River Engineering

John Fenton
Institute of Hydraulic and Water Resources Engineering
Vienna University of Technology
June 20, 2011

Table of Contents
Table of Contents

1 Introduction
1.1 The nature of what we will and will not do illuminated by some aphorisms and some people
1.2 The problem of flow in a river . . . . . . . . . . . . . . . . . .

.
.

3
3
4

2 River hydraulics
2.1 A note on terminology in the English language:
2.2 Summary
. . . . . . . . . .
2.3 The one-dimensional equations of hydraulics
2.4 Structures, controls, and boundary conditions

.
.
.
.

10
. 10
. 10
. 12
. 50

.
.

62
62
69

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

3 Measurement and analysis

3.1 Hydrometry and the hydraulics behind it . . . . . .
3.2 The analysis and use of stage and discharge measurements .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

4 Computational hydraulics

80

5 Sediment transport
5.1 General . . . . . . .
5.2 Initiation of motion . . . .
5.3 Bedforms and alluvial roughness
5.4 Transport formulae . . . .
5.5 Unsteady aspects
. . . .
6 River morphology
6.1 Introduction . . . . .
6.2 Planform
. . . . .
6.3 Longitudinal profile
. .
6.4 Bends
. . . . . .
6.5 Channel characteristics . .
6.6 Bifurcations and confluences

.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

81
. 81
. 81
. 81
. 81
. 81

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

82
. 82
. 82
. 82
. 82
. 82
. 82

7 River engineering
7.1 Introduction . . . . . . . .
7.2 Bed regulation . . . . . . .
7.3 Discharge control . . . . . .
7.4 Water level control . . . . . .
7.5 Water quality control . . . . .
7.6 River engineering for different purposes

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

83
. 83
. 83
. 83
. 83
. 83
. 83

.
.
.
.
.
.

.
.
.
.
.
.

References

84

Chapter 1

Introduction

1.1 The nature of what we will and will not do illuminated by some aphorisms
and some people
"There is nothing so practical as a good theory" stated in 1951 by Kurt Lewin (D-USA, 1890-1947): this is
essentially the guiding principle behind these lectures. We want to solve practical problems, both in professional
practice and research, and to do this it is a big help to have a theoretical understanding and a framework. We claim
to aspire to deep connected simplicity rather than shallow disparate complexity.
"The purpose of computing is insight, not numbers" the motto of a 1973 book on numerical methods for
practical use by the mathematician Richard Hamming (USA, 1915-1998?). That statement has excited the opinions
of many people (search any three of the words in the Internet!). However, in contradiction to Hammings assertion,
numbers are often important in engineering, whether for design, control, or other aspects of the practical world.
A characteristic of many engineers, however, is that they are often blinded by the numbers, and do not seek the
physical understanding that can be a valuable addition to the numbers. In this course we are not going to deal with
many numbers. Instead we will deal with the methods by which numbers could be obtained in practice, and will
try to obtain insight into those methods. Hence we might paraphrase simply: "The purpose of this course is insight
into the behaviour of rivers; with that insight, numbers can be often be obtained more simply, cheaply, and reliably.
"It is EXACT, Jane" a story told to the lecturer by a botanist colleague. The most important river in Australia is
the Murray River, 2 375 km (Danube 2 850 km), maximum recorded flow 3 950 m3 s1 (Danube at Iron Gate Dam:
15 400 m3 s1 ). It has many tributaries, flow measurement in the system is approximate and intermittent, there is
huge biological and fluvial diversity and irregularity. My colleague, non-numerical by training, had just seen the
demonstration by an hydraulic engineer of a computational model of the river. She asked: "Just how accurate is
your model?". The engineer replied intensely: "It is EXACT, Jane".
Nothing in these lectures will be exact. We are talking about the modelling of complex physical systems.
A further example of the sort of thinking that we would like to avoid: in the area of palaeo-hydraulics, some
Australian researchers made a survey to obtain the heights of floods at individual trees. This showed that the palaeoflood reached a maximum height on the River Murray at a certain position of 18.01 m (sic), Having measured the
cross-section of the river, they applied the Gauckler-Manning-Strickler Equation to determine the discharge of the
prehistoric flood, stated to be 7 686 m3 s1 (sic) ...
William of Ockham (England, c1288-c1348): "Occam"s razor is the principle that can be popularly stated as
"when you have two competing theories that make similar predictions, the simpler one is the better." The term razor
refers to the act of shaving away unnecessary assumptions to get to the simplest explanation, attributed to 14thcentury English logician and Franciscan friar, William of Ockham. The explanation of any phenomenon should
make as few assumptions as possible, eliminating those that make no difference in the observable predictions
of the explanatory hypothesis or theory. When competing hypotheses are equal in other respects, the principle
recommends selection of the hypothesis that introduces the fewest assumptions and postulates the fewest entities
while still sufficiently answering the question. That is, we should not over-simplify our approach.
In general, model complexity involves a trade-off between simplicity and accuracy of the model. Occams Razor
is particularly relevant to modelling. While added complexity usually improves the fit of a model, it can make the
model difficult to understand and work with.
The principle has inspired numerous expressions including "parsimony of postulates", the "principle of simplicity",
the "KISS principle" (Keep It Simple, Stupid). Other common restatements are:
Leonardo da Vinci (I, 14521519, worlds most famous hydraulician, also an artist) his variant short-circuits the
need for sophistication by equating it to simplicity "Simplicity is the ultimate sophistication".
Wolfang A. Mozart (A, 17561791) "Gewaltig viel Noten, lieber Mozart", soll Kaiser Joseph II. ber die erste der
groen Wiener Opern, die "Entfhrung", gesagt haben, und Mozart antwortete: "Gerade so viel, Eure Majestt,
als ntig ist." (Emperor Joseph II said about the first of the great Vienna operas, "Die Entfhrung aus dem Serail",
"Far too many notes, dear Mozart", to which Mozart replied "Your Majesty, there are just as many notes as are
3

necessary"). The truthfulness of the story is questioned - Joseph was more sophisticated than that ...
Albert Einstein (D-USA,1879-1955): "Make everything as simple as possible, but not simpler."
Karl Popper (A-UK, 1902-1994) argued that a preference for simple theories need not appeal to practical or
aesthetic considerations. Our preference for simplicity may be justified by his falsifiability criterion: We prefer
simpler theories to more complex ones "because their empirical content is greater; and because they are better
testable". In other words, a simple theory applies to more cases than a more complex one, and is thus more easily
falsifiable. Popper coined the term critical rationalism to describe his philosophy. The term indicates his rejection
of classical empiricism, and of the classical observationalist-inductivist account of science that had grown out of
it. Logically, no number of positive outcomes at the level of experimental testing can confirm a scientific theory
(Humes "Problem of Induction"), but a single counterexample is logically decisive: it shows the theory, from
which the implication is derived, to be false. For example, consider the inference that "all swans we have seen are
white, and therefore all swans are white," before the discovery of black swans in Australia. Poppers account of
the logical asymmetry between verification and falsifiability lies at the heart of his philosophy of science. It also
inspired him to take falsifiability as his criterion of demarcation between what is and is not genuinely scientific:
a theory should be considered scientific if and only if it is falsifiable. This led him to attack the claims of both
psychoanalysis and contemporary Marxism to scientific status, on the basis that the theories enshrined by them are
not falsifiable.
Falsifiability, however, seems not so important a principle when it comes to engineering. Another objection is
that it is not always possible to demonstrate falsehood definitively, especially if one is using statistical criteria to
evaluate a null hypothesis. More generally it is not always clear, if evidence contradicts a hypothesis, that this
is a sign of flaws in the hypothesis rather than of flaws in the evidence. However, this is a misunderstanding of
what Poppers philosophy of science sets out to do. Rather than offering a set of instructions that merely need
to be followed diligently to achieve science, Popper makes it clear in The Logic of Scientific Discovery that his
belief is that the resolution of conflicts between hypotheses and observations can only be a matter of the collective
judgment of scientists, in each individual case.
Thomas Kuhn (USA, 1922-1996): In The Structure of Scientific Revolutions argued that scientists work in a series
of paradigms, and found little evidence of scientists actually following a falsificationist methodology. Poppers
student Imre Lakatos (H-UK, 1922-1974) attempted to reconcile Kuhns work with falsificationism by arguing
that science progresses by the falsification of research programs rather than the more specific universal statements
of naive falsificationism.
Kuhn argued that as science progresses, explanations tend to become more complex before a paradigm shift offers
radical simplification. For example Newtons classical mechanics is an approximated model of the real world.
Still, it is quite sufficient for most ordinary-life situations.
Popper seems to have anticipated Kuhns observations. In his collection Conjectures and Refutations: The Growth
of Scientific Knowledge (Harper & Row, 1963), Popper writes, "Science must begin with myths, and with the
criticism of myths; neither with the collection of observations, nor with the invention of experiments, but with the
critical discussion of myths, and of magical techniques and practices. The scientific tradition is distinguished from
the pre-scientific tradition in having two layers. Like the latter, it passes on its theories; but it also passes on a
critical attitude towards them. The theories are passed on, not as dogmas, but rather with the challenge to discuss
them and improve upon them."
Another of Poppers students Paul Feyerabend (A-USA, 1924-1994) ultimately rejected any prescriptive methodology, and argued that the only universal method characterizing scientific progress was "anything goes!"
The utility of deep simplicity compared with shallow complexity
In view of the above, the lecturer prefers to approach problems with the former, deep simplicity, rather than the
latter, shallow complexity. The following lecture notes will be a reflection of that. Unfortunately "deep" sometimes
looks complicated at first, which can be off-putting. However, once a principle or a practice is established or
understood, it seems simple in retrospect.

1.2 The problem of flow in a river

The governing equations are the three-dimensional equations of mass and momentum conservation, the latter called
the Navier-Stokes equations, which are partial differential equations that describe the distribution of velocity and
pressure throughout a field of flow. They are valid at small (microscopic) scales and large (atmospheric and
oceanic) scales. For most practical problems in water flow, the flow becomes unstable: paradoxically, the viscosity
4

whose primary role is as a momentum diffusion or smoothing phenomenon, actually causes the flows to be unstable
and to become turbulent. Most flows of interest to civil engineers are actually turbulent boundary layer flows.

1.2.1

Physical modelling

The problem of a turbulent flow with a free surface in the presence of irregular and moveable boundaries is so
complicated that one effective way of modelling it is to build a physical scale model, which reproduces most of
the real phenomena. This has been the traditional way of much hydraulics. There are problems, however, in that
in a scale model, different physical properties of the fluid and the flow have different relative importances, and the
model is an approximate one. In the spirit of modelling, however, that is a legitimate approach and in hydraulics it
has often provided answers where otherwise none could be obtained.

1.2.2

Computational fluid mechanics

It is possible to apply computational fluid mechanics (CFD) to solve such problems, which usually means the
numerical solution in time and space of three-dimensional partial differential equations.
Direct numerical simulation (DNS) is the most basic method where the Navier-Stokes equations are solved
numerically in time, and the instability of the flow and the resulting turbulence is simulated. It captures all of
the relevant scales of turbulent motion, so no model is needed for the smallest scales. This approach is extremely
expensive, if not intractable, for complex problems on modern computing machines, hence the need for models to
represent the smallest scales of fluid motion.
Reynolds-averaged NavierStokes equations (RANS) is the oldest approach to turbulence modeling. The velocities and pressure are written as a sum of slowly-varying plus rapidly-varying turbulent components. The equations
are averaged mathematically over a time scale large compared with the turbulent fluctuations, which gives equations for the slowly-varying terms, but which introduces new apparent stresses known as Reynolds stresses. It
is necessary to introduce phenomenological models for such quantities. Such models include using an algebraic
equation for the Reynolds stresses which include determining the turbulent viscosity, and depending on the level
of sophistication of the model, solving transport equations for determining the turbulent kinetic energy and dissipation. Models include k (Spalding) and the Mixing Length Model (Prandtl).
Large eddy simulation (LES) is a technique in which the smaller eddies are filtered and are modelled using a
sub-grid scale model, while the larger energy carrying eddies are simulated. This method generally requires a
more refined mesh than a RANS model, but a far coarser mesh than a DNS solution.

Vortex method is a grid-free technique for the simulation of turbulent flows. It uses vortices as the computational
elements, mimicking the physical structures in turbulence.

1.2.3

Mathematical/theoretical modelling

For most hydraulics problems the irregularity of geometry, the large size, and the lack of data available, mean that
there is an imbalance between the little that is known about a problem and the sophisticated methods that can be
applied to simulate it. In this course we are going to take a different path, and use mathematical modelling. A
better title might be "theoretical modelling". There is almost an inverse correlation between the sophistication of
a model and the understanding that emerges: simple models provide insight, complicated models provide many
results, but insight is harder to obtain.
A mathematical model uses mathematical language to describe a system. Eykhoff in 1974 defined a mathematical
model as "a representation of the essential aspects of an existing system (or a system to be constructed) which
presents knowledge of that system in usable form". For our purposes, that is a quite useful definition.
Often when engineers analyze a system to be controlled or optimized, they use a mathematical model. In analysis,
engineers can build a descriptive model of the system as a hypothesis of how the system could work, or try to
estimate how an unforeseeable event could affect the system. Similarly, in control of a system, engineers can try
out different control approaches in simulations.
Black-box and Clear-box modelling: Mathematical modelling problems are often classified into black box or
white box (also called glass box or clear box) models, according to how much a priori information is available
of the system. A black-box model is a system of which there is no a priori information available. A clear-box
model is a system where all necessary information is available. Practically all systems are somewhere between the
black-box and clear-box models, for which the term grey box can be used.
Usually it is preferable to use as much a priori information as possible to make the model more accurate. Therefore
the clear-box models are usually considered easier, because if one has used the information correctly, then the
5

model will behave correctly.

In black-box models one tries to estimate both the functional form of relations between variables and the numerical
parameters in those functions. Using a priori information we could end up, for example, with a set of functions that
probably could describe the system adequately. If there is no a priori information we would try to use functions as
general as possible to cover all different models. An often used approach for black-box models are neural networks
which usually do not make assumptions about incoming data. The problem with using a large set of functions to
describe a system is that estimating the parameters becomes increasingly difficult when the amount of parameters
(and different types of functions) increases. This is usually (but not always) true of models involving differential
equations.
Any model which is not pure clear-box contains some parameters that can be used to fit the model to the system
it describes. If the modelling is done by a neural network, the optimization of parameters is called training. In
more conventional modelling through explicitly given mathematical functions, parameters are determined by curve
fitting.
As the purpose of modelling is to increase our understanding of the world, the validity of a model rests not only on
its fit to empirical observations, but also on its ability to extrapolate to situations or data beyond those originally
described in the model. One can argue that a model is worthless unless it provides some insight which goes beyond
what is already known from direct investigation of the phenomenon being studied.

1.2.4

Nature of problems
Input

System

Output

Physical parameters or
numerical coecients
Physical equations or
assumed equations

Figure 1.1:

Nature of a physical system with input and output

Consider a physical system such as that shown in Figure 1.1, which is simplified here so as to show a single timedependent input, and also just one output quantity. In Clear-box modelling, such as the hydraulic model of a river,
the model equations are known to some extent, and the physical parameters which are used by those equations are
known, probably to a lesser extent. In a Black-box model, such as a neural network or a unit-hydrograph model
of a catchment, the equations of the system are assumed generic response equations, linear or nonlinear, which are
expressed in terms of numerical coefficients.
Simulation is when both the system and the inputs are given, and it is required to calculate the responses, which
might be, for example include, a water level hydrograph at a station, such as at Hainburg an der Donau. A common
procedure in hydraulics is to assume values for physical parameters such as resistance of the stream (which might
include, for example, telephoning your friend who worked on a similar river and asking her/him what they used for
resistance ...). Rather more sophisticated is the numerical determination of physical (clear-box) or computational
(black-box) parameters. This is the process of System Identification, where measured values of input and output
are used to determine what the system characteristics are. This is also known as the solution of an Inverse problem.
After the execution of this phase, then simulation can be performed with rather more confidence.

Bibliography of useful references

Tables 1.1-1.4 show some of the many references available, some which the lecturer has referred to in these notes
or in his work. The last column shows whether the source is the lecturer (JDF), the library in the Institute of
Hydraulic Engineering (E222) or the University Library (UBTUW).

Reference

Comments

English
Chanson, H. (1999), The Hydraulics of Open Channel Flow,
Arnold, London.

Good introduction, also sediment aspects

UBTUW
JDF

Chaudhry, M.H. (1993), Open-channel flow, Prentice-Hall.

Very good readable but technical book

E222
JDF

Chow, V.T. (1959), Open-channel Hydraulics, McGraw-Hill,

New York.

Classic, now dated, not so readable

E222
JDF

Francis, J.R.D. & Minton, P. (1984), Civil Engineering Hydraulics, fifth edn, Arnold, London.

Good elementary introduction

JDF

French, R.H. (1985), Open-Channel Hydraulics, McGraw-Hill,

New York.

Wide general treatment

E222
JDF

Henderson, F.M. (1966), Open Channel Flow, Macmillan, New

York.

Classic, high level, readable

UBTUW
JDF

Jain, S.C. (2001), Open-Channel Flow, Wiley.

High level, but terse and readable

JDF

Julien, P.Y. (2002), River Mechanics, Cambridge.

Ditto more applications to

morphology

E222
UBTUW
JDF

Montes, S. (1998), Hydraulics of Open Channel Flow, ASCE,

New York.

Encyclopaedic & Definitive

JDF

Townson, J.M. (1991), Free-surface Hydraulics, Unwin Hyman,

London.

Simple, readable, mathematical

E222
JDF

Vreugdenhil, C.B. (1989), Computational Hydraulics: An Introduction, Springer.

Simple introduction to computational hydraulics

E222
JDF

The Austrian and International

classic of mathematical hydraulics

UBTUW

Deutsch
Forchheimer, Ph. (1930), Hydraulik, Teubner, Leipzig

Naudascher, E., (1992), Hydraulik der Gerinne und Gerinnebauwerke, Springer Verlag, Wien, New York

UBTUW

Preiler, G., Bollrich, G., (1985) Technische Hydromechanik,

VEB Verlag f
ur Bauwesen, Berlin

UBTUW

Table 1.1: Introductory and general references

Reference

Comments

Boiten, W. (2000), Hydrometry, Balkema

A modern treatment of river

measurement

JDF

Bos, M.G. (1978), Discharge Measurement Structures, second

edn, International Institute for Land Reclamation and Improvement, Wageningen.

Good encyclopaedic treatment

of structures

E128

Bos, M.G., Replogle, J.A. & Clemmens, A.J. (1984), Flow Measuring Flumes for Open Channel Systems, Wiley.

Good encyclopaedic treatment

of structures

Fenton, J.D. & Keller, R.J. (2001), The calculation of streamflow from measurements of stage, Technical Report 01/6, Cooperative Research Centre for Catchment Hydrology, Monash
University.

Two level treatment - practical

aspects plus high level review of
theory

JDF

Novak, P., Moat, A.I.B., Nalluri, C. & Narayanan, R. (2001),

Hydraulic Structures, third edn, Spon, London.

Standard readable presentation

of structures

E222

Table 1.2: Books on practical aspects, flow measurement, and structures

Reference

Comments

Cunge, J.A., Holly, F.M. & Verwey, A. (1980), Practical Aspects

of Computational River Hydraulics, Pitman, London.

Thorough and reliable presentation

JDF

Dooge, J.C.I. (1987), Historical development of concepts in open

channel flow, in G. Garbrecht, ed., Hydraulics and Hydraulic
Research: A Historical Review, Balkema, Rotterdam, pp.205
230.

Interesting review

JDF

Flood Studies Report (1975), Flood Routing Studies, Vol. 3,

Natural Environment Research Council, London.

A readable overview

E222
JDF

Lai, C. (1986), Numerical modeling of unsteady open-channel

flow, in B. Yen, ed., Advances in Hydroscience, Vol. 14, Academic.

Good review

UBTUW
JDF

Liggett, J.A. (1975), Basic equations of unsteady flow, in K.

Mahmood & V. Yevjevich, eds, Unsteady Flow in Open Channels, Vol. 1, Water Resources Publications, Fort Collins, chapter
2.

Readable overview

JDF

Liggett, J.A. & Cunge, J.A. (1975), Numerical methods of solution of the unsteady flow equations, in K. Mahmood & V.
Yevjevich, eds, Unsteady Flow in Open Channels, Vol. 1, Water Resources Publications, Fort Collins, chapter 4.

Readable overview

JDF

Miller, W.A. & Cunge, J.A. (1975), Simplified equations of unsteady flow, in K. Mahmood & V. Yevjevich, eds, Unsteady Flow
in Open Channels, Vol. 1, Water Resources Publications, Fort
Collins, chapter 5, pp. 183257.

Readable

JDF

Price, R.K. (1985), Flood Routing, in P. Novak, ed., Developments in hydraulic engineering, Vol. 3, Elsevier Applied Science,
chapter 4, pp. 129173.

The best overview of the

advection-diusion approximation for flood routing

E222

Skeels, C.P. & Samuels, P.G. (1989), Stability and accuracy

analysis of numerical schemes modelling open channel flow, in
Maksimovic & M. Radojkovic, eds, Computational Modelling
C.
and Experimental Methods in Hydraulics (HYDROCOMP 89),
Elsevier.

Review

JDF

Table 1.3: References on flood & wave propagation theoretical and computational

Reference

Notes

The full equations for wave propagation and flood routing

Cunge, J.A., Holly, F.M. & Verwey, A. ( 1980) Practical Aspects of
Computational River Hydraulics, Pitman, London.

The best explanation of this field

Liggett, J.A. (1975) Basic equations of unsteady flow, Unsteady Flow

in Open Channels, K.Mahmood & V.Yevjevich (eds), Vol.1, Water Resources Publications, Fort Collins, chapter 2.

A little disappointing, but the next

best explanation

Liggett, J.A. & Cunge, J.A. (1975) Numerical methods of solution

of the unsteady flow equations, Unsteady Flow in Open Channels,
K.Mahmood & V.Yevjevich (eds), Vol.1, Water Resources Publications,
Fort Collins, chapter 4.
The advection-diusion approximation for flood routing
Price, R.K. (1985) Flood Routing, Developments in Hydraulic Engineering, P.Novak (ed.), Vol.3, Elsevier Applied Science, chapter4,
pp.129173.

The best overview

Dooge, J.C.I. (1986) Theory of flood routing, River Flow Modelling

and Forecasting, D.A. Kraijenho & J.R. Moll (eds), Reidel, chapter3,
pp.3965.

A good general study

Numerical methods fundamentals

Liggett, J.A. & Cunge, J.A. (1975) Numerical methods of solution
of the unsteady flow equations, Unsteady Flow in Open Channels,
K.Mahmood & V.Yevjevich (eds), Vol.1, Water Resources Publications,
Fort Collins, chapter 4.
Noye, B.J. (1976) International Conference on the Numerical Simulation of Fluid Dynamic Systems, Monash University 1976, NorthHolland, Amsterdam; Noye, B.J. (1981) Numerical solutions to partial
dierential equations, Proc. Conf. on Numerical Solutions of Partial
Dierential Equations, Queens College, Melbourne University, 23-27
August, 1981, B.J. Noye (ed.), North-Holland, Amsterdam, pp.3137;
Noye, B.J. (1984) Computational techniques for dierential equations,
North-Holland, Amsterdam; Noye, J. & May, R.L. (1986) Computational Techniques and Applications: CTAC 85, North-Holland, Amsterdam.

All oer a simple introduction to finite dierence methods;

Smith, G.D. (1978) Numerical Solution of Partial Dierential Equations, Oxford Applied Mathematics and Computing Series, Second Edn,
Clarendon, Oxford.

A more detailed introduction to finite dierence methods

Morton, K.W. & Baines, M. (1982) Numerical methods for fluid dynamics, Academic; Morton, K. & Mayers, D. (1994) Numerical solution of partial dierential equations : an introduction, Cambridge;
Morton, K.W. (1996) Numerical solution of convection-diusion problems, Chapman and Hall, London; Richtmyer, R.P. & Morton, K.W.
(1967) Dierence Methods for Initial Value Problems, Second Edn, Interscience, New York.

All are rather more comprehensive, describing some more general

methods

Table 1.4: Useful references

Chapter 2

River hydraulics

2.1 A note on terminology in the English language:

Throughout there will be use of apparently different terms for the single main quantity that we will be talking about
the stream. Let us be clear what is meant see Table 2.1. The material in this course, which is called "river
engineering", is applicable to all of those, as is the field of "open channel hydraulics", or "Gerinnehydraulik" in
German.
Open channel

The generic term for a flow of water with a free surface

Stream

Generic term, used in passing when a certain shortness & vagueness is necessary

Canal

A constructed channel, generally prismatic and with a relatively firm boundary

Channel

A generic term for an extended depression along which water passes

Waterway

Stream on which boats navigate, may be natural or constructed

Natural streams (decreasing size)

River

A natural channel that is large or medium-sized

Creek

USA and AUS a small river, but in England a tidal inlet, rather dierent

Rivulet

Small river

Brook

Smaller stream

Beck

Northern England where the Angles and Saxons landed; from their word Bach

Ditch, Furrow

Small channel, probably excavated

Gutter

The channel at the side of a road pavement

Table 2.1: Terminology various names for what is essentially the same thing

2.2 Summary
Initially we consider the two equations for mass conservation, where almost no essential approximations are necessary:
1. Conservation of mass in the flow in the channel: one of the few exact equations in hydraulics (the hydrostatic
pressure equation is another one).
2. Conservation of mass of sediment the Exner equation (Felix Maria von Exner-Ewarten , A, 1876-1930): also
exact for uniform materials; the Wahl-Wiener who presents these lectures will defend this equation against
attacks by foreign barbarians during the last 10 years.
Then for momentum conservation, consider the hierarchy of river models, starting with the Navier-Stokes equations, where the models become simpler as we pass down the list:
1. The Navier-Stokes equations: the three-dimensional partial differential equations of fluid mechanics; solving
them is not feasible for most river problems.
2. Three-dimensional turbulent flow equations: where the turbulence is modelled. Solving them is only slightly
more feasible for most problems.
3. Conservation of moment of momentum, allowing for vertical variation of pressure and velocity: introduced by
Steffler in Canada, but it has never been exploited. One could develop a 2-D river model in which secondary
currents were described, and which could simulate differential erosion on both sides of the river.
4. Boussinesq approximation: essentially a one-dimensional model, but which allows for variation of pressure
across the flow due to curved streamlines.
5. Two-dimensional vertically-averaged river model: The real geometry of a meandering river can be included,
10

but no secondary currents are possible and erosion predictions are unreliable.
6. One-dimensional model using curvilinear co-ordinates: The long-wave equations for rivers that are curved in
plan your lecturer ...
7. One-dimensional model for straight streams: the long-wave equations, shallow-water equations, "Saint-Venant"
equations, the basis of most hydraulics. The solutions of these equations are much misunderstood, and some
of the most elementary deductions and interpretations are wrong.
8. The low-inertia approximation: A rather simpler but still a good approximation
9. Advection-diffusion flood routing
10. Muskingum-Cunge routing
11. Kinematic wave routing
12. Black-box flood routing
Figure 2.1 shows how all the theories relate to each other. The arrows generally show the direction of increasing
approximation. The assumptions made in each case are shown as text without a box.

