to accelerate layer 2 (Liu 2001).
Note: The momentum transfer discussed above happens in the vertical direction. It is
useful to point out now that later on in this material it is shown that there is momentum
transfer between the main channel and floodplain in the horizontal direction which causes
70
an additional resistance to the main channel flow. That is, there is a shear stress in the
imaginary wall between the main channel flow and the floodplain flow. In this case the
momentum is transferred in horizontal direction (perpendicular to the flow direction) and
particle moving from the main channel to floodplain will speed up the flow velocity in the
floodplain, which can be seen as additional stress acting on the boundary.
The horizontal instantaneous velocity fluctuation of the fluid parcel in layer 2 is (see Fig.
6-2).
dz
dv
l v v ' v
m 2 1
(6-4)
and assuming the vertical velocity fluctuation having the same magnitude
dz
dv
l ' w
m
(6-5)
where the negative sign is due to downward movement in this case. The turbulent shear
stress is now (see Eq. (6-3)):
2
2
m t
dz
dv
l ' w ' v

,
_

(6-6)
The very useful definition called kinematic eddy viscosity (m
2
s
-1
) is:
dz
dv
l
2
m
(6-7)
which has the same unit than the kinematic viscosity used in the shear stress in laminar
flow in Eq. (6-1). The turbulent shear stress can now be defined in a similar way compared
to laminar (viscous) shear stress:
dz
dv
t
(6-8)
6.2.3 Total shear stress
The total shear stress is the sum of viscous shear stress (it is acting also in turbulent flow)
and the turbulent shear stress:
71
dz
dv
dz
dv
t
+ +

(6-9)
6.2.4 Bottom shear stress
The derivation of the bottom shear stress is completely analogous to the derivation of the
Chezy coefficient in the previous material of this course.
Fig. 6-3. Fluid force and shear stress (adapted from Liu 2001).
Assuming uniform flow and balancing the forces acting on the darker area in Fig. 6-3 we
get
o z
xS ) z h ( g sin x ) z h ( g x (6-10)
which leads to
o z
S ) z h ( g (6-11)
where S
o
is the bottom slope. The bottom shear stress (z=0) in uniform flow is thus
o b
ghS (6-12)
72
Bottom shear stress
b
for an irregularly shaped cross-section can be easily derived by
taking into account that the shear stress is acting on the whole wetted perimeter P over the
whole length x.
sin x gA x P
b
(6-13)
and by remembering that hydraulic radius R=A/P the bottom shear stress equation will be
o b
gRS (6-14)
For wide channels R is around h and then (6-12) and (6-14) are almost equal. Moreover, in
the case that energy slope (friction slope) S
f
is known or can be calculated, S
o
in Eq. (6-14)
can be replaced by S
f
.
6.2.5 Friction velocity
The bottom shear stress is often converted to so called friction velocity v
*
, which is
defined by
o
b
*
gRS v

(6-15)
The unit of v
*
is the unit of velocity (m s
-1
) and it denotes the fluid velocity very close to
the bottom (see Eq. (6-24)).
6.3 Velocity distribution, bed roughness and friction factors in unvegetated
channels
6.3.1 Viscous sublayer
On the bottom of an unvegetated channel there is no turbulence and the turbulent shear
stress
t
=0, and therefore in a very thin layer above the bottom, viscous stress is dominant,
and hence flow is laminar in that layer called viscous sublayer.
73
Above the viscous sublayer the turbulent stress dominates and the total stress can be
calculated from Eq. (6-16), i.e. as the sum of viscous and turbulent shear stresses.
o t v z
S ) z h ( g + (6-16)
The total stress decreases linearly towards zero when approaching the water surface (z
approaches water depth h). The distribution of the shear stress as a function of z is
completely analogous with the hydrostatic pressure p=g(h-z) with the exception of the
viscous sublayer close to the bottom.
6.3.2 Scientific classification of flow layers in unvegetated channels
According to Liu (2001) the so called scientific classification of flow layers is the one
shown in Fig. 6-4.
Fig. 6-4. Scientific classification of flow regions in unvegetated channels (not to scale).
Adapted from Liu (2001).
Near the bottom there is the thin viscous sublayer where there is almost no turbulence.
Measurements show that the viscous shear stress in this layer is constant and equal to the
bottom shear stress
b
. Flow in this layer is laminar. In the transition layer viscosity and
74
turbulence are equally important. In the turbulent logarithmic layer viscous shear stress is
negligible and the turbulent shear stress is equal to the bottom shear stress. The Prandtl's
mixing length theory was developed for this layer and it leads to the logarithmic velocity
profile as shown later on. The turbulent outer region consists of about 80-90 % of the total
region and velocity is relatively constant due to the strong mixing of the flow (Liu 2001).
6.3.3 Bed roughness k
s
in unvegetated channels
Bed roughness k
s
is very often also called roughness height (karkeuskorkeus in Finnish) or
equivalent roughness (ekvivalentti karkeus).
One important question is how to determine the bed roughness k
s
. Nikuradse made his
experiments by glueing grains of uniform size to pipe surfaces. In completely flat bed
consisted of uniform spheres the bed roughness k
s
would be the diameter of the grains.
This cannot be found in nature, where the bed material is composed of grains with
different size and bottom itself is not flat but it includes ripples or dunes (see Fig. 6-5).
According to Liu (2001), the following k
s
values have been suggested based on different
type of experiments:
concrete bottom k
s
=0.001 - 0.01 m
flat sand bed k
s
= (1 - 10)d
50
bed with sand ripples k
s
=(0.5 - 1.0)H
r
: H
r
= ripple height (m)
Fig. 6-5. Ripples and dunes (adapted from Liu 2001).
75
6.3.4 Characterisation of smooth and rough flow in unvegetated channel
It is necessary to characterise the flow as hydraulically smooth or rough since it influences
e.g. the thickness of the viscous sublayer etc. A very big series of experiments were
carried out by Nikuradse for pipe flows. He introduced the concept of equivalent grain
roughness k
s
, which is usually called bed roughness for open channel flow. Based on the
experiments it was found that the following criteria can be used to characterise if flow is
hydraulically smooth, rough or in the transitional zone:
flow al transition lly hydraulica ; 70
k v
5
flow rough lly hydraulica ; 70
k v
flow smooth lly hydraulica ; 5
k v
s *
s *
s *

(6-18)
where v
*
is the friction velocity calculated using Eq. (6-15) and is the kinematic
viscosity.
6.3.5 Logarithmic velocity distribution in turbulent layer
It was described in section 6.3.2 that the measurements show that the turbulent shear stress
is constant in the turbulent logarithmic layer and it equals the bottom shear stress. By
assuming that the mixing length is proportional to the distance to the bottom, l
m
=z ( is
von Karman constant), Prandtl obtained the logarithmic velocity profile. By the
modifications done to the original theory, the logarithmic velocity profile applies also in
the transitional layer and in the turbulent outer layer. Measured and computed velocities
show reasonable agreement. This means that from an engineering point of view, two
different velocity profiles need to be considered (Liu 2001):
76
logarithmic velocity distribution, which covers the transition layer, turbulent
logarithmic layer and turbulent outer layer from Fig. 6-4.
velocity profile in the viscous sublayer
In the turbulent layer the total shear stress is assumed to contain only the turbulent shear
stress. The total shear stress increases linearly with depth
)
h
z
1 ( ) z (
b t
(6-19)
According to Prandtl's mixing length theory shown in section 6.2.2 and Eq. (6-6)
2
2
m t
dz
dv
l

,
_

(6-6)
Now comes out the modification of the original theory. Instead of assuming that the
mixing length l
m
=z like Prandtl assumed, l
m
is replaced by equation
5 . 0
m
)
h
z
1 ( z l (6-20)
It is possible to combine (6-20) and (6-6) to get
z
v
z
/
dz
dv
* b

(6-21)
Eq. (6-21) can be integrated from z
0
to h to get the logarithmic velocity profile.

