Numerical Modelling and Hydraulics

Numerical
Modelling and
Hydraulics
Nils Reidar B. Olsen
Department of Hydraulic and

Environmental Engineering
The Norwegian University of
Science and Technology
3rd edition, 9. March 2012.
ISBN 82-7598-074-7
Numerical Modelling and Hydraulics
Foreword
The class Numerical Modelling and Hydraulics is a new name for the
old course Hydroinformatics, which was offered for the first time in the
spring 2001 at the Norwegian University of Science and Technology. It is
an undergraduate course for the 3rd/4th year students. The prerequisite
was a basic course in hydraulics/hydromechanics/fluid mechanics, that
includes the derivation of the basic equations, for example the continuity
equation and the momentum equation.
When I started my employment at the Norwegian University of Science
and Technology, I was asked to teach the course and make a plan for its
content. The basis was the discontinued course River Hydraulics,
which also included topics on limnology. I was asked to include topics on
water quality and also on numerical modelling. When adding topics to a
course, it is also necessary to remove something. I have removed some
of the basic hydraulics on the momentum equation, as this is taught in
other courses the students had previously. I have also removed parts of
the special topics of river hydraulics such as compound sections and
bridge and culvert analysis. The compound sections hydraulics I believe
can not be used in practical engineering anyway, as the geometry is too
simplified compared with a natural river. The bridge analysis is based on
simplifications of 1D flow models for a 3D situation. In the future, I
believe a fully 3D model will be used instead, and this topic will be obsolete. Some of the topics on marine engineering have been removed, as a
new course Marine Physical Environment at Department of Structural
Engineering at NTNU is covering these subjects. This course also contains some ice hydraulics and related cold climate engineering, topics
which has not been included in the present text.
The resulting course included classical hydraulics, sediment transport,
numerics and water quality. It was difficult to find one textbook covering
all topics. The books were also very expensive, so it was difficult to ask
the students to buy several books. Instead I wrote the present notes. I
want to thank the Department for giving me time for this, and hope the
book will be of interest for the students.
I also want to thank all the people helping me with material, advice and
corrections to the book. Dr. Knut Alfredsen has provided advice and
material on the numerical solution of the Saint-Venants equation and on
the habitat modelling. Prof. Torkild Carstens has given advice on jets,
plumes and water abstraction. Prof. Liv Fiksdal provided advice about
water biology and Mr. Yngve Robertsen has given advice on the flood
wave formulas. I also want to thank my students taking the course in the
spring 2001, finding to a large number of errors and making suggestions
for improvements. For an earlier version, Prof. Hubert Chanson provided
useful corrections.
The new name reflects the focus of numerical models and hydraulics.
The word Hydroinformatics is very broad and covers a large number of
topics not included in the present book. In addition to numerical models,
also some topics of Hydraulics are covered, for example flood waves,
sediment transport, stratified flow and physical model tests.
In memory of Prof. Dagfinn K. Lysne
Table of content
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Classification of computer programs . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2. River hydraulics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Uniform flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Friction formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Singular losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Critical flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
2.5 Steady non-uniform flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13
2.6 Waves in rivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
2.7 The Saint-Venant equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19
2.8 Measurements of water discharge in a natural river . . . . . . . . . . . . . . .22
2.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24
3. Numerical modelling of river flow in 1D . . . . . . . . . . . . . . . . . . . . . . . . . .26
3.1 Steady flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26
3.2 Unsteady flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
3.3 Unsteady flow - kinematic wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28
3.4 Unsteady flow - Saint-Venands equations . . . . . . . . . . . . . . . . . . . . . .31
3.5 Hydrologic routing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38
3.6 HEC-RAS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39
3.7 Commercial software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39
3.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40
4. Dispersion of pollutants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42
4.2 Simple formulas for the diffusion coefficient . . . . . . . . . . . . . . . . . . . . .42
4.3 One-dimensional dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44
4.4 Jets and plumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45
4.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .47
5. Dispersion modelling in 2D and 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . .48
5.1 Grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .48
5.2 Discretization methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52
5.3 The First-Order Upstream Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . .53
5.4 Spreadsheet programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .55
5.5 False diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .57
5.6 The Second Order Upstream Scheme . . . . . . . . . . . . . . . . . . . . . . . . .58
5.7 Time-dependent computations and source terms . . . . . . . . . . . . . . . . .60
5.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .62
6. Numerical modelling of water velocity in 2D and 3D . . . . . . . . . . . . . . . .64
6.1 The Navier-Stokes equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .64
6.2 The SIMPLE method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65
6.3 Advanced turbulence models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .68
6.4 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .70
6.5 Stability and convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .72
6.6 Free surface algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .76
6.7 Errors and uncertainty in CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .79

6.8 SSIIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .81
6.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .81
7. Physical limnology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83
7.2 Circulation in non-stratified lakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83
7.3 Temperature and stratification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .84
7.4 Wind-induced circulation in stratified lakes . . . . . . . . . . . . . . . . . . . . . .87
7.5 Seiches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .89
7.6 River-induced circulation and Coriolis acceleration . . . . . . . . . . . . . . .90
7.7 Density currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .91
7.8 Intakes in stratified reservoirs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .92
7.9 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93
8. Water biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .94
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .94
8.2 Biochemical reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .94
8.3 Toxic compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .95
8.4 Limnological classifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .97
8.5 The nutrient cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .97
8.6 QUAL2E . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .100
8.7 Phytoplankton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101
8.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .104
9. Sediment transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .106
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .106
9.2 Erosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .108
9.3 Suspended sediments and bed load . . . . . . . . . . . . . . . . . . . . . . . . . . .110
9.4 1D sediment transport formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .112
9.5 Bed forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .114
9.6 CFD modelling of sediment transport. . . . . . . . . . . . . . . . . . . . . . . . . . .116
9.7 Reservoirs and sediments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .117
9.8 Fluvial geomorphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .119
9.9 Physical model studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .123
9.10 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .126
10. River habitat modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .128
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .128
10.2 Fish habitat analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .128
10.3 Zero and one-dimensional hydraulic models . . . . . . . . . . . . . . . . . . .130
10.4 Multidimensional hydraulic models . . . . . . . . . . . . . . . . . . . . . . . . . . .131
10.5 Bioenergetic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .131
10.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .131
Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .134
Appendix I. Source code for explicit solution of Saint-Venants equations . .139
Appendix II. List of symbols and units . . . . . . . . . . . . . . . . . . . . . . . . . . . . .142
Appendix III. Solutions to selected problems . . . . . . . . . . . . . . . . . . . . . . . .144
Appendix IV: An introduction to programming in C . . . . . . . . . . . . . . . . . . .153
1. Introduction
1.1 Motivation
In todays society, environmental issues are an important concern in
planning projects related to water resources. Discharges of pollutants
into rivers and lakes are not allowed, unless special permission is given
by the appropriate authority. In an application for discharge into a receiving water body, an assessment of potential damages must be included.
A numerical model is useful in the computation of the effects of the pollution.
Over the last years, flooding of rivers and dam safety have been major
issues in Norway. The new regulations for planning, construction and
operation of dams has increased demands for dam safety. All Norwegian dams will have to be evaluated with regards to failure, and the
downstream effects have to be assessed. In this connection, flood zone
mapping of most major rivers have to be undertaken, and this will create
considerable work for hydraulic engineers in years to come.
The last twenty years have also seen the evolution of computers into a
very applicable tool for solving hydraulic engineering problems. Many of
the present-day numerical algorithms were invented in the early 1970s.
At that time, the computers were still too slow to be used for most practical flow problems. But in the last few years the emergence of fast and
inexpensive personal computers have changed this. All the numerical
methods taught in this course are applied in programs running on a PC.
The most modern numerical models often have sophisticated user interfaces, showing impressive colour graphics. People can easily be led to
an understanding that the computer solves all the problems with minimum knowledge of the user. Although present day computer programs
can compute almost all problems, the accuracy of the result is still uncertain. An inexperienced user may produce convincing and impressive
colour figures, but the accuracy of the result may still not be good
enough to have a value in practical engineering. It is therefore important
that the user of the computer programs has sufficient knowledge of both
the numerical methods and their limitation and also the physical processes being modelled. The present book therefore gives several chapters on processes as basic hydraulics, limnology, sediment transport,
water quality etc. The knowledge should be used to provide reasonable
input for the numerical models, and assessing their results. Many empirical formulas are given, providing further possibilities for checking the
result of the numerical method for simpler cases.
The numerical methods also have limitations with regards to other
issues, for example modelling of steep gradients, discontinuities, processes at different scales etc. The numerical models itself may be prone
to special problems, for example instabilities. Often, a computer program
may not include all processes occurring in the water body. The user
needs to be aware of the details of the numerical methods, its capabilities and limitations to assess the accuracy of the results.
1.2 Classification of computer programs

There exist a large number of computer programs for modelling fluvial
hydraulics and limnology problems. The programs have varying degree
of sophistication and reliability. The science of numerical modelling is
progressing rapidly, making some programs obsolete while new programs are emerging.
The computer programs can be classified according to:
- what is computed
- how many dimensions are used
- particulars of the numerical methods
Many computer programs are tailor-made for one specific application.
Examples are:
- Water surface profiles (HEC2)
- Flood waves (DAMBRK)
- Water quality in rivers (QUAL2E)
- Sediment transport and bed changes (HEC6)
- Habitat modelling (PHABSIM)
This is particularly the case for one-dimensional models, developed several years ago, when the computational power was much less than
today. There are also more modern one-dimensional programs, using
more sophisticated user interfaces and also including modules for computing several different problems. Examples are:
- HEC-RAS
- MIKE11
- ISIS
In recent years, a number of multi-dimensional computer programs have
been developed. These also often include modules for computing several different processes, for example water quality, sediment transport
and water surface profiles.
Multi-dimensional programs may be:
- two-dimensional depth-averaged
- three-dimensional with a hydrostatic pressure assumption
- fully three-dimensional
There also exist width-averaged two-dimensional models, but these are
mostly used for research purposes.
The three-dimensional models solve the Navier-Stokes equations in two
or three dimensions. Sometimes the equations are only solved in the
horizontal directions, and the continuity equation is used to obtain the
vertical velocity. This is called a solution with a hydrostatic pressure
assumption. The fully 3D models solve the Navier-Stokes equations also
in the vertical direction. This gives better accuracy when the vertical
acceleration is significant.
The various algorithms used by these types of programs are described
in the following chapters, together with the physics involved.
2. River hydraulics
The classical river hydraulics described in this chapter forms the basis
for the numerical modelling of flood waves and river pollutant dispersion.
In this chapter, a hydrostatic pressure is assumed in the vertical direction, and also that the water flow is one-dimensional.
2.1 Uniform flow

The definition of uniform flow can be visualized by looking at water flow
in a very long flume, where the water depth and velocities are constant
at any point over the length of the flume. In a natural river this never
occurs, but the concept is useful for developing hydraulic engineering
formulas. Fig. 2.1.1 shows a section of a wide channel, with forces on
the water. I is the slope of the water surface and h is the water depth.
The forces on the water volume in the direction parallel to the river bed/
surface will be:
g
U
Bed shear: F b = x
Gravity: F g = g x V = gIxh
Fig. 2.1.1 Forces on a

water volume in uniform flow
The direction of the flow is called x, h is the water depth, I is the slope of
the water surface, gx is the component of the gravity in the x-direction
and is the shear stress on the bed. Setting the two forces equal to each
other gives the formula for the bed shear stress:
= ghI
(2.1.1)
The density of water is denoted , and g is the acceleration of gravity. A

method to compute I is presented in the next chapter.
gx
gz
The vertical velocity profile in a river with uniform flow can be described
by boundary layer theory. Early experiments were carried out by
Nikuradse (1933) using uniform spheres, and later Schlichting (1936)
using particles of varying shapes. The experiments produced the following formula for the vertical velocity profile for uniform flow (Schlichting,
1979):
30y-
U- = --1- ln ----------
ks
u*
gx = gsin= gtan =
gI for small angles
(2.1.2)
U is the velocity, and it is a function of the distance, y, from the bed. The
parameter is an empirical constant, equal to 0.4. The formula only
applies for rough surfaces, and ks is a roughness coefficient. It is equivalent to the particle diameter of the spheres glued to the wall to model
roughness elements. The variable u* is the shear velocity, given by:
u* =
--
(2.1.3)
Eq. 2.1.2 is also called the logarithmic profile for the water velocity.
Schlictings formulas were based on data from experiments done in air,
but since non-dimensional parameters were used, the results worked
very well also for water flow. Schlichting found the wall laws applies for
all boundary layers, also for non-uniform flow, as long as only the velocities very close to the wall are considered.
To use the formula, the next question is which roughness to choose.
There exist a number of different relations between the effective roughness and the grain size distribution on the river bed. Van Rijn (1982)
found the following formula, based on 120 flume data sets:
k s = 3d 90
(2.1.4)
The variable d90 denotes the grain size sieve where 90 % of the material
is finer. Van Rijn reported that there were large uncertainties in this formula, and that the number 3 was an average value where the data set
suggested a variation range between 1 and 10. Other researchers have
used different formulas. Hey (1979) suggested the following formula
based on data from a natural river with coarse material, and laboratory
experiments with cubical/spherical elements:
k s = 3.5d 84
(2.1.6)
Kamphuis (1974) made a new formula based on his flume experiments.

The obtained:
k s = 2d 90
(2.1.7)
The value 2 varied between 1.5 and 2.5 in the experiments. Kamphuis
used a zero reference level of 0.7d90, which will affect the results.
Schlichting (1979) carried out laboratory experiments with spheres and
cones. Using 45 degree cones placed right beside each other, the ks
value was equal to the cone height. In other words, it is difficult to obtain
an accurate estimate of the ks value.
2.2 Friction formulas

A number of researchers have developed formula for the average velocity in a channel with uniform flow, given the water depth, water slope and
a friction factor. The formulas are empirical, and the friction factors often
depend on the water depth. The most common formulas are:
Mannings formula:
2 1
--- ---
Mannings formula:
1 3 2
U = --- r h I
n
(2.2.1)
The hydraulic radius, rh, is given by:
rh = A
--P
(2.2.2)
where A is the cross-sectional area of the river and P is the wetted

perimeter.
Often the Stricklers M value is used instead of Mannings coefficient, n.
The relation is:
Some Mannings coefficients:

Glass models:
0.009-0.01
Cement models:
0.011-0.013
Concrete lined channels:
0.012-0.017
Earth lined channels:
0.018-0.04
Rock lined channels:
0.025-0.045
Earth lined rivers:
0.02-0.05
Mountain rivers with stones:
0.04-0.07
Rivers with weed:
0.05-0.15
M = 1--n
(2.2.3)
giving:
U = M
2 1
--- --3 2
rh I
(2.2.4)
Given the grain size distribution on the bed, the Mannings friction factor
can be estimated by the following empirical formula (Meyer-Peter and
Mller, 1948):
26 M = -------------( d 90 )
1
--6
(2.2.5)
Given the water velocity and the friction factor, the formulas can be used
to predict the water depth. Together with the continuity equation (2.2.6)
the formulas can also be used to estimate the water surface slope or the
friction loss for non-uniform flow. Thereby the water elevations can be
found. A further description is given in Chapter 3.
Continuity equation:
q = Uh
(2.2.6)
The water discharge pr. unit width of the river is often denoted q.
2.3 Singular losses

Using the Energy Equation/Bernoullis Equation for the water flow in a
river, it is possible to compute the energy loss and the water surface
location. The friction loss is given by the roughness of the river bed.
There are also other energy losses, called singular losses. These are
identified with particular constructions in the rivers, for example a bridge
pier, or a river bend. The head losses are associated with eddies generated around the loss point, usually in connection with flow expansion. A
recirculation zone forms, dissipating energy. The head loss, hf, can be
computed as:
2
U
h f = k -----2g
(2.3.1)
where k is a head loss coefficient related to the geometry of the river/

obstruction.
In natural rivers, it is often difficult to identify singular losses and assign a
value to each loss. Instead, a different Mannings friction factor is often
used, where the effective friction factor is used, as a combination of the
singular losses and the friction loss. The friction factor is then found by
calibration.
It is possible to use Eq. 2.3.1 for river contractions, for example in connection with bridges and bridge piers. However, the head loss coefficient
is difficult to find without using measurements in the field/lab.
10
2.4 Critical flow

Looking at one-dimensional water flow in a channel, there are two types
of energy: Kinematic energy due to the water velocity, and pressure due
to the water depth, y, and weight of the water. The sum of these energies is called the specific energy height of the section, E [meters]:
2
E = EP + EU = y + U
-----2g
(2.4.1)
The specific energy may change along the length of the river, depending
on the water discharge, roughness, bed slope etc. If the water is given a
specific energy, Eq. 2.4.1 can be used to find the water depth:
2
-----y = EU
2g
(2.4.2)
Introducing the continuity equation (2.2.6), the equation can be written:

2
q
y = E ----------2
2gy
(2.4.3)
or
2
3
2
q
y Ey + ----- = 0
2g
(2.4.4)
This third-order equation has three possible solutions. Only two are
physically possible. If we solve for the specific energy, we get:
2
q
E = y + ----------2
2gy
(2.4.5)
The minimum specific energy to transport a given volume of water is

obtained by derivation of Eq. 2.4.5, and setting the result to zero:
2
d----E- = 1 ------q- = 0
3
dy
gy
(2.4.6)
yc =
The Froude number:
UFr = --------gy
q---g
(2.4.7)
Using the continuity equation, the equation above can also be written:
Uc
----------- = 1
gy c
(2.4.8)
The term on the left side of the equation is also called the Froude
number. For a minimum amount of specific energy, the Froude number
is unity, as given in Eq. 2.4.8. If the Froude number is below unity, the
flow is subcritical. The flow is supercritical if the value is higher than unity
Supercritical flow is very seldom encountered in natural rivers. It exists in
water falls or rapids. If supercritical flow occur in a river with stones on
11
the bed, usually a hydraulic jump is formed. Normally, the flow in a river
is subcritical.
The Froude number is important for the numerical solution of equations
for water depth in a river. If critical flow occur, instabilities often emerge
in the numerical algorithms.
Irregular cross-sections
The derivation above is valid for channels with rectangular crosssections. For a channel with irregular cross-section, the water velocity is
replaced by Q/A. The specific energy for the section becomes:
2
Q
E = y + -----------2
2gA
(2.4.9)
The minimum specific energy for the section is obtained by derivation

with respect to y and setting this to zero, similar to what was done for the
rectangular channel:
2
Q dA
dE
------- = 1 --------3- ------- = 0
dy
gA dy
(2.4.10)
For small changes in the water level, the incremental cross-sectional

area, dA is given by
dA = Bdy
(2.4.11)
The top width of the cross-section is denoted B. Inserted into Eq. 2.4.10,
this gives:
2
Q
B = 1
---------3
gA
(2.4.12)
The square root of the left side of the equation is the Froude number for
a cross-section with a general complex geometry:
2
Fr =
Q
B = ---------U---------3
A
gA
g --B
(2.4.13)
The hydraulic jump

The hydraulic jump is a transformation of the water flow from supercritical to subcritical flow. Visually, it looks like a standing wave in the channel. The water level is higher downstream than upstream.
It is possible to derive a relationship between the water level upstream,
h1, and downstream, h2, of the jump.
The momentum equation gives the force on an obstacle as:
F = Q ( U 2 U 1 ) + g ( A 2 y 2 A 1 y 1 )
(2.4.14)
The parameters y is the effective height of the hydrostatic pressure.

For a rectangular channel where F=0 and y= 1/2 y:
12
1
0 = Q ( U2 U 1 ) + --- g ( A 2 y 2 A 1 y 1 )
2
(2.4.15)
Divide by the width and the density:
1
2
2
2 1
1- + 1--- g ( y 2 y 2 )
0 = q ( U 2 U 1 ) + --- g ( y2 y 1 ) = q ---- ---2
1
2
y 2 y 1 2
(2.4.16)
Solve with respect to q2:
1
--- g ( y 22 y 21 )
2
q = ---------------------------- =
1 1-
--- y - ---
2 y1
1
--- g ( y 22 y 21 )
2
--------------------------1-
1- --- ---y y
1
2
(2.4.17)
Use the definition of the Froude number:

2
Fr 2
2
U
q= -------2- = ------3
gy 2
gy 2
(2.4.18)
Solves the equation with respect to q2, and eliminates q2 with the
equation above:
3
Fr 2 gy 2
1--2
2
g ( y2 y1 )
2
= ---------------------------- =
1-
1- --- ---y y
1
2
1--g ( y 2 y 1 ) ( y 2 + y 1 )y 1 y 2
1
2
------------------------------------------------------------ = --- g ( y2 + y 1 )y 1 y 2
2
( y2 y1 )
(2.4.19)
y1 y2
y
y y
1
2
- = 1--- ( y 2 + y 1 ) ----1- = 1--- ----1- ----1- + 1
Fr 2 = --- ( y 2 + y 1 ) --------3
2
2
2
2 y2 y2
y
y
2
(2.4.20)
Solves with respect to y1/y2:
2
y----1- + y----1- 2Fr 2 = 0
2
y 2
y2
(2.4.21)
Solves the second order equation:
y1
1
2
----- = --- ( 1 + 8Fr 2 1 )
2
y2
(2.4.22)
Given the water level and the Froude number downstream of the jump,
the water level upstream of the jump can be computed. It can be shown
13
that a formula where the indexes 1 and 2 are changed is also valid.
2.5 Steady non-uniform flow

It is possible to compute the water surface of a steady non-uniform flow
by analytical formulas, as long as the cross-sectional shape is rectangular. The formulas are derived in the following. However, if the cross-section does not have rectangular shape, the water surface location have to
be computed numerically. This is described in Chapter 3.
U12/2g
y1
Energy line
If
U22/2g
Water
surface
I0
y2
Bed
z1
z2
Datum
Figure 2.5.1 A longitudinal profile of a

channel is shown, between the two
cross-sections 1 and 2. The distance
between the sections is dx. The water
depths are denoted y and the bed level
elevation is denoted z. The bed slope
is denoted I0, and the slope of the
energy line is denoted If. Note the
energy line is located a distance above
the water level, equal to the velocity
height, U2/2g.
dx
The derivation of the formula for the water depth is based on Fig. 2.5.1.
The total height of the energy line is denoted H, so that:
2
H = z+y+U
-----2g
(2.5.1)
The energy slope, If, can be computed from for example Mannings formula. For our case, the slope can be written as:
2
2
H2 H1
d-
d-
U
U
--------z + y + ------ = I 0 +
I f = ------------------- =
y + ------
dx
dx
dx
2g
2g
(2.5.2)
The term in the brackets most right in the equation is the specific energy,
E, of the flow. The term can be rewritten:
2
2
d-
dE dy
dy dy 1 dU
----y+U
------ = dE
------- = ------- ------ = ------ ------ + ------ ----------
dx
dy dx
dx dy 2g dy
2g
dx
(2.5.3)
The continuity equation gives:

2
2
U = Q
-----2
A
Derivation with respect to y gives:
(2.5.4)
14
2
2
d Q
2 d
2 dA
2 3 dA
dU
---------- = ------ -----2- = Q ------- ( A ) ------- = 2Q A ------dy
dA
dy
dy
dy
A
(2.5.5)
The final formula needed is for a general cross-sectional shape:
dA
------- = B
dy
(2.5.6)
This is inserted into Eq. 2.5.5, and the result of this in Eq. 2.5.3, giving:
2
dy
1
d-
dy
2 3
2
U
----y + ------ = ------ 1 + ------ ( 2Q A B ) = ------ ( 1 Fr ) (2.5.7)
dx
2g
dx
dx
2g
Inserted into Eq. 2.5.2, the result is:
dy
2
I f = I 0 + ------ ( 1 Fr )
dx
(2.5.8)
which can be rearranged to:
If I0
dy
------ = ----------------2
dx
1 Fr
(2.5.9)
In a wide, rectangular channel, the Froude number can be written:
U - = ---------------Q Fr = --------gy
By gy
(2.5.10)
Similarly, the friction slope, If, can be expressed as a function of constants and y, using Mannings Equation:
2
4
--2 3
10
-----2 2 3
U - = --------------------Q
I f = -----------M y
(2.5.11)
M B y
The water depth is here denoted y. It is used instead of the hydraulic

radius, meaning the formula is only valid for wide, rectangular channels.
Eq. 2.5.10 and Eq. 2.5.11 can be inserted into Eq. 2.5.9, resulting in a
formula for the change in the water depth only being a function of constants and y. This can then be integrated analytically to compute functions for changes in water depths in wide channels with constant width.
For more complex cases it is possible to use a spreadsheet to compute
the water surface. Such cases can be channels with varying widths or
non-rectangular cross-sections. An example of such a spreadsheet is
given in the figure below. Note, for sub-critical flow we start with the
downstream cross-section, which is no. 1 in the table.
15
Crosssection no.
Depth, y
(given)
y1+y1
U
Continuity
Energy
slope, If
Froude
number, Fr
Definition
Eq. 2.5.11
Eq. 2.5.9
0.000429
-0.04213
Eq. 2.5.10
3
4
1.0
0.957
The method of invoking

more iterations is dependent on the particular
spreadsheet program. For
Lotus 123, use F9 on the
keyboard repeatedly. For
MS Excel, use the menu
Tools, Options, Calculations, and cross off Iterations, and give a number
in the edit-field, for example 50.
1.0
0.322
Also note that the water depths are calculated in the cross-sections. The
other parameters in the table are computed as an average value
between cross-sections. This means the water velocity in the table is a
function of two water depths. Iterations are therefore necessary.
The numbers in the spreadsheet is an example with a water discharge of
1 m3/s in a 1 m wide channel, a Manning-Strickler value of 50 a slope of
0.001 and a dx of 50.
Classification of surface profiles
Bakhmeteff (1932) proposed a classification system for water surface
profiles, which is included in almost all textbooks on water surface profiles. The system is useful for understanding water surface profiles, but
the classifications itself is rarely used in engineering practice. The profiles are classified according to the bed slope, the critical slope and the
water depth, as given in Table 2.5.1. The water depth is denoted y, the
slope is denoted I, and E is the specific energy of the water. Subscripts 0
denotes the bed and c denotes critical slope/depth. The figures at the
right of the table show longitudinal profiles of the surface profiles. The
lines for the critical depth and normal depth are also given. The normal
depths are found by using for example Mannings equation, given the
roughness, bed slope and water discharge. The critical depth is found
from Eq. 2.4.7.
:
The system classifies each
surface profile with a letter
Letter
Stands for
Bed slope
and a number. The letter is
only a function of the slope
M
Mild
Subcritical
of the river at the given discharge. The letters given in
S
Steep
Supercritical
the table on the right are
used.
C
Critical
Critical
The number is an index for
the actual Froude number in
the channel. Subcritical flow
is denoted 1 while supercritical flow is denoted 3. The
index 2 can mean either
Horizontal
Horizontal
Adverse
Adverse
16
supercritical or subcritical, depending on the letter. This is given in more

detail in Table 2.5.1.
Table 2.5.1. Classification of water surface profiles
y0
yc
yc
y0
yc
yc
Channel
slope
Profile
type
Depth range
Fr
dy/
dx
dE/
dx
Mild
M1
y>y0>yc
<1
>0
>0
I0<Ic
M2
y0>y>yc
<1
<0
<0
y0>yc
M3
y0>yc>y
>1
>0
<0
Steep
S1
y>yc>y0
<1
>0
>0
I0>Ic
S2
yc>y>y0
>1
<0
>0
y0<yc
S3
yc>y0>y
>1
>0
<0
Critical
I0=Ic
y0=yc
C1
y>yc
<1
>0
>0
C3
y<yc
>1
>0
<0
Horizontal
H2
y>yc
<1
<0
<0
I0=0
H3
y<yc
>1
>0
<0
Adverse
A2
y>yc
<1
<0
<0
I0<0
A3
y<yc
>1
>0
<0
M1
M2
M3
S1
S2
S3
C1
C3
H2
H3
A2
yc
A3
Examples of water surface profiles: When a mild-sloping river flows

into a reservoir, we have an M1 curve. The water profile just upstream a
spillway can be M2, H2 or A2. Profiles just before a hydraulic jump can
be M3, H3 or A3. The S1 profile can be found after a hydraulic jump in a
steep channel.
2.6 Waves in rivers

It is possible to derive formulas for waves travelling up or down a channel with constant width and slope. Rough estimation of such waves can
thereby be made, and the formulas can also be used to evaluate the
results from numerical models. Two types of waves can be described:
- kinematic waves
- dynamic waves
Both types are derived and discussed in the following.
Dynamic wave
An equation for the dynamic wave is derived by looking at the longitudinal profile of the wave given in Fig. 2.6.1:
17
Fig 2.6.1 Definition sketch

of a wave travelling down a
river. The upstream depth
and velocity are denoted h1
and U1, respectively and the
downstream depth and
velocity are denoted h2 and
U2 respectively. The velocity
of the wave is denoted c.
c
y1
U1
U2
y2
A steady situation is obtained if the reference system moves along the

channel with velocity c. The water velocities upstream and downstream
of the wave becomes U1 - c and U2 - c, respectively. The momentum
equation gives:
2
1 2
2
1 2
( U 1 c ) y 1 + --- gy 1 = ( U 2 c ) y 2 + --- gy 2
2
2
(2.6.1)
Together with the continuity equation:
( U 1 c )y 1 = ( U 2 c )y 2
(2.6.2)
Solving the continuity equation with respect to U2:
y
U 2 = ( U 1 c ) ----1- + c
y2
(2.6.3)
Inserting this into the momentum equation, eliminating U2:

2
y1
1
2
2
2
( U 1 c ) y 1 + --- g ( y 1 y 2 ) = ( U 1 c ) ----- + c c y 2 (2.6.4)
2
y2
Eliminating the cs on the right hand side, and moving the term to the left
side:
y1 2
1
2
2
2
( U 1 c ) y 1 ----- y 2 + --- ( g ( y 1 y 2 ) ) = 0
y 2
2
(2.6.5)
Simplifying the first term and changing the second term:

2
y 1
( U 1 c ) y 1 ----1- + --- ( g ( y 1 y 2 ) ( y 1 + y 2 ) ) = 0
y 2 2
(2.6.6)
Further simplifying the first term and changing the sign of the second
term:
1
2 y1
( U 1 c ) ----- ( y 2 y 1 ) --- ( g ( y 2 y 1 ) ( y 1 + y 2 ) ) = 0
y2
2
Dividing by the second part of the first term:
(2.6.7)
18
1 y2
2
( U 1 c ) = --- ----- g ( y 1 + y 2 )
2 y1
(2.6.8)
Taking the square root of each side, multiplying with -1 and moving U1 to
the other side:
g y2
c = U 1 --- ----- ( y 2 + y 1 )
2 y1
(2.6.9)
If the wave is small, y2 and y1 are approximately equal. The equation

then becomes:
If Eq. 2.6.10 is combined
with the definition of the
Froude number, the following formula is
obtained:
U
c = U --------Fr
If Fr < 1, c can be both
positive and negative. In
subcritical flow, a wave
can then travel both
upstream and downstream a channel.
If Fr > 1 then c will always
be positive. For supercritical flow, the wave can
only travel downstream.
c = U gy
(2.6.10)
Kinematic wave
The formula for a kinematic wave with speed c, is derived from the continuity equation, looking at a similar situation as in Fig. 2.6.1. This gives:
c = dQ
------dA
(2.6.11)
The differential of Q can be derived from Mannings Equation (Eq. 2.2.1)

for a wide, rectangular channel, differentiated with respect to the water
depth:
1 2
1 5
1---
2
-- 1 --2- --3-
1 --2- --3-
d
d
5 3
d
BI
dQ
------- = ------ ( AU ) = ------ By --- I y = ------ B --- I y = -------- --- y
dy
dy n
n 3
dy n
dy
(2.6.12)
or rewritten:
1 2
--- --5 BI 2 3
dQ = --- ------ y dy
3n
(2.6.13)
The formula for the area of the cross-section is then differentiated with
respect to y:
dA = Bdy
(2.6.14)
Inserting dA and dQ from the two equations above into Eq. 2.6.11 gives:
1--- 2---
51 2 3
5
c = --- --- I y = --- U
3n
3
(2.6.15)
which is the formula for the kinematic wave.