11

Real stream
Boussinesq approximations,
non-hydrostatic, can describe
transition between sub- and
super-critical flow

Assume pressure hydrostatic

Two-dimensional equations, possibly
including moment of momentum, to
include secondary flows, not yet
established
Vertically averaged
Two-dimensional equations, no
secondary flows

Assume stream curvature small

Assume stream straight, no cross-stream variation

One-dimensional long wave equations

for curved streams, e.g. 2.3.1.3, eqn
(2.4)

One-dimensional long wave equations,

eqns (2.22) and (2.23)

Characteristic formulation, 2.3.12,

reveals speed of front of disturbances
Finite-dierence-based numerical
methods, Preissmann Box Scheme
Characteristic-based
numerical models
Assume small disturbances

Assume flow and disturbances both slow

Linearised Telegraphers equation, 2.3.13.1,

eqn (2.84), reveals nature of propagation of
disturbances

Slow-change/slow-flow approximation, 2.3.8,

eqn (2.67), simple equation & computational
method

Both assumptions: small disturbances and slow flow

Linearised, and slow-change/slow-flow,
Advection-diusion equation, 2.3.9, eqn
(2.70), reveals nature of most disturbances
Assume diusion small
Interchange of time & space dierentiation,
Muskingum-Cunge routing
Assume diusion zero
Kinematic wave approximation

Figure 2.1:

Hierarchy of one-dimensional open channel theories and approximations

2.3 The one-dimensional equations of hydraulics

Initially we will develop a model using the mass conservation equations for water and for soil, and for momentum conservation we will use approximation 7, which is relatively simple to obtain, and provides many useable
results from the one-dimensional long-wave equations. The approach is to consider the conservation of mass and
streamwise momentum along the stream. In developing the model for water and sediment flow in a river, we do not
consider details of motion in the plane across the stream. It will be found that this approach requires surprisingly
few essential approximations the model is actually a good one, rather better than is commonly believed.
There may be one surprise there will be no mention of energy, other than to show how it contributes nothing
extra, and that momentum is to be preferred.
Initially we make the traditional approximation that all rivers are straight. Later we will relax this and consider
meandering streams.

12

In the past there has been little attention paid to defining co-ordinates or a control volume. Many derivations use
as the fundamental space variable x some sort of distance along the channel bed, which is ambiguous for channels
of arbitrary shape, and what is worse, this curved co-ordinate ideally requires special treatment and consideration
of the curvature of the axis. It is easier instead to use cartesian co-ordinates, for which we use x the horizontal
distance along the stream, y the horizontal transverse co-ordinate, and z the vertical, relative to some arbitrary
origin.

2.3.1

Mass conservation of water and soil

ix

A + A

Q + Q

z
Qb
y

Ab
Ab + Ab

Qb + Qb

Figure 2.2:

Elemental length of channel showing control volumes

Consider Figure 2.2, showing an elemental slice of channel of length x with two stationary vertical faces across
the flow. It includes two different control volumes. The surface shown by solid lines includes the channel cross
section, but not the moveable bed, and is used for mass and momentum conservation of the channel flow. The
surface shown by dotted lines contains the soil moving as bed load and extends down into the soil such that there
is no motion at its far boundaries. Each is modelled separately, subject to a mass conservation equation, and each
to a relationship that determines the flux.
The fluid in the channel might carry a suspended sediment load. Here, to good approximation, the concentration
of suspended sediment will be considered constant so that the density of the fluid-solid suspension is constant,
the same as for any fluid entering the control volume from rainfall, seepage, or tributaries, so that we can consider
conservation of volume rather than of mass.
Physically-oriented derivation
This is a simpler and more intuitive derivation. Afterwards, a more mathematical derivation will be presented, as
it provides something of a guide for when we consider momentum.
On the upstream vertical face at any instant, there is a volume flux (rate of volume flow) Q, and on the downstream
face Q + Q, so that
Net volume flow rate of fluid leaving across vertical faces = Q =

Q
2
x + terms in (x) .
x

If rainfall, seepage, or tributaries contribute an inflow volume rate i per unit length of stream, the volume flow
rate of this other fluid entering the control volume is i x. If the sum of these two contributions is not zero, then
volume of fluid is changing inside the elemental slice, so that the water level will change. The rate of change
with time t of fluid volume inside the control volume can be expressed most simply in terms of rate of change of
cross-sectional area as x A/t. Equating this to the net rate of fluid entering the control volume, dividing by
x and taking the limit as x 0 gives
A Q
+
= i.
(2.1)
t
x
Remarkably for hydraulics, this is almost exact. The only approximation has been that the channel is straight.
The second flow, the bed load, is composed of larger soil particles. The interstices between them are assumed to be
occupied by the same fluid as in the channel. In the absence of any detailed mechanism of transport or dilation, it
13

is assumed that the porosity of the bed load is constant. The composite bulk density of the bed load b is composed
of the main solid constituent of the bed load, the water in the pores, and the suspended particles in the pore water,
and it is assumed to be constant. The bed has a cross-sectional area Ab , the bulk volumetric flow rate is Qb , and
there is an inflow of mass rate m
i per unit length, possibly due to deposition or erosion. Mass conservation is
calculated following the same reasoning for this control volume as for that of the channel, giving
Ab Qb
m
i
.
+
=
t
x
b

(2.2)

The volume transport rate used here is the bulk flow rate; it is related to the volume flow rate of solid matter Qs used
in transport formulae, by Qs = Qb (1 ), where is the porosity of the bed load. This is a slight generalisation
of Exners law.
Derivation using integral formulation with vectors
There is insight to be gained by repeating the derivation using a general integral formulation for the conservation of
any quantity. Here it will be considered initially for mass, later it will be used for momentum, when it is simpler to
use this approach from the start. Consider the mass conservation equation for an arbitrarily moving and deforming
control surface CS and control volume CV (?, 3.2 & 3.3):
Z
Z
d
dV +
ur
n dS = 0,
(2.3)
dt
CV
CS
{z
}
| {z }
|
Total mass in CV

Rate of flow of mass

across boundary

where t is time, dV is an element of volume, ur is the velocity vector of the fluid relative to the local element of the
is a unit vector with direction normal to and directed outwards,
control surface, which is possibly moving itself, n
and dS is an elemental area of the surface, as are shown on figure 2.3. The quantity urn = ur
n is the component

ur
urn = ur
n
dS
Figure 2.3: The normal component of fluid velocity relative to the control surface is urn = ur
n, hence the rate of volume
n dS
transport is ur

of relative velocity normal to the surface at any point. It is this velocity that is responsible for the transport of any
quantity across the surface, mass or momentum in our case.
Now consider the control surface to be the elemental slice in figure 2.2, bounded by two fixed vertical planes
transverse to the channel, and bounded above by the free surface, which is free to move, and below by the soil
surface, which is similarly free to move. An element of volume in the first integral in equation (2.3) is dV =
x dA, where dA is an element of cross-sectional area. The whole term becomes
Z
d
A
x
dA = x
,
dt
t
A

where it is necessary to use partial differentiation as A is a function of x as well. We have expressed the rate of
change of mass in the control volume in terms of the rate of change of cross-sectional area.
Considering the second integral in equation (2.3), on the free surface we have chosen the control surface to coincide
with the free surface of the fluid, so that particles on the free surface remain on the free surface, and the relative
n 0 everywhere on that surface. Similarly particles on the interface between water
normal velocity urn = ur
and soil, or on any solid boundaries in the flow also define the control surface, and there is also no contribution to
mass or momentum transport. We include the usually poorly-known inflow from rain, seepage or tributaries in an
ad hoc manner, by just writing the contribution as x i, negative as it is an inflow.

On the upstream face of the control surface, which is fixed in space and through which fluid moves, ur
n = u ,
14

where u is the x component of fluid velocity, so that the contribution due to flow entering the control volume is
Z
u dA = Q,
A

where Q is the total volume flow rate or discharge in the channel. Similarly the downstream face contribution is
+ (Q + Q), where we use a Taylor expansion to write it as

Q
(Q + Q) = Q +
x + . . . .
x

Combining the contributions from the rate of change of mass in the control volume and the various contributions
which cross the control surface, dividing by x and taking the limit x 0 we have the unsteady mass
conservation equation
A Q
+
= i,
t
x
the same as equation (2.1) obtained by simpler but less-general means.
Exactly the same procedure would be followed in the case of the bed-load, which would give equation (2.2).
Streams curved in plan
z
Centre of curvature: n = r = 1/

Surface mid-point: n = nm

nR

nL
Centroid: n = n

Figure 2.4:

Section of curved river, showing important dimensions and axes

? have considered streams curved in plan (i.e. almost all rivers!) and obtained the result:

nm A Q
+
= i,
(2.4)
1
r
t
s
where nm is the transverse offset of the centre of the river surface from the curved streamwise reference axis s,
which generally follows the path of the river and r is the radius of curvature of that axis. Usually nm is small
compared with r, and the curvature term is a relatively small one, however for a highly meandering river there will
be an effect on flood propagation velocities.
Upstream Volume
The mass conservation equation (2.1) suggests the introduction of a function V (x, t) which is the volume upstream
of point x at time t, such that for the channel flow
Z x
V
V
i dx0 Q.
(2.5)
= A and
= Q (x0 , t) +
x
t
x0

The derivative of volume with respect to distance x gives the area, as shown, while the time rate of change of
volume upstream is given by the rate at which the volume is increasing due to inflow, minus the rate at which
volume is passing the point and therefore no longer upstream. Substituting for A and Q into equation (2.1),
shows that it is identically satisfied. The addition of Q (x0 , t) has been suggested by Fatemeh Soroush (Personal
communication, 2011), and gives a definition which is an improvement on an earlier one used by the lecturer. In
the case of the bed load, similarly introducing the upstream volume of the bed load Vb (x, t) such that
Z x
Vb
Vb
m
i 0
(x ) dx0 Qb ,
(2.6)
= Ab and
= Qb (x0 , t) +
x
t
b
then the mass conservation equation (2.2) is identically satisfied. The use of these V and Vb will reduce the
15

total number of equations used from four to two, and enable the reduction of the combined problem to a single
third-order differential equation.
Use of stage instead of cross-sectional area
We usually work in terms of water surface elevation stage , which is easily measurable and which is practically
more important. We make a significant assumption here, but one that is usually accurate: the water surface is
horizontal across the stream. Now, if the surface changes by an amount in an increment of time t, then the
area changes by an amount A = B , where B is the width of the stream surface. Taking the usual limit of small
variations in calculus, we obtain A/t = B /t, and the mass conservation equation can be written
B

Q
+
= i.
t
x

(2.7)

The discharge Q could be written as Q = U A, where U is the mean streamwise velocity over a section, and
substituted into this. However, we believe the discharge to be more practical and fundamental than the velocity,
and that will not be done here. When it comes to sediment transport, the velocity is important, but it can always be
obtained from the discharge and the area.

2.3.2

Momentum conservation equation for channel flow

The equation
The conservation of momentum principle is now applied to the mixture of water and suspended solids in the main
channel for a moving and deformable control volume (?, 3.2 & 3.4):
d
dt

u dV +

CV

u ur
n dS = P,

(2.8)

CS

where u is the fluid velocity, and P is the force exerted on the fluid in the control volume by both body forces,
which act on all fluid particles, and surface forces which act only on the control surface. Note how similar the left
side is to the left side of the mass conservation equation (2.3). It was a conservation equation for mass, integrating
the mass per unit volume throughout the control volume and its flux across the control surface. Equation (2.8)
is a conservation equation for momentum, obtained by integrating the contributions of fluid momentum per unit
volume, the quantity u.
Contributions to force P
1. Body force: for the straight channel considered, the only body force acting is gravity; this force per unit mass is
g, the gravitational acceleration vector. The
R total force on the fluid in the control volume is the integral of this
with respect to mass, dm = dV , giving CV g dV = gV . We will consider only the x-component of the
momentum equation, which have chosen to be horizontal, as g only has a component in the z direction, there
will be no contribution from gravity to our equation! The manner in which gravity acts is to cause pressure
gradients in the fluid, giving rise to the following term, due to pressure variations around the control surface.
2. Surface forces: we separate these into pressure and shear forces, which act normally and tangentially to the
surface respectively.
a. Pressure the direction of the pressure force on the fluid at the control surface is
n, where n
is the
outward-directed normal; its local magnitude is p dS, where p is the pressure and dS an
elemental
area
R
of the control surface, as in figure 2.3. Hence, the total pressure force is the integral CS p
n dS. This
is difficult to evaluate for arbitrary control surfaces, as the pressure and the non-constant unit vector have
to be integrated over all the faces. A simpler derivation is obtained if the term is evaluated using Gauss
Divergence Theorem:
Z
Z
p
n dS = p dV,
CS

CV

where p = (p/x, p/y, p/z), the vector gradient of pressure. This has turned a complicated
surface integral into a volume integral of a simpler quantity.
b. Shear forces: there is little that we can say that is exact about the shear forces. We will consider these when
we make several practical assumptions.

16

Collecting terms
Substituting the force contributions into the momentum equation (2.8) and taking just the x component:
Z
Z
d
u dV + u ur
n dS =
dt
CV
CS
{z
}
{z
} |
|
Unsteady term

Fluid inertia term

p
dV + Horizontal component of shear force
x
on control surface
CV
|
{z
}
| {z }

Pressure gradient term

(2.9)

Horizontal shear force term

This expression, for an arbitrary control surface, has required no additional assumptions or approximations. Now,
however, we consider the section of channel with the control surface as shown in figure 2.2. To obtain a useable
expression it is necessary to make hydraulic approximations for all terms after the first.
Hydraulic approximations
1. Unsteady term
The element of volume, as for the mass conservation equation, is dV = x dA, and the term contribution can
be written
Z
d
Q
x
u dA = x
,
(2.10)
dt
t
A

where the integral A u dA has a simple and practical significance it is just the discharge Q, so that the
contribution of the term can be written simply as shown, but where again it has been necessary to use the
partial differentiation symbol, as Q is a function of x as well. No additional approximation has been made in
obtaining this term. It can be seen that the discharge Q plays a simple role in the momentum of the flow.
2. Fluid inertia term
R
The second term on the left of equation (2.8) is
u ur
n dS, has its most important contributions from
CS

the stationary vertical faces perpendicular to the main flow. If there is also fluid entering or leaving from
rainfall, tributaries, or seepage, there are contributions over the other faces, which we shall include below in an
approximate manner. Considering these contributions separately:
a. Stationary
vertical faces: on the upstream face, ur
n = u, giving the contribution to the term of
R
A u2 dA. The downstream face at x + x has a contribution of a similar nature, but positive, and
where all quantities have increased over the distance x. The net contribution, the difference between the
2
two, after neglecting terms like (x) , can be written
Z

u2 dA.
x
x
A

+u0 ,
In fact, the u term means that there are contributions from the turbulent fluctuations. If we write u = u
0
2
where u
is the mean value in time, and u is the turbulent fluctuation, then substituting into u and taking
the mean value in time, using the fact that u0 = 0 by definition, the term becomes
Z

u
2 + u02 dA.
x
x
2

It can be shown that this is the only term in the momentum equation (2.9) where turbulence plays a role. To
evaluate this integral it is necessary to have a detailed knowledge of the flow and turbulence distribution over
the cross-section. As we have not solved the Navier-Stokes equations or the 3-D turbulent flow equations,
that information is not available, and so here we make the first major hydraulic assumption. The crude but
simple approach is to assume that u
2 and u02 are both proportional to the square of the mean velocity over
2
the section (Q/A) , and to write the integral in terms of a Boussinesq coefficient :
Z

Q2
u
2 + u02 dA
.
A

Traditionally the coefficient was used to make some allowance for the non-uniformity of velocity distribution. ? suggested that it should be modified to include also the effects of turbulence as shown. A typical
value is = 1.05. The contribution to the equation is then simply but approximately.

Q2
x

.
(2.11)
x
A
17

It will be shown below, that the whole term is of a relative magnitude that of the square of the Froude number, and its effects are small in many situations. It is useful to retain the , unlike many presentations that
implicitly assume it to be 1.0, as it is a signal and reminder to us that we have introduced an approximation.
b. Lateral momentum contributions: the remaining contribution to momentum flux is from inflow such as
tributary streams, or, although less likely, from flow seeping in or out of the ground, or from rainfall. In
obtaining the mass conservation equation above, these were lumped together as an inflow i per unit length,
such that the mass rate of inflow is i x, (i.e. an outflow of i x) and if this inflow has a mean
streamwise velocity of ui before it mixes with the water in the channel, the contribution is
x i i ui ,

(2.12)

where i is the Boussinesq momentum coefficient of the inflow. The term is unlikely to be known accurately
or to be important in most places.
3. Pressure gradient term
R the hydrostatic approximation
The contribution is CV p/x dV , the volume integral of the streamwise pressure gradient. The hydraulic
approximation has a problem, because we have not attempted to calculate the detailed pressure distribution
throughout the flow. However, in most places in most channel flows the length of disturbances is much greater
than the depth, so that streamlines in the flow are only very gently curved, and the pressure in the fluid is
accurately given by the equivalent hydrostatic pressure, that due to a stationary column of water of the same
depth. Hence we write for a point of elevation z,
p = g Depth of water above point = g( z),
where is the elevation of the free surface above that point. The quantity that we need is the horizontal pressure
gradient p/x = g /x, and so the streamwise pressure gradient is entirely due to the slope of the free
surface. The contribution is
Z
Z

dV x g
dA x gA ,
(2.13)
x
x
x
A

where any variation with y has been ignored, as the surface elevation usually varies little across the channel,
and so /x is constant over the section and has been able to be taken outside the integral, which is then
simply evaluated.
4. Horizontal shear force term
This term is probably the least well-known of all, and yet it is important in the equation. Other presentations
assume that the slope of the channel is small, but that is unnecessary, and we shall not be forced to make
it. The results will be applicable to steep spillways, inter alia. However, we are about to adopt a significant
approximate model of the process of resistance to motion. ? recommended use of the Weisbach formulation for
bed resistance, but that suggestion has been almost entirely ignored by the profession. The Gauckler-ManningStrickler equation continues to be widely used, with the disadvantages noted by the Committee. Using the
Weisbach formulation here provides additional insights into the nature of the equations and some convenient
quantifications of the effects of resistance.
Consider the equation for the shear force on a pipe wall e.g. ?:
2
(2.14)
V ,
8
where is the dimensionless Weisbach resistance factor, and V is the spatial and temporal mean velocity along
the pipe. If the pipe is uniform and long, then is constant everywhere, both around the perimeter of the flow
as well as along the pipe. The channel is a more complicated situation.
Consider a vertical rectangular elemental prism shown in figure 2.5. The bed has a slope angle of in the x
direction, which in many hydraulic problems is small. The transverse slope may be quite large however. The
total shear force is s P , where s is locally downstream along the bed and P is an element of perimeter.
This gives us

Shear force on prismatic element = V 2 s P,

(2.15)
8
in which the mean velocity tangential to the boundary V over the prism is assumed to be related to U , the
mean x-component of velocity in the elemental prism, by V = U/ cos , as suggested in the figure. Now, the
horizontal component of shear force, which is what we require, is obtained by multiplying the term (2.15) by
cos = x/s, giving
=

Horizontal shear force on prismatic element =

18

2
V x P.
8

y
Slope /x

s
(a) Side elevation
Figure 2.5:

(b) Looking down the channel

Element of channel showing dimensions and shear force from the bed

Summing around the perimeter, expressing it as an integral in perimeter distance, gives

Z
x
Total horizontal shear force on control surface =
V 2 dP.
8
P
Now we make some gross approximations. We replace V , the mean velocity parallel to the bed, by U/ cos ,
as suggested in the figure, and if we write 1/ cos2 = sec2 = 1 + tan2 the integral becomes
Z

x
U 2 1 + tan2 dP.
Total horizontal shear force on control surface =
8
P
As we usually do not know the variation of or the distribution of U or even in many cases the variation of the
bed topography to give , we write it as the total perimeter P multiplied by a product of the mean values of the
three terms in the integrand:
2
Z

Q
U 2 1 + tan2 dP
1 + S2 P,
A
P
where the quantities shown with tildes are mean values for the section: is the mean value of the dimensionless
resistance factor around the perimeter, and the mean slope term S is the mean downstream slope evaluated
around the wetted perimeter. In almost all river situations, where S is usually poorly known, it is so small that
it plays no role here. In situations such as on steep spillways, where it is important, it is actually well-defined.
Collecting terms gives

x Q |Q|
Total horizontal shear force on control surface =
2 1 + S2 P,
8
A
where the Q2 term have been replaced by Q |Q|, so that the direction of the stress is opposite to the flow,
allowing also for tidal flow situations where the flow reverses:
It is convenient for all subsequent work to introduce the symbol :

=
(2.16)
1 + S2 ,
8
so that the result can be written
Q |Q|
(2.17)
Total horizontal shear force on control surface = x 2 P.
A
Collection of terms and discussion
Now all contributions from the hydraulic approximations to terms in equation (2.9) are collected, using equations
(2.10), (2.11), (2.12), (2.13), and (2.17), and bringing all derivatives to the left and others to the right, all divided
by x, gives the momentum equation:

Q2

Q |Q|
+ i i ui
(2.18)
+

+ gA
= P
t
x
A
x
A2
19

At this stage the non-trivial assumptions in the derivation are stated, roughly in decreasing order of importance:
1. Resistance to flow is modelled by an empirical law, such that it is proportional to wetted perimeter and mean
flow velocity squared. The law can incorporate finite effects of viscosity inasmuch as that affects the resistance
coefficient, but no attempt is made to incorporate viscous or turbulent shear terms in the momentum balance.
The Navier-Stokes equations are not being used.
2. All surface variation is sufficiently long that the pressure throughout the flow is given by the hydrostatic pressure
corresponding to the depth of water above each point.
3. Effects of curvature of the stream course are ignored. Its effects are of the order of the ratio of the width of the
stream to its radius of curvature.
4. In the momentum flux term the effects of both non-uniformity of velocity over a section and turbulent fluctuations are approximated by a momentum or Boussinesq coefficient. Nowhere else is any special treatment for
turbulence necessary, so we have developed a model including turbulence.
5. Surface elevation across the stream is constant.
6. Suspended load concentration is a constant over the section at a particular time and place.
In the resistance term, the only place necessary, allowance has been made for the stream being on a finite slope
no limitation on slope has been required.
Expressing A/x in terms of surface slope /x and mean bed slope S
B
At x
At x + x

A
Z
y

Figure 2.6:

Two channel cross-sections separated by x

In the momentum equation (2.18), if expanded, there would be a mixture of derivatives of area A/x and surface
elevation /x. We choose to work with the more practical quantity of surface elevation, and in so doing, are
forced to consider the bottom geometry in greater detail.
The cross-section of a river in Figure 2.6 shows how ambiguous and possibly non-unique the concept of the
bottom of the stream may be. In a distance x the surface elevation may change by an amount as shown,
so that the contribution to the increase in cross-section area A is B , where is usually negative as
the surface drops downstream.
The change in the bed is Z, which in general varies across the section, with
R
contribution to A of B Z dy, the area between the solid and dotted lines on the figure corresponding to
the bed at the two locations. The minus sign is because, if the bed drops away and Z is negative, as usual, the
contribution to area increase is positive. Combining the two terms,
Z
A = B Z dy
(2.19)
B

In practice the precise details of the bed are rarely known. For the latter contribution, the integral of the change in
bed elevation across the stream, we introduce the symbol S for the mean downstream bed slope across the section
such that
Z
1
Z
S =
dy,
(2.20)
B
x
B

where the negative sign has been introduced such that in the usual case when Z decreases with x, this mean
downstream bed slope at a section is positive. In an important problem where bed details might be known, this
can be evaluated. In the usual case where bed topography is poorly known, a reasonable local approximation or
assumption is made. Using equations (2.19) and (2.20) we can write
A = B + B S x,
20

where in a distance x the mean bed level across the channel then changes by S x under the water. In the
rare case where the sides of the stream are vertical diverging or converging walls, an extra term would have to be
included. Taking the usual calculus limit, we obtain

A
=B
+S ,
(2.21)
x
x
which might have been able to have been written down without the mathematical details.