,
_

0
*
z
z
ln
v
) z ( v (6-22a)
The integration constant z
0
is the elevation corresponding to zero velocity (see Fig. 6-6).
Note: Most often the "universal" logarithmic velocity distribution is shown in the form:
5 . 8
k
z
ln
1
v
) z ( v
s *
+

,
_

(6-22b)
77
See next page for comparison of (6-22a) and (6-22b)!
Note the similarity to the calculation of aerodynamic resistance r
a
needed in the
estimation of potential evapotranspiration rate using the Penman-Monteith equation! In
these calculations the roughness length is usually around 10-13 % of the crop height!
Fig. 6-6. Velocity distribution in hydraulically smooth and rough flow (graphs not to
scale).
The integration constant z
0
(m) is here based on the study conducted by Nikuradse for pipe
flows.

'

flow al transition lly hydraulica ; 70

k v
5 k 033 . 0
v
11 . 0
flow rough lly hydraulica ; 70
k v
k 033 . 0
flow smooth lly hydraulica ; 5
k v
v
11 . 0
z
s *
s
*
s *
s
s *
*
0
(6-23)
NOTE! Eqs. (6-22a) and (6-22b) are exactly the same since (6-22b) is valid from z-values
which give v(z) greater than or equal to zero. v(z) given by (6-22b) is zero when
) flow rough for ( z k 033 . 0 e k ) 0 v ( z
0 s
5 . 8
s


(6-22c)
78
The friction velocity v
*
is the fluid velocity very close to the bottom and it is the flow
velocity at elevation z=z
0
e

, i.e.
*
e
0
z z
v v

(6-24)
6.3.6 Velocity profile in the viscous sublayer
In the case of hydraulically smooth flow there is a viscous sublayer whereas in the rough
flow there is a height z
0
with zero flow (Figs. 6-6 and 6-7). Viscous shear stress is constant
in the viscous sublayer and it is equal to the bottom shear stress as shown in Fig. 6-4.
b
dz
dv

(6-25)
Eq. (6-25) can be integrated assuming that velocity v=0 when z=0. Integration gives
z
v
z ) z ( v
2
*
b

(6-26)
giving a linear velocity distribution in the viscous sublayer. The thickness of the viscous
sublayer can be obtained by finding the z value where the logarithmic velocity distribution
intersects the linear distribution giving a theoretical thickness of the viscous sublayer

(Liu 2001).
*
v
6 . 11

(6-27)
Summary of the velocity profiles in the hydraulically smooth and rough flow is given in
Fig. 6-7.
79
Fig. 6-7. Summary of the velocity profiles in hydraulically smooth and hydraulically rough
flow (in the graph u(z) refers to v(z) and u
*
refers v
*
).
6.3.7 Drag force
Flowing water moving past an object will exert a force called drag force
2
D D
Av C
2
1
F (6-28)
where C
D
is the coefficient and A is the projected area of the object to the flow direction.
Values of C
D
are usually around 1.0...2.0 but experiments are needed for more accurate
determination of the values. Drag force comes partly from the skin friction when water
moves around the object and partly from the pressure the moving water exerts on the
object.
Correspondingly to the drag force another force called lift force F
L
can be defined. It has
the same shape than the drag force:
2
L L
Av C
2
1
F (6-29)
where C
L
is the lift coefficient which also need to be determined experimentally.
80
6.3.8 Chzy, Manning and Darcy-Weissbach friction coefficients
The three most often used equation for calculating friction losses h
f
(Eqs. 6-30) are the
ones known as Chzy, Manning and Darcy-Weissbach equations.
Weissbach Darcy
g 2
v
R 4
L
f h
Manning
R
Lv
n h
Chzy
R
Lv
C
1
h
2
DW f
3 / 4
2
2
f
2
2
f

(6-30)
where L is the length of the river section and R is hydraulic radius. C is the Chzy
coefficient, n =1/k
St
is the Manning coefficient (the inverse of n is k
St
called the Strickler
coefficient) and f
WD
is the Darcy-Weissbach friction coefficient originally developed for
pipe flow. Note the h
f
is linearly related to f
WD
but to C
2
and n
2
.
There is the following connection between the different friction coefficients:
2 2 3 / 1
DW
gC 8 n gR 8 f

(6-31)
The three different friction factors are related to calculation of average velocity v (m s
-1
) in
the river as follows:
Weissbach Darcy RS
f
g 8
v
Manning S R
n
1
v
Chzy RS C v
f
DW
f
3 / 2
f

(6-32)
where S
f
is the energy slope (friction slope) and in uniform flow S
f
can be replaced by the
bottom slope S
o
.
81
The next step is to provide relationships between bed roughness k
s
and different friction
factors in unvegetated channels.
Chzy coefficient versus bed roughness k
s
Liu (2001) has derived the following relationships for Chezy coefficient C and bed
roughness k
s
in a broad channel (bottom width B
B
>>h, i.e. R can be replaced by h). If one
compares the well-known Chzy equation
f
RS C v (6-31)
and the equation for friction velocity v
*
, (note S
o
replaced by S
f
)
f *
gRS v (6-15)
it is possible to derive a connection between C, v
*
and average velocity v.
g
v
v
C
*
(6-33)
In the next step it is necessary to calculate the average velocity from the logarithmic
velocity profile by integrating it from z
0
to h (Liu 2001).
)
e z
h
ln(
v
h
z
1 )
z
h
ln(
v
dz )
z
z
ln(
h
v
dz ) z ( v
h
1
v
0
* 0
0
*
h
0
z
0
* h
0
z

1
]
1

(6-34)
In this way it is possible to relate the Chezy coefficient, water depth h and roughness
height (bed roughness) k
s
and critical velocity v
*
as follows (Liu 2001):

'

,
_

,
_

70
k v
; flow rough lly Hydraulica
k
h 12
log 18
5
k v
; flow smooth lly Hydraulica
3 . 3
hv 12
log 18
)
e z
h
ln(
g
C
s *
s
s * *
0
(6-35)
82
where the equations for z
0
, (6-23) have been utilized. The log used in the equations refers
to Briggs logarithm.
Bottom friction coefficient f versus roughness k
s
Liu (2001) shows derivation of a dimensionless friction coefficient f which corresponds to
the Darcy-Weissbach friction coefficient f
DW
originally derived for pipe flow. Here a
different symbol f is used since the derivations do not originate from theory for pipe flows.
The derivation is based on examining the forces acting on a grain resting on the bed. The
drag force is slightly modified from (6-28) by multiplying the depth averaged velocity v
by an empirical coefficient to take into account the fact that the true velocity near the
grain on the bottom is somehow related to v. Then it is possible to examine the shear stress

b
acting on the grain by saying that the horizontal force is the drag force acting on A'
which is the projected area of the grain to the horizontal plane:
2
D D
) v ( A C
2
1
F (6-36)
2 2 2
D
D
b
fv
2
1
v )
' A
A
C (
2
1
' A
F
(6-37)
where f is empirical friction coefficient corresponding the Darcy-Weissbach coefficient in
pipe flow.
)
' A
A
C ( f
2
D
(6-38)
Eq. (6-38) is not useful and therefore the following derivation is needed by utilising Eqs.
(6-37), (6-15) and (6-31):
2
o
2
o
2
b
C
g 2
RS C
gRS 2
v
2
f