Wave shape
Eq. 2.6.10 shows that the wave speed will increase with larger depth.
19
This is also given from Mannings Equation (2.2.1), as the increased

water velocity will lead to increased depth. The shape of a wave will
thereby change as it moves downstream the channel. Assuming the
front and the end of the wave have smaller depths, and the maximum
depth is at the center of the wave, two phenomena will take place:
1. The wave front will become steeper
2. The wave tail will become less steep
This can be seen in Fig. 2.6.2:
Figure 2.6.2. Change of

wave shape over time.
The initial wave is shown
on the left. The water is
flowing from the left to the
right. The right figure
shows the shape of the
wave after some time. The
front is steeper and the tail
is less steep.
2.7 The Saint-Venant equations

A general flood wave for a one-dimensional situation is described by the
Saint-Venant equations. Then the water velocity and water depth can
vary both in time and along the one spatial dimension. The Saint-Venant
equations is also called the full dynamic equation for computation of 1D
flood waves.
The equation is derived by looking at a section of a channel as given in
Fig. 2.7.1. The channel has a slope I, water depth y, width B, water discharge Q and velocity U.
y1
P1
y2
P2

channel, between two sections 1 and
2. The distance between the sections
is x. The water depth is denoted y.
The bed slope is denoted I0.
x
x
A continuity equation for this situation is first derived. Looking at the
water discharge going out and in of the volume in the time period t, we
obtain:
dQ
V = Q Q + ------- x t = dQ
------- xt
dx
dx
(2.7.1)
This amount of water must be equal to the volume change caused by ris-
20
dy
dt
ing/falling of the water level: ------ :
dy
V = ------ Bxt
dt
(2.7.2)
Noting that B also may change as the water level rise/fall, one can
instead write the equation as:
dA
V = ------- xt
dt
(2.7.3)
Where A is equal to the cross-sectional area of the flow. Note, in a rectangular channel, B is constant, and Eq. 2.7.2 is obtained.
Combining Eq. 2.7.1 and 2.7.3, the following equation is obtained:
Continuity equation:
dA
------- + dQ
------- = 0
dt dx
(2.7.4)
This is the continuity equation often used in connection with solving the
Saint-Venant equation. The Saint-Venants equation itself is derived from
Newtons second law:
= ma
(2.7.5)
For the section in Fig. 2.7.1, The acceleration term on the right hand
becomes:
dU
ma = yBx ------dt
(2.7.6)
Then the external forces on the volume are given. There are four forces:
1. Gravity component:
F g = gyBxI 0
(2.7.7)
This is the same component as used for the derivation of the formula for
the bed shear stress for uniform flow.
2. Bed shear stress:
F b = Bx
(2.7.8)
The term is negative, as the force is in the negative x-direction. Often, a

friction slope is introduced:
I f = -------gy
(2.7.9)
The friction slope is computed from an empirical friction formula, for

example the Mannings equation. The term then becomes:
21
F b = gyI f Bx
(2.7.10)
3. Pressure gradient:
The pressure gradient is due to the different water level on each side of
the element. The hydrostatic pressure force on a water cross-section in
a rectangular channel is given as:
1
2
F = --- gBy
2
(2.7.11)
For the control volume in Fig. 2.7.1, there are two forces, one on each
side of the volume. The total force from the pressure gradient must
therefore be the sum of these two hydrostatic forces.
1
2 1
2
F p = --- gBy 1 --- gBy 2
2
2
(2.7.12)
The depth can be written as a function of the surface slope:

2
gB 2
dy
F p = ---------- y y + ------ x
2
dx
2
gB 2 2
dy
dy
F p = ---------- y y 2 ------ yx ------ x
2
dx
dx
(2.7.13)
(2.7.14)
The last term contains a small number squared, so this is much smaller
than the second last term. The last term is therefore neglected. This
gives:
dy
F p = g ------ x ( By )
dx
(2.7.15)
The term is negative, because a positive depth-gradient will cause a

pressure force in the negative x-direction.
4. Momentum term:
The momentum equation for the control volume is:
dU
F m = Q ( U 1 U 2 ) = UBy U U + ------- x
dx
(2.7.16)
dU
F m = UBy ------- x
dx
(2.7.17)
The negative sign is because a positive velocity gradient will cause more
momentum to leave the volume than what enters. This causes a force in
the negative x-direction.
Setting the sum of all the four forces equal to the acceleration term, one
obtains:
F g + F b + Fp + F m = ma
(2.7.18)
22
or:
dy
dU
dU
gyBxI 0 gyI f Bx g ------ x ( By ) UBxy ------- = yBx ------dx
dx
dt
(2.7.19)
Saint-Venant equation:
This can be simplified to:
dy
dU
g ( I 0 I f ) g ------ = U ------- + dU
------dx
dx dt
(2.7.20)
The Saint-Venant equations must be solved by a numerical method.

This is described in Chapter 3.
2.8 Measurements of water discharge in a natural river

There are several methods to measure the water discharge in a river.
The most common method is to use a current meter and measure
directly at various points in a cross-section of a river. This can be done at
several water discharges, and a rating curve can be obtained, where the
water discharge as a function of the water elevation is given. Using daily
observations of the water levels, the curve can give a time series of the
water discharge. This is used to predict floods and average water discharge in the river for use in hydropower plants or water supply.
Figure 2.8.1 Cross-section with four measured
velocity profiles. The
average velocity in each
profile is computed and
multiplied with the depth
and the width, w, of each
sector. This is then
summed up over all the
profiles to get the water
discharge.
w3
1
Direct measurements of discharges in rivers today is usually done by an

ADCP instrument. The ADCP is an abbreviation for Acoustic Doppler
Current Profiler. The ADCP works by sending out an acoustic signal
from the instrument into the water. Small particles in the water reflect the
signal back to a receiver on the instrument. The signal will be a function
of the speed of the particle relative to the instrument. By measuring at a
large number of points in the cross-section, the discharge can be computed.
The ADCP is usually fitted on a boat which is dragged across the water
surface in the cross-section. The beam of the instrument points vertically
downwards, and measures several points in the vertical profile. An echosounding device is usually also included in the instrument, measuring
the water depth. Modern ADCPs also have a bottom tracking device,
measuring the distance the instrument has traversed. A GPS can also
be fitted with the instrument, enabling correlation between measurements and for example a digital terrain model.
Figure 2.8.2 Picture of boat with ADCP for laboratory use.
Figure 2.8.3 Example of results from the ADCP measurements
23
24
Another method is to use a tracer, for example radioactive material or a

chemical that is easily measured for small concentrations in water. A
known quantity, m, of the tracer is dumped in the river. Further downstream, where the tracer is completely mixed in the water, the concentration, c, is measured over time. The amount of tracer can be computed
as:
m =
cQ dt
(2.8.1)
Assuming steady flow, the water discharge, Q, is constant. The quantity

of tracer and the integral of the concentration measurement is known, so
the discharge can be computed as:
mQ = -------- c dt
(2.8.2)
The method is expensive, as tracer is lost for each measurement. Also,

the tracer may not be environmentally friendly. Therefore, the method is
only used very seldom, and in situations where it is difficult to do pointvelocity measurements. This can for example be during floods.
2.9 Problems
Problem 1. Uniform flow
A natural river with depth of 2 meters, has an average water velocity of 3
m/s. What is the maximum and minimum energy gradient? Is the flow
supercritical or subcritical? Is this possible to see directly in the field?
Problem 2. Compound channel
A channel with the following cross-sectional geometry is considered:
C
5m
3m
8m
15 m
10 m
The channel can be divided in three sections, A, B and C. Section B is

the main channel, and sections A and C are the overbanks. The slope of
the channel is 1:500.
Compute the water discharge in the channel, given a Manning-Strickler
coefficient of 60 for the whole channel.
Then, compute the water discharge in the channel given a ManningStrickler coefficient of 60 for the main channel and 40 for the overbanks.
25
Problem 3. Steady non-uniform flow

Compute the water surface profile, using a spreadsheet, behind a low
run-of-the river dam. The width of the river is 30 meters, the water discharge is 50 m3/s, the dam height is 10 meters and the rivers slope is
1:400.
What happens if the slope is 1:100? Compute the water surface profile
for this situation too.
Problem 4: Location of a hydraulic jump.
The figure shows a sketch of a longitudinal profile of the bed and water
surface profile. Water is let out of a bottom outlet downstream a dam, at
point A. At point B, a hydraulic jump occur. The water levels at point A
and C are given, together with the water discharge. The question is to
find the distance between A and B.
Data:
Water level (gate opening) at A: yA=0.4 meters
Water level at C: yC=3 meters
Discharge: Q=6 m2/s.
Manning-Strickler coefficient: 50.
Problem 5. Discharge measurement using tracer
2 kg of tracer is dumped in a stream. Several km downstream, the following concentration is measured:
c, (ppm)
1000
time (minutes)
0
55
Compute the water discharge in the stream.
56
26
3. Numerical modelling of river flow in 1D

The 1D numerical model is the most commonly used tool in Hydraulic
Engineering for evaluating effects of flood waves in rivers. Dispersion of
pollutants in rivers is also mostly done using 1D models.
3.1 Steady flow

The steady flow computation uses the continuity equation and an equation for the friction loss to compute the velocity and location of the water
surface. Mannings formula (Eq. 2.2.1) is most commonly used.
The computation starts with measuring the geometry of a number of
cross-sections in the river. The distance and elevation of a number of
points in the cross-sections are recorded. The distances between the
cross-sections are also measured.
Figure 3.1.1. Sketch of a
river with crosssections.
For each section, a curve is made with the wetted area as a function of
the water level. This curve is used in the following computations.
depth
Figure 3.1.2. Generation

of the curve for the
cross-sectional area as
a function of the water
depth. The curve is used
as geometry input for the
numerical models.
area
The computation of the water elevation usually starts with a given value
downstream, as this is the controlling value for subcritical flow. Then the
water elevation of the upstream cross-section is to be found. The procedure outlined in Chapter 2.5 can be used, applying Eq. 2.5.9. Alternatively, a variation of the method is given in the following.
One problem is how to
determine the friction
coefficient, M. Usually, a
calibration procedure has
to be done, where the
results are compared with
measured water levels,
and the M values adjusted
to fit the data. Some data
programs include automatic calibration procedures using this concept.
The energy loss between the cross-sections can be found by solving

Mannings equation with respect to the friction slope:
2
U I f = ------------4
--2 3
(3.1.1)
M R
The hydraulic radius, R, is found from a similar curve as given in Fig.

3.1.2. Looking at Fig. 3.1.3, the water surface elevation difference, z,
between the cross-sections can then be found using the Energy equation, giving:
27
U2 U1
z 1 z 2 = z = I f x + -------- ---------
2g
2g
U12/2g
z1
(3.1.2)
Energy line
If
2
U2 /2g
Water
surface

channel is shown, between two crosssections 1 and 2. The distance
between the sections is dx. The elevation of the water is denoted z. The
slope of the energy line is denoted If.
z2
Bed
Datum
x
The distance, x, between the cross-sections is given by the user. The
water velocities at the two cross-sections 1 (upstream) and 2 (downstream) is computed from the water continuity equation:
U = Q
---A
(3.1.3)
where the water discharge, Q, is a constant

The procedure is then to guess a value of z, giving the water level for
both cross-sections. Then A and R are taken from the curves for the two
cross-sections, and the average values used in Eq. 3.1.1. The velocities
in Eq. 3.1.2 are computed from Eq. 3.1.3. Eq. 3.1.2 then gives a new
estimate of z. After a few iterations, the values of z should be the
same as the previous iteration, and the procedure has converged. The
solution method is usually not sensitive to the initially guessed value, so
for example z=0.0 could be used.
Control sections
Eq. 3.1.6 gives the changes in the water surface between two cross-sections. The question is then which cross-section should be used to start
the computations.
For supercritical flow, the water surface is mainly determined by the
upstream flow. Remember, the kinematic wave theory showed that a
wave could not propagate upstream in supercritical flow. For subcritical
flow, the flow is usually determined by the downstream water level. This
means that the computations should start upstream and move downstream in supercritical flow. For subcritical flow, the computations should
start downstream and move upstream.
Before starting the computations, the controlling sections should be
determined. A typical example is a critical flow section, where the flow
28
goes from subcritical to supercritical. The computations can then start in

this section and move upstream in the subcritical flow. It can also move
downstream in the supercritical flow.
Another control for the water levels are reservoirs and lakes. Then the
water surface is given. Usually, the flow is subcritical upstream of the
lake. Then the computation starts in the lake and moves upstream.
One of the main problems is when the flow turns from supercritical to
subcritical. A hydraulic jump will then form, and special algorithms based
on the momentum equation have to be used. Often, these algorithms are
not very stable, and it may be problematic to get a solution.
3.2 Unsteady flow

The one-dimensional water flow is governed by the Saint-Venant equations: water continuity:
----A- + -----Q- = 0
t x
(3.2.1)
and conservation of momentum:
y
U-----U- + U -----+ g ----- g ( Ib I f ) = 0
x
x
t
(3.2.2)
The 1D models can be classified according to how many terms are used
in Eq. 3.2.2. Solving the full Saint-Venant equations are described in
Chapter 3.4. One simplification is to neglect the two first terms. This is
called the equations for the diffusive wave:
y
g ----- g ( I b I f ) = 0
x
(3.2.3)
If the first three terms in Eq. 3.2.2 is neglected, the kinematic wave equation emerges:
g ( Ib If ) = 0
(3.2.4)
Additionally, the continuity equation (3.2.1) is solved.

The solution of the kinematic wave equation is described in Chapter 3.3.
The Muskingum method, or Hydrological routing, only uses the continuity equation. This method is described in Chapter 3.5.
3.3 Unsteady flow - kinematic wave

The simplest method to compute unsteady flow in a river is by the kinematic wave equation. There are several solution methods for this equation. Two methods are described here:
- Solution by differentials
- Analytical solution
Solution by differential is the standard form of solving wave equations.
However, the kinematic wave is so simple that it is also possible to use
29
an analytical solution.
Analytical solution
The solution method is based on Eq. 2.6.9, the formula for the wave
velocity, c:
c = KU
(3.3.1)
In Chapter 2 this equation was derived for a wide, rectangular channel,

giving K=5/3. For a natural channel, K, may vary between 1.3 and 1.6.
The algorithm is based on tracking points in the hydrograph of the wave.
For each point, the water depth and water velocity is computed, based
on Mannings formula and the continuity equation. Then Eq. 3.3.1 is
used to compute the wave speed. The time to travel a given distance to
a downstream cross-section is then computed for each point in the
hydrograph. An example is given in Table 3.3.1, taken from a spreadsheet. The spreadsheet is computed from left to right.
Table 3.3.1 Analytical computation of kinematic wave
Time for X=0

given as input
data (min)
Discharge
Given as
input data
(m3/s)
Velocity
from
Mannings
formula (m/s)
Depth, from
continuity
equation
(m)
Wave
speed, c,
From Eq.
3.3.1 (m/s)
Time for
X=5000
meters
(min)
Time for
X=10000
meters
(min)
100
2.30
0.58
3.83
21.8
43.5
10
200
3.03
0.883
5.05
26.5
43.0
20
300
3.56
1.126
5.94
34.0
48.1
30
400
4.00
1.330
6.66
42.5
55.0
40
300
3.56
1.12
5.94
54.0
68.1
50
200
3.03
0.88
5.05
66.5
83.0
60
100
2.30
0.58
3.83
81.8
103.5
The result is in the two columns to the right in the spreadsheet, 5000 and
10 000 meters downstream. The time, T, in these columns are computed
by the following equation:
T = T0 + X
--c
(3.3.2)
The use of the Mannings formula in the table is derived using the continuity equation to eliminate the water depth:
1
--2
Q
U = MI --------
UB
2
--3
1--- 2--- --5 MI 2 Q 3

or U = -----------------
2--
B3
(3.3.3)
30
Solution by differentials
There exist more involved methods to compute the kinematic wave. The
continuity equation and a formula for the normal depth in a reach is then
used. Note that Eq. 3.2.4 for the kinematic wave gives that the energy
slope is equal to the bed slope. This means the flow is uniform, and a
friction formula can be used, for example Mannings formula. If the velocity in this formula is replaced by Q/A, and the definition of the hydraulic
radius is used, the following derivation can be made:
2
U
QP
-
I 0 = I f = ---------2- = -------------2
-----
3
3
Mr
AA M
(3.3.4)
2--
3
P 5 --5- Q
A = ---------------3- 3-- ----
I 0 10 M 5
(3.3.5)
The equation can be differentiated with respect to time:

2--
5 2--
5 dQ
3 P dA
------------- = --- ---------------Q
3
3
5 ----dt
dt
- ---
10 5
I0 M
(3.3.6)
The continuity equation can be used:
-----Q- + ----A- = 0
x t
(3.3.7)
Combining Eq. 3.3.6 and Eq. 3.3.7 gives:

2--
2
5
--
P - 5 -----Q-----Q- + 3--- ---------------= 0
3 3 Q
t
x 5 ------ ---
I 0 10 M 5
(3.3.8)
Assuming the term in the bracket is constant, the equation can be solved
using first-order differences for time and second-order differences for
space. A notation of two subscripts is used, where the first subscript, i,
denotes the space direction and the second, j, the time:
-----Q- Q
i + 1, j Q i 1, j
-------------------------------------x
2x
(3.3.9)
Q i, j + 1 Q i, j
Q
------- -----------------------------t
t
(3.3.10)
Using these two equations we can transform Eq. 3.3.8 to:
31
2
---
2
--- Q
Q i + 1, j Q i 1, j 3 P 5
i, j + 1 Q i, j
-= 0
- Q i, j 5 ------------------------------------------------------------------- + --- ---------------3- 3--t
5 ----2x
I 0 10 M 5
(3.3.11)
Index i is used for the space dimension and j for the time. Given an initial
situation, Eq. 3.3.11 can be solved with respect to Qi,j+1 to give a formula
for the discharge at a node as a function of the discharges at the nodes
in the previous time step:
3
-----10
Q i, j + 1
3
--5
--- ( Q
5 I0 M 5
i + 1, j Q i 1, j )
------------------------------------------ t
Q
= Qi, j --- ---------------i
,
j
2
3
2x
--5
P
(3.3.12)
The equation can be solved numerically using a spreadsheet, if there

exist simple formulas for P as a function of the discharge. One axis in the
spreadsheet is the distance x, and the other axis is the time.
Example: Solution by differentials.
Compute the water discharge at cross-section 5 at time step 11 minutes,
when water discharge at time step 10 minutes are given as:
Cross-section 4: Q=203 m3/s
Cross-section 5: Q=195 m3/s
Cross-section 6: Q=188 m3/s.
Assume a time step of one minute, and that the cross-sections are 200
meters apart. The bed slope is 1:400 and the Manning-Stricklers coefficient is 66. The wetted perimeter is 54 meters at cross-section 5 for this
discharge.
Solution: We give the numbers to equation 3.3.12:
3
-----10
Q 5, 11
3
---
1 -
5
-------2
x66
--
400
5--- ------------------------------5 ( 188 203 )
= 195
195 ---------------------------- ( 60 )
2
3
2x200
--5
54
The disharge is 208 m3/s.

Discussion
One simplification for the kinematic wave is assuming uniform flow on a
reach. The water level is then a unique function of the water discharge. A
rating curve showing the water levels at a gauging station during the
passing of a flood, will get different values for the rising and the falling
limb of the hydrograph. This is shown in Fig. 3.3.1. However, because of
the simplifications, the kinematic wave method is not able to model this
effect.
32
Figure 3.3.1. Rating curve

during passing of a flood
wave. The kinematic wave
model gives the same values for the rising and falling
of the hydrograph. The
observations give different
stage/discharge observations in the rising and falling
of the hydrograph.
ations
observ
wave
c
i
t
a
kinem
ations
observ
Q
Another point to note is that the differential solution method introduces
some errors, causing the maximum discharge for a wave to be dampened. This is not observed in the quasi-analytical solution method. However, it is noticed in field data. The damping of a real flood wave must not
be confused with the damping introduced by inaccuracies in the numerical algorithm.
3.4 Unsteady flow - Saint-Venant equations

Solving the full Saint-Venant equations is done for dam-break modelling
and other flood problems where there is a rapid change in the water
depth over time, and the water discharge is significantly higher than the
available calibration data.
The approach to solving the Saint-Venant equations is more involved
than solving the kinematic wave equation. There exist a number of different solution methods, which can be divided in two groups:
- Explicit methods
- Implicit methods
Implicit/Explicit method
When the differences in space are to be computed, the question is if the

values in time step j or time step j+1 should be used. If the values in time
step j is used, an explicit solution is given. If the values at time step j+1
are used, an implicit solution is given. An implicit solution is more stable
than an explicit solution, and longer time step can be used. An explicit
solution is simpler to program.
33
The explicit computational molecule is seen in Fig. 3.4.1.

Figure 3.4.1 Nodes for
computation molecule. The upper figure
shows an explicit molecule and the lower figure
shows an implicit molecule. For the explicit
molecule, the value at
node (i,j) is computed
from nodes at the previous time step, j-1. The i
nodes in the space
direction is the different
cross-sections.
Time
j
j-1
i-1
i+1
Space
Time
j
j-1
i-1
i+1
Space
Often, the gradients are computed as a combination of values at time

step j and time step j-1. A weighting factor, , is then used, where the
final solution is times the gradients at time step j, plus (1-) times the
gradients at time step j-1. This means if is 1, an implicit solution is
given, and if is 0 the solution is explicit. If is between 0 and 1, the values at both time steps will be used. The method is then still said to be
implicit. The DAMBRK program uses a default value of 0.6 for , equivalent of using 60 % of the value at time step j and 40 % of the value at
time step j-1.
Both methods are described in the following.
Explicit method
The explicit solution is easier to solve than the implicit method. This can
be seen from Fig. 3.4.1. The initial water discharge in the river is known,
so the first computation starts with the next time step. For each crosssection, i, the discharge can be computed from the discharges at the
previous time step. This is repeated for all cross-sections, and the discharges at the time step j is computed by one sweep. Then the computation proceeds to the next time step.
What is needed is a formula for the water discharge at time step j as a
function of the discharges at time step j-1. This is obtained by discretizing Eq. 3.2.1 and Eq. 3.2.2. For a rectangular channel with depth y and
width B, the continuity equation, Eq. 3.2.1, becomes
y
( yU )
y
y
U
B ----- + B --------------- = B ----- + B U ----- + y ------- = 0
t
x
t
x
x
(3.4.1)
The Bs are eliminated and y and U is taken to be the value at the previous time step. The following differentials are used:
34
U i + 1, j 1 U i 1, j 1
-----U- = --------------------------------------------------2x
x
(3.4.2)
y i + 1, j 1 y i 1, j 1
----y= ----------------------------------------------x
2x
(3.4.3)
----y- = y--------------------------i, j y i, j 1
t
t
(3.4.4)
y i + 1, j 1 y i 1, j 1
U i + 1, j 1 U i 1, j 1
y i, j y i, j 1
--------------------------- + U i, j 1 ------------------------------------------------ + y i, j 1 --------------------------------------------------- = 0
2x
2x
t
(3.4.5)
The equation is solved with respect to the water depth at time step j:
t
y i, j = y i, j 1 ---------- [ U i, j 1 ( y i + 1, j 1 y i 1, j 1 ) + y i, j 1 ( U i + 1, j 1 U i 1, j 1 ) ]
2x
(3.4.6)
In a similar way, the Saint-Venant equation itself (Eq. 3.2.2) can be discretized as:
( U i + 1, j 1 U i 1, j 1 )
U i, j U i, j 1
------------------------------- + U i, j 1 ------------------------------------------------------ +
2x
t
y i + 1, j 1 y i 1, j 1
g ------------------------------------------------ = g ( I 0 I f )
2x
(3.4.7)
where
U i, j 1 U i, j 1
I f = ---------------------------------The source code in C for
the explicit solution of
Saint-Venant equations is
given in Appendix I.
4--2 3
M ri
(3.4.8)
is taken from Mannings equation.

Eq. 3.4.6 can be solved with respect to Ui,j:
( U i + 1, j 1 U i 1, j 1 )t
U i, j = U i, j 1 U i, j 1 ------------------------------------------------------------2x
gt ( y i + 1, j 1 y i 1, j 1 )
-------------------------------------------------------------+ gt ( I 0 I f )
2x
The explicit procedure then becomes:
1. Guess starting values of U and y along the channel.
2. Determine inflow values of U and y
3. Repeat for each time step
4. Repeat for each cross-section of a time
(3.4.9)
35
5. Compute the water level, y, from Eq. 3.4.6

6. Compute If from Eq. 3.4.8
7. Compute the water velocity, U, from Eq. 3.4.9
End of repetition 4
End of repetition 3
The explicit procedure will become unstable if the time step is chosen
too large. The Courant criteria says the time step should be smaller than
what would be required to make a water particle pass from one crosssection to another:
x
t < --------------------( U + c)
(3.4.10)
Figure 3.4.2 Example of

a flood wave routed
downstream a channel
using the full SaintVenant equations and
an explicit solver. The
channel has a slope of
1:200, and a Mannings M
value of 30 was used. The
vertical axis is the water
discharge, and the horizontal axis is the time in
seconds. The different
curves show the
hydrograph at 0, 1250,
2500 and 5000 meters
downstream. The initial
hydrograph has the triangular shape. A time step
of 3 seconds was used.
The source code written in
C to generate this figure is
given in Appendix I.
Control volume approach

The explicit procedure is still fairly unstable in the form given above. An
improvement in stability can be obtained by considering a control volume
approach when discretizing the continuity equation:
Fig. 3.4.3. Figure for
control volume
approach to discretization of continuity
equation
j
j-1
Ui-1
i-1
i
x
Ui
i+1
x
Fig. 3.4.3 shows a longitudinal part of the river, with three cross-sections: i-1, i and i+1. It also shows two water surfaces. One surface is at
time step j-1, and the other is at time step j. The purpose of the algorithm
is to compute the water level at section i, for time step j. This is done on
the basis of the fluxes in and out of the volume upstream of i:
36
Inflow: I= Ui-1 * yi-1
or
( U i 1, j 1 + U i 1, j ) ( y i 1, j 1 + y i 1, j )
I = -------------------------------------------------- ---------------------------------------------2
2
Outflow: O = Ui * yi
(3.4.11)
or
( U i, j + U i, j 1 ) ( y i, j + y i, j 1 )
( y i, j + y i, j 1 )
O = ------------------------------------ --------------------------------- = U i --------------------------------2
2
2
(3.4.12)
The overbar denotes average values over the time step.

The difference between the inflow and outflow is equal to the volume of
the water between the surfaces at time j and j-1, during one time step.
This gives:
( y i, j + y i, j 1 )
I U i --------------------------------- t = 0.5 ( y i, j y i, j 1 )2x
2
(3.4.13)
or
y i, j
U
I y i, j 1 -----i x
------
2 t
= ----------------------------------------------U
x
------ + -----i
t 2
(3.4.14)
This equation can be used to compute the water level at section i, starting from the upstream end going downstream. Then the values of I and
Ui are known. This can be done after computing the velocities by solving
the Saint-Venant equation. This procedure is implemented in the source
code given in Appendix I
Implicit method
The procedure starts with discretization of each differential terms of Eq.
3.2.1 and Eq. 3.2.2 in space and time according to the figure below:
Figure 3.4.4. Discretization of terms between
cross-section i and i-1,
computing for time step
j. Time step j-1 is the previous time step. This is
called a four-point difference scheme.
cross-section
i-1
cross-section
i
Qi-1,j
Qi,j
time step j
reach i
Qi-1,j-1
Qi,j-1
time step j-1
The choice of using the water discharge instead of the velocity as a variable has been reported to give better stability.
The equations are transformed so that all the terms only have the two
37
independent variables, Q and y. The rating curve where the cross-sectional area is given as a function of the water depth (Fig. 3.2.2) is also
used.
After multiplication with A, the first term in Eq. 3.2.2 becomes:
Q i, j + Q i 1, j Q i, j 1 Q i 1, j 1
U
A ------- = --------------------------------------------------------------------------------t
2t
(3.4.15)
Using the chain rule and the continuity equation, the second term
becomes:
2
Q
------
2
A
-----UA
(
U
)
1
AU
= --- -------------- = --- --------------x
2 x
2 x
(3.4.16)
Using finite differences, the term is transformed to:

2
Q
Q
Q
Q
------ ------
------
------
A i, j A i 1, j
A i, j 1 A i 1, j 1
-----U= ----------------------------------------------- + ( 1 ) ------------------------------------------------------------AU
2x
2x
x
(3.4.17)
y = y j + ( 1 )y j 1
Note the weighting factor, , described below Fig. 3.4.1. The third term
becomes:
y i, j y i 1 , j
y i, j 1 y i 1 , j 1
y
Ag ----- = Ag ---------------------------- + Ag ( 1 ) ----------------------------------------x
x
x
(3.4.18)
The friction loss term is discretized by solving the Mannings equation:
gAQ j Q j ( 1 )gAQj 1 Q j 1
gAI f = ------------------------ + ----------------------------------------------------4
--2 2 3
M A R
4
--2 2 3
(3.4.19)
M A R
The variables are here averaged between section i and i-1.

The same procedure is used for the continuity equation (3.2.1). The transient term becomes:
----A- = A
i, j + A i 1, j A i, j 1 A i 1, j 1
------------------------------------------------------------------------------t
2t
(3.4.20)
The second term:
Q i, j 1 Q i 1 , j 1
i, j Q i 1, j
-----Q- = Q
------------------------------ + ( 1 ) -------------------------------------------x
x
x
(3.4.21)
Additionally, there may be lateral inflow/outflow.

Evaluation of the equations
After the terms in the continuity and momentum equations are replaced
38
by Eq. 3.4.18 and 3.4.19, the two equations will be in the form:
f c ( Q, y ) = A c Q i + B c Q i 1 + C c y i + D c y i 1 + E c = 0
(3.4.22)
f m ( Q, y ) = A m Q i + B m Q i 1 + C m y i + Dm y i 1 + E m = 0
(3.4.23)
where A, B, C, D and E are functions of constants and variables at the
previous time step and the indexes c and m denote the continuity and
momentum equation, respectively. Note, all variables at time step j-1 are
known. The variables at time step j are unknown. The momentum equation and the continuity equation are applied to all reaches in the river
between the cross-sections. Given boundary conditions, there are the
same number of unknown as equations.
There are a number of methods for solving the equations. The two main
groups are direct methods and iterative methods. In direct methods, the
equations are set up in a matrix, which is inverted to get the solution. In
iterative methods, guesses are made for the variables, and Equations
3.4.7 and 3.4.8 are modified to get formulas for improvement of the
guessed values. One of the most used method is the Newton-Raphson
formula, an iterative method where the following formula is used to get a
better estimate of Q
f ( Qn )
Q n + 1 = Q n -------------f ( Q n )
(3.4.23)
The same equation would apply for y instead of Q when computing the
water depth.
Stability problems
Experience with solving the equations shows that convergence problems occur where there is supercritical flow. Many computer programs
therefore implement special algorithms to deal with this situation. Some
are more successful than others. If the algorithm in a given program
fails, it is possible to avoid the problem by modifying the friction factor,
M, so that only subcritical flow is present in the problematic area. The
flood wave will then get a slower translation speed as the velocities are
reduced. This must be taken into consideration when evaluating the
results.
Boundary conditions
The upstream and downstream cross-section will need boundary conditions. There are several options:
1. User-specified values
This could typically be results from a dam break computation, which
gives the water discharge as a function of time at the upstream boundary. Or if the downstream value is located in a lake or in the ocean, then
the water elevation there is known.
2. A rating curve
39
This can be used if for example a weir is present at the downstream

boundary, giving a unique relationship between discharge and water
level.
3. Uniform flow approximation
The Mannings formula can be used to find the discharge if uniform flow
is assumed.
4. Zero gradient boundary condition.
This is typically used at the downstream boundary. The depth and/or discharge are assumed to be the same at the downstream boundary crosssection as the second most downstream cross-section.
The choise of boundary conditions must be made based on the available
data and the hydraulic conditions in the river.
3.5 Hydrologic routing

A simplified river routing method is the Muskingum algorithm. It is
derived from the water continuity equation and some assumptions about
the volume of the water in the river:
The volume, V0, of water in the reach is assumed to be proportional to
the steady water discharge, Qo, flowing out of the reach:
V 0 = KQ
(3.5.1)
where K is a proportionality coefficient, with units [sec].

For unsteady flow, the water discharge into the reach is not always
equal to the discharge out of the reach. If the inflow is denoted Qi, the
water continuity defect is assumed to take up a volume equal to:
i
V D = KX ( Q Q )
(3.5.2)
X is another geometrical proportionality coefficient (dimensionless).