2.3.3

Forms of the governing equations

In equation (2.18) weignore the

last term, the inflow momentum term i i ui as it is small and poorly known.
Expanding the term Q2 /A /x, gives a term in /x which we also neglect, as we usually know little about
how varies. There are now x-derivatives of both A and in the equation, which are not independent. We use
equation (2.21) to eliminate first one and then the other to give two alternative forms of the momentum equation
governing flows and long waves in waterways. In both cases, we restate the corresponding mass conservation
equation, using (2.1) and (2.7), to give the pairs of equations:
Formulation 1 Long wave equations in terms of area A and discharge Q
Eliminating /x gives the equations in terms of A and Q:
A Q
+
= i,
t
x

Q Q
gA
Q2 A
Q |Q|
Q
.
+ 2
+
2
= gAS P
t
A x
B
A
x
A2

(2.22a)
(2.22b)

Formulation 2 Long wave equations in terms of stage and discharge Q

Now eliminating A/x, but retaining A in all coefficients, as it can be calculated in terms of :

1 Q
i
+
= ,
t
B x
B

Q |Q|
Q Q
Q2 B
Q2 B
Q
.
+ 2
+ gA 2
= 2 S P
t
A x
A
x
A
A2

(2.23a)
(2.23b)

Nature of equations
Each of the two formulations is in the form of a pair of equations that can be written as a vector evolution equation
u
u
+C
= r (u) ,
t
x
where u is the vector of unknowns, for example, [, Q], C is a 2 2 matrix with algebraic coefficients, and r is
the vector of right side terms, due to inflow, slope, and resistance.
It can be shown that the system is hyperbolic, although this mathematical terminology seems not very useful for us.
The implication of that is that solutions are of a wave-like nature. We will see that the behaviour of disturbances is
more complicated than we might expect or is often stated. This arises because the right sides are functions of the
dependent variables, that we have written here as r (u). In particular we will see that a common interpretation of
the system in terms of characteristics, with the solution that of travelling waves with simple properties, is incorrect.
The solution is actually more complicated: disturbances travel at speeds which depend on their length, and show
diffusion as well.
Comparison with previous presentations
The questions arise, "how does this relate to other presentations of results?", and, "has the present approach actually
done anything new?" It is strange, but there
are almost no presentations in textbooks of the equations in this
standard form. Most retain a term Q2 /A /x. As we have seen, the expansion of this is non-trivial. Textbooks
also present results in terms of a depth-like quantity h (called "the depth"), which is presumably only useful for
canal problems where the bottom is flat, and the depth at the centre, can be defined. We will not deal with
it, as boundary conditions are almost always expressed in terms of surface elevation . Presentations able to be
recommended include those of ? and ?.
The above presentation has given a couple of additions, which might be useful in practice:
1. The use of the Boussinesq momentum coefficient , reminding us that the term is being modelled approximately only. Some (?, ?) have included it, but also included equations without it. The lecturers claim that
21

it is better to include it, rather than implicitly assuming it to be 1.0000 might be viewed cynically, when later
in the lectures it will be suggested that terms in are of magnitude of the Froude number, and can usually be
neglected anyway equivalent to making the = 0 "approximation"!
as defined in equation (2.20), in the right sides of the momentum
2. The use of mean downstream bed slope S,
equations (2.22b) and (2.23b): this is the only way in which
variable
bed topography has entered explicitly.

Textbook presentations that do not expand the term Q2 /A /x do not confront this problem. In practice, however, S is often be calculated very approximately as, for example, even the water surface or even
surrounding land elevation difference between two stations divided by the distance between them.
3. The resistance term: in equations (2.22b) and (2.23b) it appears as P Q |Q| /A2 ,which has a relatively
simple interpretation as a dimensionless coefficient multiplied by the perimeter (over which the resistance
acts), multiplied by the mean velocity in the stream, squared. This clearly follows from the approximate
physical law assumed. There is no limitation on slope: these are long wave approximations, not shallow slope
approximations.
Alternative use of energy conservation
The question also arises how would the use of energy conservation relate to this work? Consider the Energy
conservation equation for the flow in the body of the channel (?, 3.6):
Z
Z

e dV + (p + e) ur
n dS = Losses,
(2.24)
t
CV

CS

where e is the internal energy

per unit mass of fluid, which in hydraulics is the sum of potential and kinetic energies
e = gz + u2 + v 2 + w2 /2. Proceeding in a manner similar to the momentum equation above, we obtain
0
gB
+
t
2 t

Q2
A

+
x

Q3
Q2 |Q|
,
gQ + 1 2 = gHi i e P
2A
A3

(2.25)

where 0 and 1 are energy transmission coefficients defined by

n =

1
(Q/A)

n+2

(u2 + v 2 + w2 ) un dA,

Hi is the head of the incoming flow, and e is a dimensionless energy loss coefficient. Equation (2.25) has two
time derivative terms (which are clearly related to the rate of change of potential and kinetic energy at a section).
It is not as simple as the momentum conservation equation, which is in terms of Q/t. To eliminate /t and
A/t we use two different forms of the mass conservation equation, then, neglecting inflow and 1 /x terms
gives

Q |Q|
Q 0 + 31 Q Q
Q2 B
Q2 B
.
+
+ gA 1 2
= 1 2 S e P
0
t
2
A x
A
x
A
A2
This can be compared with the momentum equation (2.23b) for no inflow:

Q |Q|
Q Q
Q2 B
Q2 B
Q
.
+ 2
+ gA 2
= 2 S P
t
A x
A
x
A
A2
It can be seen that the structure of the two equations is the same, but they are slightly different in numerical
detail. The leading term of the energy formulation is 0 Q/t; as 0 1.05, this gives a slightly different time
scale from the momentum equation. As the coefficients 0 , 1 and are all just greater than 1, the numerical
coefficients of the other terms are also just slightly different. In general, e is different from . It is the assertion
in these lectures, however, that the processes of energy dissipation, as given by e , are more complicated than the
boundary stresses that give rise to the coefficient . The momentum losses of the latter are a local phenomenon, to
do with the local nature of the boundary and flow at the boundary only, whereas energy losses are more complicated
and more distributed, in boundary layers, separation zones, vortices, and turbulent and viscous decay. Accordingly,
the coefficient should be easier to quantify than e , and the momentum approach is to be preferred. It is also
possible that the momentum response of the flow to boundary changes would be faster than energy processes,
which consist of more complicated 3-D motions, and might take further downstream to adjust, as suggested in
figure 2.7. The momentum model is simpler, and so, following Occams Razor, we are inclined to follow that path.

2.3.4

Hydraulic resistance Weisbach-Chzy, Gauckler-Manning-Strickler

? started a well-known paper on resistance to flow with the words:

22

Equilibrium velocity profile

Equilibrium velocity profile

20 depths

Continuing adjustments
to energy cascade

Small bed forms

Figure 2.7:

Large bed forms

Hypothesised faster response of momentum than energy, here at a change in roughness

For many centuries, if not millennia, the resistance to flow in open channels has claimed the attention of
hydraulic engineers.
And then:
In this period of wide-spread scientific development, a review of existing knowledge with regard to openchannel hydraulics might appear to be as lacking in timeliness as it is in glamour. Yet a glance at publications on the subject during the past decade or so will reveal many a significant anomaly. Momentum and
energy analyses are at the same time over-simplified and confused with one another. Representation of parameters by symbols is mistaken for empirical formulation of functions. Flow formulas are sometimes said
to involve the Froude number when they do not, and yet the Froude number is as frequently ignored when it
is actually essential. Boundary texture and cross-sectional non-uniformity are often discussed without distinction. Roughness measures are mistaken for resistance coefficients and are permitted to reflect not only
viscous effects but even changes in depth. In a word, principles of mechanics that have proved useful in
other phases of fluid motion are still not generally applied to problems of open-channel flow,which perforce
remain subject to empirical solution all too often without even the guidance of physical reasoning.
He then went on:
However, it need only be recalled that the Chzy equation was unheard of two centuries ago; the Manning
type of formula, one century; the Reynolds-number resistance diagram only a half century; and the KrmnPrandtl relationships little more than a quarter century yet all are in regular use today.
Were he alive today, he would despair, for the apparent convergence in development that he wrote about, has all
seemed to fizzle out.
? wrote, warning about the use of the Gauckler-Manning-Strickler (G-M-S) formula:
Many engineers have become accustomed to using Mannings n for evaluating frictional effects in open
channels. At the present stage of knowledge, if applied with judgement, both n and are probably equally
effective in the solution of practical problems. The design engineer who prefers to use n in her/his computations should continue to do so, but he/she should recognize the limitations on her/his method ... It is
believed that experimental measurements of friction in open channels over a wide range of conditions are
better correlated and understood by the use of . Furthermore, is commonly used by engineers in many
other branches of engineering and probably provides the only basis for pooling all experience on frictional
resistance in both open and closed conduits. It is recommended, therefore, that engineering teachers and
research workers emphasize the use of the friction factor ...
This has been almost completely ignored. What the Task Force was advocating was the use of the Weisbach
approach, which we shall now consider.
Dimensional analysis
Consider the dimensional quantities governing the problem of the shear stress on the boundary of a stream which
is straight: shear, where all quantities are taken to be section-averaged ones; the problem of variation around
the boundary is too difficult for us at this stage: mean shear stress , discharge Q, cross-sectional area A, wetted
perimeter P , surface width B, mean roughness size k, gravitational acceleration g, density , and dynamic viscosity

23

. From the 9 variables and 3 dimensions, from the Buckingham Theorem there are 6 dimensionless quantities.
If we use as repeating variables , Q, and A, the remaining physical quantities whose importance we are to
examine, are as shown in the first column of table 2.2. The dimensionless results are shown in column 3. However
in most cases the physical significance of the terms is not clear. Multiplying by the factors in column 4 gives the
dimensionless numbers in column 5, with more-easily-understood significance.

Physical
quantity

Dimensionless
Factor
with , Q,
and A

Result

Significance

(Q/A)2

(Q/A)2

P/ A

BP/A

B/ A

1/2

B/P

k/ A

= k/(A/P )

1/3

gA/B
(Q/A)2

= 1/F 2 where F is the Froude number: eects

of gravity

P
Q

= 1/R where R is the Reynolds number: eects

of viscosity

g A
(Q/A)2

A
Q

= , as introduced in equations (2.14) and

(2.16)
Aspect ratio of section: roughly width/depth

Top width / Bottom width

Relative roughness: roughness size / mean depth
normal to bed

Table 2.2: Dimensional analysis of resistance in a channel

Resistance coefficient : The dimensional analysis has shown how important is the quantity , that was introduced in equations (2.14) and (2.16). Traditionally, however, originating from pipe flows, the quantity considered
is = 8, the Darcy/Weisbach coefficient. In pipe flows it is used as
L V2
,
(2.26)
D 2g
where H is the head loss in a pipe of length L and diameter D. (There is a point of view that g plays no role in
confined flow such as in a pipe and should not be used here instead, gH should be considered as a unity, the
energy loss per unit mass of fluid, such that the total rate of energy loss in the pipe is QgH).
H =

There is an interesting aside here, relating head loss and boundary shear stress. To do that, we actually
use the integral momentum theorem applied to the control volume in a length of uniform pipe of length L
inclined at an angle between stations 1 and 2, where here it is easier to consider a control volume with
faces across the pipe rather than vertical as we did for the channel:
D2
(p1 p2 )
+
{z 4 }
|

Net force due to pressure

As L sin = z1 z2 we obtain

D2
g
L sin
}
| 4 {z

Component of weight force

along pipe

| DL
{z }

Force due to
boundary shear

p1
p2
4 L
z1 +
z2 +
= H =
,
(2.27)
g
g
gD
and so the quantity that we refer to as head loss has actually been obtained from momentum considerations.
Comparing equations (2.26) and (2.27) we obtain equation (2.14)
2
V .
8
It would probably have been more fundamental to have started by assuming = V 2 , the use of /8
came about because of using head and a circular pipe.
=

Relative roughness k/(A/P ): It can be seen that the relative roughness has appeared in the form k/(A/P ),
24

where we might think of the quantity A/P as the mean depth measured from the river bed, the important quantity
in considering boundary resistance. Of course, many streams are wide, P B, such that A/P is equal to the
mean depth A/B.
Reynolds number R: Also in the last row of the table, the term which is the inverse of the Reynolds number, what
appears as U D in many definitions of Reynolds number, velocity multiplied by transverse dimension, in this case
is Q/P .
Remaining quantities: the remaining quantities 2 , 3 and 5 are unimportant for pipe flows. For channel flows
it is possible to use as a basis the extensive experimental results obtained in the early 20th century for pipes.
0.1

Laminar flow | Transition Zone

Completely turbulent
0.07

0.05
0.02
0.01

Weisbach
friction
factor

Relative
roughness

0.001

0.0001

0.01
0.00001

103

104

105
106
Reynolds Number R=UD/

107

108

Figure 2.8: The Moody diagram, showing graphically the solution of the Colebrook & White equation for as a function of
R with a parameter the relative roughness = d/D.

The most important experimental results for industrial pipes were obtained by Colebrook and White (?, ?). They
obtained results for as a function of relative roughness = k/D, where D is pipe diameter (A/P = D/4 for
circular pipes) and Reynolds number R = U D/. They presented an equation connecting , , and R, which was
implicit in , the Colebrook-White equation. In the pre-computer age, that implicit relation was difficult to solve
numerically. That has probably been the main reason that the approach has not been adopted. ? plotted the results
from that equation, as shown in Figure 2.8.There are, however, explicit solutions that are not as well-known as they
should be.
In channels, where the Reynolds number is usually so large that the flow is fully turbulent, ?obtained for on the
basis of a trapezoidal shape,
1
=
log2
10
2.032

1
12.

k
A/P

ln

1.3
k
1
12. A/P

(2.28)

where the rounded factor of 1/12. comes from a recommendation from ?. This simple approximation could have
been the basis for much hydraulics, as it gives an explicit formula for the resistance factor in terms of the relative
roughness k/(A/P ). Of course, determining the equivalent roughness k is a difficult problem, yet the knowledge
of the manner of variation with roughness gives us an important tool. This is plotted in figure 2.9. It is notable
just how little the resistance coefficient varies over a very wide range of roughnesses over a 100-fold roughness
variation the resistance varies by a factor of only 3.
A more general approximation has been obtained by ?, including the results from a number of open channel
experiments. Converting to natural logarithms here for simplicity, and rounding a coefficient ? gave the formula
here recommended for use in calculating the resistance coefficient:
=
ln2

1.33

0.9 .
1
P
k
+
2.
12. A/P
Q
25

(2.29)

0.06

0.04

0.02

0
1/1000
Figure 2.9:

1/100
Relative roughness k/(A/P )

1/10

Theoretical variation of with relative roughness

Yen stated that such a formula was applicable for = k/ (A/P ) < 0.05 and R = Q/P > 30 000. If one does
know the effective roughness size k, this provides the best way of calculating the resistance coefficient.

2.3.5

Steady uniform flow

The simplest model of a river is where the channel is prismatic, with a constant bed and surface slope S0 , and the
flow is steady (/t = 0) and uniform (/x = 0). In equation (2.16) we introduced the symbol , where for
typical streams, = /8, such that in the long wave equations (2.22b) or (2.23b), the resistance term appeared
as P Q |Q| /A2 . In this case of steady uniform flow, /x = S = S0 , all terms except the gravity and
resistance terms disappear from the momentum equation to give
Q2
= gAS0 ,
(2.30)
A2
giving the Weisbach-Chzy equation for the mean velocity U :
r
r
gA
A
Q
(2.31)
U=
=
S0 = C
S0 ,
A
P
P
p
p
where we have introduced
the Chzy coefficient C = g/ = 8g/. The most notable features are that the
p

velocity varies like A/P and like S0 .

P

There is an interesting deduction from equation (2.31). We calculate the square of the Froude number (widely used
in open channel hydraulics) from it:
U2
S0 B
F2 =
=
,
gA/B
P

and as for wide rivers B P we have the conclusion that F 2 S0 /, and as S0 is constant, and to first
approximation is constant, so that Froude number is roughly constant for any given stream, independent of the
flow rate.
Gauckler-Manning-Strickler formula
The resistance formulation most widely used in practice is the Gauckler-Manning-Strickler approach. To relate
this to our formulation above, we consider its fundamental expression for a steady uniform flow:
2/3 p
2/3 p
A
Q
1 A
U=
S0 = kSt
S0 ,
(2.32)
=
A
n P
P

in terms of the Manning coefficient n, which is dimensional, with units of L1/3 T, and the Strickler coefficient
simply kSt = 1/n. The use of the symbol kSt for the Strickler coefficient is an unfortunate typographical coincidence with the mean roughness size k. Comparing this equation (2.32) with the Weisbach-Chzy equation (2.31),
the main difference is that here the velocity is modelled as varying like (A/P )2/3 , rather than (A/P )1/2 . For the
two formulations (2.31) and (2.32) to agree, can be expressed as
1/3
1/3
g
g
P
P
= 2 = gn2
= 2
.
(2.33)
C
A
kSt A
26

0.25
0.20

n g
k1/6

Yen, eqn (**)

Bridging eqn (**)
Strickler eqn with **=0.128

0.15
0.10
0.05
0.00
1/1000

1/100
1/10
Relative roughness k/(A/P )

Figure 2.10: Theoretical variation of Mannings n with relative roughness,

showing its relative insensitivity to depth as measured by A/P

Now using this relation we obtain a theoretical expression for the variation of Mannings n. Solving (2.33) and
using = /8 gives
s
1/6
A
n=
,
(2.34)
8g P
into which we substitute Keulegans expression for , (2.28), or Yens equation (2.29) with infinite Reynolds
number to give the theoretical expression for the variation of the apparent Manning n with relative roughness
k/(A/P ):

n g
0.40

(2.35)

1/6 .
1/6
k
k
1
k
ln 12.
A/P
A/P
This is plotted in figure 2.10, and it can be seen that over a wide range of relative roughnesses expected in nature,

there is little variation of n g/k 1/6 ,leading to the widely-supposed justification for the G-M-S formula that the
value of n is relatively invariant with depth. As an approximation, from the figure we might take the value of

n g/k 1/6 to be constant at 0.125, leading to the approximate formula for n:

k 1/6
n 0.125 .
g

(2.36)

Such an equation, with a numerical value of g implicitly substituted, is known as the Strickler equation, where k
is taken to be the diameter of the bed material, such as d50 , the diameter of the bed material such that 50% of the
1/6
material by weight is smaller. If we substitute g = 9.8 ms2 we obtain n 0.040 d50 , where d50 is in metres,
typical of the versions of the Strickler equation presented by ?, but where some confusion follows from using d50
also in millimetres and feet.
This enables us to propose a formula by substituting the Strickler-type formula (2.36) into the G-M-S expression
(2.32) to give the formula for discharge in terms of known channel quantities:
Q
1
U=
=
A
0.125 k 1/6

A
P

2/3

p
gS0 .

(2.37)

In real rivers, however, n varies with depth it generally changes, as the nature of the roughness and the crosssection vary with depth. It is not a fundamental quantity it was not obtained from the boundary friction, although
equation (2.34) shows how it can be so related. It has no theoretical justification it is an empirical approach which
was convenient in a pre-computer age to to describe the problem approximately and conveniently, mimicking the
behaviour of the Weisbach form. If we accept that in more general problems the resistance coefficient varies, then
there is no real reason to use Mannings n rather than the more fundamental or .
27

The G-M-S formula is widely used and abused. The manner in which a value of n is adopted in practice is often
most unsatisfactory. Although equation (2.36) does provides a basis for calculating the resistance coefficient from
knowledge of the bed material, other more typical ad hoc approaches include:
An Australian approach using the telephone (You did some work on River X years ago. What do you think
n is on River Y (10km from X) for the reach between A and B?).
Tables of values in books on open channels, given channel conditions, for example pp116-123 of ?, ? for
US rivers, or ? for New Zealand Rivers. The photograph in figure 2.11 is of the Columbia River at Vernita
in Washington, taken from Barnes. It has the lowest n = 0.024 of all the examples in the book. It is 500m
wide and has boulders of diameter . . . well it doesnt say. Pity, because it is the roughness size to river depth
which determines the resistance characteristics. Presenting a picture is not much help one has little idea of
the underwater conditions and how variable they are.

Figure 2.11:

Example of a figure from Barnes (1967)

Conveyance and Friction Slope

It is convenient to introduce the conveyance K (units of discharge L3 T1 ), so that the resistance term in the
momentum equations appears as
Q |Q|
Q |Q|
P
= gA 2 .
(2.38)
A2
K
From equation (2.33), the various definitions of K become
K=

g A3
=C
P

A3
A5/3
1 A5/3
=
k
,
=
St
P
n P 2/3
P 2/3

(2.39)

such that in the form of equation (2.38) it is a convenient shorthand that includes the resistance coefficient of
whatever law is being used, plus cross-sectional geometry terms.
A simple and important result for steady uniform flow is that
Q = K0

S0 ,

(2.40)

where the subscript 0 denotes the conveyance corresponding to the normal depth h0 of the uniform flow.
Many textbook presentations write the friction term in terms of a dimensionless quantity Sf = Q |Q| /K 2 , called
the "friction slope", possibly better known as "resistance slope", so that the resistance term in the momentum
equations appears as gASf . It is also known as the "slope of the energy grade line", or the "head gradient", which
gives an uninformative and misleading picture, for in our momentum-based approach it is neither of those things.
Computations of steady uniform flow
As we know, steady flow does not change with time; uniform flow is where the depth does not change along the
waterway. For this to occur the channel properties also must not change along the stream, such that the channel is
prismatic, and this occurs only in constructed canals. However in rivers if we need to calculate a flow or depth, it
is common to use a cross-section which is representative of the reach being considered, and to assume it constant
for the approximate application of theory. This is the simplest problem we consider!
The Weisbach and Chzy equations and the Gauckler-Manning-Strickler forms (2.31) and (2.32) are formulae for
28

the discharge Q in terms of resistance coefficient, slope S0 , area A, and wetted perimeter P :
r
r 1/2
p
8g A
8g A
S0 = A
S0
Weisbach-Chzy :
Q=A
P
P
2/3 p
A A
S0
G-M-S :
Q=
n P

(2.41a)
(2.41b)

in which both A and P are functions of the flow depth. Each equation show how flow increases with cross-sectional
area and slope and decreases with wetted perimeter. If we know or can assume the resistance parameter, this is
straightforward. However if we have to calculate it from the Yen formula (2.29), which is in terms of Q, it has to
be done iteratively see example 2.1 below.
Trapezoidal sections
B
h

Figure 2.12:

Trapezoidal section showing important dimensions

Most canals are excavated to a trapezoidal section, and this is often used as a convenient approximation to river
cross-sections too. In many of the problems in this course we will consider the case of trapezoidal sections. We
will introduce the terms defined in Figure 2.12: the bottom width is W , the depth is h, the top width is B, and the
side slope, unusually defined to be the ratio of H:V dimensions is . From these the following important section
properties are easily obtained:
Top width :
Area :
Wetted perimeter :
(Exercise: Obtain these relations).

B = W + 2h
A = h (W + h)
p
P = W + 2 1 + 2 h.

(2.42a)
(2.42b)
(2.42c)

The hydraulically-wide channel

Many rivers and canals are sufficiently wide that we can neglect the effects of the sides and for purposes of simple
calculations we say A Bh and P B, so that we model the section as a rectangle that is much wider than it is
deep, and it is called hydraulically-wide. We introduce the discharge per unit width q = Q/B, which is equal to
the integral of velocity over the depth. In this case equations (2.41a) and (2.41b) become
r
8g 3/2 p
Weisbach-Chzy :
q=
S0 and
(2.43a)
h

p
1
G-M-S :
q = h5/3 S0 ,
(2.43b)
n
and we see that q h1.5 or h1.67 . An immediate deduction is, for example, that if a flow increases by a factor of
10, the the depth increases by a factor of approximately 102/3 4.6, while a 100-fold increase in flow means an
increased depth with a factor of 1002/3 22.
Computation of discharge
This is done by evaluating any of the equations. If the numerical value of the resistance coefficient is to be
calculated from the Yen formula (2.29) it has to be done iteratively.
Example 2.1 Calculate the flow in a river of slope S0 = 103 , approximated by a trapezoidal section with bottom width
10 m, side slopes 2 : 1, and depth 1 m, if the mean roughness size is 1 cm, using
(a) the Yen formula (2.29) and the Weisbach-Chzy formula (2.41a) we use the viscosity of water at 20 C = 1
106 m2 s1 .
(b) the Strickler formula (2.36) and the G-M-S formula (2.41b) combined to give equation (2.37) above

With h = 1 m, equation (2.42b) gives A = 1 (10 + 2 1) = 12.0 m2 , equation (2.42c) gives P = 10 + 2 1 + 22 1 =

14. 47 m.

29

(a) Weisbach
Firstly we calculate the , neglecting the viscosity term
=
ln2
Calculate Q

1.33
1.33
0.9  = 2  1 0.01  = 0.0279

k
1
P
ln 12. 12/14.47
12. A/P + 2. Q

Q=A

u
8g A
8 9.8 12
S0 = 12.0
0.001 = 18.3 m3 s1
P
0.0279 14.47

Repeat (i) and (ii), including that result in the viscosity term
=
ln2

1
0.01
12. 12/14.47

1.33
0.9  = 0.0280,

1106 14.47
+ 2.
18.3

and we see that the viscosity correction is not worth incorporating.

(b) G-M-S/Strickler/Equation (2.37)

 2/3

2/3
s

A
12
A
12
Q=
gS
=
9.8 0.001 = 18.1 m3 s1
0
0.125 k1/6 P
0.125 0.011/6 14.47

Computation of normal depth

If the discharge, slope, and the appropriate roughness coefficient are known, equations (2.41a)-(2.41b) is a transcendental equation for the normal depth h, which can be solved by the methods for solving transcendental equations. A simple method is to use direct iteration, whereby we arrange the equations in the form h = f (h), and
solve by repeated evaluation. For this to converge, f (h) must be a relatively slowly-varying function. Here we
develop a procedure which seems to work well:
In the case of hydraulically-wide channels, we neglect the contributions of the side slopes so that neither the wetted
perimeter P nor the breadth B vary much with depth h. Hence the quantity A(h)/h also does not vary strongly
with h. Hence we can rewrite the G-M-S expression:

1 A5/3 (h) p
S0 (A(h)/h)5/3
Q=
h5/3 ,
S
=
0
2/3
n P (h)
n
P 2/3 (h)

where most of the variation with h is contained in the last term h5/3 , and by solving for that term we can re-write
the equation in a form suitable for direct iteration

3/5
Qn
P 2/5 (h)
h=

,
A(h)/h
S0
where the first term on the right is a constant for any particular problem, and the second term is expected to be
a relatively slowly-varying function of depth, so that the whole right side varies slowly with depth a primary
requirement that the direct iteration scheme be convergent and indeed be quickly convergent.
Here, in addition, we include the Strickler formula (2.36), which is equation (2.37) re-written
3/5

P 2/5 (h)
Q k 1/6

.
h = 0.125

A(h)/h
S0 g

(2.44)

Experience with typical trapezoidal sections shows that this works well and is quickly convergent. However, it also
works well for flow in circular sections such as sewers, where over a wide range of depths the mean width does not
vary much with depth either. For small flows and depths in sewers this is not so, and a more complicated method
such as the secant method might have to be used.
Example 2.2 Calculate the normal depth in the previous example 2.1, for the calculated discharge of 18.1 m3 s1 using the
G-M-S formulation, with equation (2.44). We pretend we do not know the answer of 1 m and for an initial estimate we choose
1.5 m. We have A = h (10 + 2 h), P = 10 + 4.472 h, giving the scheme

3/5
Q k1/6
(10 + 4.472 h)2/5
h =
0.125

10 + 2 h
S0 g


3/5
(10 + 4.472 h)2/5
18.1 0.011/6

=
0.125
10 + 2 h
0.001
9.8
=

4.125

(10 + 4.472 h)2/5

10 + 2 h

30

and starting with h = 1.5 we have the sequence of approximations: h = 0.979 m, h = 1.002 m, h = 1.001 m, and the method
has converged to this result, quite satisfactorily in its simplicity and speed (it did not converge to the exact result because the
specified discharge was rounded).