(6-39)
where it has been assumed that flow is uniform. Eq. (6-39) can finally be converted to
hydraulically smooth and rough flow conditions by utilising (6-35).
83

'

1
]
1

,
_

1
]
1

,
_

70
k v
; flow rough lly Hydraulica
k
h 12
log
0555 . 0
5
k v
; flow smooth lly Hydraulica
3 . 3
hv 12
log
0555 . 0
C
g 2
f
s *
2
s
s *
2
*
2
(6-40)
Friction coefficient equation (6-40) together with Eq. (6-37) provide in some cases a
useful way to calculate the bottom shear stress
b
.
Colebrook-White equation relating f
WD
and k
s
was originally developed for pipe flows and
it relates the friction coefficient f
WD
with k
s
, Reynolds number Re and D is hydraulic
diameter of the pipe.

,
_

+
2
s
DW
1
DW
c
D / k
f Re
c
log 2
f
1
(6-41)
where the coefficient c
1
=2.51 and c
2
=3.71 are based on experiments in pipe flows. The
same values are used for open channel flow. Moreover, an effective hydraulic diameter
D
eff
is used instead real pipe diameter by introducing an empirical correction factor f
PL
, i.e.
D
eff
=f
PL
D (i.e. D
eff
is used in Eq. (6-41) but in the final equation (6-42) the pipe diameter D
is used):

,
_

+
2 PL
s
DW PL
1
DW
c f
D / k
f f Re
c
log 2
f
1
(6-42)
The first term on the right hand side of (6-42) refers to hydraulically smooth flow and it
can be neglected in hydraulically rough turbulent flow. Moreover, in open channel flow
the D is replaced by D=4R in open channel flow leading to relationship between f
DW
and
bed roughness k
s
and hydraulic radius R.
84

,
_

,
_

s
PL
2 PL
s
DW
k
R
f 84 . 14 log 2
c f
D / k
log 2
f
1
(6-43)
where f
PL
is usually 0.827. Eq. (6-43) resembles the equation derived by Liu in (6-40) but
with the comment that f
DW
=4f (the factor 2 before log-term raised to exponent 2 causes the
conversion factor 4). For wide channels R is almost the same than h.
Meyer-Peter empirical relationship between k
St
and d
90
Assuming that k
s
=d
90
, it is possible to use the Meyer-Peter empirical relationship between
k
St
(=1/n) and k
s
:
6 / 1
s
St
k
26
n
1
k (6-44)
6.3.9 Velocity distribution in unvegetated channels
Example 6-1. Consider an unvegetated wide channel with the following input data (see
Table 6-1):
________________________________________________________________________
Table 6-1. Input data for calculation of velocity distribution in unvegetated channel in the
case of rippled and flat bed.
________________________________________________________________________
85
The example is calculated for two different cases: rippled bed with ripple height H
r
=0.05
m and it is assumed that k
s
=H
r
for rippled bed. For flat bed k
s
is assumed to be 2.5d
50
.
In the calculation of the logarithmic velocity profile it is necessary to know the friction
velocity v
*
and z
0
in equation (6-22a) or (6-22b)

,
_

0
*
z
z
ln
v
) z ( v (6-22a) 5 . 8
k
z
ln
1
v
) z ( v
s *
+

,
_

(6-22b)
Note that (6-22a) and (6-22b) give identical result since there is a relationship between k
s
and z
0
, i.e. (6-22b) gives zero velocity at location
) flow rough for ( z k 033 . 0 e k ) 0 v ( z
0 s
5 . 8
s


(6-22c)
The calculation of friction velocity is carried out using Eq. (6-15) taking into account that
for a wide river or floodplain hydraulic radius R can be replaced by water depth h. Bottom
slope S
o
is replaced by friction slope S
f
. It is also useful to calculate bottom shear stress
b
from Eq. (6-14). Calculation of z
0
is carried out using (6-23).
86
________________________________________________________________________
Table 6-2. Calculation of additional variables needed in computation of velocity
distribution in unvegetated channel in the case of rippled and flat bed.
________________________________________________________________________
The additional variable shown in Table 6-2 are the thickness of the viscous sublayer for
smooth flow (Eq. (6-27)), test criteria v
*
k
s
/ for Eq. (6-18) to check the type of flow.
The velocity distribution is shown as numbers in Table 6-3 and in Fig. 6-8a for the whole
depth range (01 m) and in Fig. 6-8b for the narrow range close to the bottom.
Fig. 6-8a. Velocity distribution in rippled and flat bed: water depth h=1.0 m, S
f
=0.0001.
87
Fig. 6-8b. Velocity distribution in rippled and flat bed near the bottom: water depth h=1.0
m, S
f
=0.0001.
________________________________________________________________________
Table 6-3. Velocity distribution in unvegetated channel; water depth h=1.0 m, S
f
=0.0001,
ripple height H
r
=k
s
=0.05 m and k
s
=2.5d
50
= 0.0075 m for flat bed (d
50
=0.3 mm).
________________________________________________________________________
As shown in Figs. 6-8a and b and in Table 6-3, the difference between the velocity
88
distribution in rippled and flat bed is quite large and the difference stems from the very
narrow range near the bottom (Fig. 6-8b) where v(z) increases fast.
Average velocity v
A
can be calculated by numerical integration of the velocity distribution
but very accurate results can be obtained using Eq. (6-34).
)
e z
h
ln(
v
v
0
*
A