The total volume of water in the reach is therefore: V=V0+VD
o
V = KQ + KX ( Q Q ) = K ( XQ + ( 1 X )Q ) )
(3.5.3)
The change in water volume in a reach between one time step j, and the
next time step j+1 becomes:
i
V j + 1 V j = K { [ XQ j + 1 + ( 1 X )Q j + 1 ] [ XQ j + ( 1 X )Q j ] }
(3.5.4)
From the continuity equation, the volume change can also be written:
1 i
1 o
i
o
V j + 1 V j = --- ( Q j + Q j + 1 )t --- ( Q j + Q j + 1 )t
2
2
(3.5.5)
40
Combining the two equations and solving with respect to Qoj+1, the following equation is obtained:
o
Qj + 1 = C1 Qj + 1 + C2 Qj + C3 Qj
(3.5.6)
Expressions for the Cs as functions of X, K and t are:
t 2KX C 1 = -----------------------------------2K ( 1 X ) + t
(3.5.7)
t + 2KX C 2 = -----------------------------------2K ( 1 X ) + t
(3.5.8)
2K ( 1 X ) tC 3 = -----------------------------------2K ( 1 X ) + t
(3.5.9)
By combining Eq. 3.5.4 and 3.5.5, the following expression for K is

obtained:
The parameters in the
Muskingum method are
based on observed discharges. If floods higher
than the observed values
are to be computed, the
coefficients may not be correct. The Muskingum
method can then not be
used.
0.5t [ ( Q j + Q j + 1 ) ( Q j + Q j + 1 ) ]
K = --------------------------------------------------------------------------------------i
i
o
o
X ( Q j + 1 Qj ) + ( 1 X ) ( Q j + 1 Q j )
(3.5.10)
The Muskingum method determines the C values from observed

hydrographs. The numerator and denominator in Eq. 3.5.10 is computed
and plotted for each time interval for different values of X. The value giving the straightest line is chosen, and K is then the slope of this line.
From these values of K and X, Eq. 3.5.7-9 are used to compute C1-3.
When the Cs are given, the water discharge can be determined from
Eq. 3.5.6.
3.6 HEC-RAS
HEC is an abbreviation for Hydrologic Engineering Center. It is a part of
US Army Corps of Engineers. Over the years, the organization has
made several computer programs for water flow problems, named
HEC1, HEC2 etc. HEC2 computed the water surface profile for a steady
water flow in a natural river in one dimension. The solution procedure followed the theory in Chapter 3.2.
HEC-RAS is freeware, and
can be downloaded from the
Internet. It is much used by
Norwegian consulting companies and water authorities.
The original HEC2 program did not have a user interface. It read an
ASCII input file with all the necessary information about water discharge,
friction factors, geometry etc. The result was an output file with the computed water levels. Later, a graphic user interface for the program was
made, with interactive input of data and visualization of results. This version is called HEC-RAS. The most recent version of HEC-RAS includes
algorithms for computing unsteady flow, including mixed flow regime
between supercritical and subcritical flow. It also has a dam break analysis module, and connection to GIS programs. Version 4 came out in
2008 and has sediment transport with movable bed and water quality
computations.
41
3.7 Commercial software

Because the initial HEC2 program was limited in functionality and graphics, there was a period of time in the 1990s where there were a market
for commercial software for hydraulic engineering. The purpose was
often analysis of flooding and dam-breaks. Several commercial programs were developed, and some are briefly described below.
DAMBRK
DAMBRK is made by the US National Weather Service. It consist of two
parts:
- A program for computing the outflow hydrograph from a reservoir
where the dam is breaking
- Routing of the hydrograph downstream, by solving the Saint-Venant
equations.
The program is in principle freeware, and the source code is publicly
available. The original version did not have a user interface, but commercial companies have made user interfaces and these are sold
together with the program. An example is BOSS DAMBRK.
There also exist two programs for separate computations of the dam
break hydrograph and the flood routing, made by the same organization.
The program computing the hydrograph is called BREACH. The program computing only the routing of the flood wave is called DWOPPER.
MIKE11
MIKE 11 is used by Norwegian consulting companies

and The Norwegian Water
Resources and Energy
Directorate for flood zone
mapping.
ISIS is the most used tool by

consulting companies in the
UK for river flow computatitons.
. MIKE 11 is made by the Danish Hydraulic Institute. It is a one-dimensional program with both steady state water surface profile computation
and solution of the full Saint-Venant equations. The program has a
graphical user interface, and program includes connections with GIS
systems. MIKE 11 has a number of different add-on modules, computing
for example:
- rainfall/runoff
- water quality
- sediment transport
- groundwater
The modules makes MIKE 11 very well suited for a solving a number of
different hydraulic river problems.
ISIS
ISIS is made by Hydraulic Research Wallingford, in the UK. It is similar in
functionality to MIKE11, with a graphical user interface and computational modules for one-dimensional steady and unsteady flow. It also has
a number of add-on modules.
3.8 Problems
Problem 1. Coefficients
Derive formulas for the coefficients C1, C2, and C3 in Eqs. 3.5.7-9. This
is done by combining Eq. 3.5.4 and 3.5.5, eliminating the volumes V,
and solving with respect to Qot+1, The resulting equation is compared
42
with Eq. 3.5.6, giving the coefficients.

Problem 2. Flood wave
Water is released from a dam, with the outflow hydrograph given in the
figure below. The river downstream is 50 m wide, and has stones with
d90 of 0.2 meter. The river slope is 1:300. A town is located 2 km downstream of the dam. When will the water start to rise in the town? And
when will the peak of the flood wave reach the town?
Q (m3/s)
60
20
Time (min)
10
70
180
Use a computer program solving the Saint-Venant equation to compute

the maximum water discharge at the town.
43
4. Dispersion of pollutants
4.1 Introduction
Numerical algorithms for

solving the water velocity
equations in 3D can be difficult to learn. It is sometimes
easier to learn the algorithms when looking at dispersion of a pollutant than
computing the water velocity. This is the reason why
the dispersion processes
are described before the
computation of the water
flow.
Dispersion is a combination
of the two processes: convection and diffusion. Convection is pollutant transport
by the time-averaged water
velocity. This is relatively
straightforward to compute.
For our purposes, diffusion
is caused by turbulence and
velocity gradients, which are
more complicated.
The most common method
of modelling diffusion is by
use of a turbulent diffusion
coefficient, , defined as:
F
--A = --------- dc
------
dx
(4.1.1)
F is the flux of a pollutant

with concentration c, passing through an area A, and x
is the direction of the flux
transport.
The picture shows the confluence of two

small rivers in Costa Rica with different
water quality. The branch on the left part of
the picture comes from an area with volcanic activity. (Photo: N. Olsen)
4.2 Simple formulas for the diffusion coefficient

The diffusion coefficient, , for diffusion of a toxic substance, can be set
equal to the eddy-viscosity, T, of the water. The relationship between
the variables are given by the following formula:
= -----TSc
(4.2.1)
Sc is the Schmidt number. This has been found to be in the range of 0.51.0, but more extreme values have been used. In the following, we
assume a value of 1.0, meaning the turbulent diffusion coefficient is
equal to the eddy-viscosity.
Classical hydraulics give a number of empirical and semi-analytical formulas for the eddy-viscosity in rivers or lakes. For a river, the eddy-viscosity, T, is used in the definition of the shear stress in a fluid:
dU
= T ------dy
The equation can be solved with respect to T:
(4.2.2)
44
T = ---------dU
------dy
(4.2.3)
For a wide rectangular channel, Schlichtings wall laws (Eq. 2.1.2) gives
the variation of the velocity with the depth. The vertical velocity gradient
can be obtained by derivation of Eq. 2.1.2, with respect to the distance
above the bed, y:
u
dU
------- = -----*dy
y
(4.2.4)
For a wide, rectangular channel, the shear stress increases linearly from
the surface to the bed:
= 0 1 --y-
(4.2.5)
The water depth is denoted h. Using the definition of the shear velocity,
the equation can be rewritten:
2
= u * 1 --y-
(4.2.6)
Inserting Eq. 4.2.4 and Eq. 4.2.6 into Eq. 4.2.3:
T = u * y 1 --y-
h
(4.2.7)
The average value over the depth is obtained by integrating Eq. 4.2.7
over the depth:
h
1 u *
2
T = --- --------- ( hy y ) dy = --- u * h = 0.067u * h

h h
6
(4.2.8)
Equation 4.2.8 is derived for an idealized case, with a wide straight

channel with rectangular cross-section. Naas (1977) measured the
eddy-viscosity in a number of natural rivers, and suggested the following
formula instead:
T = 0.11u * h
(4.2.9)
Note that this formula is based on the vertical velocity gradients, so this
gives the eddy-viscosity in the vertical direction.
The shear stress on a lake will also introduce turbulence in the water. A
similar approach can be used as for a river, but now the shear stress is
acting on the water surface instead of the bed. The result is a formula
similar to Eq. 4.2.8 and 4.2.9, but a different empirical coefficient may be
used.
45
4.3 One-dimensional dispersion

A typical example of a one-dimensional dispersion problem is pollution
from a point-source into a river where the vertical and lateral mixing is
large. Typical concentration profiles at several points in the river downstream of the spill is given in Fig. 4.3.1:
c
t
spill
x=1km
x=2km
x=3km
x=5km
Figure 4.3.1 Time-series of concentrations at several locations downstream of a spill
The figure shows the two main processes:

- Convective movement of the point of maximum concentration
- Diffusion of the spill, with reduction of the maximum concentration
The transport can be described by a convection-diffusion equation for
the pollutant concentration, c:
dc
d dc
dc
------ + U ------ = ------ ------
dx
dx dx
dt
(4.3.1)
The problem is to find the correct value of the longitudinal diffusion coefficient, . The coefficient is not a function of small-scale turbulent processes. Instead, mixing in the longitudinal direction is often caused by
convective movements due to lateral velocity gradients. The diffusion
coefficient for a one-dimensional model of a river will be much larger
than the small-scale turbulent diffusion used in multi-dimensional models.
Some researchers have developed empirical formulas for the longitudinal dispersion coefficient:
McQuivey and Keefer (1974):
Q
= 0.058 -----IB
(4.3.2)
Fisher et. al. (1979)

2
( UB )
= 0.011 --------------Hu *
(4.3.3)
46
Q is the water discharge in the river, with slope I, width B, and depth H.
U is the water velocity and u* is the shear velocity.
The convection-diffusion equation can be solved analytically, assuming
constant values of velocity and diffusion coefficient, giving (Chapra,
1997, p.182):
( x Ut )
c 0 L --------------------4t
c ( x, t ) = ---------------- e
2 t
(4.3.4)
The initial concentration is denoted c0, and the length of the spill in the
river is denoted L.
In a natural river, the simplifications used to derive Eq. 4.3.4 are not
valid. The equation may give a rough estimate of the concentration, but
to get better accuracy, it is necessary to solve the convection-diffusion
equation numerically.
Looking at measurements of pollution concentration in a river, the profiles in Fig. 4.3.1 have an additional feature: a prolonged tail. It is caused
by storage of pollution in recirculation zones and dead waters along the
river. This effect is difficult to take into account using one-dimensional
models. Instead, it is possible to model the river using a three-dimensional model, where the water flow field is modelled, including recirculation zones. This approach also have the advantage that the uncertainty
with the longitudinal diffusion coefficient is eliminated. It is computed as
a part of the solution of the equations. Research is ongoing in this area.
4.4 Jets and plumes

Jets and plumes are water entering a reservoir or a lake, for example
from a river or a wastewater outlet. For an idealized case, there exist formulas for the dispersion of jets and plumes. The formulas can be derived
analytically based on the momentum equation, but empirical coefficients
are required when computing dispersion because of turbulence.
Close to the outlet, the momentum of the water will be a dominating
force determining the flow field. This is then called a jet. Fig. 4.4.1 shows
a jet close to the outlet:
Figure 4.4.1 Velocity profiles at the core of a jet
outlet. The entrance profile
is denoted A, the core
region ends at profile C.
Profile B has uniform distribution at the core, but
curved distribution at the
edge. Profile D has reduced
maximum velocity, as it is
located after the core
region.
Assuming a uniform velocity profile at the outlet, the jet will have a core
where the velocity distribution changes. The length of the core is approximately 6 times the diameter of the outlet. In the core, the maximum
velocity is constant. The water from the jet is mixed with the surrounding
water, reducing the velocity in this area. The reduction of the velocity will
47
finally take place also at the center of the jet. This point forms the end of
the core.
After the core, the jet may moves in various directions, depending on:
- The geometry around the plume
- The density difference between the inflowing and the surrounding
water
- The density stratification in the surrounding water
- The velocity field in the surrounding water
- The turbulence in the surrounding water.
A flow situation dominated
by the momentum of the
inflowing water is often
called a jet. If the flow situation is dominated by the
density difference between
the inflowing and receiving
water, this is often called a
plume.
If the velocity of the surrounding water is very strong, this may also affect
the jet in the core region.
For an idealized case, it is possible to derive formulas for the velocity
and effective discharge after the outlet. Assuming the receiving water
has no velocity, turbulence or density stratification, and its density is the
same as the water in the jet, the momentum of the water stays the same
in a cross-section of the jet. It is also assumed the jet will not interact
with any geometry. The momentum equation, together with experiments
then give the following equations for a jet from a circular pipe (Carstens,
1997):
2
2
u
r
----------- = 1 + -------------------
2
u max
0.016x
(4.4.1)
u max
x 1
----------- = 6.4 -----
d 0
u0
(4.4.2)
xQ- = 0.42 -------d

Q0
0
(4.4.3)
The velocity is denoted u, r is the distance from the centerline of the

plume, x is the distance from the pipe outlet, d0 is the diameter of the
pipe and Q0 is the water discharge out of the pipe. Because the surrounding water will be mixed into the plume, the total water discharge, Q,
in the plume will increase with the distance from the outlet.
The formulas are empirical. Fischer et al (1979) came up with slightly different coefficients: 6.2 instead of 6.4 in Eq. 4.4.2, and 0.28 instead of
0.42 in Eq. 4.4.3.
If the receiving water does not have the same density as the water from
the pipe, the plume will rise or sink. A densimetric Froude number, Fr, is
often used to derive formulas for the plume.
u0
Fr = --------------------------------------
res
0
------------------- res gd 0
(4.4.4)
It is assumed that plume water density, 0, has a lower value than the
recipient water density, res. The formulas below are given by Rouse et.
al. (1952) from experiments:
48
1
--3
2
--3
z-
u- = 4.3Fr ------ d 0 e
u0
5
--3
2--3
res
z
------------------ = 9Fr -----
d0
0
2---
r
96 ----2
z
(4.4.5)
r
71 ----2
z
(4.4.6)
5
---
z- 3
Q- = 0.18Fr 3 ------- d 0
Q0
(4.4.7)
It is assumed the pipe releases the water in the vertical direction.

Detailed derivations of the formulas, together with equations for more jet/
plume cases are given by Fisher et. al. (1979).
4.5 Problems
Problem 1. Dispersion in a river
Two thousand litres of a toxic chemical is spilled from a factory into a
river during ten minutes. The river has a water discharge of 20 m3/s, an
average depth of 2 meters, and average width of 30 meters and a slope
of 1:63. Ten kilometer downstream of the factory is a city. Compute the
concentration of the chemical in the river at the city as a function of time.
Problem 2. Dispersion of a plume
A plume rises from a hydropower plant outlet into the sea. The sea water
has a salinity of 3 %, and a density of 1023 kg/m3. The water discharge
is 50 m3/s from a tunnel with diameter 3 meters. The water from the tunnel has a density of 1000 kg/m3. The lake is 30 meters deep at the outlet
point. What is the velocity of the water 20 meters right above the outlet?
Assume no initial water currents or vertical stratification.
What will happen if there are currents in the lake?
49
5. Dispersion modelling in 2D and 3D

CFD is about computing a
bit and then guessing a bit,
for so to compute a bit more
and see if one has guessed
correctly. What differentiates the various algorithms
is where one is computing
what and with which numbers - student answering an
exam in Hydroinformatics.
Grid classifications
Dispersion modelling in 2D or 3D is conducted using a science called

computational fluid dynamics, or CFD. This chapter gives an introduction
to CFD, applied to modelling dispersion of pollutants.
5.1 Grids
A basic concept of CFD is to divide the fluid geometry into elements or
cells, and then solve an equation for each cell. In the following text, the
word cell will be used instead of element, to avoid confusion with the
finite element method. The algorithms described in the following chapters are based on the finite volume method
Grids can be classified according to several characteristics:
shape
orthogonality
structure
blocks
grid movements
nesting
outblocking
The shape of the cells is usually triangular or quadrilateral:
Fig. 5.1.1. Grid shapes
Triangular and quadrilateral shapes

The orthogonality of the grid is determined by the angle between crossing grid lines. If the angle is 90 degrees, the grid is orthogonal. If it is different from 90 degrees, the grid is non-orthogonal.
Fig. 5.1.2. Grid orthogonality
Orthogonal grid
Pronounciation of the
Greek letters:
: ksi
: psi
: zeta
Non-orthogonal grid
For non-orthogonal grids, a non-orthogonal coordinate system is often

used to derive terms in the equations. The coordinates then follow the
grid lines of a structured grid. The three non-orthogonal coordinate lines
are often called ,,, corresponding to x,y and z in the orthogonal coordinate system. Fig. 5.1.3 shows the two systems in 2D. Additionally, the
third direction will be z in the cartesian system and in the computational
domain.
50
Figure 5.1.3 Body-fitted

coordinate system. The
two coordinate systems
are shown in two dimensions, where some values
of and are given in
brackets
y
(4,4)
(2,4)
x
(3,3)
(4,1)
(1,1)
(2,1)
(3,1)
The directions along the computational domain are often called (,) in
2D as shown in the figure above, or (,,) in 3D, where the last index is
the vertical direction. In the computational domain, the distance between
the grid lines are often set to unity, so it is easy to calculate gradients of
variables. It means all will be unity.
An important definition is the notation of the variables at a cell. Instead of
using x,y and z directions, the non-orthogonal cell now uses the directions north, south, east, west, bottom and top. Another definition is to
use indexes, as in tensor notation. Then direction 1 is east-west, direction 2 is north-south and direction 3 is vertical. Using tensor notation,
(,,) can also be written (1,2,3).
Grid structure
Grids can be structured or unstructured. Often a structured grid is used

in finite volume methods and an unstructured grid is used in finite element methods. However, this is not always the case. The figure below
shows a structured and an unstructured grid. In a structured grid it is
possible to make a two-dimensional array indexing the grid cells. If this is
not possible, the grid is unstructured.
Figure 5.1.4 Grid structure
Unstructured grid
Structured grid
Level 2
Almost all grids using triangular cells are unstructured.

It is possible to connect several structured grids. Each grid is then called
a block, and the result is called a multi-block grid.
Adaptive grid
The grid may also move during the computations. A grid that moves
according to the solution of the equations is called an adaptive grid.
Typical examples are vertical movements due to changes in water lev-
51
els, or changes in bed levels due to erosion or sedimentation. The grid

may also move laterally, for example modelling a meandering river.
Examples are given in Chapter 9. All these movements are due to
changes in the geometry of the computational domain. Grid movements
can also be internal. Often, a higher grid density is wanted in large gradients, and algorithms have successfully been used to change the internal
grid structure to change the cell sizes according to for example velocity
gradients.
High gradients may also necessitate the use of a nested grid. This
means that a grid with small cells are located inside a coarser grid. A
typical example is computation of pollutants from a point source in a
lake. The lake itself is modelled with a coarse grid, while the concentration around the point source is modelled with a finer grid, nested inside
the coarse grid.
Outblocking is a procedure where cells in a structured grid are made
inactive. This makes it easier to generate structured grids in a complex
geometry.
Grid Qualities
The accuracy and convergence of a finite volume calculation depends
on the quality of the grid. Three grid characteristics are important:
Grid qualities
- non-orthogonality
- aspect ratio
- expansion ratio
The non-orthogonality of the grid line intersections is the deviation from
90 degrees. If the grid line intersection is below 45 degrees or over 135
degrees, the grid is said to be very non-orthogonal. This is a situation
one should avoid. Low non-orthogonality of the grid leads to more rapid
convergence, and in some cases better accuracy.
The aspect ratio and expansion ratio is described in the figure below:
Figure 5.1.5 Expansion/aspect ratio definition:
yA
xA
The figure shows two grid cells, A and B. The length of the cells are xA
and xB.
The expansion ratio of the grid at these cells is xA/xB.
The aspect ratio of the grid at cell A is xA/yA.
The expansion ratio and the aspect ratio of a grid should not be too
great, in order to avoid convergence problems and inaccuracies. Aspect
ratios of 2-3 should not be a problem if the flow direction is parallel to the
longest side of the cell. Experience shows that aspect ratios of 10-50 will
give extremely slow convergence for water flow calculations. Expansion
ratios under 1.2 will not pose problems for the solution. Experience also
shows that expansion ratios of around 10 can give very unphysical
results for the water flow calculation.
52
Grid generation
Some kind of geographical information is required to make a grid of a
natural river or lake. Often a map can be used. The first step is usually to
determine the points at the edges of the grid. Then the internal grid intersections are made.
Two of the most commonly used methods to generate internal points in a
structured grid are called Transfinite interpolation and Elliptic grid generation. These are described in the following.
Transfinite interpolation
In a transfinite interpolation, the grid lines on two opposing edges are
connected with straight lines. An example is given in Fig. 5.1.6. The
method is well suited for modelling rivers, as the straight lines can be
cross-sections. Note that the straight lines are only generated in one
direction. For the right figure in Fig. 5.1.6, the lines in the longitudinal
direction are not straight.
Fig. 5.1.6. Grid generated with
transfinite interpolation.
Elliptic grid generation

Some times a smoother grid is required, than the result of the Transfinite
interpolation. The elliptic grid generation solves a differential equation for
the location of the grid intersections:
2
i = 0
The grid given in Fig. 5.1.7 is made by this method.
Fig. 5.1.7. Grid made by an elliptic grid generation.
53
5.2 Discretization methods

The discretization described here is by the control volume method.
Steady state dispersion is governed by the convection-diffusion equation
for the concentration, c, of the pollutant:
c
U i ------- = ------- -------
x i
x i x i
(5.2.1)
The left side of the equation is the convective term, and the right side of
the equation is the diffusive term.
The main point of the discretization is:
Discretization is:
The new function can be thought of as a weighted average of the concentration in the neighbouring cells. For a two-dimensional situation, the
following notation is used, according to directions north, south, east and
west:
:
cn
cw
cp
cs
To transform the partial differential equation into a new equation where

the variable in one cell is a function of the variable in the neighbour cells
ce
ae : weighting factor for cell e

aw : weighting factor for cell w
an : weighting factor for cell n
as : weighting factor for cell s
ap = ae+aw+an+as
The formula becomes:

Figure 5.2.1 Discretization
molecule. Computation of concentration, c, in the center cell,
a w c w + ae ce + an cn + as cs
p, as a function of the concentrac p = -------------------------------------------------------------------tion in the neighbouring cells n,
ap
s, e and w.
(5.2.2)
The weighting factors for the neighbouring cells ae, aw, an and as are
often denoted anb
What we want to obtain are formulas for anb.
In a three-dimensional computation, the same principles are involved.
But two more neighbouring cells are added: t (top) and b (bottom),
resulting in six neighbour cells. The simple extension from 2D to 3D is
one of the main advantages of the finite volume method.
There are a number of different discretization methods available for the
control-volume approach. The difference is in how the concentration on
a cell surface is calculated. Some methods are described in the following.
Note that the methods are based on the physics of the dispersion and
flow processes. They are not mathematically based, or derived from the
convection-diffusion equation (5.2.1).
54
Development of CFD
algorithms was initially
done in aeronautics.
The fluid was air, and
the methods were then
called upwind instead of
upstream. Both expressions are used, meaning the same.
5.3 The First-Order Upstream Scheme

For a non-staggered grid, the values of the variables are given in the
center of the cells. Using the finite volume method, it is necessary to estimate variable values on the cell surfaces. The main idea of the upstream
methods is to estimate the surface value from the upstream cell. The first
order method uses information in only one cell upstream of the cell surface. In other words: the concentration at a cell surface for the first-order
upstream method is the same as the concentration in the cell on the
upstream side of the cell side.
The control volume method is based on continuity of sediments. The
basis of the calculation is the fluxes on a cell surface. The surface area
is denoted A; the velocity at the surface, normal to it, is denoted U; c is
the concentration at the surface, and is the turbulent diffusion at the
surface.
The convective flux is calculated as: U * A * c
The diffusive flux is calculated as: * A * dc / dx
(5.3.1)
(5.3.2)
The term dc/dx is calculated as the concentration difference between the

cells on each side of the surface, divided by the distance between the
centres of the cells. Looking at the west side of cell p, Fig. 5.3.1 explains
the variable locations and the fluxes in the center cell.
Figure 5.3.1 Fluxes
through the walls of
the center cell in a
computational molecule. The cells have a
width of dx and a height
of dy. Note, the diffusion
coefficient and the
velocities are given on
the boundary between
the cells. The concentrations are computed in
the centre of each cell.
Also note that this is a
two-dimensional situation. The area of a cell
surface is therefore
equal to dx or dy, multiplied with a unit depth in
the third dimension.
cn
n
Un
Ue
Uw
dy
cw
cp
ce
Us
s
cs
dx
The flux, Fw, through the west side of cell P then becomes:
Aw ( cw cp )
F w = U w A w c w + w ----------------------------dx
(5.3.3)
where Aw is the area of the cell wall on the west side, equal to y times
the height of the wall. For the other sides, the following fluxes are
obtained:
55
Ae ( cp ce )
F e = U e A e c p + e --------------------------dx
As ( cp c s )
F s = U s As c p + s -------------------------dy
An ( cn cp )
F n = U n A n c n + n --------------------------dy
(5.3.4)
(5.3.5)
(5.3.6)
Sediment continuity means the sum of the fluxes is zero, in other words:
F w F e + Fn F s = 0
(5.3.7)
This gives the following equation:
A
A
A
w
A
----------e- + U s A s + s ------s + n -----n- c p
+
U
A
+
w
e
e
e
dx
dx
dy
dy
A
A
An
As
w
e
= U w A w + w ------ c w + e ------ c e + U n An + n ------ c n + s ------ c s
dx
dy
dx
dy
(5.3.8)
When we compare Equation 5.2.2 with Equation 5.3.8, we see they are
the same. The concentration in Cell P is a function of the concentration
in the neighbouring cells. The resulting weighting factors are:
Aw
Ae
As
An
a p = w ------ + U e A e + e ------ + U s A s + s ------ + n -----dx
dx
dy
dy
Aw
a w = U w A w + w -----dx
Ae
a e = e -----dx
As
a s = s -----dy
An
a n = U n A n + n -----dy
(5.3.9)
(5.3.10)
(5.3.11)
(5.3.12)
(5.3.13)
The water continuity equation for the grid cell is:
Uw Aw Ue Ae + Un An Us As = 0
(5.3.14)
U w A w + U n An = U e A e + U s A s
(5.3.15)
or:
If the above equation is inserted into the expression for ap, the equation
a p = a e + a w + a s + an
(5.3.16)
is verified to be correct.
Note that the equations above are only valid if the velocity flows in the
56
same direction as given on the arrows in Fig. 5.3.1.

Example:
Particles are deposited in a river with constant width and depth. A 2D situation is assumed, averaged over the width. The north-south direction is
used as the vertical direction, so the indexes s and n are replaced by b
and t. Further simplifications are:
- uniform the water flow
- negligible horizontal diffusion
Then the vertical velocity is equal to the sediment particle fall velocity, w.
Also, the grid can be made orthogonal and two-dimensional, so Ae = Aw
= dy, and Ab = At = dx. The weighting factors then become:
ae = 0.0
aw = U dy
ab = dx/dy
at = w dx + dx/dy
ap = U dy + w dx + 2 dx/dy
(5.3.17)
If for example, U is 2 m/s, w is 0.01 m/s, is 0.01 m2/s, and the river
depth is 4 meters, we may assume 10 cells in the vertical direction, giving dy = 0.4 m. Modelling a reach of 1 km with 100 cells, gives dx=10 m.
The coefficients becomes:
ae = 0.0
aw = 0.8
ab = 0.25
at = 0.35
ap = 1.4
(5.3.18)
The numbers can be inserted in a spreadsheet and the problem solved.

This is described in the next chapter.
The Power-Law Scheme
The Power-Law Scheme is a first-order upstream scheme where the diffusive term is multiplied with the following reduction factor:
f = ( 1 0.1 Pe )
(5.3.19)
where Pe is the Peclet number, given by:
Pe = Ux
----------
(5.3.20)
The Peclet number is the ratio of convective to diffusive fluxes. The factor f is always between 1 and 0. The diffusive term will be reduced for
flows where the convection is large compared with the diffusion.
5.4 Spreadsheet programming

Given formulas for the water flow field and the turbulence, it is possible
to make a spreadsheet for calculation of the pollutant concentration. One
57
application is to calculate the trap efficiency of a sand trap. Then a twodimensional width-averaged approach is used. A structured orthogonal
grid is used, where each cell in the grid is simulated by a cell in the
spreadsheet. If a uniform water velocity and turbulence field can be
assumed in the vertical direction, then the same anb coefficients can be
used for all the cells. A more advanced approach is to use a logarithmic
velocity distribution, and a given distribution of the eddy-viscosity.
If the simpler approach is used, the coefficients anb can be calculated
before the programming starts. This is based on the formulas given previously, and a chosen number of grid cells. The grid is structured,
orthogonal and all cells have the same size.
A figure of this spreadsheet is given below, with 8 cells in the vertical
direction, and 9 cells in the horizontal direction. The size of the grid can
of course be changed according to the dimensions of each problem.
A
1
B
0
C
0
D
0
E
0
F
0
G
0
H
0
I
0
J
0
K
0
10
The cells marked X are inflow boundary conditions. These are the
A2..A10 cells. A concentration value is given in these cells. A constant
value can be given, or it is possible to use a formula for the vertical distribution of the concentration.
The cells marked 0 is the boundary condition at the water surface. This
is zero.
The cells marked Y is the outflow boundary condition. If the horizontal
diffusion is assumed to be zero, the values in these cells will not affect
the computation. If the horizontal diffusion is non-zero, a zero gradient
boundary condition can be used. Then the formula in these cells should
be:
Cell K2: =J2
Cell K3: =J3
...
Cell K10: =J10
58
The cells marked Z is the bed boundary condition. A formula for the equilibrium sediment concentration can be used, for example van Rijns formula, given in Chapter 9.6. However, often the shear stress is below
critical at the bed of the sand trap, giving zero concentration. This will not
be correct, as the sediment concentration at the bed will always be
higher than the cell above. Therefore, the concentration can be set equal
to the concentration in the cell above.
(Note that a more detailed calculation will give a very low diffusion coefficient close to the bed of the sand trap, meaning the concentration in the
cell above the bed is independent of the concentration in the bed boundary. For a simplified calculation, the diffusion coefficient is significant.
Then the same result is obtained if zero gradient boundary condition is
used.)
Cell B10: =B9
Cell C10: =C9
...
Cell J10: =J9
The discretized formula now has to be given in all the remaining interior
cells. As an example, the following data is assumed:
aw = 0.1
an = 0.2
as = 0.006
ae = 0.002
ap = 0.308
Starting in cell B2, we give the following formula:
+(0.1*A2+0.2*B1+0.006*B3+0.002*C2)/0.308
The method of invoking
more iterations is dependent on the particular
spreadsheet program. For
Lotus 123, use F9 on the
keyboard repeatedly. For
MS Excel, use the menu
Tools, Options, Calculations, and cross off Iterations, and give a number
in the edit-field, for example 50.
This formula is copied to all the interior cells, from cell B2 to J9. Afterwards, the calculation has to be repeated some times to get convergence.
The trap efficiency is calculated by first summing the inflow and the outflow:
Inflow: sum of cells A2..A10
Outflow: sum of cells K2..K10
Trap efficiency = (Inflow-Outflow)/Inflow
Running this case with varying number of grid cells will give different
result for the trap efficiency. The next chapter explains why.
5.5 False diffusion

False diffusion is due to the approximations in the convective terms in
the discretization schemes. More specifically, how the concentration on
the cell side is calculated. The effect is best shown with a coarse grid
and steep gradients, for example the following situation:
A sand trap is calculated with 5 x 5 cells in the vertical and horizontal
direction, respectively. The cells are 1 meter high and 5 meters long.
The water velocity is 0.3 m/s, and the sediment fall velocity is 6 cm/s.
The sediments are added from a point source at the water surface, over
59
a length of 5 meters. There is no turbulence! This gives the following

theoretical concentration profile in the flow:
Figure 5.5.1. Profile of

the real concentration
in the geometry
The concentration is unity along a band in the flow. The concentration is

zero elsewhere.
The first-order upstream scheme is used to calculate the concentration.
The following coefficients are obtained:
an = 0.06 * 5 = 0.3
aw = 0.3 * 1 = 0.3
as = 0.0
ae = 0.0
ap = 0.6
This gives an/ap = 0.5 and aw/ap = 0.5. In other words, the concentration
in a cell is the average of the concentration in the cell above and in the
cell upstream. The boundary condition is a concentration of unity in one
cell at the surface, and zero concentration in the other surface cells and
the inflow cells. The following result is obtained:
1.0
Figure 5.5.2. Grid
with computed concentration values
0.0
0.5
0.25
0.125
0.0625
0.0313
0.25
0.25
0.188
0.125
0.0781
0.125
0.188
0.188
0.156
0.117
0.0625
0.125
0.156
0.156
0.137
0.0313
0.0781
0.117
0.137
0.137
0.0156
0.0469
0.0820
0.109
0.123
The maximum concentration has decreased from unity to 0.137 at the

bed. It has also been smeared out over several cells at the bed.
False diffusion can be avoided by:
- aligning the grid with the flow direction
- increasing the number of grid cells
- using higher order schemes
The last option is described in the following chapter.
60
5.6 The Second Order Upstream Scheme