2.3.6

Steady gradually-varied flow

A common problem in river engineering is, for example, how far upstream water levels might be changed, and
hence flooding possibly enhanced, due to downstream works such as the installation of a bridge or other obstacles.
Far away from the obstacle or control, the flow may be uniform, but generally it is variable. The transition between
conditions at the control or point of known level, and where there is uniform flow is described by the GraduallyVaried Flow Equation, which is an ordinary differential equation for the water surface height. The solution will
approach uniform flow if the channel is prismatic, but in general we can treat non-prismatic waterways also. The
steady flow approximation is often used as a first approximation, even when the flow is unsteady, such as in floods.
The differential equation
Consider the mass conservation equation for steady flow, when /t 0, and equation (2.1) becomes
dQ
= i,
dx
with the solution obtained by integration
Q(x) = Q0 +

i dx,

x0

where at an upstream station x0 the discharge is Q0 , the extra discharge just being given by the integral of the
inflow i.
The momentum equation (2.23b) for Q/t = 0 and Q/x = i, and assuming Q positive, becomes

Q2 B d
Q
Q2
Q2 B
gA 2
= 2 S P 2 2 i,
A
dx
A
A
A

(2.45)

which is a first-order ordinary differential equation for (x), provided we have evaluated Q(x), and that we know
how the geometric quantities A and B depend on surface elevation at each point. This is the Gradually-Varied
Flow Equation. The equation may be solved numerically using any of a number of methods available for solving
ordinary differential equations which are described in books on numerical methods. These methods are usually
accurate and can be found in many standard software packages.
It is surprising that books on open channels do not recognise that the problem of numerical solution of the
gradually-varied flow equation is actually a standard numerical problem, although practical details may make
it more complicated. Instead, such texts use methods such as the Direct step method and the Standard step
method, which can become complicated. There are several software packages such as HEC-RAS which use such
methods, but solution of the gradually-varied flow equation is not a difficult problem to solve for specific problems
in practice if one recognises that it is merely the solution of a differential equation.
In sub-critical (relatively slow) flow, the effects of any control can propagate back up the channel, and so it is that
the numerical solution of the gradually-varied flow equation also proceeds in that direction. On the other hand, in
super-critical flow, all disturbances are swept downstream, so that the effects of a control cannot be felt upstream,
and numerical solution also proceeds downstream from the control.
No inflow:
If there is no inflow, i = 0 and Q = Q0 , a constant, throughout. Dividing both sides of equation (2.45) by gA
gives
d
S P/B
S P/B
=
= F2
,
(2.46)
2
dx
1 F
1/F 2
where, unusually for lectures on flow with a free surface, it has taken us until now to define the Froude number
2

F2 =

Q2 B
(Q/A)
=
,
gA3
g (A/B)

the ratio of the mean velocity Q/A squared to g times the mean depth A/B. In this course we call F 2 the Froude
number, and not F , as the latter quantity occurs quite rarely, and F 2 expresses the real relative importance of
inertia terms.
31

The Gradually-Varied Flow Equation (GVFE) in the form of equation (2.46) seems simple deceptively simple.
For example, can be taken as a constant; S might be a function of x, but we probably do not have enough
information to express it as a function of ; many open channels are much wider than their depth, and so P
B and P/B 1. This leaves most of the functional variation with on the right side in the term 1/F 2 =
gA3 /Q2 B in which, for practical river problems the dependence of A and B on the local elevation is actually
quite complicated.
Constant slope:
As a special case, consider a channel with a bed of constant slope S = S0 . It is simpler to use as a variable the
depth of flow h, where h = Z, where Z is the elevation of the bed at a section, so that dZ/dx = S0 . Equation
(2.46) becomes
d
S0 P/B
dh dZ
dh
=
+
=
S0 =
.
dx
dx
dx
dx
1/F 2
Solving for dh/dx and introducing the conveyance gives the GVFE for a prismatic canal of constant slope:
S0 Q2 /K 2
dh
.
=
dx
1 F 2

(2.47)

Properties of gradually-varied flow and the governing equations

Normal depth h0
h(x)

Figure 2.13: Subcritical flow retarded by a gate, showing typical behaviour of the free surface and, if the channel is prismatic,
decaying upstream to normal depth

The equation and its solutions are important, in that they tell us how far the effects of a structure or works in or
on a stream extend upstream or downstream.
It is an ordinary differential equation of first order, hence one boundary condition must be supplied to obtain
the solution. In sub-critical flow, this is the depth at a downstream control; in super-critical flow it is the depth
at an upstream control.
If the channel is prismatic, far from the control, the flow is uniform, and the depth is said to be normal.
In general the boundary depth is not equal to the normal depth, and the differential equation describes the
transition from the one to the other. The solutions look like exponential decay curves, and below we will show
that they are, to a first approximation. Figure 2.13 shows a typical sub-critical flow in a prismatic channel,
where the depth at a control is greater than the normal depth.
The differential equation is nonlinear, and the dependence on h is complicated, such that analytical solution is
not possible, and we will usually use numerical methods.
However, a small-disturbance approximation can be made, the resulting analytical solution is useful in providing us with some insight into the quantities which govern the extent of the upstream or downstream influence.
If the flow approaches critical flow, when F 2 1, then dh/dx , and the surface becomes vertical. This
violates the assumption we made that the flow is gradually varied and the pressure distribution is hydrostatic.
This is the one great failure of our open channel hydraulics at this level, that it cannot describe the transition
between sub- and super-critical flow.
Approximate analytical solution by linearising
Whereas the numerical solutions give us numbers to analyse, sometimes very few actual numbers are required,
such as merely estimating how far upstream water levels are raised to a certain level, the effect of downstream
works on flooding, for example. Here we introduce a different way of looking at a physical problem in hydraulics,
32

where we obtain an approximate mathematical solution so that we can provide equations which reveal to us more
of the nature of the problem than do numbers. This work was originally done by ?.
This is carried out by linearising about the uniform flow in a prismatic channel, i.e. by considering small
disturbances to that flow. Consider the water depth to be written
h(x) = h0 + h1 (x),

(2.48)

where we use the symbol h0 for the constant normal depth, and h1 (x) is a relatively small departure of the surface
from that. We use the governing differential equation (2.47) (although our notation has obscured the fact that F
and K are functions of h). Substitute equation (2.48) into the equation and writing numerator and denominator as
Taylor series in h1 :

S0 Q2 /K 2 0 + terms in h21
S0 Q2 /K 2 0 + h1 dh
dh0 dh1

d
.
+
=
dx
dx
(1 F 2 )0 + h1 dh
(1 F 2 ) 0 + terms in h21

Now, as h0 is constant, dh0 /dx = 0. Also, from equation (2.40), S0 Q2 /K 2 0 = 0 and the first term in the

numerator is zero. Now evaluating d S0 Q2 /K 2 /dh = 2Q2 /K 3 dK/dh, and considering just the first term
top and bottom, neglecting all higher order powers of h1 as it is small, we find
2Q2 /K03 (dK/dh)0
dh1
= 0 h1 ,
h1
dx
1 F02
where
0 =
and where we have used Q2 /K02 = S0 .

2S0
1 dK
,
1 F02 K dh 0

(2.49)

(2.50)

Equation (2.49) is a linear differential equation which we can solve analytically by separation of variables, giving
h1 = Ce0 x ,

and h = h0 + Ce0 x ,

(2.51)

where C is a constant which would be evaluated by satisfying the boundary condition at the control, and where 0
is a constant decay rate given by equation (2.50).
This shows that the water surface is actually approximated by an exponential curve passing from the value of depth
at the control to normal depth. As dK/dh is positive, and for subcritical flow 1 F02 is also positive, equation
(2.50) shows that 0 is positive, and far upstream as x , the water surface decays to normal depth. For
supercritical flow, 1 F02 < 0, 0 is negative, and the water surface approaches normal depth downstream.
Now we obtain an approximate expression for the rate of decay 0 . From the Gauckler-Manning-Strickler formula
for a wide channel, a common approximation, we can show that K h5/3 , dK/dh 5/3 h2/3 , and for slow
flow F02 1, we find
10 S0
0
.
(2.52)
3 h0
The larger this number, the more rapid is the decay with x. The formula shows that more rapid decay occurs with
steeper slopes (large S0 ), and smaller depths (h0 ). Hence, generally the water surface approaches normal depth
more quickly for steeper and shallower streams, and the effects of a disturbance can extend a long way upstream
for mild slopes and deeper water.
Let us use equation (2.52) to calculate the distance upstream that the disturbance decays by 1/2, that is, exp (0 x) =
0.5. We find
x
3 ln 0.5 1
10 S0 x
= ln 0.5 giving
=
.
3 h0
h0
10 S0
For S0 = 104 and a stream 2 m deep, the distance is 4 km. For the stream disturbance to decay to 1/16 = (1/2)4
of the original, this distance is 4 4 km = 16 km. These are possibly surprising results, showing how far the
backwater effect extends.
Numerical solution of the gradually-varied flow equation
Consider the gradually-varied flow equation (2.47)
dh
S0 Q2 /K 2
=
dx
1 F 2
33

where F 2 (h) = Q2 B(h)/gA3 (h). The equation is a differential equation of first order, and to obtain solutions
it is necessary to have a boundary condition h = h0 at a certain x = x0 , which will be provided by a control.
The problem may be solved using any of a number of methods available for solving ordinary differential equations
which are described in books on numerical methods.
The direction of solution is very important. For mild slope (sub-critical flow) cases the surface decays somewhat
exponentially to normal depth upstream from a downstream control, whereas for steep slope (super-critical flow)
cases the surface decays exponentially to normal depth downstream from an upstream control. This means that to
obtain numerical solutions we will always solve (a) for sub-critical flow: from the control upstream, and (b) for
super-critical flow: from the control downstream.
The simplest (Euler) scheme to advance the solution from (xi , hi ) to (xi + xi , hi+1 ) is

Eulers method

xi+1
hi+1

where xi is negative for subcritical flow,

dh
S0 Q2 /K 2 (hi )
2
hi + xi
= hi + xi
+ O (xi ) .

2
dx i
1 F (hi )
xi + xi ,

(2.53a)
(2.53b)

This is the simplest but least accurate of all methods yet it might be appropriate for open channel problems
where quantities may only be known approximately. One can use simple modifications such as Heuns method
to gain better accuracy, or use Richardson extrapolation or even more simply, just take smaller steps xi .
Richardson extrapolation There is an interesting method for obtaining more accurate solutions from computational schemes for almostany physical
problem. Applied to the Euler scheme in the present context, as the

2
local truncation error is O (xi ) , so that after a number of steps proportional to 1/xi , the actual solu

1
tion has an error O (xi ) so that n = 1, a first-order scheme. In the present problem, if we use a constant
space step to obtain the first solution 1, then another constant space step half that /2, requiring twice the
number of steps, then r = 1/2, and it can be shown that a better estimate of the solution is
hi 2h2i (/2) hi (),

(2.54)

which is trivially applied to each or any of the steps. The notation h2i (/2) is intended to show that the same
point in physical space is used; with half the step size it will now take twice the number of steps to reach that
point.
Heuns method In this case the value of hi+1 calculated from Eulers method, equation (2.53b), is used
as a first
estimate of the depth at the next point, written hi+1 , then the value of the derivative at that point xi+1 , hi+1
is calculated. Heuns method is then to use the mean slope over the step, the mean of the initial value and
that at the other end of the interval calculated by the Euler step. Then, the change over the step is calculated,
multiplying that mean slope by the step length. That is,
hi+1

dh
dh
+
dx (xi ,hi )
dx (xi+1 ,h )
i+1

2
2
xi S0 Q /K (hi ) S0 Q2 /K 2 (hi+1 )
3
= hi +
+
+ O (xi ) .

2
2
2
1 F (hi )
1 F (hi+1 )
xi
hi +
2

(2.55)

Now, the error of a single step is proportional to the third power of the step length and the error at any point
will be proportional to the second power.
Neither of these two methods are presented in hydraulics textbooks as alternatives, yet they are simple
and flexible, and reveal the nature of what we are doing. The step xi can be varied at will, to suit possible
irregularly spaced cross-sectional data. In many situations, where F 2 1, we can ignore the F 2 term in the
denominators, giving a notationally simpler scheme.
Predictor-corrector method Trapezoidal method This is simply an iteration of the last method, whereby
the step in equation (2.55) is repeated several times, at each stage setting hi+1 equal to the updated value of
hi+1 . This gives an accurate and convenient method, and it is surprising that it has not been used.
Higher-order methods One of the aims here has been to emphasise that all that is being done is to solve
numerically a differential equation, and any method can be used, for which reference can be made to any book
on numerical solution of ordinary differential equations. There are sophisticated methods such as high-order
Runge-Kutta methods and predictor-corrector methods. However, in the case of open channel hydraulics there
will usually be some variation of parameters along the channel that such sophistication is unnecessary.
34

Accurate: 4th order Runge-Kutta

RK-4, n in error by 5%
Euler, eqn (2.53)
Euler-Richardson, eqns (2.53) & (2.54)
Direct step, eqn (2.57)
Modified direct step, eqn (2.58)
Heun, eqn (2.55)
Trapezoidal, eqn (2.55)+

2.6

2.4

Surface
elevation
(m)

2.2

1.8

1.6
-1000

-800

-600

x (m)

-400

-200

Figure 2.14: Backwater curve computed with various schemes; the dashed line is the surface
for uniform flow.

0.01

Elevation 0
Error
(m)
-0.01

-0.02

-0.03

-800
Figure 2.15:

-600

x (m)

-400

-200

Errors for the different schemes; symbols as for Figure 2.14.

35

Comparison of schemes
To compare the performance of the various numerical schemes, Example 10-1 of ? was solved using each. All
quantities specified by Chow were converted to SI units and rounded to the numbers shown here: a flow of
11.33 m3 s1 passes down a trapezoidal channel of gradient S0 = 0.0016, bed width 6.10 m and channel side
slopes 0.5, g = 9.8 m s2 , the quantity or = 1.10, and Mannings n = 0.025. At x = 0 the flow is backed
up to a depth of 1.524 m, and the backwater curve was computed for 1000 m. Results for the water surface profile
are shown in Figure 2.14, while Figure 2.15 shows the errors. Some 10 computational steps were used for each
scheme.
The basis of accuracy is shown by the solid line, from a highly-accurate Runge-Kutta 4th order method (see, for
example, p372 of ?, ?, or almost any other book on numerical methods). It is not recommended here as a method,
however, as it makes use of information from three intermediate points at each step, information which in nonprismatic channels is usually not available. The rest of the results are shown in reverse order of accuracy. The
dotted line is that with the same numerical method, but where the roughness n was changed by -5%, to give an
idea of the effect of uncertainty in knowledge of that quantity. The maximum error, of about 3 cm, in the normal
depth, is greater than any of the other methods, so that a preliminary conclusion is that if the roughness is not
known to within 5%, almost certainly the case in practice, then any of the methods can be used.
It can be seen that Eulers method, eqn (2.53), was the least accurate, as expected. As it is a first-order scheme,
halving the step size would halve the error. Actually doing just that and then applying Richardson extrapolation,
equation (2.54), gave the second most accurate of all the methods, with an error of about 1 mm. The most accurate
of all was the Trapezoidal method, namely using Heuns method, equation (2.55) and iterating the final step. All
the other methods, Heuns, and the two inverse formulations, equations (2.57) and (2.58), gave errors intermediate
between the two extremes. It is interesting that the two traditional methods were accurate, notably the traditional
inverse formulation over the modified version presented in this work; and the Trapezoidal method, the basis for the
so-called "standard" method.
The results show the disadvantages of the inverse formulation (Direct Step), that the distance between computational points becomes large as uniform flow is approached, and the points are at awkward distances. In this
example relatively few steps were chosen (roughly 10) so that the numerical accuracies of the methods could be
distinguished visually. The computational effort was very small indeed.
In this problem the analytical solution (2.51) gave poor results. This was because the depth at the control was rather
larger (50%) than the normal depth, and the linearisation adopted, for small departures from normal depth, was not
accurate. In general, however, it does give a simple approximate result for the rate of decay and how far upstream
the effects of the control extend. For many practical problems, this accuracy and simplicity may be enough.
Traditional methods
Here we present methods for comparison as they are given in textbooks.
U22 /2g
h2

Sf x

Total energy line

U12 /2g

Sub-critical flow
h1

S0 x
x
2

1
Figure 2.16:

Elemental section of waterway

Derivation of the gradually-varied flow equation using energy Consider the elemental section of waterway
of length x shown in Figure 2.16. We have shown stations 1 and 2 in what might be considered the reverse
order, but for the more common sub-critical flow, numerical solution of the governing equation will proceed
back up the stream. Considering stations 1 and 2:
Total head at 2 =
Total head at 1 =
36

H2
H1 = H2 HL ,

and we introduce the concept of the friction slope Sf which is the gradient of the total energy line such that
HL = Sf x. This gives
H1 = H2 Sf x,
and if we introduce the Taylor series expansion for H1 :
H1 = H2 + x

dH
+ ...,
dx

substituting and taking the limit x 0 gives

dH
= Sf ,
dx

(2.56)

an ordinary differential equation for the head as a function of x, and we use the approximation that the friction
slope is given by Q2 /K 2 .
Direct step method Textbooks do present the Direct Step method, which is applied by taking steps in the height
and calculating the corresponding step in x. It has practical disadvantages, such that it is applicable only to
prismatic sections, results are not obtained at specified points in x, and as uniform flow is approached the x
become infinitely large. However it is a surprisingly accurate method.
The reciprocal of equation (2.56) is
dx
1
= .
dH
Sf
The numerical method as set out in textbooks is to approximate the differential equation (2.56) by the finite
difference expression
xi

=
=

Hbi
S0 Q2 /K 2 (Hb )
Hbi

2
S0 12 Q2 Ki2 + Ki+1

(2.57a)
(2.57b)

where the overbar in equation (2.57a) indicates the mean of the friction slope at beginning and end of the
computational interval, which finds its mathematical expression in equation (2.57b), where the shorthand Ki
has been used for K (Hbi ).
While this is a plausible approximation, it is not mathematically consistent. It is an apparent attempt to
develop a Trapezoidal method. Applying Heuns method as formally presented in equation (2.55) automatically
leads to the Trapezoidal scheme which in this case gives

1
Hb,i
1
3
xi+1 = xi +
+
)
+
O
(H
,
(2.58)
b,i
2
2
S0 Q2 /Ki2 S0 Q2 /Ki+1
The term O (. . .) is a Landau order symbol, showing in this case that the local truncation error is proportional
to the third power of the step, which is a strong result and explains the accuracy of the method. Since the use of
a step size of Hb,i over the whole computational domain requires a number of steps proportional to 1/Hb,i ,
2
the global error in this case will be of order (Hb,i ) , thus the global error, or accumulated error at the end
of that integration interval will be of this order, so that halving the step should improve the global accuracy by
about a factor of 4.
In view of the method presented here, the method is no longer applicable only to prismatic sections, but
the practical disadvantages remain that results are not obtained at specified points in x, and as uniform flow is
approached the x become infinitely large.
Standard step method The nomenclature "standard" is not very descriptive. Presumably it refers to finding the
solution for at specified values of x, rather than the other way round, for which the term "direct", as above,
is even worse. This is an implicit method, requiring numerical solution of a transcendental equation at each
step. It can be used for irregular channels, and is rather more general. In this case, the distance interval x is
specified and the corresponding depth change calculated. In the Standard step method the procedure is to write
H = Sf x,
and then write it as
H2 (h2 ) H1 (h1 ) =
37

x
(Sf1 + Sf2 ) ,
2

for sections 1 and 2, where the mean value of the friction slope is used. This gives

Q2
x
Q2
+ Z2 + h2 =
+ Z1 + h1
(Sf1 + Sf2 ) ,
2
2gA2
2gA21
2

where Z1 and Z2 are the elevations of the bed. This is a transcendental equation for h2 , as this determines A2 ,
P2 , and Sf2 . Solution could be by any of the methods we have had for solving transcendental equations, such
as direct iteration, bisection, or Newtons method.
Although the Standard step method is an accurate and stable approximation, the lecturer considers it unnecessarily complicated, as it requires solution of a transcendental equation at each step. It would be much simpler
to use a simple explicit Euler or Heuns method as described above.

2.3.7

Long wave equation in terms of volume upstream V

It is possible to simplify the usual approach of considering two long wave equations in two unknowns by introducing a quantity that satisfies one of the equations exactly. We consider the (A, Q) formulation of the long wave
equations, equations (2.22). In partial derivative terms we introduce the function V (x, t) which is the total volume
of fluid upstream of point x at time t, and use equations (2.5) in 2.3.1.4 so that we write:
V
x
V
t

= A and

(2.59a)

= Q(x0 , t) +

x0

i(x0 ) dx0 Q,

(2.59b)

where Q(x0 , t) is the inflow at x0 , the upstream end of the flow domain, which has been introduced by Fatemeh
Soroush (2011, Personal Communication), improving on an earlier definition. With this addition V can now more
consistently be understood as the total volume in the channel. The derivative of volume V with respect to x is
clearly the local cross-sectional area,
R x as shown, while the rate of change of V with time is equal to the rate at
which flow is entering, Q(x0 , t) + x0 i(x0 ) dx0 minus the discharge Q at point x which is the rate at which flow is
passing point x, and hence is no longer upstream.
Equations (2.59) can be solved to give
A =

V
x

and

Q = Q(x0 , t) +

(2.60a)
Z

x0

i(x0 ) dx0

V
,
t

(2.60b)

so that both A and Q have been expressed as derivatives of V . Substituting into the mass conservation equation
(2.22a):

Z x

V
0
i dx
+
= i,
t x
x
t
x0
so that the left side becomes i, which equals the right side, and the mass conservation equation is identically
satisfied! The momentum conservation equation (2.22b), with all signs reversed becomes:
2
R x

Z
i dx0 V /t
2V
Q 2V
gA 2 V
Q2
+
+
2

=
P

gA
S
+

2
t2
A xt
A2
B
x2
(V /x)

i 0
Q
dx + 2 i. (2.61)
t
A

where for notational simplicity, single symbols Q and A have been retained in coefficients of derivatives. The
momentum equation has become a second-order partial differential equation in terms of the single variable V .
This makes theoretical manipulations easier and provides insight into the nature of solutions of the equation. This
is anticipated by the comment that the first three terms on the left actually look like a wave equation the traditional
forms of equations (2.22) and (2.23) do not look like wave equations. However the single equation certainly does
look complicated it will be of most use when we consider approximations.
The case of no inflow i = 0 is rather simpler, and is adequate for our purposes. Equation (2.61) becomes

2V
Q 2V
gA 2 V
(V /t)2
Q2

+
2

= P
+

2 gAS.
2
2
2
t
A xt
A
B
x
(V /x)

(2.62)

This equation, which is a single equation in a single unknown and could be used for flood routing and wave
propagation studies. Here we will use it first to obtain an approximate equation which is then solved to give
solutions that show the real behaviour of long waves.
38

2.3.8

The slow-change/slow-flow approximation

In many river and canal problems, such as flood propagation, change is slow and the flow is relatively slow such
that the Froude number is small. In the momentum equations we now neglect the time derivatives and terms of the
order of the Froude number squared. This is usually a very good approximation.
Formulation 1 area A
Consider Formulation 1 for the momentum equation (2.22b) in terms of area A,
gA A
Q |Q|
= gAS gA 2 ,
B x
K
where we have written the resistance term using the conveyance K as introduced in equation (2.38). Restricting
our attention to uni-directional flows, such that Q |Q| = Q2 we find that the momentum equation becomes simply
r
1 A
Q = K S
,
(2.63)
B x
an explicit formula for discharge in terms of resistance and geometrical quantities. Substituting for Q from equation
(2.63) into the corresponding mass conservation equation (2.22a) gives
r
!
1 A

+
K S
= i,
(2.64)
t
x
B x
so that we have a single partial differential equation in the geometric quantity area A.
Formulation 2 surface elevation
If, instead, we use Formulation 2 for the momentum equation (2.23b) in terms of stage the low-inertia approximation is simply
r

(2.65)
Q=K ,
x
which seems surprisingly simple. We might have written this last equation down assuming that it was a plausible
engineering approximation it is a surprise that it is actually an approximation to the momentum equation and
actually a very good one. Our only problem seems to be to get used to the minus sign under the square root; we
just have to remember that for uni-directional flow /x is always negative. Substituting into the corresponding
mass conservation equation (2.23a) gives
r
!

1
i
+
K
= ,
(2.66)
t
B x
x
B
another single partial differential equation in terms of elevation . It is remarkable that these simple equations are
accurate enough for most routing purposes.
Formulation 3 upstream volume V
If we were only given information in terms of water level variation at boundaries and only required water level
variations as output, these equations (2.64) and (2.66) might be enough for our purposes. However, often information in the form of an inflow hydrograph in terms of Q(x0 , t) is given, difficult to include in these. Now our
concept of upstream volume V becomes useful. Here we can just substitute A = V /x into equation (2.64) to
give
!
r
2V
2V

= i,
+
K S
tx x
B x2
and integrating with respect to x gives us the equation in terms of V :
s
Z x
1 2V
V
=
i dx,
+ K(Vx ) S
t
B(Vx ) x2
x0

(2.67)

where we have shown that both surface width B and conveyance K are functions of the cross-sectional area,
which we write as A = V /x = Vx . This Volume Routing equation might be useful in a range of hydrologic
and hydraulic computations, replacing the solution of the long wave equations. It is a nonlinear partial differential
equation which is a single equation in a single variable. The only approximations that have been made is that F 2
is small and variation is slow.
39

Boundary conditions:
At an upstream boundary x0 we might have a given inflow as a function of time Q(x0 , t). From equation (2.60b),
when x = x0 we have simply V (x0 , t) = 0 for all t, which makes physical sense there is never any volume
upstream of the upstream boundary.
At control points, such as a downstream structure we will usually have some relationship between Q and A such
as provided by weir formulae, or resistance formulae, which give V /t as a function of V /x there. As part of
the solution we will have to differentiate numerically to give the latter and then integrate to give the updated value
of V . At open boundaries, where there is no control point, we may simply be able to apply the partial differential
equation as if it were an interior point.
Initial conditions:
To provide the initial conditions in the interior of the computational region, we consider the steady state form
of equation (2.67), noting that V /t is not zero for steady flow, as volume is continually passing, but instead
V /t = Q0 where Q0 is the steady flow (here for i = 0), giving:

d2 V
Q20

=
B(V
)
S

.
x
dx2
K 2 (Vx )

(2.68)

This is just a low-inertia version of the gradually-varied flow equation (2.46)

S P/B
S P/B
d
=
= F2
.
2
dx
1 F
1/F 2
Equation (2.68) is slightly more complicated than the conventional formulation, as we have to solve the secondorder equation numerically, which is usually done by introducing a subsidiary variable dV /dx which is A in our
formulation.