(6-34)
In the case of rippled bed the average velocity is v
A
=0.42 m s
-1
and for flat bed v
A
=0.57 m
s
-1
.
Influence of bottom roughness k
s
on average velocity calculated using Eq. (6-34) is shown
in Table 6-4.
________________________________________________________________________
Table 6-4. Depth-averaged velocity v
A
as a function of bottom roughness k
s
: water depth
h=1.0 m, S
f
=0.0001.
________________________________________________________________________
The results of Table 6-4 show that when flow is smooth the depth-averaged velocity is
v
A
=0.905 m s
-1
and for first rough flow case (k
s
=0.004 m), v
A
= 0.62 m s
-1
, i.e. about 31 %
smaller. The results of Table 6-4 show how crucial it is to be able to evaluate the bed
roughness correctly. Comparison of the depth-averaged velocity calculated from velocity
distribution is also compared to velocity given by the Darcy-Weissbach equation (6-32).
89
The results given in the last three columns in Table 6-4 show that the difference between
the values is quite small. Calculation of the Darcy-Weissbach friction factor f
DW
is carried
out using Eq. (6-43) and assuming that f
PL
=0.827 and R is replaced by h. The results are
very close to each other for rough flow conditions, i.e. conditions prevailing in natural
channels.
6.3.10 Exercise
Calculate velocity distribution v(z) (show also as a graph) in unvegetated channel using
the following input data. Assume that the bottom is rippled and k
s
=H
r
.
Calculate average velocity by integrating numerically the velocity distribution over the
depth h. Calculate also average velocity using Eq. (6-34) and compare the result with
average velocity calculated using the Darcy-Weissbach equation.
6.4 Velocity distribution, roughness height and friction factors in vegetated
channels
6.4.1 Unsubmerged-submerged flow, deflected-undeflected vegetation
In the estimation of the influence of vegetation on friction coefficient f
DW
it is very
important to know if vegetation is rigid or is deflected as flow velocity in increasing.
90
Fig. 6-9. Example of deflected vegetation (experiments carried out by J.
Jrvel/HUT/Laboratory of Water Resources).
Fig. 6-10. Example of rigid and deflected vegetation (experiments carried out by J.
Jrvel/HUT/Laboratory of Water Resources).
6.4.2 Classification of flow layers in vegetated channel
According to El-Hakim and Salama (1992), three different flow layers can be
distinguished in the case of submerged vegetation as shown in Fig. 6-11. The
corresponding velocity distribution is also shown.
91
Fig. 6-11. Classification of three flow layers and the corresponding velocity distribution
in the case of totally submerged and deflected vegetation (adapted from El-Hakim and
Salama, 1992).
Zone 1 refers to flow inside the canopy. The thickness of this layer is h
p
, which denotes
the deflected height of vegetation. In the case of stiff plants h
p
would not be influenced by
the velocity, but for flexible plants like shown in Fig. 6-11, the deflected height decreases
as velocity increases. The mathematical formulation of the velocity distribution is
discussed in Section 6.4.4.
Zone 2 is an intermediate layer (transition zone) between zone 1 and zone 3. In zone 3 the
velocity distribution is assumed to be logarithmic. According to Fig. 6-11, it is assumed
that the logarithmic velocity distribution is valid in zone 3, but the zero point of velocity is
assumed to be inside the vegetation (broken line). The variable h' is called the shift of the
zero point from h
p
in logarithmic velocity distribution, i.e. the zero velocity is at height
z=h
p
-h'. A special velocity, v
hp
is velocity just at the top of the deflected height h
p
.
Fig. 6-12 shows a dye experiment where rhodamin is inserted at height which is turbulent
zone 3.
92
Fig. 6-12. Dye experiment showing the effect of turbulent flow (= indication of the
magnitude of the turbulent shear stress) (experiments carried out by J.
Jrvel/HUT/Laboratory of Water Resources).
6.4.3 Roughness height k
p
in vegetated channels
In the case of vegetation it is useful to define a vegetation roughness height k
p
which
basically corresponds the roughness height k
s
of the unvegetated channels. The parameter
k
p
takes into account both the influence of bottom and the vegetation. The dependence of
vegetation roughness height k
P
on the deflected vegetation height h
p
has been studied in
very many laboratory/field experiments. No universal agreement exists but Stephan (2001)
has collected the results of many experiments. Some of the results have been collected to
Table 6-5.
93
______________________________________________________________
Table 6-5. Connection between deflected vegetation height h
p
and vegetation roughness
height k
p
and h' (= the shift of the zero point from h
p
in logarithmic velocity distribution).
______________________________________________________________
______________________________________________________________
According to Kouwen et al. (1969), k
p
can be calculated from Eq. (6-45):
)
h
p
h
1 )(
1
C 1 (
p p
e h k

(6-45)
where h is water depth and C
1
is an empirical coefficient (1.52 < C
1
< 2.69).
6.4.4 Velocity distribution in vegetated channels
The computation of velocity distribution inside the vegetation in zone 1 and in the
turbulent zone 3 above the canopy will be discussed next (see Fig. 6-11). The methods
used to calculate the velocity distribution above the vegetation (canopy) will be discussed
first and thereafter in connection with each method the calculation of velocity distribution
inside the canopy will be discussed.
Method of Plate and Quraichi (1965)
In the method of Plate and Quraichi (1965, ref. Stephan 2001), calculation of logarithmic
velocity distribution above the canopy is shown in Eqs. (6-46). The location of height
94
where v(z)=0 is shown to be h
p
+k
p
, i.e. above the deflected height of the canopy, which
makes the equation slightly different from other methods presented below.
p p
p p
p
p
*
h 15 . 0 k
k h ) 0 v ( z
) 0 v ( z z ;
k
h z
ln
1
v
) z ( v

+
>

,
_

(6-46)
The calculation of velocity profile inside the canopy is not shown in the reference. The
friction velocity v
*
is calculated here (and in all other models as well) by computing first
bottom shear stress
b
just above the canopy from Eq. (6-12)
o b
ghS by replacing h
by h-h
p
and S
0
by friction slope S
f
(slope of the line). Friction velocity v
*
is calculated
from (6-15), i.e. v
*
=(
b
/)
1/2
.
Method of Kouwen et al. (1969)
In the method of Kouwen et al. (1969, ref. Stephan 2001), calculation of velocity profile
above the canopy is carried out using Eqs. (6-47)

> +

,
_

p
h
0
*
2
2
C
p
2
p *
vdz
h v
1
C
e k ) 0 v ( z
) 0 v ( z z ; C
k
z
ln
1
v
) z ( v
(6-47)
where k
p
is calculated using Eq. (6-45). The exact evaluation of coefficient C
2
is not
described in Stephan (2001). However, in the method it is assumed that velocity inside the
canopy is constant and calculated from Eq. (6-48).
95
*
1 p
v C t tan cons ) z ( v (6-48)
and by assuming constant velocity in the canopy, C
2
from (6-47) can be approximated by
h
h C
h v
h v C
dz ) z ( v
h v
1
C
p 1
*
p
*
1
p
h
0
p
*
2

(6-49)
However, as shown in Example 6-2, Eq. (6-49) leads to quite small values of C
2
and
therefore, the height z(v=0) is quite high and velocity calculated using Eq. (6-47) leads to
small velocities just above the canopy. In Example 6-2 this problem is solved by finding a
C
2
value which gives a continuous velocity profile, i.e. find C
2
that gives v(z=h
p
) values
from Eq. (6.47), which is greater than v
p
(z). Without further details of the method of
Kouwen, its use seems to be very problematic.
Method of Haber (1982)
In the method of Haber (1982, ref. Stephan 2001), calculation of velocity profile above the
canopy is carried out using Eqs. (6-50)
*
v
hp
v
p
*
hp
p *
e k ) 0 v ( z
) 0 v ( z z ;
v
v
k
z
ln
1
v
) z ( v


> +

,
_

(6-50)
where v
hp
is the velocity at height z=h
p
, i.e. at the height of the deflected canopy (see Fig.
6-10). Stephan (2001) does not mention the specific method used by Haber (1982) to
calculate v
hp
and in this case the method given by Kouwen et al. (1969) is used (basically
the same equation than 6-48).
96
*
1 hp
v C v (6-51)
Inside the canopy the velocity profile v
p
(z) is calculated from

,
_

1
p
h
z
1
hp
p
e
v
) z ( v
(6-52)
where
1
is a vegetation parameter which usually is around 0.40-1.56. An equation for
calculating
1
is also given
p
2 / 1
2
H
1 v W
1
h
L 4
g 2 A C