The Second-Order Upstream (SOU) method is based on a second-order
accurate method to calculate the concentration on the cell surfaces. The
method only involves the convective fluxes, and the diffusive terms are
calculated as before. The following figure shows the calculation of the
concentration on the west side of cell p: side W:
Concentration
Figure 5.6.1 Defini- W
c
tion sketch for
concentration estimation at the wall,
cw
W, for the SOU
scheme.
West wall W
Cell ww
Cell w
Cell p
dy
cww
U
dx
dx
The cell on the west side of cell w is called cell ww. The concentration in
this cell is denoted cww. The concentration in cell w is denoted cw and
the concentration on side W of cell p is denoted cW. The SOU scheme
uses the concentration in cell ww and cell w to extrapolate linearly to
side W. Given the width of the cell in the x-direction is dx, and the height
in the y-direction is dy, it is possible to derive a formula for the concentration on side W by triangulation:
w
c c ww
c w c ww
------------------------- = -------------------dx + 0.5dx
dx
(5.6.1)
or
c
NN
N
WW W
EE
3
1
= --- c w --- c ww
2
2
(5.6.2)
Equation 5.6.1 is only valid if the cells are of equal size. If the expansion
ratio is different from unity, a separate formula needs to be applied,
where the coefficients 3/2 and are given as a function of the expansion ratio.
The calculation molecule now gets nine cells, as shown in the figure to
the left
SS
The flux through the west side of Cell P then becomes:
Figure 5.6.2 SOU ninepoint calculation molecule
Aw ( cw cp )
3
1
F w = U w A w --- c --- c ww + w ----------------------------2 w 2
dx
For the other sides, the following fluxes are obtained:
(5.6.3)
61
Ae ( cp ce )
3
1
F e = U e A e --- c --- c w + e --------------------------dx
2 p 2
As ( cp cs )
3
1
F s = U s As --- c --- c + s -------------------------dy
2 p 2 n
An ( cn cp )
3
1
F n = U n A n --- c n --- c nn + n --------------------------2
2
dy
(5.6.4)
(5.6.5)
(5.6.6)
Again, the equations are only valid if the velocity vectors are in the same
direction as in Fig. 5.3.1. The weighting factors becomes:
Aw 1
3
a w = --- U A w + w ------ + --- U e A e
2 w
dx 2
1
a ww = --- U A w
2 w
Ae
a e = e -----dx
a ee = 0
An 1
3
a n = --- U A n + n ------ + --- U As
2 n
dy 2 s
1
a nn = --- U A n
2 n
A
a s = s ------s
dy
a ss = 0
(5.6.7)
(5.6.8)
(5.6.9)
(5.6.10)
(5.6.11)
(5.6.12)
(5.6.13)
(5.6.14)
For the SOU scheme, Equation 5.3.1 now becomes:
a w c w + a e c e + a n c n + a s c s + a ww c ww + a nn c nn
c p = --------------------------------------------------------------------------------------------------------------------------ap
(5.6.3)
The formula is used for a two-dimensional situation. In 3D, the terms for
top and bottom is also added, giving four extra coefficients: at, att, ab,
abb.
5.7 Time-dependent computations and source terms

The derivations given previously has been made under the assumption
of a steady state condition. Often, it is necessary to compute the concentration changes over time. In Chapter 5.3, the fluxes into cell P was
equal to the fluxes out of the cell. In a time-dependent situation, this may
not be the case. To derive an equation for this problem, we look at what
happens during a time step t, between time t and t-1. The mass
change, m, in cell P would then be:
( c p, t c p, t 1 )Vp
m = ---------------------------------------t
(5.7.1)
62
The volume of cell P is then denoted Vp

The fluxes in and out of cell P can then be computed as previously, but it
is necessary to multiply with the time step, to get the mass instead of the
flux. Equation 5.3.7 becomes:
tF w tF e + tF n tF s = m
(5.7.2)
Combining Eqs. 5.7.1 and 5.7.2, we obtain:
anb cnb ap cp
nb
( c p, t c p, t 1 )V p
= F w F e + F n F s = ---------------------------------------t
(5.7.3)
The left side of the equation is similar to the steady state situation. For a
time-dependent computation, the term on the right hand side emerges.
The equation can be rewritten:
ap c
( c p, t c p, t 1 )V p
= a w c w + a e c e + a n c n + a s c s ---------------------------------------p
t
(5.7.4)
The anb and ap coefficients are now the same as for the steady equation.
The additional term on the right side is called a source term. In a computer program these has to be calculated. The most commonly way of
doing this for Eq. 5.7.4 is by dividing the source term in two, according
to:
( c p, t c p, t 1 )V p
S C + S P c p, t = ---------------------------------------t
(5.7.5)
The source terms are:
c p, t 1 V p
S C = --------------------t
(5.7.6)
This term only depends on known variables, as the concentration at the

previous time step is known.
V
S P = -----pt
(5.7.7)
The final equation can then be written in the following way, where the anb
and ap coefficients are the same as for the steady case:
aw cw + a e ce + a n cn + a s cs + S C
c p = -------------------------------------------------------------------------------( ap + SP )
(5.7.8)
The equation can be solved similarly as for a steady case. However,
63
because of the new dimension, time, it is difficult to do this in a spreadsheet. A computer program is often necessary.
The transient convection-diffusion equation can be written:
c
c
----- + U ------- = ------- -------
x i x i
t
i x i
(5.7.9)
5.8 Problems
The following three line styles
are used in the sketches in
the problems 1-6:
Inflow
Outflow
Wall
Problem 1. Grid for reservoir

Make a sketch of a 2D structured non-orthogonal grid with 300 cells for
the geometry given below.
Problem 2. Structured grid for groyne

Make a sketch of a 2D structured non-orthogonal grid with 300 cells for
the geometry given below, without using outblocking.
Problem 3. Grid with outblocking

Make a sketch of a structured non-orthogonal grid with 300 cells for the
geometry given below, but use outblocking for the groyne.
64
Problem 4. Grid for bay

Make a sketch of a structured non-orthogonal grid with 600 cells for the
geometry given below.
Problem 5. Cylinder grid

Make a paper sketch of a structured non-orthogonal grid with 2400 cells
for a circular cylinder with diameter 1.0 meters placed vertically in the
centre of a 4 meter wide and 7 meter long flume.
Problem 6. Dispersion of particles
Particles are dumped in a river with depth 2 meters, velocity 1 m/s and a
slope of 1/300. The particles have a fall velocity of 0.01 m/s. A water
intake is located 1 km downstream of the dumping place. What is the
percentage of particles passing the intake? Assume no resuspension of
the particles, and that the particles are added close to the water surface.
65
6. Numerical modelling of water velocity in

2D and 3D
Claude Louis Marie

Henri Navier was professor at cole Polytecnique
in Paris from 1819 to
1831. He derived the
Navier-Stokes equations
in 1822, 23 years before
Stokes. Prof. Navier also
worked on road and
bridge constructions, and
derived theories for suspension bridges.
This chapter describes the solution procedures for the Navier-Stokes

equations. These equations describe the water velocity and turbulence
in a river or a hydraulic system.
6.1 The Navier-Stokes equations

The Navier-Stokes equations describe the water velocity. The equations
are derived on the basis of equilibrium of forces on a small volume of
water in laminar flow:
U i
U i
U U
1
--------- + U j --------- = --- ------- P ij + ---------i + ---------j
x j
x j
t
x j x i
(6.1.1)
For turbulent flow, it is common to use the Reynolds averaged versions

of the equations. The Reynolds averaging is described first.
We are looking at a time series of the velocity at a given location in turbulent flow:
Figure 6.1.1 Time series

of water velocity, for
definition of velocity
fluctuations, u.
Ut
U
u
Time
Turbuilence is a cascade
of eddies, from large
eddies to small eddies.
The largest eddies take
their energy from the main
flow. The large eddies
give energy to smaller
eddies etc., until the eddy
size is at the Kolmogorov
micro scale. There, the
energy is killed by viscosity. Prof. H. Wengle
The velocity, Ut is divided into an average value U, and a fluctuating

value u. The two variables are inserted into the Navier-Stokes equation
for laminar flow, and after some manipulations and simplification the
Navier-Stokes equation for turbulent flow emerges:
U i
U i
1
--------- + U j --------- = --- ------- ( P ij u i u j )
x j
x j
t
(6.1.2))
P is the pressure and ij is the Kronecker delta, which is 1 if i=j and 0 otherwise. The last term is the Reynolds stress term, often modelled with
the Boussinesq approximation:
U U
2
u i u j = T ---------i + ---------j --- k ij
3
x j x i
(6.1.3)
The variable k is the turbulent kinetic energy. This is further described in
66
Chapter 6.3. Inserting Equation 6.1.3 into Equation 6.1.2 and regrouping
the variables:
U i
U i
U j
U i
1
2
--------- + U --------- = --- ------- P + --- k ij + T --------- + T -------- x j
3
x j
x i
t
j x j
(6.1.4)
There are basically five terms: a transient term and a convective term on
the left side of the equation. On the right side of the equation there is a
pressure/kinetic energy term, a diffusive term and a stress term.
Important note:
The convective and diffusive term are solved with the same methods as
the solution of the convection-diffusion equation for dispersion modelling
in Chapter 5. The difference is that the pollution concentration is
replaced by the velocity.
The stress term is sometimes neglected, as it has very little influence on
the solution for many cases. The pressure/kinetic energy term is solved
as a pressure term. The kinetic energy is usually very small, and negligible compared with the pressure.
A difference between Eq. 6.1.4 and the convection-diffusion equation for
sediments is the diffusion coefficient. Eq. 6.1.4 includes an eddy-viscosity instead of the diffusion coefficient. The relationship between these
two variables is:
T = Sc
(6.1.5)
where Sc is the Schmidt number. This is usually set to unity, meaning

the eddy-viscosity is the same as the turbulent diffusivity.
This leaves the problem of solving the pressure term. Several methods
exist, but with the control volume approach, the most commonly used
method is the SIMPLE method. This is described in the next chapter.
6.2 The SIMPLE method

Modern CFD using the
finite volume method, the
SIMPLE method and the
k- turbulence model was
pioneered by a group of
researchers in the early
1970s at Department of
Mechanical Engineering,
Imperial College, London.
The group included
researchers as D. B.
Spalding, B. E. Launder,
S. V. Patankar and W.
Rodi. The algorithms are
used in most commercial
CFD programs today.
SIMPLE is an abbreviation for Semi-Implicit Method for Pressure-Linked

Equations. The purpose of the method is to find the unknown pressure
field. The main idea is to guess a value for the pressure and use the continuity defect to obtain an equation for a pressure-correction. When the
pressure-correction is added to the pressure, water continuity is satisfied.
To derive the equations for the pressure-correction, a special notation is
used. The initially calculated variables do not satisfy continuity and are
denoted with an index *. The correction of the variables is denoted with
an index . The variables after correction do not have a superscript. The
process can then be written:
P = P* + P
Uk = Uk * + Uk
(6.2.1)
(6.2.2)
P is the pressure and U is the velocity. The index k on the velocity

denotes direction, and runs from 1 to 3 for a 3D calculation.
Given guessed values for the pressure, the discretized version of the
Navier-Stokes equations is:
67
ap U k, p =
anb Uk, nb + Bu
nb
A P
---------
k
k

(6.2.3)
The convective and diffusive terms have been discretized as described

in Chapter 3. The variable B contains the rest of the terms besides the
convective term, the diffusive term and the pressure term. In the pressure term, Ak is the surface area on the cell wall in direction k, and is
an index for the grid, described in Chapter 5. Looking at a pressure difference between two neighbour cells, will be unity.
The discretized version of the Navier-Stokes equations based on the
corrected variables can be written as:
a p U k, p =
anb Uk, nb + Bu
nb
P
A k ------

(6.2.4)
If this equation is subtracted from Equation 6.2.3, and the two Equations
6.2.2 and 6.2.3 are used, the following equation can be made for the
velocity correction in cell P:
Exam suggestions from
students on what SIMPLE
stands for:
Semi-Implicit Multiple Pressure Loss Equation
Surface IMplied Pressure
Level Elements
Semi-Imperfect Pressure
Link Estimation
A k P
U k = ----- --------
a p
(6.2.5)
A simplification has then been made to neglect the first term on the right
side of Eq. 6.2.4. The SIMPLEC method instead uses the following formula:
Ak
P
U k = ------------------------------- --------

p nb
nb
(6.2.6)
Equations 6.2.5 and 6.2.6 give the velocity-corrections once the pressure-corrections are known. To obtain the pressure-corrections, the continuity equation is used for cell P, where the water fluxes through each
cell side are summed up:
Ak Uk
nb
The term
Ak Uk + Ak U k
*
nb
= 0
(k=1,2,3)
(6.2.7)
nb
Ak U k is equal to the water continuity deficit in each cell,

*
nb
from the previous iteration. We denote this expression to be V.

The expression for the velocity correction from Equation 6.2.5 is inserted
into Equation 6.2.7, eliminating it as unknown. Summing up over each
side of the cell, the pressure correction gradient can be discretized for
one side as follows:
68
A k P
Ak
Side east: A k U k = A k ----- -------- = ------ ( P p P e )
a p
ap
(6.2.8)
Using this formula for four sides in a 2D situation, Eq. 6.2.7 can be written:
2
Aw
A
------ ( P w P p ) + -----s- ( P s P p ) +
ap
ap
2
(6.2.9)
Ae
A
------ ( P e P p ) + -----n- ( P n P p ) + V = 0
ap
ap
The result is an equation where only the pressure-correction is unknown:
anb Pnb + b
a p P p =
(6.2.10)
nb
The index 0 is used to indicate the new set of a0nb coefficients. The
source term, b, in Eq. 6.2.10 will be the water continuity deficit V from
the guessed velocity field. When water continuity is satisfied, this term is
zero, and there are no more corrections to the pressure.
The following formula is derived for a0e:
2
Ae
a e = -------a p, e
(6.2.11)
A similar equation holds for the other a0nb coefficients. The index e is
then replaced by w, n, s, t or b. The ap,e factor is the average ap value in
cell p and cell e.
Equation 6.2.10 is solved in the same way as the other equations.
The procedure is therefore:
Guess a pressure field, P*
Calculate the velocity U* by solving Equation 6.2.3
Solve equation 6.2.10 and obtain the pressure-correction, P
Correct the pressure by adding P to P*
Correct the velocities U* with U using equation 6.2.5.
Iterate from point 2 to convergence
An equation for the pressure is not solved directly, only an equation for
the pressure-correction. The pressure is obtained by accumulative addition of the pressure-correction values.
The SIMPLE method can give instabilities when calculating the pressure
field. Therefore, the pressure-correction is often multiplied with a number
below unity before being added to the pressure. The number is a relaxation coefficient. The value 0.2 is often used. The optimum factor depend
on the flow situation and can be changed to give better convergence
rates. Relaxation coefficients are further described in Chapter 6.5.
Regarding the difference between the SIMPLE and the SIMPLEC
69
The book Numerical Heat

Transfer and Fluid Flow,
by S. V. Patankar, is one
of the most readable texts
on CFD, and provides an
excellent introduction to
this science.
Patankar taught CFD at
the Norwegian University
of Science and Technology in 1977.
method, the SIMPLEC should be more consistent in theory, as a more

correct formula is used. Looking at Equations 6.2.5 and 6.2.6, the SIMPLE method will give a smaller correction than the SIMPLEC method, as
the denominator will be larger. The SIMPLE method will therefore move
slower towards convergence than the SIMPLEC method. If there are
problems with instabilities, this can be an advantage.
A more detailed description of the SIMPLE method is given by Patankar
(1980).
Most pressure-correction methods for incompressible flow follows algorithms similar to SIMPLE. There are algorithms involving more correction
steps, for example SIMPLER and PISO. Note that the different method
will only affect the convergence speed and the stability of the solution.
The accuracy of the results will not be directly affected, as long as the
methods are based on water continuity.
6.3 Advanced turbulence models

The following chapter is a brief overview of advanced turbulence models. The reader is referred to White (1974) and Rodi (1980) for more
detailed description of turbulence and turbulence models.
In chapter 6.1, the Boussinesq approximation was introduced for finding
an expression for the Reynolds stress term:
The kinematic viscosity is
a fluid property, while the
turbulent eddy-viscosity
depends of the velocity
field.
U U
2
u i u j = T ---------i + ---------j --- k ij
x
3
x
j
i
(6.3.1)
T is the turbulent eddy viscosity.

Some simpler turbulence models were described in Chapter 3. These
models require calibration before being used on new cases. They are
also based on algebraic relations, and no differential equations are
solved when computing the eddy-viscosity. The models are then often
called zero-equation models.
A more advanced approach is to use more complex methods to computed the eddy-viscosity. One option is to use a partial differential equation for T. One of the more popular approaches is the Spalart-Allmaras
model:
The Spallart-Allmaras model
The model (Spallart and Allmaras, 1994) is essentially a convection-diffusion equation for the eddy-viscosity, where different terms are included
to take special physical phenomena into account.
T
T
T 2
T
T
1
--------- + U j --------- = c b1 S T + --- ----- T --------- + c b2 T --------- c w1 f w ----x j
x
x j
x j
d
t
(6.3.2)
The first term on the right side is the production of turbulence. There are
several ways this can be modelled. One is:
70
U j U U
S = --------- ---------j + ---------i
x i x i x j
(6.3.3)
The next term on the right side of Eq. 6.3.2 is the diffusion of turbulence.
The last term is related to damping of turbulence close to the wall. The
distance to the wall is given as d. Spalart and Allmaras suggested the
following function for fw:
fw = g
1--6
1 + c w3 6
------------------6
6
g + c w3
g = r + c w2 ( r r )
T
r = -------------2 2
S d
(6.3.4)
The remaining parameters are constants:

cb1 = 0.1355, cb2 = 0.622, = 2/3, cw1 = 3.28, cw2 = 2, cw3 = 0.3, =0.4
The k- model
Instead of solving only one equation for the eddy-viscosity, it is possible
to use two partial differential equations. The most popular two-equation
model is the k- model (Jones and Launder, 1973). The k- model computes the eddy-viscosity as:
2
k
T = c ----
(6.3.5)
k is turbulent kinetic energy, defined by:
1
k --- u i u i
2
(6.3.6)
k is modelled as:
k
T k
k
----- + U j ------- = ------- ----- ------- + P k
x j
x j k x j
t
(6.3.7)
where Pk is the production of turbulence, given by:
U j U U
P k = T --------- ---------j + ---------i
x i x i x j
(6.3.8)
The dissipation of k is denoted , and modelled as:

2
T
----- + U j ------- = ------- ----- ------- + C 1 -- Pk + C 2 ----x j

x j x j
k
k
t
(6.3.9)
The constants in the k- model have the following standard values:
71
The main advantage of

the k- model is the almost
universal constants. The
model can thereby be
used on a number of various flow situations without calibration. For river
engineering this may not
always be the case,
because when friction
along the bed is influencing the flow field, the
roughness of the bed also
needs to be given. If the
roughness can not be
obtained from direct
measurements, it has to
be calibrated with measurements of the velocity.
c = 0.09
C1 = 1.44
C2 = 1.92
= 1.0
= 1.3
(6.3.10)
As seen from Equation 6.3.1, the eddy-viscosity is isotropic, and modelled as an average for all three directions. Schall (1983) investigated
the eddy-viscosity in a laboratory flume in three directions. His work
shows that the eddy-viscosity in the streamwise direction is almost one
magnitude greater than in the cross-streamwise direction. A better turbulence model could therefore give more accurate results for many cases.
More advanced turbulence models
To be able to model non-isotropic turbulence, a more accurate representation of the Reynolds stress is needed. Instead of using the Boussinesq
approximation (Equation 6.3.1), the Reynolds stress can be modelled
with all terms:
uu vu wu
u i u j = uv vv wv
(6.3.11)
uw vw ww
The following notation is used: u is the fluctuating velocity in direction 1,
v is the fluctuating velocity in direction 2 and w is the fluctuating velocity
in direction 3.
The nine terms shown on the right hand side of Equation 6.3.8 can be
condensed into six different terms, as the matrix is symmetrical. A Reynolds stress model will solve an equation for each of the six unknown
terms. Usually, differential equations for each term are solved. This
means that six differential equations are solved compared with two for
the k- model. It means added complexity and computational time.
An alternative is to use an Algebraic Stress Model (ASM), where algebraic expressions for the various terms are used. It is also possible to
combine the k- model with an ASM to obtain non-isotropic eddy viscosity (Rodi, 1980).
An even more advanced method is to resolve the larger eddies with a
very fine grid, and use a turbulence model only for the smaller scales.
This is called Large-Eddy Simulation (LES). If the grid is so fine that subgrid eddies do not exist because they are dissipated by the kinematic
viscosity, the method is called a Direct Solution (DS) of the NavierStokes equations.
Note that both LES and especially DS modelling require extreme computational resources, presently not available for engineering purposes.
6.4 Boundary conditions

Boundary conditions for the Navier-Stokes equations are in many ways
similar to the solution of the convection-diffusion equation. In the following text, a division in four parts is made: Inflow, outflow, water surface
and bed/wall.
72
Inflow
Dirichlet boundary conditions have to be given at the inflow boundary.
This is relatively straightforward for the velocities. Usually it is more difficult to specify the turbulence. It is then possible to use a simple turbulence model, like Equation 3.4.1 to specify the eddy-viscosity. Given the
velocity, it is also possible to estimate the shear stress at the entrance
bed. Then the turbulent kinetic energy k at the inflow bed is determined
by the following equation:
k = ----------- c
(6.4.1)
This equation is based on equilibrium between production and dissipation of turbulence at the bed cell.
Given the eddy-viscosity and k at the bed, Equation 6.3.2 gives the value
of at the bed. If k is assumed to vary linearly from the bed to the surface, with for example half the bed value at the surface, Equation 6.3.2
can be used together with the profile of the eddy-viscosity to calculate
the vertical distribution of .
Outflow
Zero gradient boundary conditions can be used at outflow boundaries for
all variables. A boundary condition where the gradient is specified is
often called a von Neumann condition.
Water surface
Zero gradient boundary conditions are often used for . The turbulent
kinetic energy, k, can set to zero. Rodi (1980) gives an alternative
expression for computing k at the water surface. Symmetrical boundary
conditions are used for the water velocity, meaning zero gradient boundary conditions are used for the velocities in the horizontal directions. The
velocity in the vertical direction is calculated from the criteria of zero
water flux across the water surface.
Bed/wall
The flux through the bed/wall is zero, so no boundary conditions are
given. However, the flow gradient towards the wall is very steep, and it
would require a significant number of grid cells to dissolve the gradient
sufficiently. Instead, a wall law is used, transformed by integrating it over
the cell closest to the bed. Using a wall law for rough boundaries (Schlichting, 1980)
1
30y
U---= --- ln ---------
ks
u*
(6.4.2)
also takes the effect of the roughness, ks, on the wall into account. The
velocity is denoted U, u* is the shear velocity, is a coefficient equal to
0.4 and y is the distance from the wall to the centre of the cell.
The wall law is used both for the velocities and the turbulence parameters. The use on turbulence parameters is described in more detail by
Rodi (1980). For the velocities, the wall shear stress is a force on a cell,
and it is computed as a sink term in the Navier-Stokes equation. The
73
force is computed by rearranging Eq. 6.4.2:
U
2
F = A = u* A = A --------------------------
--1- ln 30 ---y-
k s
(6.4.3)
The area of the cell at the bed is denoted A, while y is the distance from
the center of the cell to the bed. The velocity in the bed cell is denoted U.
6.5 Stability and convergence

The solution method described previously are iterative. The principle is
to guess a starting value for the variables and then iterate to get a better
solution. The procedure is illustrated in Fig. 6.5.1.:
Figure 6.5.1 Convergence graph for the
velocity in one cell.
Velocity
Correct value
guessed
value
Iterations
Convergence criteria
In an iterative procedure, some criteria has to be met to decide if the
solution is converged. Several different criteria exist, based on computation of a residual. The residual is a measure of how large the deviation is
between the correct value and the values in the current iteration. A low
residual indicates that convergence is reached.
One formula for the residual, r, is given in Eq. 6.5.1:
n
ap Up anb Unb Sourcep

1
r = ------------------------------------------------------------------------------nF c U c
(6.5.1)
Fc is a characteristic flux, and Uc is a characteristic velocity. The values

at the inflow boundary are often used. The total number of cells in the
grid is denoted n.
Example: We are looking at one 2D cell with the following parameters:
aw = 2.0, Uw = 1.0
ae = 0.5, Ue = 0.3
an = 1.0, Un = 0.4
as = 0.3, Us = 0.7
74
For simplicity, we assume that the source term is zero.

The first time cell p is computed, the following values are obtained:
ap = 3.8. Up = 0.73
The velocity in cell P is computed from Eq. 5.2.2, where the concentration is replaced by the velocity.
After the cell is computed, the east and north cells are recomputed and
get different values. They now have values:
Ue = 0.35
Un = 0.45
The contribution of this cell to the residual of Eq. 6.5.1 becomes:
a p U p a nb U nb = 3.8x0.73 2x1 + 0.5x0.35 + 1x0.45 + 0.3x0.7

= 0.075
Note that the values from the previous iteration is not used.
Another convergence criteria is based on the difference in the values
between two iterations:
n
Ui Ui 1
1
r = ------------------------------nU c
(6.5.2)
The disadvantage with using Eq. 6.5.2 is that the residual can go to zero
even if the solution is not converged. This is illustrated in Fig. 6.5.2:
velocity in one cell.
The correct value is not
obtained, but still the
new values in the iterations are the same as in
the previous iteration.
Velocity
Correct value
guessed
value
Iterations
The reason for this can be that the velocity-correction equation from the
SIMPLE method may change the velocities back to what they were
before the application of the solver. Also, possible bugs in the program
can give the same problem. Many CFD programs therefore prefer to use
Eq. 6.5.1 instead of Eq. 6.5.2.
Instabilities
It is not always that the iterative solution method is successful in obtaining convergence. The system of equations may be unstable. The con-
75
vergence graph for one cell may then look like what is given in Fig. 6.5.3.
The values may oscillate, and more and more extreme values are produced. Often, one defines a solution crash as the residuals becoming
above a high value, for example 1010.
There are several methods to prevent instabilities and accelerate convergence. Some are further described in the following.

velocity in one cell,
where there are instabilities and non-convergence..
Velocity
Correct value
guessed
value
Iterations
Relaxation
The main principle in the solution of the equations are to obtain an
improvement of a guessed velocity field. Starting with the guessed values, several iterations are done to improve the result. For each iteration,
a new guess is made. Let us say that we have finished iteration i-1 and i,
and we are looking at what variables, v, we should use when starting
iteration i+1. An obvious choice is of course the variables at iteration i.
However, introducing the relaxation coefficient, r, instead we use:
v = r * vi + (1-r) * vi-1
(6.5.3)
The relaxation coefficients should normally be between 0 and 1.

Relaxation will give a slower convergence speed towards the final solution, but there will be less instabilities. If the solution diverges or does not
converge because of instabilities, a normal measure is to lower the
relaxation coefficients.
Multigrid and block-correction
The purpose of the multigrid methods is to speed up the convergence of
the solution. The main principle is a division of the grid several coarser
sub-grids. This is shown in the figure below:
76
Figure 6.5.4 Grid structure for multi-grid

method
Coarse grid
Fine grid
The solution is first iterated once on the coarse grid. It is then extrapolated to the finer grid, where it is iterated further. Then it is extrapolated
to the finest grid, where more iterations are done. The solution is interpolated to the medium fine grid, where more iterations are done. Then this
is interpolated to the coarse grid, and computed again with more iterations. The sequence is repeated until convergence. There are transformation functions moving the variables from one grid to the other by
extrapolation/interpolation. When the finest grid has a high amount of
cells, it is possible to have several grids with varying number of grid
cells.
A version of the multigrid method is called block-correction. For a twodimensional situation, the grids then look like this:
Figure 6.5.5 Grid structure for multiblock method. Original grid to the left.
The iterations are started on the original grid. Then all variables are
summed in a slice of the grid, so that a one-dimensional grid emerges.
This is solved, and the result is used to correct the original values. This
is repeated in all directions, shown here with two coarse grids, for a twodimensional situation.
The Rhie and Chow interpolation
Using a non-staggered variable location, all variables are calculated in
the centre of the cells. This causes oscillations in the solution and instabilities. The staggered grid was invented to avoid these oscillations.
Then the pressure is calculated between the centres of the grid cells.
There are several problems with this variable arrangement, especially
for non-orthogonal grids. The Rhie and Chow interpolation was invented
to avoid the instabilities and still use a non-staggered grid. The interpolation gives the velocity on the cell surface. This velocity is used to calculate the flux on the cell surface.
A derivation of the Rhie and Chow interpolation procedure is fairly
involved, and the reader is referred to Rhie and Chow (1983). The main
idea is to use information about pressure gradients in staggered and
non-staggered positions. The resulting interpolation formula is a function
of the linearly interpolated velocity plus a term dependent on the pressure gradients, cell areas and the ap coefficient.
The Rhie and Chow interpolation can be interpreted as an addition of 4th
order artificial diffusion. However, no adjustment coefficients are used.
77
The Rhie and Chow interpolation is used in most CFD programs using
the finite volume method, the SIMPLE algorithm and non-staggered
grids.
Note that the Rhie and Chow interpolation has given problems for some
cases where there are large source terms in the Navier-Stokes equations (Olsen and Kjellesvig, 1998b). In these cases there are significant
forces on the water in addition to the pressure, for example from gravity.
Upstream methods and artificial diffusion
The discretization schemes given in Chapter 5 are all fairly stable for the
calculation of sediment concentration. However, other schemes developed earlier were not so stable. The classical example is the central-difference scheme. In this method, the flux on a cell wall is calculated by
interpolation between the cells on the two sides. The figure below shows
the estimation.
Figure 6.5.6 Estimation of concentration for
the central scheme
cP
Concentration
cW
cW
Cell w
Cell p
The fluxes through the sides are then:
Aw ( cw cp )
F w = 0.5U w A w ( c w + c p ) + w ---------------------------dx
Ae ( cp ce )
F e = 0.5U e A e ( c p + c e ) + e --------------------------dx
As ( cp cs )
F s = 0.5U s A s ( c p + c s ) + s -------------------------dy
An ( cn cp )
F n = 0.5U n A n ( c n + c p ) + n --------------------------dy
(6.5.2)
(6.5.3)
(6.5.4)
(6.5.5)
Applying the continuity equation (Eq. 5.3.7), and the same method as in
Chapter 5.3, the coefficients become:
Aw
a w = 0.5U w A w + w -----dx
Ae
a e = 0.5U e A e + e -----dx
An
a n = 0.5U n A n + n -----dy
(6.5.6)
(6.5.7)
(6.5.8)
78
A
a s = 0.5U s A s + s ------s
dy
Aw
Ae
a p = 0.5U w A w + w ------ + 0.5U e A e + e ------ +
dx
dx
As
An
0.5U s A s + s ------ 0.5U n A n + n -----dy
dy
(6.5.9)
(6.5.10)
Applying continuity, the following simplification can be done for ap:
A
A
A
A
a p = w ------w + e -----e- + s ------s + n -----ndx
dx
dy
dy
(6.5.11)
Looking at for example ae, if the diffusivity is low compared with the
velocity, there is a chance that ae can become negative. Also, the effective weighting factor is actually ae/ap. When the diffusion becomes small,
the effective factor becomes very large, as ap is only a function of the diffusion. A large negative number for the weighing factor is not physically
realistic, and this causes instabilities.
The minimum value of the diffusivity before instability occur can be calculated by setting the weighting factor to zero.
A
a e = 0.5U e A e + e -----e- = 0
dx
(6.5.12)
This gives the following theoretical minimum viscosity to avoid instabilities:

Artificial/false diffusion:
Many people confuse the
difference between artificial
and false diffusion. Artificial
diffusion can be seen as a
kind of fudging factor, to get
a stable solution. False diffusion is due to inaccurate
approximation in the discretization method.
e, min = 0.5U e dx
(6.5.13)
Schemes based on the central-difference scheme or similar ill-formulated numerical schemes may require adding extra diffusivity to the solution in order to become stable. This is called artificial diffusion, and
comes in addition to the physical diffusivity. The disadvantage with adding artificial diffusivity is that the increased diffusivity it may give a different final result than what the natural diffusion would give.
6.6 Free surface algorithms

The ability to compute flow with a free surface is important in hydraulic
engineering. The free surface algorithms can be classified according to
how many dimensions are used. For two-dimensional depth-averaged
computations, similar algorithms as described in Chapter 3 can be used.
However, there exist a large number of different algoritms for computing
the free surface in 2D.
For computing the free surface in 3D, the different algorithms can be
classified according to if an adaptive grid is used or not.
Fixed grid algorithms
The fixed grid algorithms will in general compute a two-phase flow. The
two phases will be water and air. The algorithms must determine where
the location of the free surface is within the grid. Some cells will be completely filled with water, and others completely filled with air. The remaining cells will be partially filled with water and air.
79
One of the most commonly used algorithms is called a Volume Of Fluid