400

300

Q
(m s ) 200
3 1

100
(a) Steep slope 0.001
Inflow
Long wave eqns
Eqn (2.67) (obscured)
0

24

48
Time (h)

72

(b) Gentle slope 0.0001

Inflow
Long wave eqns
Eqn (2.67) (obscured)
96 0

24

48
Time (h)

72

96

Figure 2.17: Inflow and computed outflow hydrographs for two model rivers
with (a) a steep slope and (b) a gentle slope

Consider the two cases in figure 2.17 computed for a hypothetical trapezoidal channel with bottom width 40 m,
side slopes H:V 0.5, Mannings n = 0.035, and with an inflow hydrograph with a base flow of 100 m3 s1 , time to
peak 24 h, and peak flow of 400 m3 s1 . The figure shows the inflow and outflow hydrographs of a 100 km reach
obtained from numerical solutions of the nonlinear long wave equations plus solutions of the slow change / slow
flow equation, an approximation to the full equations. Part (a) of the figure is for a steep bed slope of 0.001, (with
a Froude number of the initial flow F 2 0.1 and remember that F 2 is relatively independent of flow, so it would
40

hold for the flood peak too) (b) for a gentler slope of 0.0001, when F 2 0.01. The accuracy of the approximation
is remarkable, especially for the steeper slope and larger Froude number.

2.3.9

A further approximation the advection diffusion equation

We have seen that in the case of a common situation in river engineering, where the change due to input is slow and
where the Froude number of the flow is not large, that the equations simplify. Now we obtain a further approximation by linearising, to show that the equations exhibit advection-diffusion behaviour, and simple deductions can be
made about the nature of wave propagation in waterways.
Advection-diffusion equation
This is a an approximation which was first obtained by Hayami in 1951 which gives considerable insight into the
nature of flood and wave propagation in rivers. It can be obtained by linearising any of the low-inertia equations
(2.64), (2.66) or (2.67) about a uniform flow of slope S0 , to give an equation in terms of depth or discharge. It
is slightly more general to take the upstream volume formulation (2.67) and to linearise that. In fact, we already
have the solution to hand, for if we ignore the inertial terms in the linear Telegraphers equation (2.84), simply by
ignoring second time derivative terms and those inertial terms with a coefficient , giving

gA0 2
2
= 0,
(2.69)
+c

t
x
B0 x2
where as before, is the perturbation upstream volume. This can be written

2
+ |{z}
c
= |{z}

t
x
x2
Kinematic
wave speed

(2.70)

Diffusion
coefficient

such that the equation governing the propagation of floods and long waves for low Froude number is an advectiondiffusion equation,where the coefficients c and are functions of the underlying uniform flow. The kinematic wave
speed is given by equation (2.86), but we simplify it here by neglecting variation of with Q:

3
1 A0 (P )
c = U0 1
.
(2.71)
2
3 0 P0 A 0
It can be shown in terms of the conveyance that
c=

dQ0 p dK
S0 dK
= S0
=
,
dA0
dA 0
B0 dh 0

(2.72)

where Q0 = K0 S0 ,and dK/dh is the derivative of the conveyance with respect to the depth h.
The coefficient of the second derivative in equation (2.70) is a diffusion coefficient with units of L2 T1 and is
given by
gA0 /B0
Q0
K0
gA0 /B0 U0
,
=
=
=
(2.73)
=
2
2gS0
2B0 S0
2B0 S0

on using the uniform flow relationship Q0 = K0 S0 .

The behaviour of solutions
Solutions of equations with advection and diffusion terms are well known, and this makes possible a physical
description of wave motion in waterways. Solutions of the advection equation

+c
=0
(2.74)
t
x
are easily shown to be a waveform, any waveform, travelling in the positive x direction with velocity c. That is,
= f (xct) is a solution of equation (2.74), where f () is any function. Also, solutions of the diffusion equation

2
(2.75)
= 2
t
x
have a simple mathematical structure. It can be shown that any disturbance of a certain wavelength is damped at
an exponential rate where the exponent is inversely proportional to the square of the wavelength, such that short
disturbances and irregularities are damped very quickly, and the process of diffusion is such as to smooth solutions.
In the case of the advection-diffusion equation (2.70) the processes are combined, such that waves are advected at
a speed c and the wave profile undergoes diffusion at the same time. This is illustrated in Figure 2.18, which is the
solution for an infinite channel with a single disturbance which is carried downstream and diffused.
41

Figure 2.18:

Analytical solution of advection-diffusion equation

It is useful to get a feel for nature of wave propagation in waterways. Here we provide a simple tool for estimating
the relative importance of diffusion. If we were to scale the advection-diffusion equation (2.70) such that it was
in terms of dimensionless variables X = x/L, where L is the length of reach, such that X goes from 0 to 1,
and a dimensionless time scale T = tc/L,such that T = 1 is the time taken for a wave to transit the pool, then
substituting into equation (2.70) gives

2
dT
dX
2 dX
, then
+c
=
T dt
X dx
X 2 dx
1
2 1
c
, and multiplying through:
+c
=
T L
X L
X 2 L2

2
+
=
T
X
cL X 2
and the importance of the diffusion term to the other terms is the dimensionless Pclet number /cL. This looks
like the inverse of a Reynolds number (which is correct the Reynolds number is the inverse of a dimensionless
viscosity or diffusion number). Now we substitute in the approximations for a wide channel, where A0 = B0 h0 ,
where h0 is the depth, giving from equations (2.71) and (2.73):
A measure of the importance of diffusion =

U0 A0 2
1 h0
=

.
cL
2B0 S0 3U0 L
3 S0 L

This shows us the effects of diffusion very simply, as h0 is the depth of the stream, and S0 L is the amount by
which a stream drops over the reach of interest, we have
A measure of the importance of diffusion

1 Depth of stream

.
3
Drop of stream

This shows a surprising result, that for streams which are steep and the water depth relatively shallow, such as
steep mountain streams, the gravity and friction terms are in balance, and in this high friction limit, apparently
paradoxically, the flood wave moves as a kinematic wave with little diminution. On the other hand, where slopes
are gentle and friction less, such as in irrigation channels, the effects of diffusion are stronger. This is an unusual
and unexpected result, and opposes the intuitive practical interpretation of the behaviour of waves in gently-sloping
rivers and channels that they travel without a great deal of diminution due to friction. In fact, waves in a waterway
with a mild slope may be markedly diminished in height and spread out much more in space and time.
We have shown that it is possible to formulate a low-inertia model such that it is a good approximation in most
cases of rivers and canals, and hence is worthy of further examination, as it is considerably simpler than the full
equations, both in presentation and numerical properties. It shows that waves in waterways with finite friction
travel downstream
at a speed roughly equal to 3/2 Water speed, and not upstream and downstream at a speed
given by c = g Mean depth, widely used for back of the envelope calculations. In reality, however, the
diffusion coefficient can be large, and the equation behaves more like the diffusion equation, where disturbances
are diminished and spread out in space as time passes. In many situations the effects of diffusion are dominant.
42

2.3.10

An even further approximation Muskingum-Cunge routing

? proposed the scheme from Muskingum methods from hydrologic storage routing,
Q(x + , t + ) = C1 Q(x, t + ) + C2 Q(x, t) + C3 Q(x + , t),

(2.76)

where is a step size in x, is a step size in t, expressing the solution at one point of a computational rectangle
in (x, t) space, at a later point and time, and the quantities are given by
2K X
+ 2K X
, C2 =
,
2 K (1 X) +
2 K (1 X) +

.
where K = , and X = 12 1
c
BcS0
C1 =

and C3 =

+ 2 K (1 X)
,
2 K (1 X) +

(2.77)
(2.78)

If we take a bi-dimensional power series expansion of equation (2.76) with (2.77) and (2.78) about the point (x, t),
we find that to lowest order it is actually satisfying the differential equation
Q 2 Q
Q
+c
+
= O (, ) ,
(2.79)
t
x
c t x
then the scheme is a solution of a modified diffusion equation. It can be compared with the advection-diffusion
equation (2.70), where we write it using not just the perturbed volume potential but the discharge Q:
Q
2Q
Q
(2.80)
+c
2 = 0.
t
x
x
Thus we see that the equation which Muskingum-Cunge routing is solving is (2.79) instead of (2.80). We can see
how the two are related. Consider the advection-diffusion equation in the low-diffusion approximation where
is small, where the equation becomes the kinematic wave equation, and so we can consider the operator /t +
c/x to be small, and we replace one derivative /x in the second derivative by 1/c /t, which immediately
gives us the left side of equation (2.79). Hence we see that the Muskingum-Cunge equation is actually a lowdiffusion approximation to the advection-diffusion equation. The interchanging of time and space differentiation
has meant that one can solve the problem by simple wave-like interpolation methods, rather than by approximating
the diffusion operator, however one has to solve a transcendental equation at each point and step. It is not surprising
that in the lecturers experience it gives poor results for small slopes, where diffusion is large. The lecturer has
shown that the method contains significant diffusion, inherent in the approximation leading to equation (2.79),
which is greatest for small slopes and faster variation.

1.6
1.4

Decay rate

Mi C
Ai D

1.2
Ratio of
M-C value
to A-D value

1
0.8

Wave speed

cM C
cA D

0.6
0.4
0.2
0

0
0.2
0.4
0.6
0.8
1
Dimensionless diusion-frequency coecient N = a0 /c20
Figure 2.19: Propagation speeds and downstream decay rates predicted by the Muskingum-Cunge
approximation compared with those from the Advection-Diffusion equation

Figure 2.19 shows propagation speed and downstream decay rate of this approximation. It is clear that its behaviour
is less than wonderful.
43

2.3.11

The most approximate approximation of all the Kinematic wave equation

Consider the advection-diffusion equation (2.70) in the limit where the diffusion can be ignored, for streams that
are shallow and steep, which gives equation (2.74)

+c
= 0,
(Kinematic wave equation)
t
x
which is the kinematic wave equation. Solutions of this are disturbances travelling without diffusion, with a local
speed given by the local kinematic wave speed. In fact, this equation was obtained by linearising about a uniform
flow, so strictly speaking c should be considered a constant, which we will now denote by c0 . In this case the
equation has the exact solution (x, t) = f (x c0 t), where f is any function of the argument, so that any
disturbance travels without change. We can show this by considering the solution at a point x at a later time t + ,
where the solution is f (x c0 (t + )). Writing the argument as f ((x c0 ) c0 t), we can see that the solution
at the later time (x, t + ) = (x c0 , t), namely the solution that was originally upstream a distance c0 ,
showing that the solution is a simple wave of translation. It would be possible to solve problems approximately
by using a local value of c and using this "upwinding" form of solution, and writing the updated solution at x at
t + that which was upstream a disatnce c,where c now varies along the channel. This, however is a highly
approximate method, and not generally to be recommended.

2.3.12

The characteristic formulation of the equations

t
(xj , tn + )
U + C
U C

tn +

1
tn
xj1 x+
j
Figure 2.20:

xj

x
j

xj+1 x

Part of (x, t) plane, showing information arriving from upstream and downstream at velocities U C

The two long wave equations (2.23), which are partial differential equations, can be expressed as four ordinary
differential equations. Two of the differential equations are for paths for x(t), a path known as a characteristic:
dx
= U C,
dt
where U = Q/A is the mean fluid velocity in the waterway at that section and the velocity C is
r

gA
+ U 2 2 ,
C=
B

(2.81)

often described as the "long wave speed". It is, as equation (2.81) shows, actually the speed of the characteristics
relative to the flowing water. There are two contributions of C, corresponding to both upstream and downstream
propagation of information. Two characteristics that meet at a point are shown on Figure 2.20. The "downstream"
or "+" characteristic has a velocity at any point of U + C. In the usual case where U is positive, both parts
are positive and the term is large. As shown on the diagram, the "upstream" or "-" characteristic has a velocity
U C, which is usually negative and smaller in magnitude than the other. Not surprisingly, upstream-propagating
disturbances travel more slowly. The characteristics are curved, as all quantities determining them are not constant,
but functions of the variable A, B, and Q.
The other two differential equations for and Q can be established from the long wave equations:

Q
Q |Q|
Q
d dQ
Q2 B
B C
+ i C ,
+
= 2 S P
A
dt
dt
A
A2
A

(2.82)

On each of the two characteristics given by the two alternatives of equation (2.81), each of these two equations
holds, taking the corresponding plus or minus signs in each case. To advance the solution numerically means that
44

the four differential equations (2.81) and (2.82) have to be solved over time, usually using a finite time step .
Figure 2.20 shows the nature of the process on a plot of x against t.
The usual computational problem is, for a time tn+1 = tn +, and for each of the discrete points xm , to determine

the values of x+
m and xm at which the characteristics cross the previous time level tn . From the information about

+
and Q at each of the computational points at that previous time level, the corresponding values of +
m , m , Qm ,

and Qm are calculated and then used as initial values in the two differential equations (2.82) which are then solved
numerically to give the updated values (xm , tn+1 ) and Q (xm , tn+1 ), and so on for all the points at tn+1 .
The advantage of characteristics has been believed to be that numerical schemes are relatively stable. The lecturer
is unconvinced that they are any more stable then simple finite difference approximations to the original partial
differential equations, but this remains to be proved conclusively.
The use of characteristics has led to a widespread misconception in hydraulics where C is understood to be the
speed of propagation of waves. It is not it is the speed of characteristics. If surface elevation were constant
on a characteristic there would be some justification in using the term wave speed for the quantity C, as disturbances travelling at that speed could be observed. However as equation (2.82) holds, in general neither (surface
elevation the quantity that we see), nor Q, is constant on the characteristics and one does not have observable
disturbances or discharge fluctuations travelling at C relative to the water. While C may be the speed of propagation of information in the waterway relative to the water, it cannot properly be termed the wave speed as it would
usually be understood.

2.3.13

The behaviour of floods and waves in rivers

Linearised equation the Telegraphers equation

As in the analytical solution for steady flow of 2.3.6.3, here we linearise the equation by considering small
perturbations about a uniform flow of cross-sectional area A0 and discharge Q0 in a channel of constant slope,
S = S0 . We write V = A0 x Q0 t + , where is a small parameter such that is a small perturbation in the
volume potential about the uniform flow. In this case the two physical quantities A and Q are given by

V
= A0 + ,
x
x

V
Q =
= Q0
t
t
A =

When we substitute into equation (2.62) and express all nonlinear terms as series we ignore terms in 2 and higher.
In fact, the linearised version of the left hand side can simply be written down, as the coefficients of the second
derivatives on the left become simply their values for the undisturbed uniform flow. The right side is more difficult.
The quantity P is written as the two-dimensional Taylor series

(P )
(P )
(A A0 ) +
(Q Q0 ) + . . .
A
Q

(P )

= 0 P0 +
P0
+ ...,

A 0 x
Q 0 t
= 0 P0 +

where the resistance coefficient may also be a function of the discharge in cases (a) where bed forms can develop,
and (b) where it depends on the Reynolds number, as shown on the Stanton-Moody plot of the Colebrook-White
equation, figure 2.8. The velocity squared term becomes

Q0
t
A0 +
x

!2

Q0
A0

1
1+

Q0

A0

!2

= U02 1 2
2
+ O 2 ,
Q0 t
A0 x

where we have introduced U0 = Q0 /A0 . After introducing the symbols

(P )A =

(P )
A 0

and Q =
45

,
Q 0

the whole term on the right of equation (2.62) becomes

P0 Q
2
0 P0 + (P )A
U02 1 2
gS0 A0 +
x
t
Q0 t
A0 x
x
= 0 P0 U02 gS0 A0

2
2
+0 P0 U02

Q0 t
A0 x

+U02 (P )A
P0 Q
x
t

gS0 .
x

There are no terms at zeroeth order 0 on the left side, so the term on the right at this order has to be satisfied itself,
giving
g A0
U02 =
S0 ,
(2.83)
0 P0
which is the uniform flow relationship. Now considering all the terms on the left and right of equation (2.62) at 1
and bringing them together, and occasionally using equation (2.83) to eliminate U02 , we obtain the equation

2
2

2
2
(2.84)
+c
+ 2 + 2U0
(C02 2 U02 ) 2 = 0,
t
x
t
xt
x
where it has been convenient to introduce the symbols whose significance is described immediately below. This
is a Telegraphers equation, with a first-order "kinematic" part containing the effects of friction, and second order
contributions corresponding to the wave dynamics. Such an equation was obtained by ? and ? for wide channels.
The symbols introduced are, although the names and significance are not yet obvious, and will be justified further
below:
1. Decay rate :
gS0
=
U0

1 Q0
1 Q0
gS0 0
1+
=
1+
2 0 Q 0
A0 /P0
2 0 Q 0

(2.85)

which has dimensions of T1 , and measures the rate at which disturbances decay; it increases with both slope
and roughness and decreases with depth. In river flow problems this can usually be simplified, as the Reynolds
number of the flow is so large that there is no longer any variation with it, and /Q = 0.
2. Kinematic wave speed c:

(P )
1 A0
1 3
0 P0 A 0
3

c = U0
,
(2.86)

2
1 Q0
1+ 2
0 Q 0

is the speed of waves when all dynamic terms, of the order of F 2 are unimportant, which will be proved below.
This is a slight generalisation of the usual Kleitz-Seddon equation to allow for the resistance coefficient being
a function of discharge, which states that
dQ(A)
c=
,
dA
where the Q(A) is the discharge given by any of the usual resistance formulae.
For convenience, the symbol is introduced such that
c = U0 ,

(2.87)

where is roughly 3/2 from equation (2.86), and its most important variation is caused by the change of
perimeter with area. The role of this term as a "kinematic wave speed" will first be established below.
3. Dynamic wave speed C0 :
q

C0 = gA0 /B0 + 2 U02 ,

(2.88)

is the speed of waves when there is no friction or bed load, which will bepshown below. This is the formula for
the speed of characteristics, but the term "wave speed" widely used for gA/B is not justified, as is just part
of the expression (2.88), and it will be seen that even C0 itself is the wave speed only in the limit of no friction.
46

Behaviour of solutions modal analysis

We now obtain solutions of equation (2.84), which we believe to be a good model for small disturbances to the
flow. Unfortunately simple deductions from the form of the partial differential equation seem not to be possible.
However, we obtain useful insight by assuming that, as any disturbance can be written as a Fourier series, we
can consider just one term of that Fourier series and examine how it propagates, namely its speed of propagation
and its decay rate as it travels. This is only possible because the equation is linear in . ? have done this for a
two-equation formulation.
It is simpler to write the variation in x in complex
form, such that we represent all variation along the waterway

as the periodic wave exp (i kx), where i= 1 and k = 2/L, where L is a wavelength we are considering. The
exponential can be expanded: exp (ikx) = cos kx+i sin kx, thereby revealing more its periodic form. We write
for the general solution:
= exp (ikx t) = exp (ikx) exp (t) ,
(2.89)
showing the periodic behaviour in x, and where the behaviour in time t is contained in the term . The real part
of shows how the solution decays or grows in time, the imaginary part determines how the solution oscillates in
time. We substitute (2.89) into (2.84) and find that it satisfies the equation exactly, giving a quadratic equation for
:

2 2 ( + ikU0 ) + 2ikc + k 2 C02 2 U02 = 0

with solution:

q
= + i kU0 i k2 C02 2 + 2i kv

(2.90)

c = U0 + v,

(2.91)

in which the substitution for c has been made:

where v will be seen to be the speed of propagation of waves relative to the advection velocity U0 in the limit
of large resistance in the channel. These definitions of C0 in equation (2.88) and v here, are such that has been
subsumed in them.
It is possible to express equation (2.90) in standard complex notation (i.e. Real+i Imaginary). The symbol =
1 is introduced, +1 for downstream-propagating waves and 1 for upstream, and the symbol also introduced
for notational convenience
q
2
= (k 2 C02 2 ) + (2kv)2 .

It is insightful to consider real and imaginary parts. Substituting = R +iI into equation (2.89)
= exp (ikx t) = exp (ikx) exp (R t iI t)

= exp (R t) exp (ikx iI t) = exp (R t) exp ik x i I t ,
k
and the role of the two parts becomes clear:
1. Real part: R = < () is decay rate of the disturbance. The real part of equation (2.90) gives
< ()
=1

q
/ 2 + k 2 C02 / 2 1,

(2.92)

giving a formula for the rate at which disturbances decay. In most situations we expect this to be positive such
that they are actually damped. The details of this are not as important as the physical result that the rate of
decay depends on the wavelength in the form of k different wavelength components decay at different rates.
2. Imaginary part: the propagation velocity of the wave is I /k = = () /k, and from equation (2.90), if we write
this as
= ()
= U0 + C,
k
showing that there is an advection velocity U0 relative to which the waves propagate up- and down-stream at
speed C where
s

2
1

=
+1
,
(2.93)
C0
kC0
2 (kC0 )2
which is a function of the wavelength, expressed in terms of k = 2/L. This means that for waves travelling
on streams with resistance there is no such thing as a unique long wave speed, and waves of different lengths
travel at different speeds, contrary to many writings.
47

These propagation characteristics of the waves depend on just two dimensionless parameters /kC0 and kv/,
each of which contains the wavenumber k. It is more revealing to eliminate that between them, giving the two
parameters /kC0 and v/C0 :
1. The relative importance of resistance: = /kC0 = / 0 , using the dynamic wave frequency 0 = kC0 ,.
For flows that are not fast, equation (2.85) gives
s

L
L S0 0
S0 0
=

,
A0 A0
kC0
2 B
2D0
P
0

where the symbol D0 has been introduced for the mean depth D0 = A0 /B0 and where D0 A0 /P0 for many
channels that are wide. We will refer to as the resistance-slope-wavelength parameter, as it increases with
each of those quantities.
2. Vedernikov number (2.95), Ve = v/C0 , which is a function of the Froude number from equations (2.87) and
(2.91):
v
( ) F0
Ve =
=q
,
(2.94)

C0
1 + 2 F 2
0

so that as 1.5 to 1.7, and 1.05, for relatively slow flows, Ve is roughly 0.5F0 and we can think of it as
a measure of the velocity of flow in the channel.

(a)Decay rate for downstream-travelling waves

1

Ve = 1/8, F0 0.25
Ve = 1/4, F0 0.5

Dimensionless
decay
rate
0.5
< (0 ) /

Ve = 1/2

Ve = 1

(b) Wave speed

Ve = 1

Dimensionless
wave speed
C/C0
0.5

Ve = 1/2
F0 0.5, Ve = 1/4

F0 0.25, Ve = 1/8
0
0.01

0.1
1
10
Resistance-slope-wavelength parameter = /kC0

100

Figure 2.21: Wave propagation parameters and their dependence on the resistance-slope-wavelength parameter and Vedernikov number: (a) Decay rate for downstream-travelling waves, (b) Wave speed

Now we can examine the nature of wave propagation by plotting the decay rate and propagation velocity as functions of these two parameters, given in figure 2.21. Part (a) shows the variation of the dimensionless decay rate
< () / for downstream-travelling waves from equation (2.92), while (b) shows the dimensionless wave speed
C/C0 from equation (2.93).
Decay rate of waves:
Considering the decay rate, figure 2.21(a) shows that for small values of the decay rate depends on the Vedernikov number Ve . This can be explored mathematically by writing equation (2.92) as a power series in , valid
48

when that is small, which gives

< (0 )
= 1 Ve + O 2 .

For downstream-propagating waves = +1, and the curves on the left side of figure (a) tend to this value, 1 Ve .
For upstream propagating waves, = 1, the dimensionless decay rate is 1 + Ve , larger than for downstream
waves.
Possibly the most important result seems to be that for Ve = 1, < (0 ) = 0, and there is no decay. This is the
boundary of instability, and here can be established by performing simple algebraic operations for the stability
condition < () 6 0 that leads to the criterion for stability
Ve =

v
1,
C0

(2.95)

where v/C0 is the Vedernikov number Ve . (?) first obtained solutions like (2.90) for a wide rectangular channel
with = 1, and obtained the equivalent criterion for stability that S0 4 in present terms. Substituting into this
equation gives the condition for stability
The criterion v C0 for stability (equation 2.95)can be further examined. Using the definitions of v and C0 from
equations (2.88) and (2.91) and equation (2.94) gives for stability:
c2 2cU0 + U02 gD0 .

The kinematic wave speed c is proportional to the mean fluid velocity U0 , from equations (2.86) and (2.87),
c0 = U0 , which gives the stability condition:
F02 =

U02
1
2
.
gD0
(2 1)

(2.96)

For a wide channel, the Weisbach and Chzy expressions have = 3/2, while if Gauckler-Manning-Strickler is
used = 5/3, giving the stability criteria:
1/2
,
9/8
3/7
Gauckler Manning Strickler F02
.
25/21
F02

Weisbach Chezy

(2.97)
(2.98)

The denominator in the two cases goes to zero at 1.13 and 1.19 respectively. It has been suggested (?) that
incorporates turbulence effects as well, and a typical value might be greater than 1.05, such that the denominator
in both (2.97) and (2.98) is small and the limiting values of Froude number obtained might be large. These limits
are plotted in Figure 2.22. The stability depends strongly on the resistance law and the velocity distribution. If the
channel were of finite width, the value of would be smaller, and equation (2.96) predicts a larger limiting value
of F0 for instability. Above this value the wave system is unstable, and we have the development of Roll Waves.
Propagation speed of waves:
Now the possibly more important quantity of wave speed can be examined. Considering figure 2.21(b) it can
be seen that for small values of resistance-slope-wavelength, all waves travel at the same speed C0 , however, in
general this is not the case. This contradicts one of p
the bases of much hydraulics, where it is widely stated and
believed that all long waves travel at the same speed, gA0 /B0 . The figure shows that this is simply not the case,
and the propagation speed is a function of the resistance-slope-wavelength parameter and the Vedernikov number,
and takes on very different values.
For small , equation (2.93) can be shown to have the solution

1
C
=1
1 Ve2 2 + O 4 ,
C0
2
q

and in the limit of very small we get C = C0 = gA0 /B0 + 2 U02 from equation (2.88), so that the
speed is indeed
independent of wavelength in this limit. In the case = 1 we recover the widely-believed result
p
that C = gA0 /B0 .
For large , however, equation (2.93) gives

1 Ve2
C
4
= Ve 1 +
+
O
1/
,

C0
2 2
49

Unstable

3
Froude
number
F0

Stable
2

= 3/2, Weisbach-Chezy
= 5/3, Gauckler-Manning-Strickler

0
1

1.05

1.1

1.15

1.2

Velocity distribution parameter

Figure 2.22: Stability limits for flows in wide channels and their dependence on velocity distribution parameter and the
exponent in the friction formula

as can be seen on the figure. For large values of resistance-slope-wavelength, we then have C C0 Ve = v, such
that the velocity of waves is U0 v, which is the so-called kinematic wave speed c in the downstream case.