,
_

(6-53)
where C
W
is the vegetation drag coefficient (around 1.0-1.5), A
v
is the horizontal projected
area of vegetation (perpendicular to flow direction, i.e. the area of vegetation that the flow
"sees" when flowing through the canopy), L
H
is the Prandtl's mixing length and
1
an
empirical parameter. However, the use of Eq. (6-53) is not straightforward.
Method of El-Hakim and Salama (1992)
In the method of El-Hakim and Salama (1992, ref. Stephan 2001), calculation of velocity
above the canopy is carried out using Eq. (6-54).
3
A
3
B
p
p 3
3
*
e k ) 0 v ( z
) 0 v ( z z ;
k
z
ln
B
1
A
v
) z ( v

>

,
_

+
(6-54)
where parameter A
3
is 1.84 and B
3
=0.35 (El-Hakim and Salama 1992). Inside the canopy
the following equation is used:
97
1
B
p
1
hp
p
h
z
A
v
) z ( v

,
_

(6-55)
where A
1
and B
1
are empirical parameters. The values used by the original reference are
not known and the following parameters are used here: B
1
is assumed to be around 1.3-2.0
and A
1
is scaled in such a way that v
p
(z) at height z=h
p
from Eq. (6-55) will give the same
value than v(z) obtained at height z=h
p
from Eq. (6-54) thus giving a continuous velocity
profile at the height of the deflected vegetation. This leads to
hp
p
1
v
) 54 6 _( from _ ) h ( v
A

(6-56)
Method of Tsujimoto et al. (1992)
In the method of Tsujimoto et al. (1992, ref. Stephan 2001), calculation of velocity above
the canopy is carried out using Eq. (6-57).
p
hp
*
v
T , hp
v
hp
*
T , hp
hp
hp ) p
*
h
I e I
) 0 z ( v
v
v
I
I h z (
ln
1
v
) z ( v
+

,
_

+
(6-57)
where v
hp,T
is the velocity at the height of the deflected canopy as calculated by Tsujimoto
(1992) and I
hp
is the mixing length. I
hp
and v
hp,T
calculated from relatively complicated
expressions shown in (6-58).
98
y x
p
v
v W
f
p
p p T , hp
p
p
f p
p x
p T , hp
*
hp
a a
d
C
gS 2
' v
) h h ( 1 ' v v
) h h ( 1
1
1
2
h
S ) h h (
log 85 . 0 32 . 0
h a
1
) ' v v (
v
I

,
_

+
+

,
_

,
_

(6-58)
where S
f
is the slope of the energy line (friction slope), a
x
and a
y
distances of canopy
elements in the direction of flow and in the direction perpendicular to flow (m),
respectively, d
p
is the diameter of the canopy element (m), v
p
' is called characteristic
velocity inside the canopy,
v
is vegetation density parameter (m
-1
) and C
W
is the drag
coefficient of vegetation. As shown by Eqs. (6-58), the vegetation density is explicitly
included in the model.
Velocity inside the vegetation is computed from Eq. (6-59):
, , )
p
h z (
p
p
p
e ) h h ( 1
' v
) z ( v

+ (6-59)
where the characteristic velocity v
p
' is given in (6-58). According to Example 6-2 (section
6.7.2), Tsujimoto's method gives considerably higher values than the above mentioned
methods.
99
Method given by Stephan (2001)
Stephan gives an equation for calculating the velocity distribution above the canopy:

+
> +

,
_

5 . 8
m , p m , p
m , p
m , p
*
e h h ) 0 v ( z
) 0 v ( z z ; 5 . 8
h
h z
ln
1
v
) z ( v
(6-60)
where h
p,m
can be taken as h''=h
p
-h'. The velocities calculated by this method are
comparable, or even higher than the values given by the method of Tsujimoto et al. (1992).
6.4.5 Influence of vegetation density on friction factors in vegetated channels
In unsubmerged case the vegetation friction coefficient f
V
is usually computed from
Eq. (6-61):
W v
y x
W p
V
C h 4
a a
C hd 4
f (6-61)
where h is water depth, a
x
and a
y
distances of canopy elements in the direction of flow and
in the direction perpendicular to flow (m), respectively, d
p
is the diameter of the canopy
element (m) and C
W
is the drag coefficient of vegetation (usually around 1.0-1.5) and
v
is
vegetation density parameter (m
-1
) as calculated in (6-58). In the case of vegetation,
bottom friction can usually be neglected and velocity v
V
in the vegetated zone can be
calculated by combining (6-61) into Darcy-Weissbach equation shown in (6-32) and by
setting R=h indicating that it is question of floodplain flow:
v W
f
f
V
V
C
gS 2
ghS 8
f
1
v

(6-62)
100
It is useful to notice that basically v
V
is the same velocity, v
p
' that was computed in (6-58).
v
p
' is called characteristic velocity inside the canopy.
In submerged cases Wu et al. (1999) give an equation for velocity inside the canopy v
V,S
:
v W
f
p
S , V
C
gS 2
h
h
v