(VOF) method. The method introduces a variable called the volume of
fluid, defined as:
Vw F = -------------------Va + Vw
Vw is the volume of water in a cell and Va is the volume of air in the cell.
The parameter r will therefore be 1 when the cell is completely filled with
water and 0 if a cell is completely filled with air.
The VOF ratio is computed by solving a convection-diffusion equation:
F
F
------ + U ------- = ------- -------
x
x
x
t
i i
i
i
(6.6.2)
Based on the F values in all the cells, the location of the free surface
must be determined. The reconstruction of the surface is not trivial, and
there are several different methods that can be used.
The VOF method is used in Flow-3D.
A more recent method that has attracted attention in research communities is the Level Set method. Instead of solving an equation for the volume of fluid, an equation for the distance, L [m], to the water surface is
used. A convection equation for this distance is given as:
L
L
------ + U ------- = 0
t
i x i
(6.6.2)
The equation is solved with similar methods as given in Chapter 5, as

this is a convection-diffusion equation where the diffusive term is omitted.
The advantage of using the level set method instead of the volume of
fluid method is that it is easier to compute the location of the free water
surface once the equations are solved.
Adaptive grid algorithms
Adaptive grid algorithms will change the grid so that the free surface
always is aligned with its top. All the cells will thereby always be filled
with water. No cells are thereby wasted by filling it with air. The method
therefore needs less cells than the fixed grid algorithms. Another advantage is that inaccuracies can occur when the cells are partially filled with
water. Also, the grid close to the surface will be aligned with the flow,
reducing false diffusion.
A disadvantage with the adaptive grid methods is that they may be more
unstable than the fixed grid methods.
In an adaptive grid algorithm, the free surface is given an initial value.
The algorithms will compute changes in the free surface, and adjust the
grid accordingly. The adjustments are done in small steps, to prevent
instabilities.
The adaptive grid methods can be further classified from which equations are solved to compute the changes in the water levels. One
80
method uses the water continuity equation in the cells close to the surface (Olsen and Kjellesvig, 1998). Normally, the water continuity defect
is used in the SIMPLE algorithm to compute the pressure. Instead, the
pressure is computed by linear interpolation between the cell below and
the surface. This approach also introduces the gravity in the NavierStokes equations. The gravity is a large source term, causing instabilities in the solution. Therefore, a very short time step has to be used. An
example is given in Fig. 6.6.1, where the coefficient of discharge for a
spillway is computed.
5 ms
0.5 sec.
50 ms
93 sec.
Figure 6.6.1 Longitudinal profile of the water level and velocities for computation of coefficient of discharge for a spillway. The numbers show the computed time.
The main problems with the method is the stability and the need for
extremely short time steps. When modelling a river over several weeks,
it is necessary to use longer time steps. Then the gravity can not be
used in the Navier-Stokes equation. The alternative is to use the Energy
equation instead of the continuity equation to compute the changes in
the free surface. Then the computed pressure field is used to estimate
the location of the surface, according to the following equation:
dpdh = ----:
g
(6.6.3)
It is then assumed that one location in the grid is kept at a known elevation. The elevation difference, dh, between this location and another cell
in the grid can be computed from Eq. 6.6.3, given the pressure difference dp between the cells. The method is very stable and can be used
with large time steps. However, the water close to the surface must have
a hydrostatic pressure for Eq. 6.6.3 to be valid. Therefore, very steep
surface slopes, like in Fig. 6.6.1 can not be computed. The location in
natural rivers and channels can be computed.
6.7 Errors and uncertainty in CFD

As shown in previous chapters, there are a number of uncertainties in
CFD modelling, and approximations in the algorithms leads to some
81
errors in the results. The European Research Community on Flow, Turbulence and Combustion (ERCOFTAC) published Best Practice Guidelines for CFD, where the errors are classified according to the following
list:
1. Modelling errors
2. Errors in the numerical approximations
3. Errors due to not complete convergence
4. Round-off errors
5. Errors in boundary conditions and input data
6. Human errors due to inexperience of the user
7. Bugs in the software
Modelling errors are errors introduced when modelling the real world
with a number of mathematical equations. Typical modelling errors are
using a one-dimensional formulation if there are three-dimensional
effects affecting the problem. Another example is the assumption of an
isotropic turbulent eddy-viscosity made for example in the standard k-e
model. Non-isotropic effects may affect the results in some cases.
Errors due to numerical approximations are often introduced when discretization the equations. False diffusion is a typical error in the numerical approximations.
Many times an iterative solver is used for the equations. Sometimes the
results are used even if the solution is not fully converged. This could be
the case if proper convergence criteria are not used. Also, for timedependent computations, convergence may not be reached for each
time step.
Round-off errors are due to limitations in the accuracy of the microprocessors of the computers. Most numerical programs nowadays use 64
bits floating point numbers with 12 digits accuracy, and then this is usually not a serious problem. However, earlier 32 bits programs often used
numbers with only 6 digits accuracy, and then round-off errors could be
significant.
Errors in the boundary conditions is one of the most common problems
in CFD modelling. Computing flow in complex geometries the grid has to
follow the water level and river bed completely. This is sometimes difficult. Also there may be problems with deciding boundary conditions for
example for the roughness. Inflow boundary conditions are also uncertain. This applies for the distribution of the velocity at the inlet cross-section and the turbulence there. For sediment computations, the amount of
sediment inflow may be uncertain. Also, the empirical formula for sediment concentration close to the bed is often not very accurate.
Human errors due to inexperience of the user is often a likely problem.
Experience on CFD modelling is scarce, and it is easy to make errors
when choosing among different parameters and algorithms in the CFD
model.
There will always be bugs in every software. An estimate often used is
one bug pr. 1000-10 000 lines for a commercial program. A typical CFD
program may have 100 000 - 1 million lines of code. It is therefore likely
that most CFD program has a fair number of bugs. A CFD program is
usually improved relatively frequently. Every time a new algorithm is
made in the program, it is possible that bugs are introduced. This may be
problematic to detect, as it is often difficult to predict how the new algorithm with interact with the older algorithms.
82
6.8 SSIIM
There exist a large number

of CFD programs. Some are
tailor-made for hydraulic
engineering, for example
TELEMAC, SSIIM and
SMS/TABS (made by US
Army Corps of Engineers).
Others are general-purpose
programs, that can be used
for gas, oil, multiphase flow
etc. Some of the most used
general-purpose programs
are PHOENICS, FLUENT,
CFX, FLOW3D, and STARCD. Web addresses to more
information about the programs can be found at:
folk.ntnu.no/nilsol/cfd
SSIIM is an abbreviation for Sediment Simulation In Intakes with Multiblock option. The program solves the Navier-Stokes equations in a
three-dimensional non-orthogonal grid, using the k- turbulence model
and the SIMPLE method to compute the pressure. The program also
solves convection-diffusion equations for various water quality constituents, like sediments, temperature, algae, nutrients, pollutants etc. Timedependent changes in bed and surface levels are computed.
The program was originally designed to compute sediment transport for
hydropower intakes. Later it has been expanded to areas of river morphology, hydraulic structures like spillways, head loss in contractions
etc.; general water quality, density stratification, wind-induced currents,
special algae algorithms etc.
The program has a graphical user interface with an interactive grid editor, containing several algorithms simplifying the constructions of the
grid. The main program contains graphical presentation of results in multiple dimensions, which can be run simultaneously with the solution of
the differential equations. A separate program is included to view
coloured surfaces in three dimensions. This is based on the OpenGL
graphics library.
The program runs on Windows, and can be down loaded from the Internet: folk.ntnu.no/nilsol/ssiimwin. The Users Manual gives more details,
and can be downloaded from the same web page.
6.9 Problems
Problem 1. Navier-Stokes solver
Apply the SSIIM model to the example with the sand trap: Tutorial 1 in
the Users Manual. Note how many iterations are required for convergence of the Navier-Stokes equations, and how long time it takes on
your PC.
Problem 2. Multigrid
Repeat the calculation in Problem 1, but this time use block-correction
for all equations. How many iterations are now needed, and how long
time does this take?
Problem 3. Relaxation coefficients
Repeat the calculation as in Problem 1, but this time change the relaxation factors first to 1.0 for all the equations. How many iterations are
needed?
Again, change the relaxation coefficients to 0.5 for all equations. How
many iterations are needed?
Change the relaxation coefficients to 0.3 for velocity, 0.1 for pressure
and 0.2 for k and . How many iterations are needed?
Problem 4. The Rhie and Chow interpolation
Repeat the calculation in Problem 1, but this time reduce the influence of
the Rhie and Chow interpolation by the using the F 21 data set. Set this
to 0.5 for one run and 0.0 for the next run. How does the Rhie and Chow
83
interpolation affect the resulting trap efficiency? And how many iterations
are needed?
Problem 5. Upstream boundary conditions
Repeat the calculation in Problem 1, but this time use uniform upstream
and downstream velocity profile. This is done by using the G 7 data sets.
How does this affect the convergence rate and the resulting trap efficiency? What are the reasons for the change?
Problem 6. Bed roughness
Repeat the calculation in Problem 1 twice, using a roughness of 2 cm
and 0.1 mm. This is done by giving the roughness in the F 16 data set.
How does the roughness affect the trap efficiency and calculation time?
What is the reason for the change?
Problem 7. Initial values
Calculate the water flow for the fish farm tank example with SSIIM. How
many iterations are needed? Remove the G 8 data set with initial velocities in the control file. What happens in the calculations and why? How is
it possible to make the calculations converge without lowering the initial
water velocities?
Problem 8. Stability
Implement the central scheme for Problem 1. Vary the amount of diffusion, and see how the result changes. What is the minimum amount of
diffusion to get a stable solution? How does this compare with the real
diffusion?
84
7. Physical limnology
7.1 Introduction
Limnology is the science of processes in lakes, including water quality,
temperature, ice, water currents etc. A large number of words explaining
the processes has been made. Also various classification systems have
been developed. This chapter focuses on the hydraulic and temperature
processes. Biological processes and classification systems for lakes are
given in Chapter 8.
7.2 Circulation in non-stratified lakes

The term circulation is often used for water currents in a lake. Velocities
in a lake are often wind-induced. The water will then follow a circulation
pattern, moving with the wind at the water surface and a return current is
formed close to the bed. However, the term circulation is also used if the
water currents are due to inflow/outflow, and the water moves in almost
straight lines. Fig. 7.2.1 shows a typical profile of wind-induced velocity
in a shallow lake:
Figure 7.2.1. Longitudinal
profile of a lake with windinduced currents. The velocity profile is shown, where
the water close to the surface moves with the wind,
and a return current is
formed close to the bed.
Wind
Velocity profile
Return current
The magnitude of the wind-induced currents will depend on the wind

speed. The most common approach to calculate the currents is to compute a shear stress from the wind on the surface of the lake. And then
use this shear stress to calculate the currents. The wind-induced shear
is given by:
2
= c 10 a U a
(7.2.1)
The wind speed is denoted Ua, a is the air density (around 1.2 kg/m3)
and c10 is an empirical coefficient. The coefficient will be different
depending on which elevation the wind is measured. The c10 coefficient
is given for wind speeds taken 10 meters above the water surface. There
are a number of empirical formulas for c10, for example as given by
Bengtsson (1973)
c 10 = 1.1*10
(7.2.2)
If the wind persist for a long time, the water surface will not be horizontal
any more. The slope, I, can be computed from equilibrium forces on a
water element, similar to what was done for a river in Chapter 2. The
slope becomes:
85
I = --------gh
(7.2.3)
where h is the water depth.

The wind will also induce velocity gradients and turbulence in the lake.
Classical hydraulics will give the following formula for the turbulent mixing coefficient, (derivation given in Chapter 4.2):
= u * h
(7.2.4)
The shear velocity is denoted u*, h is the water depth and is an empirical coefficient. For rivers, has been found to be 0.11. This value was
used successfully for lakes by Olsen et. al. (2000) modelling a small reservoir in Wales, UK. However, when modelling Loch Leven in Scotland
(Olsen et. al. 1998), the formula gave too high diffusion.
Table of water density as a
function of temperature:
Temp (0C)
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
7.3 Temperature and stratification
Density (kg/m3) The water density in freshwater lakes and reservoirs is mainly a function
of the temperature, as long as the sediment concentration is reasonably

low. Maximum density is at 4oC, with lower densities at higher and lower
temperatures. Stratification of the lake/reservoir will therefore occur for
some vertical temperature distributions.
999.87
999.97
1000.0
999.97
999.88
999.73
999.52
999.27
998.97
998.62
998.23
997.80
997.33
996.81
996.26
995.68
The specific heat for water is 4182 W/(kg0Cs) The temperature changes
in the water close to the surface can be computed from the energy balance across the water surface. The sources/sinks of energy are:
- solar short-wave radiation
- atmospheric longwave radiation
- longwave black radiation from the water
- conduction
- evaporation
The formula for the surface flux, I, in Watt/m2, can be written according
to Chapra (1997) and Henderson-Sellers (1984):
temperature flux
I =
short-wave irradiance
IrB
long-wave irr. from atm.

long-wave irr. from water
conduction
evaporation
+ ( T air + 273 ) ( A + 0.003 e air ) ( 1 R )

( T w + 273 )
0.136c 1 ( 1 + 0.437U 2 ) ( T w T air )

0.136 ( 1 + 0.437U2 ) ( e w e air )
(7.3.1)
In the formula, Ir is the irradiance, B is a reduction factor and T is the
temperature. The irradiance terms follow Stefan-Bolzmanns law, where
is the Stefan-Bolzman's constant 5.67x10-8 W/(m2K4) and is the
emissivity of water (~0.97). The emissivity is a correction factor for the
water not being a perfect emitter of radiation. A is an empirical coefficient
between 0.5 and 0.7 and eair is the air vapour pressure. R is a reflection
86
coefficient, which usually is very small (around 0.03). The parameter c1

is Bowens coefficient (62 Pa/oC), U2 is the wind speed 2 meters above
the water surface and ew is the saturation vapour pressure at the water
surface. The subscript w denotes water and air the air.
Short-wave irradiance intensity is usually measured in
mol-photons/m2/s, or
W/m2. The conversion
between the units is:
1 mol photons/s = 0.3 W
The short-wave irradiance is from the sun, and depends on several

factors:
- Variation over the year
- Variation over the day
- Latitude of lake
- Shading by clouds
- Reflection from the water surface
All these parameters can be estimated and put into the factor B. The first
three factors can be taken from tables. The cloud shading is determined
by the weather conditions. And the reflection from the water surface
depends mostly of the solar altitude. If this is above 20o, the reflection is
less than 10 %.
The long-wave irradiance is black-body heat emission from the atmosphere and the water. The term contains attenuation from the atmosphere
and reflection.
The conduction and convection term describes physical processes at
the water surface. The processes are similar to convection and diffusion,
as described in Chapter 4. Eq. 7.3.1 gives an empirical formula for these
processes.
Evaporation/condensation at the water surface affect the temperature,
as the specific energy of a certain mass of water is different depending
on if it is in liquid or gas form. The term in Eq. 7.3.1 also gives an empirical estimate of the amount of evaporation, as it is a function of the wind
speed.
Limnological classifications
The temperature profile of a lake will vary over the year, depending on
the heat flux at the water surface. In different climates, there will be different types of stratification. The science of limnology has given several
definitions to classification of lakes and stratification layers.
In a temperate climate with warm summers and cold winters, the vertical
temperature profiles are given in Fig. 7.3.1.
Figure 7.3.1. Vertical temperature profiles for a dimitic lake

in the summer and the winter
Summer
Winter
The upper layer close to the water surface is called epilimnion. The
layer close to the bed is called hypolimnion. In deep lakes, the hypolim-
87
nion will hold a temperature of 4oC throughout the year. The layer
between the epilimnion and hypolimnion is called the metalimnion. During the summer, the vertical temperature gradient is usually large in this
layer. The thermocline is located in this layer, marking the difference
between the warm upper water and the cold water close to the lake bottom.
In temperate climates, the winter is cold and the summer is warm. The
stratification will follow Fig. 7.3.1. In the summer, the warm water close
to the surface is lighter than the cold bottom water. In the winter, the
water close to the surface is below 4oC, and is lighter than the bottom
water. These two situations give a stable water body, and the stratification prevents mixing from taking place. However, during spring and fall,
the water temperature will at some point in time be 4oC over the whole
depth of the lake. Then vertical circulation may occur. The water from
the bottom may rise to the surface, if wind-induced currents are present.
This process is called a spring/fall overturn.
The science of limnology has also provided classifications of the lakes
according to the overturns. If there is no overturn due to the surface
water being too cold the whole year, the lake is called amictic. If only
one overturn occur during the year, the lake is called monomictic. This
may be due to the lake being so warm that no winter stratification is
formed. The lake is then called warm monomictic. If the lake is so cold
that no stable summer stratification occur, and only one overturn takes
place in the summer, the lake is called cold monomictic. The cycle
described in Fig. 7.3.1 with two overturns is present in dimictic lakes.
There are also lakes with multiple overturns, called polymictic lakes.
This is due to small changes in seasonal temperature and strong winds.
Turbulence damping
When horizontal water currents occur in a lake, there will be velocity gradients producing turbulence. In a stratified lake, the turbulent eddies will
be dampened by the stratification. A formula for the damping of the turbulence is often given by the following formula for the turbulent diffusion
coefficient, (Rodi, 1980):
= 0 ( 1 + Ri )
(7.3.2)
0 is the original turbulent diffusion, when not taking the stratification into
account. Ri is the Richardson number, given by:
----g z
Ri = --- --------------2 U
------ z
(7.3.3)
The formula is often used in numerical models (Olsen and Tesaker,

1995; Olsen et. al. 1999; Olsen and Lysne, 2000)
Various values of the constants a and b are used by different reserachers. Rodi (1980) recommends the values given by Munk and Anderson
(1948):
= -0.5 and = 10.0, computing the diffusion for the velocity.
= -1.5 and = 3.33, computing the diffusion for other variables.
88
Olsen and Lysne (2000), however, found better correspondence with

measurements when using the values:
= -1.3 and = 10.0, for all variables.
More advanced methods of taking the stratification into account when
modelling turbulence is given by Rodi (1980).
7.4 Wind-induced circulation in stratified lakes

Water currents in a stratified lake will be influenced by the density variation in two ways:
1. Vertical velocities will be dampened
2. Turbulence will be dampened
Looking at a deep stratified lake in the summer, where wind-induced currents occur, the circulation pattern given in Fig. 7.4.1 will emerge:
Figure 7.4.1. Windinduced circulation in a
stratified lake. The arrows
show the direction of the
velocity. Note this is a stationary situation.
Wind
line
Thermoc
The slope of the water surface can be computed by looking at the forces
on the water body, as given in Fig. 7.4.2:
Figure 7.4.2. Forces on a
water body. The hydrostatic
pressure is denoted p, L is
the length of the reservoir,
is the shear stress and H is
the water depth.
pl
Hl
Hr
pr
The force from the wind is given as:
F = L
(7.4.1)
The force from the hydrostatic pressure is:
( Hr + Hl )
1
2 1
2
F = --- gH r --- gH r = g ----------------------- ( H r H l )
2
2
2
(7.4.2)
We define the average water depth as:
( Hr + Hl )
H = ---------------------2
(7.4.3)
89
The surface slope is given as:
Hr Hl
I = ----------------L
(7.4.4)
If Eq. 7.4.3 and Eq. 7.4.4 is inserted into Eq. 7.4.2, we obtain:
F = gHLI
(7.4.5)
Setting this equal to the wind shear stress in Eq. 7.4.1, gives the following equation for the water surface slope:
I = ---------gH
(7.4.6)
The slope of the thermocline, I, can be computed from looking at equilibrium of forces in the horizontal direction on the hypolimnion (Fig. 7.4.3)
Figure 7.4.3. Forces on

the hypolimnion. The left
side is denoted L, and the
right side denoted R. The
density difference between
1 and 2 is denoted . The
pressure at the bed is
denoted p. On the upper figure, p is equal to 1gH,
where H is the water depth.
Thermoc
line
2
pL
p1,L
p1,R
pR
1
Thermoc
line
2 hypolimnion
pL
(p1,R-p1,L)
pR
Thermoc
line
2
pL
(p1,R-p1,L)
pR
The water below the thermocline is heavier than the water above. The
density difference is denoted , and given by:
= 2 1
(7.4.7)
90
It is assumed that the shear stress between the epilimnion and the
hypolimnion is negligible. The water flow direction above and below the
thermocline is the same.
Looking at the forces on the epilimnion only, the situation will be similar
to the derivation of the water surface slope. The difference is that the
water density is replaced by the density difference, , and the wind force
is replaced by the pressure difference pR-pL. The force from the pressure difference becomes:
F = ( p R p L )H = 1 gILH :
(7.4.8)
The height from the bed to the thermocline is denoted H.

The force from the density difference is the same as Eq. 7.4.5, only the
density is replaced by the density difference, and the water depth is
replaced with H:
F = gILH
(7.4.9)
Equilibrium of forces means the forces in Eq. 7.4.8 and 7.4.9 are the
same. Assuming H/H can be approximated to be unity, the following
equation is obtained:
1
I = I ----
(7.4.10)
The negative sign comes from the fact that the thermocline slopes in the
opposite direction of the lake water surface.
The density difference is much smaller than the water density itself, so
the slope of the thermocline is orders of magnitude larger than the water
surface slope. During strong winds it may happen that the slope
becomes so large that the cold water below the thermocline reaches the
water surface.
7.5 Seiches
Assuming we have a situation like given in Fig. 7.4.1, and the wind
speed suddenly drops to zero, the thermocline and the water surface will
start to oscillate. Fig. 7.5.1 shows the movement:
Figure 7.5.1. Surface and
internal seiches in a stratified lake. The thermocline
is drawn with a line. The
slope of the water surface is
exaggerated compared to
the slope of the thermocline.
91
The movements of the water surface is called surface seiches. The

movement of the thermocline is called internal seiches.
In Fig. 7.5.1, the thermocline is drawn with a straight line. This will not be
the case, as numerical and physical models shows that the water close
to the thermocline moves more like a wave, as shown in Fig. 7.5.2:
Figure 7.5.2. Movement of the thermocline during an internal seiche.
The upper figure shows the initial situation. The middle figure shows the thermocline some time after the wind has
stopped. The lower figure shows the situation later.
Speculations:
Numerical experiments
has shown that under special conditions, an internal
seiche may diffuse
upwards to the water surface. If this would be the
case, a sudden upwelling
of water to the water surface would occur, with
reasonably high horizontal
velocity. Local observations of such phenomena
in a lake with otherwise
quiet water surface could
be interpreted in imaginative ways...
The movement of the internal seiche is associated with considerable

horizontal velocities at the thermocline.
7.6 River-induced circulation and Coriolis acceleration

A river feeding water into a lake or a reservoir will create water currents
even if there is no wind present. A particular case is an ice-covered lake,
where this will be the dominant forcing mechanism for the circulation. If
the lake is stratified, the current may form plume inside the lake. An
example is given in Fig. 7.6.1, showing the velocity pattern close to the
water surface in Lake Sperillen in Norway during winter. The plume follows the right (top) side of the lake, due to the Coriolis acceleration.
Fig. 7.6.1. Velocity vectors close to the surface in

Lake Sperillen, during the
winter. The lake is then covered with ice. The river is
flowing in from the right and
out to the left.
Coriolis
The effect of the earths rotation is most pronounced for large lakes with
stratification. However, the effect may also be present in large non-stratified lakes.
92
The Coriolis acceleration affect the water movement by the following formula:
4
a = f U = ------ sin U
T
T is the time it takes for one earth rotation, i.e. 24 hours or 86 400 sec.,
is the latitude of the lake and U is the water velocity. In the northern hemisphere, the acceleration will always be to the right. An example for a
lake in southern Norway, the latitude is 60 degrees, and f will be
1.26x10-4.
Slope of water surface
For a straight channel, the water surface slope would tilt slightly due to
the Coriolis acceleration. The cross-directional slope, Ic, would be equal
to the ratio of the Coriolis acceleration to the gravity:
fU
I c = -----g
The velocity of the current is denoted U and f is the Coriolis factor. The
bottom of the current will also get a cross-directional slope, Ib. Looking at
the cross-directional balance of forces, similar to what was done for the
thermocline, it is possible to derive a formula for Ib:
I b = ----- I c = fU
---------
g
As previously, is the water density, and is the density difference
between the water in the current and the water below.
7.7 Density currents

If the water flowing into the lake/reservoir has a different density than the
lake water, a density current is formed. The density difference can be
due to:
1. Temperature variations
2. Sediment concentration
3. Content of salt in sea water
Combinations of the effects are also observed. If the temperature
causes the density of the inflowing water to be lower than in the lake, the
current will move close to the water surface. If the density is higher than
the lake water, the current will move along the lake bed. In a stratified
lake, it is possible that the current may move down into the water body to
a temperature similar to the inflowing water.
A density current caused by high sediment concentrations is called a
turbidity current. The turbidity current can transport sediment a long
way into the reservoir, and even cause deposits in front of an intake of a
large reservoir. The process also redistributes sediments from the river
mouth to the deeper part of the lake/reservoir.
93
Figure 7.7.1. Turbidity current entering a reservoir.

The plunge point is often
visible at the water surface,
as the river water contains
more sediments and has a
different colour than the
water in the reservoir.
River
Plunge point
Reservoir
Dam
Intake
The sediment-laden river water often has a darker colour than the
cleaner reservoir water. It is therefore sometimes possible to observe
the plunge point at the water surface.
7.8 Intakes in stratified reservoirs

When water is abstracted from a linearly stratified reservoir, the situation
given in Fig. 7.8.1 may occur.
Figure 7.8.1. Water
abstraction from a stratified reservoir. The depth is
H, the discharge is Q and d
is the height of the layer with
abstracted water.
Only the water in a layer at the same level as the intake is abstracted.
The thickness, d, of the layer for a two-dimensional situation with a line
abstraction can be computed from:
qd = k 1 --N
(7.8.1)
where q is the water discharge/meter width, and k1 is an empirical constant, between 3 and 5 (Steen and Stigebrandt, 1980). N is the BruntVisl frequency, given by:
d
2
N = --g- ----- dz
(7.8.2)
If the water discharge is above a critical value, qc, then water will be
abstracted from the whole depth. The following formula is used to find qc
(Carstens, 1997):
q c = 0.32NH
(7.8.3)
94
Eq. 7.8.1 is derived theoretically from an idealized 2D situation. In practice and intake will be limited in width, and 3D effects will be important.
Steen and Stigebrandt (1980) developed formulas for a 3D situation
where the abstraction gate had a significant size, of width B and height
h:
1
--3
Q
-
d = k 3 -------------- B2 hN 2
2
(7.8.4)
It was assumed that B was much smaller than the dam length and h was
much smaller than the dam height. The constant, k3, was found by
experiments to be 0.74 (Carstens, 1997) for a gate close to the bed or
the water surface, and 1.2 for a gate midway between the water surface
and the bed (Carstens, 1997; Steen and Stigebrandt, 1980). Steen and
Stigebrandt (1980) also developed more complex formulas for the thickness of the abstraction layer for cases when the outlet size was not relatively small compared to the other dimensions.
7.9 Problems
Problem 1. Thermocline
A lake with depth 100 meters has a thermocline 7 meters below the
water surface. The temperature is 15 0C above the thermocline and 5 0C
below it. A wind from the north lasts for five days. The wind speed is 15
m/s. The lake is 5 km. long in the north-south direction. Compute the
depth of the thermocline at the northern and southern side of the lake.
Problem 2. Water intake
Water is taken from a reservoir to a treatment plant to be used for municipal water supply. The intake is located 10 meters below the water surface, with a width of 2 meters and a height of 1 meter. The discharge is 2
m3/s. The temperature gradient of the reservoir is 0.5 0C/m, and the temperature at 10 meter is 11 degrees. A pollutant is spilled on the water
surface. There is no wind at the time of spill, so on average for the whole
lake, the pollutant is only mixed in the upper 1 meter layer. At the intake,
water is abstracted from a layer with vertical magnitude h. Will this be
so large that polluted water enters the intake?
95
8. Water biology
8.1 Introduction
Water quality in lakes and reservoirs is often of interest for engineering
purposes. An example is determination of location of water intakes, to
prevent excessive pollutants in the inflowing water. Sometimes algae
accumulate on one side of a lake. What is the reason for the variation in
the spatial distribution of the algal concentration? Another example is the
assessment of the capacity of a lake/reservoir to receive waste water.
The dispersion of the various components was discussed in Chapter 4
and 5, predicting the variation of the concentration as a function of turbulence, water currents etc. The variation of a component can be computed solving the convection-diffusion equation of its concentration.
However, the concentration of a component is also a function of biochemical reactions. However, biochemical reactions can also affect the
variation of a variable. A toxic substance may change as a function of
time due to for example sunlight. Or oxygen may be consumed by bacteria or generated by algae. These biochemical reactions are discussed in
the following chapters.
8.2 Biochemical reactions

Some atomic weights in
units grams/mol:
Hydrogen: 1
Carbon: 12
Nitrogen: 14
Oxygen: 16
Stochiometry
Stochiometry is used to quantify the different components in a biological
or chemical reaction. An example is the photosynthesis, where carbon
dioxide and water are used to produce organic material and oxygen:
CO 2 + H 2 O C 6 H12 O 6 + O 2
(8.2.1)
The balanced equation can be written:
6CO 2 + 6H 2 O C 6 H 12 O 6 + 6O 2
(8.2.2)
The molecular weight of carbon is 12 g/mol and oxygen is 16 g/mol. The

weight of a carbon dioxide molecule is then 44 g/mol. The weight of the
oxygen molecule is 32 g/mol. So for example 44 grams of carbon dioxide
will produce 32 grams of oxygen.
Reaction kinetics
Reaction kinetics is used to describe how fast a biochemical process
takes place. There are a number of different rate laws describing the
processes, classified by the order of the kinetics:
The reaction coefficients, k, Zero-order kinetics:
are often given in units
1/day. Remember to convert
dc
this to 1/seconds, when
------ = k or
using computer programs
dt
where this time step is used.
First-order kinetics:
c = c 0 kt
(8.2.3)
96
dc
------ = kc or
dt
c = c0 e
kt
(8.2.4)
Second-order kinetics:
2
dc
------ = kc or
dt
1
c = c 0 ------------------1 + kc 0 t
(8.2.5)
The equations describe a situation with only one variable. For some
cases there are multiple water quality constituents. The change in one
variable may be a function of the concentration of other variables. However, often there is a limiting variable determining which process is taking place. The reaction is then dependent on this variable.
Temperature dependency
Most biochemical processes depend on temperature. For example, algal
growth will increase significantly when a lake is heated by the sun during
summer. The following formula is often use to estimate the growth
increase/decrease as a function of temperature, T:
k = k 20
( 20 T )
(8.2.6)
The reaction rate at 20 oC, k20, is often used as a basis. The parameter
is specific for the reaction. Typical values are slightly above unity.
Some examples are given in Table 8.7.1. The processes in the table are
further described in the following chapters.
Discretization
Biochemical reactions are included in the convection-diffusion equation
when computing the dispersion of a water quality parameter. This is
done by including the left side of Eq. 8.2.3-5 in the source term. If firstorder kinetics is used, the equation can be written:
c
U i ------- = ------- ------- kc
x i
x i x i
(8.2.7)
When the equation is discretized, it becomes:
ap c
= a w c w + a e c e + a n c n + a s c s kcp V p
(8.2.8)
Vp is the volume of the cell. The last term on the right side is due to the
biochemical reactions. Note that this term is negative, so that it is in principle possible to get a negative concentration. This can lead to instabilities in the solution and unphysical results. To avoid this, the same
procedure as in Chapter 5.7 can be used. The equation is then written:
( a p + kV p )c
= a w c w + ae ce + an cn + as cs
(8.2.9)
In this way, negative concentrations will be avoided.

The time term can also be added, similarly as described in Chapter 5.7.
97
8.3 Toxic compounds

Toxic compounds are chemical substances not naturally occurring in the
river/lake water, causing a hazard to humans and animals drinking the
water.
Individual types of toxics will not be described here, and the dangerous
concentrations need to be determined for each component. But a few of
the processes affecting the toxic concentration are discussed.
Sorption
Toxic substances often attach to organic and inorganic sediments, suspended in the water. When the sediments settle, some of the toxic substance will be removed. The total concentration of toxics, c, in a water
body is therefore the sum of the dissolved toxic concentration, cd, and
the concentration attached to particles, cp. The fractions of the two components as a function of the total concentration is given by:
In the spring of 1995 there
was a large flood in the rivers flowing into Lake Mjsa
in Norway. The flood caused
extensive damage to infrastructure. It was also feared
it could lead to decreased
quality of the water in the
lake, which was used for
water supply to the main cities in the area. This did not
happen, as much of the polluting compounds sorbed to
sediments and deposited at
the bottom of the lake.
c
1
F d = ----d- = -------------------c
1 + Kd cs
c
Kd cs
F p = ----p- = -------------------c
1 + Kd cs
(8.3.1)
(8.3.2)
The index d denotes the dissolved fraction and p denotes the fraction
attached to the suspended particles. The particle concentration is
denoted cs. The partition coefficient, Kd, may range between 0.0001 and
1000. The value depends on the type of toxic and the composition of the
suspended particles. There exist empirical formulas for Kd for some
types of toxics/sediments. Otherwise, the coefficient must be determined
by a laboratory analysis.
Photolysis
Translation to Norwegian:
Biodegradation: Biologisk
nedbrytning
Partition coefficient:
Partisjonskoeffisient
Reaeration: Lufting
Sorption: Sorpsjon
Sunlight may cause toxic chemicals to undergo a transformation to other

compounds. The decrease in concentration is given by first-order kinetics (Eq. 8.2.4). The reaction rate, kp, is a function of the specific chemical
and the sunlight. There exist tables for kp at the water surface. But as the
light penetrates the water body, it will be dampened. This process must
be taken into account when evaluating the effect of the photolysis. The
computation of the irradiance damping must also take into account that
each photolysis reaction requires light at a specific wavelength. The
damping computation must be specific for the particular light frequency.
Hydrolysis
Hydrolysis is a transformation of the toxic to other components, usually
by acid or bases in the water. The reaction can be computed by firstorder kinetics, and the reaction rate, kp, is usually in the order of 10-7 to
0.1/day. It will be a function of the pH of the water and the chemical composition of the toxic.
Biodegradation
In biodegradation processes, toxic substances are reduced to other
compounds by organic material. Usually, different types of bacteria are
involved in the process.
98
8.4 Limnological classifications

Depending on the nutrition inflow to a lake, there may be more or less
organic material and processes. A lake with little nutrients and organic
material is called oligotrophic. If the nutrition inflow is high and large
amount of organic material is present in the lake, it is called eutrophic.
The state in between oligotrophic and eutrophic is called mesotrophic.
A quantitative distinction between the types are often related to the
phosphorous concentration. The mesotropic lake has a concentration
between 10 and 30 mg/l. Lower phosphorous concentrations are found
in oligotrophic lakes, and higher concentrations in euthropic lakes.
The water body of a lake can be divided in several zones, depending on
the biology of the lake. The littoral zone is the shallow part of the lake
close to land, where plants may grow. Classification also depends on the
amount of light penetrating the water body. Since the algae requires light
for photosynthesis, the production of organic components only take part
in the upper zone. This is called the euphotic zone. The aphotic zone
is below. The light penetration is here so small that photosynthesis is not
possible, and only decomposition processes take place.
Water quality is defined from the concentration of various components in
the water. For example, the water should have above a certain amount
of oxygen, and the concentration of toxic components should be below a
threshold value. The component is then a water quality indicator. Certain species of algae has also been used for this purpose. The water
engineer is faced with the question of determining the concentration of
the various water quality constituents.
More details on the typical processes and water quality parameters in a
lake is given in the next chapter.
8.5 The nutrient cycle

Most computer programs for water quality models the nutrient cycle in a
river/lake using oxygen, phosphorous, nitrogen and algae/bacteria. Fig.
8.5.1 shows a schematic view of the components involved:
CBOD
Settling
Settling
Air
Algae
Org. P
Oxygen
Diss. P
Org. N
NH3
NO2
Figure 8.5.1. A model of the components in a river/lake
NO3
99
P is phosphorous, N is nitrogen and the oxygen is dissolved in the water.