Both these results can be identified on the figure. The first is the traditional result, that waves propagate at the
"long wave speed", C = C0 , where C0 is actually modified by velocity terms, equation (2.88), which we see here
is valid only for small . For larger values, the waves do not travel at C0 , but at speeds intermediate between that,
and in the limit, the kinematic wave speed C = v , or, C0 = v0 . This too, is widely known, but the general result,
that wave speed depends on wave length seems also to be widely unknown.
Conclusions
The long wave equations have been widely interpreted as a simple hyperbolic system where disturbances travel at
a constant wave speed. That is incorrect. In fact the system is an advective-diffusive-dispersive system, where the
propagation speeds of disturbances and their rates of decay all depend on the wave length of the disturbances. The
behaviour is much more complicated than the simple hyperbolic inferences suggest.
Any numerical solution of the equations should give the correct behaviour; what is not correct are simple inferences
as to the behaviour based on the constant wave speed belief, and an accompanying acceptance that effects of
resistance are small. They are not. And, they are complicated.

2.4 Structures, controls, and boundary conditions

Generally the treatment of boundary conditions in the one-dimensional open channel equations is decidedly nontrivial.

2.4.1

Upstream boundary conditions

At the upstream boundary x = x0 we often have the condition that the inflow is known as a function of time,
Q (x0 , t) = Qin (t). Typical conditions at an upstream boundary are shown on the left side of figure 2.23. If we
do not use the method of characteristics, but instead some finite difference scheme, the solution at time t may be
known at a finite number of computational points x0 , x1 , . . . shown by the crosses, and it is required to calculate
the solution at time t + . If we use, for example, and Q as the dependent variables, we still have to determine
the value of the other variable at the boundary, as only one condition can be specified in a sub-critical flow. We can

50

(xi , t + )
t+

dx
dt

0
C

= u + c0

dx
dt

C+

= u c0

t
x0
Figure 2.23:

x1

x+

xi

Computational points and typical characteristics on (x, t) axes

use the mass conservation equation (2.23a)

1 Q
i

+
= ,
t
B x
B
and from the point values of Q at t an estimate of Q/x can be made and a value of /t calculated and used
to give a prediction of (x0 , t + ). It is clear that only simple low-accuracy numerical approximations are being
made. Interior values are updated using approximations to the two partial differential equations at the interior
points.
0
If the method of characteristics were to be used, then a characteristic such as C
would be used, and for a finite
difference approximation of the characteristic equation such as (2.82), the value of (x0 , t + ) calculated. Again,
low-order numerical approximations would be made.
Rx
If the upstream volume were used as a dependent variable, then as generally V /t =
i dx0 Q, at the upstream
boundary we obtain the ordinary differential equation dV (x0 , t) /dt = Qin (t), which can be solved by simple
or by higher order methods.

2.4.2

Downstream boundary conditions

If anything, the downstream condition is more difficult. If it were say, the sea with tides or a reservoir with a
specified surface level, (L, t), that is relatively simple, and the other condition would be obtained from application
of the momentum equation in terms of Q/t there. Alternatively, the method of characteristics could be used,
this time with a C+ characteristic. The third alternative of using V is rather less simple, as specification of as a
function of time implies that A(L, t) can be calculated, giving the derivative V /x as a function of time.
More complicated is the condition if there is a control at the downstream point, such as a weir or sluice, where Q
is a known function of . In this case it is necessary to use the momentum equation in terms of Q/t to give an
updated approximate value of Q(L, t + ) and the equation of the control solved to give the updated value of .
In the case of V this will be a value of V /x.
At open boundaries, where an arbitrary end to the computational domain is proposed, the problems are greater,
although we may simply be able to apply the partial differential equation as if it were an interior point, which is a
point that the lecturer has yet to resolve satisfactorily.

2.4.3

Hydraulic structures on channels

Hydraulic structures are anything that can be used to divert, restrict, stop, or otherwise manage the natural flow of
water. They can be made from materials ranging from large rock and concrete to obscure items such as wooden
timbers or tree trunks.
A dam, for instance, is a type of hydraulic structure used to hold water in a reservoir as potential energy, just as
a weir is a type of hydraulic structure which can be used to pool water for irrigation, establish control of the bed
(grade control) or, as a new innovative technique, to divert flow away from eroding banks or into diversion channels
for flood control.
The main texts to which reference can be made are ?, French (1985?), Henderson (1966?), Novak (2001?).
Overshot gate - the sharp-crested weir
The conventional analysis of flow over a sharp-crested weir is one of the great misleading results in hydraulic
engineering, which is widely known to be so, yet every 51
textbook reproduces it. Instead of the actual pressure distribution shown dashed in Figure 2.24, a hydrostatic distribution is assumed, shown dotted, which gives maximum
pressure at the crest, whereas it is actually zero there, then Bernoullis law is applied, and then it is assumed that
the fluid velocity is horizontal only (whereas it is actually vertical at the crest!). Results obtained for the horizon-

Assumed pressure distribution

Hypothetical actual pressure distribution
D

Figure 2.24:

Side view of sharp-crested weir

0.8

Accurate computations
Classical theory

0.6
z/h 0.4

0.2

0.0

Figure 2.25:

0.2

0.4

0.6

0.8
1

u/ gh

1.2

1.4

1.6

Horizontal velocity distribution over crest - simple theory and accurate computations

hence we obtain the well-known 3/2 law:

q = C gH 3/2 .

If we now consider a finite width of channel B, finite width of weir b, and a finite apron height P , with now a total
discharge Q we obtain
Q
p
= f (b/B, H/P, b/P ).
b gH 3

Following traditional forms we write the quantity on the right in terms of a discharge coefficientCD as

2
Q
p
= CD 2.
(2.99)
3
3
b gH

It is a strong temptation to incorporate the 2/3 and the 2 in with the CD , as both are artefacts of the inappropriate
use of Bernoullis theorem. In 1924 the Schweizerischer Ingenieur- und Architekten-Verein (SIA), the Swiss Society
of Engineers and Architects, published a Standard containing the following expression for CD , which we have here
converted from the dimensional expression presented in ? and a particular case presented by Rehbock:
!

2
4
2 !
1 b
b
6 5 (b/B)2
H/P
1+
+ 0.615
, (2.100)
CD = 0.578 + 0.037
B
2 B
1 + H/P
0.047H (g/ 2 )1/3 + 1.6

1/3
in which the last term in the first bracket allows for real fluid effects. The dimensionless expression 0.047H g/ 2
appeared as 1000 H in the original, where 1000 is actually a dimensional one of units m1 . Here we have re52

expressed it in terms of the only possible combination of gravitational acceleration g and viscosity , assuming
that values of g = 9.8 ms2 and = 106 were used to arrive at the value of 1000 m1 .

An approximate expression from equations (2.99) and (2.100) is Q = 0.6 gb (1 zc )3/2 ,where 1 is the upstream surface elevation and zc is the elevation of the crest.
Triangular weir
An analysis as misguided as the traditional one produces the result quoted by ? on p352, where some results of ?
are quoted. The expression is

8p
2g tan H 2.5 ,
Q = CD
15
2
where CD is the coefficient of discharge, and where is the angle between the sides of the triangle. A typical result
from ? is that CD is roughly 0.58, which has been found to agree well with experiment (see ?). We prefer to write
the expression as
2.5
Q = 0.44 tan
gH ,
2
suggesting that it is empirical.
Rapidly-varying flow problems
In the following sections we will be considering steady rapidly-varied flow. The horizontal distances are small
enough that we ignore effects of friction and slope, as they are roughly in balance anyway, and are not so important
for these problems. The energy and momentum fluxes are:

Q2
Energy flux at a section :
Q g +
= Constant,
2 A2

2
+ Q
= Constant,
Momentum flux at a section :
A g h
A2

is the depth of the centroid below the surface. As Q is usually

where is the elevation of the free surface and h
constant along the channel, we can usually consider the total head to be a constant:
+

Q2
= Constant.
2g A2

It is convenient to consider the Specific Head, the head relative to the bottom of a channel, defined by
H =h +

Q2 1
,
2g A2 (h)

(2.101)

where h is the depth of flow and area A is a function of h. As the total head is constant, and we can write = Z +h,
where Z is the elevation of the channel bottom, we have
H + Z = constant.

(2.102)

Broad-crested weirs critical flow as a control

A typical relationship for H(h) is plotted in figure 2.26(a), where both sub- and super-critical limbs of the curve
can be seen. Now consider a sub-critical flow at 1 such as that shown in part (b) of the figure, where the bottom
of the stream rises by an amount . As total head is constant, from equation (2.102) H1 + Z1 = H2 + Z2 , and
H2 = H1 , and this is shown in part (a) (note that the change of is plotted horizontally!), and the depth h
also decreases as shown, from h1 to h2 .
Consider now the case where the step is high enough that the flow is brought to critical there, h2 = hc , where
such an hc is shown on figure 2.26(a). If the step were raised even more, then the depth of flow over the raised bed
would remain constant at hc and the upstream depth would have to increase so as to make the flow possible. The
step is then acting as a weir, controlling the flow.
Consider the situation where the bed falls away after the horizontal section, such as on a spillway, as in figure 2.27.
The flow upstream is subcritical, but the flow downstream is fast (supercritical). Somewhere between the two, the
flow depth must become critical - the flow reaches its critical depth at some point on top of the weir, and the weir
provides a control for the flow. In this case, the total head upstream (the height of the upstream water surface above
the sill) uniquely determines the discharge, and it is enough to measure the upstream surface elevation where the
flow is slow and the kinetic part of the head negligible to provide a point on a unique relationship between that
53

(a)

(b)

Flow
depth
h

h1
2

h2

h2

H =h+
hc

h1
Q2 1
2g A2 (h)

Specific head H

Figure 2.26: (a) Head-surface relation for a particular discharge and cross-section, showing the change in depth if as in (b)
the bottom rises by

hc
hc
hc
Figure 2.27:

Spillway as broad-crested weir

Small energy loss

hc

h1
hc

Figure 2.28:

hc

Broad-crested weir in a canal showing substantial recovery of energy

head over the weir and the discharge. No other surface elevation need be measured. Such a horizontal flow control
is called a broad-crested weir.
In recent years there has been a widespread development of broad-crested weirs placed in canals where the flow is
subcritical both before and after the weir, but passes through critical on the weir, as shown in figure 2.28. There is
a small energy loss after the flume. The advantage is that it is only necessary to measure the upstream head over
the weir.We can apply simply theory of critical flows to give the discharge per unit width q:
q=
where

q
p
3 q 8
3/2
gh3c = g 23 H1 = 27
gH1
H1 = h1 +

q2
,
2gh21

the head upstream. As H1 is a function of h1 , the relationship for discharge q is not simply a 3/2 power law, but
for sufficiently low upstream Froude number the deviation will not be large. In practice the discharge for the weir
54

as a whole is written, introducing various discharge coefficients ?,

3/2
Q = 0.54Cs Ce Cv b gh1 .
Free overfall

hc hb

Figure 2.29:

Free overfall, showing location of critical depth hc and brink depth hb

Consider the free overfall shown in Figure 2.29. It can be shown that the discharge per unit width is q =
3/2
1.65 ghb .
Sluice gate or undershot gate

h1
h2

Figure 2.30:

Sluice gate for tutorial example

Now we consider the case where the gate is not drowned, but a stream of supercritical flow can exist for some
distance downstream. Applying the energy theorem between a point at section 1 and one at 2:
2
2
1 q
1 q
= gh2 +
.
gh1 +
2 h1
2 h2
Solving for q we obtain:
r
2g h1 h2

q=
h1 + h2
which it is easier to write in dimensionless form:

h2 /h1
p
=p
.
3
1 + h2 /h1
2gh1
In practice we know the upstream depth h1 and the gate opening D, such that
q

(2.103)

h2 = Cc D,
where Cc is the coefficient of contraction. A number of theoretical and experimental studies have been made of
this, but the variation is not large. ? shows some of these but ends up recommending a constant value of 0.61.
Equation (2.103) is a convenient way of representation, as h1 is given, and so we find that the dimensionless
discharge is simply a function of h2 /h1 , the depth ratio, or D/h1 . In fact, the variation is rather linear if we
expand the denominator of the right side of equation (2.103) using the binomial theorem, we obtain

h2
1 h2
q
p
=

+
.
.
.
,
1

h1
2 h1
2gh31
and we find that to first order, for small gate openings,
p
q 2gh1 h2 ,

and the result is almost obvious, that the discharge per unit width is given by a velocity corresponding to the full
upstream head of water multiplied by the downstream depth of flow.
55

Drowned sluice gate

h1
h2
h0
D
hc

Figure 2.31:

A drowned sluice gate

Consider Figure 2.31, a rather abstract representation of the highly turbulent flow just downstream of the gate,
where the gate is drowned. In this case the analysis follows that suggested on p208 in ?. Applying the energy
theorem between a point at section 1 and one at c:
gh1 +

q
h1

= gh0 +

q
hc

(2.104)

where we recognise that the pressure contribution is given by the hydrostatic force of depth h0 but that the velocity
contribution is due to flow in the jet only. Solving for q we obtain:
s
r
2g
h1 h0
.
(2.105)
q=
h1 hc

h21 h2c
And now we apply the momentum theorem between sections c and 2:

q2
q2
+ 12 gh20 =
+ 1 gh2 .
hc
h2 2 2

Again solving for q we obtain

q=

g
2

hc h2 (h22 h20 )
.
h2 hc

(2.106)

(2.107)

Now if we eliminate q between the two equations (2.105) and (2.107) we obtain the quadratic in h0 :
h20 +

4h21 hc (h2 hc ) (h1 h0 )

h22 = 0,
h2
(h21 h2c )

(2.108)

and if we introduce the depth scale h4 defined by

h4 =

1 hc /h2
,
hc
1 (hc /h1 )2

(2.109)

equation (2.108) becomes

h20 + 4h4 (h1 h0 ) h22 = 0,
and the solution is :
h0 = 2h4 +

q
4h24 4h1 h4 + h22 .

(2.110)

In practice we know the upstream and downstream depths h1 and h2 as part of the computations. If the gate
opening is D, also presumed to be known, then
hc = Cc D,
where Cc is the coefficient of contraction which we may assume to be 0.6. Hence we can calculate h4 from (2.109),
h0 from (2.110), and the discharge per unit width from (2.105) or (2.107).
56

2.4.4

Interior conditions

Obstacles such as bridge piers

When water in a river or canal flows past an obstacle such as a bridge pier or woody debris, there are accompanying
momentum and energy losses, and the water level is different on either side of the obstacle. In sub-critical flow the
water level is higher upstream, this backwater causing a greater risk of flooding. In super-critical flow, the level is
higher downstream.
The problem has not been researched as much as it might and there has been some ambiguity about the governing
principles. As early as the 1850s energy conservation was used. In the early 20th century some large experimental
campaigns were carried out, but results obtained were empirical. They were also widely variable, and it has been
difficult to use them for a particular application. In more recent times there has been a gradual shift towards using
momentum conservation. The resistance force of an obstacle to a flow causes an immediate momentum loss to the
flow. In comparison most energy losses are distributed, in boundary layers, shear layers, separation zones, vortices,
and turbulent and viscous decay in the wake after the obstacle. Momentum loss is easier to measure, in the form
of the force on an obstacle, and easier to model, in terms of drag coefficients.
Here we consider the backwater caused by finite obstacles, provided that they do not change the overall nature
of the flow. This means, for example, that situations are excluded where an embankment partially blocks the
waterway and alters the flow from a generally one-dimensional one.
The main problem is in calculating the force on an obstacle. This work separates it into two parts, one due to
viscous and form drag, the second, termed free surface resistance, is due to the variation of the free surface around
the obstacle, for which an approximate theory is developed. Applying momentum conservation gives an implicit
equation for the backwater in terms of the force on the obstacle. Then approximate explicit formulae are obtained
which show considerable sensitivity to geometry and flow, such that it is not surprising that experimental results
have shown much scatter.
Some implications for practice are discussed. A hierarchy of different levels of treatment is suggested, depending
on the importance of backwater and the importance of the project. It is suggested that in any investigation the
magnitude of the backwater be calculated approximately using the approximate theory. If small it can be neglected
or added to the boundary resistance, for which formulae are given.
Figure 2.32(a) shows a typical problem of a sub-critical flow with a surface-piercing pier that has caused the mean
water level behind it to rise to some excess or backwater height above the water level determined by conditions
further downstream. Far upstream the flow is unaffected. As it approaches the obstacle, the water depth gradually
increases, until it reaches the value required to overcome the momentum loss at the obstacle. The flow divides
around it so that there may be a stagnation point on the upstream face with a small mound just upstream. It is a
local effect only, so that a traditional slowly-varying one-dimensional approach to the hydraulics is still possible
until just before the mound, at a section denoted by 1 in figure 2.32. As the flow divides and accelerates around
the pier, the surface level drops, and waves may form. Generally the level becomes a minimum beside the pier,
or a little way downstream. By the end of the pier, any momentum loss has already occurred, however most
energy loss is still to occur in the wake region downstream. The flow has separated from the sides of the obstacle,
leaving a relatively stagnant zone behind it. The mean water level just behind the pier may be lower than that
further downstream, however it recovers quickly. Section 2 on the figure is deemed to be at a point some little
way downstream where the flow can again be considered one-dimensional for hydraulics purposes. The solid
line on the figure shows the actual water level on the sides of the obstacle, and on the axis of the obstacle, upand down-stream. Its rapid variation shows the local dynamical effects near the obstacle. The dashed line shows
the streamwise variation of the mean surface level across the stream, which is not a visible quantity, but which
determines the overall momentum balance at a section.
Part (b) of figure 2.32 shows the physical idealisation adopted to calculate the backwater , assuming flow in a
horizontal prismatic frictionless channel with an arbitrary origin, common in other rapidly-varying flow problems.
Defined as = 1 2 , it is positive for the more common case of sub-critical flow.
Drag force
The turbulent, three-dimensional, rapidly-varying free-surface flow near obstacles in practice is so complicated that
at the time of writing it is not yet feasible to calculate the forces on them from fluid-mechanical principles. The part
of the force due to viscous and form drag TD can be estimated using the empirical expression from conventional
fluid mechanics for the drag on a bluff body (see for example, 7.6 of White 2003?): TD = 12 CD u2 a, where
is fluid density, CD is drag coefficient, u is the local horizontal fluid velocity impinging on the object, and a
is the projected area of the object normal to the flow direction. For values of CD reference can be made to most
introductions to fluid mechanics, however they are almost all for the case of fully-immersed bodies with few shear
57

Surface if no obstacle: slowly-varying flow

Surface along axis and sides of obstacle
Mean of surface elevation across channel

(a) The physical problem, longitudinal section showing backwater

at obstacle decaying upstream to zero

(b) The idealised problem, uniform channel with no friction or slope

Figure 2.32:

A typical physical problem flow past a bridge pier

or free-surface effects. The general backwater problem is more complicated as the incident flow velocity on a
body depends on its position in the flow, whether near the bottom, surface, or extending across the whole flow. The
obstacle may be a compound body with different components, such as the piers and roadway of a bridge and each
element may be subject to a different velocity. For example, the appropriate velocity for a vertical pier extending
over the whole depth would be the mean velocity in the flow; that for a bridge deck would be the surface velocity,
while for a vane or block near the bed it would be rather smaller. All those could be plausibly combined linearly.
Here for a single element a simple dimensionless factor is introduced here such that the drag force is written in
terms of the mean velocity in the channel Q/A:
TD = 12 CD a

Q2
.
A2

(2.111)

where the mean velocity incident on the body is given by u2 = Q2 /A2 , where Q is the discharge in the channel
and A its cross-sectional area. The coefficient depends on the bodys position in the flow; for a pier that extends
from top to bottom it could be calculated from knowledge of the turbulent logarithmic velocity distribution, or
simply lumped in with the drag coefficient.
Free surface resistance
When flow passes a body that penetrates or approaches the free surface, the pressure variations around it cause
surface disturbances whose magnitudes depend on the Froude number. Near the front, the stagnation effect leads
to an increase in the water level locally so that there is a crest near the upstream face. Around the sides, where
the water accelerates to pass around the body, a wave system originates, as seen in figure 2.32 for a fast flow.
Part of the resistance of the obstacle, then, is determined by the shape of the free surface immediately around
the obstacle. This suggests that that component could be described as "free surface resistance" rather than the
term "wave resistance" used in naval architecture. Eventually the momentum possessed by the wave train passes
into the water by breaking and dissipation (except for any water that actually breaks onto the river banks) and is
incorporated in the momentum of the downstream flow.
Here the effects of the free surface are assumed given by the net hydrostatic force on the obstacle as given by the
varying water level around it. In the momentum approach of HEC (2002?, equations 5-1 and 5-3) a similar model
is used. There the gradually-varied flow equations are solved numerically between the front and rear faces. Here it
58

is suggested that the variation is rapid enough that that is questionable and instead a lumped local approach can be
used.
The horizontal hydrostatic force on an arbitrary curved surface is given by a surprisingly-simple expression (see,

for example, White 2003?, 2.6, or almost any other book on elementary fluid mechanics). It is gah,where
a is the

projected area surface perpendicular to the force direction (streamwise here) and h is the depth of the centroid of
the projected area. The net force on the obstacle due to free surface effects is the difference between contributions
on the upstream (US) and downstream (DS) projections:

TS = g ah
g ah
.
(2.112)
US
DS
For relatively small backwater the
between the two contributions can be written as the first term of the
difference

/, where is the free surface elevation. The resulting expression can be

Taylor series: g (US DS ) ah
shown to be
TS = ga (US DS ) ,
(2.113)

in terms of the water level difference between the upstream and downstream ends, which depends on the length of
the disturbance around the body. For a slow flow when the wavelengths are short there may be several wave crests
and troughs around a pier, although they might be barely perceptible; for a faster flow the wavelength and height
are larger; in general the water level at the rear of the body is an oscillatory function of stream velocity, and so is the
net force on the body. This is well-known for the resistance of ships, which are designed to minimise such forces.
For bluff objects such as bridge piers the effect should be rather more marked. With increasing fluid velocity, the
oscillations should increase in height and length, until there is a single wave trough near the downstream face,
such that the length of the pier is half the wavelength, when the force should show a local maximum. For higher
velocities and wavelengths the trough would be further downstream from the pier and there should be no more
oscillations in the force until the flow approaches critical beside the pier.
A crude but simple approximation is now made. It is assumed that, in keeping with the wavelike nature of disturbances, the water surface around the pier is in the form of a cosine wave with an amplitude proportional to the
overall level difference between upstream and downstream such that = 2 +c cos kx, where c is a dimensionless factor, k is the wave number, and x has origin at the upstream face. It is expected that c will be 1 or
slightly greater so as to agree with the actual free surface on the upstream face at x = 0. On the downstream face
where x = L in reality there may be a finite level drop as water passes around the corner from the side of the pier,
which is subsumed in the coefficient c here. According to this model the difference between upstream and downstream levels US DS =c (1 cos kL), where L is an effective obstacle streamwise length. Substituting into
equation (2.113), this means that
TS = gac (1 cos kL) .
(2.114)
For small waves on a slow flow one can imagine short periodic waves around the side of an obstacle, and a cosine
wave may be quite a good approximation. For more dramatic flows, such as that shown in Figure 2.32, it will be
less so, nevertheless we proceed with it as a model.