(6-63)
indicating that velocity inside the canopy is increased as the submergence ratio, h/h
p
is
increased. In the derivation of Eq. (6-63), the vegetation was assumed to be rigid and here
it is assumed that the same equation can be used for deflected vegetation as well.
However, this assumption has not been verified.
6.4.6 Velocity distribution in the case of submerged vegetation
Example 6-2. The input data needed to calculate the velocity distribution in the case of
submerged vegetation is shown in Table 6-6. The methods used are: Plate and Quraichi
(1965), Kouwen et al. (1969), Haber (1992), El-Hakim and Salama (1992), Tsujimoto at
al. (1992) and Stephan (2001).
________________________________________________________________________
Table 6-6. Input data needed to calculate velocity distribution in the case of submerged
vegetation.
________________________________________________________________________
101
The equations used to calculate the velocity distribution above the canopy and inside the
canopy are shown in section 6.4.4. Velocity profile inside the vegetation is not calculated
in the methods of Plate and Quraichi (1965) and Stephan (2001).
The friction velocity v
*
is needed in the equations used to calculate the logarithmic
velocity distribution above the canopy. v
*
can be computed from Eq. (6-15) by taking into
account that R can be replaced h-h
p
(water depth above the canopy) and S
f
is used.
________________________________________________________________________
Table 6-7. Shear stress and friction velocity v
*
at the height of deflected canopy h
p
the
canopy friction.
________________________________________________________________________
Table 6-8 shows the parameters needed in different methods. The parameter C
2
needed in
the method of Kouwen et al. (1969) is not calculated from Eq. (6-49) but a value for C
2
is selected in such a way that the velocity distribution curve is continuous throughout the
profile. Calculation of vegetation density
1
needed in Haber's method is given as input
data, not calculated from Eq. (6-53). The vegetation density
v
(calculated from Eq. (6-58)
is needed in the method of Tsujimoto et al. (1992) and the parameters needed are plant
diameter d
p
, distance between plants in horizontal direction, a
x
and a
y
, respectively.
102
________________________________________________________________________
Table 6-8. Parameters needed in the calculation of velocity distribution inside and above
the canopy.
________________________________________________________________________
The velocity distribution inside the canopy is based on computing the velocity at the
height of the deflected canopy, v
hp
, either using method of Kouwen et al. (1965) given in
Eq. (6-51) or by equation (6-58) by Tsujimoto et al. (1992) for the corresponding value
v
hp,T
. The characteristic velocity v
p
' inside the canopy is calculated using equation shown
in (6-58).
The results of the comparison are shown in Table 6-9 and in Fig. 6-16.
103
________________________________________________________________________
Table 6-9. Comparison of different methods to calculate velocity distribution inside and
above submerged vegetation.
________________________________________________________________________
104
Fig. 6-13. Comparison of velocity distribution inside and above submerged vegetation.
According to Table 6-9 and Fig. 6-13 the methods give quite different velocity
distributions especially above the canopy. Methods of Tsujimoto et al. (1992) and Stephan
(2001) give much higher velocities above the canopy compared to the four other methods.
Table 6-9 shows the average velocity inside the canopy using four different methods and
above the canopy using six different methods. The depth averaged velocity in submerged
case, v
A,S
is calculated as
h
) h h ( v h v
v
p ABOVE p IV
S , A
+
(6-64)
where v
IV
is the average velocity inside the canopy, v
ABOVE
is the average velocity above
the canopy. v
IV
and v
ABOVE
are calculated numerically.
105
6.4.7 Exercise
Calculate velocity distribution in the case of submerged submerged vegetation using the
method of Tsujimoto et al. (1992). The input data is given below:
Influence of the bottom friction need not be considered in the vegetated case.
Assume that the width of the channel is very big compared to depth: both a) and b)
6.5 Interaction between main channel flow and floodplain flow
6.5.1 Momentum transfer and shear stress between main channel and flood plain
In section 6.2.2 the momentum transfer in vertical direction was discussed and it was
shown that the momentum transfer creates a turbulent shear stress. Basically the same
principle holds between main channel and floodplain. Velocity in the main channel is
usually much higher compared to velocity in the floodplain (this velocity can be calculated
from Eq. (6-62)). In this case the momentum is transferred in horizontal direction
(perpendicular to the flow direction) and particle moving from the main channel to
floodplain it will speed up the flow velocity in the floodplain. The velocity difference
induces momentum transfer between the main channel and the floodplain and this causes
additional turbulent shear stress at the interface between the floodplain and the main
channel. One can visualise this as an "imaginary wall" and the shear stress can be
converted to additional friction coefficient of the imaginary wall, f
IW
. This friction
coefficient reduces the main channel flow velocity near the interface and hence the
conveyance capacity of the main channel will also be reduced.
106
Assuming that gravity is the force acting in the flow direction and it is balanced by the
shear stresses in the bottom of the main channel and in the imaginary walls in the left and
right hand side of the main channel, it is possible to write an equation (see Fig. 6-14):
x h x h x P xS gA
L , I L , b R , I R , b m m , b f
+ + (6-65)
where A is the cross-sectional area of the main channel, x is the length of the river
section studied, S
f
is friction slope,
b,m
is the bottom shear stress of the main channel,
b,R
and
b,L
are the shear stresses of the imaginary walls (interfaces) at the right- and left-hand
sides, respectively. P
m
is the wetted perimeter of the main channel, and h
I,R
and h
I,L
are the
water depths at the interface in the right- and left-hand sides, respectively. Assuming a
symmetric case, Eq. (6-65) can be converted to
, ,
m , b m f
T , I
I , b
P gAS
P
1


(6-66)
where,
b,I
is the shear stress (N m
-2
) acting on the interface and P
I,T
is the total length of
the interfaces (=h
I,R
+h
I,L
).
Fig. 6-14. Shear stresses acting in the interfaces between main channel flow and the
floodplains at the left- and right-hand sides.
107
6.5.2 Exercise
Calculate main channel and floodplain velocities and discharges in the cross-section
shown below using either the combination method (example 6-4) or Nuding's method
(example 6-5). The influence of the imaginary wall has to be taken into account. Calculate
also the discharge when the influence of the interaction zone and vegetation is neglected
and the average velocity is assumed to be the same both in the main channel and in the
floodplain.
Input data shown below:
108
6.6 Methods used to evaluate the influence of vegetation on conveyance
capacity of the main channel and the floodplains
The next step is to calculate the friction factor f
IW
of the imaginary walls between the main
channel and the floodplains and include this friction in the calculation of the main channel
conveyance capacity. In the computation it is assumed that the influence of the imaginary
wall friction is not taken into account in the calculation of the floodplain flow due to the
small velocity and great friction caused by the vegetation.
The idea is to calculate separately the main channel average flow velocity as influenced by
the bottom roughness and the additional friction of the interaction zone between main
channel and floodplain (see Fig. 6-15). The horizontal velocity distribution is shown in
Fig. 6-16 indicating that near the interface the velocity of the main channel is reduced due
to the influence of the friction of the imaginary wall.
109
Three different methods are introduced to take into account the effect of the momentum
transfer between main channel and floodplain:
1 Combination of different methods (ref. Nuding 1991)
2 Nudings method (Nuding 1991; Helmi 2002)
3 Merten's method (Nuding 1991; Jrvel 1998; Helmi 2002)
Equations used to calculate the friction factor and its influence of main channel
conveyance are shown in section 6.7.4 in the solution of the examples.
Fig. 6-15. Example of cross-section with floodplain on the left-hand side. The wetted
perimeter of the imaginary wall is h
3
= water depth in the floodplain.
110
Fig. 6-16. Velocity distribution in the main channel in the floodplain.
6.6.1 Calculation of composite roughness using iterative method
Example 6-3. In natural channels the bottom roughness k
s
is not constant throughout the
cross-section, but varies as shown in Fig. (6-17). It would be possible to calculate an
average roughness by weighting each roughness area denoted by k
s,i
proportional to the
length it takes from the total wetted perimeter:

N
1 i
i
N
1 i
i i , s
av , s
P
P k
k (6-67)
111
Fig. 6-17. Basic principle of the method for calculating composite friction factor f
DW
and
friction factor f
DW,i
for each section i (adapted from Nuding 1991).
However, it is more preferable to use an iterative method (ref. Nuding 1991) to take into
account the fact that different roughness areas contribute to total flow depending on the
friction factor f
DW,i
related to the corresponding k
s,i
. It means that each section with
different k
s,i
is also related to the corresponding A
i
and R
i
. In the calculation of Darcy-
Weissbach friction factor f
Dw
from Eq. (6-43) hydraulic radius is needed as input data.
When the bottom has several sections with different k
s,i
it is necessary to know the
corresponding hydraulic radius R
i
. In the iteration method suggested here the problem is
solved using the following procedure:
0 Give initial value for all R
i
, i=1, N, where N is the number of different type of
roughness areas:

N
1 i
i TOT
TOT
TOT
0 , i
P P
N , 1 i ;
P
A
R
(6-68)
1 Calculate f
DW,i,n
and a new estimate for the hydraulic radius from Eqs. (6-69):
112

,
_

+

+
i n , i , DW
n , i , DW TOT
n , i 1 n , i
i , s
n , i
n , i , DW
P f
f A
R 5 . 0 R
N , 1 i ; )
k
R
84 . 14 log( 2
f
1
(6-69)
where subindex n refers to iteration number. Repeat Eq. (6-69) until the successive
hydraulic radius values do not change more than a predefined iteration stopping criteria
(e.g. 0.001 m). In (6-69) it has been assumed that f
PL
=1.0.
2 Composite roughness f
DW
and average velocity v can be calculated from Eqs. (6-70)
and (6-71):
N , 1 i ;
P f
P
f
1
i i , DW
TOT
DW