Oxygen processes
Oxygen dissolved in the water is seen as a main indicator of water quality. The oxygen concentration is often measured as a fraction of the saturation concentration, or a saturation concentration deficit. Clean water
will usually have high oxygen concentrations, or close to zero concentration deficit.
One of the most important processes for life is the photosynthesis. The
process transforms carbon dioxide and water to organic substances and
oxygen. In water, the process takes place in chlorophyll of the algae.
Algal processes will be further discussed in Chapter 8.7.
Oxygen is used by plants and bacteria in respiration and consumption of
nutrients. Usually, there are many types of plants and nutrients, with different types processes taking place. It is therefore difficult to derive analytical formulas for the processes. Instead, an empirical approach is
used. A sample is taken from the water, and the oxygen consumption is
measured in the laboratory as a Biochemical Oxygen Demand (BOD).
The BOD will be dependent on many factors, and empirical coefficients
are required when modelling the process. Usually, the oxygen demand
is from organic material. Since this is based on carbon, the term CBOD
is also used. BOD for Nitrogen processes is termed NBOD, and this will
be discussed later.
The consumption of nutrients by bacteria/algae can be described by
first-order kinetics:
Note that this equation is
independent of the concentration of bacteria/
algae.
dL
------ = k d L
dt
(8.5.1)
L is the concentration of nutrients, and kd is an empirical reaction coefficient.

If dissolved oxygen is consumed completely from the water, anaerobic
conditions occur. Then new processes with different types of bacteria
will take place. Such processes may create toxic substances and fish
deaths. It is therefore important that an oxygen concentration above zero
is maintained in rivers and reservoirs/lakes.
The bacteria/algae consuming the oxygen can be floating in the water, or
it can be attached to sediments at the river/lake bed. In the last case, the
term Sediment Oxygen Demand (SOD) is often used instead of BOD.
Chapra (1997) developed relatively complex algorithms to compute
SOD.
The oxygen in the air above the water surface will be able to replenish
the river/lake water if the oxygen concentration, o, is below the saturation concentration os. A large number of formulas have been made to
quantitatively determine the reaeration. The general formula is given as:
do
------ = k a ( o s o )
dt
(8.5.2)
The oxygen saturation concentration is denoted os. This is the maximum

oxygen concentration in the water. If the concentration goes above this
value, gas bubbles are formed, and the oxygen disappears to the atmosphere. The oxygen saturation concentration depends on:
100
The development of CFD

models enables the estimation of turbulence close to
the water surface. Future
reaeration formulas will
probably be based directly
on these turbulence parameters: the turbulent kinetic
energy or the eddy-viscosity.
- water temperature
- salinity
- altitude
The value of the oxygen saturation concentration will be in the order of
10 mg/liter (Chapra, 1997). Often, the oxygen deficit, D, is modelled
instead, where D=os - o.
The reaeration coefficient, ka depends on the oxygen flux through the
water surface film. It is also influenced by the oxygen mixing by turbulence below the surface. The present empirical formulas contain indirect
parameters of the turbulent mixing, for example the water velocity. A formula for the reaeration coefficient for a river is given by OConnor and
Dobbins (1958):
0.5
U k a = 3.93 --------1.5
h
(8.5.3)
The formula is not dimensionless, and U is given in m/s and h is the

water depth in meters. The coefficient ka is given in units day-1.
For lakes, an empirical formula for ka, has been developed by Broecker
et al. (1978).
0.864U
k a = --------------------wh
(8.5.4)
Uw is the wind speed 10 meters above the water surface in m/s.

There exist a large number of other empirical formulas for the reaeration
process.
Nitrogen
NH4+ = ammonium
NH3 = ammonia
The source of nitrogen in the water is usually organic material. The

material decomposes in an ammonification process, leading to formation of ammonium (NH4+). The ammonium can react with dissolved oxygen in a nitrification process to form nitrite (NO2-). The nitritification
process can further transform the nitrite to nitrate (NO3-). Since the nitrificaion process uses oxygen, the process has a nitrogen oxygen
demand (NBOD). If not, or only small amounts of dissolved oxygen is
present, a denitrification process can take place. The nitrate is then
transformed to nitrite and to nitrogen in gaseous form. The nitrogen gas
may be lost to the atmosphere.
There are several implications of the processes on the water quality:
1. If nitrogen is the limiting nutrient for organic growth, abundant nitrogen
may cause increased eutrophication.
2. High concentrations of nitrate (above 10mgN/l) in drinking waters can
cause disease for very young children.
3. Dissolved ammonia may form ammonia gas (NH3), which may be
toxic to fish. This only happens at high temperatures and high pH in the
101
water. (above 200 C and pH above 9).

The nitrogen processes can be modelled by the equations given below,
using the following subscripts: o=organic, a=ammonium, i=nitrite and
n=nitrate.
d (N )
dt o
d (N )
dt a
d N
( )
dt i
d (N )
dt n
= ko a No
(8.5.5)
= ko a N o ka i Na
(8.5.6)
= ka i N a ki n Ni
(8.5.7)
= ki n Ni
(8.5.8)
Note that the stochiometry coefficients are not included. The equations
can be solved analytically for a simple system of one well-mixed lake/
reservoir. Often, an equation for dissolved oxygen is solved simultaneously.
Phosphorous
Phosphorous is usually the limiting nutrient for plant growth in fresh
water. The growth process is then only a function of the phosphorous
concentration. The phosphorous may be dissolved in the water or it
may be present in organic material. The organic material consume the
dissolved phosphorous and then settle to the bottom of the lake/river.
These processes will decrease the concentration of dissolved phosphorous in the water. However, when the organic material decomposes, the
phosphorous may again be released and dissolved in the water.
Formulas for depletion of dissolved phosphorous is derived from growth
formula for biological material. Stochiometry is used to determine how
much the phosphorous concentration is reduced as a function of the
organic growth.
d P
(
) = k op Porganic
d t organic
d ( PO ) = k P
4
op organic kc A
dt
(8.5.9)
(8.5.10)
Porganic is the organic phosphorous in the algae, PO4 is the dissolved

phosphorous. The decomposition rate of organic phosphorous is
denoted kop, and k is the algae growth factor. The algae concentration is
denoted cA, and the fraction of phosphorous in the algae is .
In many lakes there is relatively little phosphorous occurring naturally,
leading to small concentrations of organic material. Over the last decades there has been an increase in phosphorous inflow into many lakes
in Europe. Often, fertilization from agriculture is the cause. The
increased phosphorous loading leads to accelerated growth of organic
material. This eutrophication has negative effects on the water quality.
102
8.6 QUAL2E
The QUAL2E program is made by the US Environmental Protection
Agency. It is designed to model water quality in rivers in one dimension.
The hydraulic computation is based on a steady calculation. The convection-diffusion equation is solved for a number of water quality parameter. Biological reactions are included in the equations, together with
interaction between the parameters.
The program models a number of water quality variables: Temperature
is modelled with surface fluxes as given in Chapter 7.3. Algae is modelled, including growth, predation, settling and scour. The cycles of nitrogen, phosphorous, carbonaceous BOD and oxygen are modelled,
according to the models described in Chapter 8.5.
The program is freeware, and can be downloaded from the Internet,
including users manuals. A Windows user interface is included. Table
8.6.1 gives default values of important input parameters:
Table 8.6.1 Parameters used in QUAL2E
Process
Algal growth
Equation
number
8.7.7
Reaction
rate name
Default
reaction
rate (day1)
Max-min.
reaction
rate (day-1)
Temperature
coefficient
2.5
1.0-3.0
1.047
0.005
0.005-0.5
1.047
kd
0.0
0.0-10.0
1.047
kNO>NH4
0.0
0.0-10.0
1.047
0.0
0.0-10.0
1.024
Algal respiration rate

BOD decay
Organic nitrogen decay
Organic nitrogen settling
8.5.1
8.5.5+6
8.5.5
Ammonia oxidation
8.5.6+7
kNH4>NO2
0.0
0.0-10.0
1.083
Nitrite oxidation
8.5.7+8
kNO2>NO3
2.0
0.0-10.0
1.047
8.5.9+10
kop
0.0
0.0-10.0
1.047
Organic phosphorous decay
8.7 Phytoplankton
The most important types of plankton for water quality is free-flowing
algae. Algae are plants with one or more cells, living in water. Two
groups of algae exist:
- periphyton: algae attached to the river/lake bed
- phytoplankton: free flowing algae
The most common species of phytoplankton in freshwater can be classified in three main groups:
- Cyanobacteria
- Flagellates
103
- Diatoms
Cyanobacteria often have gas vesicles, variable in size. The buoyancy
of the algae can thereby be changed and vertical movement take place.
The main process in the algae is photosynthesis, and an appropriate
amount of light is necessary. If too little light is present, the algae will
move toward the water surface. And if the light is too strong the algae
will want to move downwards, where the turbidity of the water cause
decreased light intensity. The change in phytoplankton concentration
may be due to the rise/fall velocity together with wind-induced currents
and turbulence. Another process is algal growth and predation by zoo
plankton.
Rise/sink velocity
Kromkamp and Walsby (1990) developed formulas for cyanobacteria
buoyancy based on laboratory experiments:
The formula given by Kromkamp and Walsby was
developed from data using
an algae from the lake
Gjersjen in Norway
I
a, 1 = a, 0 + t k 1 ------------ k 2 I 24 k 3
I+K
(8.7.1)
a is the algae density, t is the time step and k1, k2, k3 and K are constants. I24 is the average irradiance over the last 24 hours. I is the irradiance, with maximum value at the water surface, and decreasing values
downward in the water body, as shading occur. Bindloss (1976) investigated the damping of the irradiance for lakes in the UK, and found the
following relationship to compute the specific light transmission coefficient, kl:
k l = 0.0086c + 0.69
(8.7.2)
where c is the algal concentration. The irradiance is dampened by a factor f, given by the following formula:
f =
k l z
(8.7.3)
The summation is over all layers with magnitude z from the surface
down to the level y.
The fall/rise velocity, w, of the algae is calculated from Stoke's equation:
2 w a
w = d g -----------------18 w
(8.7.4)
where a and w is the algal and water density, and is the kinematic
water viscosity, evaluated as:
6 0.55234 0.026668T
= 10 e
(8.7.5)
where T is the temperature in degrees Centergrade.

The flagellates also seek optimum light intensity. Instead of changing
their buoyancy, the flagellates have flagelles, enabling movement. This
is similar to fins on a fish, but much lower velocities are produced. The
following formula is often used for flagellates:
104
I opt I
w = w max ----------------
I ref
(8.7.6)
I is the actual irradiance, Iopt is the optimum irradiance, Iref is a reference

irradiance and wmax is the maximum rise/sink velocity for the flagellate.
The maximum rise/sink velocity of phytoplankton are in the order of one
meter/hour.
The Diatoms have a specific weight slightly higher than water, giving a
constant fall velocity. The density can not be changed, and the diatoms
do not have flagelles. The only way to move diatoms upwards is by turbulence.
The size of the phytoplankton are often in the order of micrometer. A
microscope is necessary to identify the different species. Certain types
of cyanobacteria - Microcystis - form groups or colonies. The rise/sink
velocity is a function of the algae buoyancy and the group diameter. A
larger group will thereby get a much higher fall/rise velocity, giving
increased efficiency in search for optimum light. Groups of up to 2 mm in
size have been observed.
Growth
Time series of phytoplankton concentrations in a lake show variations
over the year. Also, the type of algae changes. Diatoms may be dominant in the spring, followed by cyanobacteria in the summer. This all
depends on temperature, nutrients, shading, grazing by zoo plankton
etc.
The algal growth can be computed by use of first-order kinetics:
c = c0 e
kt
(8.7.7)
The growth rate coefficient, k, may be around 1.0/day for optimum conditions. The biomass will then increase by 200 % in one day, and by a factor 1000 in one week. Usually, optimum conditions do not exist, as the
algae need several nutrients to achieve maximum growth. The limiting
nutrient for Cyanobacteria and Flagellates in freshwater is most often
phosphorous. Silica is often the limiting nutrient for Diatoms.
If sufficient amount of nutrients are present, the growth can be limited by
light. Reynolds (1984) found the following formula to estimate the growth
coefficient in lakes in the UK:
f
---
b
k = a1 e
(8.7.8)
The damping factor, f, is given by Eq. 8.4.3, and a and b are constants
(0.4 and 9.0) as found by Reynolds (1984).
Modelling of phytoplankton
Phytoplankton can be computed in models having from zero to three
spatial dimensions. A zero-dimensional model assumes complete mixing
in the lake/reservoir, and predicts algae concentration over time. Onedimensional models are used for rivers and sometimes for lakes. Then
105
horizontal layers are often used, and complete mixing within each layer
is assumed. Three-dimensional CFD models have been used to predict
spatial variation of algae in lakes (Hedger et. al. 2000) and reservoirs
(Olsen et. al. 2000). The CFD model predicts velocities in all spatial
directions on a three-dimensional grid. Wind-induced circulation can be
modelled, together with effects of inflowing/outflowing water. The convection-diffusion equation for algae concentration can be solved including sink/source terms for algal growth and settling/rise velocities.
Nutrients, light and temperature can be modelled simultaneously in timedependent calculations, enabling modelling of most of the important
processes affecting the algae.
8.8 Problems
Problem 1. Completely mixed lake:
A lake with volume 1 million m3 has received phosphorous from a sewage plant for several years, with an amount of 2000 kg/day. Assume a
loss rate of 0.1/day, and compute the average phosphorous concentration in the lake.
One day an improvement of the sewage treatment plant was made,
causing the loading to decrease to 300 kg/day. How long time will it take
before the phosphorus concentration is below 20 % of the original value?
Problem 2. The Streeter-Phelps equations
In 1925, a study of the water quality in the Ohio river in the USA was
published by H. W. Streeter and Earle B. Phelps. Their paper was a
landmark for water quality modelling. Using first-order kinetics, they
modelled BOD and the oxygen deficit (DO) in the water as a function of
sewage released to the river. The following formulas were used:
dL
------ = k d L
dt
dD
------- = k a D + r oc k d L
dt
where D is the oxygen deficit: os-o. The concentration of BOD is denoted
L, kd is a reaction coefficient for consumption of oxygen and ka is a reairation coefficient. The parameter roc is the stochiometry coefficients for
how much oxygen is consumed for each unit of BOD. Assume a value
of 0.1 for the current problem.
Solve the equations using a spreadsheet for a 100 km long river with a
water discharge of 100 m3/s, where the upstream sewage outflow is 30
kg/s. Present longitudinal profiles of BOD and oxygen concentration for
a steady situation. Assume kd = 0.8 day-1. The river is 2 meters deep
and 200 meters wide.
Compare the results with the analytical solutions of the equations:
L = L0 e
k
----d x
U
106
D = D0 e
k
----a x
U
kd
ka
x
---- x
c 0 k d ----e U e U
+ --------------ka kd
Problem 3. The nitrogen cycle

Use a spreadsheet to compute the variation over time of various forms of
nitrogen in a well-mixed lake, according to Eq. 8.5.5-8.5.8. Use the reaction rates and initial concentrations given in Table 8.6.1.
Table 8.8.1 Parameters for problem 3
Reaction
rate name
Reaction
rate (day1)
Organic nitrogen decay
kNO>NH4
0.15
Organic N
0.008
Ammonium oxidation
kNH4>NO2
0.15
Nitrite
0.0
Nitrite oxidation
kNO2>NO3
0.5
Nitriate
0.008
Ammonium
0.0
Process
Variable
Initial
conc.
Problem 4. Stratified lake

A 22 meter deep lake is thermally stratified, with a 1 meter thick thermocline 7-8 meters below the water surface. It is assumed that the water
on both sides of the thermocline is well-mixed, and the mixing through
the thermocline is according to a diffusion coefficient, , equal to 1.3x102 2
m /s.
Initially, the lake is fully saturated with oxygen. Assume an oxygen concentration of 9 mg/l. It is assumed that the wind mixes oxygen into the
water, so the water above the thermocline stays saturated with oxygen.
A the lake bottom, there is a sediment oxygen demand of 0.3 (m2day)-1.
What is the concentration of oxygen below the thermocline after a long
time?
107
9. Sediment transport
9.1 Introduction
Sediments are small particles, like sand, gravel, clay and silt. The water
in a river has a natural capacity of transporting sediments, given the
velocity, depth, sediment characteristics etc. Man-made structures in a
river may change the sediment transport capacity over a longer part of
the river, or locally. Erosion may take place in connection with structures,
such as bridges, flood protection works etc. The hydraulic engineer has
to be able to assess potential scour problems. During a flood, the risk for
erosion damages is at its highest.
Sediments cause many problems when constructing hydropower plants
and irrigation projects in tropical countries. Deposition and filling of reservoirs is one problem, and the water intake has to be designed for handling the sediments. The sediments reaching the water turbine may
cause wear on the components, as shown in the picture below.
The picture shows erosion on
hydraulic machinery: the
blades leading the water
towards the turbine. Photo:
N. Olsen.
In recent years, the topic of polluted sediments has received increased

interest. Organic and toxic substances may attach to inorganic sediment
particles and affect the water quality. Erosion of old polluted sediments
may create a hazard. Sediments also affects the natural biochemistry of
shallow lakes.
The origin of sediments vary according to different climates. In tropical
countries, rock decompose naturally. Water in form of rain and temperature fluctuations, together with chemical reactions cause cracks in the
rock. The weathering cause a layer of particles to be formed above the
solid rock. The particles close to the surface have the longest exposure
to weathering, and have the smallest grain sizes. Larger stones are
formed closer to the bedrock. Over time, the finer particles are removed
by rain, causing erosion of the rock. The process is relatively slow, as
vegetation cover prevents erosion from taking place. Where man
removes the vegetation, there may be accelerated erosion.
The sediments are transported by the rivers and streams through the
catchment. The annual sediment transport in a river, divided by its catchment area is called sediment yield. Typical numbers for tropical coun-
108
The picture on the

right is taken from a
volcano in Costa
Rica. The fine material of the volcano lies
on steep slopes. It is
relatively unprotected
by vegetation. The
rain fall then causes
relatively extreme
erosion rates. Such
areas often have gullies, as shown on the
picture. Photo: N.
Olsen.
tries are around 100-2000 tons/km2/year.

Sediment yield
In some countries, like Norway, the geology is very much influenced by

the ice-age. The glaciers removed the upper layer with soil, leaving solid
bedrock for a large part of many catchments. The sediment yield is
therefore much lower, almost zero many places. The main sediment
source in Norway is from the glaciers.
109
9.2 Erosion
The initial step to the science of sediment transport in a river is looking at
forces on a sediment particle resting on the bed. The purpose is to find a
method for determining when the particle will be eroded. There are four
forces influencing the stability of the particle (Fig. 9.2.1) resting on a bed
where the water has a velocity U.
- Gravity: G
- Drag: D
- Lift: L
- Friction: F
L
U
Assuming the particle has diameter d, the forces can be written:
G = k 1 ( s w )gd
D
(9.2.1)
4
-- 2 --3- 2
2 3 2
2
D = k 2 u d = k 2 IM r d = k 2 --------- M r d k 3 d (9.2.2)
gr
2 2
G
Fig. 9.2.1 Forces on a
particle in a stream
The density of a sediment

particle is often set to 2.65
times the water density
2
2
--d----- 2
1--d----- 2 3
1--2
L = C w u = C w IM r k 4 d
2 L 4
2 L 4
(9.2.3)
The friction force, F, is a function of the force pushing the particle downwards, multiplied with a friction coefficient. This friction coefficient is the
same as tangens to the angle of repose of the material, .
3
F = ( G L ) tan ( ) = tan ( ) [ k 1 g ( s w )d k 4 d ]
(9.2.4)
Some constants are used: k1 for the shape factor of the particle, k2 is a
drag coefficient and CL is a lift coefficient. The critical shear stress on the
bed for movement of a particle is denoted c.
Equilibrium of forces along the direction of the bed gives:
F = D
(9.2.5)
Using Eq. 9.2.4 and 9.2.2, this gives

3
tan ( ) [ k 1 g ( s w )d k 4 d ] = k 3 d
(9.2.6)
The equation is solved with respect to the particle diameter:
c
c
- = ------------------------------d = ----------------------------------------------------------------k 1 tan ( )
g ( s w )*
g ( s w ) --------------------------------k 3 + k 4 tan ( )
(9.2.7)
The parameter, *, was found experimentally by Shields (1937), and can

be taken from Fig. 9.2.2:
110
Fig. 9.2.2 Shields graph,

giving the critical shear
stress for movement of a
sediment particle
*= ----------------------------g ( s w )d
The value on the horizontal axis is the particle Reynolds number, given
by:
d --u* d
R* = -------- = ----------
(9.2.8)
The viscosity of water is denoted , d is the particle diameter, u* is the

shear velocity and is the shear stress on the bed.
Shields graph can be used in two ways: If the bed shear stress, , in a
river is known, Eq. 9.2.7 can be used to determine the stone size that will
not be eroded. Or if the particle size on the bed is known, Eq. 9.2.7 can
be used to compute the critical shear stress for movement of this particle. In both cases, Eq. 9.2.7 is used, where the parameter * is found by
Fig. 9.2.2.
Cohesive sediments
under 0.1 mm
If the particle is very small, under 0.1 mm, there are also electrochemical
forces occurring. The sediments are then said to be cohesive. The critical shear stress then depend on the chemical composition of the sediments and the water.
Example: A channel with water depth 2 meters and a slope of 1/1000 is
covered with stones of size 0.03 m. Will the stones be eroded or not?
First, the bed shear is computed:
1
= ghI = 1000x9.81x2x ------------ = 20Pa
1000
Then the particle Reynolds number is computed.
20 d --0.03 ----------u* d
1000
R* = -------- = ----------- = --------------------------- = 4243
6
10
The Shields diagram gives the Shields coefficient as 0.06. The critical
shear stress for the particle is then:
*
c = g ( s w )d = 0.06x9.81x ( 2650 1000 )x0.03 = 29Pa
111
We see that the critical shear stress for the particle is above the actual
shear stress on the bed. The particle will therefore not be eroded.
Sloping bed
The decrease, K, in critical shear stress for the sediment particles as a
function of the sloping bed was given by Brooks (1963):
sin sin - 2
2
tan - 2
sin sin - + --------------------- ---------K = ---------------------+
cos
tan
tan
tan
Critical shear for sloping

banks
(9.2.9)
The angle between the flow direction and the channel direction is
denoted . The slope angle is denoted and is a slope parameter. The
factor K is calculated and multiplied with the critical shear stress for a
horizontal surface to give the effective critical shear stress for a sediment
particle.
U
Cross-section
Plan view
Fig. 9.2.3. Plan view (left) and cross-section (right) for explanation of the angles and in the formula for the decrease of the
critical shear stress on the bed.
Looking at the bank of a straight channel, where the water velocity is

aligned with the channel direction, is zero. Eq. 9.2.9 is then simplified
to:
tan
K = cos 1 -----------
tan
(9.2.10)
The slope parameter, is slightly higher than the angle of repose for the
material (Lysne, 1969). A value of 50 degrees was used by Olsen and
Kjellesvig (1999) computing bed movements in a sand trap.
More recently, Dey (2003) developed another formula for K:
0.745
0.372
1
K = 0.954 1
---
-
A very good handbook for

design of scour protection
works in Norwegian is written by Fergus et. al. (2010).
(9.2.11)
The angles and are here not defined in the same way as Brooks. The
angle is the bed slope normal to the direction of the velocity vector.
While the angle is the bed slope in the direction of the velocity vector.
Bihs and Olsen (2011) obtained fairly good results using this formula to
compute local scour around an abutement in a channel.
112
9.3 Suspended sediments and bed load

When the bed shear stress exceeds the critical value for the bed particles, there will be a sediment transport in the river. The particles will roll
along the bed or jump up into the flow. The latter process is called saltation. The length of the jump will depend on the fall velocity of the particles. Fig. 9.3.1 gives the fall velocity for quartz spheres at 20 oC.
Fig. 9.3.1 Fall velocity of

quartz spheres in water.
The horizontal axis is the
diameter of the spheres,
and the vertical axis is the
fall velocity
cm/s
mm
The sediments will move close to the bed or in suspension, depending
on the particle size and the turbulence in the water. The Hunter Rouse
parameter (Eq. 9.3.1) is often used to determine the vertical distribution
of the sediment concentration profile:
wz = -------u *
(9.3.1)
The fall velocity of the particles is denoted w, is a constant equal to 0.4

and u* is the shear velocity. High values of z indicates the fall velocity is
high compared with the turbulence. The sediments will then move close
to the bed. Low values of z indicates high amount of turbulence compared with the fall velocity, and the distribution becomes more uniform.
Hunter Rouse also developed formulas for the vertical distribution of the
concentration, c(y):
Figure 9.3.2. The vertical
distribution of sediment concentration for some chosen
values of z
h y- ----------a - z
c--------( y -) = --------- y h a
ca
(9.3.2)
The water depth is denoted h, y is the distance from the bed and a is the
distance from the bed where the reference concentration, ca, is taken.
Often a is set to 0.05h. The vertical distribution of sediment concentration for some values of z is given in Fig. 9.3.2, by using Eq. 9.3.2.
Sediment transport capacity
The river will have a certain sediment transport capacity, given its
hydraulic characteristics and the sediment particle size. Supplying more
sediments than the transport capacity leads to sedimentation, even if the
critical shear stress for the particles is exceeded. Less available sediment than the transport capacity leads to erosion.
Ralph A. Bagnold did

most of his sediment
research on sand transported by the wind. His initial field work was done in
the desert of North-Africa.
When the second world
war broke out, his knowledge of the desert was
used by a special regiment in the allied forces in
Egypt, that he commanded (Bagnold, 1990).
Hans Albert Einstein
was son of the famous
Albert
Einstein,
the
founder of the theory of
relativity. One day while
Hans Albert was at university, Einstein senior asked
his son what he intended
to choose as the topic for
his research. He then
answered the science of
sediment transport. The
father replied that he also
thought of this when he
was young, but he considered it to be too difficult.
113
Initial studies of sediment transport was done by Bagnold (1973), looking

as sand transported by wind in the desert. Bagnold divided the transport
into two modes: Bed load and suspended load. The bed load rolled
along the ground, and the suspended load was transported in the air.
For some reason, the same approach was used in water. The problem is
that there is no clear definition of the difference between the two transport modes, agreed by most researchers. Some definitions are:
1. According to Einstein
Hans Albert Einstein, was a prominent researcher on sediment transport. According to him, the bed load is transported in a distance two particle diameters from the bed, as the transport mode was by sliding or
rolling.
2. According to van Rijn
Van Rijn started from Bagnolds approach, and derived formulas for how
far the bed load particles would jump up into the flow. This distance is far
greater than predicted by Einsteins approach
3. According to measurements
Sediment transport in a river is often determined by measurements. A
water bottle is lowered into the river, and water with sediments is
extracted. The sediment concentration is determined in a laboratory. The
water bottle is not able to reach all the way down to the bed, so there will
be an unmeasured zone 2-10 cm from the bed. Often the measured sediment will be denoted suspended load, and the unmeasured load
denoted bed load.
4. According to Hunter Rouses example
Because Hunter Rouse showed an example where the reference level
for the concentration was 5% of the water depth, some people assume
the suspended load is above this level and the bed load is below.
As the definition of bed load and suspended load is not clear, it is necessary for the engineer to require further specification of the definition
when using sediment transport data where the terms are used
9.4 1D sediment transport formulas

There exist a large number of sediment transport formulas. Some of the
formulas are developed for bed load, and some for total load. All formulas contain empirical constants, so the quality of the formula depend on
the data set used to calibrate the constants. In other words, some formulas work well for steep rivers, and some for rivers with smaller slopes,
finer sediments etc. The formulas give very different result for the same
case, and there is often an order of magnitude between lowest and highest value. It is therefore difficult to know which formula to use. Different
researchers also have varying opinions and preferences as to what formula to use.
The formulas can be divided in two groups: Bedload formulas and total
load formulas. The bedload formulas are developed for data sets where
only bedload occur. When used in situations where the sediment transport is mainly suspended load, the formulas may give very inaccurate
results. The total load formulas should work for both modes of transport.
114
Sediment transport
formulas:
Engelund/Hansen
Danish researchers looking
at rivers with relatively fine
sediments and mild energy
gradient.
Ackers&White
British researchers, using
data from 925 individual
sediment transport experiments to find the constants
in their equation.
Mayer-Peter&Muller
Swiss researchers, working
mostly on rivers with steep
slopes, where most of the
material moved close to the
bed.
Commonly used formulas for total load are:

Engelund/Hansens (1967) formula:
q s = 0.05 s U
d 50
----------------------- ---------------------------------g ( s w )d 50
g -----s- 1
3
--2
(9.4.1)
The sediment transport, qs, is given in kg/s pr. m width. U is the velocity,
s is the density of the sediments, w is the density of the water, is the
shear stress on the bed, g is the acceleration of gravity and d50 is the
average sediment diameter. This version of the formula works in the SI
system of units.
The Ackers&White (1973) formula requires five steps, given in the following. Note that the logarithm in the function has base 10, log10.
1. Compute a dimensionless particle size:
D gr
s w
g -----------------
w
= d -------------------------2
1
--3
(9.4.2)
For uniform grain sizes, the mean particle diameter, d, is used. For
graded sediments, the d35 value is used.
2. Compute four parameters, m, n, A and C, to be used later:
if Dgr > 60, the particle sizes are said to be coarse:
n = 0.0
A = 0.17
m = 1.5
C = 0.025
(9.4.3)
if Dgr is less than 60, but larger than 1, the sediments are medium sized:
n = 1.0 - 0.56 log Dgr
0.23- + 0.14
A = -----------D gr
m = 9.66
---------- + 1.34
D gr
(9.4.4)
2
log C = 2.86 log D gr ( log D gr ) 3.53

if Dgr is less than 1, the sediments are under 0.04 mm. It is assumed that
cohesive forces may occur, making it difficult to predict the transport
capacity. However, the transport capacity is then usually much larger
than what is available for the river, so the sediment load is limited by the
supply.
3. The mobility number is then computed (note simplification if n=0):
115
n
u*
1n
U
F gr = ---------------------------------- --------------------------------10h
s w
32 log ---------
gd ----------------d
w
(9.4.5)
The water depth is denoted h.

4. The sediment concentration, c, is then given in weight-ppm:
s w
d -----------------
w
F gr m U n
c = -------------------------- C ------- 1 ---- u *
h
A
(9.4.6)
5. The concentration is multiplied with the water discharge (in m3/s) and
divided by 103 to get the sediment load in kg/s.
Bed load formula
Mayer-Peter&Mullers formula for bed load
A formula for bed load was given by Mayer-Peter and Mller (1948):
1 w grI 0.047g ( s w )d 50
q s = --- -------------------------------------------------------------------2
g
1----3
3 s
w
0.25 w -----------------
s
3
--2
(9.4.7)
The hydraulic radius is denoted r.

These are some of the most well-known formulas, together with Einsteins formula and Tofalettis formula. The latter two are fairly involved,
and are only used by some computer programs. There also exist a large
number of other formulas giving more or less accurate answers.
Which formula to use?
The question remains on which formula to use. Three approaches exist:

1. Some formulas work better in particular situations, for example steep
rivers etc. The problem with this approach is the difficult and inaccurate
classification of the formulas.
2. Do a measurement in the river, and use the formula that best fits the
result (Julien, 1989). The problem is the difficulty of obtaining a good
measurement.
Because of confusion on
Imperial and metric units,
and also because of misprinting, many textbooks do
not give the correct sediment transport formulas
3. Use several formulas, and choose an estimate close to the average

value.
The final approach depend on the information available and the experience and knowledge of the engineer.
116
9.5 Bed forms

Sediment particles moving on an initially flat bed may generate bed
forms. Sediments forms small bumps on the bed with regular shape and
interval. The following classification system is used for different types of
bed forms:
Four different bed forms:
The Norwegian words for
ripples are riller. Dunes are
called dyner or sandbanker. Bars are called
sandbanker or grusrer.
Antidunes are called motbanker or motdyner.
1. Ripples
2. Dunes
3. Bars
4. Antidunes
The first three types of bed forms occur in subcritical flow. Sediments
deposit on the down side of the bed form and erode from the upstream
side. The ripples are fairly small, with a height under 3 cm, and occur
only on sediment finer than 0.6 mm. The bars are much larger, with
heights similar to the water flow depth. The dunes have a size between
the bars and the ripples.
The antidunes are different in that they occur only in supercritical flow. A
hydraulic jump is formed between the bedforms. Deposition takes place
at the front of the dune, and the downstream side erodes. The bedform
itself therefore may move upstream, even though the individual grain
sizes move downstream. Antidunes may also move downstream or be
stationary.
There exist a large number of methods for prediction of bed forms, predicting both the type and size. Unfortunately, as for the sediment transport formulas, the various methods give highly different answers. An
example of a bed form predictor is given by van Rijn (1987), estimating
the bed form height, of dunes:
D 50 0.3
--- = 0.11 --------1e

h
h
c
------------2 c
c
25 -----------
c
(9.5.1)
where h is the water depth.