To determine k, if a dimensional analysis is performed for the idealised problem of small short two-dimensional
waves of wavenumber k on a flow of speed U , the relationship is obtained that kU 2 /g = , where is a constant.
Hence we write k = g/U 2 and use it in equation (2.114). Introducing the dimensionless symbol CS gives
TS

= gaCS ,

where

L
CS = c (1 cos kL) = c 1 cos 2
,
(2.115)
F h

a function of the square of the Froude number F = U/ gh and L/h, the ratio of streamwise dimension to water
depth h at the obstacle. This shows the unexpected behaviour, that for small F , the argument of the cosine function
becomes large and small variations in F will cause the cosine function and hence CS to oscillate between values
of 0 when a crest is also near the downstream face, and 2c when a trough is there. It will be seen that this has the
capacity to explain fluctuations in experimental results.
Total resistance force
Combining equation (2.111) for the fluid dynamic drag and equation (2.115) for the free surface resistance gives
the total resistance force of an obstacle in a stream flow:
T = 12 CD a1

Q2
+ ga1 CS ,
A21

(2.116)

in which a = a1 and A = A1 , respectively the projected area of the obstacle perpendicular to the flow direction,
and the flow cross-section, both given by the elevation of the free surface at the upstream section 1. Figure 2.32
59

shows how the depth and velocity at 1 are those most closely approximating the flow that actually impinges on the
body, rather than that at 2.
The coefficient CS as it appears in equation (2.116) is another form of resistance coefficient. What has customarily
been measured or used in calculations, has been the apparent drag coefficient, which includes both fluid dynamic
and free surface resistance, written here as CD0 :
T = 12 CD0 a

Q2
.
A2

(2.117)

In applying this, Gippel et al. (1996) used the values of a and A upstream at 1. On the other hand several people
have used the downstream conditions at 2. There is a practical reason for that, as in sub-critical flow problems
it is the depth there that is known, determined by conditions even further downstream. However, that can be
incorporated in the analysis, and it seems that it is more rational to use the actual fluid velocity that impinges on
the body, given more accurately by the conditions at section 1, as in equation (2.116).
Application of momentum theorem
The momentum flux M at any section of gradually-varying open channel flow is, assuming a hydrostatic pressure
distribution,

Q2

M = gAh +
.
(2.118)
A
is the depth of the centroid of the
where g is gravitational acceleration, A is cross-sectional area of the channel, h
is the first moment of area of the section about a transverse axis at the
section below the surface, such that Ah
water level, and is a Boussinesq coefficient allowing for non-uniformity of velocity over the section.
The conservation of momentum theorem states that M1 M2 T = 0. Substituting expression (2.118) for M at
1 and 2 and the expression for total resistance force T from equation (2.116), and dividing by gives

+Q
gAh
A

12 CD a1

Q2

gAh +
A 2
1

Q2
ga1 CS ( 1 2 ) = 0.
A21

(2.119)

1 are known functions

For sub-critical flow and known downstream conditions at point 2, and where both A1 and h
of 1 , this is a transcendental equation for 1 that can be solved numerically for a given flow and channel crosssection. For super-critical flow, where the flow depth increases after the obstacle, one would instead calculate 2
from a knowledge of upstream conditions at 1.
Approximate explicit solutions
Numerical solution of equation (2.119) is not difficult, and any root-finding method could be used, such as trialand-error, bisection, the secant method or Newtons method, although for the latter the derivative of the left side
with respect to 1 is complicated. Instead, to obtain more physical insight it is useful to obtain approximate explicit
solutions, which will be done here.
Consider the physical quantities associated with the two water levels, at 1 and 2. The change of surface elevation
is usually small relative to the overall depth. The momentum equation (2.119) is now written for sub-critical
flow with all quantities at section 1 expressed as Taylor series in terms of the usually-known conditions at section
2:
A1

Ah 1
a1

= A2 + B2 + 12 B20 2 + O 3 ,

2 + A2 + 1 B2 2 + O 3
= Ah
2

= a2 + b2 + O 2 ,

(2.120a)
(2.120b)
(2.120c)

where B is the surface width, B 0 = dB/d is its derivative with respect to surface elevation which is a function of
the side slopes at the stream banks. The Landau order symbols O (. . .) denote the order of the neglected terms in
each case.
Substituting into the momentum equation (2.119) and using power series operations such as use of the binomial
60

theorem, gives the series in the dimensionless backwater / (A2 /B2 ), where A2 /B2 is the mean depth:

b2 A2
2 +
1 F22 2 CS + 2 2
+
A2 /B2
a2 B2

B20 A2
A2 b2
2
1
1
2 CS
2 + F2
A2 /B2
2B22
B2 a2
3

= O , 2 2 ,
(2.121)

where 2 = 12 CD 2 F22 , a dimensionless expression of the force on the structure, 2 = a2 /A2 is the blockage
ratio of the obstacle were the surface level to be that at the downstream section 2; and F22 = Q2 B2 /gA32 , the square
of the Froude number there.
First-order solution
It is insightful to consider the first-order solution, for which we recognise that the dimensionless backwater
/(A2 /B2 ) and the dimensionless force 2 are the same order of magnitude, so ignoring all second-order terms
in equation (2.121) by taking just the terms linear in 2 and /(A2 /B2 ), the lowest-order solution is obtained

A2 /B2

=
=

2
1 F22 2 CS
a2
F22
1
2 CD A 1 F 2 C .
2
2 S
2

(2.122)

This is a generalisation of Fenton (2003?), an explicit approximation for the relative change of surface level in terms
of the stream conditions and the geometry and resistance characteristics of the obstacle. For sub-critical Froude
numbers, such that F22 + 2 CS < 1, the denominator is positive and so is the backwater 2 . If F22 + 2 CS 1,
the flow approaches choking, such that the backwater is large, and the theory breaks down.
Conclusions
A model has been proposed for the force on an obstacle due to surface variation around it and approximate formulae
for backwater have been obtained in terms of drag and free surface resistance coefficients. Experimental results
agree with the predictions from this model, which shows how they may be more variable than would be the case for
prototypes, which may explain why there are few results for practical applications. However the present work also
does not give results in the form of numerical values of resistance coefficients; it provides a basis for understanding.
A hierarchy of different levels of treatment is suggested, depending on the importance of backwater and the importance of the project. Initially an approximate solution to the problem can be used to estimate the backwater
relative to the effects of boundary resistance. If the backwater is small, it can be neglected. If it is not so small,
a correction can be made to the boundary resistance coefficient used, which is usually not known accurately anyway. If the backwater is relatively large, and the project important, there are two main implications. The first is
that in a numerical model the spatial discretisation must be changed and the obstacle has to be treated as a discontinuity of elevation in the stream. The second is that the problem of evaluating the magnitude of the backwater is
difficult, as there are few reliable results, and any that are available are for quite particular geometries. The best
solution would be to perform load tests on a model obstacle which, it is suggested, can be done in a relatively short
laboratory installation.

61

Chapter 3

Measurement and analysis

3.1 Hydrometry and the hydraulics behind it

3.1.1

Definitions

In English, the traditional word used to describe the measurement of water levels and flow volumes is "Hydrography". That is a substantial ambiguity, for that word is also used for the measurement of water depths for navigation
purposes, has been so used since the great navigators of the eighteenth century, and organisations with names like
"*** Hydrographic Service" are usually only concerned with the mapping of an area of sea and surrounding coastal
detail.
Here we are going to follow Boiten (2000?), who provides a refreshingly modern approach to the topic we are
concerned with, calling it Hydrometry the measurement of water, which in the past has received little research.
Below, a practitioner will be called a hydrometrician, but the term hydrograph will be retained for a record, either
digital or graphical, of the variation of water level or flow rate with time.
There is a website with much information of a rather practical nature in the Water Measurement Manual on
http://www.usbr.gov/pmts/hydraulics_lab/pubs/wmm/. It is remarkable, however, that a field so important has
received such little benefit from hydraulics research.

3.1.2

Water levels

Water levels are the basis for any river study. Most kinds of measurements, such as discharges, have to be related
to river stages (the stage is simply the water surface height above some fixed datum). Both stage and discharge
measurements are important. Often, however, the actual discharge of a river is measured rarely, and routine measurements are those of stage, which are related to discharge.
Water levels are obtained from gauges, either by direct observation or in recorded form. The data can serve several
purposes:
By plotting gauge readings against time, the hydrograph for a particular station is obtained. Hydrographs of a
series of years are used to determine duration curves, showing the probability of occurrence of water levels at
the station or from a rating curve, the probability of discharges.
Combining gauge readings with discharge values, a relationship between stage and discharge can be determined, resulting in a rating curve for the station.
Apart from use in hydrological studies and for design purposes, the data can be of direct value for navigation,
flood prediction, water management, and waste water disposal.
Methods
Most water level gauging stations are equipped with a sensor or gauge and a recorder. In many cases the water
level is measured in a stilling well, thus eliminating strong oscillations.
Staff gauge:
This is the simplest type, with a graduated gauge plate fixed to a stable structure such as a pile, bridge pier, or a
wall. Where the range of water levels exceeds the capacity of a single gauge, additional ones may be placed on the
line of the cross section normal to the plane of flow.
Float gauge:
A float inside a stilling well, connected to the river by an inlet pipe, is moved up and down by the water level.
Fluctuations caused by short waves are almost eliminated. The movement of the float is transmitted by a wire
passing over a float wheel, which records the motion, leading down to a counterweight.
Pressure transducers:
The water level is measured as an equivalent hydrostatic pressure and transformed into an electrical signal via a
62

semi-conductor sensor. These are best suited for measuring water levels in open water (the effect of short waves
dies out almost completely within half a wavelength down into the water), as well as for the continuous recording
of groundwater levels. They should compensate for changes in the atmospheric pressure, and if air-vented cables
cannot be provided air pressure needs to be measured separately.
Bubble gauge:
This is a pressure actuated system, based on measurement of the pressure which is needed to produce bubbles
through an underwater outlet. These are used at sites where it would be difficult to install a float-operated recorder
or pressure transducer. From a pressurised gas cylinder or small compressor gas is led along a tube to some point
under the water (which will remain so for all water levels) and bubbles constantly flow out through the orifice. The
pressure in the measuring tube corresponds to that in the water above the orifice. Wind waves should not affect
this.
Ultrasonic sensor:
These are used for continuous non-contact level measurements in open channels. The sensor points vertically down
towards the water and emits ultrasonic pulses at a certain frequency. The inaudible sound waves are reflected by
the water surface and received by the sensor. The round trip time is measured electronically and appears as an
output signal proportional to the level. A temperature probe compensates for variations in the speed of sound in
air. They are accurate but susceptible to wind waves.
Peak level indicators:
There are some indicators of the maximum level reached by a flood, such as arrays of bottles which tip and fill
when the water reaches them, or a staff coated with soluble paint.
Presentation of results
Stage records taken along rivers used for hydrological studies, for design of irrigation works, or for flood protection require an accuracy of 2 5 cm, while gauge readings upstream of flow measuring weirs used to calculate
discharges from the measured heads require an accuracy of 2 5 mm. These days almost all are telemetered to a
central site. There is a huge volume of electronic hydrometry data being sent around Victoria.
Hydrographs, rating tables, and stage relation curves are typical presentations of water level data:
Hydrograph when stage records or the discharges are plotted against time.

Rating table at many gauging stations water levels are measured daily or hourly, while discharges are measured some times a year, using direct methods such as a propeller meter. From the corresponding water levels
from these, and possibly for others over years, a stage-discharge relationship can be built up, so that the routine
measurement of stage can be converted to discharge. We will be considering these in detail.
Stage relation curves from the hydrographs of two or more gauging stations along the river, relationships can
be formulated between the steady flow stages. These can be used to calculate the surface slope between two
gauges, and hence, to determine the roughness of the reach. Under unsteady conditions the relationship will be
disturbed. We will also be considering this later.

3.1.3

Discharge

Flow measurement may serve several purposes:

information on river flow for the design and operation of diversion dams and reservoirs and for bilateral agreements between states and countries.
distribution and charging of irrigation water
information for charging industries and treatment plants discharging into public waters
water management in urban and rural areas
reliable statistics for long-term monitoring.
Continuous (daily or hourly) measurements are very useful.
There are many methods of measuring the rate of volume flow past a point, of which some are single measurement
methods which are not designed for routine operation; the rest are methods of continuous measurements.
Velocity area method (current meter method)

63

Figure 3.1:

Traditional manner of taking current meter readings. In deeper water a boat is used.

The area of cross-section is determined from soundings, and flow velocities are measured using propeller current
meters, electromagnetic sensors, or floats. The mean flow velocity is deduced from points distributed systematically over the river cross-section. In fact, what this usually means is that two or more velocity measurements are
made on each of a number of vertical lines, and any one of several empirical expressions used to calculate the mean
velocity on each vertical, the lot then being integrated across the channel.
Calculating the discharge requires integrating
the velocity data over the whole channel - what is required is the
R
area integral of the velocity, that is Q = u dA. If we express this as a double integral we can write
Q=

Z(y)+h(y)
Z

u dz dy,

(3.1)

Z(y)

so that we integrate the velocity from the bed z = 0 to the surface z = h(y), where h is the local depth and where
our z is a local co-ordinate. Then we have to integrate these contributions right across the channel, for values of
the transverse co-ordinate z over the breadth B.
Calculation of mean velocity in the vertical
The first step is to compute the integral of velocity with depth, which hydrometricians think of as calculating
the mean velocity over the depth. Convention in hydrometry is that the mean velocity over a vertical can be
approximated by
1
u = (u0.2h + u0.8h ) ,
(3.2)
2
that is, the mean of the readings at 0.2 of the depth and 0.8 of the depth. ? has developed some families of methods.
Consider the law for turbulent flow over a rough bed, which can be obtained from the expressions on p582 of ?:
u
z
u=
(3.3)
ln ,

z0
where u is the shear velocity, = 0.4, ln() is the natural logarithm to the base e, z is the elevation above the bed,
and z0 is the elevation at which the velocity is zero. (It is a mathematical artifact that below this point the velocity
is actually negative and indeed infinite when z = 0 this does not usually matter in practice). If we integrate
equation (3.3) over the depth h we obtain the expression for the mean velocity:
1
u
=
h

Zh

u dz =

ln

h
1 .
z0

(3.4)

Now it is assumed that two velocity readings are made, obtaining u1 at z1 and u2 at z2 . This gives enough
information to obtain the two quantities u / and z0 . Substituting the values for point 1 into equation (3.3) gives
us one equation and the values for point 2 gives us another equation. Both can be solved to give the solution
1
u2 u u
2
1
z1
u
u2 u1
.
(3.5)
=
and z0 =
u1

ln (z2 /z1 )
z2
64

It is not necessary to evaluate these, for substituting into equation (3.4) gives a simple formula for the mean velocity
in terms of the readings at the two points:
u
=

u1 (ln(z2 /h)+1) u2 (ln(z1 /h)+1)

.
ln (z2 /z1 )

(3.6)

As it is probably more convenient to measure and record depths rather than elevations above the bottom, let
h1 = h z1 and h2 = h z2 be the depths of the two points, when equation (3.6) becomes
u
=

u1 (ln(1 h2 /h)+1) u2 (ln(1 h1 /h)+1)

.
ln ((h h2 ) / (h h1 ))

(3.7)

This expression gives the freedom to take the velocity readings at any two points, and not necessarily at points
such as 0.2h and 0.8h. This might simplify streamgauging operations, for it means that the hydrometrician, after
measuring the depth h, does not have to calculate the values of 0.2h and 0.8h and then set the meter at those points.
Instead, the meter can be set at any two points, within reason, the depth and the velocity simply recorded for each,
and equation (3.7) applied. This could be done either in situ or later when the results are being processed. This has
the potential to speed up hydrographic measurements.
If the hydrometrician were to use the traditional two points, then setting h1 = 0.2h and h2 = 0.8h in equation
(3.7) gives the result
u
= 0.4396 u0.2h + 0.5604 u0.8h 0.44u0.2h + 0.56u0.8h ,
(3.8)
whereas the conventional hydrographic expression is (see e.g. #7.1.5.3 of ?):
u
= 0.5 u0.2h + 0.5 u0.8h .

(3.9)

The nominally more accurate expression, equation (3.8), gives less weight to the upper measurement and more to
the lower. It might be useful, as it is just as simple as the traditional expression, yet is based on an exact analytical
integration of the equation for a turbulent boundary layer.
This has been tested by taking a set of gauging results. A canal had a maximum depth of 2.6m and was 28m wide,
and a number of verticals were used. The conventional formula (3.2), the mean of the two velocities, was accurate
to within 2% of equation (3.8) over the whole range of the readings, with a mean difference of 1%. That error
was always an overestimate. The more accurate formula (3.7) is hardly more complicated than the traditional one,
and it should in general be preferred. Although the gain in accuracy was slight in this example, in principle it is
desirable to use an expression which makes no numerical approximations to that which it is purporting to evaluate.
This does not necessarily mean that either (3.2) or (3.8) gives an accurate integration of the velocities which were
encountered in the field. In fact, one complication is where, as often happens in practice, the velocity distribution
near the surface actually bends back such that the maximum velocity is below the surface.
Integration of the mean velocities across the channel
The problem now is to integrate the readings for mean velocity at each station across the width of the channel.
Here traditional practice seems to be in error often the Mean-Section method is used. In this the mean velocity
between two verticals is calculated and then multiply this by the area between them, so that, given two verticals i
and i + 1 separated by bi the expression for the contribution to discharge is assumed to be
1
bi (hi + hi+1 ) (ui + ui+1 ) .
4
This is not correct. From equation (3.1), the task is actually to integrate across the channel the quantity which is
the mean velocity times the depth. For that the simplest expression is the Trapezoidal rule:
Qi =

1
bi (ui+1 hi+1 +ui hi )
2
To examine where the Mean-Section Method is worst, we consider the case at one side of the channel, where the
area is a triangle. We let the waters edge be i = 0 and the first internal point be i = 1, then the Mean-Section
Method gives
1
Q0 = b0 u1 h1 ,
4
while the Trapezoidal rule gives
1
Q0 = b0 u1 h1 ,
2
which is correct, and we see that the Mean-Section Method computes only half of the actual contribution. The
same happens at the other side. Contributions at these edges are not large, and in the middle of the channel
Qi =

65

the formula is not so much in error, but in principle the Mean-Section Method is wrong and should not be used.
Rather, the Trapezoidal rule should be used, which is just as easily implemented. In a gauging in which the lecturer
participated, a flow of 19.60 m3 s1 was calculated using the Mean-Section Method. Using the Trapezoidal rule,
the flow calculated was 19.92 m3 s1 , a difference of 1.6%. Although the difference was not great, practitioners
should be discouraged from using a formula which is wrong.
An alternative approach
It is strange that only very local methods are used in determining the vertical velocity distribution. ? considered
velocity distributions given by the more general law, assuming an additional linear and an additional quadratic
term in the velocity profile:
u
z
+ a1 z + a2 z 2 ,
(3.10)
u=
ln

z0
and using a global approximation method, where a function was assumed which could describe all the velocity
profiles on all the verticals, and then this was fitted to the data. An example of the results is given in figure 3.2.

Figure 3.2:

Cross-section of canal with velocity profiles and data points plotted transversely, showing fit by global function

Dilution methods
In channels where cross-sectional areas are difficult to determine (e.g. steep mountain streams) or where flow
velocities are too high to be measured by current meters dilution or tracer methods can be used, where continuity
of the tracer material is used with steady flow. The rate of input of tracer is measured, and downstream, after total
mixing, the concentration is measured. The discharge in the stream immediately follows.
Integrating float methods
There is another rather charming and wonderful method which has been very little exploited. At the moment it
has the status of a single measurement method, however the lecturer can foresee it being developed as a continuing method, although some research has been attempted, and always difficulties (light, weather, etc.) proved
insurmountable.
Theory
Consider a single buoyant particle (a float, an orange, an air bubble), which is released from a point on the bed. We
assume that it has a constant rise velocity w. As it rises it passes through a variable horizontal velocity field u(z),
where z is the vertical co-ordinate. The kinematic equations of the float are
dx
= u(z),
dt
dz
= w.
dt
Dividing the left and right sides, we obtain a differential equation for the particle trajectory
w
dz
=
,
dx
u(z)
however this can be re-arranged as:
Zh
0

u(z) dz =

ZL

w dx = wL,

where the particle reaches the surface a distance L downstream of the point at which it was released on the bed,
and where we have used a local vertical co-ordinate z with origin on the bed and where the fluid locally has a depth
66

h. The quantity on the left is important - it is the vertical integral of the horizontal velocity, or the discharge per
unit width at that section. We can generalise the expression for variation with y, across the channel, to write
Zh

u(y, z) dz = wL(y),

Z(y)

where Z(y) is the z co-ordinate of the bed. Now, if we integrate across the channel, in the co-ordinate direction y,
the integral of the left side is the discharge Q:

Q=

ZB Zh

u(y, z) dz dy = w

0 Z(y)

ZB

L(y) dy,

where B is the total width of the channel. Hence we have an expression for the discharge with very few approximations:
ZB
Q = w L(y) dy.
0

If we were to release bubbles from a pipe across the bed of the stream, on the bed, then this is
Q = Bubble rise velocity area on surface between bubble path pattern and line of release.
This is possibly the most direct and potentially the most accurate of all flow measurement methods!
Ultrasonic flow measurement

Figure 3.3:

Array of four ultrasonic beams in a channel

Mean velocity along beam path

Unfortunately, in all textbooks and International Standards a constant velocity is assumed - precisely what is being
sought to measure, and totally ignoring the fact that velocity varies along the path and indeed is zero at the ends!
Here we include the variability of velocity in our analysis.
Consider a velocity vector inclined to the beam path at an angle . If the velocity is u(s), showing that the velocity
does, in general, depend on position along the beam, then the component along the path is u(s) cos . Let c be
the speed of sound. The time dt taken for a sound wave to travel a distance ds along the path against the general
direction of flow is dt = ds/ (c u(s) cos ). If the path has total length L, then the total time of travel T1 is
obtained by integrating to give
ZT1
ZL
ds
T1 = dt =
,
(3.11)
c u(s) cos
0

and repeating for a traverse in the reverse direction:

T2 =

ZT2
0

dt =

ZL

ds
.
c + u(s) cos

67

(3.12)

Now we expand the denominators of both integrals by the binomial theorem:

1
T1 =
c

ZL
ZL
1
u(s)
u(s)
1+
1
cos ds and T2 =
cos ds,
c
c
c
0

(3.13)

where we have ignored terms which contain the square of the fluid velocity compared with the speed of sound.
Evaluating gives
ZL
ZL
L
L
1
1
T1 = + 2 u(s) cos ds and T2 = 2 u(s) cos ds.
(3.14)
c
c
c
c
0

Adding the two equations and solving for c and re-substituting we obtain
ZL

u(s) cos ds = 2L2

T1 T2

(T1 + T2 )

2.

(3.15)

is a measure of velocity. As
It can be shown that the relative error of this expression is of order (
u/c) , where u
u
1 ms1 and c 1400 ms1 it can be seen that the error is exceedingly small. What we first need to compute
the flow is the integral of the velocity component transverse to the beam path, for which we use the symbol Qz ,
the symbol with subscript suggesting the derivative of discharge with respect to elevation:
Qz =

u(s) sin ds.

(3.16)

Now we are forced to assume that the angle that the velocity vector makes with the beam is constant over the path
(or at least in some rough averaged sense), and so for constant, taking the trigonometric functions outside the
integral signs and combining equations (3.15) and (3.16) we obtain
Qz = 2 tan L2

T1 T2

(T1 + T2 )2

(3.17)

This shows how the result is obtained by assuming the angle of inclination of the fluid velocity to the beam is
constant, but importantly it shows that it is not necessary to assume that velocity u is constant over the beam path.
Equation (3.17) is similar to that presented in Standards and trade brochures, and implemented in practice, but
where it is obtained by assuming that the velocity is constant. It is fortunate that the end result is correct.
Vertical integration of beam data
The mean velocities on different levels obtained from the beam data are considered to be highly accurate, provided
all the technical problems associated with beam focussing etc. are overcome, and the streamflow has a constant
angle to the beam. The problem remains to calculate the discharge in the channel by evaluating the vertical
integral of Qz , which, as shown by equation (3.16), is the integral along the beam of the velocity transverse to the
beam. The problem is then to evaluate the vertical integral of the derivative of discharge with elevation:

Q=

Zh

Qz (z) dz,

(3.18)

where in practice the information available is that Qz = 0 on the bottom of the channel z = 0 and the two to
four values of Qz which have been obtained from beam data, as well as the total depth h. It is in the evaluation
of this integral that the performance of the trade and scientific literature has been poor. Several trade brochures
advocate the routine use of a single beam, or maybe two, suggesting that that is adequate (see, for example, Boiten
2000?, p141). In fact, with high-quality data for Qz at two or three levels, there is no reason not to use accurate
integration formulae. However, practice in this area has been quite poor, as trade brochures that the author has
seen use the inaccurate Mean-Section Method for integrating vertically over only three or four data points, when
its errors would be rather larger than when it is used for many verticals across a channel, as described previously.
This seems to be a ripe area for research.
Acoustic-Doppler Current Profiling (ADCP) methods
In these, a beam of sound of a known frequency is transmitted into the fluid, often from a boat. When the sound
strikes moving particles or regions of density difference moving at a certain speed, the sound is reflected back and
received by a sensor mounted beside the transmitter. According to the Doppler effect, the difference in frequency
between the transmitted and received waves is a direct measurement of velocity. In practice there are many particles
68

in the fluid and the greater the area of flow moving at a particular velocity, the greater the number of reflections
with that frequency shift. Potentially this method is very accurate, as it purports to be able to obtain the velocity
over quite small regions and integrate them up. However, this method does not measure in the top 15% of the depth
or near the boundaries, and the assumption that it is possible to extract detailed velocity profile data from a signal
seems to be optimistic. The lecturer remains unconvinced that this method is as accurate as is claimed.
Electromagnetic methods
Coil for producing magnetic field
Signal probes

Figure 3.4:

Electromagnetic installation, showing coil and signal probes

The motion of water flowing in an open channel cuts a vertical magnetic field which is generated using a large coil
buried beneath the river bed, through which an electric current is driven. An electromotive force is induced in the
water and measured by signal probes at each side of the channel. This very small voltage is directly proportional
to the average velocity of flow in the cross-section. This is particularly suited to measurement of effluent, water
in treatment works, and in power stations, where the channel is rectangular and made of concrete; as well as in
situations where there is much weed growth, or high sediment concentrations, unstable bed conditions, backwater
effects, or reverse flow. This has the advantage that it is an integrating method, however in the end recourse has to
be made to empirical relationships between the measured electrical quantities and the flow.

3.2 The analysis and use of stage and discharge measurements

Almost universally the routine measurement of the state of a river is that of the stage, the surface elevation at
a gauging station, usually specified relative to an arbitrary local datum. While surface elevation is an important
quantity in determining the danger of flooding, another important quantity is the actual flow rate past the gauging
station. Accurate knowledge of this instantaneous discharge - and its time integral, the total volume of flow - is
crucial to many hydrologic investigations and to practical operations of a river and its chief environmental and
commercial resource, its water. Examples include decisions on the allocation of water resources, the design of
reservoirs and their associated spillways, the calibration of models, and the interaction with other computational
components of a network.

3.2.1

Stage discharge method

The traditional way in which volume flow is inferred is for a Rating Curve to be derived for a particular gauging
station, which is a relationship between the stage measured and the actual flow passing that point. The measurement
of flow is done at convenient times by traditional hydrologic means, with a current meter measuring the flow
velocity at enough points over the river cross section so that the volume of flow can be obtained for that particular
stage, measured at the same time. By taking such measurements for a number of different stages and corresponding
discharges over a long period of time, a number of points can be plotted on a stage-discharge diagram, and a curve
drawn through those points, giving what is hoped to be a unique relationship between stage and flow, the Rating
Curve, as shown in Figure 3.5. If the supposedly unique relationship between the flow rate and the stage is written
Qr (), subsequent measurements of the surface elevation at some time t, such as an hourly or daily measurement,
are then used to give the discharge:
Q(t) = Qr ((t)).
It is assumed that for any stage reading, which is the routine periodic measurement, the corresponding discharge
can be calculated. This is what happens when the stage is read and telemetered to a central data management
authority. From the rating curve for that stage, the corresponding discharge can be calculated. This is very widely
used and is the routine method of flow measurement. It begs a number of questions.
69

Stage
(m)

Discharge ( m3 s1 )

Figure 3.5:

Rating curve (stage-discharge diagram) and data points to which it has been fitted

Stage

Actual flood event

Steady flow rating curve

A measured stage value

Qfalling Qrated

Qrising

Discharge

Figure 3.6: Stage-discharge diagram showing the steady-flow rating curve and an exaggerated looped trajectory of a particular flood event, which can be due to two effects: changing roughness and unsteady flow effects

There are several problems associated with the use of a Rating Curve:
The assumption of a unique relationship between stage and discharge may not be justified.