(6-70)
TOT
TOT
TOT
WD
f TOT
P
A
R
f
S gR 8
v

(6-71)
3 The contribution areas A
i
can be calculated from Eq. (6-72).
N ,.., 1 i ; P R A
i i i
(6-72)
Example 6-3. Calculate the individual friction coefficients f
DW,i
for the cross-section
shown in Fig. (6-18).
113
Fig. 6-18. Cross-section used in example 6-3 to calculate the composite roughness.
________________________________________________________________________
Table 6-10. Input data for calculation of the composite roughness iterative method.
________________________________________________________________________
The cross-section has six different k
s,i
values and the corresponding wetted perimeters P
i
.
The results of the iteration are shown in Table 6-11 indicating that 5 iterations are needed
in this case.
114
_______________________________________________________________________
Table 6-11 Results of the calculation of the composite roughness iterative method.
________________________________________________________________________
6.6.2 Comparison of two different methods to evaluate the influence of effect of
floodplain on main channel conveyance
Two different methods will be compared to calculate the influence of the momentum
transfer between main channel and floodplain on conveyance of the main channel.
1 Combination of different methods (ref. Nuding 1991)
2 Nudings method (Nuding 1991; Helmi 2002)
The methods will be presented in the examples 6-4 and 6-5 shown in this section. The case
calculated is described below:
Problem formulation for Examples 6-4 and 6-5. Calculate average velocity and
conveyance capacity of the main channel and the floodplain in a cross-section shown in
Fig. 6-19. In the calculation section B
4
is assumed to be very small so that the left-hand
side of the floodplain is assumed to be vertical. The input data is given in Table 6-12.
Bottom width B
B
is 3.0 m. Four other sections (areas) 14 are denoted in the cross-
115
section. Section 4 is inclined in Fig. 6-19, but in examples 6-4 and 6-5 the section is
assumed to be vertical (a very small value given for B
4
).
Fig. 6-19. Cross-section of examples 6-4 and 6.5.
Nudings method takes this into account the influence of branches in plants and
vegetation density parameter
v,N
is calculated in Nuding's method from Eq. (6-73)
y x
z
y
z p
N , v
a a
a
a
d d +
(6-73)
where a
x
and a
y
are distances of canopy elements in the direction of flow and in the
direction perpendicular to flow (m), respectively, d
p
is the diameter of the canopy element
(m). d
z
is the thickness of the branches of the plant (horizontal elements) and the distance
of the branches in vertical direction is assumed to a
z
.
________________________________________________________________________
Table 6-12. Input data used in the comparison of different methods to evaluate the
influence of the momentum transfer between main channel and floodplain on the average
velocity and conveyance capacity of the main channel.
________________________________________________________________________
116
________________________________________________________________________
The combination method does not take into account the influence of branches in plants and
vegetation density parameter
v
is calculated using equation shown in Eq. (6-58). In this
material it is assumed that a so called equivalent diameter is used in the combination
method. The equivalent diameter d
p,e
is calculated by approximating the equivalent
diameter with the nominator of Eq. (6-73)
z
y
z p e , p
a
a
d d d + (6-74)
which is shown in Table 6-12 and this value is used in the combination method.
Example 6-4. Solution of the cross-section shown in Fig 6-19 and in Table 6-12 using the
combination method. The method has been collected from various equations referenced by
Nuding (1991). In the first step discharge in the main channel is calculated without taking
into account the influence of the shear stress of the imaginary wall between main channel
and floodplain. The results are shown in Table 6-13. Friction factor f
DW
is calculated from
117
Eq. (6-43) assuming that f
PL
=1.0. Hydraulic radius R
M
is calculated using the wetted
perimeter of the main channel, P
MAIN
(bottom B
B
+ wetted perimeters of sections 1 and 2).
The velocity calculated without interaction term is denoted by v
M,0
.
________________________________________________________________________
Table 6-13. Combination method. Calculation of Darcy-Weissbach friction factor,
hydraulic radius, average (ideal) velocity v
M,0
and discharge in the main channel without
taking into account the momentum transfer between main channel and floodplain.
________________________________________________________________________
________________________________________________________________________
In the next step it is necessary to calculate the floodplain velocity and discharge using the
vegetation density defined as input data in Table 6-12. Hydraulic radius of the floodplain,
R
F
, is calculated first and then the friction factors corresponding to vertical wall (section 4)
and bottom of the floodplain (section 3) from Eq. (6-43). Since the k
s
-values are the here
the same, the friction factors are the same for sections 3 and 4. Vegetation friction factor
f
V,C
is calculated from
y x
e , p
C , v
C , v W F C , V
a a
d
C R 4 f


(6-75)
where the equivalent plant diameter d
p,e
is calculated from Eq. (6-74). Friction factor for
the floodplain f
F,C
(vegetation and bottom) is calculated from the floodplain bottom
friction f
FB,C
and vegetation friction f
V,C
.
118
C , V C , FB P
P
C , F
f f ) 1 (
1
f
1
+

(6-76)
where
p
is Kaiser's volume ratio (usually around 0.001..0.02, ratio of the volume of plant
elements from the total volume under water). Floodplain bottom friction f
FB,C
is calculated
by weighted average of friction factors for sections 3 and 4 (here the same)
4 , DW 4 3 , DW 3
4 3
C , FB
f P f P
P P
f
1
+
+
(6-77)
Average velocity in the floodplain v
F,C
is finally computed by slightly modifying Eq. (6-
32) or Eq. (6-62):
C , F
f F
C , F
f
S gR 8
v (6-78)
________________________________________________________________________
Table 6-14. Combination method. Calculation of average velocity and discharge in the
floodplain.
________________________________________________________________________
________________________________________________________________________
In the combination method the friction factor of the imaginary wall, f
IW,C
is assumed to be
composed of two parts: a) vegetation induced shape friction f
S,0
and b) influence of the
interaction f
I,C
.
119
C , I 0 , S C , IW
f f f + (6-79)
f
S,0
-values have been reported to lie between 0.06 and 0.10 (0.09 assumed here). f
I,C
is
computed by a two-stage procedure. Compute first the maximum value for f
I,C
using the
ratio of velocities v
M,0
(ideal velocity in the main channel without interaction term) and
floodplain velocity v
F,C
.