The bed form height can be used together with the bed sediment grain
size distribution to compute an effective hydraulic roughness (van Rijn,
1987):
25
----------
k s = 3D 90 + 1.1 1 e
(9.5.2)
where is the bedform length, calculated as 7.3 times the water depth.
The formula can be used to compute the velocities and the water surface
elevation. However, the resulting shear stress is only partially used to
move sediments. When computing the effective shear stress for the sediment transport, the shear due to the grain roughness only should be
used. This can be computed by using partition coefficients. The partition
of shear stress used to move the particles divided by the total shear
stress is denoted:
s
p = ----s = --------------
d + s
(9.5.3)
117
Here, s denotes the shear stress due to the roughness of the sediment
particles, and d denotes the shear stress due to the roughness of the
dunes.
Using several empirical formulas, the following equation can be derived
to compute p:
p = ( pr )
0.25
(9.5.4)
where pr is the partition of the roughness, given as:
k s, particles
3D 90
p r = ----------------------- = --------------------------------------------------------25
k s, total
--------

3D 90 + 1.1 1 e
Many sediment discharge

formulas are derived from
laboratory experiments. It is
very difficult, almost impossible, to scale the size of the
bed forms in the laboratory.
This may be one of the reasons why the many sediment discharge formulas
give different results.
(9.5.5)
In a laboratory experiment with movable sediments, the bed is often flattened before the experiment starts. The bedforms will grow over time,
until they get the equilibrium size. The roughness will therefore also vary
over time in this period.
9.6 CFD modelling of sediment transport

Sediment transport is often divided in two: suspended load and bed load.
The bed load can be computed with specific bed load formulas, for
example (van Rijn, 1987):
c 2.1
------------c
qb
----------------------------------------- = 0.053 ---------------------------------------------------0.1
(
)g
)g
0.3
s
w
1.5
s
w
-
D 50 ------------------------D 50 ------------------------2
w
w
(9.6.1)
The sediment particle diameter is denoted D50, is the bed shear stress,
c is the critical bed shear stress for movement of sediment particles, w
and s are the density of water and sediments, is the viscosity of the
water and g is the acceleration of gravity.
Suspended load is computed using the algorithms given in Chapter 5.
The convection-diffusion equation for suspended load is solved:
c
c
c
c
----- + U i ------- + w ----- = ------- T ------- + S
x
z
x
x
t
i
i
i
(9.6.2)
The fall velocity for the particles is denoted w. This is a negative number
if the z direction is positive upwards. S is a source term which can be
used to prescribe a pick-up flux from the bed. An alternative method to
model resuspension of sediments, is to give a a boundary condition near
the bed. The most commonly used method is to use van Rijns (1987)
formula:
118
c bed
c 1.5
------------D 50
c
= 0.015 --------- ------------------------------------------------------0.3
a
1
---
( s w )g 3
-
D 50 ------------------------2
w
(9.6.3)
The sediment particle diameter is denoted d, and a is a reference level

set equal to the roughness height
It is also possible to adjust the roughness in the computation of the water
velocities according to the computed grain size distribution at the bed
and the bed forms (Eq. 9.5.2) (Olsen, 2000).
When the sediments are prescribed in the bed cell according to Eq.
9.6.3, sediment continuity is not satisfied for this cell. There may therefore be sediment deposition or erosion. The continuity defect can be
used to change the bed. A time-dependent computation can compute
how the geometry changes as a function of erosion and deposition of
sediments.
CFD modelling of sediment transport has currently been done by a
number of researchers on many different cases: sand traps, reservoirs,
local scour, intakes, bends, meandering channels etc. However, the science is still at a research stage, and it is not yet much used in consulting
work.
9.7 Reservoirs and sediments

A river entering a water reservoir will loose its capacity to transport sediments. The water velocity decreases, together with the shear stress on
the bed. The sediments will therefore deposit in the reservoir and
decrease its volume.
In the design of a dam, it is important to assess the magnitude of sediment deposition in the reservoir. The problem can be divided in two
parts:
1. How much sediments enter the reservoir
2. What is the trap efficiency of the reservoir
In a detailed study, the sediment grain size distribution also have to be
determined for question 1. Question 2 may also involve determining the
location of the deposits and the concentration and grain size distribution
of the sediments entering the water intakes.
In general, there are two approaches to the sedimentation problem:
1. The reservoir is constructed so large that it will take a very long time
to fill. The economical value of the project will thereby be maintained.
2. The reservoir is designed relatively small and the dam gates are
constructed relatively large, so that it is possible to remove the sediments regularly by flushing. The gates are opened, lowering the
water level in the reservoir, which increases the water velocity. The
sediment transport capacity is increased, causing erosion of the
119
deposits.
A medium sized reservoir will be the least beneficial. Then it will take relatively short time to fill the reservoir, and the size is so large that only a
small part of the sediments are removed by flushing.
The flushing has to be done while the water discharge into the reservoir
is relatively high. The water will erode the deposits to a cross-stream
magnitude similar to the normal width of the river. A long and narrow reservoir will therefore be more effectively flushed than a short and wide
geometry. For the latter, the sediment deposits may remain on the sides.
The flushing of a reservoir may be investigated by physical model studies.
Another question is the location of the sediment deposits. Fig. 9.7.1
shows a longitudinal profile of a reservoir. There is a dead storage below
the lowest level the water can be withdrawn. This storage may be filled
with sediments without affecting the operation of the reservoir.
Figure 9.7.1 Longitudinal
profile of a reservoir. HRW
is the highest regulated
water level. LRW is the lowest regulated water level.
The reservoir volume below
LRW is called the dead storage, as this can not be
used.
HRW
Inflow
LRW
Dead storage
Sediment load prediction

Rough estimates of sediment
load may be taken from
regional data. Often the sediment yield in the area is
known from neighbouring
catchments. It is then possible
to assess the seriousness of
the erosion in the present
catchment and estimate rough
figures of sediment yield. The
land use, slope and size of the
catchment are important factors.
For a more detailed assessment, measurements of the
sediment concentration in the
river have to be used. Sediment concentrations are
measured using standard
sampling techniques, and
water discharges are
recorded simultaneously. The
measurements are taken at
Sediment sampler
120
varying water discharges. The values of water discharge and sediment

consternations are plotted on a graph, and a rating curve is made. This
is often on the form:
b
Q s = aQ w
(9.7.1)
Qs is the sediment load, Qw is the water discharge and a and b are constants, obtained by curve-fitting.
Fig. 9.7.2 Example of sediment rating curve. The
sediment load is on the vertical axis, and the water discharge on the horizontal
axis. The points are measured values, and the line is
Eq. 9.7.1
Note: Often, most of the
sediments are transported
during the larger floods
The annual average sediment transport is obtained by using a timeseries of the water discharge over the year together with Eq. 9.7.1.
9.8 Fluvial geomorphology

The river plays an essential role in shaping the landscape, in the process
of transporting eroded sediments to the sea. The transport mechanisms
are described differently depending on the river characteristics. In the
upper part of the catchment, the creeks will have relatively large slopes.
The transport capacity will often exceed the amount of available material, leaving bedrock or larger stones on the river bed. Closer to the sea,
the river gradient will be lower, and the river will not be able to transport
large-sized material.
Alluvial river
The regime theory was

developed by British water
engineers in India at the end
of the 19th century.
An alluvial river is formed where the bed material have sufficient magnitude to enable free vertical bed fluctuations. Also, the bed material has a
lower size than what is given by Shields curve, so there is continuous
sediment transport. The bed fluctuates depending on the sediment concentraion. If the supply of sediment is larger than the capacity, the bed
rises. If the supply is lower then the capacity, the bed is lowered. If the
bed is in equilibrium, the river is said to be in regime. There exist theory
for the relationship should be between different parameters for the channel. This is called regime theory (Blench, 1970). An example is a relationship between the water velocity, U, and the water depth, d:
U = 0.84d
0.64
(9.8.1)
The formula is given in British Imperial units, where the velocity is in feet
pr. seconds and the water depth is in feet. Also formulas exist where the
river width and slope can be computed.
121
Paving is called dekklag

in Norwegian. Meander is
called elveslyng.
Many steeper rivers are not in regime. Very little sediments may be
transported during normal and low flow. Only during large floods, the bed
material moves. These rivers often have large stones at the bed. They
are said to be paved. The paving protects the underlaying sediments.
Only during very large floods the paving will be removed, and then large
changes in the bed geometry may occur. In general, the bed changes
and shape of a river usually change mostly during large floods.
Meandering rivers
The river may also move sideways (laterally), as the classical meandering pattern evolves. As the water velocity is higher closer to the water
surface, than at the bed, a vertical pressure gradient will be formed when
the water meets an obstacle or a river bank in a bend. The result is a
downwards velocity component, causing a secondary current. (Fig.
9.8.1)
Figure 9.8.1 Cross-sectional view of a river bend,

where the arrows indicate the
secondary current
Outer bank
Erosion
Inner bank
Deposition
The flow pattern causes the sediment transported on the bed to move to
the inside of the curve. The shape of the resulting cross-section is given
in Fig. 9.8.1, with the deeper part on the outside of the curve. Over time,
sediments deposit on the inside of the curve and erosion will take place
on the outside. Looking at a plan of the river, the meanders will move
outwards and downstream. There exist different classification systems
for the plan form (Schumm et. al., 1987).
Figure 9.8.1. Definition of
sinuosity of a river. The
figure shows a plan view of
the river. The length of the
centerline is denoted C, and
the length along the valley is
denoted L. The sinuosity is
then the ratio between these
numbers: C/L.
Centerline
A river can be classified according to its sinuosity (Fig. 9.8.2). If the sinuosity is below 1.05, the river is straight. Some researchers classify a
meandering river by a sinuosity above 1.25.
122
Figure 9.8.3. Plan view of a

meandering river. The
location of the apex and
cross-over is shown. The
maximum depth is at the
outer side of the bend at the
apex.
Apex
Cross-over
Cross-over is called brekk

in Norwegian. The area at
the apex with the scour is
called kulp.
Maximum depth
Useful terminology on meandering rivers is given in Fig. 9.8.3, giving the

location of the apex and the cross-over. The secondary currents will
have maximum strength at the apex, giving the largest scouring and
maximum depth at this location, on the outside of the bend. This is also
shown in Fig. 9.8.4, giving a longitudinal profile of a meandering river
and explaining the word thalweg.
Figure 9.8.4.
Thalweg
The upper figure shows a

plan view of the meandering
river. The broken line shows
the location of the maximum
depth of the river. This is
also called the thalweg.
The lower figure shows a
longitudinal profile of the
river, with the water surface
and the bed level at the thalweg. The thalweg level
shows a pattern with highest
value at the cross-over and
lowest level at the apex.
Water
surface
Bed
level
Apex
Cross-over
There are many factors affecting the meander formation, magnitude of

the sinuosity etc. for a river: Valley slope, size of sediments, sediment
discharge and water discharge. Cohesion of the bank material also
seems to be important, as it affects the bank stability. Vegetation along
the river is then also important.
Experiments on meandering channels
A natural meandering channel often do not show regular patterns. This is
due to inhomogenities in bed material, local geology, vegetation, manmade scour protection etc. To study the meander formation, flume studies in the laboratory have been carried out. An example is large flume
studies at Colorado State University in the 1970s. (Schumm et. al.,
1987). The flume was 28 meters long, and was initially filled with sand. A
straight channel was moulded in the sand, and a water was then run
through it. After some time, a meandering pattern emerged. This case
was later computed using a CFD model (Olsen, 2003). The results are
123
shown in Fig. 9.8.5, 9.8.6 and 9.8.7.

Figure 9.8.5 Plan view of
velocity vectors in a
meandering river, modelled with CFD. The black
arrows shows the vectors
at the water surface, and
grey arrows close to the
bed.
Note the velocity vectors close to the bed points more towards the inside
of the curve than the vectors at the surface. This corresponds with Fig.
9.8.1.
Figure 9.8.6 Plan view of

water depths in a meandering river, modelled
with CFD. The values are
given in cm.
1
5
9
3
1
7
1
1
5
7
The water depth in Fig. 9.8.6 is largest at the outside of the bend, corresponding with Fig. 9.8.3
Figure 9.8.7 Plan view of the meandering pattern in a channel computed with a CFD model. The case is a
physical model study by Zimpfer (1975).
If a meander short-cuts like shown in Fig. 9.8.8, this is often called a

chute. Chutes can also be seen in Fig. 9.8.5 and Fig. 9.8.6 by close
examination.
Figure 9.8.8. Plan view of
the meandering river,
where a chute has formed.
chute
124
Braided rivers
Beside being straight or meandering, the river may have a third plan
form: braided. This usually takes place at a steeper valley slope than
the meandering plan form. Also, braided river systems often occur in
reaches where the sediment transport capacity is lower than the sediment inflow, so that a net deposition occurs.
A river can evolve from a meandering planform to braided. Then chutes
form first and grow larger. The braided river does not follow a regular
pattern, but consists of several parallel channels, separating and joining
each other.
The flume study described in the figures above also evolved into a braiding pattern. After the meandering channel had evolved, the meander
bends became very large and channel cutoffs emerged. When several
cutoffs had formed, the resulting plan form was braided (Zimpfer, 1975).
CFD modelling of a braided system is shown in Fig. 9.8.9. The original
meandering pattern can also be seen.
Figure 9.8.9 Plan
view of depth-average water velocities
in a meandering
river, where cutoffs
have started to
make a braided plan
form.
Smaller islands in the river are often called bars.
9.9 Physical model studies

A physical model is an important tool to estimate effects of sediment
transport for engineering purposes. There exist a large number of scaling laws that has to be used according to the purpose of the study. A
detailed description of the different methods are given by Kobus (1980).
Water flow
The water discharge in the physical model is usually determined by the
the Froude law, based on the similarity between momentum and gravitational forces. The Froude number should be the same in the model as in
the river:
Scaling the water discharge in the physical
model
UFr = --------gh
(9.9.1)
Given the geometrical scale, the Froude number determines the water
discharge in the model. The next problem is to determine the correct
roughness in the physical model The roughness will affect the shear
forces on the bed and thereby the energy slope in the model. The shear
force is affected by viscosity and the Reynolds number. The Darcy-
125
Weissbachs diagram for the friction coefficients can be used to compute

the physical model friction factors. This friction factor should be the same
in the physical model as in the prototype. Given the two Reynolds numbers in the physical model and the prototype, the diagram can be used to
compute the relative roughness (ks/rh roughness to hydraulic radius) of
the physical model, given the similar parameter for the prototype.
The scaling of the sediments is more difficult.
Erosion
If the main topic of the investigation is the computation of maximum
scour or erosion, Shields graph may be used.
Scaling sediments for erosion studies
-.
* = ---------------------------g ( s w )d
(9.9.2)
A simplified approach is to say that * should be the same in the prototype as in the physical model. Eq. 9.9.2 is then solved with respect to the
particle diameter, giving the following equation:
dm
m ( s, p w )
------ = ------------------------------- p ( s, m w )
dp
(9.9.3)
The density of the sediments in the prototype is denoted s,p, and s,m is
the density of the sediments in the model. The computation involves the
shear stress and particle diameter in the prototype and in the model.
Subscripts m and p are used for model and prototype, respectively. It is
assumed that the particle Reynolds numbers are so great that the *
value is the same in prototype and model. The equation is also only valid
for particle sizes and bed shear stresses close to erosion. For finer particles, a more involved approach must be used.
Suspended sediments
Scaling sediments for
suspended load
If the main topic of the investigation is suspended sediments, the Hunter

Rouse number, z, is usually used:
wz = -------u *
(9.9.4)
If the Hunter-Rouse number is the same in the prototype and the model,
then this gives the fall velocity of the particles in the physical model. The
sediment diameter and density then have to be chosen accordingly.
Scaling time
To model the time for the movement of the sediments, the ratio of sediment transport to volume of sediments is used:
TQ
----------s
V
(9.9.5)
T is the time, Qs is the sediment load and V is the volume of the material
being transported. The scaling factor for time, st, is given as
126
T model
s t = ---------------------T prototype
A sediment transport formula can then be used to give the sediment discharge pr. unit width, qs. If s is the geometric scale (for example 1:20),
then Eq. 9.5.5 and 9.5.6 can be combined to:
q s, prototype 2
s t = -------------------------- s g
q s, model
(9.5.6)
Example: Find the scaling time for a physical model with sediments of
size 4 mm in the prototype and 2.5 mm in the physical model. The water
velocity is 0.2 m/s in the model, and the water depth is 0.2 m. The prototype sediments have a density of 2.65 kg/dm3, and the model sediments
have a density of 1.05 kg/dm3. Manning-Stricklers friction factor is 40,
and the model scale is s=0.015 (1:66.67).
Solution: First, the sediment discharge is computed in both the prototype
and the model, using Engelund-Hansens formula. This gives:
qs = 0.030 kg/s/m (model)
qs = 0.191 kg/s/m (prototype)
The time scale becomes:
0.191
2
1s t = ------------- ( 0.015 ) = 0.001435 = -------0.030
697
A simulation time of 1 hour in the lab will be similar to 697 hours in the
prototype.
The accuracy of the scaling of the time will not be better than the accuracy of the sediment transport formula used.
Multiple sediment processes
If the topic of the investigation involves both suspended sediments, erosion and sediment transport, then the different methods of scaling the
sediments may give different model sediment characteristics. It may
therefore not be possible to model all transport modes and processes.
This has been one of the motivations for developing CFD models with
sediment transport processes.
127
Fig. 9.9.1 Photo of the physical model

made at SINTEF, Norway, for the Kali
Gandaki hydropower plant in Nepal. The
model was used to investigate sediment
flushing from the hydropower reservoir.
The dams are shown in the lower left
corner. The water is fed into the model at
the upper left part of the picture. The
model is approximately ten meters long
from the dam to the end shown on the
right. The model was built of concrete
according to the prototype bed, and filled
with sand to the level of the spillway
crest. The model was then carefully filled
with water, and then the flushing started
by opening the dam gates. As the run-ofthe-river reservoir is fairly narrow, much
of the sediments was removed by the
flushing
Other problems
Scaling down finer sediments can result in particles with cohesive
forces. This would be the case for particles under 0.1 mm. To avoid this,
it is possible to use a distorted physical model. Then the vertical scale is
larger than the horizontal scale. However, this will also distort all secondary currents, which will not be correctly modelled.
Another problem is to scale bedforms. Dunes and ripples occur at different hydraulic conditions, and it is almost impossible to get for example
the ratio bedform height to water depth to be the same in the physical
model and the prototype. Also, bedforms may occur only in the physical
model and not in the prototype. The bedforms will then cause different
effects for energy loss and sediment transport capacity in the physical
model and the prototype.
9.10 Problems
Problem 1. Channel design
A discharge channel from a power station is 30 m wide and has a rectangular cross-section. The bed slope is 1:200 and the maximum water discharge from the power station is 100 m3/s. To prevent erosion of the
channel, stones are used at the bed of the channel. How large must the
stones be? Assume uniform grain size distribution for the stones.
Problem 2. Sediment transport
Problem 2 is taken from
the Iffezheim dam in the
River Rhine in Germany (Kuhl, 1992),
where this solution was
chosen.
The construction of a dam in a river caused erosion downstream. What

would be the reason?
The river authorities solved the problem by adding gravel to the river,
downstream of the dam. What would be the required amount of gravel,
when the water discharge was assumed to be 1500 m3/s, the water
depth 3.5 meters, the width 100 meters and the slope 1:1600 The gravel
size was identical to the sediment size at the bed: a diameter of 20 mm.
Can you think of an alternative solution to the problem?
128
Problem 3. Sediment load estimation

Estimate the annual sediment load in a river, given the flow duration
curve below, and the rating curve in Fig. 9.7.1. How much of the sediments are transported by the highest floods, occurring under 5 % of the
time?
3000
m3/s
2000
Flow duration
curve
1000
100 %
50 %
0 % (of time)
Problem 4. Physical model study

A physical model study of a reservoir is conducted. The scale is 1:30.
The water discharge in the prototype is 300 m3/s. What is the discharge
in the model?
The average dimensions of the physical model is 10 m long, 5 m wide,
and 10 cm deep. Suggest a sediment type and size for the physical
model, when investigating sedimentation of prototype silt particles of 0.3
mm.
Suggest a sediment type for modelling flushing of the same silt.
If a water discharge over time were to be modelled, including both sedimentation and erosion, what kind of sediment should be chosen?
Problem 5. HEC-6
Compute the trap efficiency over time in a water reservoir built in a river
with slope 1:200. The water discharge is 100 m3/s, the river width is 30
meters, the dam height is 30 meters. Assume a constant reservoir width
of 200 meters. The inflow sediment is silt with particle diameter 0.3 mm,
and the concentration is 2000 ppm.
Problem 6. Sediment load formulas
Compute the sediment transport capacity pr. m. width in an alluvial channel, with the following data:
U=2.5 m/s
Depth, y = 1.5 m
Manning-Strickler coefficient: 50
Sediment size: 1 mm
Use both Engelund-Hansens and van Rijns formulas.
129
10. River habitat modelling

10.1 Introduction
Earlier methods of evaluating impacts of river regulations on fish was dubbed
BOGSAT: Bunch Of Guys
Sitting Around a Table, discussing the effects of the
regulation. The purpose of
River Habitat Modelling is to
improve the scientific background for the evaluation.
The science of River Habitat Modelling has evolved over the last ten
years. The main purpose is to assess the effect of habitat for fish, mostly
salmon, in relation to river regulations. Hydropower production has often
been the cause of changes in the water discharge. River Habitat Modelling aims to quantify the effects of changes in the river flow conditions
and geometry, to assess the impact of regulations on the fish production.
In Norway, this has been used to determine minimum flow regulations in
rivers. Also, it is used for assessing environmental effects from hydropeaking.
The science has also lead to

increased cooperation and
10.2 Fish habitat analysis
understanding between biologists and engineers workThe basis of the currently used method is that fish will seek an optimum
ing in rivers.
condition with respect to:

- Feeding
- Energy to stay in the river
- Spawning
- Protection from predation
The factors vary depending on the age of the fish. Often, the critical age
for salmon will be the juvenile stage. The studies often look at the rearing
and growing areas of the fish.
Looking at river regulations, the main changes in the river will be the
physical habitat. The main factors often used are:
Main factors affecting
the fish:
- Water depth
- Water velocity
- Substrate/cover
Substrate is usually determined by a characteristic size of the stones on
the river bed. The stones provide cover/hiding places for the fish. However, plants with leaves extending out over the river will also provide
cover.
The effect of feeding is neglected in simpler models. Then, it is assumed
the fish prefers optimum values of the parameters given above.
The currently most used methodology for fish habitat analysis is called
the Instream Flow Incremental Methodology (IFIM). It can be divided in
the following stages:
Instream Flow Incremental Methodology

(IFIM):
1. Selection of representative reach of the river

2. Counting the fish at different locations, recording habitat parameters
3. Generating habitat preference functions
4. Measuring habitat parameters at different discharges
5. Generating spatially distributed habitat as a function of discharge
6. Computing the habitat as a function of a time series of discharge
The selection of representative reach of the river has to be done with
respect to typical values of depth, velocity and substrate. Often, fish biologists provide input for the selection.
130
The preferred method for counting the fish is by diving. The diver covers
the whole area of the selected reach, and put a marker on the river bed
for each fish observation. Afterwards, the depth, velocity and substrate is
recorded for each marker. It is also possible to identify fish locations
from the surface, but then it is often more difficult to spot the fish. Then
the velocity, depth and substrate is divided into groups, and it is counted
how many fish there is in each group. This is shown in Fig. 10.2.1:
Figure 10.2.1. Generation of HSI index
for velocity. The
number of fish in each
velocity group is
counted (left figure),
and scaled to max
value of unity to
become the HSI index
(right)
no. fish
HSI
12
0.5
Velocity (m/s)
5
10
Velocity (m/s)
5
10
The HSI index indicates the preferred velocity for the fish. Another question is how much area of a preferred velocity there is in the river. This is
called availability of velocity. The availability is computed the same way
as the HSI index, but on the vertical axis is the area of a given velocity
interval. This can also be scaled to unity as the maximum area.
A preference index, D, is made from the HSI index and the availability
function, by using the following formula:
rp D = -----------------------------( r + p ) 2rp
(10.2.1)
where p is the available habitat of a defined velocity range, and r is the

HSI index. Both p and r are scaled so their values range between zero
and 1.
Three regions are made:
- Preferred habitat, where most of the fish observations are
- Avoided habitat, where there are no observations
- Indifferent, the region between
Preferred habitat is if D is above 0.2. If D is below -0.2, this is avoided
habitat. The area between -0.2 and 0.2 is indifferent. An example is
given in Fig. 10.2.2.
It is also possible to make an index by combining the curves for velocity,
depth and substrate.
Given the indexing system, a map of preferred, indifferent and avoided
habitat is made. This is called a habitat map, and an example is given in
Fig. 10.2.3. The habitat will vary according to the water discharge. Summing up the preferred area in a time series of water discharge, a measure of fish habitat for a given river regulation is made. Effects of changes
in the regulations can thereby be computed, but using different regulated
discharges over the year.
131
Figure 10.2.2 Generation of preference curve

as a function of velocity
Preference function
1
Velocity
The preference curve is a

biological model for the
fish habitat. There also exist
other biological models,
based on different
approaches.
-1
Preferred
Avoided
Indifferent
The method will, however, only work if it is possible to estimate the
velocity and depth as a function of the water discharge. This has to be
made in preferably three dimensions, as the spatial variation of velocity
near the bed is required. The fish is often found near the bed. The following two chapters shows the two main methods:
- Measurements and zero/one-dimensional models
- Multidimensional numerical models
This is described in the following chapters.
Figure 10.2.2 Velocity vector map (right) and computed habitat (left) near the bed for juvenile salmon
in a fish farm tank, seen from above. The water is entering on the left side, creating the current shown. The
outlet is in the center of the tank, at the bottom. The fish farm tank is 1.5 meters deep and 2.7 meters wide
(Olsen and Alfredsen, 1994). Most of the area has preferred velocity, except for the entrance and exit region
where the velocity is too high, and the corners where the velocity is lower than preferred.
10.3 Zero and one-dimensional hydraulic models

Initial work on habitat hydraulics started with measuring the water velocities in the characteristic reach of the river. Measurements were made in
multiple cross-sections, or transects. A two-dimensional map of the
velocities and depths were then obtained for a given discharge. This was
repeated for several water discharges. An interpolation function was
used for discharges between the measured ones.
132
The interpolation function could be fairly involved, and often a onedimensional backwater program was used. Various weighing functions
were calibrated for the velocity distribution in the lateral direction. Examples of computer programs using this procedure is RIMOS and PHABSIM.
There were two problems with this procedure:
1. A large number of field measurements had to be carried out
2. The calibration of the interpolation functions were only valid for the
geometry where the measurements were taken. In other words, it
was not possible to estimate the effects of changes in the geometry.
To solve this problem, multi-dimensional models have to be used.
10.4 Multidimensional hydraulic models

The multidimensional numerical models are two-dimensional depthaveraged or three dimensional. The models solve the Navier-Stokes
equations using for example the methods given in Chapter 6. The models can compute the effect of changes in the bed geometry, as the
geometry is given as input for the grid. This is very useful for assessment
of fish habitat studies for river regulations, as a mean of improving the
habitat is often to create small dams or obstacles in the flow. The water
depth is thereby increased, and also the variation in water velocity.
10.5 Bioenergetic models

At the time this book was
made, the bioenergetic
models were still on the
research stage.
Bioenergetic models are the latest in a succession of habitat assessment methodologies. The theory is to compute how much energy the fish
uses in different locations of the flow, as a function of water velocity and
possibly turbulence. As opposed to earlier studies, the food intake is also
taken into account. The fish feed on small organisms, which have different concentrations in various locations of the river. This means the fish
will receive more food/energy in some locations than in others. The models also assess how much energy is consumed by the fish. This is a
function of the velocity in the river. Using three-dimensional numerical
models, it is possible to compute the food concentration, water velocity
and turbulence over the whole three-dimensional river body. The optimum location for the fish can then be estimated, with the assumption
that the fish will seek a maximum possible difference between the
energy intake and consumption.
10.6 Problems
Problem 1. Preference curves
The figures below give depth and velocity in a representative reach of
Sokna River in Norway, at a discharge of 10 m3/s. The figure with the
dots provide locations of fish observations. Make preference curves for
the fish, for both depth and velocity.
133
Velocity (cm/s) near

the bed for a discharge of 10 m3/s.
4
7
7
4
12
10
7
10
15
12
15
10
10
10
10
10
12
7
4
**
Fish observations
* **
**
* * ** * *** *
* * **
* * * *** *
* ** **
*
*** * *
* ** * *
2.6
2.2
Depth in meters, for 10 m3/s

1.5
2.6
2.6
2.2
2.2
2.2
2.6
2.2
2.2
2.6
2.6
2.2
1.9
1.9
2.9
1.5
3.6
1.9
Problem 2. Habitat map

Make habitat maps for Sokna River, for 10 m3/s, for both water depth
and velocity.
134
Problem 3. Habitat map for changed velocity

Compute the habitat maps when the discharge is lowered to 3 m3/s. The
velocity and depth is given below. Is the habitat improved or has it deteriorated?
Water depth in
meters for 3 m3/s
2.1
1.4
2.4
2.4
2.4
2.4
2.1
2.4
2.1
2.4
2.1
2.1
2.1
2.4
1.7
1.7
2.8
1.4
3.5
3
Velocity in cm/s near

the bed at 3 m3/s.
3
3
4
2
3
4
5
6
4
4
4
3
Problem 4. Habitat improvement

Suggest measures to improve the habitat in the river, and methods to
document the improvements before they are carried out.
135
Literature
Ackers, P. and White, R. W. (1973) "Sediment Transport: New Approach
and Analysis", ASCE Journal of Hydraulic Engineering, Vol. 99, No.
HY11.
Bagnold, R. A. (1973) The nature of saltation and of bed-load transport
in water, Proceedings of the Royal Society of London, A332. pp473504.
Bagnold, R. A. (1988) The physics of sediment transport by wind and
water, A collection of hallmark papers, ASCE Publications.
Bagnold, R. A. (1990) Sand, wind, and war: memoirs of a desert
explorer, Tucson : University of Arizona Press.
Bakhmeteff (1932) Hydraulics of open channels, McGraw-Hill Book
Company, New York.
Bengtsson, L. (1973) "Wind Stress on Small Lakes", Tekniska Hgskolan, Lund, Sweden.
Bihs, H. and Olsen, N. R. B. (2011) Numerical Modeling of Abutment
Scour with the Focus on the Incipient Motion on Sloping Beds, Journal
of Hydraulic Engineering.
Bindloss, M. (1976) "The light-climate of Loch Leven, a shallow Scottish
lake, in relation to primary production of phytoplankton", Freshwater
Biology, No. 6.
Blench, T. (1970) Regime theory design of canals with sand beds,
ASCE Journal of Irrigation and Drainage Engineering, Vol. 96, No. IR2,
Proc. Paper 7381, June, pp. 205-213.
Bowles, C., Daffern, C. D. and Ashforth-Frost, S. (1998) "The Independent Validation of SSIIM - a 3D Numerical Model", HYDROINFORMATICS '98, Copenhagen, Denmark.
Brooks, H. N. (1963), discussion of "Boundary Shear Stresses in Curved
Trapezoidal Channels", by A. T. Ippen and P. A. Drinker, ASCE Journal
of Hydraulic Engineering, Vol. 89, No. HY3.
Brethour, J. M. (2002) Transient 3-D model for lifting, transporting and
depositing solid material, Proceedings from the Fifth International Conference on Hydroinformatics, Cardiff, UK.
Carstens, T. J. (1997) Class notes in fluvial hydraulics - hydraulics of
receiving waters, Department of Hydraulic and Environmental Engineering, The Norwegian University of Science and Technology. (In Norwegian)
Chapra, S. C. (1997) "Surface water-quality modelling", McGraw-Hill,
ISBN 0-07-115242-3.
Demuren, A. O. and Rodi, W. (1984) Calculation of turbulence-driven
secondary motion in non-circular ducts, Journal of Fluid Mechanics, vol.
140, pp. 189-222.
Dey, S. (2003) Threshold of sediment motion on combinded transverse
and longitudinal sloping beds, Journal of Hydraulic Research, Vol. 41
136
(4), pp. 405-415.