Discharge is rarely measured during a flood, and the quality of data at the high flow end of the curve might be
quite poor.
It is usually some sort of line of best fit through a sample made up of a number of points - sometimes extrapolated for higher stages.
It has to describe a range of variation from no flow through small but typical flows to very large extreme flood
events.
There are a number of factors which might cause the rating curve not to give the actual discharge, some of
which will vary with time. Factors affecting the rating curve include:
The channel changing as a result of modification due to dredging, bridge construction, or vegetation growth.
Sediment transport - where the bed is in motion, which can have an effect over a single flood event, because
the effective bed roughness can change during the event. As a flood increases, any bed forms present will tend
to become larger and increase the effective roughness, so that friction is greater after the flood peak than before,
so that the corresponding discharge for a given stage height will be less after the peak. This will contribute to
70

a flood event showing a looped curve on a stage-discharge diagram as is shown on Figure 3.6.
Backwater effects - changes in the conditions downstream such as the construction of a dam or flooding in
the next waterway.
Unsteadiness - in general the discharge will change rapidly during a flood, and the slope of the water surface
will be different from that for a constant stage, depending on whether the discharge is increasing or decreasing,
also contributing to a flood event appearing as a loop on a stage-discharge diagram such as Figure 3.6.
Variable channel storage - where the stream overflows onto flood plains during high discharges, giving rise
to different slopes and to unsteadiness effects.
Vegetation - changing the roughness and hence changing the stage-discharge relation.
Ice - which we will ignore.
Some of these can be allowed for by procedures which we will describe later.
The hydraulics of a gauging station
1
High flow

Flood

2
3

Low flow
4
5
Distant
control

Channel control
Gauging Local
station control
Channel control

Larger body
of water

Figure 3.7: Section of river showing different controls at different water levels with implications for the stage discharge
relationship at the gauging station shown

A typical set-up of a gauging station where the water level is regularly measured is given in Figure 3.7 which
shows a longitudinal section of a stream. Downstream of the gauging station is usually some sort of fixed control
which may be some local topography such as a rock ledge which means that for relatively small flows there is a
relationship between the head over the control and the discharge which passes. This will control the flow for small
flows. For larger flows the effect of the fixed control is to drown out, to become unimportant, and for some other
part of the stream to control the flow, such as the larger river downstream shown as a distant control in the figure,
or even, if the downstream channel length is long enough before encountering another local control, the section
of channel downstream will itself become the control, where the control is due to friction in the channel, giving
a relationship between the slope in the channel, the channel geometry and roughness and the flow. There may be
more controls too, but however many there are, if the channel were stable, and the flow steady (i.e. not changing
with time anywhere in the system) there would be a unique relationship between stage and discharge, however
complicated this might be due to various controls. In practice, the natures of the controls are usually unknown.
Something which the concept of a rating curve overlooks is the effect of unsteadiness, or variation with time. In
a flood event the discharge will change with time as the flood wave passes, and the slope of the water surface will
be different from that for a constant stage, depending on whether the discharge is increasing or decreasing. Figure
3.7 shows the increased surface slope as a flood approaches the gauging station. The effects of this are shown on
Figure 3.6, in somewhat exaggerated form, where an actual flood event may not follow the rating curve but will in
general follow the looped trajectory shown. As the flood increases, the surface slope in the river is greater than the
slope for steady flow at the same stage, and hence, according to conventional simple hydraulic theory explained
below, more water is flowing down the river than the rating curve would suggest. This is shown by the discharge
marked Qrising obtained from the horizontal line drawn for a particular value of stage. When the water level is
falling the slope and hence the discharge inferred is less.
The effects of this might be important - the peak discharge could be significantly underestimated during highly
dynamic floods, and also since the maximum discharge and maximum stage do not coincide, the arrival time of the
peak discharge could be in error and may influence flood warning predictions. Similarly water-quality constituent
loads could be underestimated if the dynamic characteristics of the flood are ignored, while the use of a discharge
hydrograph derived inaccurately by using a single-valued rating relationship may distort estimates for resistance
coefficients during calibration of an unsteady flow model.
71

Rating curves representation, approximation and calculation

14
13
12
Stage 11
(m)
10
9
8
7

500
Figure 3.8:

1000
1500
Discharge (Q m3 s1 )

2000

2500

Rating curve using natural (Q, ) axes

Figure 3.8 shows the current rating curve for the Ovens River at Wangaratta in Victoria, Australia, where flow
measurements have been made since 1891. There are a couple of difficulties with such a curve, including reading
results off for small flows, where the curve is locally vertical, and for high flows where it is almost horizontal.
A traditional way of overcoming the difficulty of representing rating curves over a large range has been to use
log-log axes. However, this has no physical basis and has a number of practical difficulties, although it has been
recommended by International Standards.Hydraulic theory can help here, for it can be used to show that the stagedischarge relationship will tend to show stage varying approximately like Q1/2 , for both cases:

1. Flow across a U-shaped (parabolic) weir, the approximate situation for low flow at a gauging station, when a
local control such as a rock ledge controls the flow, and
2. Uniform flow down a U-shaped (parabolic) waterway for large flows, when the local control is washed out and
the waterway acts more like a uniform flow governed by Mannings law.

14
13
12
Stage 11
(m)
10
9
8
7

Figure 3.9:

10

30
20
Q ( m3 s1 )1/2

40

50

The same rating curve as in figure 3.8 but using ( Q, ) axes

In these cases, both parts of the relationship would plot as (different) straight lines on

72

Q, axes. In figure 3.9

we plot the results from the above figure on such a square root scale for the discharge, and we see that indeed
at both small and large flows the rating curve is a straight line. This means that simpler procedures of numerical
approximation and interpolation could be used. Sometimes

results have to be taken by extrapolating the curve. If

this has to be done, then linear extrapolation on the
Q, axes might be reasonable, but it is still a procedure to
be followed with great caution, as the actual geometry for above-bank flows can vary a lot.
A number of such practical considerations are given in ?.

3.2.2

Stage-discharge-slope method

This is presented in some books and in International Standards, however, especially in the latter, the presentation is
confusing and at a low level, where no reference is made to the fact that underlying it the slope is being measured.
Instead, the fall is described, which is the change in surface elevation between two surface elevation gauges and
is simply the slope multiplied by the distance between them. No theoretical justification is provided and it is
presented in a phenomenological sense (see, for example, ?).
Although the picture in Figure 3.7 of the factors affecting the stage and discharge at a gauging station seems
complicated, the underlying processes are capable of quite simple description. In a typical stream, where all wave
motion is of a relatively long time and space scale, the governing equations are the long wave equations, which
are a pair of partial differential equations for the stage and the discharge at all points of the channel in terms of
time and distance along the channel. One is a mass conservation equation, the other a momentum equation. Under
the conditions typical of most flows and floods in natural waterways, however, the flow is sufficiently slow that
the equations can be simplified considerably. Most terms in the momentum equation are of a relative magnitude
given by the square of the Froude number, which is U 2 /gD, where U is the fluid velocity, g is the gravitational
acceleration, and D is the mean depth of the waterway. In most rivers, even in flood, this is small, and the
approximation may be often used. For example, a flow of 1 ms1 with a depth of 2 m has F 2 0.05.
In 2.3.8 we saw that under these circumstances, a surprisingly good approximation to the momentum equation of
motion for flow in a waterway is the simple equation:
p
(3.19)
Q(t) = K((t)) /x(t),

where we have shown that the slope of the free surface might be a function of time. This gives us a formula for
calculating the discharge Q which is as accurate as is reasonable to be expected in river hydraulics, provided we
know
1. the stage and the dependence of conveyance K on stage at a point from either measurement or the G-M-S or
Chzys formulae, and
2. the slope of the surface
Stage-conveyance curves
Equation (3.19) shows how the discharge actually depends on both the stage and the surface slope, whereas traditional hydrography assumes that it depends on stage alone. If the slope does vary under different backwater
conditions or during a flood, then a better hydrographic procedure would be to gauge the flow when it is steady,
and to measure the surface slope , thereby enabling a particular value of K to be calculated for that stage. If this
were done over time for a number of different stages, then a stage-conveyance relationship could be developed
which should then hold whether or not the stage is varying. Subsequently, in day-to-day operations, if the stage
and the surface slope were measured, then the discharge calculated from equation (3.19) should be quite accurate,
within the relatively mild assumptions made so far. All of this holds whether or not the gauging station is affected
by a local or channel control, and whether or not the flow is changing with time.
This suggests that a better way of determining streamflows in general, but primarily where backwater and unsteady
effects are likely to be important, is for the following procedure to be followed:
1. At a gauging station, two measuring devices for stage be installed, so as to be able to measure the slope of the
water surface at the station. One of these could be at the section where detailed flow-gaugings are taken, and
the other could be some distance upstream or downstream such that the stage difference between the two points
is enough that the slope can be computed accurately enough. As a rough guide, this might be, say 10 cm, so
that if the water slope were typically 0.001, they should be at least 100 m apart.
2. Over time, for a number of different flow conditions the discharge Q would be measured using conventional
methods such as by current meter. For each gauging, both surface elevations would be recorded, one becoming
the stage to be used in the subsequent
relationship, the other so that the surface slope S can be calculated.
p
Using equation (3.19), Q = K() S , this would give the appropriate value of conveyance K for that stage,
automatically corrected for effects of unsteadiness and downstream conditions.
73

3. From all such data pairs (i , Ki ) for i = 1, 2, . . ., the conveyance curve (the functional dependence of K on
) would be found, possibly by piecewise-linear or by global approximation methods, in a similar way to the
description of rating curves described below. Conveyance has units of discharge, and as the surface slope is
unlikely to vary all that much, we note that there are certain advantages in representing rating curves on a plot
using the square root of the discharge,
and it my well be that the stage-conveyance curve would be displayed

and approximated best using ( K, ) axes.

4. Subsequent routine measurements would obtain both stages, including the stage to be used in the stageconveyance relationship, and hence the water surface slope, which would then be substituted into equation
(3.19) to give the discharge, corrected for effects of downstream changes and unsteadiness.
If hydrography had followed the path described above, of routinely measuring surface slope and using a stageconveyance relationship, the science would have been more satisfactory. Effects due to the changing of downstream controls with time, downstream tailwater conditions, and unsteadiness in floods would have been automatically incorporated, both at the time of determining the relationship and subsequently in daily operational practice.
However, for the most part slope has not been measured, and hydrographic practice has been to use rating curves
instead. The assumption behind the concept of a discharge-stage relationship or rating curve is that the slope at a
station is constant over all flows and events, so that the discharge is a unique function of stage Qr () where we
use the subscript r to indicate the rated discharge. Instead of the empirical/rational expression (3.19), traditional
practice is to calculate discharge from the equation
Q(t) = Qr ((t)),

(3.20)

thereby ignoring any effects that downstream backwater and unsteadiness might have, as well as the possible
changing of a downstream control with time.
In comparison, equation (3.19), based on a convenient empirical approximation to the real hydraulics of the river,
contains the essential nature of what is going on in the stream. It shows that, although the conveyance might be
a unique function of stage which it is possible to determine by measurement, because the surface slope will in
general vary throughout different flood events and downstream conditions, discharge in general does not depend
on stage alone.

3.2.3 Looped rating curves correcting for unsteady effects in obtaining discharge
from stage
In conventional hydrography the stage is measured repeatedly at a single gauging station so that the time derivative
of stage can easily be obtained from records but the surface slope along the channel is not measured at all. The
methods of this section are all aimed at obtaining the slope in terms of the stage and its time derivatives at a single
gauging station. The simplest and most traditional method of calculating the effects of unsteadiness has been the
Jones formula, derived by B. E. Jones in 1916 (see for example ?, ?). The principal assumption is that to obtain the
slope, the x derivative of the free surface, we can use the time derivative of stage which we can get from a stage
record, by assuming that the flood wave is moving without change as a kinematic wave (Lighthill and Whitham,
1955?) such that it obeys the partial differential equation:
h
h
+c
= 0,
(3.21)
t
x
where h is the depth and c is the kinematic wave speed. Solutions of this equation are simply waves travelling
at a velocity c without change. The equation was obtained as the last of a series of approximations in Section
2.3.11. The kinematic wave speed c is given by the derivative of flow with respect to cross-sectional area, the
Kleitz-Seddon law
1 dQr
1 dK p
S,
(3.22)
c=
=
B d
B d

where B is the width of the surface and Qr is the steady rated discharge p
corresponding to stage
p , and where we

have expressed this also in terms of the conveyance K, where Qr = K() S, and the slope S is the mean slope
of the stream. A good approximation is c 5/3 U , where U is the mean stream velocity.

The Jones method assumes that the surface slope S in equation (3.19) can be simply related to the rate of change
of stage with time, assuming that the wave moves without change. Thus, equation (3.21) gives an approximation
for the surface slope: h/x 1/c h/t. We then have to use the simple geometric relation between surface
such that we have the approximation
gradient and depth gradient, that /x = h/x S,
S =

h
1 h
= S
S +
x
x
c t
74

and recognising that the time derivative of stage and depth are the same, h/t = /t, equation (3.19) gives
r
1
Q = K S +
(3.23)
c t
If we divide by the steady discharge corresponding to the rating curve we obtain
r
1
Q
= 1+
Qr
cS t

(Jones)

In situations where the flood wave does move as a kinematic wave, with friction and gravity in balance, this theory
is accurate. In general, however, there will be a certain amount of diffusion observed, where the wave crest subsides
and the effects of the wave are smeared out in time.

To allow for those effects ? provided the theoretical derivation of two methods for calculating the discharge. The
derivation of both is rather lengthy. The first method used the full long wave equations and approximated the
surface slope using a method based on a linearisation of those equations. The result was a differential equation for
dQ/dt in terms of Q and stage and the derivatives of stage d/dt and d2 /dt2 , which could be calculated from
the record of stage with time and the equation solved numerically. The second method was rather simpler, and was
based on the next best approximation to the full equations after equation (3.21). This gives the advection-diffusion
equation
h
h
2h
,
(3.24)
+c
=
t
x
x2
where the difference between this and equation (3.21) is the diffusion term on the right, where is a diffusion
coefficient (with units of L2 T1 ), given by
K
p .
=
2B S

Equation (3.24), is a consistent low-inertia approximation to the long wave equations, where inertial terms, which
are of the order of the square of the Froude number, which approximates motion in most waterways quite well.
However, it is not yet suitable for the purposes of this section, for we want to express the x derivative at a point in
terms of time derivatives. To do this, we use a small-diffusion approximation, we assume that the two x derivatives
on the right of equation (3.24) can be replaced by the zero-diffusion or kinematic wave approximation as above,
/x 1/c /t, so that the surface slope is expressed in terms of the first two time derivatives of stage. The
resulting expression is:
h
h
2h
+c
= 2 2,
t
x
c t
and solving for the x derivative, we have the approximation

h
1 h
d2 h
= S
S +
3 2,
x
x
c t
c dt
and substituting into equation (3.19) gives
v
u
1 d
d2
Q = Qr ()u
1
+

u |{z}
u
cS dt
c3 S dt2
tRating curve
|
{z
} | {z }
S =

Jones formula

(3.25)

Diffusion term

where Q is the discharge at the gauging station, Qr () is the rated discharge for the station as a function of stage,
S is the bed slope, c is the kinematic wave speed given by equation (3.22):
p
S dK
1 dQr
=
,
c=
B d
B d
in terms of the gradient of the conveyance curve or the rating curve, B is the width of the water surface, and where
the coefficient is the diffusion coefficient in advection-diffusion flood routing, given by:
=

K
Qr
p =
.
2B S
2B S

(3.26)

In equation (3.25) it is clear that the extra diffusion term is a simple correction to the Jones formula, allowing for
the subsidence of the wave crest as if the flood wave were following the advection-diffusion approximation, which
is a good approximation to much flood propagation. Equation (3.25) provides a means of analysing stage records
and correcting for the effects of unsteadiness and variable slope. It can be used in either direction:
75

 If a gauging exercise has been carried out while the stage has been varying (and been recorded), the value of
Q obtained can be corrected for the effects of variable slope, giving the steady-state value of discharge for the
stage-discharge relation,
And, proceeding in the other direction, in operational practice, it can be used for the routine analysis of stage
records to correct for any effects of unsteadiness.
The ideas set out here are described rather more fully in ?.
An example
A numerical solution was obtained for the particular case of a fast-rising and falling flood in a stream of 10 km
length, of slope 0.001, which had a trapezoidal section 10 m wide at the bottom with side slopes of 1:2, and a
Mannings friction coefficient of 0.04. The downstream control was a weir. Initially the depth of flow was 2 m,
while carrying a flow of 10 m3 s1 . The incoming flow upstream was linearly increased ten-fold to 100 m3 s1
over 60 mins and then reduced to the original flow over the same interval. The initial backwater curve problem
was solved and then the long wave equations in the channel were solved over six hours to simulate the flood. At
a station halfway along the waterway the computed stages were recorded (the data one would normally have), as
well as the computed discharges so that some of the above-mentioned methods could be applied and the accuracy
of this work tested.

100

Actual flow
From rating curve
Jones formula
Eqn (3.25)

90
80
70
60
Flow
( m3 s1 )

50
40
30
20
10
0
0

Time (hours)
Figure 3.10: Simulated flood with hydrographs computed from stage record using three levels
of approximation

Results are shown on Figure 3.10. It can be seen that the application of the diffusion level of approximation of
equation (3.25) has succeeded well in obtaining the actual peak discharge. The results are not exact however,
as the derivation depends on the diffusion being sufficiently small that the interchange between space and time
differentiation will be accurate. In the case of a stream such as the example here, diffusion is relatively large, and
our results are not exact, but they are better than the Jones method at predicting the peak flow. Nevertheless, the
results from the Jones method are interesting. A widely-held opinion is that it is not accurate. Indeed, we see here
that in predicting the peak flow it was not accurate in this problem. However, over almost all of the flood it was
accurate, and predicted the time of the flood peak well, which is also an important result. It showed that both before
and after the peak the discharge wave led the stage wave, which is of course in phase with the curve showing
76

the flow computed from the stage graph and the rating curve. As there may be applications where it is enough to
know the arrival time of the flood peak, this is a useful property of the Jones formula. Near the crest, however, the
rate of rise became small and so did the Jones correction. Now, and only now, the inclusion of the extra diffusion
term gave a significant correction to the maximum flow computed, and was quite accurate in its prediction that the
real flow was some 10% greater than that which would have been calculated just from the rating curve. In this
fast-rising example the application of the unsteady corrections seems to have worked well and to be justified. It is
no more difficult to apply the diffusion correction than the Jones correction, both being given by derivatives of the
stage record.

3.2.4

The effects of bed roughness on rating curves

Introduction

16
15
14
13
12
Stage 11
(m) 10
9
8
7
6
5

5000

10000
15000
Discharge (Q m3 s1 )

20000

Figure 3.11: Flood trajectories for Station 41 on the Red River, showing raw data corrected
data was everywhere obscured

We have considered the detailed records for 1995 and 1996 for the Red River, Viet Nam. The trajectories, plotted
on Stage/Discharge axes on Figure 3.11, show considerable loopiness. In the expectation that the cause was the
unsteady wave effects as described above, we applied the theory described above, assuming for simple purposes,
a slope S0 = 0.0015, and using the recorded values of stage, depth, area, and breadth. Everywhere, no visible
correction was made, and the methods of the section above have failed to describe the loopiness. It seems that the
loopiness in this case must be due to effects of the bed topography and friction changing with flow. This is the
more common situation, and it is bed roughness changes which will more often be the cause of loopiness.
? have written on the nature of the relationship between stage and discharge when the river bed is mobile, when
bed forms may change, depending on the flow, such that in a flood there may be different bed roughness before
and after, and the stage-discharge trajectory may show loopiness, as shown in Figure 3.11 above. This suggests a
different approach to the relationship between stage and discharge observed in a river, when the river itself is the
control, rather than a hydraulic structure.
Let us consider the mechanism by which changes in roughness cause the flood trajectories to be looped, by considering a hypothetical and idealised situation. In Figure 3.12 is shown a plot of Stage versus Discharge. The rating
curves which would apply if the bedforms were held constant are shown for a flat bed and for various increasing bedform roughness. In the left corner is a stage-time graph with three flood events, one small, one large, and
one not yet completed. The points labelled O, A, ..., G are also shown on the flood trajectory, showing the actual
relationship between stage and discharge at each time.
O: flow is small, over a flat bed.
A: a small flood peak has arrived. The flow is not enough to change the nature of the bed, and the flood
trajectory follows the flat-bed rating curve back down to a smaller flow, and then back up to a larger flow.
77

Stage

Large bedforms
C
D

Medium bedforms
Small bedforms
Flat bed

B C D EF G
Time

Stage

Flood trajectory
Rating curves for constant roughness

O
Discharge
Figure 3.12:

Rating curves for different bedforms and a looped flood trajectory

B: the bed is no longer stable and bed forms, with increasing roughness, start to form. Accordingly, the carrying
capacity of the channel is reduced and the stage increases.
C: the flood peak has arrived, and the bed forms continue to grow, so that a little time later the stage is a
maximum.
D: the bedforms have continued to grow until here, although the flow is decreasing.
E: the flow has decreased much more quickly than the bed-forms can adjust, and the point is only a little below
the rating curve corresponding to the largest bedforms.
F: over the intervening time, flow has been small and almost constant, however the time has been enough to
reduce the bed-forms. Now another flood starts to arrive, and this time, instead of following the flat-bed curve,
it already starts from a finite roughness, such as we seem to see in Figure 3.11.
G: after this, the history of the stage will still depend on the history of the flow and the characteristics of the
bed-form rate of growth.
A theoretical approach
This process could be modelled mathematically. If we had an expression for the rate of growth of roughness as a
function of discharge, depth, and roughness itself, say,
d
= f (Q, h, ),
dt

(3.27)

where Q is discharge, h is depth, and is some measure of roughness, and if we had a channel friction law such
as Gauckler-Manning-Strickler, which we write in a general way:

Q = F (h, , S),

(3.28)

then if we assume that slope is constant, as it almost always is except when unsteady effects are important, then
differentiating (3.28) with respect to time,
dQ
dt

=
=

F
h
F
h

dh F d
+
dt
dt
dh F
+
f (Q, h, ),
dt

having substituted equation (3.27), giving us an ordinary differential equation for Q as a function of t provided we
know the stage hydrograph h(t), which is what is usually measured.
78

In general the condition of the river bed, whether smooth, or with dunes, antidunes or standing waves, will depend
on the time history of the flow. That is, the roughness now depends on the preceding conditions for a period of
time.
An empirical approach
This immediately suggests the concept of using some form of convolution, where the effects of preceding events
are incorporated in an integral sense. In hydrology, this is most familiar as the unit hydrograph, see for example,
Chapter 7 of ?. In the case of a river with data such as in Figure 3.11 we suggest the following nonlinear influence
function, where the discharge Q at time t can be written as nonlinear functions of stage at previous times t ,
t 2, etc.:
Q(t) = a00

+a01 (t) + a02 2 (t) + . . .

+a11 (t ) + a12 2 (t ) + . . .
+a21 (t 2) + a22 2 (t 2) + . . .
+....

If only the first line of that equation had been taken, then that is a conventional rating curve, where the discharge
at t is assumed to be a function of the stage at t.
Now, the procedure would be to calculate the coefficients aij from a long data sequence, using least squares
methods. This could then be then applied to future events so that a better prediction of the actual discharge could
be obtained.

79

Chapter 4

Computational hydraulics

4.0.5

Steady flow

4.0.6

Unsteady flow

Finite differences Forwards-time-centred space

The slow-change/slow-flow approximation

80

Chapter 5

Sediment transport

5.1 General
5.2 Initiation of motion
5.3 Bedforms and alluvial roughness
5.4 Transport formulae
5.5 Unsteady aspects

81

Chapter 6

River morphology

6.1 Introduction
6.2 Planform
6.3 Longitudinal profile
6.4 Bends
6.5 Channel characteristics
6.6 Bifurcations and confluences

82

Chapter 7

River engineering

7.1 Introduction
7.2 Bed regulation
7.3 Discharge control
7.4 Water level control
7.5 Water quality control
7.6 River engineering for different purposes
7.6.1

Flood control and drainage

7.6.2

Navigation

7.6.3

Hydropower

7.6.4

Water supply

7.6.5

Waste discharge

7.6.6

Crossings of other infrastructure

7.6.7

Soil conservation

7.6.8

Nature preservation and restoration

83

References
From Wikipedia:
UNESCO- und DIN-Definition
Die Hydrographie ist "die Wissenschaft und Praxis der Messung und Darstellung der Parameter, die notwendig
sind, um die Beschaffenheit und Gestalt des Bodens der Gewsser, ihre Beziehung zum festen Land und den
Zustand und die Dynamik der Gewsser zu beschreiben." (Definition nach United Nations Economic and Social
Council, 1978)
Diese Definition ist auch in der DIN 18709-3 festgehalten. In der DIN werden die Bezeichnungen Gewsservermessung, Seevermessung und Binnengewsservermessung einheitlich mit "hydrographic surveying" ins Englische
bertragen.
Englische IHO-Definition
Hydrography: That branch of applied sciences which deals with the measurement and description of the features of
the seas and coastal areas for the primary purpose of navigation and all other marine purposes and activities inter
alia- offshore activities, research, protection of the environment, and prediction services. (Definition: International
Hydrographic Organisation, IHO) Eine neue Definition von Hydrography wird momentan in der IHO diskutiert
(April, 2009).
sterreicher NORM-Definition
Die Hydrographie ist jener Teil der Hydrologie der sich mit der quantitativen Erfassung und Beschreibung des
Wasserkreislaufes auf, unter und ber der Erdoberflche und mit der Behandlung der damit zusammenhngenden
Fragen beschftigt. (Ergnzende NORM B 2400 zur NORM EN ISO 772)
Diese wohl in sterreich und in der Schweiz bekannte Definition ist in Deutschland unblich und trifft auch nicht
die Bedeutung des englischen Wortes "Hydrography".
Bifurcations and confluences this should go after the structures part

84