,
_

IW
2
C , F
2
0 , M
max , C , I
P v
v
0135 . 0 log 18 . 0 f (6-80)
where P
IW
is the wetted perimeter of the imaginary wall (=h
2
in this case). In the next step
it is necessary to check if the width of the vegetation is smaller than a maximum width
b
W,F,max
.
f
max , C , I
2
C , F
max , F , W
gS
f v
5 . 17 b (6-81)
If the true width (here B
3
+ B
4
) is smaller than b
W,F,max
, f
I,C
is smaller than the maximum
value f
I,C,max
(linear decrease). The results for the example are shown in Table 6-15. Here
true width is greater than b
W,F,max
and therefore f
I,C
gets the maximum value f
I,C,max
=0.094
and the total friction factor is calculated from (6-79) when f
S,0
=0.09.
________________________________________________________________________
Table 6-15. Combination method. Calculation of the friction factor in the imaginary wall
between main channel and floodplain.
________________________________________________________________________
________________________________________________________________________
120
In the next step the iterative procedure described in example 6-3 (section 6.7.3) and in Eqs.
(6-68)-(6-69) is applied for the main channel to take into account the influence of the
interaction zone. The result of the iteration is shown in Table 6-16.
________________________________________________________________________
Table 6-16. Combination method. Calculation of main channel friction using iterative
method. Influence of friction factor in the imaginary wall is taken into account.
________________________________________________________________________
________________________________________________________________________
Note that in the iteration only the friction factor of the main channel is changed using Eq.
(6-69), and the friction factor of the interaction zone is not iterated. Finally, the main
channel friction factor is calculated from Eq. (6-70) where the interaction term is included.
P
TOT
is here the sum of the main channel wetted perimeter P
MAIN
and the wetted perimeter
of the interaction zone P
IW
(=h
2
in this case). Average velocity in the main channel is
calculated from Eq. (6-71). As shown in Table 6-17, the reduction due to the influence of
the imaginary wall is 21 % compared to case when this term is neglected.
________________________________________________________________________
Table 6-17. Combination method. Calculation of main channel velocity and discharge
when shear stress in the imaginary wall is taken into account.
________________________________________________________________________
121
________________________________________________________________________
Example 6-5. Solution of the cross-section shown in Fig 6-19 and in Table 6-12 using the
Nuding's method. The equations have been shown in Nuding (1991) and Helmi (2002).
The first step is the calculation of the main channel velocity by neglecting the interaction
term. The results are shown in Table 6-18. The result is the same as that shown in Table 6-
13 for the combination method, i.e. v
M,0
=1.265 m s
-1
.
________________________________________________________________________
Table 6-18. Nuding's method. Calculation of main channel velocity by neglecting the
influence of the interaction zone.
________________________________________________________________________
________________________________________________________________________
In the second phase slightly different method is used to calculate average velocity in the
floodplain depending on the vegetation density parameter
v,N
calculated from Eq. (6-73).
In the case of dense vegetation,
v,N
1.0, floodplain velocity vF,N is calculated from
N , v W
f
N , F
C
gS 2
v

(6-82)
If vegetation is assumed to be sparse,
v,N
<1.0, then the friction factor of the bottom of
floodplain, f
FB,N
(Eq. (6-43)) and friction factor f
V,N
for sparse vegetation are calculated.
122
N , v W F N , V
C h 4 f (6-83)
where h
F
is water depth at the location of the interaction zone. The total floodplain friction
factor f
F,N
(bottom + vegetation) is calculated from Eq. (6-76) (basically the same
equation that is used in the combination method). The velocity in the floodplain is then
calculated from (6-78). In (6-76) and (6-78) the friction values are taken from Nuding's
method, i.e. f
BF,N,
f
F,V
and f
F,N
are used.
________________________________________________________________________
Table 6-19. Nuding's method. Calculation of friction factor, velocity and discharge in the
floodplain.
________________________________________________________________________
________________________________________________________________________
The third phase of Nuding's method is to calculate the friction factor of the imaginary
wall. The calculation of f
IW,N
is based on estimating the effective width b
EFF
of the
interaction zone using the procedure shown in Eq. (6-84). First a test parameter b
N
is
calculated and if b
N
a
y
, then b
B
=a
y
, otherwise b
B
=b
N
. If b
B
b
v
(true width of the
vegetation zone), then b
C
=b
v,
otherwise b
C
=b
B
. Finally if b
C
0.15h
F
(water depth at the
location of the interaction zone), then b
EFF
= 0.15h
F
, otherwise b
EFF
= b
C
. However, in the
case that vegetation is bushes, then b
EFF
= b
N
.
123
(6-84)
d
p,y
in Eq. (6-84) can be taken as d
p,e
from (6-74). The friction factor f
IW,N
(denoted fJ in
table 6-20) is calculated from
N , M
EFF
F
F
2
N , F
0 , M
N , IW
b
b
h
R
v
v
log 4 f
1
1
]
1

,
_

(6-85)
where b
M,N
is calculated from b
M,N
=A
M
/h
F
and if b
M,N
is greater than the true surface width
of the main channel, b
M
, then b
M,N
=b
M
. A
M
is the cross-sectional area of the main channel.
________________________________________________________________________
Table 6-20. Nuding's method. Calculation of the friction factor of the imaginary wall
between the main channel and the floodplain.
________________________________________________________________________
________________________________________________________________________
124
In the fourth phase of Nuding's method the iterative procedure shown in section 6.7.3
(example 6-3, Eqs. (6-68) and (6-69)) is used to find the composite friction factor for the
main channel. Note that in the iteration only the friction factor of the main channel is
changed using Eq. (6-69), and the friction factor of the interaction zone is not iterated.
The iteration procedure is shown in Table 6-21 and totally six iteration are needed to reach
convergence.
________________________________________________________________________
Table 6-21. Nuding's method. Calculation of main channel friction factor using iterative
method.
________________________________________________________________________
________________________________________________________________________
The total friction factor of the main channel including the interaction wall friction factor is
calculated from (6-70) and velocity in the main channel from (6-71). The results are
shown in Table 6-22.
________________________________________________________________________
Table 6-22. Nuding's method. Calculation of velocity and discharge in the main channel
when the friction factor of the imaginary wall is taken into account.
________________________________________________________________________
________________________________________________________________________
125
REFERENCES
Abbott, M.B. (1979). Computational hydraulics; Elements of the Theory of Free Surface Flows, Pitman
Publishing Limited, London.
Amein, M.M. and Fang, C.S. (1970). Implicit flood routing in natural channles, JHYD, ASCE, 96, No.
HY12, December.
Chow, V.T. (1959). Open channel hydraulics, McGraw-Hill, New York.
Cunge, J.A. (1975). Applied mathematical modelling of open channel flow, Chapter 10 of Unsteady
Flow In Open Channels, Water Resources Publications, Fort Collins, Colorado.
Cunge,J.A., Holly, F.M., Verwey, A. (1980). Practical aspects of computational river hydraulics.
Pitman Advanced Publishing Program, London.
Graf W.G. and Altinakar, M.S. (1998). Fluvial hydraulics. Flow and transport processes in channels of
simple geometry. John Wiley & Sons. 680 p.
Hamill, L. (2001). Understanding hydraulics. Palgrave, Great Britain. 608 p.
Holly, F.M. and Preissmann, A. (1977). Accurate calculation of transport in two dimensions, JHYD,
ASCE, 103, No. HY11, November.
Liggett, J.A. (1975). Basic equations of unsteady flow, Chapter 2of Unsteady Flow In Open Channels,
Water Resources Publications, Fort Collins, Colorado.
Liu , Z. (2001). Sediment transport, Aalborg University, excellent lecture notes.
Stephan, U. (2001). Zum Fliewiderstandsverhalten flexiber vegetation. Dissertation, Institut fr
Hydraulik, Gewsserkunde und Wasserwirtschaft, Technische Universitt Wien.
Other good books:
Nezu, I. and Nakagawa, K. (1993). Turbulence in open channels. A.A. Balkema, Rotterdam.