van Dorn, W. (1953) "Wind stress on an artificial pond", Journal of
Marine Research, No. 12, pp. 249-276.
Einstein, H. A., Anderson, A. G. and Johnson, J. W. (1940) A distinction
between bed-load and suspended load in natural streams, Transactions
of the American Geophysical Unions annual meeting, pp. 628-633.
Einstein, H. A. and Ning Chien (1955) "Effects of heavy sediment concentration near the bed on velocity and sediment distribution", UCLA Berkeley, Institute of Engineering Research.
Engelund, F. (1953) "On the Laminar and Turbulent Flows of Ground
Water through Homogeneous Sand", Transactions of the Danish Academy of Technical Sciences, No. 3.
Engelund, F. and Hansen, E. (1967) "A monograph on sediment transport in alluvial streams", Teknisk Forlag, Copenhagen, Denmark.
Fergus, T., Hoseth, K. A. and Sterb, E. (2010) Handbook of rivers
(Vassdragshndboka), Tapir forlag, Norway, ISBN 978-82-519-2425-2
(in Norwegian).
Fisher, H. B., List, E. J., Koh, R. C. Y., Imberger, J. and Brooks, N. H.
(1979) Mixing in Inland and Coastal Waters, Academic Press.
Fischer-Antze, T., Stsser, T., Bates, P. and Olsen, N. R. B. (2001) "3D
numerical modelling of open-channel flow with submerged vegetation",
IAHR Journal of Hydraulic Research, No. 3.
Fisher, H. B., List, E. J., Koh, R. C. Y., Imberger, J. and Brooks, N. H.
(1979) Mixing in inland and costal waters, Academic press, New York.
French, R. H. (1994) Open-channel hydraulics, McGraw-Hill.
Graf, W. H. and Altinakar, M. S. (1996) Fluvial hydraulics, Vol. 2, Flow
and transport processes in channels of simple geometry, Wiley Publishers.
Hedger, R. D., Olsen, N. R. B., Malthus, T. J. and Atkinson, P. M. (2002)
"Coupling remote sensing with computational fluid dynamics modelling
to estimate lake chlorophyll-a concentration", Remote Sensing of Environment, Vol. 79, No. 1, pp. 116-122.
Henderson-Sellers, B. (1984) Engineering Limnology, Pitman Publishing Limited, ISBN 0-273-08539-5.
Hervouet, J-M. and Petitjean, A. (1999) Malpasset dam-break revisited
with two-dimensional computations, IAHR Journal of Hydraulic
Research, Vol. 37, No. 6.
Hey, R. D. (1979) Flow resistance in gravel-bed rivers, ASCE Journal
of Hydraulic Engineering, Vol. 105, No. 4, April.
Jones, W. P. and Launder, B. E. (1972) The prediction of Laminarization with a Two-Equation Model of Turbulence, International Journal of
Heat and Mass Transfer, Vol. 15, pp.
Kamphuis, J. W. (1974) , IAHR Journal of Hydraulic Research, Vol. 12,
No. 2.
137
Kawai, S. and Julien, P. Y. (1996) Point bar deposits in narrow sharp

bends, IAHR Journal of Hydraulic Research, No. 2.
Kobus, H. (1980) Hydraulic modelling, German Association for Water
Resources and Land Improvement, Hamburg, Germany.
Kreyszig, E. (1983) Advanced Engineering Mathematics, John Wiley &
Sons Publishers.
Kuhl, D. (1992) 14 years artificial grain feeding in the Rhein, downstream the barrage Iffezheim, 5th Int. Symp. on River Sedimentation,
Karlsruhe, Germany.
Lysne, D. K. (1969) Movement of sand in tunnels, ASCE Journal of
Hydraulic Engineering, Vol. 95, No. 6, November.
Mayer-Peter, E. and Mueller, R. (1948) "Formulas for bed load transport", Proceedings from the Second Meeting of the International Association for Hydraulic Structures Research, Stockholm, Sweden.
McQuivey, R. S. and Keefer, T. N. (1974) Simple method for predicting
dispersion in streams, ASCE Journal of Environmental Engineering, pp.
997-1011.
Melaaen, M. C. (1992) "Calculation of fluid flows with staggered and
nonstaggered curvilinear nonorthogonal grids - the theory", Numerical
Heat Transfer, Part B, vol. 21, pp 1- 19.
Munk, W. H. and Anderson, E. R. (1948) Notes on the theory of the
thermocline, Journal of Marine Research, Vol. 1.
Naot, D. and Rodi, W. (1982) Calculation of secondary currents in channel flow, ASCE Journal of Hydraulic Engineering, No. 8, pp. 948-968.
Nikuradse, J. (1933) Flow in rough pipes, ?? Report no. 361.
OConnor, D. J. and Dobbins, W. E. (1958) Mechanisms of reaeration in
natural streams, Trans. ASCE, no. 123, pp. 641-684.
Olsen, N. R. B. and Melaaen, M. C. (1993) "Three-dimensional numerical modelling of scour around cylinders", ASCE Journal of Hydraulic
Engineering, Vol. 119, No. 9, September.
Olsen, N. R. B., Jimenez, O., Lovoll, A. and Abrahamsen, L. (1994) "Calculation of water and sediment flow in hydropower reservoirs", 1st. International Conference on Modelling, Testing and Monitoring of
Hydropower Plants, Hungary.
Olsen, N. R. B. and Alfredsen, K. (1994) "A three-dimensional model for
calculation of hydraulic parameters for fish habitat", IAHR Conference on
Habitat Hydraulics, Trondheim, Norway.
Olsen, N. R. B. and Skoglund, M. (1994) "Three-dimensional numerical
modelling of water and sediment flow in a sand trap", IAHR Journal of
Hydraulic Research, No. 6.
Olsen, N. R. B. and Stokseth, S. (1995) "Three-dimensional numerical
modelling of water flow in a river with large bed roughness", IAHR Journal of Hydraulic Research, Vol. 33, No. 4.
Olsen, N. R. B. (1999) "Computational Fluid Dynamics in Hydraulic and
138
Sedimentation Engineering", Class notes, Department of Hydraulic and

Environmental Engineering, The Norwegian University of Science and
Technology.
Olsen, N. R. B. and Kjellesvig, H. M. (1998) "Three-dimensional numerical flow modelling for estimation of maximum local scour depth", IAHR
Journal of Hydraulic Research, No. 4.
Olsen, N. R. B. and Kjellesvig, H. M. (1998) "Three-dimensional numerical flow modelling for estimation of spillway capacity", IAHR Journal of
Hydraulic Research, No. 5.
Olsen, N. R. B. (1999) "Two-dimensional numerical modelling of flushing
processes in water reservoirs", IAHR Journal of Hydraulic Research,
Vol. 1.
Olsen, N. R. B. and Kjellesvig, H. M. (1999) "Three-dimensional numerical modelling of bed changes in a sand trap", IAHR Journal of Hydraulic
Research, Vol. 37, No. 2. abstract
Olsen, N. R. B. and Lysne, D. K. (2000) "Numerical modelling of circulation in Lake Sperillen, Norway", Nordic Hydrology, Vol. 31, No. 1.
Olsen, N. R. B., Hedger, R. D. and George, D. G. (2000) "3D Numerical
Modelling of Microcystis Distribution in a Water Reservoir", ASCE Journal of Environmental Engineering, Vol. 126, No. 10, October.
Olsen, N. R. B. (2003) "3D CFD Modeling of a Self-Forming Meandering
Channel", ASCE Journal of Hydraulic Engineering, No. 5, May.
Patankar, S. V. (1980) "Numerical Heat Transfer and Fluid Flow",
McGraw-Hill Book Company, New York.
Reynolds, C. S. (1984) "The ecology of freshwater phytoplankton", Cambridge University Press, Cambridge, UK.
Rhie, C.-M, and Chow, W. L. (1983) "Numerical study of the turbulent
flow past an airfoil with trailing edge separation", AIAA Journal, Vol. 21,
No. 11.
van Rijn, L. C. (1982) Equivalent roughness of alluvial bed, ASCE Journal of Hydraulic Engineering, Vol. 108, No. 10.
van Rijn, L. C. (1987) "Mathematical modelling of morphological processes in the case of suspended sediment transport", Ph.D Thesis, Delft
University of Technology.
Rodi, W. (1980) "Turbulence models and their application in hydraulics",
IAHR State-of- the-art paper.
Rouse, H. (1937) "Modern Conceptions of the Mechanics of Fluid Turbulence", Transactions, ASCE, Vol. 102, Paper No. 1965.
Rouse, H., Yih, C. S. and Humphreys, H. W. (1952) Gravitational convection from a boundary source, Tellus, pp. 201-210.
Schall, J. D. (1983) Two-dimensional investigation of shear flow turbulence in open-channel flow, PhD dissertation, Colorado State University, USA.
Schlichting, H. (1936) Experimental Investigations of Roughness, Proc.
Soc. Mech. Eng., USA.
139
Schlichting, H. (1979) "Boundary layer theory", McGraw-Hill.

Schumm, S. A., Mosley, M. P. and Weaver, W. E. (1987) Experimental
fluvial geomorphology, John Wiley & Sons Publishers.
Shields, A. (1936) Use of dimensional analysis and turbulence research
for sediment transport, Preussen Research Laboratory for Water and
Marine Constructions, publication no. 26, Berlin (in German).
Seed, D. (1997) "River training and channel protection - Validation of a
3D numerical model", Report SR 480, HR Wallingford, UK.
Spallart and Allmaras (1994) A one-equation turbulence model for aerodynamic flows, La Recherche Aerospatiale, no 1. pp 5-21.
Speziale, C. G. (1987) On nonlinear K-l and k-e models of turbulence,
Journal of Fluid Mechanics, vol. 178, pp. 459-475.
Stigebrandt, A., (1978) Three-dimensional selective withdrawal, Report
No. STF60 A78036, The Norwegian River and Harbour Laboratory,
Trondheim, Norway.
Steen, J-E. and Stigebrandt, A. (1980) Topological control of threedimensional selective withdrawal, Second Int. Symp. on Stratified
Flows, Trondheim, Norway, pp.447-455.
Streeter, H. W. and Phelps, E- B. (1925) "A study of the pollution and
natural purification of the Ohio River", US Public Health Service, Washington DC, Bulletin 146.
Vanoni, V., et al (1975) "Sedimentation Engineering", ASCE Manuals
and reports on engineering practice - No54.
Wu, J. (1969) "Wind Shear Stress and Surface Roughness at Air-Sea
Interface", Journal of Geophysical Research, No. 74, pp. 444-455.
Yang, T. C. (1973) "Incipient Motion and Sediment Transport", ASCE
Journal of Hydraulic Engineering, Vol. 99, No HY10.
Zimpfer, G. L. (1975) Development of laboratory river channels, PhD
Thesis, Department of Earth Resources, Colorado State University.
140
Appendix I. Source code for explicit solution

of Saint-Venant equations
#include "stdio.h"
#include "math.h"
main()
{
FILE *in;
FILE *out;
int i,j,k,n;
double timein[100];
double qinn[100];
double depth[200][2];
double velocity[200][2];
double time, y, u, alpha, beta, q, dummy;
double timestep = 3.0;
double deltax = 50.0;
double slope = 0.005;
double manning = 30.0;
int sections = 99;
in = fopen("inflow","r");
out = fopen("outflow","w");
fclose(out);
/* reading time series */
n=0;
for(j=0;j<100;j++) {
if(fscanf(in,"%lf %lf",&timein[j], &qinn[j]) != 2) break;
n++;
}
fclose(in);
/* initialization */
y = pow (qinn[0]/(manning*sqrt(slope)),0.6);
u = qinn[0] / y;
for(i=0;i<sections+1;i++) {
depth[i][0] = y;
depth[i][1] = y;
velocity[i][0] = u;
velocity[i][1] = u;
}
/* main loop */
time = timein[0];
for(j=0;j<=1000;j++) {
141
/* boundary conditions */
for(k=0;k<n;k++) if(timein[k]>time) break;
beta = (timein[k] - time) / (timein[k] - timein[k-1]);
q = qinn[k-1] * beta + qinn[k] * (1.0-beta);
y = pow (q/(manning*sqrt(slope)),0.6);
u = q / y;
velocity[0][0] = u;
velocity[0][1] = u;
depth[0][0] = y;
depth[0][1] = y;
velocity[sections][0] = velocity[sections-1][1];
depth[sections][0] = depth[sections-1][1];
/* first computation of the water depth, according to Eq. 3.4.5 */
for(i=1;i<sections;i++) {
depth[i][1] = depth[i][0] - timestep / (2.0 * deltax) * (velocity[i][0] *
(depth[i+1][0] - depth[i-1][0]) + depth[i][0] *
(velocity[i+1][0] - velocity[i-1][0]));
}
/* compute the water velocity */
dummy = - velocity[i][0] * timestep * 0.5 / deltax
* (velocity[i+1][0] - velocity[i-1][0]);
dummy += - 9.81 * (timestep * 0.5 / deltax) * (depth[i+1][0] depth[i-1][0]);
dummy += 9.81 * slope * timestep;
dummy += - velocity[i][0] * velocity[i][0] * timestep * 9.81
/ (pow(depth[i][0],1.3333) * manning * manning);
velocity[i][1] = velocity[i][0] + dummy;
}
/* depth according to continuity - control volume approach*/
u = 0.5 * (velocity[i][0] + velocity[i][1]);
depth[i][1] =
(0.25 * (velocity[i-1][1] + velocity[i-1][0]) *
(depth[i-1][1]+depth[i-1][0])
+ depth[i][0] * (deltax / timestep - 0.5 * u ))
/ (deltax / timestep + 0.5 * u );
}
/* updating variables */
velocity[i][0] = velocity[i][1];
depth[i][0] = depth[i][1];
}
142
/* printing */
time = time+timestep;
out = fopen("outflow","a");
fprintf(out,"%lf ", time);
fprintf(out,"%lf ",velocity[0][1]*depth[0][1]);
fprintf(out,"%lf ",velocity[1][1]*depth[1][1]);
fprintf(out,"%lf ",velocity[sections/4][1]*depth[sections/4][1]);
fprintf(out,"%lf ",velocity[sections/2][1]*depth[sections/2][1]);
fprintf(out,"%lf ",velocity[sections-1][1]*depth[sections-1][1]);
fprintf(out,"\n");
fclose(out);
}
}
INFLOW FILE:
0.0 10.0
100.0 15.0
200.0 20.0
300.0 15.0
400.0 10.0
10000.0 10.0
143
Appendix II List of symbols and units

The units are in brackets []
Latin
A
B
c,C
c
d
D
E
F
g
h
I
Ib,I0
If, Ie,
I
k
k
k
K
M
n
N
p
P
P
q
qs
Q
r
r
r
t
T
U
u
u*
V
V
v
W
w
x
X
y
y
z
area [m2]
width of a river/channel [m]
constants in k- turbulence model [dimensionless]
concentration [ppm, kg/m3, dimensionless]
sediment particle diameter [m]
habitat preference index
specific energy height [m]
flux [kg/s]
acceleration of gravity [m/s2]
water depth [m]
slope [dimensionless]
bed slope [dimensionless]
friction or energy slope [dimensionless]
heat flux
friction loss coefficient [dimensionless]
turbulent kinetic energy [m2/s2]
coefficient for biological reactions [1/day]
coefficient in hydrologic routing method [s]
Manning-Stricklers friction coefficient [m1/3/s]
Mannings friction coefficient [s/m1/3]
Brunt-Visl frequency [1/s]
habitat availability index
pressure [N/m2]
wetted perimeter [m]
water discharge/width [m2/s]
sediment discharge/width [kg/s/m]
water discharge [m3/s]
hydraulic radius [m]
distance from centerline of plume [m]
HSI index (habitat hydraulics)
time [sec.,days]
temperature [0C, 0K]
average velocity [m/s]
fluctuating velocity [m/s]
shear velocity [m/s]
average velocity in direction 2 [m/s]
volume [m3]
fluctuating velocity [m/s]
average velocity in vertical direction [m/s]
fluctuating velocity in vertical direction [m/s]
spatial variable [m]
coefficient in hydrologic routing method [dimensionless]
spatial variable [m]
water depth [m]
spatial variable - vertical distance [m]
Greek
dissipation of turbulent kinetic energy [m2/s2]

specific density for water [N/m3]
144
w
s
constant in velocity wall law (0.4) [dimensionless]

dynamic viscosity of water [Ns/m2] (1.3x10-3 at 20 0C)
kinematic viscosity of water [m2/s] (1.3x10-6 at 20 0C)
turbulent eddy-viscosity [m2/s]
density of water [kg/m3]
density of water [kg/m3]
density of sediments [kg/m3] (often set to 2.65 kg/dm3)
shear stress [N/m2]
dimensionless shear stress [dimensionless]
coordinate system direction 1 [dimensionless]
diffusion coefficient [m2/s]
145
Appendix III Solutions to selected problems

Chapter 2, Problem 2
First, compute the cross-sectional area:
A = 8x2+10x+15x2 = 96 m2
The wetted perimeter is:
P = 2+15+3+10+3+8+2 = 43 m
The hydraulic radius is:
r = A
--- = 96
------ = 2.23m
P
43
Mannings formula gives the water velocity:
1--- 2--2 3
2---
1 - 2.23 3 = 4.5 m/s

U = MI r = 60 -------500
The water discharge is:
Q = UA = 4.5x96 = 440 m3/s
When the Manning-Strickler friction coefficient varies, the same procedure and formulas have to be used for each part of the cross-section.
The following table gives the results:
Section A
Section B
Section C
Area
30
50
16
Wetter perimeter
17
16
10
Hydraulic radius
1.76
3.12
1.6
Velocity
2.6
5.73
2.44
Discharge
78
286
39
The total discharge is the sum of the discharges in the three parts.
Q = 78+286+39=403 m3/s.
146
First, find the water level at B, from the equation for the hydraulic jump:
h
1
2
-----B- = --- ( 1 + 8Fr C 1 )
2
hC
The velocity at C:
Q = 6--- = 2 m/s
U C = ----yC
3
The Froude number at C is:
2
Fr C
2
UC
2
= -------= 0.136
- = ---------------gy C
9.81x3
The depth is then given by:
h
1
1
2
2
-----B- = --- ( 1 + 8Fr C 1 ) = --- ( 1 + 8x0.136 1 ) = 0.222
2
2
hC
and
h B = 0.222h C = 0.222x3 = 0.667m

The main method to compute the distance from A to B is to start at A and
compute the water surface profile. When it reaches 0.667m, this will be
the location of the jump.
The following formula is used:
If I0
dy
------ = ----------------2
dx
1 Fr
It is converted to:
2
( 1 Fr )
x = y ---------------------If
This has to be solved in a table, where If and Fr are computed. If is computed by Mannings formula:
2
U I f = -----------4
--2 3
M y
147
Table 1:
y
If
Fr2
0.4
0.05
0.425
14.11
0.25
47.8
9.4
9.4
0.05
0.475
12.63
0.172
34.24
9.7
19.1
0.05
0.525
11.43
0.123
25.36
9.9
29.0
0.05
0.575
10.43
0.0910
19.28
10.0
39.0
0.037
0.6185
9.70
0.0714
15.51
7.5
46.5
0.03
0.652
9.2
0.0599
13.24
6.13
52.6
The distance from A to B is 53 meters.

Chapter 2, problem 5
The integral of the concentration over time becomes:
I =
c dt
= ct = 500x1ppmminutes
The discharge is:

3
kgm
m
l
m- = m
2kg - = 4000 -------Q = ------------ = -----------------= 4 --------- = 0.067 ------ = 67 min
min
s
s
I
500min
------------------ c dt
6
10
The continuity equation gives the water velocity as:
Q- = ----------20 - = 0.33 m
---U = ----s
By
2x30
The shear stress is:
1
= gyI = 1000x9.81x2x ------ = 313Pa
63
The shear velocity is:
u* =
--- =
313- = 0.56 m
-------------s
1000
Computing the diffusion coefficient. Using Fischer et al. (1979): (Eq.

4.3.3)
148
( UB )
( 0.33x30 )
= 0.011 --------------- = 0.011 --------------------------- = 0.96
Hu *
2x0.26
The initial concentration is equal to the mass divided by the volume of
water in the river in the time period:
m
2000
c 0 = ---------- = --------------------------------------------- = 0.000167 = 167ppm
Qt
1000x ( 20x10x60 )
These numbers we use in the general equation for the concentration
downstream of a spill: (Eq. 4.3.4)
2
c0 L
c ( t ) = ---------------- e
2 t
x Ut ) (--------------------4t
167x200
= ----------------------- e
2 0.96t
( 10000 0.33t ) --------------------------------------3.84t
This equation is plotted below:
The densimetric Froude number is:
50
-----------------------------------u0
( 3.14x1.5x1.5 ) - = 8.7
Fr = ---------------------------------------- = ------------------------------------------------------- 1000-
res 0
1023
-----------------------------------------------9.81x3
res gd 0
1023
The velocity is computed from:
149
2
1-- --3 20 3
u- = 4.3 ( 8.7 ) -------- 3

u0
0
96 --------2
20
= 1.01
The velocity is equal to the initial velocity, or 7 m/s. The fact that the
velocities are the same in this case is coincidental.
The shear stress on the lake is computed from Eq. 7.2.1 as
2
= c 10 a U a = 1.1x10 x1.2x15 = 0.3Pa

The slope of the water surface is given by Eq. 7.2.3:
7
0.3
I = ---------- = -------------------------------------- = 3x10
gh
1000x9.81x100
The density at 15 oC is approximately 999.12 kg/m3 according to the

table in Chapter 7.4, and the density for 5 oC is 999.99 kg/m3.
The slope of the thermocline is given by Eq. 7.4.10:
1
1000
7
4
I = I ----- = 3x10 x --------------------------------------- = 3.5x10
999.99 999.12
The thermocline will tilt around an axis in the middle of the lake. The dis-
150
tance from the middle of the lake to the far ends is half the length of the
lake. The rise/fall in the thermocline at these locations will be:
L
5000
4
h = --- I' = ------------ x3.5x10 = 0.87m
2
2
We start by computing the density gradient from the temperature gradient. The temperature gradient at 10 meter is 0.5 degrees/meter. There
will therefore be 4 meters between 10 and 12 degrees. The table in
Chapter 7.3 gives the densities at these temperatures as 999.73 and
999.52. The density gradient becomes:
kg
kg
999.52 -----3- 999.73 -----3kgd
m
m = 0.05 25 ---------- = ----------------------------------------------------3
dz
4m
m
The Brunt-Visl frequency is given as:
2
d
9.81
N = --g- ------ = ------------ ( 0.0525 ) = 0.000515Hz
1000
dz
(7.8.2)
This is inserted into the formula for the thickness of the abstraction layer:
1
--3
Q
-
d = k 3 -------------- B2 hN 2
2
3 2
--32m
------
s
- = 15m
= 1.2 ----------------------------------------------------------- ( 2m ) 2 x1m ( 0.000515Hz )
(7.8.4)
The thickness of the abstraction layer is both above and below the
intake. Therefore, it will only reach half the thickness above the intake, or
7.5 m. This is below 9 meters, which means the intake will not take water
from the upper 1 meter layer.
A stable condition means that the loss is equal to the inflow: 2000 kg/
day. The loss rate is 0.1, meaning 2000 kg/day is 10 % of the total. The
total phosphorous is therefore 2000/0.1 = 20 000 kg. The concentration
is:
20000kg - = 2x10 5 = 20ppm

c = ------------------------6
3
10 x10 kg
The steady state concentration for an inflow of 300 kg/day is similarly 3
ppm. The reduction will therefore be 17 ppm. 20 % of the original value
is 4 ppm. After time t, there will be 4-3=1 ppm left of the reduction. The
time is described by the following equation:
151
c = c0e
kt
1 = 17e
0.1t
17
t = 10 ln ------ = 28days
1
Assume uniform flow and that the bed shear stress is equal to the critical
shear stress given by Shields graph. Also, assume the channel is wide,
so that the hydraulic radius is equal to the water depth. This gives six
unknown and six equations. The unknown are:
- water depth h
- water velocity U
- particle diameter d
- bed roughness ks
- Manning-Stricklers friction factor M
- bed shear stress
The six equations are:
- Water continuity: Q=BhU
= ghI
k s = 3d 90
U = M
(2.1.1)
(2.1.4)
2 1
--- --3 2
rh I
(2.2.4)
M = 26
------
(2.2.5)
1
--6
ks
c
*= ----------------------------g ( s w )d
From Shields graph
The equations are best solved in the following manner:

1. Guess a Mannings coefficient, M.
2. Computer the water velocity using Eq. 2.2.4
3. Compute the water depth using the continuity equation
4. Compute the shear stress using Eq. 2.2.1.
5. Compute R* in Shields diagram
6. Find from Shields diagram
7. Compute d from the equation in Shields graph
8. Compute ks using 2.1.4
9. Compute Mannings M value from Eq. 2.2.5.
10 Check if M is equal to what was assumed in point 1. If it is very different, do another iteration starting from point 1, with the M value from
point 9.
Using this method we get:
1. M=30 (guess)
2. U=2.6 m/s
3. y = 1.3 m
4. = 60 N/m2
5. R* = 18 000
6. = 0.06
152
7. d = 0.06 m
8. ks = 0.18 m
9. M = 34
10. Iteration with M=34 gives d=0.06 m.
First, the particle fall velocity is obtained from Fig. 9.3.1 to be 0.15 m/s.
Then some basic parameters are computed:
Friction slope:
2
U - = --------------------2.5 - = 0.001456
I f = -----------4
--2 3
M y
50 x1.5
4
--3
Bed shear:
= gyI f = 1000x9.81x1.5x0.001456 = 21.42745673 N/m2

Shear velocity:
u* =
--- =
21.42
------------- = 0.146381203 m/s
1000
Particle Reynolds number:
u* d
0.14638x0.001- = 146
R * = -------= ----------------------------------6
10
Shields curve gives the Shields parameter: C = 0.05
Critical shear stress:
c = Cg ( s w )d = 0.05x9.81x ( 2650 1000 )x0.001

= 0.809 N/m2
The suspension parameter:
w - = -------------------------0.15 - = 2.5618
Z = -------u*
0.4x0.1464
Engelund-Hansens formula:
q s = 0.05 s U
d 50
----------------------- ---------------------------------g
(
s w )d 50
g -----s- 1
w
3
--2
153
q s = 0.05x2650x2.5
21.43
0.001
------------------------------------ ----------------------------------------------------------9.81 ( 2650 1000 )0.001
9.81 2650
------------ 1
1000
3--2
= 9.9 kg/m/s
Then we use van Rijns formula for bed material load:
c 2.1
------------c
qb
----------------------------------------- = 0.053 ---------------------------------------------------0.1
0.3 ( s w )g
1.5 ( s w )g
-
D 50 ------------------------D 50 -------------------------w
w2
21.43
0.809- 2.1
-------------------------------0.809
2650 1000 )9.81-
- 0.001 1.5 (-------------------------------------------q b = 0.053 ---------------------------------------------------------------------------0.1
1000
0.3 ( 2650 1000 )9.81
--------------------------------------------0.001
6 2
1000x ( 10 )
qb = 0.002295 m2/s
Van Rjin gives qb in m2/s. This is transformed to kg/s by multiplying it
with the sediment density: 2650 kg/m3, giving:
qb = 6.082 kg/s/m
Then we use Van Rijns formula for suspended load. We assume the reference level at the bed is equal to 5 % of the water depth, or 1.5x0.05 =
0.075 m. The reference concentration is:
c bed
c 1.5
21.43
0.809- 1.5
-------------------------------------------D 50
c
0.809
- = 0.015 0.001
------------- -------------------------------------------------------------------------------= 0.015 --------- ------------------------------------------------------0.3
0.3
a
0.075
1
1
---
--( s w )g 3
(-------------------------------------------2650 1000 )9.81- 3

------------------------ D 50
0.001
2
6 2
1000x ( 10 )
cbed = 0.009756758
(volume fraction)
The sediment transport caused by this reference concentration only

takes place above the reference level. To compute the suspended sediment transport, we can divide the water column in layers. For convenience, let us assume the first layer is twice as high as the reference level.
This is then 10 % of the water depth, or 0.15 m. And that there are three
more layers of equal size:
154
(1.5 - 0.15)/10 = 0.45 m

We then use the Hunter Rouse formula to compute the concentration in
the center of each cell, and the logarithmic velocity profile to compute
the velocities. This can then be multiplied with the height of each cell to
give an estimate of the suspended load.
Table 2:
Cell
no.
Distance
from bed
Velocity
Hunter Rouse
concentration
Cell
height
Cell flux
1.275
2.77
6.22E-06
0.45
7.7E-06
0.825
2.61
0.0003168
0.45
0.00037
0.375
2.32
0.00883
0.45
0.00923
0.075
1.73
0.075
0.130
The sum of the cell fluxes is: 0.14. This can be multiplied with the bed
concentration to give the sediment transport in m2/s. Then multiplied
with the density, we get the sediment transport in kg/s/m:
qs = 0.14* 0.009756758*2650 = 3.6 kg/s/m.
Note that the height of the cell closest to the be has only been set to
0.075 m, although it is 0.15 meters. This is because the sediment transport below the reference level is considered bed load, and computed by
the bed load formula.
Total load according to van Rijn is then:
qt = 3.6 + 6.08 = 9.7 kg/s/m.
Also note that there are only four layers in the vertical direction. If 11 layers had been used, we would have gotten 2 % higher result for the suspended load.
In our case, the results by van Rijns method and Engelund-Hansens formula are very similar. This is normally not the case. Often, results from
different sediment discharge formulas may deviate with a factor 2-3 or
more.
155
Appendix IV: An introduction to programming in C

This chapter is written as a brief introduction to C programming for students who has not had previous courses on programming.
A computer program is written as text file using an editor. In the text file,
different commands tells the computer what to do. This text file is often
called a source code. The text file needs to be transformed into a computer program. The file containing the computer program is often called
the executable file. The transformation of the text file to the executable
file is done using a compiler. Depending on the language of the text file,
different compilers are used. For the C language, we need a C compiler.
In the present course we will use the LCC compiler, which is freeware.
The C program consists of the following structure:
main () {
Variable declarations
Commands on what to do
}
Note the brackets, which are necessary. Omitting one leads to the compiler giving an error message, and the executable is not produced.
Variable declarations
To make the program compute something, we need to declare variables.
In a spreadsheet, a variable is named after its address, for example b12.
In C, we can give any name to a variable. For example:
int counter;
double flux;
The variable named counter is declared as an integer, and the variable
flux is declared as double, a floating point number with 12 digits precision. Note the ; after each declaration.
It is also possible to declare arrays of integers or floats:
double velocity[100], depth[100];
If we have 100 cross-sections in a river, the velocity in section 14 is
given as velocity[14].
We can also use multi-dimensional arrays:
double discharge[2][100];
In C, the first number in the array is zero, and the last array in the arrays
above is 99, if decreased with 100 elements.
156
We can also declare files with different names, used to read input-data
from and write output-data to:
FILE * input:
FILE * result;
Note the * that needs to be included for the files.
A file needs to be opened before we can read from it. The syntax for this
is:
input = fopen("inflow","r");
Similar for a file we want to write to:
result = fopen("outflow","w");
The file names are here given as inflow and outflow. These are text files.
After writing to a file, it needs to be closed for other programs to read
from it:
fclose(result);
To read information from a file to a variable, the following syntax is used
:
fscanf(input,"%lf",&depth[1]);
Note the syntax. If this is not correct, the compiler will produce an error
message, or the program will not work.
Similarly, the following syntax is used to write to files:
fprintf(result,"%lf ",velocity[0]);
Note the & is used when reading data and not when writing data to files.
Commands
The program is made up of a series of commands. They will be carried
out in the same order as given in the file. There are several types of
commands.
Variables can be initialized using the = operator. For example:
velocity[0] = 2.0;
Variables can be incremented. For example, in the following, the variable counter will increase its value by one:
counter = counter + 1;
If count was 3 before this line, it will be 4 afterwards. Incrementing an
integer with one can also be written
counter++;
This does the same as the line before.
157
A typical command is to write a formula:

velocity[1] = discharge / depth[1];
A formula can be long and complex, and one can use * for multiplication,
and + and - for addition and subtraction. Also, one can use brackets in
complex formulas, to several levels:
velocity[1] = discharge / (depth[1] + 1.0e-20);
The above formula is a trick to avoid errors if depth[1] is zero.
Repetition of formulas over many cross-sections is done in a loop. There
are different types of loops, but the for loop is often used. The following
loop repeats the above formula for all the 100 cross-sections in the river,
starting from cross-section no. 0 to cross-section no. 99:
for(counter = 0; counter <100; counter++) {
velocity[counter] = discharge / (depth[counter] + 1.0e-20);
}
Several formulas can be used inside a loop, and it is also possible to
nest several loops inside each other.
An if sentence can be used to give a special logic. For example:
if ( froude > 1.0 ) {
depth = 2.0;
}
An if sentence can also be used to break out of a loop:
for(counter = 0; counter <100; counter++) {
if ( depth[counter] < 1.0e-20) break;
velocity[counter] = discharge / depth[counter];
}
This loop will be stopped if depth is below 10-20, even if counter has not
reached 99. The program will then jump out of the loop and do the next
thing in the file.
Remember that the syntax is very important. One wrong type of brackets, for example, will give a compiler error.
The introduction given above is only a very small fraction of all the commands, declarations etc. available in C. For more details, a textbook on
C programming should be consulted.

Numerical Modelling and Hydraulics

Uploaded by

Copyright:

Available Formats

Numerical Modelling and Hydraulics

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Numerical Modelling and Hydraulics

Uploaded by

Copyright:

Available Formats

Numerical

Department of Hydraulic and

3rd edition, 9. March 2012.